Conversion of a Chaperonin GroEL-independent Protein into an Obligate Substrate*

Background: Chaperonin GroEL and GroES (GroE) assist a subset of proteins in the cell. Results: Conversion of a GroE-independent MetK into an obligate MetK occurred; GroE dependence was correlated with the propensity to form protein aggregates. Conclusion: Subtle differences, even at single point mutations, determine the GroE dependence. Significance: Buffering the aggregation-prone mutations by GroE plays a role in maintaining diversity of proteins. Chaperones assist protein folding by preventing unproductive protein aggregation in the cell. In Escherichia coli, chaperonin GroEL/GroES (GroE) is the only indispensable chaperone and is absolutely required for the de novo folding of at least ∼60 proteins. We previously found that several orthologs of the obligate GroE substrates in Ureaplasma urealyticum, which lacks the groE gene in the genome, are E. coli GroE-independent folders, despite their significant sequence identities. Here, we investigated the key features that define the GroE dependence. Chimera or random mutagenesis analyses revealed that independent multiple point mutations, and even single mutations, were sufficient to confer GroE dependence on the Ureaplasma MetK. Strikingly, the GroE dependence was well correlated with the propensity to form protein aggregates during folding. The results reveal the delicate balance between GroE dependence and independence. The function of GroE to buffering the aggregation-prone mutations plays a role in maintaining higher genetic diversity of proteins.

Protein folding is basically a spontaneous process (1). However, folding often competes with a side reaction, intermolecular aggregate formation, which usually impairs the functions of proteins (2,3). Indeed, a global aggregation analysis of thousands of Escherichia coli proteins, using a reconstituted cellfree translation system, revealed that ϳ30% of proteins are aggregation-prone (4). The cellular milieu, where proteins and other molecules exist in highly crowded conditions, further increases the risk of aggregate formation (5,6). In addition, most of the amino acid mutations in proteins are deleterious, and eventually result in aggregate formation (7).
To counteract the inherent tendency toward protein aggregation, cells have evolved a variety of chaperones (8). The canonical chaperones assist the folding of many proteins by preventing irreversible aggregate formation (3,9,10). In vitro, most of the aggregation-prone proteins, which were revealed by the reconstituted cell-free translation system (4), are rescued by one or more conserved chaperones, such as GroEL/GroES (GroE) 2 or DnaK/DnaJ/GrpE (DnaK system) (11). Overall, the effects of GroE and the DnaK system are similar, but certain fractions of proteins are biased toward either GroE or DnaK, probably reflecting the chaperone specificities (11). In vivo, recent proteome-wide analyses to identify the substrate proteins of chaperones confirmed the commonality and the specificity among the chaperone substrates (12)(13)(14). These previous observations suggested that determinants for the chaperone requirements would be included in their amino acid sequences, although the mechanism by which each chaperone recognizes non-native proteins in a different manner remains to be elucidated. To decode the determinants for the chaperone requirements, uncovering the relationship between chaperones and their obligate substrates is a key approach. The chaperonin GroE is a highly conserved molecular chaperone and is the only chaperone essential for E. coli viability (15). In E. coli cells, GroE interacts with ϳ250 proteins (12), including ϳ60 obligate substrates termed Class IV substrates, which are inherently aggregation-prone (13). The in vivo GroE dependence of Class IV substrates is not conserved among species; Class IV homologs (orthologs) in GroE-lacking organisms, such as Ureaplasma urealyticum, fold to the native state in GroE-depleted E. coli cells (13,16). A recent study using structural homologs of obligate GroE substrates revealed that the folding properties were different between the GroE-dependent and GroE-independent proteins (16). Further comparative analyses using such homologs with different GroE requirements will be necessary to reveal the GroE dependence at an amino acid sequence level.
MetK, one of the Class IV substrates, is a good choice to analyze the determinants of the GroE requirement among homologs. The GroE requirement of MetK in E. coli (EcMetK) has been analyzed very well both in vitro and in vivo (12,13,17). EcMetK, which is an essential protein for methionine biosynthesis, links a well characterized phenotype of GroE-depleted cells; the loss of functional MetK in GroE-depleted E. coli induces the overexpression of the MetE protein (17)(18)(19). The MetK ortholog in U. urealyticum (UuMetK), which shares 45% identity with EcMetK, does not require GroE in its folding process and complements the MetE overexpression in GroE-depleted cells (13,17).
In this study, we investigated the key features that define the GroE dependence, by using the MetK proteins. Chimera or random mutagenesis analyses revealed that independent multiple point mutations, and even single mutations, were sufficient to convert the GroE-independent UuMetK into the GroE-dependent MetK. Strikingly, the GroE dependence was well correlated with the propensity to form protein aggregates during folding. The results suggest that the maintenance of GroE independence is marginal, and thus provide insight into protein evolution buffered by chaperones.

EXPERIMENTAL PROCEDURES
Plasmids-The plasmids pMCS, pMCS-HA, pMetE240-ParM-FLAG, and pMetE-sfGFP were constructed previously (13,17,20,21). The UuMetK gene optimized for E. coli codon use was chemically synthesized in the previous study (13). To construct chimeric metK genes, each region was independently amplified from pMCS-EcMetK and pMCS-UuMetK and combined by PCR. Local or point mutations were introduced by whole plasmid PCR, using PrimeSTAR MAX DNA polymerase (Takara) and DNA primers with the corresponding mutations. The resultant genes encoding chimeric or mutated MetKs were digested and then cloned into the NdeI/BamHI sites of pMCS-HA. The tac promoter and the HA tag at the C terminus were attached by the cloning. To construct the plasmids for protein purification, the metK genes were cloned into the NdeI/XhoI sites of pET15b(ϩ) (Novagen). To construct the plasmids for the colony assay, the BamHI/XhoI sites in pACY-CDuet-1 (Novagen) were replaced with the BglII/XhoI sites of pMCS-UuMetKHA, by digestion and ligation.
Proteins-E. coli BL21(DE3) cells harboring each metK plasmid were grown at 37°C in LB medium with 50 g/ml ampicillin. At log phase, protein overexpression was induced by 1 mM isopropyl ␤-D-thiogalactopyranoside. EcMetK and UuMetK were overexpressed for 3 h at 37°C. UuMetK mutants (K215F and K215F/I216F) were expressed by overnight induction at 18°C because they formed inclusion bodies at 37°C. The cells were then harvested, suspended in lysis buffer (20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 10% glycerol), and sonicated (Branson Sonifier). Soluble fractions were obtained by centrifugation at 15,000 ϫ g for 30 min. The proteins were applied to nickelnitrilotriacetic acid-agarose resin (Qiagen), which was washed and eluted with lysis buffer containing 50 or 400 mM imidazole, respectively. The eluted fractions were dialyzed against 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, and 10% glycerol. The MetK concentration was typically Ͼ400 M after dialysis.
In vivo GroE Requirement Assay-The in vivo GroE requirement of each metK gene was assessed by the previously reported procedure, with slight modifications (13). E. coli MGM100 (MG1655 groE::araC-P BAD -groE (Kan r ) (22)) cells harboring each MetK plasmid were grown in LB medium, with 50 g/ml ampicillin and 0.2% arabinose, at 37°C to log phase. After washing, the cells were diluted into LB with 1 mM diaminopimelate, 50 g/ml ampicillin, and either 0.2% arabinose or 0.2% glucose. The dilution ratios were 1:5,000 for the 0.2% arabinose condition and 1:500 for the 0.2% glucose condition. The cells were cultivated for 5.5 h after the sugar shift, and then were harvested. The total quantities of the cells were adjusted with a buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA), and then the cells were disrupted by sonication (Branson Sonifier). Soluble fractions were obtained by centrifugation at 15,000 ϫ g for 30 min. The total and soluble extracts were examined by SDS-PAGE, and the MetK levels were detected by immunoblotting, using an anti-HA monoclonal antibody (conjugated with HRP) (Sigma).
Light Scattering Assay-Protein aggregation during refolding was monitored by light scattering for 20 min at room temperature (23). The light scattering intensity was measured with an FP-6500 spectrofluorometer (JASCO), with both excitation and emission at 640 nm. The purified recombinant MetK variants (100 M) were denatured by 6 M guanidine hydrochloride and were refolded by dilution to 5 M in 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1 mM EDTA, and 1 mM dithiothreitol.
Colony Assay to Screen MetK Mutations That Convert the GroE Requirement-E. coli MGM100 cells harboring both pMetEp-sfGFP and the metK genes, under the tac promoter cloned in the pACYC vector, were cultured at 37°C on LB plates with 50 g/ml ampicillin, 12.5 g/ml chloramphenicol, and 0.2% arabinose. Colonies were transferred onto gauze (BEMCOT, Asahi Kasei Fibers) to make replicas on LB plates with 1 mM diaminopimelate, 50 g/ml ampicillin, 12.5 g/ml chloramphenicol, 1 mM isopropyl ␤-D-thiogalactopyranoside, and 0.2% glucose. After an incubation at 37°C for 5 h, the GFP fluorescence of each colony was detected with a trans-illuminator (BioSpeed).

Systematic Chimeric Analysis between EcMetK and UuMetK-
The sequence similarities between EcMetK and UuMetK are very high, and they share 45% amino acid identity (Fig. 1A). In particular, the middle (M) domain, which corresponds to residues 221-301 of EcMetK, is well conserved, with 74% identical residues (Fig. 1A). We divided MetK into three domains, including the N-terminal and C-terminal sides of the M domain (N, M, and C domains, Fig. 1B). To identify which domains are critical for the GroE dependence of EcMetK, all combinations of chimeric MetKs (EEU, EUE, UEE, EUU, UEU, and UUE) were constructed (Fig. 1B).
To evaluate the in vivo GroE requirement of the MetK chimeras, we applied the method using the conditional GroE expression strain developed previously (13). This method utilizes the following characteristic: the obligate GroE substrates become aggregated or are degraded under the GroE-depleted conditions (13,22). Various MetKs, including the chimeras under the tac promoter, were expressed in the conditional GroE strain MGM100, in which the expression of GroE is controlled by the arabinose-inducible BAD promoter (22). To avoid potential aggregate formation due to MetK overexpression, leaky expression by the tac promoter was used in this study. The MetK proteins in the total and soluble fractions were separated by centrifugation and detected by Western blotting. EcMetK was degraded under the GroE-depleted conditions, whereas UuMetK remained in the soluble fraction even under these conditions (Fig. 1B). We also observed that the UuMetK expression suppressed the overexpression of MetE (data not shown) (13), which is one of the hallmarks of GroE-depleted E. coli (17)(18)(19), indicating that the soluble UuMetK was functional. These results confirmed that EcMetK is GroE-dependent but UuMetK is not, as described previously (13). We investigated the MetK chimeras using the in vivo GroE dependence assay and found that, strikingly, all of them were degraded under the GroE-depleted conditions, indicating that all of the chimeric MetKs depend on GroE for folding.
However, we noticed that several chimeras (e.g. EEU, EUE) tended to be degraded even under the normal GroE conditions, suggesting that their stabilities were lower than those of their authentic counterparts, EcMetK and UuMetK. These results revealed that at least one replacement of the GroE-independent UuMetK domain with the GroE-dependent EcMetK domain confers the GroE dependence on the chimeric MetKs in vivo. The results also implied that all three domains have determinants that confer the GroE dependence.
Single or Double Amino Acid Substitutions Are Sufficient to Convert UuMetK into an Obligate GroE Substrate-Next, we sought to identify the critical regions that confer the GroE dependence. We studied the UEU chimera because it shares 94% amino acid sequence identity with wild-type UuMetK (Figs. 1A and 2A). In the crystal structure of EcMetK (Protein Data Bank (PDB) code: 1fug), the M domain is composed of three ␤-sheets (␤8 -␤10) and two ␣-helices (␣5 and ␣6). Among these secondary structures, ␤-sheet 9 and ␣-helix 5 are identical between UuMetK and EcMetK. Therefore, we substituted the amino acids in the regions of the other three secondary structures (␤8, ␣6, and ␤10, Fig. 2A) with the corresponding EcMetK amino acids. Again, all three mutants 2A). In contrast, the variant altered in the residues from 247 to 249 (U-E 247-249 -U), which are not conserved between EcMetK and UuMetK and are not included in any secondary structures, did not change the GroE-independent property of UuMetK ( Fig. 2A).
Aggregation Propensities of GroE-dependent UuMetK Mutants-Almost all GroE obligate substrates show strong aggregation propensities, according to the global aggregation analysis under the chaperone-free conditions (4,13). Thus, we investigated the in vitro aggregation properties of representative UuMetK mutants, K215F and K215F/I216F, which were converted to GroE-dependent substrates, by a conventional light scattering assay. Under the conditions where EcMetK formed aggregates but wild-type UuMetK did not, the purified K215F and K215F/I216F UuMetK mutants aggregated (Fig.  3). The double mutant, K215F/I216F, was more prone to aggregation than the single K215F mutant (Fig. 3), suggesting an additive effect of each mutation on the aggregation propensity. These results suggest that the in vivo GroE dependence of UuMetK mutants correlates with the aggregation propensities of the proteins.
Screening System to Identify Other GroE-dependent UuMetK Mutants-The above results, which were restricted to the M domain, showed that point mutations, including single amino acid substitutions, convert the in vivo GroE requirement of UuMetK. Besides the M domain, the replacement of the N or C domain also converted the GroE requirement (Fig. 1B), indicating that the determinants that confer the GroE dependence on UuMetK are distributed throughout the entire UuMetK sequence. To identify these determinants, we developed a screening system to select the GroE-dependent UuMetK proteins from randomly mutated UuMetK proteins.
To evaluate the GroE requirement of the numerous UuMetK random mutants, we established a screening system that can distinguish the UuMetK mutants that confer GroE dependence from wild-type UuMetK, at the colony level (Fig. 4A). We focused on one of the E. coli phenotypes associated with the GroE depletion, the massive induction of methionine synthase (MetE) (13,(17)(18)(19). The MetE overexpression is induced by the activation of the metE promoter (metEp) due to the impaired folding of EcMetK under the GroE-depleted conditions (13,17). Briefly, the impaired EcMetK activity, induced by the GroE depletion, leads to the deficiency of S-adenosyl methionine, the product of the MetK enzyme reaction, which then activates the metE promoter (17). Because we previously showed that the MetE overexpression in GroE-depleted cells is suppressed by the UuMetK expression (13, 17), we reasoned that the conversion of UuMetK to a GroE-dependent substrate would activate the metE promoter. Therefore, we constructed a GFP reporter plasmid in which the expression of superfolder GFP (sfGFP (24)) was under the control of the metE promoter (21) (Fig. 4, A and B). The cells were also transformed with additional plasmids harboring randomly mutated UuMetK genes. We expected to observe green fluorescent colonies when the cells expressed the UuMetK mutants that are converted to the GroE obligate substrates, under the GroE-depleted conditions (Fig. 4, A and B).

Identification of Mutations in UuMetK That Confer GroE
Dependence-The expression of wild-type UuMetK in this system under the GroE-depleted conditions did not generate GFP fluorescence (Fig. 4C), confirming that UuMetK complemented the impaired folding of EcMetK due to the lack of GroE. Therefore, we expressed randomly mutated UuMetK proteins, generated by error-prone PCR using Mn 2ϩ , and evaluated their GroE requirements. We obtained 809 colonies in total and then selected 169 colonies that showed bright GFP fluorescence intensities. We should note that these 809 colonies are only a small fraction of the possible mutations generated by this mutation rate. Because the simple insertion of a stop codon or mutations in the UuMetK ORF that result in decreased MetK expression would lead to the activation of metE and the bright GFP fluorescence, we cultured all of the selected colonies individually and eliminated the false positives by multiple procedures, as follows. First, the cells that expressed full-length UuMetK mutants were selected by Western blotting. Second, the UuMetK DNAs were analyzed by PCR and restriction enzyme digestion. Third, to eliminate the UuMetK mutants that lost the S-adenosyl methionine synthesis activity, the UuMetK mutants that remained soluble upon isopropyl ␤-D-thiogalactopyranoside-induced overexpression under the GroE-depleted conditions were discarded because such mutants were expected to lack the S-adenosyl methionine synthesis function. After eliminating the false positives through these selections, we finally obtained 35 mutants of UuMetK that showed the GroE requirement for folding in vivo ( Table 1).
Characteristics of Mutations That Convert UuMetK into Obligate GroE Substrates-Among the obtained 35 mutants, 54% of them (19 out of 35 clones) had only a single amino acid substitution, and the others had two or three substitutions ( Table 1). The mutation positions were widely distributed throughout the three domains (N, M, and C), although the single mutants tended to accumulate in either the M-domain or the N-terminal region (Fig. 5A and Table 1).
Next, we analyzed the correlation of the mutation positions with the secondary structures in the MetK tertiary structure (Structural Classification of Proteins (SCOP) fold ID: d.130). The extremely high homology between UuMetK and EcMetK permitted us to generate a homology-modeled UuMetK structure (Fig. 5B), based on the crystal structure of EcMetK (PDB code: 1fug (25)). Mapping of the single mutations on the UuMetK structure revealed their dispersed distribution in the UuMetK structure (Fig. 5B). The mutations were not accumulated in secondary structure regions (Fig. 5, B and C). Instead, the mutations tended to be found in the flanking regions of the secondary structures, defined as the Ϯ2 residues from the boundary of the secondary structures; 10 out of 19 single mutations and 29 out of 54 all mutation points were found in the flanking regions (Fig. 5C). Furthermore, the mutation rates in the conserved residues between EcMetK and UuMetK were remarkably higher than the identity of these proteins (45%); the ratio was 63% (12 out of 19) in the case of single mutations and 63% (34 out of 54) in the case of total mutations.

DISCUSSION
Recent proteome-wide approaches to identify in vivo GroE substrates revealed that a subset of the proteome is obligate GroE substrates, which absolutely require GroE for correct folding. EcMetK, one of the obligate GroE substrates, has a highly homologous ortholog, UuMetK in Ureaplasma, which does not require GroE for folding in E. coli (13). In this study, we tried to convert the GroE-independent UuMetK to GroE dependence by chimera analysis, site-directed amino acid substitutions, and random mutagenesis. We found that independent multiple point mutations, and even single mutations distributed throughout the entire ORF, were sufficient to convert the GroE-independent UuMetK to GroE dependence.
In addition, our results showed that the mutations that converted the GroE dependence were not rare; at least ϳ5% of the amino acids in the ORF were identified as GroE determinants, although this might be underestimated because our random mutagenesis experiment was not saturated. Based on the multiple determinants of the GroE dependence distributed throughout the entire sequence, one of the main messages of this work is that the difference between the GroE-dependent and GroE-independent properties is marginal; a delicate balance determines the GroE dependence. Although more than half of the mutations that conferred GroE dependence are single amino acid substitutions (Table 1), we detected additive effects for the acquisition of the GroE requirement. For example, the single mutations of L291V, A292S, or F293Y in UuMetK did not affect the GroE dependence, although the solubility of L291V in the GroE-normal cells was slightly lower than that of wild-type UuMetK (Fig. 2B). However, the additional A292S or F293Y mutation converted the GroE-independent L291V mutant into a GroE-dependent protein, indicating that the accumulation of mutations had additive effects. This kind of additive effect on the GroE dependence might be related to the case of maltose-binding protein (MBP) mutants. The double mutant MBP(V8G/ Y283D) has a more stringent folding property as a GroE substrate than the single mutant MBP(Y283D) (26), although MBP has not been identified as an in vivo GroE substrate protein.
What properties are affected by the amino acid substitutions in the GroE-dependent UuMetK variants? One immediate possibility to convert the GroE dependence might be the acquisition of GroEL-binding motifs, due to the mutations. However, this is unlikely because no consensus GroEL-binding motif has been identified based on primary amino acid sequences, although several attempts have been made to identify sequence motifs for GroEL substrates (e.g. Refs. [27][28][29]. Instead, we raised the possibility that the GroE dependence is correlated with the aggregation propensity because GroEL assists in protein folding by suppressing the formation of the aggregates (e.g. Refs. 23 and 30). In fact, we previously found that in vivo obligate GroE substrates are strongly aggregationprone (13). Strikingly, the in vitro aggregation analysis clearly showed that the in vivo GroE dependence is closely correlated with the inherent aggregation propensity (Fig. 3), supporting the idea that the acquisition of aggregation propensity could cause the conversion to GroE dependence.
The next point to be discussed is the determinants that confer the aggregation propensities. A previous global aggregation analysis under chaperone-free conditions only revealed that the aggregation property is partly correlated with the topologies of the secondary structures (i.e. Structural Classification of Proteins (SCOP)) (4). The bias on a tertiary structure itself is not applicable to this study because this work only considered MetK. However, we noticed that the mutation sites were enriched in the flanking regions of the secondary structures (Fig. 5C). Although there is no plausible explanation for the enrichment, mutations in the vicinity of the secondary structures might affect the folding rate of MetK. For example, the slower folding rates caused by the mutations would prolong the lifetimes of the folding intermediates, which tend to self-associate and form aggregates. The prolonged folding rate would be compensated by GroE. Indeed, prolonged folding rate caused by a single point mutation of MBP, MBP(Y283D), results in a much higher stability of the MBP(Y283D)-GroEL complex than that of wild-type MBP (31). In addition, this scenario, in which GroEL accelerates the folding rate, is consistent with the results from a recent detailed analysis using DapA, an in vivo GroE obligate substrate (22), and its homologs, in which GroE accelerates the rate of TIM barrel domain folding by catalyzing segmental structure formation (16). Alternatively, the stability of the folding intermediates may be lower in the mutants. It was recently proposed (32) that GroEL stabilizes these impaired folding intermediates, which eventually leads to the acquisition of GroE dependence.  . Random mutagenesis analysis to convert UuMetK into obligate GroE substrates. A, the screening system to identify GroE-dependent UuMetK mutants. EcMetK encoded in the genome of the E. coli MGM100 strain is impaired under the GroE-depleted conditions, and consequently, the metE promoter is activated. UuMetK suppresses the metE activation, but when mutations convert UuMetK into the obligate GroE substrate, the suppression does not occur. The activation of the metE promoter was monitored by GFP fluorescence. B, a schematic illustration of the experimental procedure. E. coli MGM100 cells harboring the pMetEp-sfGFP plasmid were transformed with a randomly mutagenized library of UuMetK. After the formation of colonies under the GroEexpressing conditions, the colonies were transferred to LB agar plates containing 0.2% glucose, which suppresses the expression of the groESL gene regulated by the araBAD promoter. After an incubation for 5 h at 37°C, the fluorescence intensities were detected by an LAS-4000 FluorImager (Fujifilm). Amp, ampicillin; Cm, chloramphenicol; DAP, diaminopimelate; IPTG, isopropyl ␤-D-thiogalactopyranoside. C, pictures of the colonies and the GFP fluorescence under the EcMetK-and UuMetK-expressing conditions.
Most protein mutations are deleterious because the small margin of protein stability prevents the acquisition of a large fraction of mutations (33). It has been proposed that chaperones have facilitated protein evolution by buffering the deleterious mutations that cause impaired folding (34). This concept, which was originally developed from studies on Hsp90 chaperones, has now been extended to the chaperonin GroEL (35)(36)(37). In particular, directed evolution studies clearly revealed that GroE overexpression can buffer the destabilizing mutations, thereby promoting the folding of compromised proteins to improve the enzyme activity (32,36). In connection with our study, we would like to note that the evolved enzymes in GroEoverexpressing E. coli gained aggregation-prone properties and were converted to GroE-dependent proteins (36). The relationship between the aggregation propensity and the GroE dependence in the directed evolution experiments is strikingly consistent with that in our experiment, supporting the conclusion that the difference between the GroE-dependent and GroEindependent properties was marginal. The function of GroE to buffer the aggregation-prone mutations therefore plays a critical role in maintaining higher genetic diversity, eventually lead-ing to the expansion of the protein world, including metabolic networks (38).
Finally, one might ask about the possibility of the opposite direction of the conversion on the GroE dependence. Is it possible to engineer the EcMetK to become a GroE-independent protein? The answer would be yes in principle because our data show that a delicate balance determines the GroE dependence. However, the conversion of GroE-dependent substrates such as EcMetK into a GroE-independent protein might be difficult because destabilizing mutations might be accumulated in the GroE substrates during protein evolution. Studies converting GroE-dependent substrates into independent ones, which are in progress, would provide further insight about how the GroE dependence is determined. A, mutation sites were colored in red or yellow for single or multiple mutations, respectively, plotted on the ORF of UuMetK. a.a, amino acids. B, the single mutations described above were plotted on a tertiary structure model of UuMetK, obtained through the SWISS-MODEL Repository (SIB) and derived from the crystal structure of EcMetK (PDB code: 1fug) (25). The regions colored red and orange represent the mutation points obtained by the random mutagenesis (Table 1) and by the amino acid substitutions (Fig.  2B), respectively. C, enrichment of mutation positions in the conserved residues and the flanking regions of secondary structures. Shown is comparison of the ratios of the conserved residue, the secondary structure, and the flanking region of the secondary structure between all amino acid residues and the mutation positions that cause conversion to the GroE requirement of UuMetK. Dark gray bars show the ratios in 19 single mutations, and light gray bars show all 54 mutation positions, including the mutants with multiple mutations (***, p Ͻ 0.01, *, p Ͻ 0.05, binomial test).