Stimulating the Substrate Folding Activity of a Single Ring GroEL Variant by Modulating the Cochaperonin GroES*

In mediating protein folding, chaperonin GroEL and cochaperonin GroES form an enclosed chamber for substrate proteins in an ATP-dependent manner. The essential role of the double ring assembly of GroEL is demonstrated by the functional deficiency of the single ring GroELSR. The GroELSR-GroES is highly stable with minimal ATPase activity. To restore the ATP cycle and the turnover of the folding chamber, we sought to weaken the GroELSR-GroES interaction systematically by concatenating seven copies of groES to generate groES7. GroES Ile-25, Val-26, and Leu-27, residues on the GroEL-GroES interface, were substituted with Asp on different groES modules of groES7. GroES7 variants activate ATP activity of GroELSR, but only some restore the substrate folding function of GroELSR, indicating a direct role of GroES in facilitating substrate folding through its dynamics with GroEL. Active GroELSR-GroES7 systems may resemble mammalian mitochondrial chaperonin systems.

The Hsp60 chaperonin family is ubiquitous among the three kingdoms of life. The paradigm Escherichia coli GroEL is the well characterized Hsp60 chaperonin that, together with cochaperonin GroES, uses ATP to assist a variety of proteins to fold in the cell (1)(2)(3)(4)(5)(6). Both GroEL and GroES are essential for cell viability (7). Fourteen identical GroEL subunits assemble into a cylindrical structure of two heptameric rings stacked back to back, with a functional central cavity at either end (8). Each GroEL subunit is organized into three domains. The apical domains, located on the opening of the rings, form the main substrate-binding site and interact with GroES. The equatorial domains, situated in the center of the cylindrical GroEL, provide most of the intra-ring and all the inter-ring subunit interactions and contain the nucleotide-binding site. The intermediate domains connect the apical and the equatorial domains, transmitting signals including substrate binding and nucleotide effect across the subunits. The surface lining the GroEL central cavity is mainly hydrophobic, and this hydrophobic character has been shown to be important for capturing and interacting with misfolding protein substrates (9 -12). The homoheptameric GroES adopts a dome-like assembly with each monomer folding into a ␤-barrel structure flanked by a structurally flexible loop termed the GroES mobile loop (13,14). In the presence of nucleotides, GroES forms a stable asymmetric bullet-shaped GroEL-GroES complex (see Fig. 1A inset) (15)(16)(17)(18). The crystal structure of GroEL-ADP 7 -GroES shows that binding of GroES converts the GroEL central cavity to an enclosed chamber (termed the cis complex) (19). Three hydrophobic GroES residues (Ile-25, Val-26, and Leu-27), of the GroES mobile loop, interact with residues from the GroEL apical domain (see Fig. 1A), and the GroES-GroEL interface is largely hydrophobic. Binding of GroES to one GroEL ring transforms the surface of the GroEL central cavity from hydrophobic, favorable for binding misfolded substrate proteins, to folding promoting hydrophilic.
The double ring assembly of GroEL is essential for the GroEassisted protein folding reaction. The substrate protein initiates folding within the isolated folding favorable environment upon displacement from the wall of the GroEL central cavity due to GroES association and is released outside the GroE capsule following GroES departure. Both dissociation and association of GroES are ATP-driven and are mediated by the intricate and highly coordinated cross-ring allostery. ATP hydrolysis in the GroES-bound GroEL ring (cis ring) weakens GroES-GroEL interaction (20); however, the hydrolysis product, GroES-Gro-EL cis -ADP 7 -GroEL trans , is still very stable with a dissociation constant (K d ) of 0.1-26 nM (16,(21)(22)(23). To finally discharge GroES from the cis ring, thereby releasing the folding substrate, binding of ATP to the other (trans) GroEL ring is required (20,24,25). The newly ATP-charged trans ring, which is prone to GroES association, becomes a "new" folding active cis chamber, whereas the collapsing cis chamber assumes inactive state. Thus, in assisting protein folding, the two GroEL rings communicate to alternate between folding active and inactive states, and GroEL functions as a two-stroke molecular machine (1, 26 -28).
Consistent with the essential role of the two ring architecture of GroEL in cycling the folding active cis chamber, a single ring (SR) 2 version GroEL (29), termed GroEL SR in our study here, is functionally impaired. Although GroEL SR and GroES may form the folding chamber for substrate proteins to undergo and complete folding, substrate proteins are not released from the chaperonin system (20, 30 -32). Dissociation of the GroEL SR -GroES complex is very slow (the half-life t1 ⁄2 of Gro-EL SR -GroES is ϳ300 min) (29), due to the lack of the allosteric effect originating from ATP binding to the absent trans ring (above). As a result, the highly stable GroEL SR -GroES complex traps the folding substrate proteins, stalling the chaperonin reaction cycle. Thus, GroEL SR is unable to substitute GroEL for cell growth (33).
Recently, active single ring chaperonin systems have been reported. The folding active GroEL SR -GroES systems consist of variants of GroEL SR , identified from genetic analysis (34 -37), or variants of GroES, identified from random mutagenesis study on the GroEL-binding GroES mobile loop (36). Reduced GroEL-GroES interaction has been implicated in these folding active GroEL SR -GroES variants; however, the basis for the mutational effect is not immediately clear. For example, the affected GroEL residues are not located on the GroEL-GroES interface and most likely interfere with the ATP-driven, highly coordinated structural transitions that are obligated in the GroE reaction. Intriguingly, the weakened GroEL-GroES interaction in these folding active single ring chaperonin systems is reminiscent of mammalian mitochondrial chaperonin mtHsp60-mtHsp10 system. mtHsp60 exists as a single heptameric ring in solution (38), and may function as single ring (23,39), although a mechanism involving a transiently associated double ring has been proposed (40,41). Moreover, the mtHsp60-mtHsp10 interaction appears to be transient in the presence of ADP (23).
In this study, we directly explored the notion that weakened GroEL-GroES interaction may restore the substrate folding function of GroEL SR . We took a unique strategy to create GroES variants with various affinities for GroEL in a systematical and controlled manner. We constructed a covalently linked groES variant, groES 7 , so that we could independently substitute residues in individual GroES subunits. We examined these GroES 7 variants for their function with both the double ring GroEL and the single ring GroEL SR . In addition to the mechanistic understanding of how a GroEL ring assists protein folding, studies of active GroEL SR -GroES systems may provide useful information on the mechanism of mtHsp60-mtHsp10 in mammalian mitochondria.

EXPERIMENTAL PROCEDURES
Construction of groES 7 and Its Mutants-To construct the covalently connected groES 7 , groES was amplified with primers including the desired restriction site sequences, to generate seven groES containing DNA fragments: groES1 (with NdeI and KpnI), groES2 (with KpnI and XhoI), groES3 (with XhoI and SalI), groES4 (with SalI and SacI), groES5 (with SacI and BamHI), groES6 (with BamHI and EagI), and groES7 (with EagI and HindIII). groES1 and groES2 were digested with KpnI, purified, and ligated to generate the groES1-groES2 fragment, which was further amplified by PCR for a sufficient quantity. Fragments of groES3-groES4 and groES5-groES6-groES7 were generated in a similar manner, and so was the groES1-…-groES7 fragment, which was digested with NdeI and HindIII and inserted into a modified pET28b vector, whose region between HindIII and Bpu was replaced with a short DNA oligonucleotide. To generate GroES 7 mutants, a mutation was first introduced into the regular groES construct using QuikChange kit (Stratagene). The mutated groES was amplified with primers including the desired restriction site sequences and ligated to a TOPO vector (Invitrogen). The circularized TOPO vector was transformed into TOP10 cells for amplification and digested with specific restriction enzymes to generate the groES mutant gene with defined restriction cuts, which was inserted via the corresponding sites into the groES 7 pET28b construct. DNA sequencing was performed on the entire groES 7 gene to confirm that mutation was only on the designated module(s). When expressed, GroES 7 has a His 6 tag on the N terminus of the protein.
Protein Expression and Purification-groES 7 pET28b was transformed into BL21DE3 for overexpression. Cells were grown in LB medium until A 600 reached 0.6, induced with 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside, and continued for 4 -5 h. Cells were re-suspended in 10 mM sodium phosphate, pH 7.5, 500 mM NaCl, and 10 mM imidazole, and lysed using a microfluidizer (Microfluidics). The clear cell lysate was loaded onto a Ni affinity column, and GroES 7 was eluted with a linear gradient of imidazole (5-500 mM). The GroES 7 -containing fractions were combined, concentrated, dialyzed to 50 mM Tris-Cl, pH 7.5, and loaded onto a FastQ column (GE Healthcare), and GroES7 was eluted with a linear gradient of NaCl (0 -1 M). The GroES 7 -containing fractions were combined, concentrated, dialyzed to 50 mM Tris-Cl, pH 7.5, and loaded onto a Superdex 200 gel filtration column (GE Healthcare).
Cell growth and protein purification of GroEL and GroEL SR were similar, as described in Ref. 42. A purification step using 30% methanol followed by an additional size exclusion chromatography was included to ensure high quality of GroEL and GroEL SR , and tryptophan fluorescence confirmed the absence of bound peptides. Protein expression and purification of wildtype and GroES mutants were based on the previous procedures (19).
ATPase Assay-The steady-state ATP hydrolysis rate was measured using the malachite green assay (43). GroEL or GroEL SR was added to TEA buffer (50 mM triethanolamine, pH 7.5, 50 mM KCl, and 20 mM MgCl 2 ) to a final concentration of 0.125 M (tetradecameric GroEL) or 0.25 M (heptameric GroEL SR ). The final concentration of GroES (heptamer) and GroES 7 proteins was 0.3 M. The solution was incubated at 25°C for 10 min. The hydrolysis was initiated by addition of 100 mM pH 7.0 ATP to a final concentration of 2 mM and followed every 2 min for 12 min using the malachite green assay.
Malate Dehydrogenase (MDH) Refolding Assay-MDH was unfolded in TEA buffer including 3 M GdmHCl to a final concentration of 36.7 M (monomeric MDH). To refold MDH, 2.75 l of unfolded MDH was diluted at 1:100 (v/v) to a final volume of 275 l of refolding solution (at 30°C) containing 50 mM TEA, pH 7.4, 50 mM KCl, 1 mM ATP, and 1 M tetradecameric GroEL (or 2 M heptameric GroEL SR ) with 4 M or without GroES. At desired time intervals, 20 l of reaction solution was removed and mixed with 1 ml of NADH assay solution (50 mM Tris-Cl, pH 7.4, 10 mM DTT, 0.2 mM NADH, 1 mM ketomalonate), and absorption at 340 nm was taken to monitor the decrease of NADH (44). As a positive control, activity of 367 nM native MDH (monomeric concentration) was measured at the same time intervals and was taken as 100% activity. Refolded and active MDH converts NADH to NAD ϩ .

Essential Role of GroES Ile-25 and Leu-27-To decrease
GroEL-GroES interaction, the three GroEL-interacting residues, Ile-25, Val-26, and Leu-27, on the GroEL-GroES interface (Fig. 1A) were individually substituted to Asp, resulting in GroESI25D, GroESV26D, and GroESL27D variants. When incubated with GroEL in the presence of ADP, both GroESI25D and GroESL27D variants could not form a stable complex with GroEL; they were largely absent by SDS-PAGE analysis on the GroEL-containing fractions (supplemental Fig. S1A). Interestingly, the GroESV26D variant, albeit containing a total of seven substitutions in heptameric form, could still form the GroEL-GroES complex, as suggested by SDS-PAGE (supplemental Fig.  S1A). Previous studies show that in forming a complex with GroEL, GroES inhibits the ATP hydrolysis rate of GroEL by ϳ50% (16,21,45,46). Neither GroESI25D nor GroESL27D variants altered GroEL ATP hydrolysis significantly (supplemental Fig. S1B), whereas the GroESV26D variant reduced GroEL ATPase function in a manner similar to the wild-type GroES. These findings are consistent with the model that both GroESI25D and GroESL27D variants did not associate with GroEL, whereas GroESV26D formed a stable complex with GroEL. To further confirm the mutational effects, the GroES variants were subjected to a substrate protein folding study. The spontaneous folding of MDH is low (ϳ5%), and the complete chaperonin system, including GroEL, GroES, and ATP, is required to achieve efficient MDH folding (47), indicating that formation of the folding active GroEL-GroES chamber is essential for MDH folding. Supplemental Fig. S1C shows that neither GroESI25D nor GroESL27D supported GroE-assisted MDH folding, suggesting that both variants were not able to form a functional complex with GroEL. As expected, folding of MDH using GroESV26D was as efficient and effective as using the wild-type GroES, indicating formation of the GroESV26D-GroEL complex. Taken together, these findings confirm the critical function of GroES residues Ile-25 and Leu-27 in GroEL-GroES interaction and indicate a minor role of GroES Val-26 in complex formation.
We next examined whether the observed mutational effects of I25D, V26D, and L27D on GroEL-GroES interaction were applicable to the GroEL SR -GroES system. GroES inhibits the ATP hydrolysis rate of GroEL SR by up to ϳ90% (29). As shown in supplemental Fig. S1B, GroESV26D decreased the ATPase activity of GroEL SR to a level comparable with that of wild-type GroES. In contrast, both GroESI25D and GroESL27D did not affect the ATP hydrolysis rate of GroEL SR . These results are parallel to those using GroEL, further substantiating the importance of Ile-25 and Leu-27, in contrast to the minimal role of Val-26, in GroEL-GroES interaction. These observations are consistent with results of sequence analysis on 31 GroES homologous proteins from prokaryotes to eukaryotes (48). Ile and Leu are highly conserved at position 25 (96% of sequences) and position 27 (94%), respectively, whereas the amino acid at position 26 is more diverse: Val (48%), Ile (29%), Leu (10%), Met (7%), Phe (3%), and Gln (3%).
When assaying for MDH folding activity, we found that none of the three mutants was able to cooperate with GroEL SR to assist MDH folding; however, the inactivation mechanisms are different. Considering the similar binding affinities of GroESV26D and GroES for GroEL, the observed limited MDH folding using GroESV26D should be most likely due to the folding MDH trapped within the nondissociable GroESV26D-Gro-EL SR complex. On the other hand, given the abolishing effect of I25D or L27D mutation on GroEL-GroES interaction, the minimal MDH activity observed with GroESI25D or GroESL27D and GroEL SR is most likely due to the lack of formation of the folding chamber, which is essential for MDH folding (47). Thus, straightforward manipulations on the GroEL-interacting residues of Ile-25, Val-26, and Leu-27 cannot confer the single ring GroEL SR the sustainable substrate folding activity because the required enclosed folding chamber either is nonrecyclable (in the case of GroESV26D) or does not form (in the case of GroESI25D and GroESL27D). Alternatively, for GroESV26D, it might also be likely that its presumably decreased affinity for GroEL, although subtle, could not completely displace the protein substrate into the central cavity for folding (see "Discussion").
Modulating GroES Affinity for GroEL-The abolished GroEL-GroES interaction observed in GroESI25D and GroESL27D may be explained by the 7-fold amplified mutational effect because a single gene modification in groES transforms into seven amino acid substitutions in the GroES heptamer. We reasoned that if some of the seven GroES subunits maintained the wild-type sequence, the resulting GroES variants would display different levels of reduced affinities for GroEL. We hypothesized that these GroEL-binding compromised GroES variants might cooperate with GroEL SR to complete the chaperonin reaction cycle.
To modify an individual GroES subunit, we took advantage of the observation that the N and C termini of neighboring GroES subunits in heptameric GroES are within 5 Å (C␣-C␣ distance) (19) (Fig. 1B) and created a synthetic gene, groES 7 , concatenating seven copies of the groES gene (Fig. 1C). The expressed protein, GroES 7 , consists of seven GroES modules, corresponding to the seven GroES subunits in the canonical GroES homoheptamer, in one contiguous polypeptide chain. Linker sequences of four amino acids (consisting of Gly, Ser, or Ala) give flexibility between subunits. Distinct restriction sites are designed to flank each groES gene, so that modifications can be introduced into the specific position(s) of the desired module(s). We used groES 7 to incorporate I25D or L27D substitutions in specific GroES modules, to create GroES 7 variants with a range of weakened affinities for GroEL.
GroES 7 Behaves Like GroES-To examine the effect of the peptide linkers tethering neighboring GroES modules, we compared the structural and functional aspects of GroES 7 with the canonical GroES heptamer. GroES 7 (theoretical molecular mass of 76,721 Da) migrated as a molecular species with the molecular size of 82.2 kDa estimated from gel filtration chromatography, whereas the molecular size of the canonical GroES heptamer (theoretical molecular mass of 72,723 Da) was estimated to be 71.3 kDa by gel filtration. Like GroES, GroES 7 formed a complex with GroEL in the presence of nucleotides, as assayed by SDS-PAGE analysis ( Fig. 2A). GroES 7 decreased ATPase activity of GroEL to a level comparable with that using GroES (Fig. 2B). Moreover, GroES 7 assisted MDH refolding in a manner similar to GroES, with comparable reaction kinetics and refolding yield (Fig. 2C). The resemblances of GroES 7 to GroES were extended to their interaction with GroEL SR . As shown in Fig. 2B, GroES 7 diminished ATPase activity of Gro-EL SR to a level similar to that using GroES (ϳ10% of the intrinsic rate of GroEL SR ). Like GroES, GroES 7 could only support the minimal MDH folding by GroEL SR (Fig. 2C). Taken together, these data show that the covalently linked single chain GroES 7 is similar to the wild-type GroES in all the measures examined. Studies using the covalently linked GroES have been  Fig. S1A. B, effect of GroES 7 on the steady-state ATPase activities of GroEL and GroEL SR . The intrinsic ATP hydrolysis rate of GroEL was set as 100% for the GroEL group, and that of GroELSR was set as 100% for the GroEL SR group. S.D. values (error bars) are calculated from multiple (at least three) experiments using proteins from different purifications. C, GroES 7 in GroEL-and GroEL SR -mediated MDH folding. At least three independent experiments were carried out using proteins from different batches of purification. Shown here are representative MDH refolding reactions.
reported, and the additional peptide linkers are found not to have functional interference (49,50).
Compared with the completely depleted inhibition of GroESI25D or GroESL27D on GroEL ATPase activity, substitutions of I25D and L27D in GroES 7 generated a range of effects on GroEL ATP hydrolysis rates. In both the GroES 7 I25D and GroES 7 L27D series, GroEL ATPase activity increased with the number of substitution (Fig. 3A). When three substitutions were incorporated into GroES 7 as in GroES 7 I25D 1,4,7 and GroES 7 L27D 1,4,7 , the inhibition changed from 50% to 10 -20% of the intrinsic GroEL ATP hydrolysis rate. Because a large inhibition may presumably be correlated with the strong GroEL-GroES interaction, these findings suggest that the substituted GroES 7 variants display different affinities for GroEL and that they do not interact with GroEL as strongly as the wild-type GroES. As expected, V26D substitutions in GroES 7 did not alter GroES function, as variants of GroES 7 V26D affected GroEL ATPase activity in a manner similar to GroES (supplemental Fig. S2A).
Given that variants of GroES 7 I25D and GroES 7 L27D inhibited GroEL ATPase activity to various extents, we next investigated how the altered ATP hydrolysis rate of the modified chaperonin system influenced MDH folding. A single substitution of GroES 7 I25D 1 slowed the MDH folding rate but with a slightly decreased folding yield over a time course of 60 min (Fig. 3B). A further substitution, GroES 7 I25D 1,4 , decelerated the rate of MDH folding drastically and reduced the final yield by ϳ28% (60% compared with 77% using the wild-type system). Folding of MDH became modest when using GroES 7 I25D 1,4,7 , the tri-substituted GroES 7 variant. The observations that both the folding rate and the final yield of MDH declined with the increasing number of the affected GroES modules were also found in the GroES 7 L27D series (Fig. 3C). Note that MDH folding appears to be more sensitive to I25D substitution in GroES 7 than to L27D substitution; with the same number of affected modules, variants of I25D substitution resulted in greater reduction of both MDH folding rate and folding yield than those of L27D substitution. These findings suggest that residue Ile-25 plays a more important role than Leu-27 in MDH folding, although their differential effect on GroEL ATP hydrolysis rate was not evident (Fig. 3A). As expected, the mono-, di-, and tri-V26D substitutions in GroES 7 did not affect MDH folding (supplemental Fig. S2C).
Increase of ATPase Rate in GroEL SR -GroES 7 Systems-Given that variants of GroES 7 I25D and GroES 7 L27D decreased inhi-bition on ATP hydrolysis of GroEL, we next investigated whether these variants restored the diminishing ATPase activity of GroEL SR -GroES. A single I25D substitution in GroES 7 alleviated the GroES-stalled ATP hydrolysis of GroEL SR (ϳ7%) to ϳ18% of the intrinsic rate of GroEL SR (Fig. 4A). The GroES 7 I25D 1,4 variant with two substitutions further relieved the cochaperonin-induced inhibitory effect, so that GroEL SR -GroES 7 I25D 1,4 hydrolyzed ATP at a rate 58% that of GroEL SR . Given that GroEL and GroEL SR hydrolyze ATP at the same rate, the ATPase activity of GroEL SR -GroES 7 I25D 1,4 was comparable with that of the canonical GroEL-GroES system. Interestingly, an additional I25D substitution did not further relieve the inhibition, and GroES 7 I25D 1,4,7 reduced the ATPase activity of GroEL SR to a similar extent, 56%, as GroES 7 I25D 1,4 . In comparison, the regained ATP hydrolysis activity appeared to increase with the number of L27D substitution, with 22, 51, and 73% GroEL SR activity for systems using GroES 7 L27D 1 , GroES 7 L27D 1,4 , and GroES 7 L27D 1,4,7 , respectively (Fig. 4A). Also note that GroES 7 L27D 1,4 regulated GroEL SR ATP hydrolysis at a rate similar to GroES regulation of GroEL. The elevated ATP hydrolysis rates of GroEL SR observed using variants of GroES 7 I25D and GroES 7 L27D suggest efficient turnover of the folding chamber of the single ring GroEL SR , or dissociation of GroEL SR and GroES 7 at a biologically relevant time frame. As expected, V26D substitutions in GroES 7 did not effectively revive the ATP hydrolysis rate of the system (supplemental Fig.  S2B), suggesting that GroEL SR remained arrested in the highly stable complex state.
Stimulating Substrate Folding Activity of GroEL SR -With the regained ATP hydrolysis using variants of GroES 7 I25D and GroES 7 L27D, the GroEL SR folding chamber presumably has a revived turnover cycle and might support MDH folding. An increase of folding yield to 32% was observed using the single I25D-substituted GroES 7 I25D 1 over a 60-min time course, from ϳ15% of GroEL SR -GroES (Fig. 4B). Most strikingly, the MDH folding yield continued to improve to Ͼ50% during the 60-min assay period when using the double-substituted GroES 7 I25D 1,4 . This improved MDH yield achieved by the single ring system corresponded to ϳ70% of the maximum folding yield (ϳ77%) using the canonical double ring GroEL-GroES system. It is noted that the MDH folding accelerated drastically after the first 10-min delay (Fig. 4B). An additional I25D substitution, GroES 7 I25D 1,4,7 , however, completely abolished the regained MDH folding activity. Intriguingly, of the three GroES 7 L27D variants tested, only GroES 7 L27D 1,4 improved MDH folding substantially with a final folding yield of ϳ30% (Fig. 4C). GroES 7 L27D 1 , which allowed ATP to hydrolyze at a rate much higher than GroES(or GroES 7 I25D 1 ), did not increase MDH folding significantly compared with GroES. MDH folding in the presence of GroES 7 V26D variants and Gro-EL SR was not improved compared with that using GroES-Gro-EL SR (supplemental Fig. S3B). Our observations that substitutions at Ile-25 in GroES 7 are more effective than those at Val-26 or Leu-27 in activating MDH folding activity of GroEL SR parallel results from genetic studies (36). In their work, Liu et al. substituted each of Ile-25, Val-26, and Leu-27 to 19 other amino acids and assayed the abilities of these GroES variants and Gro-EL SR to support cell growth. They found that only GroES variants with Ile-25 substitutions may collaborate with GroEL SR to support cell growth.

DISCUSSION
Among the various aspects that the chaperonins may contribute to substrate protein folding efficiency, the distinctive ability to provide a folding favorable isolated environment (the cis folding chamber) for protein folding is fundamental (51). Presumably, efficient turnover of the folding chamber is critical FIGURE 4. Effects of GroES 7 I25D and GroES 7 L27D variants on GroEL SR . A, steady-state ATPase activity. B and C, MDH folding activity. For the ATPase experiments, the intrinsic GroEL SR ATP hydrolysis rate was set as 100%. Experiments were repeated at least three times, and S.D. values (error bars) are show. For the MDH folding assay, activity of native MDH was set as 100%. Experiments were repeated multiple times, and data of representative runs are shown here.