INTRODUCTION

Inside the cell, the folding of many proteins requires the assistance of a class of proteins known as molecular chaperones (1-3). The best known members of this family are the chaperonins, oligomeric complexes of approximately 60 kDa/subunit, which are present in both prokaryotic and eukaryotic organisms. The most studied protein of this group is GroEL from Escherichia coli, a 14-mer of approximately 800 kDa that requires for its folding activity the presence of ATP, Mg²⁺, K⁺, and the cochaperonin GroES (2). The GroEL oligomer has a cylindrical shape and is composed of two identical heptameric rings enclosing a cavity where unfolded polypeptides can bind (4-6). The structure of GroEL has been solved at atomic resolution (7), showing that each monomer is composed of three domains: an apical domain that forms the end of the cylinder and is involved in substrate and GroES binding, an equatorial domain that provides most of the intra- and inter-ring contacts and that contains the ATP binding site, and an intermediate domain that connects apical and equatorial domains and acts as a hinge region that allows the conformational variability of the apical domains (8, 9). The structure of GroES, an heptamer of 10-kDa subunits, is also known at atomic resolution (10), where each monomer consists of a -barrel with two -hairpin loops. Under physiological conditions, GroEL and GroES form two types of ATP-dependent complexes: an asymmetric complex where one GroES binds to one end of the GroEL cylinder (11) and a symmetric complex where two GroES molecules are simultaneously binding both ends of the GroEL oligomer (12-15). Recent data indicate that the functional unit in the GroEL oligomer is each one of the GroEL cavities that are capped by the GroES oligomer (16, 17), which suggests that symmetric GroEL-GroES complexes may allow the GroEL oligomer to use both GroEL cavities simultaneously for folding. In fact, biochemical data correlate the presence of symmetric complexes with enhaced folding rates (18-22). On top of this, electron microscopy studies have shown that symmetric GroEL-GroES complexes can contain two substrate molecules simultaneously in both GroEL rings (6, 22), confirming the functional significance of these type of complexes in the reaction cycle.

The GroEL oligomer shows positive cooperativity in ATP binding within each ring and negative cooperativity between the two rings (23), so that there is an equilibrium between three different allosteric states, TT, TR, and RR corresponding respectively to free GroEL, GroEL with only one ring containing nucleotide, and GroEL containing nucleotide simultaneously in both rings. In the presence of ATP or AMP-PNP,¹ the apical domains move upward and outward, as shown by electron microscopy studies (8, 9). The conformational changes associated with each one of the three different allosteric states have also been analyzed (9). Nevertheless, these conformational changes have not been visualized in the crystal structure of ATPS-bound GroEL (24), possibly due to the lack of inter-ring communication in the GroEL mutant used to obtain the crystal (25) and/or the fact that the ATPS may cause different conformational changes than ATP (9).

Current models for the chaperonin-assisted folding propose that unfolded polypeptide binds to an open end of the GroEL oligomer, and after ATP-dependent GroES binding, the substrate is released inside the sequestered environment of the GroEL-GroES cage, where polypeptide is given a chance to fold (2, 17). Taking this into account, it might seem that GroEL could work as a single-ring protein, but in the wild type protein both rings are required to complete the reaction, mainly because GroES release (essential to discharge the sequestered substrate and allow a new cycle of polypeptide and GroES binding) requires the presence of ATP in the opposite ring (26).

Both the negative cooperativity and the ATP-dependent GroES release are a consequence of the communication between the two rings that form the GroEL oligomer. There are also additional biochemical data that reinforces the existence of communication between the two GroEL rings: (a) conformational changes in GroEL occurring after ATP binding promote release of bound substrates. Surface plasmon resonance studies (27) indicate that nucleotide binding to the first ring of GroEL leads to moderate rates of substrate release, whereas subsequent ATP binding on the second ring of GroEL gives rise to a much higher rate of release. (b) After GroES binding to one GroEL ring to form an asymmetric complex, the second ring seems to bind a second GroES molecule to form a symmetric complex with lower affinity (14). Nevertheless, the structural bases for this inter-ring communication remain essentially unknown. Some features regarding the signaling pathway can be inferred from the GroEL x-ray structure (7), where two sites of contacts between each subunit with each partner of the opposite ring are observed. Nothing is known, however, of the different behavior of both rings due to the inter-ring communication. It could be speculated that as a consequence of some kind of signaling between the two GroEL rings, each one undergoes structural rearrangements that sets them in a different conformation. In this work, we have characterized by electron microscopy and image processing a conformational change that is strictly dependent on the signal coming from the opposite GroEL ring. Biochemical analysis correlates the presence of this change with previously described phenomena that are known to be dependent on the inter-ring communication.

MATERIALS AND METHODS

GroEL mutants were produced according to Weissman et al. (28) for the single-ring mutant (SR1) and Fenton et al. (29) for the A126V mutant. Both mutants were generated by the homologous recombination technique (30). Polymerase chain reaction was carried out using Pfu polymerase (Stratagene) on the plasmid pAR3GRO (31).

E. coli GroEL and GroES were purified from a pAR3 plasmid harboring strain (31) that overexpresses both GroEL and GroES as described previously (13). Both SR1 and A126V mutants were purified using the same procedure that was used for wild type GroEL.

ATPase activity was assayed using malachite green to measure the amount of inorganic phosphate released on ATP hydrolysis. The reactions were started by the addition of GroEL or SR1 (final oligomer concentration, 0.25 and 0.5 µM, respectively) to the assay solution containing 50 mM Tri-HCl, 10 mM MgCl₂, 100 mM KCl, pH 7.5, and 2 mM ATP. At different times, aliquots of the mixture were made to react with the malachite green reagent and allowed to stand for 2 h at room temperature for the color to develop (32). Concentrations of inorganic phosphate were determined by using a standard solution of phosphate under the same conditions. Alternatively, a spectrophotometric method that includes an ATP regenerating system was used to characterize the ATPase activity of both proteins (33). Both methods gave the same results, indicating that the components of the ATP regenerating system do not interact with the GroE machinery.

Rhodanese was denatured in 8 M guanidinium chloride or 6 M urea and refolded as described previously (13) using GroEL, SR1, or A126V. All the reactions were carried out at 37 °C. For the experiments where ATP was needed, an ATP regenerating system was used as described (18).

Mitochondrial MDH (from porcine heart, Sigma) was denatured in 6 M urea at 37 °C during 30 min and diluted 100-fold in the buffer containing GroEL or SR1 also at 37 °C. The activities of MDH were assayed by measuring the decrease in absortion at 340 nm of NADH in 50 mM Tris-HCl, pH 7.5, with oxalacetate as substrate. For the experiments using ATP, an ATP regenerating system was used as described (18).

Mixtures of either GroEL or the A126V mutant with SR1 in the presence of radiolabeled rhodanese were fixed with glutaraldehyde and loaded into a size exclusion chromatography Superose 6 HR 10/30 (Pharmacia Biotech Inc.) connected to an HPLC system and equilibrated with 50 mM Tris-HCl, pH 7.5, and a flow rate of 0.250 ml/min. Rhodanese was labeled with ¹²⁵I (¹²⁵I-sodium carrier-free; Amersham, Little Chalfont, UK) by the chloramine-T method (34). For cross-linking, samples were incubated in 0.08% (w/v) glutaraldehyde (Sigma) for 20 min at 37 °C. The cross-linking reaction was stopped by adding ammonium chloride (final concentration, 40 mM).

Samples were negatively stained with 1% uranyl acetate on thin coated collodion grids previously glow discharged for 15 s. Transmission electron microscopy was performed in a JEOL 1200EX-II electron microscope operated at 120 KV. Images were digitized, and all the side views observed in the images were extracted without any a priori selection and centered using a synthetic mask. Centered particles were aligned using a free pattern algorithm (35, 36). The particles thus aligned were classified by assignment to dictionary output vectors (37) that were afterward further classified using multivariate statistical analysis (38). When the coordinates of the images for the first autovector were plotted against any of the others, the particles could be separated in different populations as described previously (6, 9). After classification, homogeneous populations (at the centered state) were aligned again to prevent the influence of the particles assigned to a different group in the previous alignment. Resolution of the final average images was estimated by the spectral signal to noise ratio method (39, 40), and the average images were filtered in each case to the resolution obtained.

RESULTS

To study the existence of conformational changes directly associated to the inter-ring communication in the GroEL double-toroid, a single-ring mutant was generated to avoid the influence of one ring upon the other. This mutant, named SR1 (28), contains four point mutations in residues that make the major contacts between the two rings. Weissman et al. (17) have shown that SR1 was able to fold rhodanese upon GroES binding, but the folded protein was not released from the mutant cavity probably because the absence of a signal coming from the nonexistent opposite ring abolished GroES release (17). However, Hayer-Hartl et al. (41) found that the single-ring mutant was able to fold rhodanese at rates similar to those of GroEL in the presence not only of ATP but also of ADP. To evaluate the mechanistic significance of the signaling between the two GroEL rings in polypeptide folding, the ATPase activity of both GroEL and SR1 in the presence or the absence of GroES (Fig. 1), and the ability to refold denatured substrates (Fig. 2) was analyzed. Regarding the ATPase activity (Fig. 1), the addition of GroES to GroEL in the presence of ATP and K⁺ caused a 43% decrease in the ATP hydrolysis, in accordance with previous results (42), something that has been interpreted as a consequence of GroES binding to GroEL. On the other hand, SR1 showed rates of ATP hydrolysis similar to those of GroEL in the absence of GroES, but in the presence of the cochaperonin the ATPase activity of SR1 is strongly reduced (84%). Therefore, under these experimental conditions, GroES dissociation from the SR1-GroES complex is considerably slowed down in accordance with previous results (17). Neither GroEL nor SR1 showed significant differences in their GroES-mediated ATP hydrolysis inhibition when unfolded rhodanese was present (Fig. 1, closed squares and circles with dashed lines). Nevertheless, some authors have found that under these same conditions SR1 is able to fold rhodanese, apparently in the absence of ATP hydrolysis (41). Therefore, to further characterize the SR1 mutant, some folding experiments were carried out (Fig. 2). Equal protein amounts of GroEL and SR1 and therefore an equal amount of "functional cavities" were compared in their capacity to fold denatured rhodanese (both with urea or guanidinium chloride as denaturant) under functional conditions (GroES, ATP, and K⁺; Fig. 2A). Surprisingly, SR1 was able to fold at rates comparable with those of GroEL, regardless of the fact that under those conditions ATP hydrolysis is almost inhibited (Fig. 1). Accordingly, when the folding reaction was carried out in the presence of ADP instead of ATP, SR1 showed similar folding kinetics as in the presence of ATP. In fact, Hayer-Hartl et al. (41) have found that unfolded rhodanese is able to induce GroES cycling, eliminating the requirement of an ATP-dependent signal from the opposite ring to release GroES and allowing the substrate to be discharged so that a new substrate and GroES binding can take place.

[View Larger Version of this Image (20K GIF file)]

[View Larger Version of this Image (16K GIF file)]

A comparison between the folding kinetics of rhodanese by GroEL and SR1 (Fig. 2A) showed that although SR1 displayed similar hyperbolic folding kinetics in the presence of both ATP and ADP, GroEL showed hyperbolic folding for ATP and a much slower kinetics for ADP. This could be explained considering that in the case of SR1, all the rings can be used for folding, and rhodanese seems to eliminate the requirement for ATP hydrolysis (41). On the other hand, GroEL incubated in the presence of ADP cannot form symmetric GroEL-GroES complexes (14, 20), making this way only half of the rings present in the assay available to bind GroES and therefore capable of folding rhodanese. The fact that faster kinetics were found with GroEL in the presence of ATP and high K⁺ can be interpreted as a consequence of the ability of GroEL, under those conditions, to form symmetric GroEL-GroES complexes (12-14), allowing both GroEL rings to be used simultaneously for folding (6, 22). These results give further support to the idea that under functional conditions, symmetric complexes are a common intermediate in the reaction. The discrepancy between these results and the ones described by Hayer-Hartl et al. (41) could be a consequence of the use of low K⁺ concentrations in their kinetics experiments comparing ATP and ADP mediated folding, because more physiological K⁺ concentrations (above 100 mM) are known to be required for a high level of symmetric complex formation (14, 20).

Although rhodanese seems to have the capacity of inducing ATP hydrolysis-independent GroES cycling, the question that arises is whether this is a general mechanism or a specific property of this substrate. As a first approximation, the same refolding experiments were carried out using MDH instead of rhodanese (Fig. 2B). SR1 was unable to fold denatured MDH in the presence of either ATP or ADP, whereas GroEL refolded MDH only in the presence of ATP. This indicates that in the absence of a "substrate-induced" GroES release, both ATP and double-ring structures are required to complete the folding reaction, and this suggests that the rhodanese-induced GroES discharge may not be a general mechanism for chaperonin-mediated folding.

The conformational changes of GroEL (Fig. 3) and SR1 (Fig. 4) induced upon nucleotide and GroES binding were analyzed by electron microscopy and image processing and compared in an attempt to find and characterize conformational rearrangements related to the signaling between the two GroEL rings. Because it has been found that the conformational changes undergone by GroEL upon nucleotide binding can be detected by electron microscopy by following changes in the shape of the apical domains depicted in the side views of GroEL (8, 9), a similar method was used in this work. When GroEL was incubated with an AMP-PNP concentration adequate to fill both GroEL rings (the RR allosteric state (23)), they showed an outward movement of their apical domains (Fig. 3B) in comparison with free GroEL (Fig. 3A). These conformational changes are similar to the ones previously described (8, 9). When GroEL was incubated under physiological concentrations of ATP and K⁺, similar conformational changes were observed (results not shown). Interestingly, the extent of the conformational changes on both GroEL rings were different (Fig. 3B), giving rise to an asymmetric particle that has been previously described for GroEL in the RR allosteric state for ATP and AMP-PNP (9). Thus, three different conformational states for a GroEL ring can be defined: (a) a "closed" state, for nucleotide-free GroEL (Fig. 3A); (b) an "open" state (top ring of Fig. 3B) in which the apical domains move upward and outward so that the ring takes a trapezoidal shape instead of the rectangular one found in a nucleotide-free GroEL ring. This open state has been previously described when only the first GroEL ring is occupied by nucleotide (TR allosteric state (9)); and (c) a "fully open" state, in which the apical domains move also outward but to a much larger extent than in the open state (bottom ring of Fig. 3B), similar to what is found when GroES binds to GroEL (Ref. 9 and Fig. 4, D and E).

[View Larger Version of this Image (62K GIF file)]

[View Larger Version of this Image (85K GIF file)]

The SR1 mutant was also incubated with AMP-PNP (Fig. 4, B and D) and ATP (Fig. 4, C and E) in the absence (Fig. 4, B and C) or presence (Fig. 4, D and E) of GroES. In the presence of both ATP and AMP-PNP (3 mM nucleotide), the SR1 oligomer takes the shape of the open conformational change detected in GroEL (Fig. 4, B and C), which is different from the nucleotide-free SR1 (Fig. 4A). As a control, SR1 was also incubated with 10 mM AMP-PNP or ATP and again only the open conformation was found (result not shown). After GroES binding (Fig. 4, D and E), the rings adopt the fully open state, showing that the SR1 mutant has not been affected in its ability to bind GroES.

All the open conformational changes found in GroEL and SR1 in the presence of ATP or AMP-PNP were compared among them (top ring of Figs. 3B and 4) and against the fully open state found in GroEL (bottom ring of Fig. 3B) by calculating the differences between all the final averages using the same procedure previously described (9). From this analysis, it was confirmed that all the open states found in SR1 and GroEL share a common structure that is clearly distinct from the conformation found in the fully open ring of GroEL (results not shown).

All these observations suggest that there is a conformational change in GroEL, consisting in large apical domain movements (the fully open state) that can only be achieved after the ring has received some kind of signal from the opposite ring (already occupied by nucleotide), giving rise to a structurally asymmetric particle. Nucleotide binding-induced conformational changes have been related to a decrease in the affinity of GroEL toward unfolded substrates (27), which differently affects both GroEL rings. The observed structural asymmetry may be responsible for the asymmetric behavior of GroEL in substrate (27) and GroES binding (14).

The effect of the open and fully open conformation on denatured rhodanese binding was tested using HPLC size exclusion chromatography (Fig. 5). These two conformations were preformed by incubating a mixture of GroEL and SR1 (equimolar in terms of number of rings) with AMP-PNP. Radioactively labeled denatured rhodanese was then added to analyze the relative affinities of the substrate toward the two types of oligomers. First, a control was carried out to see whether GroEL and SR1 showed similar binding affinity toward substrate in the absence of nucleotides, so that a direct effect of the mutations of SR1 in the apical domains could be ruled out. When equal amounts of GroEL and SR1 (that is, the same number of cavities and therefore putative substrate binding sites; see drawing in Fig. 5A) were incubated in the presence of radioactively labeled unfolded rhodanese, fixed with glutaraldehyde, and loaded into a gel filtration column, two peaks were obtained (Fig. 5A). Each peak was analyzed by native gel electrophoresis and electron microscopy, showing that the first peak corresponded to wild type GroEL and the second peak corresponded to the single-ring mutant (results not shown). The radioactivity/mass ratio of the two peaks indicated that substrate bound with approximately the same affinity to the GroEL or SR1 cavities. As a control, when the sample was not fixed with glutaraldehyde before loading in the HPLC, a similar result was obtained (not shown).

[View Larger Version of this Image (19K GIF file)]

A mixture of GroEL and SR1 (equimolar in terms of the number of rings) was then incubated in the presence of 3 mM AMP-PNP with radioactively labeled denatured rhodanese. The use of AMP-PNP is more convenient than ATP to generate a stable conformational state, because upon ATP hydrolysis different intermediates of the reaction can occur. At the AMP-PNP concentration used, GroEL binds AMP-PNP in only one of its two rings (TR allosteric state (9)), whereas SR1 adopts its nucleotide-bound state (R state; Fig. 4C). The sample was fixed and loaded into the HPLC (Fig. 5B), and the two peaks were analyzed by native gel electrophoresis and electron microscopy (results not shown), which confirmed the presence of GroEL in the TR state for the nucleotide (first peak) and SR1 in the R state (second peak). As in the control experiment, radioactivity measurements indicated that substrate bound with approximately the same affinity to all the cavities present in the assay. This means that the open conformational change present in GroEL modifies the affinity for substrate at a similar extent as the open state in SR1. When the same GroEL/SR1 mixture was incubated in 10 mM AMP-PNP (Fig. 5C), GroEL showed by electron microscopy the presence of an open conformation in one ring and a fully open conformation in the other ring (Fig. 3B), whereas an open conformation was found in SR1 (Fig. 4B) (results not shown). Interestingly, the radioactivity/mass ratio was approximately 2:1 in favor of the single-ring mutant. Therefore, the large rearrangements in the apical domains at the fully open state dramatically reduced the affinity of this ring toward the substrate when compared with a SR1 mutant that can only generate an open conformational change, due to the lack of inter-ring communication.

The affinity of GroES to bind to the open and fully open states of the GroEL rings was also compared using electron microscopy. GroEL was incubated with different concentrations of AMP-PNP and limited amounts of GroES. The percentage of each species (free GroEL and the GroES-bound forms) could be estimated by electron microscopy because under our experimental conditions mostly side views were obtained (20). Although this technique does not provide precise measurements, it indicates the relative affinity of GroEL and SR1 toward GroES. Again, the use of AMP-PNP instead of ATP allows analysis of a stable particle rather than a continuous reaction where GroES is being bound and discharged from GroEL. A first control experiment was carried out to detect whether the binding affinities of the open conformations of GroEL and SR1 toward GroES were similar. GroEL and SR1 were mixed in a 1:1 oligomer molar ratio, and limited amounts of GroES (1:1 oligomer molar ratio with SR1, not enough GroES oligomers to cap all the rings present in the solution) were added in the presence of 3 mM AMP-PNP. When analyzed by electron microscopy, all the GroEL molecules showed an open conformational change in only one of the two rings of the oligomer, whereas the nucleotide-free ring was in a closed conformation (TR state). All the SR1 molecules showed an open state (R state). As summarized in the Scheme I, three different possibilities could be presented depending on the relative affinity of the SR1 ring and the nucleotide-bound GroEL ring toward GroES. Given the limited amount of GroES, if the nucleotide-bound GroEL ring had higher affinity toward GroES than the SR1 ring, most of the GroES oligomers would be bound to the GroEL ring in the R state (option A). If the SR1 ring had higher affinity toward GroES than the nucleotide-bound GroEL ring, most of the GroES oligomers would be bound to the SR1 ring (option B). Finally, if both SR1 and the nucleotide-bound GroEL ring had similar affinity toward GroES, the cochaperonin would cap the SR1 ring and the nucleotide-bound GroEL ring with the same probability (option C). As shown in Table I, the percentage of GroES-bound particles was similar for SR1 and GroEL, indicating that GroES had distributed randomly between the two oligomers. Therefore, it is deduced that GroES binds with approximately the same affinity to GroEL in the TR state and to SR1 in the R state.

Scheme I.

Table I. GroES binding to the open and fully open conformations GroEL (oligomer final concentration, 0.8 µM) was mixed with SR1 (oligomer molar ratio, 1:1 GroEL:SR1) and GroES (oligomer molar ratio, 1:1 GroEL:GroES) in the presence of 3 mM AMP-PNP. Type of side views Percentage of side views GroEL SR1 Free oligomers 57.8  ± 4.5 54.7  ± 4.5 Asymmetric complexes 42.2  ± 4.4 45.3  ± 4.5 Symmetric complexes 0.0

To test the differences in affinities between the open and the fully open conformations in GroEL toward GroES, similar experiments were carried out. GroEL was incubated with 10 mM AMP-PNP, so that one ring was in an open conformation and the other in a fully open conformation (RR state (9)) (Table II). To this mixture, a limited amount of GroES oligomers was added (1:1 GroEL/GroES oligomer molar ratio; not enough GroES to cap all the GroEL rings). Three different behaviors could be expected (Scheme II): (a) the fully open state binds GroES with lower affinity than the open state. Thus, GroES would bind preferentially to the open ring of the GroEL oligomers to form asymmetric GroEL-GroES complexes, and only afterward, if free GroES was still be present, it would bind to an asymmetric complex to form a symmetric GroEL-GroES complex. Thus, the majority of the population would be asymmetric GroEL-GroES complexes and few symmetric complexes, and free GroEL would be found (option A); (b) the fully open state binds GroES with higher affinity than the open state (option B) so that two GroES oligomers would tend to bind to the GroEL oligomer to form symmetric complexes. Thus, the majority of the population would be a mixture of symmetric GroEL-GroES and free GroEL (because of the limited amount of GroES); (c) the fully open state binds GroES with the same affinity as the open state (option C), so that an equimolar mixture of asymmetric complexes, symmetric complexes, and free GroEL would be found. Table II shows that asymmetric GroEL-GroES complexes represent most of the population, indicating that the fully open conformation binds GroES with lower affinity than the open state. It has been previously described that inter-ring communication is responsible for the decrease of GroEL affinity toward the binding of a second GroES molecule (14). The movement of the apical domains in a fully open conformation, as a consequence of the signaling coming from the opposite ring, gives a structural basis to those observations; when the apical domains move farther outward and upward, they bind GroES with a reduced affinity.

Table II. GroES binding to the open and fully open conformations in GroEL and A126V GroEL or the mutant A126V (oligomer final concentration, 0.8 µM) were mixed with GroES (oligomer molar ratio, 1:1 GroEL or A126V:GroES) in the presence of 10 mM AMP-PNP. Type of side views Percentage of side views GroEL A126V Free oligomers 19.6  ± 2.9 31.2  ± 3.9 Asymmetric complexes 71.8  ± 3.7 27.5  ± 3.6 Symmetric complexes 8.5  ± 1.4 41.3  ± 4.4

Scheme II.

To further characterize the relationship between inter-ring signaling and the appearance of the fully open conformation, a GroEL mutant with affected inter-ring communication was designed using the information provided by the x-ray structure of the GroEL double mutant R13G/A126V (7, 29). Aharoni and Horovitz (25) have shown that this double mutation in GroEL affects the negative cooperativity between the two rings with respect to ATP so that both rings switch independently from the T to the R state. Both mutations, by looking the double mutant x-ray structure, seemed to be good candidates to be involved in the modification of the inter-ring signaling. However, because the properties of the R13G mutant regarding the ATPase activity, GroES and poypeptide binding, and protein folding are similar to that of native GroEL (29), it can be deduced that the A126V mutation is the one involved in the disruption of the signal between both GroEL rings. A chaperonin with the A126V mutation was then generated to study the effect of this mutation on the structural and functional asymmetry of GroEL.

The conformational changes undergone by the A126V mutant after AMP-PNP binding to both rings (Fig. 3C) were analyzed by electron microscopy and image processing as described for wild type GroEL (Fig. 3B). A homogeneous population of A126V oligomers with both rings filled with nucleotide was prepared by incubation with 10 mM AMP-PNP. To make sure that both GroEL rings contained nucleotide (RR state) and because GroES binding is nucleotide-dependent, the same sample used for image processing was incubated with GroES, and a homogeneous population of symmetric A126V-GroES was obtained (results not shown). When analyzed, the A126V mutant in the RR state was found to undergo an open conformational change in both rings, and no fully open conformation was observed at all (Fig. 3C). The open conformation in both rings of the mutant was similar to the one found in one of the rings of wild type GroEL (top ring of Fig. 3B) and in the SR1 mutant (Fig. 4, B and C). A comparison between this average image and all the previous ones was carried out by calculating their differences, and the similarity between the open conformation in A126V and the open conformation found in SR1 and GroEL was confirmed (results not shown). It is reasonable to suggest that the absence of the fully open conformation in any of the A126V rings can be related with the lack of inter-ring cooperativity found in the double mutant R13G/A126V (26). It can also be suggested that some kind of inter-ring communication (absent in the A126V mutant) is required to induce the fully open state in the apical domains. The affinity of the A126V mutant in the RR state toward unfolded rhodanese was tested by comparing it with the SR1 mutant in the open state (R state). Unlike the GroEL oligomer in the RR state (Fig. 5C), no decrease in rhodanese binding affinity was observed (Fig. 5D). This suggests that (in the absence of the fully open conformation) both rings of the A126V mutant have the same affinity toward substrate. Similarly, GroES binding to the A126V mutant was also analyzed as in the case of wild type GroEL by incubating A126V in the RR state in the presence of a limited amount of GroES (Table II). In contrast to the result obtained with GroEL, where mostly asymmetric GroEL-GroES complexes were found (indicating a preferential binding to the one of the two GroEL rings), in the case of A126V an almost equimolar mixture of free GroEL, asymmetric and symmetric A126V-GroES complexes was obtained. Therefore, both A126V mutant rings seemed to bind GroES with the same affinity. Taken together, these results confirm that the fully open conformation of the native GroEL oligomer in the RR state is responsible for the decrease in the affinity toward substrate and GroES.

To further characterize the A126V mutant, the folding of denatured rhodanese was tested in the presence of both ATP and ADP. The results obtained showed hyperbolic kinetics in the presence of ATP and slower kinetics in the presence of ADP (results not shown), similar to what was found for wild type GroEL and in contrast to what was obtained for SR1 (Fig. 2A). This suggests that some kind of signaling is maintained between the rings of the A126V mutant, which was confirmed when no A126V-GroES symmetric complexes were obtained by incubating A126V with GroES and 10 mM ADP (results not shown). MDH folding assays were carried out with the A126V mutant, and the results obtained revealed that the mutant was able to fold MDH with behavior similar to that of wild type GroEL (results not shown). In the single-ring mutant, the absence of inter-ring communication prevented GroES recycling and abolished MDH refolding (Fig. 2B). Because MDH renaturation is obtained with the A126V mutant, it can be deduced that GroES recycling is occurring, and therefore, some kind of communication between the rings is taking place, albeit not enough to induce the fully open conformation in one of them.

DISCUSSION

Recent biochemical experiments (16, 17) indicate that the 7-mer GroEL ring is the functional unit in protein folding. After polypeptide binding to an open GroEL ring, GroES binding releases the substrate inside the GroEL-GroES cage where folding can occur. Nevertheless, both rings are required to complete the cycle, because at least the GroES discharge requires some kind of signal coming from the opposite GroEL ring (26). In fact, this work shows that a single-ring mutant, which lacks the inter-ring signal, is unable to release and rebind GroES and therefore to fold substrates like MDH. Nevertheless, rhodanese has been described to induce GroES binding and release in the absence of ATP hydrolysis (41), probably reflecting a specific effect of this substrate. From the results obtained when comparing the rhodanese folding rates using GroEL and SR1 in the presence of ATP and ADP (Fig. 2A), it can be deduced that under functional conditions symmetric GroEL-GroES complexes enhance folding rates by being able to use both cavities simultaneously to fold substrate (6, 22).

Apart from biochemical data, there is not much information on the structural bases for the transmission of information between the two GroEL rings. Nevertheless, it could be speculated that one ring modifies the other ring conformation so that the last one shows a different behavior toward the ligands. By comparing the nucleotide-promoted conformational changes in GroEL and a single-ring mutant (lacking the inter-ring communication effects) (Figs. 3 and 4), three distinct ATP and AMP-PNP induced conformational changes have been detected, according to the extent of the movements in the apical domain: closed, open, and fully open conformations. The fully open conformation, which was strictly dependent on the signal coming from the nucleotide-filled opposite GroEL ring, binds substrate (Fig. 5) and GroES (Table II) with lower affinity than the closed and open states, which correlates well with the data from other authors on the effect of allosterism in substrate (27) and GroES binding (14).

The atomic structure of GroEL has been obtained using a double mutant (R13G/A126V) (7, 43). This mutant is functional in ATP hydrolysis and protein folding (29) but has been found to lack negative cooperativity between the two GroEL rings (25). The results obtained with the A126V mutant could explain the disruption of the inter-ring cooperativity and allows proposal of an inter-ring signaling pathway involved in the generation of the fully open conformation. This substitution generates a more stable structure because V126 (in -helix 5, according to (43)) establishes hydrophobic interactions with Leu⁴²⁶ and Leu⁴²⁹, which are located in the loop between -helices 14 and 15 and close to amino acid Glu⁴³⁴. This residue, located at the beginning of -helix 15, is involved in the contact between subunits of different rings (7). The stabilization of this region of the GroEL structure by the A126V mutation could partially affect the inter-ring communication by generating a more rigid structure. The comparison of the structures in the RR state of both GroEL and A126V oligomers (Fig. 3, B and C) confirms that the A126V mutation affects the inter-ring signaling (as in the double mutant R13G/A126V) and supports the relationship between inter-ring communication and the fully open conformation. The disruption in the inter-ring signaling is partial, however, because although the A126V mutant is unable (unlike native GroEL) to induce the fully open conformation in one of the rings when both of them are filled with AMP-PNP, it behaves in other ways like native GroEL because: (a) symmetric GroEL-GroES complexes cannot be obtained in the presence of ADP and (b) GroES is released in the presence of ATP, as indicated by the folding of MDH. Therefore, several pathways of signaling between the two rings may exist.

The crystal structure of GroEL shows that the inter-ring contacts are made through two groups of residues (7), each one interacting with a different subunit of the opposite ring. The first of these groups of residues is composed of amino acids Arg⁴⁵², Glu⁴⁶¹, Ser⁴⁶³, and Val⁴⁶⁴, and their mutations give rise to the stable single ring studied here. The other group of residues is composed of amino acids Lys¹⁰⁵, Ala¹⁰⁹, and Glu⁴³⁴, which seem to communicate with the ATP binding site through the -helix 4 (residues 89-108). This -helix has been proposed to be one of the possible pathways in the communication to the opposite subunit of the conformational changes induced by ATP binding (8). We can speculate that the asymmetric oligomer obtained in conditions close to those found in vivo, in which one of the rings is in the open state and the other in the fully open state, could be formed as follows: the signal generated in the nucleotide binding pocket (residues 86-90) by ATP binding to one of the rings would be transferred through the -helix 4 to residues Lys¹⁰⁵, Ala¹⁰⁹, and Glu⁴³⁴ (or some of them) and then to the contacting residues of the opposite subunit (the interaction of Glu⁴³⁴ and Lys¹⁰⁵ may be a good candidate), which would communicate the signal through the equivalent -helix of that subunit to the nucleotide binding pocket where nucleotide is bound. The signal would be transferred to the apical domain through movements in the hinge regions of the intermediate domain, generating in this ring the fully open conformation described in this work. The mutation of residues 452, 461, 463, and 464 may disrupt this pathway by generating a stable single ring, although it cannot be ruled out that the inter-ring communication could be performed through these mutated residues.

The results obtained with ATP and AMP-PNP (Ref. 9 and this work) show that the fully open state appears only when both rings contain nucleotide. In the case of ATP, under conditions similar to those found in vivo and while ATP is being hydrolyzed, a large percentage of the GroEL side views observed by electron microscopy correspond to asymmetric particles in which one of the rings is in a fully open conformation. This correlates with a functional asymmetry in substrate and GroES binding. We can speculate that because both rings are filled with nucleotide (a situation that is present in vivo), ATP binding (26) in one ring would induce the release of GroES and the caged polypeptide in the other ring by generating a fully open conformation that has less affinity toward GroES and substrate. This fully open conformation would be maintained in this ring until the ADP and P_i generated in the previous hydrolysis is replaced by ATP. After binding of new polypeptide and GroES molecules, ATP binding in this ring would induce the fully open conformation in the other ring and the cycle would continue. The asymmetry of the GroEL rings, which is maintained by nucleotide-exchange and inter-ring communication, would only reflect the structural consequences of the "switching" mechanism between the two GroEL rings, which would allow them to be always at a different stage of the folding cycle, thus behaving like the cylinders of an engine.