Biochemical Characterization of Symmetric GroEL-GroES Complexes EVIDENCE FOR A ROLE IN PROTEIN FOLDING*

When chaperonins GroEL and GroES are incubated under functional conditions in the presence of ATP (5 m M ) and K (cid:49) (150 m M ), GroEL-GroES complexes appear in the incubation mixture, that are either asymmetric (1:1 GroEL:GroES oligomer ratio) or symmetric (1:2 GroEL: GroES oligomer ratio). The percentage of symmetric complexes present is directly related to the [ATP]/[ADP] ratio and to the K (cid:49) concentration. Kinetic analysis shows that there is a cycle of formation and disappear- ance of symmetric complexes. A correlation between the presence of symmetric complexes in the incubation mix- ture and its rhodanese folding activity suggests some active role of these complexes in the protein folding process. Accordingly, under functional conditions, symmetric complexes are found to contain denatured rho- danese. These data suggest that binding of substrate inside the GroEL cavity takes place before the symmet- ric complex is formed. Under physiological conditions, spontaneous folding of polypeptides is often inefficient, due to competing aggregation pathways. A group of proteins, termed chaperonins, have been shown to assist the folding of polypeptides in vitro and in vivo (Ellis , 1987). Chaperonins belong to an ubiquitous group of proteins that shows a high degree of sequence identity (Hartl They are large oligomeric complexes which are divided into two families: the Hsp60 family present in bacteria and eukaryotic organelles and the Tcp-1 family of proteins (distant homologues of Hsp60) found in archaebacte-ria and eukaryotic cell cytosol Then, denatured (equimolar to was added and the folding was carried out as described previously. Isolation of Symmetric GroEL-GroES of symmetric

Under physiological conditions, spontaneous folding of polypeptides is often inefficient, due to competing aggregation pathways. A group of proteins, termed chaperonins, have been shown to assist the folding of polypeptides in vitro and in vivo (Ellis et al., 1987). Chaperonins belong to an ubiquitous group of proteins that shows a high degree of sequence identity . They are large oligomeric complexes which are divided into two families: the Hsp60 family present in bacteria and eukaryotic organelles and the Tcp-1 family of proteins (distant homologues of Hsp60) found in archaebacteria and eukaryotic cell cytosol (Gupta, 1995). The best known member of the bacterial and organelle chaperonins is GroEL from Escherichia coli. GroEL is a 14-mer of approximately 800 kDa whose atomic structure has already been resolved at 2.8-Å resolution (Braig et al., 1994). The GroEL cylinder binds unfolded polypeptides into its central cavity (Braig et al., 1993). This fact suggests that GroEL functions by providing a cavity where the substrate can fold properly thus avoiding aggregation, being released later in a native form. Nevertheless, recent studies seem to indicate that several rounds of ATP-hydrolysis dependent binding and release of the substrate are necessary for folding (Weissman et al., 1994). The conformation of the GroEL-bound polypeptide resembles that of the molten globule state (Martin et al., 1991).
For the chaperonin-dependent folding of many substrates, GroEL requires the presence of GroES, a heptameric ring of about 70 kDa (Chandrasekhar et al., 1986). GroES binds GroEL in the presence of nucleotides. GroEL residues essential for GroES binding have been determined by mutational analysis and location inside the GroEL crystal structure . The same residues involved in binding GroES are part of the polypeptide-binding site. Therefore, GroES may act in protein folding by directly displacing bound polypeptide by competition. In the presence of GroES, the hydrolysis of ATP by GroEL, which is K ϩ dependent, does not proceed linearly and distint phases can be discerned (Todd et al., 1993). The ATP hydrolysis by GroEL is coupled to cycles of release and rebinding of GroES from GroEL (Martin et al., 1993). Rebinding of GroES may occur either to the substrate-bound or the substrate-free ring of GroEL.
Under functional conditions, GroEL and GroES form two distint types of complexes (Llorca et al., 1994;Harris et al., 1994;Azem et al., 1994;Schmidt et al., 1994;Todd et al., 1994) where one ring of GroES binds to one side of the GroEL oligomer (GroEL:GroES 1:1, asymmetric complexes from now on) or one GroES binds to each side of the GroEL oligomer (GroEL: GroES 1:2, symmetric complexes from now on). Initially, only asymmetric particles were described as part of the functional holo-chaperonin (Langer et al., 1992). Not much is known about the symmetric complex and its possible role in the chaperoninassisted protein folding (Todd et al., 1994;Engel et al., 1995;Hayer-Hartl et al., 1995). A biochemical analysis of the symmetric GroEL-GroES complexes has been carried out. Some evidences suggesting the active role of symmetric GroEL-GroES complexes in protein folding have been obtained. Folding experiments show that both symmetric and asymmetric complexes are always present in the incubation mixture under functional conditions and that, regardless of the percentage of each type of complex, initial folding rates are similar, suggesting that both complexes could be implicated in the folding cycle. The presence of substrate in the symmetric complexes has been tested using labeled unfolded rhodanese. Despite some suggestions indicating that binding of unfolded substrate to symmetric complexes occurs in the outer surface of the chaperonin (Azem et al., 1994), it has been demonstrated that GroEL binds substrate inside its central cavity (Braig et al., 1994). Unfolded substrate present in symmetric complexes probably binds within their central cavity when in the form of asymmetric complexes since, coupled to ATP hydrolysis, GroEL continuously binds and releases GroES (Martin et al., 1993).

EXPERIMENTAL PROCEDURES
Purification of GroEL and GroES-E. coli chaperonins were obtained from a pOF39 plasmid harboring E. coli strain that overexpresses both GroEL and GroES (Fayet et al., 1986), as described previously (Llorca et al., 1994).
Sample Preparation and Electron Microscopy of GroEL-GroES Complexes-GroEL (0.3 M final concentration) and GroES (1:2 molar ratio) were incubated in 50 mM Tris-HCl, pH 7.5, with different concentrations of MgCl 2 , KCl, and ATP as indicated in each experiment. Complexes were incubated at room temperature. For experiments with labeled rhodanese, complexes were obtained after cross-linking as indicated below. GroEL-GroES complexes were negatively stained with freshly made 2% uranyl acetate on thin carbon-coated collodion grids previously glow-discharged for 15 s. Transmission electron microscopy was performed in a JEOL 1200EX electron microscope operated at 120 kV.
Rhodanese Refolding Assay-Rhodanese from bovine liver (Sigma) was denatured in 6 M guanidinium chloride. Refolding assays were carried out according to Llorca et al. (1994). For some experiments, final guanidinium chloride concentration in the refolding assay was reduced from the standard 100 mM down to 10 mM. 125 I Labeling of Rhodanese-Rhodanese was labeled with 125 I ( 125 Isodium carrier-free; Amersham, Little Chalfont, UK) by the chloramine-T method, according to Hunter (1978).
Cross-linking of GroEL and GroEL-GroES Complexes with Glutaraldehyde-Oligomers of GroEL (0.2 M final concentration) either alone or with 125 I-labeled rhodanese (equimolar to GroEL) were incubated in 50 mM Tris-HCl, pH 7.5, in the presence of 0.08% (w/v) glutaraldehyde (Sigma) for 20 min at 37°C. The cross-linking reaction was stopped by adding ammonium chloride (40 mM final concentration).
Native PAGE of GroEL and GroEL-GroES Complexes-Electrophoresis of cross-linked proteins under native conditions was carried out in slab gels containing 4.5% polyacrylamide in TBE buffer. Gels were stained with Coomassie Brilliant Blue R-250 (Sigma). When needed, radioactivity gels were dried and Konica x-ray films were exposed. Band quantification was carried out by scanning of the film in a Color OneScanner (Apple) and processing of the image using the public domain NIH image program.

Formation of Symmetric GroEL-GroES
Complexes under Different Incubation Conditions-GroEL tetradecamer is able to bind simultaneously two heptamers of GroES in the presence of ATP building what has been called symmetric complexes (Llorca et al., 1994;Harris et al., 1994;Azem et al., 1994;Schmidt et al., 1994;Todd et al., 1994). It has been previously shown that both asymmetric and symmetric complexes appear under folding conditions, whereas only asymmetric complexes are found in the presence of ADP or ATP␥S 1 (2 mM, final concentration) (Llorca et al., 1994). As not much is known about the conditions upon what they are formed, the dependence of symmetric particle formation on ATP, Mg 2ϩ , or K ϩ concentration was studied. GroEL and GroES were incubated at room temperature for 15 min with ATP, Mg 2ϩ , and K ϩ , each one at two concentrations, a relatively low and a relatively high concentration, as indicated in Table I. Samples were then observed by electron microscopy (Fig. 1). It is important to note that under the conditions used for sample preparation, only a small percentage of the total particles (around 10%) were circular front views and the rest corresponded to the three types of side views: GroEL, asymmetric and symmetric complexes. This fact allowed the estimation of the percentage of the different structural classes in the population. While no relevant effect of ATP and Mg 2ϩ concentration was observed under the concentration range tested in the experiment, the formation of the symmetric complex showed a clear dependence on K ϩ concentration: although symmetric complexes were present either under low (5 mM) or high (150 mM) K ϩ concentrations, they were more abundant under high K ϩ concentration ( Table I).
Effect of pH on the Formation of Symmetric GroEL-GroES Complexes-It has been recently proposed that symmetric particles are present only at high pH and that they are essentially absent under more physiological conditions . Although the standard conditions of the experiments were pH 7.5, the effect of changing the pH in the formation of symmetric GroEL-GroES complexes was tested. GroEL and GroES were mixed and dialyzed against either 50 mM Tris-HCl, pH 7.0, or 50 mM Tris-HCl, pH 8.0. Afterwards, 5 mM ATP (final concentration) and 150 mM KCl (final concentration) were added to each sample and incubated for 15 min, after which they were observed by electron microscopy. In both cases, the percentage of side views corresponding to symmetric complexes was around 30% of the total number of side views, a similar 3 0 5 0 Ϯ 1 F IG. 1. Representative field of a micrograph from a negatively stained incubation mixture. Only a small percentage of front views was observed (approximately below 10%). Three structural classes of side views could be seen, corresponding to the GroEL oligomer (circle), asymmetric GroEL-GroES complexes (circle, arrowheads), and symmetric particles (circle, arrows). Bar corresponds to 50 nm. value to that obtained at pH 7.5 (Table I), indicating that the formation of symmetric GroEL-GroES complexes seemed to be not pH-dependent in the pH 7-8 range.
Kinetics in the Formation of Symmetric Complexes-To know whether the formation of the symmetric complexes was a dynamic process in which the percentage of the GroEL-GroES system varied in time, or instead a more static process in which the amount of symmetric particles remained unchanged throughout the time, a kinetic study of the symmetric complexes was carried out. Fig. 2 shows the effect of the incubation time of a GroEL-GroES reaction mixture on the percentage of symmetric complexes over the total side views observed by electron microscopy. In the presence of low K ϩ concentration (5 mM), symmetric GroEL-GroES complexes accounted for approximately 10% of the total side views observed. This percentage was independent of the presence or absence of denatured rhodanese as substrate for folding and remained quite stable for 60 min. Longer incubation times led to a progressive disappearance of the symmetric complexes. In the presence of a high K ϩ concentration (150 mM), an increasing proportion of symmetric complexes was found that reached around 80% after a 45-min incubation. Afterwards, the percentage of symmetric particles was progressively reduced until disappearance at longer incubation times. As for low K ϩ concentration, this result was essentially independent of the presence or absence of folding substrate in the reaction mixture.
Stabilization of the Symmetric GroEL-GroES Complexes-After long incubation times (around 90 min) of GroEL and GroES under conditions leading to the formation of symmetric complexes, the GroEL-GroES system reached a state where no symmetric particles could be found while there was a concomitant increment of asymmetric particles (Fig. 3A). Different approaches to obtain a stabilization of the symmetric complexes were tested. After incubating GroEL and GroES until a high percentage of symmetric particles was obtained (45 min in Fig. 3A), the incubation mixture was subjected to three treatments and the proportion of symmetric GroEL-GroES complexes was estimated by electron microscopy. When ADP was added to the system (30 mM final concentration), all symmetric particles disappeared with a parallel increment in asymmetric particles (Fig. 3B). When non-hydrolyzable ATP␥S was added (15 mM final concentration), a similar result to that of Fig. 3B was obtained (Fig. 3C). However, when EDTA was added to the incubation mixture (80 mM final concentration), the percentage of symmetric complexes in the whole population remained stable for a long period of time (more than 1 h) (Fig. 3D).
Regeneration of Symmetric GroEL-GroES Complexes after Long Incubation Times-After incubation times longer than 90 min no symmetric particles were found in the GroEL-GroES incubation mixture ( Fig. 2 and Fig. 3A). After disappearance of symmetric complexes, several experiments were carried out to find conditions leading to their regeneration (Fig. 4). When the ATP concentration was increased by adding ATP to the system (10 mM final concentration) a small increment (6%) in the percentage of symmetric particles with respect to the control was found (Fig. 4, A and B). When denatured rhodanese (equimolar to GroEL) was added to the system, a slightly higher amount (around 13%) of symmetric complexes was obtained ( Fig. 4C). This effect was enhanced (up to 37% of symmetric complexes) when a combination of extra ATP (10 mM final concentration) and denatured rhodanese was added (Fig.  4D). These results showed that as soon as the GroEL-GroES system was able to resume folding (by adding ATP and substrate) symmetric GroEL-GroES complexes were formed.
A similar result was obtained when regeneration experiments were performed under conditions that did not favor the presence of symmetric particles, that is, low ATP and K ϩ concentration (0.5 and 25 mM final concentrations, respectively) (Fig. 5). In this case, addition of ATP (10 mM, final concentration) triggered the formation of asymmetric particles (around 90%) as well as symmetric complexes (around 6%) and the concomitant disappearance of the isolated GroEL oligomer (compare Fig. 5,  A and B). The presence of rhodanese (equimolar to GroEL) alone induced a similar change (Fig. 5C). The joint addition of ATP (10 mM final concentration) and rhodanese (equimolar to GroEL) led not only to the appearance of asymmetric particles, but also to the detection of a significant proportion (up to 23%) of symmetric GroEL-GroES complexes.
Relationship between Substrate Folding and the Presence of Symmetric GroEL-GroES Complexes-The presence of symmetric complexes under folding conditions suggested the study of whether there was a more direct implication of these complexes in the folding cycle. Rhodanese folding reactions were followed in three different experimental conditions (Fig. 6): low K ϩ concentration (5 mM), high K ϩ concentration (150 mM), and high K ϩ concentration (150 mM), where the GroEL-GroES system was preincubated for 45 min at room temperature until a high percentage of symmetric particles were present in the system. Folding was measured at different times throughout the folding reaction, and for each time the proportion of symmetric GroEL-GroES complexes was estimated as percentage of side views in samples observed by electron microscopy. As shown in Fig. 6, A and B, folding reaction followed similar kinetics at low and high K ϩ concentration. During the first 20 min of the reaction, in which most rhodanese was folded, the percentage of symmetric GroEL-GroES complexes remained 10 -15% of the whole population. At longer incubation times, symmetric particles disappeared at low K ϩ concentration (5 mM), but they increased at high K ϩ concentration (150 mM) in the same way as shown in Fig. 2. When the GroEL-GroES system was preincubated (Fig. 6C) until almost 70% of the population corresponded to symmetric particles, the addition of the denatured rhodanese led to the rapid disappearance of those complexes which were replaced by asymmetric particles. Afterwards, the folding reaction and the amount of symmetric particles followed a similar scheme than in Fig. 6B.
It has been shown that substrate binding to the GroEL-GroES complex decreases its stability (Martin et al., 1993). Unfolded protein would play an active role on chaperoninassisted folding by inducing the dissociation of the GroEL-GroES complex. Consequently, the result obtained in Fig. 6C could be interpreted as a dissociation of the pre-existing symmetric GroEL-GroES complexes after addition of the unfolded rhodanese. Nevertheless, It has been recently suggested that GroEL-GroES complex rapidly dissociates upon addition of the low guanidinium chloride concentrations typically used for in vitro assays of the chaperonin activity (Todd and Lorimer, 1995). The effect of substrate addition observed by Martin et al. (1993) could then be due to the guanidinium chloride added to the incubation and not to the substrate itself. This poses the question of the eventual effect of guanidinium chloride in the result obtained in Fig. 6C. The same folding experiments of Fig.  6, A-C, were carried out, but using a lower guanidinium chloride concentration (10 mM in Fig. 6, D-F, instead of 100 mM in Fig. 6, A-C). In this case, the results obtained for the folding reaction and the appearance of symmetric particles were very similar to those obtained under standard conditions (compare Fig. 6, A and B, and D and E). On the other hand, when symmetric particles were present previously to the addition of rhodanese, the behavior was different (Fig. 6F). Addition of rhodanese did not dissociate symmetric particles indicating that the effect observed in Fig. 6C was due to the denaturant added and not to the unfolded substrate itself. The folding in Fig. 6F was similar or even faster than in the control experiment, suggesting that a population with a high proportion of symmetric GroEL-GroES complexes is able to fold denatured rhodanese at a similar or even faster rate than a population with a smaller proportion of these complexes.
The result observed in Fig. 6F suggested the study of the effect of different relative percentages of symmetric complexes present in the incubation mixture versus the initial folding rates after addition of denatured rhodanese (Fig. 7). GroEL and GroES were incubated under functional conditions and aliquots were withdrawn at different times so that different percentages of symmetric particles were expected for each aliquot,  ), and symmetric (closed circles) GroEL-GroES complexes was estimated by electron microscopy. After 90 min incubation, when no symmetric complexes could be found in the sample, three aliquots were withdrawn from the incubation mixture and (B) excess ATP (10 mM final concentration), (C) denatured rhodanese (equimolar with GroEL), or (D) both ATP and denatured rhodanese was added to each aliquot, respectively. Afterwards, the percentage of GroEL, asymmetric and symmetric complexes was estimated as in A.
as shown in Fig. 1. Unfolded rhodanese (equimolar to GroEL and using a low guanidinium chloride final concentration of 10 mM) was added to each aliquot and the percentage of symmetric particles in the mixture was estimated later by electron microscopy. For each case, the percentage of refolded rhodanese was measured after 2 min of incubation of the chaperonin folding assay. As Fig. 7 shows, initial folding rates were similar for all cases even though very different concentrations of symmetric and asymmetric complexes were present. Therefore, an incubation mixture with a low percentage of symmetric particles and high percentage of asymmetric complexes has a similar initial folding rate than an incubation mixture containing a high percentage of symmetric particles and a low percentage of asymmetric complexes. This implies that there is no data to assert that only one of the two types of GroEL-GroES complexes found under functional conditions is responsible for folding and suggest that there are no reasons to discard either symmetric or asymmetric complexes in the folding cycle.  3,5,7,15,20,30, and 60 min incubation. At the same time, the amount of symmetric complexes (closed symbols and dashed line) was estimated by electron microscopy. In experiments A-C, the final guanidinium chloride concentration was 100 mM. In experiments D-F, the final guanidinium chloride concentration was 10 mM. Reactions mixtures contained 5 mM K ϩ (A and D) or 150 mM K ϩ (B, C, E, and F). In C and F, the incubation mixture without denatured rhodanese was preincubated for 45 min, until a high percentage of symmetric complexes was present in the population. Then, denatured rhodanese (equimolar to GroEL) was added and the folding assay was carried out as described previously.

125
I Labeling of Rhodanese-Isolation of symmetric complexes was attempted by native PAGE. For that purpose, GroEL and GroES were incubated under conditions leading to a high proportion of symmetric complexes. Then, samples were applied to a native gel equilibrated with buffer containing ATP, Mg 2ϩ , and K ϩ . A control aliquot was kept at room temperature. The gel only resolved GroEL and asymmetric complexes while the control sample kept at room temperature still contained a high percentage of symmetric particles (data not shown). As shown in Fig. 3D, symmetric complexes could be stabilized by adding EDTA. A sample with a high proportion of symmetric particles and later stabilized with EDTA was also subjected to native PAGE to try resolving the three species. As before, only GroEL and asymmetric complexes were obtained (data not shown). This result suggested that in the symmetric GroEL-GroES complexes, one of the two GroES oligomers was less stably bound to GroEL in the presence of ATP, as previously indicated by Hartl and Martin (1995).
As the symmetric complexes were found to be so labile, they were cross-linked with glutaraldehyde in the conditions indicated under "Experimental Procedures." Cross-linked GroEL oligomer showed in the native PAGE a band with the same mobility than that of the non-fixed GroEL oligomer (Fig. 8A,  lanes 2 and 3). Asymmetric GroEL-GroES complexes were obtained after incubation of GroEL and GroES in the presence of 2 mM ADP and after cross-linking, a new electrophoretic band was obtained with a slower mobility than that of the GroEL oligomer (Fig. 8A, lane 5). The mild conditions used for crosslinking allowed the observation of the cross-linked complexes by electron microscopy and the population obtained was mostly asymmetric complexes (up to 90%). When GroEL and GroES were incubated in the presence of 30 mM Mg 2ϩ and 10 mM AMP-PNP and cross-linked, an approximate 1:1 ratio of symmetric and asymmetric complexes could be estimated by electron microscopy. When this sample was applied to a native PAGE, two bands with a similar concentration appeared, one corresponding with the mobility obtained for asymmetric complexes in lane 5, and another band with a slower mobility, which could be assigned to the symmetric GroEL-GroES complexes (Fig. 8A, lane 6), thus validating the relative quantification obtained by electron microscopy.
To make sure that the different mobilities were due to the presence of different GroEL-GroES complexes and not to any artifact induced by the electrophoresis, the effect of salts present in the samples and the ability of the native gel to resolve a mixed population of cross-linked GroEL, asymmetric and symmetric complexes was tested. The results indicated that mobility differences among the three species were maintained (data not shown).
The presence of rhodanese in the different populations of GroEL-GroES complexes was tested using 125 I-labeled rhodanese in the incubation mixture prior to cross-linking with glutaraldehyde. Cross-linked products were resolved by native PAGE (Fig. 8B). To test that cross-linking was just intramolecular and did not generate any artifactual binding of labeled rhodanese, GroEL was incubated with native rhodanese instead of denatured rhodanese. After cross-linking, no radioactivity co-migrating with GroEL was found (Fig. 8, A-B, lane 3). In the same way, GroEL and GroES were incubated in 5 mM ATP and 150 mM K ϩ for 40 min until a high percentage of symmetric complexes could be detected by electron microscopy. Then, native rhodanese was added and the mixture was cross-  , 10, 20, 30, 40, 50, and 60 min incubation and denatured rhodanese was added to each sample (equimolar to GroEL and under low final guanidinium chloride concentration of 10 mM). Afterwards, the percentage of symmetric GroEL-GroES complexes (closed symbols and dashed line) were estimated by electron microscopy and the folding assays were carried out for 2 min as described previously. Initial folding rates (circles) were calculated as the percentage of refolded rhodanese after 2 min of folding assay. (1:2 molar ratio) in 30 mM Mg 2ϩ and 10 mM AMP-PNP (final concentrations) until a 1:1 ratio of symmetric-asymmetric complexes were detected by electron microscopy. Then, the sample was incubated in the presence of 125 I-labeled rhodanese for 1 min and cross-linked with glutaraldehyde. 7, GroEL incubated with GroES (1:2 molar ratio) in 30 mM Mg 2ϩ , 150 mM K ϩ , and 5 mM ATP (final concentrations) for 40 min until a high percentage of symmetric complexes were detected by electron microscopy. Then, the sample was incubated in the presence of 125 I-labeled rhodanese for 1 min and cross-linked with glutaraldehyde. 8, sample prepared as 7 but the mixture was incubated for 30 min after adding the denatured rhodanese before cross-linking with glutaraldehyde. 9, sample prepared as 7 but native labeled rhodanese was added instead of denatured rhodanese. Samples 3-9 were incubated for 20 min with 0.08% glutaraldehyde at 37°C. The cross-linking reaction was stopped by the addition of ammonium chloride (40 mM final concentration). Samples were applied to a 4.5% polyacrylamide native PAGE and stained with Coomassie Brilliant Blue R-250 (Sigma). Afterwards, the gel was dried and autoradiographed. Arrowheads mark the position of the three complexes found. linked. Fig. 8B, lane 9, shows that no radioactivity was present in the band corresponding to symmetric particles. The presence of denatured substrate associated with the symmetric complexes was analyzed by incubating GroEL and GroES in 5 mM ATP and 150 mM K ϩ for 40 min. Then, denatured rhodanese was added and after 1 min, the incubation mixture was crosslinked. After detecting the presence of mostly symmetric complexes (up to 90%) by electron microscopy of the cross-linked product, the sample was applied to native PAGE (Fig. 8, A and  B, lane 7). The radioactivity label co-migrates with the band corresponding to symmetric complexes. In lane 8, after adding denatured rhodanese to the symmetric complexes, the mixture was incubated for 30 min, allowing most rhodanese to be refolded before cross-linking of the GroEL-GroES complexes. In this case, no rhodanese can be detected associated with the bands in the gel. Thus, symmetric GroEL-GroES complexes interact with the substrate following kinetics that are compatible with an active role of these particles in the chaperonin folding circle. It has been previously described that symmetric complexes formed with AMP-PNP are quite stable , probably indicating that GroES does not cycle between bound and free states since GroEL is not hydrolyzing ATP. When a mixture of asymmetric and symmetric GroEL-GroES complexes was preformed by incubation with AMP-PNP, and then, denatured rhodanese was added and cross-linked, only the band corresponding with the asymmetric complexes contained labeled-rhodanese (lane 6). This seems to indicate that (a) denatured rhodanese does not artifactually bind symmetric complexes. (b) When the symmetric complexes are stable and no hydrolysis takes place, rhodanese does not bind to symmetric complexes. This result strongly argues against the possibility that the substrate interacts in the outer surface of GroEL. Instead, rhodanese is most probably located within the central cavity of GroEL, thus suggesting that the substrate enters the chaperonin complex when, at least, one end of the GroEL oligomer is open.
The amount of labeled rhodanese present in lane 7 was estimated by densitometry and showed that 52% of symmetric complexes in that band contained rhodanese (assuming only one rhodanese molecule for each GroEL-GroES complex).

DISCUSSION
The chaperonin GroEL needs in most of the cases the interaction with the co-chaperonin GroES to carry out the folding of proteins both in vivo and in vitro. This interaction generates a GroEL-GroES complex in which one GroES heptamer binds transiently to one GroEL tetradecamer in the apical region of one of the toroids (Langer et al., 1992). This complex is commonly termed asymmetric complex. It has, however, been described that chaperonins GroEL and GroES can, under conditions leading to protein folding, form what has been called symmetric complexes in which one GroES heptamer binds to each of the GroEL toroids ( Fig. 1) (Llorca et al., 1994;Harris et al., 1994;Azem et al., 1994;Schmidt et al., 1994;Todd et al., 1994). Not much is known, however, about the parameters that govern their formation and disappearance. The assays carried out with different concentrations of ATP, Mg 2ϩ , and K ϩ have helped to define the conditions of the symmetric complex formation (Table I). No symmetric complexes are formed in the absence of K ϩ or at low ATP concentration, whereas the highest percentage appears at high concentrations of ATP and K ϩ . It has been previously shown that the GroEL ATPase activity is K ϩ dependent (Todd et al., 1993). In the presence of K ϩ , the rate of ATP hydrolysis is 10 4 -fold higher than in the absence of K ϩ . The K ϩ ion appears to exert its influence by enhancing the affinity of GroEL for ATP. The results obtained support the idea that symmetric complexes are present under conditions where GroEL is actively hydrolyzing ATP. The Mg 2ϩ variation does not exert any influence on the percentage of symmetric complexes formed, nor does the pH in the 7-8 range.
The kinetic assays performed (Fig. 2) confirm a direct relationship between the K ϩ concentration and the formation of symmetric complexes, regardless of the presence or absence of unfolded substrate. At low K ϩ concentration, a small percentage of symmetric complexes is formed which is maintained for 45 min and then a slow decrease in the percentage occurs. At high K ϩ concentrations, there is a steady increase in the percentage of symmetric complexes, reaching 80% after 45 min incubation. Afterwards, there is a slow decrease of the percentage of symmetric complexes until their disappearance after 3 h incubation, probably due to the hydrolysis of the ATP present in the incubation mixture and therefore to a low [ATP]/[ADP] ratio. In the presence of GroES and K ϩ , ATP hydrolysis by GroEL does not proceed linearly (Todd et al., 1993). Instead, three different phases can be resolved. These variations in the rate of ATP hydrolysis could also contribute to explain the different percentages of symmetric complexes at different times. Two sets of experiments were performed to test the relationship between ATP hydrolysis and the percentage of symmetric complexes (Fig. 3). The addition of ADP to an incubation mixture containing a high proportion of symmetric complexes and therefore the decrease of the [ATP]/[ADP] ratio, induces a sharp fall in their percentage until their disappearance. The same applies when non-hydrolyzable ATP␥S is added to the solution. This effect of ATP␥S is not due to the absence of ATP hydrolysis, since symmetric complexes can be obtained after incubation with a different non-hydrolyzable ATP analogue, AMP-PNP, as shown in Fig. 8. Blocking ATP hydrolysis by EDTA, a Mg 2ϩ chelating agent, stabilizes both the symmetric and the asymmetric complexes, freezing the cycle of GroEL-GroES binding and release. Interestingly, when the symmetric complexes disappear from the solution, they are replaced by asymmetric complexes which are stable for hours.
A second set of experiments have helped to establish the relationship between the [ATP]/[ADP] ratio and the appearance of symmetric complexes (Figs. 4 and 5). An assay in which GroEL and GroES were incubated with a high concentration of ATP and K ϩ , which favors the formation of symmetric complexes, was allowed to go to completion such that no symmetric complexes were present (Fig. 4). At this moment, almost all the GroEL oligomers are in the form of asymmetric complexes. The addition of ATP induces the transient formation of symmetric complexes in a percentage that is directly related to the amount of added ATP (data not shown), which reinforces the notion of the [ATP]/[ADP] ratio controlling the formation of symmetric complexes. The addition of unfolded rhodanese to the GroEL-GroES solution also generates the transient formation of symmetric complexes, probably because the protein folding process induces a burst of hydrolysis of the remaining ATP. The combined addition of ATP and unfolded rhodanese reinforces the effect of both substrates and confirms that the presence of ATP and its hydrolysis induces the formation of symmetric complexes.
A second experiment reveals that under low K ϩ and ATP concentrations in which almost no GroEL-GroES complexes are formed (Fig. 5), the increase of the [ATP]/[ADP] ratio by the addition to the incubation mixture of ATP causes only a small and transient increase in symmetric complexes but a large increase in asymmetric complexes. This is probably due to the low K ϩ concentration present that does not induce a large ATP hydrolysis and therefore the formation of the symmetric species. The same applies when unfolded rhodanese is added to the solution. GroEL starts folding rhodanese and a percentage of symmetric complexes appears, but because of the low concentration of ATP, and especially K ϩ present, there is a limited hydrolysis of ATP and therefore the GroEL-GroES complexes that are accumulated are mostly asymmetric. The same can be said when ATP and unfolded rhodanese are added to the solution. In spite of an increase of the [ATP]/[ADP] ratio, the low amount of K ϩ present does not induce a large ATP hydrolysis and therefore, although a certain percentage of symmetric complexes are transiently formed (around 20%), most of the accumulated complexes are asymmetric. The need of K ϩ for the formation of the symmetric complexes is confirmed by the fact that when only K ϩ is added to the GroEL-GroES mixture, a sudden but short appearance of symmetric complexes are observed (7% of the complexes, data not shown), in spite of the low ATP concentration (0.5 mM, final concentration) present in this set of experiments.
The addition of unfolded rhodanese in all the cases studied has generated in the subsequent folding process a large percentage of the asymmetric complexes together with a smaller percentage of the symmetric species. But are the symmetric complexes functional in protein folding or just a side pathway in the folding cycle due to the presence of an excess of ATP? Although the second option is unlikely in view of the previous results, several experiments have been performed to determine the functionality of the symmetric species. In a first set of experiments, the kinetics of the folding activity of a GroEL-GroES mixture (in the presence of a high ATP concentration) to which unfolded rhodanese is added, is compared with the percentage of the symmetric complexes formed (Fig. 6). In all the cases studied (low K ϩ concentration, high K ϩ concentration, and high K ϩ concentration where GroEL and GroES are preincubated so that a large amount of symmetric complexes is present), the kinetics of rhodanese folding follows a similar pattern in which the highest folding activity is obtained during the first minutes of the reaction and then slows down until completion. However, when the K ϩ concentration present is low, the amount of folded rhodanese is lower than in the case of high K ϩ concentration. When there is a low amount of K ϩ , a certain percentage of symmetric complexes are formed which disappear rapidly (15 min), presumably because in spite of the high ATP concentration, there is not enough K ϩ available to induce a large ATP hydrolysis. When the K ϩ concentration present is high, there is a steady increase in the formation of symmetric complexes, even after most of the denatured rhodanese has been folded, which could be explained by the high concentration of ATP. When the symmetric species are already present, the addition of unfolded rhodanese generates their sudden disappearance followed by their rapid and steady increase similar to the previous experiment. This disaggregation of the GroEL-GroES complexes has already been explained by a decrease of their stability in the presence of guanidinium chloride (Martin et al., 1993;Todd et al., 1995) and the results obtained can be explained accordingly. The results in Fig. 6, A-C, indicate that under functional conditions, symmetric complexes can accumulate but apparently they do not need to be highly populated for the folding to take place. The addition of unfolded rhodanese in the presence of a much lower guanidinium chloride concentration (10 mM instead of the typical 100 mM; Fig. 6) does not generate any significant variation in the percentage of symmetric complexes that are formed or in the folding activity of the incubation mixture. However, when the symmetric complexes are already formed, the addition of unfolded rhodanese with a low guanidinium chloride content does not generate the sudden dissociation of the symmetric complexes that is obtained in the presence of a higher guanidinium chloride concentration, which confirms the results of Todd et al. (1995). Interestingly, the presence of a high percentage of symmetric complexes in the first minutes of the incubation goes along with an apparently very fast rhodanese refolding. When the initial folding rate of the GroEL-GroES incubation system for different percentages of symmetric particles was tested, no significant differences were found. Taken together, the results obtained in the functional assays seem to indicate that (a) there is not clear data to assert that either the symmetric or the asymmetric complex are the only ones implicated in protein folding since a GroEL-GroES incubation mixture with a high percentage of symmetric particles is equally effective for folding than a mixture with a high percentage of asymmetric particles. (b) If the symmetric complexes were a dead-end particle inactive in folding, increasing amounts of symmetric complexes in the GroEL-GroES incubation would lead to a decrease in the initial folding rates as symmetric particles would not be active for folding and this does not correspond with the observed results.
Another set of experiments using 125 I-labeled rhodanese (Fig. 8) have confirmed that unfolded rhodanese (but not native) can be found bound not only to GroEL oligomers and to asymmetric complexes, but most interesting, to symmetric complexes. When stable symmetric GroEL-GroES complexes are preformed in the presence of AMP-PNP, rhodanese is not detected in these complexes. This indicates that, under functional conditions, substrate does not bind to symmetric complexes in the outer surface. Rhodanese present in symmetric complexes is most probably located within their central cavity, suggesting that substrate interacts with the asymmetric complex, where at least one end of GroEL is open, and then, the symmetric complexes are formed. This could be explained because during ATP hydrolysis, GroEL binds and releases GroES, making possible the existence of a cycle that incorporates both symmetric and asymmetric complexes.
Current models for the GroEL-GroES reaction cycle are mainly based on the folding activity of asymmetric complexes . Nevertheless, symmetric GroEL-GroES complexes may be involved in protein folding (Todd et al., 1994). The GroEL-GroES system goes through several intermediate states once the ATP hydrolysis has begun, before the cycle is completed (Todd et al., 1993(Todd et al., , 1994. Symmetric complexes could take part of this cycle as one of the intermediates. Binding of nucleotides to one of the GroEL toroids triggers a conformational change that allows GroES to bind GroEL in that end of the cylinder. The nucleotide-bound toroid is able to bind GroES. It has been shown that in the presence of ATP, its hydrolysis allows the formation of an intermediate state with both toroids containing nucleotides (Todd et al., 1994). It could be speculated that in this state both rings of GroEL could be able to bind GroES. Once the GroEL-bound ATP is hydrolyzed, an asymmetric GroEL-GroES complex with ADP bound in just one of the rings of GroEL would be the most stable complex. GroEL residues essential for GroES binding have been located inside the GroEL complex structure solved by x-ray diffraction of crystals . The same residues involved in binding GroES are part of the polypeptidebinding site. Thus, GroES may act in protein folding by directly displacing bound polypeptide, releasing the substrate inside the cavity of the symmetric GroEL-GroES complexes.
Taken together, the results described above are consistent with the possibility that GroEL-GroES symmetric complexes can act as functional units in the protein folding process. Their formation, which run parallel to that of the asymmetric complexes, is influenced by the [ATP]/[ADP] ratio, the K ϩ concentration present in the incubation mixture and to a lesser extent, by the presence of unfolded protein. Any suggestion that the symmetric complexes are only formed when a favorable GroES:GroEL molar ratio is present should be ruled out since symmetric complexes are formed even at a 1:4 GroES:GroEL molar ratio, provided there is enough ATP and K ϩ in the incubation mixture (data not shown). Nothing is known yet about whether the symmetric complexes are also found in vivo. However, the ATP, ADP, and K ϩ concentrations that have been described for E. coli cells (8 mM for ATP, 1 mM for ADP and in the range of 150 -500 mM for K ϩ ; Lehninger, (1982) and Ingraham et al. (1987)) are compatible with their existence as a part of the cycle of protein folding.