Chaperonin chamber accelerates protein folding through passive action of preventing aggregation

Abstract

The original experiments reconstituting GroEL–GroES-mediated protein folding were carried out under “nonpermissive” conditions, where the chaperonin system was absolutely required and substrate proteins could not achieve the native state if diluted directly from denaturant into solution. Under “permissive” conditions, however, employing lower substrate concentration and lower temperature, some substrate proteins can be refolded both by the chaperonin system and while free in solution. For several of these, the protein refolds more rapidly inside the GroEL–GroES cis chamber than free in solution, suggesting that the chamber may have an active role in assisting protein folding. Here, we observe that the difference is caused by reversible multimolecular association while folding in solution, an avenue of kinetic partitioning that slows the overall rate of renaturation relative to the chaperonin chamber, where such associations cannot occur. For Rubisco, reversible aggregation during folding in solution was observed by gel filtration. For a mutant of maltose-binding protein (DM-MBP), the rate of folding in solution declined with increasing concentration, and the folding reaction produced light scattering. Under solution conditions where chloride was absent, however, light scattering no longer occurred, and DM-MBP folded at the same rate as in the cis cavity. In a further test, dihydrofolate reductase, thermally inactivated in the cis cavity or in solution, was substantially reactivated upon temperature downshift in the cis cavity but not in solution, where aggregation occurred. We conclude that the GroEL–GroES chamber behaves as a passive “Anfinsen cage” whose primary role is to prevent multimolecular association during folding.

The GroEL–GroES chaperonin system provides kinetic assistance to the folding of many newly translated proteins via two successive actions (1–4). First, nonnative polypeptide is bound through exposed hydrophobic surfaces via the hydrophobic cavity lining of an open GroEL ring. This serves to prevent multimolecular aggregation and may exert an unfolding action that can reverse misfolding (5–7). Then, upon binding ATP to the same ring, GroES is recruited, its binding associated with large elevation and twisting movements of the apical domains that remove the hydrophobic binding surface, replacing the cavity lining with a hydrophilic, net negatively charged surface (8). This step, occurring in <1 s, efficiently releases the nonnative polypeptide into the GroES-encapsulated folding chamber where folding commences (6, 7, 9–11).

The nature of the assistance provided by the cis folding chamber has been a subject of considerable interest. Is the chamber passive, or does it actively assist polypeptide folding, for example via its hydrophilic character or by close physical confinement? The most accessible avenue to addressing whether there is an active process in the cavity is to compare rates of recovery of the native state inside the cis folding chamber vs. outside in free solution. The original reconstitution experiments of GroEL–GroES-mediated folding were carried out under so-called “nonpermissive” conditions of relatively high substrate concentration and temperature, under which substrate proteins misfolded and aggregated while free in solution, exhibiting an absolute dependence on the GroEL–GroES–ATP system to reach native form (12, 13). For example, after dilution of Rhodospirillum rubrum ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) (51 kDa) from denaturant at 23 °C, the protein quantitatively aggregated, but when stoichiometric GroEL was present, the protein was quantitatively captured by GroEL, and it was then efficiently renatured upon addition of GroES and MgATP (12). Later studies showed that such productive refolding critically required release into and folding within the hydrophilic GroEL–GroES (cis) chamber (9–11). “Permissive” conditions for Rubisco folding were subsequently identified, employing reduced temperature (15 °C), allowing productive folding to occur either free in solution or with the GroEL–GroES–ATP system (14, 15). Strikingly, under these conditions, the rate of renaturation by GroEL–GroES–ATP was observed to be many-fold greater than that in free solution (14, 15). What is the basis to such apparent acceleration of folding?

There are two general models to explain the faster rate of cis folding. One posits that the cis cavity behaves “actively,” via physical confinement or electrostatic wall character, to assist productive folding, effectively accelerating the folding of the caged polypeptide (16). A second model posits that the cis cavity is a passive “no-stick” chamber where the primary structure of the polypeptide governs its folding (17). In this latter model, the relative difference in rates of folding between the cis cavity and free solution under permissive conditions might be accounted for most simply by additional kinetic partitioning that occurs in free solution as the result of multimolecular association. Although such association is for the most part reversible under permissive conditions (because the native form is ultimately fairly efficiently recovered), it nevertheless does not occur in the cis cavity where only a single molecule of substrate protein can generally fit. Thus, such partitioning caused by multimolecular association in solution would slow the overall rate of folding relative to cis folding. We report evidence that such multimolecular association is indeed occurring under permissive conditions in free solution.

Results and Discussion

Faster Folding of Rubisco in the Cis Cavity vs. Free Solution Is Associated with Reversible Aggregation of Rubisco Folding in Free Solution.

In an initial test, we confirmed that refolding of R. rubrum Rubisco is indeed accelerated by a factor of 5 under the permissive conditions used by others (Fig. 1A). An even faster relative rate of recovery was mediated by the stable folding chamber of the single-ring version of GroEL, called SR1, complexed with GroES in the presence of ATP (Fig. 1A). In considering the basis for such acceleration, we noticed that the earlier studies of Rubisco folding under permissive conditions had included a 100-fold molar excess of BSA (14, 15). Consistent with its general uses to stabilize proteins and prevent nonspecific interactions, we observed that when BSA was deleted from the solution folding reaction, the recovery of native Rubisco became negligible (Fig. 1A, spont w/o BSA). This suggested that the nonnative protein was subject to multimolecular aggregation. Indeed, when the solution reaction without BSA was monitored by dynamic light scattering, an immediate and strong rise was observed (data not shown). However, even in the presence of BSA, there was significant light scattering (Fig. 1B, red). This scattering was produced by the nonnative Rubisco substrate and not by refolded Rubisco or BSA because when native Rubisco and BSA were mixed together in solution, there was no detectable light scattering (Fig. 1B, blue). The presence of higher-molecular-weight species of Rubisco under these conditions was established by gel filtration chromatography of refolding reactions of ³⁵S-radiolabeled Rubisco in the presence of BSA (Fig. 1C), revealing the presence of high-molecular-weight species whose elution peaked at the void volume of the column (≈10 MDa), and labeled species eluting broadly thereafter, corresponding to molecular sizes of several 100 kDa. During the course of the reaction, there was a decrease of the latter intermediate-sized species associated with both an increase in amount of [³⁵S]Rubisco eluting at the void volume, and an increase in native Rubisco dimer (102 kDa), eluting at ≈17 mL (Fig. 1C). The latter observation indicates that native Rubisco is being formed at least in part by the reversal of multimolecular association. In contrast with the solution reaction, refolding of Rubisco by SR1–GroES–ATP exhibited the same relatively rapid recovery of native Rubisco either with or without BSA (Fig. 1A). From earlier studies, Rubisco monomers folding in this setting are stably lodged in the cis folding chamber where multimolecular association cannot occur (11, 15) We conclude that spontaneous refolding of Rubisco in solution even under permissive conditions is complicated by multimolecular association. The GroEL–GroES cavity, by contrast, prevents such aggregation through confinement of the folding protein as a monomer.

Fig. 1.

Renaturation of Rubisco under permissive conditions while free in solution is slower than chaperonin-mediated refolding and is associated with aggregation. (A) Rubisco refolding under permissive conditions (80 nM final concentration and 15 °C), monitored by recovery of enzymatic activity either free in solution (spont), with or without 7.5 μM BSA, or as mediated by the chaperonin systems, SR1–GroES [a stable cis folding chamber (9)] or cycling GroEL–GroES, with or without 7.5 μM BSA. Apparent rate constants are indicated for the respective refolding reactions. Note that BSA does not affect the rate of the chaperonin-mediated renaturation, whereas it has a major effect on refolding free in solution. Even so, the reaction in solution containing BSA is 5- to 10-fold slower than the chaperonin-mediated reactions. Note that because native Rubisco is a homodimer, the encapsulated monomers in the SR1–GroES reaction had to be released from the complex to allow folded ones to assemble, accomplished by a brief 4 °C incubation that releases GroES (see ref. 11) before enzyme assay. (B) Dynamic light-scattering measurements during refolding of 80 nM Rubisco free in solution in the presence of 7.5 μM BSA, reported in arbitrary units (red). As a control, light scattering by the same amount of native Rubisco recovered in the reaction was measured in the presence of the same concentration of BSA, producing negligible scattering (blue). BSA alone in buffer also produced negligible scattering (black). (C) Gel filtration analyses of 80 nM [³⁵S]Rubisco carried out at 3 times during its refolding free in solution in the presence of 7.5 μM BSA. Amounts of radioactive Rubisco recovered in fractions from a Superose 6 column are reported as percentages of the input Rubisco. The void volume peak at 6 mL corresponds to ≈10 MDa. The peak at 17 mL corresponds to native Rubisco homodimer, 102 kDa, established by independent gel filtration analysis of native Rubisco (dotted line). Note that species in the 10-min experiment that are intermediate in size between the void peak and native appear to be converted at later times into the native species and into additional void volume species, apparently reflecting, respectively, reversal of aggregation to produce the native state and irreversible aggregation.

Faster Folding of Double-Mutant Maltose-Binding Protein (DM-MBP) in the Cis Cavity vs. Free Solution Is Associated with Concentration-Dependent Aggregation of the Mutant MBP in Free Solution.

Similar behavior was observed for a double-mutant form of MBP (DM-MBP, V8G/Y283D; 41 kDa), folding ≈10-fold faster with GroEL–GroES than free in solution (ref. 16 and see Fig. 2A). Strikingly, the rate of recovery of native DM-MBP free in solution was strongly dependent on its concentration (Fig. 2B), slowing significantly as the concentration was increased. Such behavior usually reflects the occurrence of multimolecular association/aggregation, and this was observed here when a solution of refolding DM-MBP (100 nM) was analyzed by dynamic light scattering (Fig. 2C). Light scattering occurred promptly upon dilution from denaturant. By contrast, the wild-type form of MBP folding free in solution did not exhibit significant light scattering (Fig. 2C), nor did it exhibit a concentration-dependent rate of recovery (Fig. 2D).

Fig. 2.

Renaturation of DM-MBP under permissive conditions is slower free in solution than chaperonin-mediated, associated with aggregation occurring free in solution as indicated by concentration dependence of rate of refolding and dynamic light scattering. (A) DM-MBP free in solution (black; spont) at 100 nM refolds at a rate that is only ≈1/10 the rate of the chaperonin reaction mediated by GroEL–GroES–ATP or by SR1–GroES–ATP (red and blue). Renaturation is monitored by an increase in tryptophan fluorescence intensity (excitation, 295 nm; emission, 345 nm), as described in Methods. (B) Rate of refolding free in solution is inversely related to the concentration of DM-MBP. (C) Dynamic light scattering, in arbitrary units, during refolding of 100 nM DM-MBP in solution. In contrast with the double mutant, wild-type MBP did not produce significant light scattering when refolding free in solution. (D) Wild-type MBP, 100 nM, does not exhibit concentration dependence of its refolding rate.

Chloride Depletion Blocks DM-MBP Aggregation During Folding in Solution and Increases the Rate of Folding to That of the Chaperonin Reaction.

Further support that aggregation of the DM-MBP in solution is responsible for its slow refolding relative to the chaperonin reaction was forthcoming when we found solution conditions that relieved the light-scattering behavior, involving omission of chloride from the refolding buffer (Fig. 3A). In the absence of chloride-dependent aggregation, the rate of recovery of DM-MBP in free solution was now equivalent to that of the GroEL–GroES–ATP and SR1–GroES–ATP reactions (Fig. 3B). Thus, for DM-MBP, the apparent acceleration in folding rate with the chaperonin reaction could be fully accounted for by slowing of the solution reaction by multimolecular aggregation.

Fig. 3.

Relief of light scattering behavior of 100 nM DM-MBP, folding free in solution, by deletion of chloride from the refolding buffer is associated with restoration of folding rate to that of the chaperonin-mediated reaction. (A) Light scattering showing that deletion of chloride from the folding buffer (replacing it with acetate) relieves light-scattering behavior of DM-MBP folding free in solution. (B) Refolding of DM-MBP free in chloride-deficient solution measured by the increase of tryptophan fluorescence intensity, showing that the rate of renaturation in chloride-deficient solution (black) is, under these conditions, the same as that of the chaperonin reactions mediated by GroEL–GroES–ATP (red) or SR1–GroES–ATP (blue), also in chloride-free conditions.

Dihydrofolate Reductase (DHFR) in the Cis Cavity but Not Free Solution Is Protected from Irreversible Thermal Inactivation Associated in Solution with Aggregation.

To inspect further for an active role of the cis cavity, we asked whether it could affect nonnative states produced via thermal unfolding of an already-produced native form. Could the cis cavity, for example, selectively protect against inactivation by actively renaturing a protein as it became thermally unfolded, compared with free solution? To test this possibility, the inactivation of human DHFR was examined when the native protein was present either in solution or inside a stable cis cavity, measuring enzyme activity after a 10-min treatment at various elevated temperatures. As shown in Fig. 4, virtually identical curves of temperature-dependent inactivation were observed (Fig. 4 A vs. B). Notably, the cis chamber used for the inactivation testing remained stable under the temperature treatment, revealed by testing in a parallel experiment (using fluorescent GroES) for any release of GroES during thermal treatment by transfer to a “trap” molecule [see supporting information (SI) Methods and Fig. S1). The results of identical thermal inactivation in solution and in the cis cavity indicate that the cis chamber has no unique ability to protect incipiently unfolding forms of DHFR, i.e., has no activity that favors restoration of the native state.

Fig. 4.

Comparison of thermal inactivation of an already-native monomeric protein, DHFR, free in solution and inside a stable cis cavity, and recovery of activity after temperature downshift in the cis cavity but not in solution. (A and B) Thermal inactivation. Native DHFR free in solution or in a cis ternary GroEL–GroES–ADP-aluminum fluoride complex was exposed to elevated temperatures for 10 min, and enzymatic assay was then immediately carried out at 23 °C. The loss of enzymatic activity with increasing temperature occurred to a nearly identical extent while free in solution (Solution) or while in the cis cavity (cis). (See Methods for formation of the cis ternary complex and SI Methods for a test that assured that the cis complex remained stable during thermal treatment.) (C and D) Recovery from thermal inactivation. After inactivation of DHFR at 50 °C (middle bars), the mixture was returned to 23 °C for 30 min and then assayed for enzymatic activity. A substantial amount of activity was recovered from the cis complex (D, right bar) but not from the solution reaction (C, right bar), where aggregation had occurred (see Results and Fig. S2).

A striking difference was observed, however, in recovery of enzyme activity in solution vs. the cis cavity when fully inactivated DHFR (50 °C, 10 min) in these two situations was downshifted to 23 °C. There was ≈50% recovery of activity inside the cis cavity (Fig. 4D), whereas none occurred in free solution (Fig. 4C). In the latter setting, the failure of recovery was associated with readily detectable aggregation as measured by light scattering (Fig. S2). Thus, in the context of thermal exposure, as with folding under permissive conditions, the major action of the cis cavity in affording recovery of the native state appears to be prevention of multimolecular association.

GroEL–GroES Cis Cavity Functions as a Passive “Anfinsen Cage.”

This principal action of the cis cavity in preventing multimolecular association represents, in effect, an extension of chaperonin action observed earlier under nonpermissive conditions; that is, under nonpermissive conditions, proteins folding free in solution were observed to undergo quantitative aggregation (e.g., refs. 12 and 13), but the GroEL–GroES system was shown to prevent that, kinetically partitioning misfolded monomers away from irreversible multimolecular association occurring while free in solution (18). Although confinement in a cis cavity thus forestalls multimolecular association, it does not appear to confer a different folding route, as also indicated by a recent hydrogen–deuterium exchange experiment observing the same development of exchange protection in a substrate protein folding inside the cis cavity as free in solution (19). Thus, although the starting state for folding in solution after dilution from denaturant may be different from that produced by release from the cis cavity wall upon ATP–GroES binding, it seems likely that the same ensembles of nonnative states are populated by rapid collapse (<1 s) in both cases, leading to similar rates and folding trajectories. Corollary to the production of a similar ensemble of nonnative states in the two locations, it should be noted that, as in solution, misfolded states can be populated inside the cis cavity, as demonstrated both by the thermal exposure experiment presented here and by a recent SR1–GroES cis folding experiment using a secretory protein, in which nonnative disulfides were observed to be transiently populated during productive folding (20). The misfolded states produced in the cis cavity are likely to be the same as states produced free in solution, but in the cis cavity they are prevented from multimolecular aggregation, giving them a chance to return to a productive route. In sum, we conclude that the cis cavity behaves as a passive chamber, likely by using its charged walls for no-stick behavior, allowing a polypeptide chain to fold according to its primary structure, as observed originally by Anfinsen and his coworkers (21). Inside the Anfinsen cage of the chaperonin, however, there is no possibility of aggregation.

Methods

Proteins.

GroEL, SR1, GroES, and GroES(98C) were expressed and purified as described in ref. 22. Human DHFR (23) and R. rubrum Rubisco (11) were expressed and purified as referenced. The mature form of wild-type MBP was expressed as a soluble protein in Escherichia coli and purified by affinity chromatography on amylose resin (New England Biolabs). The mature form of the DM-MBP (V8G/Y283D) was expressed in E. coli and was purified from inclusion bodies under denaturing conditions as described in ref. 24. GroES(98C) was labeled with fluorescein maleimide as described in ref. 22. Rubisco was labeled with ³⁵S by expressing a C-terminally His₆-tagged version in a small-scale culture in the presence of [³⁵S]methionine (25); the radiolabeled protein was purified batchwise with Talon metal affinity resin (Clontech).

Protein Refolding.

Rubisco.

Folding was carried out under permissive conditions (14, 15) at a final Rubisco concentration of 80 nM at 15 °C in buffer containing 50 mM Tris·HCl (pH 7.7), 10 mM KCl, 7 mM MgCl₂, and 2 mM DTT. For the spontaneous reaction, Rubisco was diluted 100-fold from a solution containing 5 M guanidine-HCl, 100 mM Tris·HCl (pH 7.5), 1 mM EDTA, 20 mM DTT into buffer solution with or without 7.5 μM BSA. For the chaperonin reaction, unfolded Rubisco was diluted 100-fold into buffer with or without 7.5 μM BSA containing 2 μM GroEL or SR1, to form binary complexes, and GroES and ATP were subsequently added to final concentrations of 2.5 μM and 5 mM, respectively, to commence folding. At the indicated times of refolding, an aliquot was quenched with hexokinase/glucose, and activity was determined as described under the assays. To permit active Rubisco dimer formation in the SR1 reactions, aliquots of the folding reaction were incubated at 4 °C for 10 min before assay to allow GroES to dissociate, releasing the Rubisco monomers, and allowing dimerization. Under these conditions, release and dimerization occur in <3 min.

MBP.

Folding was carried out as described at a final MBP concentration of 100 nM at 25 °C in buffer containing 20 mM Tris·HCl (pH 7.5), 200 mM KCl, and 5 mM Mg(OAc)₂. For chloride-free reactions, refolding was carried out in 100 mM Hepes (pH 7.5), 20 mM KOAc, 5 mM Mg(OAc)₂. For the spontaneous reaction, MBP was diluted 100-fold from a solution containing 6 M guanidine-HCl, 100 mM Tris·HCl (pH 7.5), 20 mM KCl. For chloride-free conditions, MBP was diluted from 9.5 M urea. For the chaperonin reaction, MBP unfolded in 6 M guanidine HCl or 9.5 M urea was diluted 100-fold into one of the above buffers with or without chloride, respectively, containing 250 nM GroEL or SR1 to form binary complexes, and GroES and ATP were subsequently added to final concentrations of 250 nM and 5 mM, respectively, to commence folding.

Assays.

Rubisco was assayed radiochemically (12). Refolding of MBP was followed by an increase of tryptophan fluorescence (excitation, 295 nm; emission, 345 nm) (26), using a PTI QuantaMaster fluorescence spectrometer. DHFR enzyme activity was assayed spectrophotometrically (19). Light-scattering experiments were carried out on a DynaPro dynamic light scattering instrument (Wyatt Technology).

Production of Refolded DHFR in Solution and in a Stable GroEL–GroES–DHFR Ternary Complex.

To produce native DHFR free in solution for thermal-inactivation testing, DHFR denatured in 6 M guanidine-HCl and 2 mM DTT was diluted 100-fold into buffer containing 100 mM [bis(2-hydroxyethyl)amino]tris(hydroxymethyl)methane (pH 6.1), 20 mM KCl, 5 mM MgCl₂, 2 mM DTT, to reach a final concentration of 2 μM, and folding was allowed to proceed to completion (30 min). To produce cis encapsulated native DHFR inside GroEL–GroES for thermal inactivation studies, guanidine-HCl-unfolded DHFR was diluted 100-fold into the same pH 6.1 refolding buffer containing 2 μM GroEL to a final DHFR concentration of 2 μM (19). Refolding in a stable ternary complex was then initiated by adding 4 μM, GroES, 5 mM ADP, 5 mM KF, and 0.5 mM KAl(SO₄)₂. ADP·AlF_x is a transition state analog that produces stable cis folding-active ternary complexes of GroEL, GroES, and substrate protein, which reaches its native state within the cis chamber (27). Only ≈50% of the DHFR in DHFR–GroES binary complexes becomes cis-encapsulated by this step, reaching the native state, with the remaining DHFR molecules bound to the ring in-trans to GroES. We observed here that trans ring-associated DHFR fails to renature during the initial step of refolding and during the subsequent steps of thermal exposure and temperature downshift, verified by examining obligate trans ternary complexes produced by reversing the order of addition, adding DHFR to preformed GroEL–GroES–ADP-aluminum fluoride complexes. These ternary complexes failed to renature DHFR both during initial formation and after downshift of temperature after thermal treatment.

Thermal Denaturation–Renaturation of DHFR in Solution and in Ternary Complex.

Aliquots of native DHFR, free in solution or inside a cis ternary complex, were exposed to 20, 35, 40, 45, or 50 °C for 10 min and then immediately assayed at 23 °C for DHFR activity. Renaturation was assessed by measuring DHFR activity in the 50 °C samples after they were returned to 23 °C and incubated for 30 min at this temperature.