Direct Observation of Chaperone-Induced Changes in a Protein Folding Pathway

See allHide authors and affiliations

Science  30 Nov 2007:
Vol. 318, Issue 5855, pp. 1458-1461
DOI: 10.1126/science.1144972


How chaperone interactions affect protein folding pathways is a central problem in biology. With the use of optical tweezers and all-atom molecular dynamics simulations, we studied the effect of chaperone SecB on the folding and unfolding pathways of maltose binding protein (MBP) at the single-molecule level. In the absence of SecB, we find that the MBP polypeptide first collapses into a molten globulelike compacted state and then folds into a stable core structure onto which several α helices are finally wrapped. Interactions with SecB completely prevent stable tertiary contacts in the core structure but have no detectable effect on the folding of the external α helices. It appears that SecB only binds to the extended or molten globulelike structure and retains MBP in this latter state. Thus during MBP translocation, no energy is required to disrupt stable tertiary interactions.

Folding pathways are traditionally studied with isolated proteins, even though the cellular environment often presents interactions with other molecules during the folding process. For instance, interactions with chaperones are often crucial to guide folding, to prevent aggregation, and to facilitate protein translocation across membranes (1, 2). However, how folding pathways are affected by chaperones remains poorly understood. For instance, it is unclear which steps of the folding process are affected by chaperones, whether the protein-chaperone structure is stable and well defined, and whether native or alternative tertiary interactions are formed in the presence of chaperones.

We addressed these questions by using optical tweezers to induce the mechanical unfolding and refolding of a single protein molecule in the absence and the presence of molecular chaperones. Single-molecule techniques can reveal folding transitions, as has been shown for isolated molecules (37), although they cannot directly correlate the transitions with molecular-level changes in protein structure. To obtain structural insights into the chaperone-free folding pathway, we combined all-atom molecular dynamics (MD) simulations with optical tweezers measurements.

We studied the unfolding and folding of Escherichia coli maltose binding protein (MBP) and its dependence on the chaperone SecB (8, 9). The interaction between SecB and MBP has been extensively studied by bulk biochemical assays, making it an ideal system for exploration at the single-molecule level. The known physiological effects of SecB are to reduce aggregation (10), to promote translocation across membranes (11), and to delay formation of the native state (12, 13). SecB binds to hydrophobic peptide regions (14, 15) and does not require a signal sequence for interaction (16). Although the structure of MBP in complex with SecB is unknown, it has been suggested that MBP forms tertiary structure elements in the presence of SecB (17). Such elements would need to be disrupted during translocation because only an extended chain can pass through the narrow SecYEG channel (18).

Individual MBP molecules were tethered between two polystyrene spheres ∼2 μm in diameter by using a 2553–base pair DNA molecular spacer to prevent undesired bead-bead interactions (Fig. 1A). Connections to the N- and C-terminal ends of wild-type MBP were obtained via an engineered biotin group and a quadruple c-myc tag, respectively. Purified constructs, here referred to as MBP, retained their ability to bind to an amylose resin, indicating proper folding. The DNA-protein tether was stretched by displacement of the pipette bead, and the resulting force on the second bead trapped in the laser focus was measured. These experiments yielded force-extension curves (Fig. 1B) showing a sudden change in extension at an average applied force of 25 ± 8 (SD) pN. When MBP was replaced with a plain biotinylated c-myc tag, there were no sudden extension changes (fig. S1), showing that they correspond to MBP unfolding events. After the unfolded MBP was relaxed by moving the beads together, it folded back to its native state, as evidenced by a second stretching experiment that yielded a similar curve. This unfold-refold cycle was reproducible several times on a single molecule until a linkage in the tether broke.

Fig. 1.

Experimental setup and SecB dependence of MBP force-extension curves. (A) MBP is tethered between two beads: One is held on a position controlled micropipette; the other is held by an optical trap allowing force detection. At the C terminus, MBP is attached via an antibody-myc-tag connection, whereas at the N terminus it is attached via streptavidin-biotin linkages to a DNA tether, which in turn is attached to the bead surface via an antibody-digoxigenin connection. (B) Force-extension curves in the absence of SecB showing unfolding at high force (red), refolding at low forces (green), and again unfolding at high force (blue). (C) Force-extension curves in the presence of SecB (0.1 μM). The second stretching curve (red) lacks the typical unfolding features, showing that stable tertiary interactions are absent. (D) Comparison between relaxation curves and WLC model. Indicated is the average extension from several curves within a force window, normalized to the extension expected from the WLC model (fig. S2). At higher forces, data and model overlap, whereas a lower value points to additional compaction of the polypeptide. The error bars indicate the standard deviations of the extension from various refolding curves.

Stretching curves of a second construct composed of four tandem MBP repeats (4MBP) reproduced the unfolding at an average force of 23 ± 4 pN (Fig. 2, A and B). As previously reported for atomic-force microscopy experiments on proteins with repeated domains (3, 6), a sawtooth pattern was observed, corresponding to the separate unfolding events of the repeats. Interestingly, however, after the force had been lowered to allow refolding, in the second stretching curve the protein often failed to unfold at the normal 25 pN (Fig. 2B, red trace), instead requiring over 40 pN. Thus, the high local concentration likely promoted aggregation (19) of the MBP repeats to produce a structure more stable than the native state.

Fig. 2.

(A) Schematic illustration of the 4MBP construct. (B) Force-extension curves in the absence of SecB. Dotted lines represent the theoretical WLC compliance of the DNA-protein construct in various stages of unfolding, as predicted by MD simulations: curve 1, DNA tether alone, representing the native or aggregated MBP state; curve 2, DNA tether plus compliant polypeptide of four times 91 residues, representing four folded MBP cores and their unfolded external α helices; and curve 3, polypeptide length is increased by 279 residues, representing the unfolding of one core structure. For each subsequent curve (4, 5, and 6) the length is increased by one additional unfolded core. After full unfolding (6) and subsequent relaxation to low force, the stretch data (red trace) follow WLC curve 1, indicating 4MBP has misfolded into a tight aggregate. (C) Force-extension curves in the presence of SecB (0.1 μM). Aggregation is prevented during relaxation, as evidenced by the low forces required for extending the polypeptide (red trace). The 4MBP results are consistent with the single MBP repeat data.

Once proper folding behavior was established in the single MBP construct, the chaperone SecB was flowed into the fluidic chamber of the optical tweezers. The first stretching curve (Fig. 1C) was similar to those without chaperone present, exhibiting a comparable average unfolding force of 24 ± 4 pN. SecB thus does not promote unfolding of native MBP, even when it is held under tension. Subsequent stretching curves did not overlap with the first stretching curve but rather followed the relaxation curve and lacked features typical of breaking tertiary interactions. SecB thus only binds fully unfolded MBP and when bound prevents the formation of any stable tertiary interactions in MBP. In addition, the MBP-SecB complex appears to form more rapidly than the fastest folding event (8). Similarly, in the aggregation transition the second stretching curve for 4MBP overlapped with the previous relaxation curve (Fig. 2C), showing that despite their stability, the aggregation interactions cannot form because of interference by SecB.

The force-extension curves show more detailed features, and the challenge is to relate these to structural changes. For example, a gradual extension increase involving several small substeps is observed below 15 pN before full unfolding at ∼25 pN (Figs. 1 and 2). The low- and high-force transitions cannot reflect a separate unfolding of structural domains; the two lobes of MBP (20) cannot unfold independently because the polypeptide runs back and forth between them.

To identify structural changes associated with transitions in the force-extension curves, we used molecular dynamics (MD) to simulate the unfolding of native MBP. When a force was applied on the MBP termini (Fig. 3A and movie S1), a five-residue segment at the N terminus unfolded first, followed by the C-terminal α helix. Continuing such unfolding simulations is problematic, because the increasing protein extension makes the simulation volume computationally intractable. To circumvent this, we deleted unfolded segments and then continued the simulations on the new construct. This method relies on the assumption that unfolded segments remain extended and do not fold back, which is reasonable given the force applied on the termini. The subsequent simulations showed that another four C-terminal α helices located on the exterior of the structure were sequentially unfolded, until in total 91 residues had unfolded and a stable core structure remained (Fig. 3B). Note that the simulations involve pulling at much faster time scales (21) and therefore act as a hypothesis that we will test against the experiments.

Fig. 3.

Results from all-atom MD simulations. (A) MBP native state. After applying a force to the C and N terminus, the red segments sequentially detached until a stable core structure remained (B). Unfolded segments were deleted from the structure to keep the simulation volume computationally tractable.

The stability of the simulated intermediate core structure implies that it might be observed as a distinct state in the tweezers measurements. The theoretical force-extension relation for this state based on the worm-like chain (WLC) model (22) indeed overlaps with the data after the low-force unfolding below 15 pN (Fig. 2 and fig. S2), indicating a good agreement between simulated and measured length changes. Also the gradual nature of the low-force extension increases is consistent with the simulated sequential unfolding. The 4MBP data show an increase in length that is four times larger at low force compared with that of native MBP, indicating that all external α helices unfold before the first core unfolding. The length changes of the high-force unfolding events agree with the unfolding of a single 279-residue core, showing that MBP fully unfolds. We note that reported stability-decreasing MBP mutations (23) are all positioned near the N terminus of the predicted core intermediate structure.

The sequential detachment of the external α helices suggests that the reverse folding process, in which they would zip back onto the core surface, might be directly observable in the force-extension experiments. We tested this idea by stopping the stretch experiment after the low-force unfolding and then bringing the beads together again. A gradual refolding of the external α helices was indeed detected directly as a near-equilibrium process, as evidenced by the shortening of the tether at nonzero loads (Fig. 4A). The native state was recovered after relaxation, as seen by the characteristic unfolding features in the subsequent stretching curve. Note that after full unfolding (Figs. 1 and 2) one does not directly observe refolding of the core nor of the external α helices during relaxation, which is consistent with the proposed folding pathway in which the slowly folding core structure must form first.

Fig. 4.

Low-force unfolding and refolding and translocation assay results. (A) Force-extension curves of 4MBP in absence of SecB. The construct was first stretched, resulting in the predicted gradual unfolding of the external α helices (Fig. 3), and then relaxed before the core structures would unfold. During relaxation the near-equilibrium refolding of the same α helices was observed directly as a shortening of the tether. The dotted lines indicate the WLC behavior: The first denotes the DNA alone, whereas in the second a 4x91 residue-compliant polypeptide was added, representing the unfolded external α helices (also fig. S2). (B) Force-extension curves of 4MBP in presence of SecB (0.1 μM). The SecB interactions do not alter this refolding transition. (C) ATPase activity during translocation of MBP stability mutants. Folding stability in kcal/mol is, respectively: 11.2 [Thr345→Ile345 (T345I)], 10.5 (wild type), 9.4 [Val8→Gly8 (V8G)], 9 [Ala276→Gly276 (A276G)], and 7.3 [Tyr283→Asp283 (Y283D)]. All mutants exhibited a similar energy requirement for SecB-mediated translocation (dark blue). The preMBP translocation kinetics was similar in the presence of SecB. Controls without SecB (light blue) or MBP showed low ATPase activity and translocation as expected. The error bars indicate the standard deviations. (D) Force-induced MBP folding and unfolding transitions observed in the experiments and their dependence on SecB.

The external α helices refolded with similar dynamics in the presence of SecB (Fig. 4B). SecB apparently does not affect this transition, despite the existence of a putative SecB binding site on the unfolded external α helices, as identified in peptide binding assays (14). These results may indicate that this refolding transition occurs faster than SecB binding. Alternatively, SecB may only bind sites buried within the core structure (15) and not efficiently bind the putative binding site on the external α helices, possibly because of the nearby bulky core.

In the experiments without SecB, when the polypeptide is relaxed after being fully unfolded (Fig. 1B), the measured extension increasingly deviates from the WLC behavior, showing lengths that are over two times smaller at the lowest forces (fig. S2 and Fig. 1D). This compaction of the polypeptide is consistent with previous bulk studies indicating an MBP molten globulelike state (23, 24), a compacted form that is held together by unstable tertiary and secondary intramolecular interactions. When such a relaxation experiment was rapidly followed by a second stretching experiment, the two resulting curves overlapped (fig. S3), which indicates that the transition is near equilibrium and that the tertiary interactions are unstable.

The curve shape for relaxation to low forces is similar in the presence of SecB (Fig. 1, C and D), showing that protein-SecB interactions do not result in an additional compaction. The MBP-SecB complex has been shown to have a strict 1:1 stoichiometry (25) and likely involves multiple contacts (26). One might therefore anticipate that their rupturing during stretching would also show up as sudden extension changes. The lack of such features suggests that MBP-SecB contacts are not stable or alternatively that binding occurs effectively at a single MBP site.

Next, we followed the transition from the compacted molten globulelike state to the natively folded state. Considering this transition as a single-barrier process, the probability to fold within a waiting time t is P(t) = 1 – ekt, where k is the folding rate. This probability can be estimated by performing multiple stretch-relaxation cycles in which the polypeptide is allowed to refold at zero load in the waiting time between the relaxation and the stretching curve. Whether folding had occurred was determined from the unfolding features during the subsequent stretching. We obtained a folding rate of 0.76 ± 0.19 s–1 in the absence of SecB (fig. S4). This value is similar to MBP folding rates found in bulk (23, 24), which is consistent with the idea that it is the slowest folding step. We could not quantify the refolding rate in the presence of SecB because this required waiting times in the order of minutes. It is clear, however, that SecB-MBP interactions maintain MBP in the molten globule-like compacted state by perturbing the formation of stable tertiary interactions.

The single-molecule results have a direct implication for MBP translocation across the cellular membrane. Whether stable tertiary interactions are actively disrupted by the translocation machinery is a matter of debate (17, 27). Our data suggest that SecB efficiently prevents any stable tertiary interactions in MBP. To test this prediction in a translocation reaction, we measured the adenosine triphosphatase (ATPase) activity during the translocation of MBP mutants with altered folding stability (17, 23). We indeed found that all mutants displayed ATPase activity similar to that of wild-type preMBP (precursor form of MBP with the signal sequence) (Fig. 4C). Control experiments lacking MBP or SecB showed low ATPase activity as expected. These results support our hypothesis that for SecB-mediated translocation only limited energy is required to disrupt SecB-MBP interactions as well as tertiary intramolecular interactions in MBP. The single-molecule data provide an estimate for this energy requirement of ∼25kBT (fig. S2), which roughly corresponds to the hydrolysis of one ATP molecule.

Taken together, the data indicate a folding pathway with a large variety of transitions and modes of folding (Fig. 4D), which are each affected differently by SecB: (i) The extended peptide is compacted to a molten globule state in either the presence or the absence of SecB. (ii) In the absence of SecB, folding proceeds from the molten globule to a core intermediate, but SecB prevents the formation of stable tertiary interactions, thereby maintaining the molten globulelike state. This effect may well be more general and apply also to other chaperones such as GroEL, for which there is support from other studies (28, 29). (iii) Once the core intermediate has formed, SecB cannot bind, and it therefore has no effect on the folding of the external α helices onto the surface of the core structure. Folding from the core to the fully folded state exhibits similarities with a nucleation-growth mode of folding (30, 31) and contrasts with the complex conformational search in the molten globulelike phase that characterizes the folding of the core. (iv) SecB also prevents the stable aggregation interactions that occur at high local MBP concentrations.

The results provide an important first step in understanding how a protein folding landscape is altered by contacts with a secondary protein. The approach is general and can be applied to interrogate other proteins and chaperone systems.

Supporting Online Material

Materials and Methods

Figs. S1 to S4


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article