First glimpses of a chaperonin-bound folding intermediate

Molecular chaperones that function by encapsulation of their substrates within a cavity formed by rings of subunits (known as “chaperonins”) are now arguably among the best-understood molecular machines, both structurally and functionally. As a consequence, we understand in detail how the Escherichia coli GroEL/ES chaperonin team harnesses ATP in an elegant allosterically regulated cycle to transiently present hydrophobic surfaces (1–3). These surfaces bind partially folded or misfolded substrate proteins and release them to wander anew over their folding landscapes in search of their biologically critical native states. Yet, despite this impressive progress in our understanding of the workings of this chaperonin machine, the subject of the actions of GroEL/ES, the partially folded polypeptide substrate, has received considerably less attention (4), largely because of technical obstacles to observing and characterizing it. An article in a recent issue of PNAS resulting from a collaboration between the laboratories of Kurt Wüthrich and Art Horwich (5) exploited newly developed NMR approaches, in a novel way, to directly observe a substrate when it is bound to GroEL.

The GroEL Cavity: A Privileged Environment for Folding

The structure of the GroEL monomer and the arrangement of monomers in the tetradecamer particle enable it to perform its key role in facilitating the folding of “reluctant folders.” Each monomer is made up of three domains. Substrates bind with high affinity to an ATP-bound ring of GroEL subunits via hydrophobic patches on their apical domains. The binding of GroES via its GroEL interactive mobile loops to the substrate-bound ring leads to a twist and an angular movement of the apical domains, sequestering the hydrophobic surfaces and creating a hydrophilically lined cavity of ≈60,000-ų interior volume. During this step, the substrate protein is released into the cavity, where its folding can begin. ATP binding in the opposite ring favors dissociation of GroES in the cis ring and thus allows the substrate to diffuse away, either successfully folded or destined to rebind to another or the same GroEL.

Chaperonins do not act on all proteins to modulate their folding. Rather, it seems that the GroEL/ES system becomes important for larger proteins that populate aggregation-prone kinetically metastable intermediates. A recent extensive proteomic study of E. coli proteins (6) found that 250 of the ≈2,400 cytosolic E. coli proteins interact with GroEL, and, of these, only 85 are obligately dependent on GroEL. Of these 85, 15 are essential proteins. Strict dependence on chaperonin is recapitulated in vitro by the ability of these proteins to fold in high yield without competing aggregation processes only when GroEL, GroES, and ATP are present. Provocatively, there is a disproportionate bias for α/β TIM barrels among the set of obligate GroEL substrates, leading one to speculate about the relationship between details of GroEL action and obstacles associated with specific protein structural classes.

How does GroEL binding influence the conformational properties of the substrate, both when it is bound and upon release? How do interactions with the chaperonin lead to more productive folding of the substrate and avoidance of “pitfalls” along its folding trajectory? Alternative models for the beneficial effects of the chaperonin system include the rather unlikely possibility that chaperonins specifically “template” secondary and tertiary structure, thus nucleating the proper fold and diminishing the probability of misfolding; the geometric reality that chaperonins are “isolation chambers,” with the consequence that the vulnerable polypeptide is prevented from encounter with others of its kind and ensuing aggregation; the potential impact of confinement of the folding polypeptide in a restricted volume surrounded by hydrophilic surfaces; and the idea that the chaperonin gives a polypeptide a fresh start, because binding mediated by hydrophobic surfaces would tend to break up any residual structure in the substrate. Although it is widely agreed that GroEL binding leads to unfolding of a substrate, there has been some debate as to whether the unfolding is active mechanical unfolding or, alternatively, whether it is a passive consequence of binding to the hydrophobic surfaces disposed at some distance from one another inside GroEL, like surface denaturation [see, for example, the recent study by Lin and Rye (7)]. In either case, the released substrate essentially starts over on its folding trajectory, but the extent to which its unfolded ensemble is altered may differ between these two models.

An exciting goal of current research on chaperone-facilitated folding is to deepen our understanding of the energy landscape sampled by a polypeptide chain under the influence of a chaperone. To do this requires us to “see” the ensemble of states populated by a protein upon binding to a chaperonin, upon release, and upon folding to its native state. Here, we face head-on the reality of the technical challenges that must be overcome. To explore how interaction with a chaperone alters the folding energy landscape of a substrate requires that we deploy methods that enable observation and characterization of species that are: (i) parts of very large complexes, (ii) ensemble-like in nature, (iii) dynamic, and (iv) poorly structured.

The New NMR: A Bigger Hammer

In the last few years, Wüthrich and co-workers (8, 9) have developed experiments that break the size barrier of NMR. In fact, they recently teamed with Horwich and colleagues (10, 11) to demonstrate that NMR could be applied to the nearly megadalton GroEL/ES complex and were able to see conformational changes in GroES caused by its binding to GroEL. Now they have directed similar approaches toward a GroEL-bound substrate, in this case dihydrofolate reductase (DHFR). NMR study of larger proteins requires isotopic enrichment and multidimensional spectroscopy, which together provide spectral dispersion and simplification. Here the focus is on the bound DHFR, which was therefore enriched in ¹⁵N and observed bound to unenriched GroEL [or a single-ring version, SR1, that is competent for binding (12)]. Next, the new experiments from Wüthrich and co-workers (8, 9) were used to address the two dominant factors that broaden resonance line widths in very large proteins: very efficient relaxation of proximal protons through mutual dipole–dipole coupling at slow tumbling rates and inadequately averaged variations of the chemical shift with different orientations of the molecule in the magnetic field, which also enhance relaxation. Horst et al. (5) deuterated the DHFR, thereby reducing the contribution of proton dipole–dipole coupling to relaxation, and applied transverse relaxation-optimized spectroscopy (TROSY) (8) and cross-correlated relaxation-enhanced polarization transfer (CRINEPT) (9), both of which take advantage of mutual cancellation of two pathways that contribute to the rate of transverse relaxation in amide groups. One can see the effect of this interference for a large protein in a simple 1D ¹H or ¹⁵N experiment: scalar coupling between the ¹H and ¹⁵N atoms in an amide group splits the ¹H and ¹⁵N resonance lines, and the two components have different line widths, particularly at slow tumbling times, because of relaxation effects. Conventional ¹H-¹⁵N correlation experiments (such as heteronuclear spin quantum coherence) are acquired with decoupling, which collapses the peaks of varying line widths, resulting in a peak whose intensity and line width retain contributions from the broad components. In a TROSY experiment, no decoupling is applied, and the single narrowest peak is selected from multiplets. This may seem counterintuitive, because a significant fraction of the NMR signal is thrown away, but for large proteins, the decrease in line width is appreciable, and a substantial augmentation of signal to noise is achieved. Similarly, the CRINEPT experiment achieves efficient signal enhancement in the case of very large slowly tumbling molecules and complexes through optimized cross-polarization. (For excellent general reviews on these techniques, see refs. 13 and 14.)

The technical aspects of the NMR methods may be daunting, but the experimental setup in Horst et al. (5) is straightforward: they diluted denatured ¹⁵N- and ²H-labeled DHFR into refolding buffer containing unlabeled GroEL or SR1. That the spectrum truly represents GroEL-bound DHFR is shown by using the same protocol to dilute denaturant away from DHFR, but in the absence of GroEL, which leads to no soluble protein. In addition, the GroEL–substrate complex is shown to be a true functional intermediate in the refolding cycle, because addition of GroES and ATP releases enzymatically active DHFR with peak dispersion and line widths characteristic of a folded protein.

The Trapped Folding Intermediate

What are the physical characteristics of GroEL-bound DHFR by NMR? First, the bound substrate does not yield the kind of well defined spectrum seen for GroES in its complex with GroEL (10). Instead, the DHFR peaks are ill-dispersed, suggesting there is no regular secondary structure, and broad, which could arise from either dynamics or static heterogeneity. Intriguingly, the overall peak volume is only ≈25% of what would be expected, for reasons that are not yet clear. However, many possibilities, not mutually exclusive of one another, come to mind. DHFR could be bound in many different modes, it could be binding and unbinding during the course of the experiment, or it may be tethered at a few sites such that it is possible to observe only some regions. On the other hand, deuterium labeling of SR1, which should reduce the efficiency of relaxation pathways for portions of DHFR that are close to SR1, had no effect on the DHFR spectrum, suggesting that there are not specific sites of binding. Surprisingly, the CRINEPT magnetization transfer for GroEL-bound DHFR is not nearly as efficient as one might predict based on the molecular size of this complex, and the 1D ¹H spectrum of the bound DHFR shows characteristics of a significantly more rapidly reorienting entity than the chaperonin should yield (the broad component of the peaks should not be discernable, but it is present). Both of these observations ultimately point to the fact that the transverse relaxation is not as fast as it would be if the substrate tumbled rigidly with the complex. Thus, the bound DHFR must have significant mobility.

How does this image of a GroEL-bound folding intermediate as a dynamic entity with heterogeneous binding to surfaces of GroEL subunits fit with previous information about the nature of chaperonin–substrate interactions? Results from peptide studies using transferred nuclear Overhauser effects (15–17) or crystallography (18) have established that various conformations can bind to GroEL; the common feature that mediates binding is exposure of a patch of hydrophobic surface (complementary to the binding surfaces on the apical domains of GroEL). Atomic structures (18, 19) implicated two helices on the inner binding surfaces of the apical domains in polypeptide binding, and mutagenesis results (20) also suggest that hydrophobic residues on an underlying extended segment are crucial for substrate binding. Thus, it would appear that there could be many different potential binding sites on a given protein substrate, that there may also be multiple binding modes on the chaperonin (21), and that the substrate may bind in many different conformations. The resulting heterogeneous and dynamic ensemble of bound states would be consistent with the NMR observations.

A noteworthy aspect of the results presented by Horst et al. (5) is that the NMR spectrum measured after release of DHFR from chaperonin is different from the spectrum of the protein that was refolded in vitro by dilution from urea, even though both protein samples are found to be enzymatically active. Some resonances are lacking, and others have shifted. Further analysis of these results could provide insight into how chaperonins influence the folding landscape of a polypeptide. Chaperonin interactions have been proposed to “smooth” the energy landscape of obligate substrates, which otherwise would have rugged energy landscapes typical of highly “frustrated” proteins (those with many competing and strong intramolecular interactions and complex topologies). The observation that a product of GroEL-mediated folding differs in some way from a product of a spontaneous refolding reaction raises the heretical possibility that complex proteins may have route-dependent final structures, and that their folding in vivo may be complex and may rely on information and interactions extrinsic to the folding chain. We eagerly anticipate future results on the well studied chaperonin machinery that will help shed light on the realities of the assisted folding of complex proteins.