Protein Folding and Unfolding at Atomic Resolution

Cell, Vol. 108, , February 22, 2002, Copyright 2002 by Cell Press Protein Folding and Unfolding at Atomic Resolution Review Alan R. Fersht 1,3 and Valerie Daggett 2,3 1 Department of Chemistry and
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Similar Documents
Cell, Vol. 108, , February 22, 2002, Copyright 2002 by Cell Press Protein Folding and Unfolding at Atomic Resolution Review Alan R. Fersht 1,3 and Valerie Daggett 2,3 1 Department of Chemistry and MRC Centre for Protein Engineering University of Cambridge Lensfield Road Cambridge CB2 1EW United Kingdom 2 Department of Medicinal Chemistry University of Washington Seattle, Washington Experiment and simulation are now conspiring to give atomic-level descriptions of protein folding relevant to folding, misfolding, trafficking, and degradation in the cell. We are on the threshold of predicting those protein folding events using simulation that has been carefully benchmarked by experiment. Protein folding and unfolding are fundamental events in the cell that have been very difficult to characterize in detail, even in vitro. However, the past decade has seen a revolution in experimental and theoretical methods that can describe folding at atomic-level resolution. Experiments using engineered mutations as precisely targeted probes (φ value analysis) define the structures of intermediates and transition states in folding and unfolding pathways at near-atomic resolution. NMR is pin- ning down the structure of the denatured state. Molecu- lar dynamics simulations can unravel whole pathways of unfolding. Most importantly, experimentalists and theoreticians are working together to solve the prob- lems. The synergy between experiment and theory is increasing as their timescales merge because of faster computers and the discovery of ultrafast folding proteins. In this review, we outline the application of these methods and their results, and how they are being trans- ferred from studies in vitro to processes in vivo. We begin by discussing the importance and relevance of studying protein folding in vitro and in silico to in vivo protein folding, misfolding, and disease. Folding In Vivo versus Folding In Vitro The first question asked by a cell biologist about in vitro protein folding studies is: do proteins fold the same way in vitro as they do in vivo? We know that proteins are slowly biosynthesized on the ribosome from the N ter- minus and that there is a host of complicated molecular chaperones in the cell that play diverse supportive roles in folding. The answer is generally, yes, small proteins up to the size of kda or so, which is the size of a domain of a larger protein, do fold in the same way. These small proteins are also the very ones whose pathways of folding we are beginning to unravel at atomic 3 Correspondence: (A.R.F.), edu (V.D.) resolution. The relevance of folding in vitro to folding in vivo was established using two small proteins, CI2 (chymotrypsin inhibitor 2, 64 residues) and barnase (Bacillus amylolquiefaciens Rnase, 110 residues), the two proteins on which many of the modern methods of studying folding were pioneered. Neither protein will fold until a stretch of C-terminal residues is free to interact with the rest of the protein (Neira and Fersht, 1999; Neira et al., 1997). Hence, those small proteins have to leave the channel in the ribosome in order to fold. However, larger proteins that are made of a series of domains may start to fold from the N terminus while the C-terminal sequences are still bound. Once CI2 and barnase have escaped the ribosome, they also fold so rapidly and bind so weakly to the key molecular chaperone GroEL that they are effectively immune to its influence. Barnase and CI2 fold with half-times of some 50 and 10 ms, respectively. They bind very weakly to the major form of GroEL in the cell, the GroEL 14 GroES 7.ATP complex, but much more tightly to GroEL, which is only transiently formed on the hydrolysis of ATP. The half-time for the ATPase activity of GroEL is many seconds, so that CI2, barnase, and other fast-folding proteins that are tran- siently bound to GroEL 14 GroES 7 complexes fold before they can enter the chaperoning cycle (Corrales and Fersht, 1996). The slowness of the ATPase activity of GroEl.GroES may indeed be a gatekeeper (Corrales and Fersht, 1996) or timer (Weissman et al., 1996) activity that has evolved to avoid GroEL being used unnecessarily. But larger, slow-folding proteins may enter the chaperoning cycles. Further, although GroEL bound barnase folds 500 times more slowly than when free in solution, it follows the same mechanism (Gray et al., 1993). We can assume that what we learn about the mechanism of folding of small, fast-folding proteins in vitro will apply to their folding in vivo and, to a large extent, to the folding of individual domains in larger proteins. The second question asked by a cell biologist about in vitro protein folding studies is: does understanding protein folding at atomic resolution have any biological relevance? There are many diseases of protein misfold- ing, aggregation, and instability that are caused by inher- ent properties of a wild-type protein, by changes in the protein environment, or by mutation. Understanding the mechanism of folding may lead to ways of designing drugs to correct these diseases. Studies of protein unfolding in vitro are very relevant since unfolding of many proteins is an important process in the cell. For example, transient unfolding can occur on transfer across membranes; there is reversible unfolding during the action of proteins such as titin; unfolding is a step in protein degradation via the proteasome; and full or partial un- folding is a key step in amyloidosis. In reality, most diseases of misfolding are diseases of unfolding; for example, transthyretin, gelsolin, and the prion protein initially fold correctly. Some of the unfolding in vivo is an active process, using methods or components that can be mimicked or reconstructed in vitro. Other processes are spontaneous and can be directly studied in vitro. Cell 574 Fundamental Problems in Studying Folding The discovery of two-state folding proteins revolutionized Protein folding involves very small overall changes in the study of protein-folding kinetics for several rea- energy, typically ranging from 1 to 15 kcal mol 1, as a sons. First, there is the very simplicity of the system that protein progresses from its denatured state, possibly lends itself to study both by simulation and experiment. via intermediates, to its native structure. These changes Second, the transition state for folding and unfolding is are equivalent to just the strength of a few hydrogen one and the same, which means that the transition state bonds; denatured structures make many interactions for folding can be analyzed by measuring or simulating with water, which almost compensate for the interactions the unfolding reaction kinetics, which will be seen later in the native structure. Further, the states popu- to be essential. Third, small, two-state proteins are the lated along the pathway are not discrete, single entities size of a single domain in a larger protein. Much of but are ensembles of structures. The denatured state advanced protein folding study nowadays is of twostate in particular is a heterogeneous collection of rapidly proteins, since they give information on the very interconverting structures, some of which have flick- early events in folding and basic principles on the folding ering native-like elements. As the protein folds, the ensembles of domains that apply to larger proteins. become tighter, to eventually give the native structure that fluctuates around the structure as seen Protein Engineering and φ Values by X-ray crystallography. The study of mechanism in The transition state does not accumulate and so, apart simple organic molecules usually deals with the forma- from a few special cases, it has to be characterized tion of strong covalent bonds and stable intermediates indirectly. How do we study these elusive transition that often can be isolated and characterized. The inter- states? There is a classical technique of chemistry, mediates in protein folding differ mainly by a myriad of value analysis, that is used to measure the extent of weak noncovalent bonds. The challenge to the experi- bond formation in a transition state (Fersht, 1999): a mentalist and theoretician is to cope with the small ener- substituent is introduced into a nonreacting part of an getic changes and to devise methods for describing the organic reagent, and the relative effects of the substituprogress of all the side chains and the backbone at ent on the energy of activation and the free energy of atomic resolution as the protein either folds or unfolds. the equilibrium give a measure of the extent of bond There are only three methods currently available for this making and breaking in the reaction. Changes in struclevel of resolution: NMR for equilibrium structures in ture are thus inferred from changes in energetics. The solution, in particular intermediates and denatured next achievement of recombinant DNA technology was states; kinetics on engineered mutants to give fine details to allow a similar, but not identical, procedure to be of the structures of transition states and intermediates (φ introduced into the analysis of side chain interactions values); and atomic-level simulation (molecular dynam- in proteins (Fersht et al., 1986, 1987). Using site-directed ics, MD). These methods are increasingly being com- mutagenesis, we alter the direct interaction of a side bined to describe complete unfolding pathways at chain of a protein with a ligand or with other residues atomic resolution, which is the subject of this review. in a protein, which will perturb the energetics of the reaction. Then, by comparing the effects of the substitu- Folding Intermediates tion on the rate and equilibrium constants of the reac- Classically, protein folding was considered to proceed tion, we can determine whether or not that interaction via a series of intermediates, with partially formed struc- is formed in the transition state of the reaction. More tural elements accruing during the folding reaction (Kim specifically, we calculate the quantity φ for protein foldand Baldwin, 1982). Much of the evidence for the neces- ing from the following equation (Fersht et al., 1986, sity of intermediates was the result of the limited choice 1992): φ G D / G N D, where G D and G N D of proteins available for study. Hence, most of the classi- are the changes in the free energies of activation and cal folding studies involved characterizing those inter- the free energies of folding caused by mutation (see mediates. However, the first impact on protein folding legend to Figure 1). Each mutation acts as probe for the of recombinant DNA technology was to allow the pro- formation of structure at the site of mutation (Fersht et duction of large quantities of proteins and their mutants al., 1992), and therefore, fine structure can be inferred that were very suitable for extended study: T4 lysozyme from energetics. A φ value of 0 implies that the structure (Alber et al., 1987), Staphylococcal nuclease (Shortle, of the transition state at the site of mutation is the same 1986), and barnase (Matouschek et al., 1989). Then, as in the denatured state, while a φ value of 1 implies small proteins, beginning with CI2, were found that fold that the structure is native-like at the site of mutation with simple two-state kinetics and no detectable inter- in the transition state. Fractional φ values imply either mediates (Jackson and Fersht, 1991). partial unfolding in the transition state, which has been verified experimentally (Fersht et al., 1994), or a mixture Folding Transition States and Two-State Folding of folded and unfolded states. The only state accessible to experimental study in a φ value analysis is the only experimental technique two-state reaction is the transition state (Figure 1). A available for fine structure analysis of transition states transition state is the highest energy point on a reaction in general, and the transition state is the only entity pathway. One of the characteristics of a transition state that can be studied in the folding of two-state proteins. is that a molecule in its transition state structure should Consequently, φ values are increasingly being applied collapse with equal frequency to its starting materials or to a wide variety of proteins, with more than 120 experi- to its products. This defining characteristic has recently mental and theoretical studies in the last three years. been shown by simulation for CI2 (Li and Shakhnovich, More sophisticated approaches extend φ values to 2001). study the movement of transition states with structure Review 575 Figure 1. Reaction Profile and φ Value Analysis Schematic profiles are sketched in red for a protein that has an alanine residue (A) in a helix, and in blue for a mutant in which the alanine is mutated to a glycine. (Left) The transition state ( ), at the top of the energy profile, has the helical region as denatured as in the denatured state D. The energy of the transition state is affected by A to G by the same energy as in D, and so the change in energy of relative to that of D, G D,is0.Thus,φ G D / G N D 0. (Right) The opposite case when the helix is fully structured in the transition state has G D G N D, and so φ 1 (modified from Fersht, 1999). The value of G D is calculated from the ratio of rate constants for folding of wild-type (k f(wt) ) and mutant (k f(mut) ) proteins [ G D RTln(k f(wt) /k f(mut) )]. The value of G N D is calculated by subtracting the free energy of folding of wild-type protein ( G N D(wt) ) from that of mutant ( G N D(mut) ). The free energies of folding are usually measured from urea-, guanidinium chloride-, or thermaldenaturation curves. φ value analysis requires measuring rate and equilibrium constants. and reaction conditions (Jager et al., 2001; Matthews atomic-level resolution of events in folding requires MD and Fersht, 1995; Oliveberg, 2001). Recently, φ values simulation, the application of Newton s laws of motion have been applied directly to analyze secondary struc- to every atom in the protein using empirical energy functions tural interactions in small proteins that have been chemically and computation (McCammon et al., 1977). For synthesized with backbone substitutions (Fergu- real proteins, the simulations are generally applied in son et al., 2001). φ values also provide benchmarks the direction of unfolding for two reasons. First, simula- for computer simulation of folding. Recently, computer tions are currently restricted to timescales of less than simulations have used φ values as distance constraints 1 s (Daggett, 2000), which is far too short for the time- for prediction of transition states in a manner analogous scale of 1 ms for the half-time of folding of most proteins. to distance constraints in NMR structure determination But, the rate of unfolding increases at high temperto (Vendruscolo et al., 2001). atures and so most proteins unfold in less than 1 ns at 225 C. Secondly, unfolding occurs from the bestcharacterized Mapping Residual Structure in the Denatured state on the pathway, the native state, in State by NMR contrast to folding, which starts from the least-known NMR spectroscopy is the best method to study denatured and very heterogeneous denatured state. However, for and partially denatured states in solution (e.g., two-state proteins, the transition state for folding and Shortle and Ackerman, 2001). The denatured state is unfolding is expected from the principle of microscopic best described as an ensemble of conformations inter- reversibility to be the same, which has been confirmed converting faster than the NMR chemical shift timescale experimentally (Itzhaki et al., 1995), and therefore, un- ( 10 3 s 1 ). At one extreme, as exemplified by CI2, the folding simulations give the structure of the folding transition denatured state is expanded. However, in general, the state for two-state folding proteins as well as the ensemble is not random but has regions that are biased unfolding transition state for multistate reactions (reviewed toward the structural preferences of the native structure by Daggett and Fersht, 2000). Intermediate as well as nonnative interactions. The extent and nature states and the unfolding pathway were the first to be of the residual structure varies with the environmental targeted for characterization via denaturing molecular conditions and typically involves transient hydrophobic dynamics simulations (Daggett and Levitt, 1992), and it clusters and residual but unstable secondary structure has now become quite a popular technique (reviewed and turns. Although these experimental methods have by Daggett and Fersht, 2000; Shea and Brooks, 2001) revolutionized our view of the denatured state, the data At this stage in the development of MD simulations, are insufficient for direct transformation into molecular it is essential to benchmark them by experiment for two models. reasons. First, the potential functions they employ are empirical and do have approximations. Second, extrapolation Molecular Dynamics Simulations of Protein from unnaturally high temperatures in silico to Unfolding/Folding Pathways experimentally accessible temperatures adds more uncertainty. Simplified models of protein folding have enriched our So far, however, there has been excellent understanding of the fundamental principles of protein agreement between simulated and measured φ values folding and can even approach atomic-level resolution as well as other comparisons. Repetitive simulation of (see Onuchic et al., 2000, for a recent review). But, full the same unfolding reaction (e.g., for CI2, Li and Daggett, Cell ; Lazaridis and Karplus, 1997) shows that there are sents a basic folding unit and as such is a model for variations in the pathway and transition state of unture folding units in larger multimodular proteins. The strucfolding, but the different structures form an ensemble of the common transition state for folding and un- that fluctuates around the experimental data, which are, folding has been studied by a variety of techniques, of course, the average over a large number of molecules. including a φ value analysis using 100 mutations span- More precise benchmarking will require developet ning the length of the protein (Itzhaki et al., 1995; Otzen ments in computer power by some three orders of magtruncated al., 1994) and the structures of a large number of nitude so that simulation can be done to 1 ms and mutants and peptide fragments (Gay et al., greater. Steps are being taken in this direction in two 1995). MD simulations started almost as early as the ways. IBM is constructing a massive parallel computer experiments (Li and Daggett, 1994; Daggett et al., 1996). the Blue Gene project ( The φ values are independent of whether unfolding or bluegene/). Pande and colleagues have distributed refolding are measured, as expected for two-state kinet- screen savers that calculate discontinuous folding tra- ics. The φ values for CI2 tend to fall between 0.2 and jectories during the idle time of PCs around the world 0.5. There are some higher values that are found in the (Zagrovic et al., 2001). However, a step forward has been helix, particularly at the N terminus, and in the sheet made recently by the discovery by the experimentalists for residues that dock with the helix. In general terms, of very fast unfolding and folding proteins that unfold the transition state for folding resembles a distorted on the 1 10 ns timescale at 100 C, so that experiment form of the native state, which appears to be increas- and simulation can be compared more directly (Mayor ingly stretched out from the N-terminal region of the et al., 2000; Ferguson et al., 2001). Importantly, these helix and where it docks onto the sheet. Secondary studies have shown that the overall unfolding pathway structure is being consolidated at the same time as long- is independent of temperature and th
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!