The transition state for integral membrane protein folding

Transcription

1 The transition state for integral membrane protein folding Paul Curnow 1 and Paula J. Booth 1 Department of Biochemistry, School of Medical Sciences, University of Bristol, University Walk, Bristol BS8 1TD, United Kingdom Edited by Alan Fersht, University of Cambridge, Cambridge, United Kingdom, and approved November 21, 2008 (received for review July 18, 2008) Biology relies on the precise self-assembly of its molecular components. Generic principles of protein folding have emerged from extensive studies on small, water-soluble proteins, but it is unclear how these ideas are translated into more complex situations. In particular, the one-third of cellular proteins that reside in biological membranes will not fold like water-soluble proteins because membrane proteins need to expose, not hide, their hydrophobic surfaces. Here, we apply the powerful protein engineering method of -value analysis to investigate the folding transition state of the alpha-helical membrane protein, bacteriorhodopsin, from a partially unfolded state. Our results imply that much of helix B of the seven-transmembrane helical protein is structured in the transition state with single-point alanine mutations in helix B giving values >0.8. However, residues Y43 and T46 give lower values of 0.3 and 0.5, respectively, suggesting a possible reduction in native structure in this region of the helix. Destabilizing mutations also increase the activation energy of folding, which is accompanied by an apparent movement of the transition state toward the partially unfolded state. This apparent transition state movement is most likely due to destabilization of the structured, unfolded state. These results contrast with the Hammond effect seen for several water-soluble proteins in which destabilizing mutations cause the transition state to move toward, and become closer in energy to, the folded state. We thus introduce a classic folding analysis method to membrane proteins, providing critical insight into the folding transition state. phi value kinetics thermodynamics protein engineering Protein folding plays a central role in biology. Folding investigations provide key information on protein structure and dynamics, while protein misfolding can have serious disease implications (1, 2). To fold correctly, proteins must overcome an activation barrier to pass through a high-energy transition state. Understanding the nature of this folding transition state is important in resolving how a protein folds to a stable and functional native structure (3, 4). The most powerful method available to probe the structure and energetics of the folding transition state combines sitedirected mutagenesis, equilibrium thermodynamics, and kinetic dataina -value analysis. This approach has revolutionized protein folding studies by providing a quantitative description of the environment experienced by individual side chains in the folding transition state and has been applied to the folding of many small, water-soluble proteins (3, 5 8). However, the -value method has yet to be applied to larger integral membrane proteins. Integral membrane proteins are a special case in protein folding because they are adapted to the lipid bilayer rather than to the cytoplasmic milieu (9). The sequences of transmembrane regions are biased in favor of hydrophobic amino acids to match the low dielectric of the membrane interior. This presents experimental difficulties for in vitro studies of protein folding as the proteins need to be solubilized in lipids or detergents and are often resistant to common denaturants such as urea. Consequently, the current understanding of membrane protein folding is poor compared with the advanced state of knowledge for water-soluble proteins. In particular, there are few quantitative studies of the kinetics and thermodynamics of membrane protein folding. Yet, understanding these folding parameters gives valuable information on a major problem: how to fold and stabilize membrane proteins for structural and functional studies. A value is a measure of the change in activation energy, relative to the change in the overall free energy of folding, induced by the directed mutation of a single amino acid. values are expected to fall between one and zero (these extreme values mean the change in the free energy of the transition state is the same as either that in the folded or unfolded state, respectively). Based on this energetic perturbation, the magnitude of can be interpreted as the extent of native interactions at the site of the mutated residue in the transition state, with a value of one indicating native energetics and interactions. Intermediate values are often observed (6), suggesting that a proportion of native-like contacts are formed or that parallel reaction pathways exist. We present a -value analysis of the major folding step of an -helical membrane protein. The protein of choice for this study is bacteriorhodopsin (br), a light-activated proton pump from the purple membrane of Halobacterium salinarum. We have previously established conditions for reversible folding of br in mixed lipid/detergent (1,2-dimyristoyl-sn-glycero-3-phosphocholine, DMPC)/3-[(3-cholamidopropyl)dimethylammonio]-1- propanesulfonate, CHAPS) micelles (10). Currently, br is the only helical membrane protein that meets the criteria for -value analysis, namely (i) it can be partially unfolded with a reduction in secondary structure to the apoprotein bacterioopsin (bo) by sodium dodecylsulfate (SDS) in a microscopically reversible, cooperative, two-state reaction (10, 11); (ii) this process can be described by linear free-energy relationships to obtain the overall free energy change of unfolding ( G u H 2 O ) and the folding and unfolding rate constants in the absence of denaturant (k f H 2 O and k u H 2 O ); and (iii) the effect of single point mutations can be interpreted in the context of a detailed native-state structure (12). br is a relatively large protein (248 aa) with seven transmembrane helices, so we begin our survey of the folding transition state by determining the values of a discrete region of the protein. We use an alanine scan to investigate a single helix of br, helix B, which is thought to form early during folding and does not make extensive interactions with the retinal cofactor. Results Kinetic and Thermodynamic Measurements. The absorbance band of the retinal chromophore of br reports on the conformational state of the protein. br is purple when folded with an absorbance Author contributions: P.C. and P.J.B. designed research; P.C. performed research; P.C. analyzed data; and P.C. and P.J.B. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1 To whom correspondence may be addressed. p.curnow@bristol.ac.uk or paula.booth@bristol.ac.uk. This article contains supporting information online at /DCSupplemental by The National Academy of Sciences of the USA BIOPHYSICS cgi doi pnas PNAS January 20, 2009 vol. 106 no

2 band at 560 nm. This 560-nm absorbance decays when br in DMPC/CHAPS is unfolded by the addition of SDS. We have previously shown that these absorption changes correlate with a reduction in secondary structure (10). The reaction we study here is the transition between the partially denatured apoprotein state in SDS (referred to as the SDS-unfolded state) and the folded br state. We have previously shown that this process corresponds to the final, major folding step of br (10) and behaves as a two-state reaction. The SDS state is not completely unfolded, but has helical content equivalent to approximately four of the native seven helices of folded br (13). Thus, the SDS-unfolded state has much more ordered structure than the urea- or guanidinium chloride-induced unfolded states frequently used in water-soluble folding studies. The folding reaction we study here is more akin to the later stages of multistate water-soluble protein folding, that is, from a structured intermediate to a folded state. The solvent for the br unfolding reaction is also relatively complex; folded br is solubilized in DMPC/CHAPS micelles and begins to unfold when SDS is added and forms mixed DMPC/CHAPS/SDS micelles. At higher SDS concentrations, SDS micelles dominate and solubilize the partly unfolded br state. Free energies of unfolding of wild-type (WT) and mutant br were obtained as described previously from equilibrium denaturation curves (11) as well as from time-resolved, stopped-flow absorption measurements of the folding and unfolding rates (10). Both the equilibrium and kinetic data fit a two-state reaction, and linear free energy relationships were observed, enabling extrapolation to zero SDS denaturant. Linear Free Energy Relationships for Alanine Mutants. -value analysis is most successful when single point mutations are introduced that perturb the overall free energy of unfolding, G H O u 2, by 0.6 kcal.mol 1 (14). On this basis, we selected nine of 24 residues within br helix B that are potentially suitable for -value analysis when mutated to alanine ( 0.6 kcal.mol 1 G H O u kcal.mol 1 ) (11). These are D36A, K41A, F42A, Y43A, T46A, T47A, I52A, F54A, and M60A; several of these would be considered relatively extreme mutations in water soluble proteins but reflect the prevalence of large aromatic side chains in integral membrane proteins. The logarithms of the folding and unfolding rate constants (k f and k u, respectively) of each mutant were linear with SDS (10), giving a characteristic chevron plot (see Fig. 1A and Table 1). Because the experimentally observed rate constant k obs is the sum of k f and k u, the characteristic downward arrow of the chevron plot can be fit by the sum of two linear functions and extrapolation of these lines to the y axis gives the folding and unfolding rates in the absence of denaturant, termed k H O f 2 and k H O u 2, respectively. These kinetic parameters were used to calculate the difference in the overall free-energy change upon mutation, G H O u 2, and the values were very similar to G H O u 2 from equilibrium experiments, showing that the extrapolations of both kinetic and equilibrium data to zero denaturant are in good agreement. The consistency of the G H O u 2 values determined from kinetic and equilibrium data, also confirms that the mutants fold by a two-state process as previously established for wild type br (see Fig. 1 B and Table 1). The refolding yield of all mutants was close to wild-type levels at 75% (see Supporting Information (SI) Fig. S1). Fig. 1C shows that destabilising mutations have equal and opposite effects on the observed folding and unfolding rates (k H O f 2 and k H O u 2, respectively). As the folding rate slows, the unfolding rate becomes faster to a similar degree, with gradients of 0.86 and 0.69, respectively. The mutations T47A and I52A induced additional kinetic phases and unusually fast unfolding rates that are omitted from the dataset. Fig. 1. Kinetics of folding and unfolding. (A) Chevron plot of a typical mutant (open circle and solid line) compared with WT br (dashed line). (B) G H O u 2 derived from kinetic chevron plots ( G chev u ) and equilibrium measurements ( G eqm u ). The negative sign of G u indicates destabilisation and a reduction in the magnitude of G u. Correlation coefficient, 0.95; gradient, (C) Linear changes in overall folding and unfolding rates, k H O f 2 (filled circles) and k H O u 2 (open circles), respectively] for destabilizing mutants. For k H O f 2, gradient, ; correlation coefficient, 0.74; for k H O u 2, gradient, ; correlation coefficient, cgi doi pnas Curnow and Booth

3 Table 1. Summary of kinetic data eqm, F chev F eqm F chev, Gu, Gu eqm, GTS-U H 2 Oeqm, mu-f chev *, Gu H 2 Ochev, mu-f H O s 1 2, mts-f*, Gu H s 1 kf H 2 O, mts-u*, ku Mutant WT ** E E D36A E e K41A E E F42A E E Y43A E E T46A E E T47A E E I52A E E F54A E E M60A E E *mts-u, mts-f and mu-f chev calculated from Eqs. 6, 7, and 8; only the magnitude of mu-f is given. G u chev and Gu eqm are the change in total free energy of unfolding determined from kinetic chevron plot ( Gu H 2 Ochev ) and equilibrium data ( Gu H 2 Oeqm ), respectively. G TS-U from Eq. 11. F chev and F eqm calculated from Eq. 12 using extrapolated chevron plot and equilibrium data, respectively. F calculated from Eq. 12 using data at SDS. Very similar values were obtained at SDS. from Eq. 14. **All errors are either standard errors from curve fitting or from propagation. Raw kinetic data of the most destabilizing mutation T46A had lower signal-to-noise than the other mutants and folding data fit less well to single exponential function, giving rise to a larger error in Gu H 2 O. T47A and I52A show unusual kinetics and are not analyzed further. Fig. 2. Changes in denaturant response upon mutation. (A) Dependence of kinetic m values, reflecting the dependence of the folding and unfolding rates on SDS (m TS-U and m TS-F, respectively) on the overall free energy ( G H O u 2 ). For m TS-U, gradient, ; correlation coefficient, 0.66; for m TS-F, gradient, ; correlation coefficient, The Inset shows the trend for the overall m value, m U-F, where gradient, ; correlation coefficient, Errors from chevron plot curve fitting. (B) Dependence of (m TS-U /m U-F )on G TS-U. G TS-U [i.e., G TS-U (WT) - G TS-U (MUT)] has a negative sign because G TS-U (MUT) G TS-U (WT). Gradient, ; correlation coefficient, Denaturant Response and Movement of the Transition State. The gradient of each arm of the chevron plot in Fig. 1A reflects the dependence of k f or k u on denaturant concentration. Fig. 2A shows that for destabilizing mutations these gradients, or m values, are linear with G u H 2 O. The absolute changes in m TS-U and m TS-F (the gradients for k f and k u, respectively) are similar and opposite in sign, although the relative change in m TS-U is greater. The overall m value for unfolding, m U-F can be calculated from the sum of m TS-U and m TS-F (Eq. 8). Thus, these folding and unfolding m values make almost equal contributions to the change in the overall m U-F with G u H 2 O. This latter change in m U-F (see Fig. 2A, Inset) reflects the dependence of G u on SDS and shows that there is a smaller variation of G u with SDS for destabilizing mutations than for WT. The ratio m TS-U /m U-F for each mutant gives, a measure of the location of the transition state along the reaction coordinate BIOPHYSICS Curnow and Booth PNAS January 20, 2009 vol. 106 no

4 relative to the native and unfolded states (4, 15, 16). We previously reported that is low for WT br (10), implying that the transition state is close to the SDS-unfolded state. This is not surprising given the amount of residual structure retained in SDS. Here, we find that shows a linear decrease with increasing G TS-U, the change in the activation energy of folding (Fig. 2B). Thus, destabilizing mutations cause an apparent movement of the transition state toward the SDS-unfolded state. This movement is correlated with slower refolding rates. Accordingly, we see an approximately equal and opposite trend in the relationship of the activation energy of unfolding, G TS-F,to m TS-F /m U-F, although the errors are much larger because of the extrapolation of the chevron plot (data not shown). -Value Analysis. Folding values were determined for each mutant and were in good agreement whether the denominator G u H 2 O was calculated from chevron plot data ( F chev ) or derived from equilibrium experiments ( F eqm ) (Table 1). values using folding and unfolding data were closely correlated (Eq. 13, Fig. S2). To avoid errors arising from extrapolation, we also calculated values using G TS-U and G u at a non-zero denaturant concentrations (0.401 and SDS, Table 1). These latter values closely matched those determined for zero denaturant, with the exception of mutant F42A. Fig. 3 shows values for the helix B mutants mapped onto the native structure of br. Most of the values are grouped close to one ( 0.8), suggesting native energetics and interactions of much of this helix in the transition state. However, mutants Y43A and T46A have lower values of 0.3 and 0.5, respectively. Such intermediate values are less straightforward to interpret, especially for polar to nonpolar amino acid changes as seen here (5). Discussion There is a dearth of information on the folding mechanisms of integral membrane proteins. We show here that it is possible to apply one of the most powerful analytical folding methods to obtain mechanistic detail for a helical membrane protein. Combined kinetic, thermodynamic, and mutagenesis studies in the formofa -value analysis have enabled us to begin mapping out the structure and energetics of the transition state for the major, final folding step of bacteriorhodopsin. This type of analysis has been applied to many water-soluble proteins, but we have used it for membrane protein folding. We find several differences in the folding of br when compared with water-soluble proteins, notably in the denaturant response and the position of the transition state with respect to the folded state. Solvent Interactions and Structured Unfolded States. There are important distinctions between the folding of membrane proteins and water-soluble proteins both in the nature of the experimental systems and the resulting data. The solvent environment and denaturant interactions are significantly different; while water-soluble proteins are surrounded by a homogenous aqueous environment, br is solubilized in mixed lipid/detergent micelles in water. Thus, there are interactions of the transmembrane domains of br with the hydrophobic chains and polar head groups of lipid and detergents as well as of the extrinsic, interhelical loops of the protein with water. The result is a complex array of protein-solvent interactions involving surface amino acid residues that are polar, apolar, and charged. A small chaotrope such as urea gives relatively unstructured states for water-soluble proteins. In contrast, the detergent SDS is used here, resulting in only partial denaturation and the interactions between SDS and br during unfolding are poorly understood (17). SDS is an anionic surfactant with a long hydrocarbon tail and is a bulky denaturant with physical properties that are quite different from those of urea. The precise structure of the SDS-unfolded state is unclear. It Fig. 3. Structural context of values. (A) most mutations give 0.8 (deep red) but values for Y43A and T46A are lower (0.3 and 0.5, respectively). (B) close up of the boxed region from (A) viewed from above shows Y43A packing against helices A and G and T46A facing helix G. Fig. 3 constructed from PDB entry 1C3W using Pymol (38) cgi doi pnas Curnow and Booth

5 cannot bind the retinal cofactor and lacks native tertiary interactions (10) but is more structured than the denatured states of most water-soluble proteins. Circular dichroism (CD) spectroscopy has shown that the SDS state has an -helical content equivalent to approximately 4 transmembrane helices (13) with a mean residue ellipticity at 226 nm for the folded and unfolded states of and deg.cm 2.dmol 1, respectively (10) (compared with less than deg.cm 2.dmol 1 for a random coil). This reduction in helix in SDS could arise from fraying of some helix ends, or from losses of some transmembrane helices, probably at the C-terminal end of the protein (13, 18, 19). The partly helical bo thus provides a different reference position to the more extensively unfolded forms of water-soluble proteins, but nonetheless enables the examination of helix helix interactions within a two-state system. SDS is expected to affect br by changing the physical properties of the DMPC/CHAPS micelle (because SDS will incorporate, forming DMPC/CHAPS/SDS mixed micelles) as well as by direct protein-sds interactions. Although these solvent properties mean a smaller denaturant concentration range is accessible by experiment for br than for water-soluble proteins, linear relationships in the form of chevron plots are observed for the folding and unfolding kinetics of the mutants studied here (see Fig. 1A). By analogy with water-soluble protein studies, these linear relationships suggest that stability correlates with the relative degree of denaturant-accessible surface area. Indeed, a direct correlation between the change in unfolding free energy and buried surface area has previously been noted for br (11). However, the change in unfolding free energy with SDS will also reflect alterations in the properties of the DMPC/CHAPS/SDS micelles that destabilize the protein. Such micelle properties include micelle size, rigidity, and the lateral pressure imposed on the protein by the lipid/detergent chains (20, 21). The linear relationships observed here give a description of denaturant behavior in a micelle system and further emphasize the experimental validity of using SDS as a denaturant in kinetic folding experiments on helical membrane proteins. Position of the Transition State on the Reaction Coordinate. The chevron plots of br exemplified in Fig. 1A are more asymmetric than those obtained for most water-soluble proteins (22), with the dependency of k f being rather shallow and that of k u being much steeper. This asymmetry reflects a different response to the denaturant during folding of the membrane protein compared with a water-soluble protein. The relative gradients of the dependence of k f and k u on SDS give a low value of approximately 0.1, indicating that the transition state for br is globally closer to the SDS-unfolded state than the folded state. Higher values of are generally seen for water-soluble proteins (23), showing that their transition states are closer to their folded states. However, because the unfolded state for br in SDS is structured, the transition state for this final folding step of br is also significantly structured. These findings agree with earlier reports on the importance of SDS in retaining a critical helical core for successful folding (13, 18, 24). The mutant br proteins also have low values and transition states close to the partly structured SDS-unfolded state, but there is an apparent movement of the transition state toward the SDS state with progressive destabilization of the folded state. The Hammond postulate (25) predicts that two consecutive states in a reaction, such as the unfolded and transition states, which have similar energies will have similar structures. Several water-soluble proteins exhibit Hammond behavior (15, 16) with destabilizing mutations moving the transition state toward the folded state and reducing the energy difference between the two states. We observe the converse for br; the folding activation energy increases despite the transition state approaching the denatured state. Our results also contrast with a recent singlemolecule force spectroscopy study (26) in which Hammond behavior was observed with a different set of br mutants. We attribute this disparity to the different unfolding methods used for force and chemical unfolding. A possible contribution to the apparent transition state movement reported here is perturbation of the SDS-unfolded state by mutation, as this unfolded state has a relatively high degree of structure and mutations can cause changes in the solvation energy of this SDS state (27). The changes in m values upon mutation (Fig. 2A) are consistent with unfolded state effects. Destabilizing mutations decrease the overall m value for the unfolding reaction, m U-F, and cause a relatively large increase the folding m value, m TS-F. This could partly arise from altered energy and interactions in the SDS-unfolded state of the mutants giving a smaller dependence of unfolding free energy on SDS. However, this effect is relatively small; m U-F values are between 87 96% of the WT value and both m TS-U and m TS-F contribute to the change in m U-F for br (28). Values and Structure in the Transition State. We observe high values for much of helix B (Fig. 3 and Table 1), which are consistent with previous suggestions (18, 29, 30) that most of this particular helix, and its structural interactions, are formed early in folding. However, the complexity of the br refolding system gives rise to some caveats. values are easiest to interpret if the free energy of the unfolded state is unaffected by mutation, but as discussed above, there are likely to be changes in the SDSunfolded state energy for br. Such changes in the SDS-bR-state free energy also account for the dependence of the m values m TS-U and m U-F on G U H20. As discussed by Fersht et al. (5), even if there are changes in the unfolded state free-energy values near zero or one are often reliable because the changes in the unfolded and transition states induced by the mutation often cancel out. Most of our values are close to one and the simplest interpretation of these values as presented here assumes any changes in unfolded state energy cancel. We also find similar values by extrapolation to zero denaturant as well as at particular SDS concentrations. However, the relatively limited range of denaturant concentrations that can be used in these br studies means values determined near the midpoint of the kinetic chevron plot are less reliable. We observe intermediate values at two positions, Y43 and T46, toward the cytoplasmic end of the helix. There are several possible explanations for intermediate values, including unfolded state effects and partial structure formation in the transition state (although the magnitude of is not linear with the extent of structure). They can also arise from other situations, such as an ensemble average of two discrete transition state populations with high and low values. Closer inspection of the region with fractional values (see Fig. 3B) shows that Y43 packs against helices A and G and T46 packs against helix G. Intriguingly, the light-induced form of bo (similar to the N intermediate of the photocycle) is characterized by movements of helices E G (31 33). In particular, the tops of helices F and G are displaced by several angstroms, with helix F tilting away from the center of the protein. A similar change may take place when the protein is denatured with SDS so that tertiary interactions between the tops of helices B, F, and G are only partially formed in the folding transition state (which would give rise to intermediate values). D36, however, has a high value of 0.9, implying native interactions at the cytoplasmic end of the helix. This is an exciting area for further investigation as we work toward a complete description of the transition state. Conclusions It has been unclear how easily the experimental methods developed for small water-soluble proteins will translate to integral membrane proteins. Here, we show that a classical approach BIOPHYSICS Curnow and Booth PNAS January 20, 2009 vol. 106 no

6 using linear free-energy relationships and site-directed mutagenesis can provide unprecedented insight into the folding transition state of br. The free energy surface of folding can be perturbed by mutation and transition state structure inferred from canonical values. However, we also find that the solvent and denaturant interactions are complex and may be quite different to those during folding of water-soluble proteins. Nonetheless, harnessing the potential of established methods to provide a full description of the folding pathway now appears to be a realistic goal for helical membrane proteins. The excellent agreement between equilibrium and kinetically derived parameters shown here, and the linear free-energy relationships in both datasets, is especially valuable and paves the way for further detailed studies in this field. Methods Expression of br Mutants. br mutants in plasmid pmpk85 (11, 34) were a kind gift from J. U. Bowie (University of California, Los Angeles, CA). H. salinarum strain L33 (which does not express WT br) was transformed with pmpk85 according to the protocol of Cline and Doolittle (35) and transformants were selected and maintained using mevinolin (Sigma) at 10 M in media plates and broth. Growth of recombinant cells and purification of recombinant br was according to Oesterhelt and Stoeckenius (36). All mutants were characterized by mass spectrometry (37). Data Collection and Analyses. The collection and analysis of kinetic data has previously been described in detail (10). Briefly, the protein was maintained in mixed micelles of DMPC/CHAPS and reversibly unfolded by the addition or removal of SDS. The kinetics of this process were determined in a stopped-flow apparatus by monitoring changes in the retinal chromophore absorbance band at 560 nm. All mutants except I52A and T47A showed folding and unfolding kinetics similar to WT br, with a single kinetic phase dominating folding and unfolding. This was further confirmed by using a photodiode array to observe simultaneous changes in absorbance at multiple wavelengths. Typically, the stopped flow was configured with a 2-mm pathlength and 2.3-nm bandwidth and protein was at 6 M. The refolding yields of all mutants were determined from the regeneration of the native chromophore using a CARY UV-Vis spectrophotometer and were comparable to WT yields. Graphs were prepared using Origin (Microcal, Inc.) or GraFit (Erithacus Software). The procedure for determining values (see SI Appendix) has been extensively described by Fersht and colleagues (22). Briefly, kinetic and equilibrium data were used to determine the effects of point mutations on the total free energy change upon unfolding ( G u H 2 O ) and the activation energy of folding ( G TS-U ) (Eq. 6, SI). The parameter is the ratio of the m value for folding (m TS-U ) to the overall m value (m U-F ), where the m values reflect the denaturant dependence of the activation free energy for folding and the overall free energy of the unfolding reaction, respectively. ACKNOWLEDGMENTS. We thank Duan Yang for mutant plasmids and help with H. salinarum transformations, Kathleen Moreton for preparing mutant protein, Jim Bowie for useful discussions and provision of plasmids and protein, and Sophie Jackson for a critique of the manuscript and analyses. We acknowledge funding from the BBSRC (BB/D001676). P.J.B. holds a Royal Society-Wolfson Research Merit Award and is a member of the E-MeP EU consortium. 1. Dobson CM (2003) Protein folding and misfolding. Nature 426: Balch WE, Morimoto RI, Dillin A, Kelly JW (2008) Adapting proteostasis for disease intervention. Science 309: Fersht AR (1994) Characterizing transition states in protein folding: An essential step in the puzzle. Curr Opin Struct Biol 5: Oliveberg M (2001) Characterisation of the transition states for protein folding: Towards a new level of mechanistic detail in protein engineering analysis. Curr Opin Struct Biol 11: Fersht AR, Matouschek A, Serrano L (1992) The folding of an enzyme I. Theory of protein engineering analysis of stability and pathway of protein folding. J Mol Biol 224: Otzen DE, Itzhaki L, ElMasry NF, Jackson SE, Fersht AR (1994) Structure of the transition state for the folding/unfolding of the barley chymotrypsin inhibitor 2 and its implications for mechanisms of protein folding. Proc Natl Acad Sci USA 91: Matouschek A, Kellis Jr, JT, Serrano L, Fersht AR (1989) Mapping the transition state and pathway of protein folding by protein engineering. Nature 340: Daggett V, Fersht AR (2003) Is there a unifying mechanism for protein folding? Trends Biochem Sci 28: Bowie JU (2005) Solving the membrane protein folding problem. Nature 438: Curnow P, Booth PJ (2007) Combined kinetic and thermodynamic analysis of -helical membrane protein unfolding. Proc Natl Acad Sci USA 104: Faham S, Yang D, Bare E, Yohannan S, Whitelegge JP, Bowie JU (2004) Side-chain contributions to membrane protein stability and structure. J Mol Biol 335: Luecke H, Schobert B, Richter, H-T, Cartailler, J-P, Lanyi JK (1999) Structure of bacteriorhodopsin at 1.55 Å resolution. J Mol Biol 291: Riley ML, Wallace BA, Flitsch SL, Booth PJ (1997) Slow -helix formation during folding of a membrane protein. Biochemistry 36: Fersht AR, Sato S (2004) -value analysis and the nature of protein folding transition states. Proc Natl Acad Sci USA 101: Matouschek A, Fersht AR (1993) Application of physical organic chemistry to engineered mutants of proteins: Hammond postulate behaviour in the transition state of protein folding. Proc Natl Acad Sci USA 90: Matouschek A, Otzen DE, Itzhaki LS, Jackson SE, Fersht AR (1995) Movement of the position of the transition state in protein folding. Biochemistry 34: Renthal R (2006) An unfolding story of helical transmembrane proteins. Biochemistry 45: Compton ELR, Farmer NA, Lorch M, Mason JM, Moreton KM, Booth PJ (2006) Kinetics of an individual transmembrane helix during bacteriorhodopsin folding. J Mol Biol 357: Booth PJ, Curran AR (1999) Membrane protein folding. Curr Opin Struct Biol 9: Booth PJ, et al. (1997) Evidence that bilayer bending rigidity affects membrane protein folding. Biochemistry 36: Kim J-M, Booth PJ, Allen SJ, Khorana HG (2001) Structure and function in bacteriorhodopsin: The role of the interhelical loops in the folding and stability of bacteriorhodopsin. J Mol Biol 308: Fersht AR (1998) in Structure and Mechanism in Protein Sci (W. H. Freeman and Co., New York). 23. Jackson SE (1998) How do small single-domain proteins fold? Fold Des 3:R81 R Booth PJ, Curnow P (2006) Membrane proteins shape up: Understanding in vitro folding. Curr Opin Struct Biol 16: Hammond GS (1955) A correlation of reaction rates. J Am Chem Soc 77: Sapra KT, Balasubramanian GP, Labudde D, Bowie JU, Muller D (2008) Point mutations in membrane proteins reshape energy landscape and populate different unfolding pathways. J Mol Biol 376: HyunJoong Joh N, et al. (2008) Modest stabilization by most hydrogen-bonded sidechain interactions in membrane proteins. Nature 453: Sanchez IE, Kiefhaber T (2003) Hammond behaviour versus ground state effects in protein folding: Evidence for narrow free energy barriers and residual structure in unfolded states. J Mol Biol 327: Lu H, Marti T, Booth PJ (2001) Proline residues in transmembrane -helices affect the folding of bacteriorhodopsin. J Mol Biol 308: Hunt JF, et al. (1997) A biophysical study of integral membrane protein folding. Biochemistry 36: Vonck J (2000) Structure of the bacteriorhodopsin mutant F219L N intermediate revealed by electron crystallography. EMBO J 19: Ludlam G, Rothschild KJ (1997) Similarity of bacteriorhodopsin structural changes triggered by chromophore removal and light-driven proton transport. FEBS Lett 407: Subramaniam S, Henderson R (2000) Molecular mechanism of vectorial proton translocation by bacteriorhodopsin. Nature 406: Isenbarger TA, Krebs MP (1999) Role of helix-helix interactions in assembly of the bacteriorhodopsin lattice. Biochemistry 38: Cline SW, Doolittle FW (1987) Efficient transfection of the archaebacterium Halobacterium halobium. J Bacteriol 169: Oesterhelt D, Stoeckenius W (1974) Isolation of the cell membrane of Halobacterium halobium and its fractionation into red and purple membrane. Methods Enzymol 31: Whitelegge JP, Gundersen CB, Faull KY (1998) Electrospray-ionization mass spectroscopy of intact intrinsic membrane proteins. Protein Sci 7: DeLano WL (2002) The PyMOL molecular graphics system cgi doi pnas Curnow and Booth