Ultra-fast mixing and protein folding
While small globular proteins typically require milliseconds to seconds to complete the process of folding, important conformational events, such as a-helix and b-hairpin formation are known to occur on a much shorter time scale. However, the initial collapse of the polypeptide chain, which plays a central role in many theoretical models of protein folding, has until recently eluded direct observation. Many proteins studied by conventional kinetic methods exhibit significant changes in spectral parameters, such as circular dichroism or intrinsic fluorescence during the first few milliseconds of refolding (the so-called burst phase), but the significance of these observations is still a matter of controversy. In particular, it is unclear whether these early events represent formation of a specifically folded intermediate, or whether they simply reflect the response of the denatured polypeptide to the change in solvent conditions dominated by non-specific hydrophobic interactions.
flow methods. Recent progress in understanding the dynamics of protein
folding has been closely linked with the development of new experimental
approaches for initating and observing conformational changes on the microsecond
time scale, using optical triggers or improved rapid mixing methods (reviewed in
et al., 1998). Although the first continuous-flow measurement of a rapid reaction has
been reported in 1923, the method has enjoyed only limited subsequent use
because of the large amount of sample consumed. We have recently shown that the
efficiency of the method can be greatly improved by using an array of detectors
(CCD) to simultaneously monitor the signal at a large number of points along the
flow channel, as outlined in Fig. 1 (Shastry,
Luck & Roder, 1998). By combining this
sensitive detection method with an efficient capillary mixer, we were able to
achieve dead times as short as 45 ms.
By contrast, conventional stopped-flow measurements rarely achieve a dead time
less than 3 ms and suffer from various artifacts that are largely absent in our
approach. This instrument has opened a new window for exploring a wide range of
biological processes, such as enzyme-catalyzed reactions and fluorescence
changes associated with protein folding or ligand binding, on the microsecond
time scale. We have recently shown that the instrument can also be adapted for
absorbance measurements, using a transmission flow cell with a path length of
Figure 1: Continuous-flow capillary mixing apparatus. a Setup for fluorescence detection. b Expanded view of mixer and flow cell.
Cytochrome c. We first applied this method to study the kinetics of folding of horse cytochrome c on the microsecond time scale (Shastry & Roder, 1998). The greatly improved time resolution of the continuous flow instrument made it possible, for the first time, to monitor the heme-induced quenching of tryptophan fluorescence associated with the initial collapse of the chain down to tens of microseconds. In combination with conventional stopped-flow measurements, we were able to account for the total fluorescence change associated with cytochrome c refolding over a time window from 45 ms to about 60 s (Fig. 2). The kinetics of folding under various initial and final conditions is dominated by a major exponential decay in fluorescence (>60% of the total signal) with a time constant of 40-60 ms, which indicates formation of a compact state separated from unfolded conformations via a discrete free energy barrier. The rate of the initial process is independent of heme ligation state, confirming that it reflects an intrinsic conformational event. The shallow temperature dependence of the rate observed suggests that chain collapse is limited by an entropic bottleneck, as predicted by some theoretical models. The barrier-limited collapse of the polypeptide chain precedes the rate-limiting step in the formation of the native structure by at least two orders of magnitude in time, which is again consistent with a minimal two-stage folding mechanism.
Figure 2: The figure illustrates the time range accessible to the continuous-flow capillary mixer developed in our laboratory (top panel). The fluorescence signal extrapolates back to the unfolded baseline signal (1.0), indicating that these kinetics account for the entire fluorescence change of horse cytochrome c during folding. The lower panel combines data from this mixer with that obtained from a conventional stopped-flow apparatus with a ~2 ms deadtime (Bio-Logic SFM4). Notice that the bulk of the fluorescence change associated with folding occurs on the submillisecond timescale, which is now accessible using our continuous-flow mixer.
Protein G domain. Many small proteins can fold in a single concerted step without detectable intermediates. One notable exception is the B1 domain of streptococcal protein G, GB1 (Fig. 3). Despite its small size (57 residues) and simple architecture consisting of a four-strand b-sheet and a flanking a-helix, GB1 exhibits a pronounced burst phase in stopped-flow fluorescence measurements under strongly native conditions (Park et al., 1997). Using our continuous-flow capillary mixing device, we recently extend these measurements down to the 100 ms range, which allowed us to directly observe the previously unresolved fluorescence changes (Park et al., 1999). The refolding reaction, initiated by a rapid drop in denaturant concentration, was monitored by measuring the increase in the fluorescence emission associated with the burial of a tryptophan side chain within the hydrophobic core of GB1.
Figure 3: Left: Ribbon diagram of GB1. Right: Folding kinetics monitored by continuous- and stopped-flow fluorescence.
Fig. 3 shows the kinetic traces obtained in a series of continuous-flow experiments at different final denaturant concentrations, along with stopped-flow traces obtained under matching conditions. Under strongly stabilizing conditions, the kinetics is dominated by an exponential increase in fluorescence with a time constant of 600 ± 20 ms, followed by a minor phase in the millisecond range. With increasing denaturant concentration, the rate-limiting process on the stopped-flow time scale slows down and gains amplitude at the expense of the fast process, whose rate remains nearly constant. The combined kinetic traces quantitatively account for the total difference in fluorescence between the native and unfolded state at equilibrium, including the previously unresolved “burst phase” signal. Thus, our observations capture the complete time course of folding observable by fluorescence, which measures the transfer of Trp43 from an aqueous to an apolar environment and concomitant formation of the hydrophobic core. Kinetic simulations indicate that the denaturant dependence of the two rate constants and their relative amplitudes are fully consistent with a three-state folding mechanism, U <=> I <=> N, where I is an obligatory folding intermediate on a direct path between the unfolded and native states. The observation of an exponential time course during the initial stages of folding confirms that formation of I from U is a barrier-crossing event, and the fact that the fluorescence signal after the initial phase approaches the level of the native state at low denaturant concentration indicates that Trp43 is already sequestered in I within a largely solvent-shielded environment. It is interesting that this compact intermediate is formed at a ~10-fold slower rate in GB1 compared to cytochrome c. Given that cytochrome c is nearly twice as large as GB1, protein size is apparently not an important factor in determining the rate of early folding events. However, this surprising result may well be related to the fact that GB1 contains extensive b-sheet structure, including a pair of parallel strands involving opposite ends of the chain, which is expected to be a slower process than coalescence of a-helices in proteins such as cytochrome c.
applications. Following previous work on early stages of folding of
cytochrome c (2) and the B1 domain of protein G (3), we are now combining
continuous-flow measurements with mutagenesis in order to characterize the
intermediates and kinetic barriers encountered during folding of cytochrome c
and staphylococcal nuclease. Other collaborative projects are aimed at surveying
early folding events in a wide range of proteins with diverse structural
properties, including a series of
bacterial immunity proteins (Capaldi
et al., 2001), b-lactoglobulin
(Kuwata et al.,
2001), ubiquitin (Shastry, Latypov &
Roder), staphylococcal nuclease (Shastry, Cheng & Roder), cytochrome c551
(with C. Travaglini, M. Brunori), the structurally unrelated cytochrome b562
(with A. Fersht), de novo designed
proteins (with W.F. DeGrado), cellular retinoic acid binding protein (with L.
Gierasch), ribonuclease A (with H.A. Scheraga), mutants of
the streptococcal proteins G and L
(with D. Baker), and the human prion protein (with R. Glockshuber). In each
case, the experiments revealed major kinetic events on the submillisecond time
scale that, in combination with stopped-flow measurements, provided a
comprehensive kinetic characterization of folding events over the time range
from tens of microseconds to minutes. We also collaborate on continuous-flow
studies of rapid ligand binding reactions, including pre-steady-state kinetic
measurements on a flavoenzyme (with James Robertson).
Structure and stability of early folding intermediates.
Cytochrome c, a well-characterized small protein with a covalently attached heme group, has served as a model protein in much of our protein folding work. Using the previously developed pulse labeling method (1, 2, 3, 4, 5, 6), we have obtained a detailed structural and kinetic description of the folding pathway for oxidized cytochrome c. Amide probes on two alpha-helices near each end of the polypeptide chain become rapidly protected from deuterium-hydrogen exchange in a common folding phase, while other amide probes remain exposed much longer, indicating that the two helices form and associate at an early stage of folding before other parts of the structure are assembled (2). Additional evidence for the rapid association of N- and C-terminal helices was obtained by pulsed hydrogen exchange experiments under variable labeling conditions, which demonstrated that amide protons on either helix are not only protected simultaneously, but also exhibit the same degree of protection (4). These experiments also show that the folding reaction of cytochrome c involves both sequential intermediates and multiple parallel pathways.
Equilibrium and kinetic molten globules.
It has been suggested that early protein folding intermediates have some of the properties of a molten globule state, a compact denatured state found for many proteins under mildly denaturing conditions. Cytochrome c is an excellent test case for this hypothesis, since detailed structural information is available for both early kinetic folding intermediates and compact equilibrium forms. In collaboration with Michel Goldberg (Pasteur), we characterized a very early folding intermediate formed within the first few milliseconds of refolding by comparing a large number of conformational probes (7), including kinetic circular dichroism (CD) in the far-UV and near UV-region of the spectrum, tryptophan-heme fluorescence energy transfer, heme absorbance, and the protection of 42 individual amide protons measured by pulsed hydrogen exchange (2, 4).
Comparison with our previous hydrogen exchange results on a non-native compact equilibrium form of oxidized cytochrome c at acidic pH (8) revealed some interesting similarities, including high secondary structure content and a compact, but loosely packed core structure. However, there are also some striking differences. The compact acid-denatured form contains highly protected amide protons in all three main helices indicative of stable hydrogen-bonded structure in a helical subdomain, whereas the earliest detectable kinetic intermediate contains no well-protected amide protons. The protection pattern observed in a subsequent intermediate formed on the 10 ms time scale indicates stable helical structure in two regions near each end of the chain. However, in contrast to the equilibrium molten globule form, there is no evidence for stable hydrogen bonds in a third helix (residues 60-69).
Heme ligation and cytochrome c folding.
The role of the heme in its various states of oxidation and ligation was addressed by optically detected stopped-flow and NMR experiments under different refolding conditions, such as variable pH, denaturant concentration and the presence of extrinsic ligand molecules (9). In the oxidized form (FeIII), one of the axial heme ligands, His 18, remains bound to the heme iron under denaturing conditions (concentrated guanidine hydrochloride solution near neutral pH), but the second axial ligand, Met 80, is replaced by a histidine side chain (His 33) in a major population of unfolded molecules. Early in folding, this non-native heme ligand becomes trapped in a partially folded intermediate with interacting N- and C-terminal alpha-helices. In a subsequent folding event on the 100 millisecond time scale, this incorrect histidine ligand is replaced by the native methionine ligand, followed by rapid formation of the remaining parts of the structure. This ligand exchange reaction is not observed under conditions that prevent the trapping of an incorrect histidine ligand (e.g. refolding at lower pH or in the presence of extrinsic ligands).
A peptide model of a folding intermediate.
In a collaborative project with Jannette Carey (Princeton), we succeeded in designing a peptide analog for an early folding intermediate of cytochrome c, confirming previous evidence that this intermediate represents a structural subdomain with two mutually interacting alpha-helices. Since the heme group is essential for folding, we used limited tryptic digestion (with prior protection of lysine residues) to isolate an N-terminal fragment that includes residues 1-38 of horse cytochrome c along with the covalently bound heme group. A C-terminal peptide fragment corresponding to residues 87-104 was prepared by peptide synthesis. The far-UV CD and proton NMR spectra of the separate fragments are consistent with largely disordered conformations with only a small tendency towards helical secondary structure (10-20%). However, when the two peptides were combined, we observed a substantial change in the far-UV CD spectrum diagnostic for the formation of helical secondary structure. Together with more detailed NMR data, the results are consistent with the formation of a high-affinity 1:1 complex between the two helical peptides. No evidence for association was found in the absence of the covalently bound heme group, confirming that the heme group is essential for the interaction between the N- and C-terminal fragments.
Folding of reduced cytochrome c probed by ligand binding and photolysis.
The time resolution of conventional kinetic protein folding experiments based on stopped-flow and pulse labeling techniques is limited by the mixing dead-time. Thus, little is known about the process of folding on a sub-millisecond time scale, even though some of the most critical structural events, such as the formation of secondary structure and hydrophobic clusters, are expected to occur in microseconds or less.
In order to improve the time resolution of kinetic folding experiments, we are developing flash photolysis and photochemical methods to trigger the folding reaction of cytochrome c by light. In collaboration with Bill Eaton (NIH) we have shown above that the folding reaction of reduced cyt c can be initiated by photodissociation of a CO ligand bound to the heme in the unfolded state, which suggested the possibility of using nanosecond laser flashes for kinetic folding experiments with improved time resolution (15). Flash photolysis of the CO-iron bond results in a pentacoordinated heme. Subsequent ligation events were monitored by time-resolved absorption spectroscopy in the heme (Soret) region. Competition among several histidine and methionine side chains of the unfolded polypeptide chain for the vacant heme binding site gives rise to a series of spectral changes in the time range from 10-8 to 10-2 s. Kinetic modeling suggests that the earliest events represent transient binding of methionines 65 and 80, which are subsequently displaced by histidines 26 and 33. A small fraction of the molecules appear to form a partially folded intermediate with a native Met 80 ligand, but complete refolding is prevented by rebinding of CO on the millisecond time scale.
Thermodynamic and kinetic role of helix packing interactions in the folding of cytochrome c.
The pairing of two alpha-helices at opposite ends of the polypeptide chain is a highly conserved structural motif found throughout the family of c-type cytochromes. We have previously shown that this interaction is a critical early event in the folding process and is responsible for the formation of a prominent folding intermediate, INC (see above). In order to gain further insight into the structural and energetic basis of this helix packing interaction and its role in folding, we prepared a series of horse cytochrome c variants in which Leu 94, a central residue at the helix contact site, was replaced by other aliphatic side chains (in collaboration with Fred Sherman, Rochester). Three variants containing L94I, L94V and L94A mutations were expressed in yeast and purified. For their thermodynamic and kinetic characterization, we relied on the intramolecular quenching of the fluorescence of Trp 59 by the covalently bound heme group as a conformational probe.
The effect of amino acid changes on protein stability was determined from changes in the chemically induced unfolding transition. The Ile and Val substitutions resulted in only minor changes in unfolding free energy (±0.5 kcal/mol relative to a free energy of 9.3 kcal/mol for the wild type), indicating that conservative mutations can be accommodated at the helix interface with only minor structural perturbations. In the case of the L94V variant, this conclusion was confirmed by the NMR solution structure analysis mentioned above (17) the effect of mutations at position 94 on the kinetics of folding and unfolding in a systematic series of stopped-flow fluorescence measurements as a function of denaturant concentration and pH. In spite of their small equilibrium effects, the Ile and Val substitutions gave rise to substantial kinetic changes. As for the wild type, the folding reaction at low denaturant concentration is characterized by two major processes, a fast phase (~20 ms) that corresponds to the helix-helix association event (formation of INC), and a slower phase (~0.5 s) that reflects the formation of the native state following dissociation of a non-native histidine ligand (9). However, in the case of the L94I and L94V variants, the fast phase has a much lower relative amplitude. These observations, as well as the denaturant-dependence of the rates and amplitudes, are consistent with a substantially reduced population of the INC intermediate compared to the wild-type protein. This effect is amplified in the case of the destabilizing L94A variant, which exhibits slower folding kinetics with even lower population of INC. These results represent our most direct evidence so far for an early folding event involving association of the N- and C-terminal helices. Since conservative mutations that have little effect on the stability of the native state can lead to major destabilization of the INC intermediate, the mutual structural recognition of the two alpha-helices during the initial stages of folding appears to involve rather specific side chain packing interactions.
Comparison of the folding mechanisms for eukaryotic and prokaryotic cytochromes c.
Aside from a handful of functionally important residues, including a Cys-X-X-Cys-His motif near the N-terminus and a methionine toward the C-terminus involved in heme attachment and ligation, the prokaryotic cytochrome c2 from Rhodobacter capsulatus has little sequence homology with the mitochondrial cytochromes c. Nevertheless, they exhibit some striking structural similarities, including two alpha-helices near opposite ends of the polypeptide chain that are tightly packed against each other in a perpendicular arrangement. Since formation and association of the N- and C-terminal helices is a critical early event in the folding of horse cytochrome c, it is possible that constraints imposed by the mechanism of folding are responsible for the conservation of this structural motif. Another attractive feature of cytochrome c2 is that the axial heme ligands, His17 and Met96, are the only histidine and methionine residues in the protein, unlike horse cytochrome c, which has additional histidines and methionines that complicate the folding process by forming non-native heme ligands. It also features a single tryptophan residue in close proximity to the heme (analogous to horse cytochrome c) that serves as a convenient fluorescence probe for monitoring the folding transition.
The kinetics of folding and unfolding of cytochrome c 2 was measured as a function of denaturant concentration, both in the presence and absence of sodium sulfate, which is known to stabilize compact protein conformations (cf. ref. 2). The results are consistent with the following four-state mechanism:
where U, I, and M represent states with one axial ligand (His17) and a vacant 6th iron coordination site, and N represents the native state with two axial ligands (His17 and Met96). I is a compact intermediate with quenched tryptophan fluorescence, and M is an almost completely folded state lacking the native Met96-iron bond. I is well populated at low denaturant concentrations, where it gives rise to a major burst phase (fluorescence decrease during the mixing dead time), and the folding rate approaches a denaturant-independent value. As in the case of ubiquitin (3), we attribute this effect to the dissociation of the methionine ligand from the heme, which precedes the main structural unfolding transition.
Folding mechanism of ubiquitin
Ubiquitin is ideally suited for NMR-based protein folding studies, due to its small size (76 residues), high stability, absence of disulfide bonds, excellent solubility and predominance of hydrogen-bonded secondary structure (a five-strand beta-sheet and a long alpha-helix). Previous pulse labeling and 2D NMR studies resulted in a detailed description of the formation of hydrogen-bonded structure during refolding (5). However, the application of optical techniques for equilibrium and kinetic folding studies is difficult in the case of ubiquitin because it lacks suitable chromophores. We remedied this problem by engineering a tryptophan into the hydrophobic core of ubiquitin, replacing a largely buried phenylalanine at position 45 (10). Our recent NMR solution structure of this F45W variant described in the next section showed that the mutation is accompanied by only minor structural perturbations.
Unfolding of the mutant protein in 6 M guanidine hydrochloride results in a large increase in tryptophan fluorescence and a shift in the emission band from 336 nm to 352 nm, which is indicative of the exposure of a largely buried tryptophan ring. This fluorescence change enables us to follow the time course of folding and unfolding by fluorescence. Systematic stopped-flow measurements over a wide range of GuHCl concentrations show a linear dependence of folding and unfolding rates on denaturant concentration, as expected for a two-state folding mechanism. However, at GuHCl concentrations below 1 M, we found clear evidence for an early folding intermediate that accumulates during the first few milliseconds of refolding. The observed change in Trp 45 fluorescence in this early folding step is consistent with the formation of a compact intermediate with a partially developed hydrophobic core. Surprisingly, this intermediate was no longer observed at lower temperature (8 °C). This apparent cold-denaturation transition is consistent with a loosely folded structure stabilized primarily by hydrophobic interactions.
NMR structure refinement of F45W ubiquitin.
The solution structure of the tryptophan-containing ubiquitin variant was determined by using 2D NMR spectroscopy in conjunction with newly developed computational approaches (11). Previous spectroscopic and biochemical results (10) indicate that replacement of the largely buried phenyl ring with an indole ring introduces little structural perturbation and results in a fully functional ubiquitin variant with only slightly lower stability. Consequently, we were able to use the structure of wild-type ubiquitin as a starting point and to limit the refinement to a 9 Å sphere around residue 45, thus avoiding the far greater effort of a de novo solution structure determination of the entire protein. Low-energy protein conformations consistent with 81 distance restraints involving Trp 45 were calculated by restrained molecular dynamics methods, using a simulated annealing protocol to sample conformational space. In order to alleviate boundary distortions between the refined and fixed portion of the protein, stochastic boundaries were incorporated into the simulation. The main criteria for assessing the accuracy of the structures is based on the comparison of experimental NOE intensities with corresponding theoretical intensities calculated from trial structures. Comparison of the average solution structure of the mutant with the wild type crystal structure reveals small rearrangements of backbone and side chain atoms in the vicinity of the mutation. The tryptophan ring is packed among hydrophobic side chains in a largely solvent-inaccessible pocket, which is consistent with its fluorescence properties (10).
Probing the folding mechanism of ubiqitin by kinetic analysis of variants with amino acid changes in the core.
The role of partially ordered intermediates in protein folding is addressed by investigating the effect of amino acid changes at a central position in the hydrophobic core of ubiquitin (Val26) on the structure and stability of the native state, and the kinetics of folding and unfolding (16). For wild-type ubiquitin and the stable variants with Ile and Leu at position 26, an intermediate with native-like fluorescence for the previously engineered Trp45 is formed in less than 2 milliseconds under stabilizing conditions. Quantitative kinetic analysis, based on a three-state model with an obligatory folding intermediate, shows that a denaturant-independent process at low guanidine HCl concentrations corresponds to the rate-determining conversion of the intermediate into the native state. In terms of its response to mutations and solvent additives, the intermediate resembles the transition state for folding. The steep decrease in the overall rate of folding observed with increasing denaturant concentration approaching the unfolding transition is attributed to the decreasing population of the intermediate as its structure is destabilized by the denaturant.
Mutation of Val26 to Ala or Gly not only results in a major loss in stability of the native state (by 3.3-6 kcal/mol), but also destabilizes the intermediate (by 2.2-4.3 kcal/mol) indicating that this region of the hydrophobic core is already consolidated in the intermediate, but not as tightly packed as in the native structure. The kinetic behavior is that of an apparent two-state reaction, even in the absence of denaturant. However, for the Ala variant the intermediate reappears on addition of sodium sulfate, demonstrating that the mutation does not prevent formation of the intermediate, but simply destabilizes it. The kinetic behavior seen here for ubiquitin and its variants is representative for the majority of small globular proteins, supporting the notion that formation of partially folded states with a subset of native-like structural elements at early stages of folding is essential for rapid and efficient protein folding.
Folding studies of staphylococcal nuclease (SNase) by pulsed hydrogen exchange and stopped-flow methods.
We have recently begun to extend the structural and kinetic approaches described above to explore the folding mechanism for another well-known model protein, SNase. Pulsed hydrogen exchange studies on uniformly 15N labeled samples showed that most amide protons in a structural subdomain consisting of a beta-barrel were already partially protected during the dead time of the quenched-flow experiment, while most of the amide sites in the predominantly alpha-helical domain were protected at a later stage. Thus, a partially folded state containing stable hydrogen-bonded structure only in the beta-barrel domain accumulates within the first few milliseconds of folding. We are currently investigating the kinetic role of this intermediate by stopped-flow fluorescence measurements on a series of mutant forms of SNase.
In a previous collaborative study on the complex of cytochrome c with a monoclonal antibody ( 13 ), we showed that hydrogen exchange measurements can provide detailed information on protein-protein interactions. In this "epitope mapping" approach, hydrogen-deuterium exchange rates in the bound form are measured indirectly by exposing the complex to D2O, followed by dissociation of the ligand protein and 2D NMR analysis under slow exchange conditions. Amide protons that are protected against exchange in the complex thus define the parts of the protein involved in binding. This protocol makes it possible to investigate small or medium-size proteins bound to substrates that are too large for direct NMR analysis.
One limitation is that only amide groups with sufficiently slow exchange rates in the dissociated ligand can be used as probes. In order to extend the method to rapidly exchanging surface protons or complexes with peptide ligands, we developed a more efficient method for quenching the exchange reaction during and after ligand dissociation, using organic solvent additives (14). We found that in a mixture containing 95% dimethyl sulfoxide and 5% D2O under optimal conditions (pH 5, 10°C), amide groups in model peptides became deuterated with time constants of 30 to 150 h, which is about 100-fold longer than in D2O. This provides ample time to record a 2D NMR spectrum on a quenched hydrogen exchange sample without loss of amide proton labels. Addition of other cosolvents (toluene and cyclohexane) resulted in improved chemical shift dispersion. We used this solvent quenching method to measure NH exchange rates for melittin (a 26 residue peptide found in bee venom), both in its free monomeric state in water at low pH, as well as in a high-affinity complex with calmodulin. Whereas exchange rates in the free peptide are characteristic for a disordered conformation, the pattern of protected amide protons in the bound form is consistent with a helical conformation of melittin. The most highly protected amide protons are found for hydrophobic residues along one side of the helix, especially in the C-terminal half of the peptide.
Return to Roder homepage
last updated 3/07 by H. Roder.