Measurement of protein unfolding/refolding kinetics and structural characterization of hidden intermediates by NMR relaxation dispersion

Abstract

Detailed understanding of protein function and malfunction hinges on the ability to characterize transiently populated states and the transitions between them. Here, we use ¹⁵N, , and ¹³CO NMR R₂ relaxation dispersion to investigate spontaneous unfolding and refolding events of native apomyoglobin. Above pH 5.0, dispersion is dominated by processes involving fluctuations of the F-helix region, which is invisible in NMR spectra. Measurements of R₂ dispersion for residues contacted by the F-helix region in the native (N) structure reveal a transient state formed by local unfolding of helix F and undocking from the protein core. A similar state was detected at pH 4.75–4.95 and determined to be an on-pathway intermediate (I1) in a linear three-state unfolding scheme (N⇆I1⇆MG) leading to a transiently populated molten globule (MG) state. The slowest steps in unfolding and refolding are N → I1 (36 s^-1) and MG → I1 (26 s^-1), respectively. Differences in chemical shift between N and I1 are very small, except in regions adjacent to helix F, showing that their core structures are similar. Chemical shift changes between the N and MG states, obtained from R₂ dispersion, reveal that the transient MG state is structurally similar to the equilibrium MG observed previously at high temperature and low pH. Analysis of MG state chemical shifts shows the location of residual helical structure in the transient intermediate and identifies regions that unfold or rearrange into nonnative structure during the N → MG transition. The experiments also identify regions of energetic frustration that “crack” during unfolding and impede the refolding process.

Within the cell, unfolding and refolding of proteins occurs constantly, and spontaneous unfolding and misfolding processes play a central role in the formation of amyloid fibrils (1). In addition, fluctuations between native protein structure and partially or fully unfolded states have functional significance for binding, allosteric regulation, translocation across membranes, protein trafficking, secretion, and degradation (2). Despite the importance of spontaneous unfolding for protein function and cellular proteostasis, little is known about the transient, partially unfolded states that are formed. Detailed structural characterization of such states is difficult because of their inherently low populations and the conformational heterogeneity present when both native and partially unfolded states simultaneously exist in solution.

Carr–Purcell–Meiboom–Gill (CPMG)-based R₂ relaxation dispersion experiments are unique in their ability to measure the kinetics of microsecond–millisecond exchange processes involving transient states with populations as sparse as 1%, and, in addition, allow determination of the associated NMR chemical shift changes (Δω) (3–5). In these experiments, the effective transverse relaxation rate is deconvoluted into the contribution from the exchange process, R_ex, which varies with the pulsing frequency in the CPMG refocusing element, and the intrinsic relaxation rate . Here, we apply a series of amide and carbonyl R₂ dispersion experiments to investigate transient unfolding and refolding events within the native (N) state free-energy landscape of apomyoglobin (apoMb).

ApoMb is an ideal model system for studies of protein folding/unfolding equilibria. The folded state is marginally stable at neutral pH, adopting a similar tertiary structure to the holoprotein except for disorder in the EF loop, F helix, FG loop, and N-terminal region of the G helix (residues 82–104), which appear to fluctuate between a folded native-like conformational substate and an ensemble of locally unfolded states (6, 7). ApoMb forms an equilibrium molten globule (MG) state at pH 4.1, amenable to direct investigation by NMR, which contains a significant number of native contacts and an overall compactness close to that of the N state (8–10). Kinetic refolding experiments show rapid formation of a burst-phase intermediate that resembles the equilibrium MG and contains native-like topology in the most structured regions (10–12). Folding to the N state from the MG is the overall rate-limiting event; although the structural changes that occur across this transition are well established, it is still unclear why folding is slow and which interactions are frustrated and responsible for the kinetic trap.

In the present work, we characterize the mechanism of transient unfolding of apoMb, identify the structural changes that occur, and measure the kinetics of the unfolding and refolding transitions. Chemical shift changes, obtained from analysis of R₂ dispersion experiments, provide insights into the structures of the two intermediates that are formed and identify regions of energetic frustration that impede protein folding.

Results

pH Dependence of N-State Structure and Dynamics:

¹⁵N and ¹³CO chemical shifts were used to probe potential changes in the structure of the N state over the pH range of interest. Most amide ¹⁵N resonances shift by less than 0.2 ppm between pH 5.90 and 4.95 (Fig. 1A), showing that the native apoMb structure is retained over this pH range. These small shifts are in marked contrast to those associated with apoMb unfolding, which moves ¹⁵N resonances by an average of 3 ppm. Resonances with ¹⁵N chemical shift changes larger than 0.2 ppm largely belong to histidines, most of which titrate over this pH range (13), and residues that contact histidines. In addition, small ¹⁵N chemical shift changes are observed for residues in the N terminus of the G helix and in sites that contact the F helix in the holoMb structure, suggesting a small pH-dependent shift in the F-helix conformational ensemble. These effects can be monitored only indirectly, because exchange broadening prevents direct observation of amide resonances of the disordered F helix in native apoMb (6). ¹³CO chemical shift changes between pH 5.9 and 4.75 (Fig. S1A) corroborate these results.

Fig. 1.

Sensitivity of N-state ¹⁵N chemical shifts and R_ex relaxation rates to changes in pH. (A) Magnitude of ¹⁵N chemical shift changes between pH 5.9 and 4.95, color-coded to highlight His (red); residues contacting His (orange); and residues contacting the disordered F helix, FG loop, and N-terminal region of helix G (blue); other residues are colored green. The solid black rectangles at the top of the figure depict helices in apoMb based on a TALOS+ (30) analysis of , ¹⁵N, ¹³CO, , and chemical shifts (6), whereas the boundaries of the F, G, and H helices in holoMb are shown as open rectangles. (B) ¹⁵N R_ex (at a static magnetic field of 11.7 T) of deuterated apoMb at 35 °C; pH 5.9 (blue), pH 4.9 (red). R_ex was estimated from the difference in R_2eff at the lowest and highest 1/τ_cp values. The green squares on the horizontal axis indicate contact sites with helix F in holoMb, and the circles identify His residues. Yellow squares indicate the N-terminal region of helix G.

Single-quantum (SQ) ¹⁵N and ¹³CO R₂ dispersion experiments were used to detect the formation of transiently populated states within the apoMb N-state ensemble in the pH range 5.2–5.9. Localized groups of residues display R₂ dispersion with greater than 2 s^-1 (Fig. 1B). These residues are located at or near sites where the F helix contacts helices C, E, and H in the native structure, suggesting that dispersion arises from millisecond timescale fluctuations in structure and/or packing of the F-helix region. Large R_ex values are also observed for residues at the N terminus of the G helix, immediately following the FG turn. The pattern of ¹³CO R_ex contributions largely matches that of ¹⁵N (Fig. S1B), with greater than average values in regions associated with helix F.

As the pH is lowered to 4.95, an additional exchange process becomes evident (Fig. 1B and Fig. S1C), associated with a large increase in R_ex for residues throughout the structure. The dramatic pH dependence of R_ex, together with the relatively small chemical shift perturbations between pH 5.9 and 4.95, is attributed to a shift in the N-state conformational ensemble such that there is an increased population of partially unfolded states at lower pH. At pH 5.9, R₂ dispersion arises predominantly from localized fluctuations of the disordered F helix, whereas global conformational fluctuations dominate at pH values below 5.

Transient Unfolding at pH Greater than 5.0.

¹⁵N and R₂ dispersion curves for perdeuterated apoMb acquired at two static magnetic fields (11.7 and 18.8 T) at pH 5.5 fitted well to a global two-site exchange model (Fig. S2) with a reduced χ² of 1.5. The small R_ex contributions in regions outside of the sites that are contacted by residues 82–104 (Fig. 1B) most likely come from the exchange process that occurs below pH 5, although long-range effects of the F-helix fluctuations cannot be entirely ruled out. The kinetics of the fluctuations in the F-helix region were determined from a clustered fit of the pH 5.5 ¹H/¹⁵N R₂ dispersion curves for residues within the contact sites of the F helix and FG loop in the native Mb structure. The fits reveal an exchange process at a rate of 533 ± 7 s^-1 between the N state and a state with 7% population (Table 1).

Table 1.

Kinetic parameters of apoMb transient unfolding and refolding measured by R₂ relaxation dispersion

Probes	pH	kN-I1 , s-1	kI1-N , s-1	kI1-MG, s-1	kMG-I1, s-1	PN	PI1	PMG
*	5.50	38 ± 1	495 ± 7	—	—	0.93 ± 0.02	0.07 ± 0.00	—
†	4.95	31.7 ± 0.5	240 ± 12	53 ± 1	26 ± 4	0.71 ± 0.03	0.09 ± 0.00	0.20 ± 0.03
3CO/15N‡	4.75	78 ± 18	650 ± 140	113 ± 16	17 ± 1	0.52 ± 0.01	0.06 ± 0.01	0.42 ± 0.01

^*Parameters from a two-state fit to amide datasets ¹H-SQ and ¹⁵N-SQ using the cluster of residues in contact with the F helix and the first helical turn of G (listed in the caption of Fig. S2). Error values were obtained using Monte Carlo sampling.

^†Parameters from a three-state fit (Scheme 1) to the data shown in Fig. 2 (¹H-SQ, ¹⁵N-SQ, ¹H¹⁵N-DQ, ¹H¹⁵N-ZQ). Error values were obtained using Monte Carlo sampling.

^‡Parameters from a three-state global fit (Scheme 1) to data shown in Fig. 3. Error values were derived from an extensive grid search of the kinetic parameter space (SI Text).

Equilibrium Unfolding Process at pH Less than 5.

Analysis of SQ ¹⁵N R₂ dispersion experiments at pH 4.75 (SI Text) showed clearly that more than two states are involved in the transient unfolding process at pH < 5. To robustly fit more complex exchange models, dispersion data were acquired on ²H/¹⁵N-labeled apoMb at pH 4.95 using ¹H SQ (4), ¹⁵N SQ (3, 4), ¹H/¹⁵N double-quantum (DQ) and ¹H/¹⁵N zero-quantum (ZQ) coherences (14), while incorporating the respective chemical shift differences Δω_H, Δω_N, (Δω_N + Δω_H), and (Δω_N - Δω_H) into a global analysis (15, 16) (Fig. 2). Because of the destabilizing effect of aliphatic deuteration (17), ¹⁵N R_ex values for perdeuterated samples at pH 4.95 were of similar amplitude to those of protonated ¹⁵N-labeled apoMb at pH 4.75.

Fig. 2.

Chemical shift differences obtained from R₂ dispersion show that apoMb populates the MG state below pH 5. (A) Representative R₂ dispersion curves for apoMb at pH 4.95, 35 °C at a static magnetic field of 18.8 T: ¹⁵N-SQ (red), ¹H-SQ (purple), ¹H¹⁵N-ZQ (green), and ¹H¹⁵N-DQ (blue), showing fits to a three-state exchange model. Data acquired at 11.7 T were included in the global fit, but are omitted from the plots for clarity. Correlation of amide ¹⁵N (B) and (C) equilibrium chemical shift differences (ppm) between the N (pH 4.95) and MG (pH 4.1) states, |Δδ_N,MG|, and the chemical shift differences |Δω_N,MG| obtained from a global fit of the dispersion curves to the linear model N⇆I1⇆MG. A line of slope 1 is shown as a visual guide. Some scatter in the correlation plots for amide ¹⁵N and resonances is to be expected because of unavoidable differences in sample conditions, required to obtain spectra of the equilibrium MG (pH 4.1, 50 °C, 10% ethanol) (8).

Dispersion curves for 89 residues were fitted using a global three-state exchange model, which gave significantly better fits than two-state exchange models according to reduced χ² values and the Akaike information criterion (SI Text). All possible three-state exchange models were evaluated, including triangular connectivity between states and linear schemes with N in the center or on the periphery. A linear exchange model (A⇆B⇆C) yielded the best global fit and physically meaningful Δω values (SI Text). The kinetics and populations are summarized in Table 1. A linear correlation is observed between the and ¹⁵N chemical shift differences (Δω_A,C) derived by fitting the dispersion curves and the equilibrium chemical shift differences (Δδ) between native apoMb (N) and the pH 4.1 MG, thus identifying states A and C as N and MG, respectively (Figs. 2 B and C). The previously unknown I1 state is therefore an on-pathway intermediate between the N and MG states (Scheme 1).

Scheme 1.

The ¹⁵N chemical shift differences between the N and I1 states (Δω_N,I1) are very small, < 0.4 ppm for most residues; residues with larger Δω values are predominantly associated with F-helix contact sites (Figs. S3 A and B and S4), suggesting that the transition to I1 involves fluctuations in the F helix or dissociation of the F helix from the protein core. In contrast, much larger ¹⁵N chemical shift differences occur in the transition between the N and MG states, with Δω(¹⁵N) > 1 ppm for the majority of resonances that exhibit dispersion (Fig. 2). Based on the small amplitude of Δω_N,I1 for both ¹⁵N and (Fig. S3A), I1 is clearly similar in overall structure to N and is formed through a local, partial unfolding process involving the NMR-invisible residues of the F helix.

A linear correlation (slope = 1.0) is observed between Δω_N,I1 and Δω values for residues in F-helix contact sites derived from the two-state clustered fit of dispersion curves at pH 5.5 (Fig. S3). This confirms that I1, formed transiently at pH < 5, corresponds to the state associated with F-helix fluctuations at pH 5.5. The unfolding rates at pH 5.5 and 4.95 are similar (Table 1), whereas the refolding rate at pH 5.5 is twice that at pH 4.95, resulting in a smaller population of the transiently unfolded F helix at pH 5.5 (7%) than for I1 at pH 4.95 (9% population).

Probing Backbone Conformation at pH 4.75 Using ¹³CO R₂ Dispersion.

Transient unfolding of secondary structure at pH 4.75 was monitored using ¹³CO R₂ dispersion (18) (Fig. 3). To provide an internal control to account for small and unavoidable differences in sample conditions, ¹³CO and ¹⁵N data were acquired on the same ¹⁵N/¹³C/¹H-labeled sample. The dispersion curves yield ¹⁵N Δω values that match those determined with solely ¹⁵N-labeled samples. Dispersion curves were analyzed for 117 residues located throughout the structure, except for the F-helix region and C terminus of the H helix where amide resonances are broadened beyond detection by conformational exchange. The ¹⁵N and ¹³CO dispersion data were fit simultaneously to a three-state model using a single set of kinetic parameters (Table 1 and Fig. 3). The ability to simultaneously fit both datasets provides evidence that the two probes report on the same three-state exchange process and supports a model of cooperative partial unfolding of tertiary and secondary structure. The exchange rates between N and I1 and the population of the MG state derived from fits of the ¹³CO/¹⁵N dataset are slightly higher than those extracted from the ¹H/¹⁵N data (Table 1), probably due to differences in sample conditions (SI Text and Fig. S1C).

Fig. 3.

Probing backbone conformational changes using ¹³CO R₂ dispersion. (A) Representative ¹³CO (black) and ¹⁵N (blue) R₂ dispersion curves collected on a single sample of ¹³C, ¹⁵N-apoMb, pH 4.75, at a static magnetic field of 18.8 T. Data acquired at a static field strength of 14.1 T were included in the global fit but are omitted from the figure for clarity. The solid curves represent three-state global fits to the exchange model N⇆I1⇆MG. (B) Correlation between Δω_N,MG (ppm) obtained from a global three-state fit to ¹³CO and ¹⁵N R₂ dispersion data at pH 4.75, and |Δδ_N,MG| determined from 3D NMR data under equilibrium conditions (pH 4.75 and 4.1 for N and MG, respectively). The line shows a linear fit to the data (R² = 0.88, slope 0.93). A description of parameter constraints and uncertainty analysis is in the SI Text.

Discussion

Exchange Rates.

Fast kinetics experiments have shown that apoMb folds by a sequential pathway (U↔I_a↔I_b↔N) involving two intermediates separated by an energy barrier (19–21). The first intermediate (I_a) is helical in the A, G, and H regions, whereas I_b contains additional structure in the B, C, and E regions (22). The k_MG,I1 folding rate from three-state fits of the dispersion curves (26 s^-1 at 35 °C; Table 1) agrees remarkably well with the folding rate for the I_b to N transition (20 s^-1 at 26 °C) determined from fast kinetics (21). The rate of apoMb unfolding to the MG state at 4.5 °C ranges from 13 s^-1 at pH 4.2 to ∼50 s^-1 at pH 3.7 (23). Given the differences in temperature and pH, these rates appear consistent with that determined here for the I1 → MG transition (53 s^-1 at pH 4.95, 35 °C).

Unfolding to the MG Occurs Through the I1 State.

The dispersion data at pH 4.75 fit best to a linear three-state exchange model, in which I1 is an on-pathway intermediate between N and MG; both unfolding and refolding transitions proceed by way of I1. Native apoMb contains a compact globular core, with secondary and tertiary structure similar to that of the holoprotein (7). However, the F helix and neighboring regions are dynamically disordered and fluctuate between a compact ordered state, in which the empty heme-binding pocket is partially closed (24), and a partially unfolded, solvent-exposed state (6). In the present work, we attribute the process that dominates R₂ dispersion at pH 5.5, and also contributes to dispersion at lower pH values, to the intrinsic fluctuations of the F helix to form state I1. The correlation between Δω, determined from a two-site exchange model at pH 5.5, and Δω_N,I1, obtained by fitting a three-site model at pH 4.75 (Fig. S3B), strongly supports the interpretation of N⇆I1 exchange as a local unfolding process. The dispersion measurements report indirectly on fluctuations of the F helix, the amide resonances of which are broadened beyond detection in the N state, through exchange contributions at F-helix contact sites. Transient unfolding of helix F does not significantly perturb the structure of the apoMb core; chemical shift changes between N and I1 (Figs. S3 and S5) are very small for the majority of residues showing that the secondary and tertiary structure of the apoMb core is preserved. Amides with the largest values of Δω_N,I1(¹⁵N) are located in F-helix contact sites (Figs. S3 and S4). Changes in ¹⁵N shifts in the transition to I1 are also observed for several residues in helix E; these likely reflect undocking of the F-helix region, which probably occupies partially the empty heme pocket of the native apoMb, where it would make direct contact with helix E (24). Some residues that directly flank helix F have Δω_N,I1(¹³CO) between 0.4–0.9 ppm, suggesting partial fraying at the C terminus of E and the N terminus of helix G in I1 (Figs. S4 and S5). Clusters of residues with Δω_N,I1(¹³CO) > 0.2 ppm are also found at F-helix contact sites in the CD region and in the H helix.

Disruption of the buried His24–His119 hydrogen bond is the primary trigger for pH-dependent unfolding of native apoMb (13). His24 has an abnormally low pK_a (< 4) in the N state and a normal pK_a in the transition state and MG (25). The His24 chemical shifts are unchanged in I1 (Δω_N,I1 ≈ 0 ppm), indicating that His24 remains in a native-like environment (with low pK_a) in the I1 intermediate. Protonation of His24, which drives unfolding (13, 25) and causes large changes in His24 amide chemical shifts, must therefore occur before the transition state between I1 and MG is reached (Fig. 4), In support, Δω_I1,MG is large for both ¹⁵N and ¹H (1.0 and 0.5 ppm, respectively).

Fig. 4.

Schematic free energy diagram for three-state transient unfolding (N⇆I1⇆MG) of apoMb at pH 4.95 (solid) and two-state unfolding (N⇆I1) at pH 5.5 (dashed). Barrier heights were calculated from the rates in Table 1 and Table S1 using transition-state theory and a standard assumption for the prefactor for protein folding (35). The rate-limiting barriers are marked with asterisks. The free energies of the I1 and MG states are 1.2 and 0.9 kcal/mol, respectively, above the N state.

Recent studies of apoMb folding and unfolding kinetics (26, 27) suggest that the initial event during unfolding and the final event during refolding involve solvation and desolvation, respectively. Indeed, it has been suggested that desolvation of backbone amides may contribute to the energy barrier in the final rate-limiting step of the folding pathway (27). Although our data do not directly probe solvation effects, it is likely that the N → I1 transition, driven by the undocking of the F helix, is accompanied by substantial hydration of the empty heme-binding pocket.

Although the secondary and tertiary structure of the compact globular core of N and I1 is very similar, I1 is thermodynamically less stable than either the native or MG states (Fig. 4). The difference in free energy between I1 and MG is small (0.3 kcal/mol) yet striking, because it suggests structural frustration in the formation of the less stable, highly native-like I1 state from the MG.

Structure of the Transient MG.

The Δω values for ¹⁵N, , and ¹³CO (Tables S2–S4) yield information about the structural changes that occur between the N and MG states during the process of folding and unfolding, at equilibrium under a single set of experimental conditions (temperature, pH, and buffer). Whereas ¹³CO chemical shifts are determined primarily by backbone conformation, and hence Δω(¹³CO) reports directly on changes in backbone dihedral angles and secondary structure, amide ¹⁵N chemical shifts are in addition highly sensitive to hydrogen bonding interactions and local environment (28, 29). Large values of Δω_N,MG(¹⁵N) for residues in the AB loop (residues 17–21), in the CD loop (residues 41–47), at the C terminus of helix D and beginning of helix E (residues 56–61), and in the GH loop (residues 117–122) indicate substantial conformational rearrangements or unfolding/folding transitions between the N and MG states (Fig. 5A). Changes in structure or packing are also evident in the helix B/helix G interface, in the middle of helix E (residues 67–70) where it contacts the AB loop in the native fold, and in helix H (residues 134, 137, 138) where it packs against the EF turn.

Fig. 5.

Sites of local unfolding and structural rearrangement in the transient MG. (A) Mapping of ¹⁵N Δω_N,MG chemical shift differences onto the structure of holoMb. The protein backbone is represented as a tube. Residues with smaller than average Δω_N,MG are colored red, whereas those with above-average shifts are colored blue and vary in thickness and color saturation according to the magnitude of Δω_N,MG. (B) Secondary structure and RCI-S² parameter (32) for the transient MG state and the N state. The secondary structure of the MG state was predicted by TALOS+ (30) from , ¹⁵N, and ¹³CO chemical shifts (derived from Δω_N,MG as described in SI Text). A more complete dataset (, ¹⁵N, ¹³CO, , shifts) was available for the N state; however, the helical boundaries and RCI-S² for N were not significantly altered when predicted from a limited set of , ¹⁵N, ¹³CO shifts. The rectangles depict the location of helical structure in each state; the thickness of each rectangle is proportional to the population of helix. The hatched lines indicate the small population of transient helical structure in the C- and E-helix regions of MG. (C) Changes in secondary structure accompanying the N⇆MG transition, mapped to the structure of holoMb. Residues predicted to be helical by TALOS+ are red. The population of helix in the MG ensemble is indicated by the tube radius, with a larger radius indicating higher population. Flexible regions with RCI S² < 0.7 are blue, and coil regions with S² > 0.7 are green. A white backbone trace indicates regions for which no predictions are available. The His24 and His119 side chains are shown as pink spheres in A and C. The figure was prepared using the program MolMol (36) from the coordinates of holoMb (Protein Data Bank ID code 1MBC).

To obtain direct insights into changes in secondary structure, backbone dihedral angles for the transient MG intermediate were determined from ¹⁵N, , and ¹³CO chemical shifts (derived from Δω_N,MG values) using TALOS+ (30). Helical structure is predicted with high confidence in several regions (Fig. 5 B and C); estimates of helix population were obtained from the average deviations of the ¹³CO shifts from sequence-corrected random coil values (31). Highly populated helices are indicated for residues 4–21 (80% population), 27–33 (65%), 105–115 (80%), and 127–138 (85%), corresponding respectively to the A helix, C terminus of B, and central regions of the G and H helices of native apoMb (Fig. 5B). Whereas the B helix is shortened in the transient MG, the A helix extends to form nonnative helix between residues 19–21. Nonnative helical structure (population 55%) is also predicted for residues 53–62, extending from helix D into the first turn of helix E. Finally, a small helical propensity (≤ 30%) is found for residues 39–43 (C helix in native apoMb) and 69–73, in the middle of helix E. The transient MG is similar in structure to that formed at pH 4.1 (see SI Text for details).

The random coil chemical shift index (RCI) order parameter (S²) (32) shows that the A, B, G, and H helical regions in the transient MG are as well ordered as in the N state (Fig. 5B), suggesting that they pack together to form a compact hydrophobic core. The other helical regions, including the weakly populated C helix, the D–E region, and the central part of helix E, have lower S² than in N, consistent with the disorder implied by their fractional helical populations. It is likely that these regions are highly dynamic and dock only transiently against the ABGH core of the MG. The RCI order parameters in the N-terminal region of helix B, in the CD region, throughout most of the E helix, at the C terminus of helix G, and in the GH turn also indicate greatly increased flexibility in the transient MG and loss of structure relative to the N state. The S² decreases slightly for Val21, suggesting dynamic fraying at the end of the nonnative A helix.

The observed changes in secondary structure and local regions of unfolding indicated by the RCI S² parameters suggest a plausible model for the unfolding process. Residues surrounding His24 at the N terminus of helix B and His 119 at the C terminus of helix G unfold during the I1 to MG transition and are disordered in the MG state. These changes reflect loss of the stabilizing His24–His119 hydrogen bond and provide graphic evidence that “cracking” (33) of the structure in this region and protonation of His24 acts as a trigger (13, 25) for apoMb unfolding. Loss of structure at the N terminus of helix B disrupts critical contacts with the N-terminal part of helix E, which also unfolds and becomes dynamically disordered. Unfolding of helix E, in turn, disrupts the packing interface for the CD loop, leading to unfolding of this entire region and formation of the nonnative D–E helix. This nonnative structure forms spontaneously (20% population) in acid-unfolded apoMb at pH 2.1 (31); the increased population in MG probably reflects stabilizing tertiary contacts in the compact MG state. Apart from a small propensity for helix between residues 69–72, most likely stabilized by transient packing of this strongly hydrophobic region onto the MG core, and the nonnative structure at the helix D/helix E junction, the entire E helix is dynamically disordered.

The MG state observed in the present experiments appears to correspond to the kinetic I_b intermediate (21, 22), containing stable helical structure in the B region and transient structure in helix C and part of helix E that would not be formed in I_a. The R₂ dispersion data provide additional structural insights into I_b, which complement those from hydrogen exchange pulse labeling (10, 22). Firstly, dispersion analysis provides direct and unambiguous evidence that the high degree of exchange protection of amide protons in the A, B, G, and H regions in the kinetic burst-phase intermediate is due to formation of stable helical structure. Second, residues that exhibit partial proton occupancies in hydrogen exchange experiments (10) are seen to be located in disordered or transiently helical regions in MG. Finally, information is available from dispersion for residues whose amides are poorly protected and are therefore not observed in hydrogen exchange experiments (10, 22). For example, dispersion reveals unfolding at the N terminus of helix B and C terminus of helix G and an extension of helical structure at the N terminus of helix H, where no hydrogen exchange probes are available. Conversely, hydrogen exchange experiments suggest formation of stable helical structure between residues 139 and 144, for which dispersion data are missing because of the broadness of the N-state resonances. Differences are observed at the C terminus of the E helix; this region is predicted to behave as dynamic coil in the transient MG state, yet displays mild H–D exchange protection in I_b. The amide closing rates involved in the formation of the kinetic intermediate indicate that the E helix is stabilized initially at its C terminus.

The R₂ dispersion methods described here should be generally applicable to study equilibrium unfolding/folding transitions in other proteins under mildly destabilizing conditions. This approach offers several advantages, including measurement of unfolding and refolding kinetics under identical conditions in a single experiment, and characterization of any intermediates formed. R₂ dispersion methods could prove particularly powerful for characterization of the earliest protein unfolding events that lead to protein aggregation, amyloid formation, and disease.

Materials and Methods

Expression and Purification.

Protein expression, purification, and NMR sample preparation are described in SI Text.

NMR Data Acquisition.

The assignment of amide and ¹³CO resonances is described in SI Text. R₂ dispersion data were acquired simultaneously on 500- and 800-, or 600- and 800-MHz Bruker spectrometers equipped with cryogenic probes, using constant time relaxation compensated CPMG pulse sequences (3). The sample temperatures were calibrated to 35 °C with perdeuterated methanol (34). SQ ¹⁵N and ¹³CO dispersion experiments (4, 18) were performed on ¹³C, ¹⁵N-labeled samples. Separate experiments were conducted on ²H, ¹⁵N-labeled apoMb using ¹H SQ (4), ¹⁵N SQ (3, 4), ¹H/¹⁵N DQ, and ¹H/¹⁵N ZQ coherences (14).

Data Fitting.

Dispersion data were fit using the in-house program GLOVE (5). The fitting parameters included the microscopic kinetic rates connecting the ground and transient states, the populations of all states, the differences in chemical shifts between each of the states (Δω), and the relaxation rate in the absence of conformational exchange, . The error in was defined as 10% of the amplitude of the R₂ dispersion curve (R_ex) because this adequately reflects the level of random noise in . A detailed description of the global optimization and model selection is given in SI Text.