HX–NMR experiments

Examination of the TROSY (transverse relaxation optimized spectroscopy) two-dimensional (2D) ¹⁵N–¹H correlation spectrum (Fig. S2a) revealed that 222 out of 240 cross-peaks were sufficiently well resolved at 40 °C to enable accurate measurements of their intensities for this study. Exchange of the amide hydrogen atoms for deuterium was monitored by recording the TROSY 2D ¹⁵N–¹H correlation spectrum as a function of time after dissolving lyophilized protein into ²H₂O buffer at pH 6.8 (meter reading) at 40 °C (Fig. S2b and c). Note that the relationship between pH and pD, pD= pH (meter reading)+0.4,⁴⁵ means that the actual acidity in solution was equivalent to pH 7.2. The precision of the HX measurements was enhanced by leaving the sample in the NMR probe for the entire time course of the experiment, 10 days.

The 90 Class I amide hydrogen atoms (NHs) exchanged within the time required to dissolve the sample, shim the magnet and collect a useful 2D spectrum, 30 min. The 79 Class II NHs underwent N95% exchange over 7 days, and almost all of the 53 Class III NHs underwent less than 65% exchange over the full 10 day time course of the experiment (see Materials and Methods). The exchange rate constants for the Class II NHs were obtained by fitting the fractional occupancy of hydrogen as a function of time to a single exponential function. Traces and fits at 0 M Gdn-HCl for a representative set of NHs are shown in Fig. 3a and b for two different time regimes, 30–1000 min and 30–10,000 min, spanned by these experiments. From these measurements, protection factors (PFs) were calculated as the ratio of the exchange rate constant of the amino acid in an unstructured peptide, k_int, to the observed exchange rate constant for a particular amino acid in the protein, k_obs, PF=k_int/k_obs. The k_int was taken as the exchange rate constant for amino acids in a random-coil conformation calculated using the program Sphere†. PFs range from 4.12×10³ for the more rapidly exchanging Class II NHs to 1.73×10⁷ for slowly exchanging NHs in the absence of Gdn-HCl (Table S2). Although there was no evidence in the NMR spectra for the timedependent changes observed in the native baseline region of the CD data (Fig. 1a and b), the fits of the hydrogen occupancies were only considered over the first 7 days of exchange. Less than 10% of the native ellipticity was lost in this time frame.

Open in a separate window

Fig. 3

Normalized intensity of ¹⁵N–¹H cross-peak versus log time plots for a representative set of NHs that exchange over the time frame. (a) 30–1000 min. (b) 30–10,000 min at 40 °C and pH 7.2 in deuterated 10 mM KPi, 50 mM KCl and 1 mM K₂EDTA. The presence of 50 mM KCl, included to mimic the buffer conditions used for the resonance assignments,⁴⁶ had no detectable effect on the rate constants for the major unfolding and refolding phases corresponding to the [N⇆N*]⇆I₂ reaction (data not shown) (Scheme 1) and, therefore, the lifetime of the native state of HisF.

EX1 versus EX2 limit: characterizing the exchange regime

To interpret the PFs for HisF, one must determine whether exchange is limited by the opening rate constant for breaking the H-bond, the EX1 limit, or by the fraction of the open state required for exchange, the EX2 limit.⁴⁷ This determination can be made by comparing the k_obs values at two different pH values.^47–50 In the EX1 limit, a linear dependence in the log–log plot is expected to intersect the y-axis at the same log unit as the x-axis and have a gradient of one.In the EX2 limit, one expects a linear dependence intersecting the y-axis at one log unit higher than the x-axis and with a gradient of one. The shift along the y-axis reflects the known acceleration of the exchange reaction with increasing pH above ~2.3.⁵¹ As shown in Fig. 4a, the raw data do not follow the expectation of either the EX1 or the EX2 regime. To probe the possible involvement of a change in stability of HisF with pH, we examined the effect of pH on the ratelimiting unfolding and refolding reactions. Although the refolding reaction is unperturbed, the unfolding reaction slows by ~5 fold when the pH is increased from 6.2 to 7.2 (Fig. S3). Relative to the free energy of the I₂ state, the N state decreases, that is, it is stabilized, by 0.90 kcal mol⁻¹ when the pH is increased from 6.2 to 7.2. The magnitude of the stabilization is precisely that would render the exchange process for the Class II NHs to obey the EX2 limit (Fig. 4b). Although HX cannot occur through the I₂ state over the time course of the experiment, N*is a potential conduit for HX in the native basin, and indeed, the slowing of the N→N* reaction at higher pH (Fig. S3) is consistent with this possibility.

Open in a separate window

Fig. 4

(a) The log–log plot of the observed exchange rate constants measured at pH 7.2 and pH 6.2 at 40 °C for amide hydrogens of HisF, before correction for the effect of pH on stability. (b) The log–log plot of the observed exchange rate constants, after correction for the effect of pH on stability (Fig. S3). The continuous line represents the EX2 limit, and the dotted line represents EX1 limit.

In the EX2 limit, the PFs reflect the free-energy difference between the closed and open forms of the H-bond forming/breaking reactions, ΔG°_HX⁵² [ΔG°_HX = −RTln(k_obs/k_int)]. The ΔG°_HX values for the Class II NHs in the absence of denaturant, range from 5 to 10.5 kcal mol⁻¹, and along with values for k_obs and k_int, are shown in Table S2. Although the exchange behavior of the Class I and Class III NHs could not be measured, if both classes exchange by an EX2 mechanism, their ∆G°_HX values would be less than 5 kcal mol⁻¹ and greater than 10.5 kcal mol⁻¹, respectively. However, as will be discussed below (see Discussion, Mechanism of HX), the accuracy of these estimates for ∆G°_HX is suspect and, therefore, should be considered as “apparent” values based upon the assumption of HX through unfolded segments of a protein.

HX protection patterns in HisF

The protection patterns for the Class I, Class II and Class III NHs are portrayed on ribbon diagrams and a 2D map for HisF in Figs. 5 and ​and6,6, respectively. The rapidly exchanging Class I NHs preferentially appear in the active-site loops at the C-termini of the β-strands, which for β4, β5 and β8 are either short α-helices (β4 and β8) or a β-hairpin (β5). Class I NHs also tend to be found at the N-termini of the a-helices, contiguous with the active-site loops. For example, residues 57–65 form a particularly long stretch of Class I NHs in α2. Class II NHs are generally found in the C-terminal segments of the α-helices, at the C-termini of five of the eight β-strands and at the N-termini of β1, β3 and β8. With the exception of the C-terminus of HisF, they also can be found in the short C-terminal loops and turns following the α-helices. These loops and turns have previously been implicated in stabilizing the TIM barrel.^27,28 Non-exchanging Class III NHs are predominantly found in the β-strands, with two or more members of this class in all eight β-strands. With the exception of α1 and α2, at least two Class III NHs are also found in the other six α-helices. Thus, the majority of the NHs in the β-barrel and segments of six of the α-helices in the external shell are highly resistant to HX in the native manifold of conformers of HisF.

Open in a separate window

Fig. 5

A three-dimensional representation of the location of the three classes of protection for HisF at pH 7.2 and 40 °C. (a) Class I: red, ΔG°_HX<5 kcal mol⁻¹; (b) Class II: yellow, ΔG°_HX≈5–7 kcal mol⁻¹; green, ΔG°_HX≈7–9 kcal mol⁻¹; blue, ΔG°_HX≈9–10.5 kcal mol⁻¹; (c) Class III: purple, ΔG°_HX > 10.5 kcal mol⁻¹. The figures were generated using PyMOL.⁵³

Open in a separate window

Fig. 6

(a) 2D representation of protection against exchange in HisF at pH 7.2 and 40 °C. Class I: red, ΔG°_HX<5 kcal mol⁻¹; Class II: yellow, ΔG°_HX≈5–7 kcal mol⁻¹; green, ΔG°_HX≈7–9 kcal mol⁻¹; blue, ΔG°_HX≈9–10.5 kcal mol⁻¹; Class III: purple, ΔG°_HX>10.5 kcal mol⁻¹. Open circles indicate either the absence of the NMR assignment or the inability to obtain accurate fits due to spectral overlap and () indicates a proline residue. (b) The secondary structure of HisF mapped on its sequence.

The Class III level of protection for the NHs of E46 at the N-terminus of β2 and E167 at the N-terminus of β6 is of special note. The H-bonds to the carbonyl oxygens in I6 and V125 are at or beyond the expected length of H-bonds, 3.64 and 3.37 Å, respectively. The N-termini of the β-strands are twisted in a fashion that enables C=O of I6 to form a second H-bond with the NH of D45 and the C=O of V125 to form a second H-bond with the NH of G166 [insets (a) and (b) in Fig. 7]. Although the length of the E46 and E167 H-bonds and the necessity to share a common acceptor with other NHs implies weaker H-bonds and, possibly, weaker protection, the absence of exchange over a period of at least 10 days demonstrates that the local environments in the native state are remarkably resistant to the conformational dynamics required to expose the NHs to deuteroxide.

Open in a separate window

Fig. 7

2D representation of the hydrogen-bonding pattern of the β-sheet in HisF. The arrows (→) depict the H-bonds, pointing from the backbone amide (NH) to the main-chain carbonyl oxygen (C=O) with the exception of I6 in β1, F77 in β3 and I198 in β7 where their respective main-chain amides are hydrogen bonded to the side-chain carboxyl of D45, D98 and D219, respectively (broken lines). The (▲) represents amino acids that are not a part of β-strands. Small circles represent main-chain positions whose side chains point outside the barrel, and large circles represent main-chain positions whose side chains pointing inside the barrel. The NHs are color-coded as Class I: red, ΔG°_HX<5 kcal mol⁻¹; Class II: yellow, ΔG°_HX≈5–7 kcal mol⁻¹; green, ΔG°_HX≈7–9 kcal mol⁻¹; blue, ΔG°_HX≈9–10.5 kcal mol⁻¹; Class III: purple, ΔG°_HX>10.5 kcal mol⁻¹. The four layers of side chains within the β-strands are indicated by broken lines. (a) and (b) in the insets illustrate the pair of H-bond donors to the I6 and V125 carbonyl acceptors (gray arrows in the main figure).

HX properties of HisF

The protection of main-chain NHs against exchange in the native basin of HisF at pH 7.2 and 40 °C falls into three classes:

1. Class I NHs exchange with solvent within the time required to initiate the exchange reaction by dilution of lyophilized protein into deuterated buffer and collect a usable 2D NMR spectrum, ~30 min. In general, the Class I NHs are found in loops connecting the C-termini of β-strands with their subsequent α-helices, that is, the loops that form the active site of HisF and the N-termini of α2, α3, α6 and α8 (Figs. 5a and ​and6).6). NMR relaxation experiments revealed dynamic behavior in this region of α2 and attributed it to the allosteric communication between the active sites of the HisF and HisH subunits in ImGPS.⁵⁹ Because the Class I NHs exchange far more rapidly than the rate of unfolding of HisF from the native state, they are most likely exchanging by an EX2 mechanism and have apparent ∆G°_HX values less than 5 kcal mol⁻¹.

2. Class II NHs exchange over the time period from ~30 min to 10 days via an EX2 mechanism. The Class II NHs are primarily found at the C-termini of the a-helices that dock on the core of the β-barrel, in the short turns linking α-helices to the subsequent β-strands and, strikingly, at the N-termini of only β1, β3 and β8 (Figs. 5b and ​and6).6). As noted above, the N-termini of β-strands are responsible for stabilizing the β-barrel, and indeed, the corresponding residues in the remaining five β-strands all fall into the Class III category. The apparent ∆G°_HX values range from 5 to 10.5 kcal mol⁻¹ for the Class II NHs (Table S2). Exchange occurs at 62 distributed sites within the native manifold without a significant exposure of buried surface area. The pH-dependent stability correction that rationalizes the EX2 behavior (Fig. 4) implies that there is another state within the native manifold, likely the N* state in Scheme 1, that serves as a conduit for the exchange of the NHs for deuterium.

3. Class III NHs do not exchange significantly over 10 days at pH 7.2 and 40 °C. Class III NHs most often appear in the eight β-strands but are also found in six of the eight canonical helices, the short loops/turns preceding β2 and β6 and in the β-hairpin following β5 (Figs. 5c and ​and6).6). With the assumption of an EX2 mechanism, the apparent ∆G°_HX values for Class III NHs are greater than 10.5 kcal mol⁻¹. If a subset of the Class III NHs exchange through partially unfolded states on the native side of the barrier, they must do so through species whose free energies are at least 10.5 kcal mol⁻¹ above the native state.

Insights into the molecular underpinnings of the HisF TIM barrel

A closer examination of the protection patterns in the β-strands, α-helices, loops and turns in HisF (Fig. 6) hints at the essential structural features that define its native thermodynamic state. We take these structural features to be delineated by H-bonds that either do not break or are not accessible to deuteroxide over the 10-day time course of the HX-NMR experiment. Nearly 90% (47 out of 53) of the Class III NHs reside in the canonical β-strands and the segments of the hydrophobic faces of six of the a-helices that dock on the β-barrel. Most notably, the eight canonical β-strands contain 57% of the non-exchanging NHs and 68% of the β-strand NHs fall into the Class III category (Fig. 7).

The interior of the HisF barrel is well described by four layers of side chains, as originally recognized by Nagano et al.⁶⁰ Each layer consists of four side chains from either the odd-numbered or the evennumbered β-strands, alternating as the layers proceed from the N- to the C-termini of β-strands (Fig. 7). Only the second and third layers each contain eight Class III NHs from the eight canonical β-strands, I7 and A8 in β1, L47 and V48 in β2, T78 and V79 in β3, V100 and S101 in β4, V126 and V127 in β5, I168 and L169 in β6, I199 and A200 in β7 and A221 and L222 in β8, that do not exchange over the period of 10 days. Examination of the structure further shows that one or two Class III side chains in six of the α-helices, L92 and I93 in α3, I116 in α4, V157 in α5, I187 and V190 in α6, A213 and F214 in α7 and L237 in α8, are closely associated with the external side chains in layers 2 and 3 of the β-barrel (Fig. S5). This almost entirely aliphatic collection of side chains linking the β-strands (all of the eight side chains pointing out of the barrel are ILVA) and α-helices (eight of nine side chains pointing toward the barrel are ILVA) may well represent one of the essential features defining the ensemble of structures comprising the native thermodynamic state of the HisF barrel. Six of the eight side chains in layers 2 and 3 inside the β-barrel are also aliphatic. The very strong bias toward aliphatic side chains for these Class III NHs may be related to the preferential partitioning of side-chain analogs into the vapor phase versus water compared to their aromatic or polar counterparts (see below).

Because the free-energy difference between the native state and the I₂ intermediate (Scheme 1) is 12.02 kcal mol⁻¹, the intervening TSE must exceed that value. Access to the PFs between 10.5 and at least 12.02 kcal mol⁻¹ on the native side of the barrier, that is, those most crucial to the integrity of the native thermodynamic state, was, unfortunately, precluded by aggregation of the protein on prolonged exposure to 40 °C and 2.0 M Gdn-HCl where such states might be revealed. Note that the discrepancy between the free-energy difference between the N and U states estimated from the equilibrium unfolding analysis, 10.5 kcal mol⁻¹, and that from the kinetic analysis of scheme 1, 20.8 kcal mol⁻¹, reflects the presence of the intermediate states with native-like CD properties.

The BASiC hypothesis and HisF

The BASiC hypothesis proposes that large clusters of branched aliphatic side chains play crucial roles in stabilizing partially folded states in TIM barrel proteins.³⁶ A survey of 55 TIM barrel proteins posited that stabilizing clusters contain at least 10 ILV side chains each of which buries at least 10 Ų in an ILV self-contact and a total of at least 500 Ų.⁶¹ Application of the BASiC hypothesis to the αTS²⁸ and sIGPS^34,62 TIM barrels found a very good correlation between HX protection in their partially folded states and large ILV clusters observed in their native conformations. The enhanced protection is ascribed to the preferential resistance of such clusters to the penetration of the water and/or hydroxide ion required for exchange of underlying NHs and is consistent with the hydrophobicity scales of Wolfenden et al.⁶³ and Kyte and Doolittle.⁶⁴

HisF has five ILV clusters, ranging from 2 to 27 side chains in size (Fig. 8a). The side chains in a 22-residue cluster (Cluster 1) are associated with NHs in the α₇(βα)₈(βα)_1–2, and the side chains in a 27-residue cluster (Cluster 2) are associated with NHs in the α₂(βα)_3–6β₇ region. A 6-residue cluster inside the barrel links β1, β2, β5, β6, β7 and β8; a 3-residue cluster links the loops after β1 and β2, and the 2-residue cluster supports the loop after α8. The NHs in Clusters 1 (Fig. 8b) and 2 (Fig. 8c) are dominated by Class II and Class III levels of protection against HX; however, the side chains in Cluster 2 are more efficiently packed with each other and a majority of the NHs display Class III levels of protection. The side chains in Cluster 1 primarily link the external loops and the C-terminus with the α-helices and β-strands in the α₈(βα)_1–2 segment, and those in Cluster 2 primarily fuse the α-helical shell to the β-barrel in the α₃(βα)_4–6β₇ segment (Fig. 8d). Consistent with this partitioning of the structure into two regions of protection against HX is the observation that two Class II NHs are found in the first core layer of the β-barrel, F77 in β3 and A220 in β8 (Fig. 7); the remainder are Class III. Although the N-terminus of the β-barrel generally stabilizes the fold, the evident dynamics at the N-termini of this pair of β-strands highlights the stronger protection for the intervening (βα)_1–2 and (βα)_4–7 regions of HisF.

Open in a separate window

Fig. 8

HisF ILV clusters represented based on their residue contact and protection pattern. (a) HisF five ILV clusters: Cluster 1 (gray), Cluster 2 (black), Cluster 3 (green), Cluster 4 (purple) and Cluster 5 (blue). Clusters of ILV side chains in the crystal structure of HisF, Protein Data Bank entry 1THF,²² were identified using in-house software⁶¹ (http://quipu.umassmed.edu/ccss/CCSS.cgi). This software identifies networks of ILV side chains each of which bury at least 10 Ų and together bury more than 500 Ų surface areas using the CSU software.⁶⁵ (b) In the α7–α2 regions, 22-residue Cluster 1 NHs are color-coded based on their ΔG°_HX pattern: yellow, ΔG°_HX≈5–7 kcal mol⁻¹; green, ∆G°_HX≈7–9 kcal mol⁻¹; blue, ∆G°_HX≈9–10.5 kcal mol⁻¹ and purple, ΔG°_HX>10.5 kcal mol⁻¹. (c) In the α2–β7 regions, 27-residue Cluster 2 NHs are color-coded based on their ∆G°_HX pattern. (d) Contact map of the two large ILV clusters illustrating the mutual contacts in the clusters spanning the α2–β7 and the α7–β2 segments of HisF. The figures were generated using PyMOL.⁵³

Three orthogonal lines of evidence support the conclusion that the α₃(βα)_4–6β₇ region is a stability core for HisF. (1) The Rosetta design software package⁶⁶ was employed to predict the most stable 2-fold symmetric HisF TIM barrel,⁶⁷ motivated by the 4+4 gene duplication studies of Richter et al.³⁷ and an earlier conjecture of 2-fold symmetry of many TIM barrels.⁶⁸ Exploring the full sequence of HisF, the cut points for the (βα)₄ modules were moved along the sequence, the modules duplicated, fused and docked and the energies were minimized. The most stable symmetric barrels invariably contained a pair of covalently connected (βα)_4–7 segments that agrees closely with the strongly protected region spanned by the large ILV cluster (Fig. 8a and c). (2) A protein engineering effort by Carstensen et al. found substantially enhanced stability for the Sym2 construct, β5/α5/β6/α6/β7/α7/β4/α4/β5/α5/β6/α6/β7/ α7/β4/α4/, that contains two (βα)_4–7 elements, one overlapping the N- and C-termini.⁴¹ The X-ray structure of the engineered construct revealed two large and symmetrical ILV clusters, each of which spans the (βα)_4–6β₇ elements (Fig. S6). (3) NMR and computational studies of the HisF domain in the heterodimeric ImGPS complex have found that binding of the substrate N′-[(5′-phosphoribulosyl) formimino]-5-aminoimidazole-4-carboxamide ribonucleotide perturbs the dynamics of the α₂(βα)_3–6β₇ and α₇(βα)₈(βα)_1–2 sides of the barrel differently, suggesting a role for ILV clusters in enabling the allosteric mechanism of this enzyme.⁶⁹

We acknowledge that the BASiC hypothesis provides an estimate of the patterns of strong protection against HX in HisF. The protection patterns illustrated in Figs. 5 and ​and66 are not nearly as distinct as the large ILV clusters illustrated in Fig. 8. The hypothesis simply provides a chemical explanation for why large ILV clusters might preferentially (but not absolutely) protect against HX. In HisF, a 6 Å shell around the two large ILV clusters captures 90% of the strongly protected (Class II and Class III) amide hydrogens. Many of these NHs are associated with non-ILV side chains, showing that proximity to the clusters can bias their HX properties. We assume that the bias, in part, reflects the involvement of their aliphatic methylenes, linking polar or charged terminal atoms to the main chain, in the ILV clusters. It is also the case that about half of the amide hydrogens within this 6 Å shell are not strongly protected. While we do not understand the reasons for the overprediction of strong protection, we attribute this behavior to the structural fluctuations that are not apparent in a crystal structure. It is evident that the BASiC hypothesis provides a simple but approximate prediction of strong protection against HX in TIM barrel proteins. Refinement of the BASiC hypothesis will require further studies of the dynamics of the protein, for example, using long-timescale molecular dynamics simulations with explicit solvent.

All of these results validate the assumption that the HX protection pattern in the native state of HisF corresponds to cores of thermodynamic stability and support the application of the BASiC hypothesis to the native conformation of the HisF TIM barrel protein.

Mechanism of HX

The mechanism by which amide hydrogens exchange with solvent has been a subject of discussion and debate since the pioneering work of Berger and Linderstrøm-Lang,⁷⁰ Linderstrøm-Lang and Schellman⁷¹ and Hvidt and Nielsen.⁷² The local unfolding model proposes that exchange takes place in segments whose exchange properties resemble those of unfolded chains.^71,73 By contrast, the broad category of solvent penetration models^74–77 supposes that water and/or hydroxide diffuses into the body of the nominally folded protein and catalyzes exchange. In both models, the H-bond is assumed to be broken in the exchange-competent state, and if H-bond reformation is rapid compared to the pseudo-first-order exchange reaction, the exchange process falls into the EX2 limit. Englander and Kallenbach, in a comprehensive analysis of both models, argued persuasively for the local unfolding model.⁷⁸ They supposed that, among other issues, diffusion of water and especially anionic hydroxide into the low dielectric interior of a protein is very unlikely. Subsequently, Englander et al. acknowledged the possibility that the depth of the H-bond from the surface of a protein and/or the packing density might also play a role in limiting exchange in the native conformation.³⁹ Indeed, the pattern of protection in HisF shows that amides buried deep in the interior of the protein invariably have high PFs, whereas the surface amides range from completely unprotected to very strongly protected. Further, the most densely packed hydrophobic region (data not shown) corresponds to the α₃(βα)_4–6β₇ segment that closely corresponds to the most strongly protected region. In those situations, access to hydroxide would be expected to involve the coordinated disruption of structure and/or multiple packing interactions, for example, fraying, in the dynamic native manifold. Further consideration of this issue⁴⁰ led to the conclusion that the HX properties of an individual NH will reflect multiple aspects of its environment and are unlikely to yield reliable generalizations.

If fraying-like mechanisms are operative for HisF, how can one reconcile the small implied changes in structure with ΔG°_HX values (at least 10 kcal mol⁻¹) approaching the free-energy difference between the N state and the I₂ state, 12.02 kcal mol⁻¹, across the major barrier? For comparison, the difference in buried surface area between the rate-limiting TSE and the native state can be calculated from the kinetic unfolding m-value, 1.7 kcal mol⁻¹ M⁻¹,⁷⁹ and is estimated to be 3200 Ų. As noted previously,⁸⁰ the values of PFs and ΔG°_HX extracted from the NMR data would be inappropriately high if the peptide models from which k_int values are derived do not accurately represent the exchange behavior of the frayed segments. Although the details of the putative fraying reactions are undoubtedly complex, the constraints on the local sequences required for the heterogeneous protection patterns observed in the α-helices and β-strands for HisF almost assuredly mean that the segments are not behaving as structureless peptides. As a consequence, the k_int values would be smaller than the values in the database employed to calculate the PFs and ΔG°_HX values, and the end result would be a substantial over-estimation of both parameters. Thus, the ΔG°_HX values reported in this study should be considered as apparent estimates of the freeenergy difference between the closed and open states of the exchangeable H-bonds in the native state of HisF. This conjecture would also hold for other proteins for which the ko_bs are independent of denaturant concentration up to the unfolding transition zone.^55,58,81,82

The possibility that the denaturant independence of k_obs is insensitive to a substantial exposure of buried surface elsewhere in HisF, as modeled by Wooll et al., is very unlikely.⁸³ The k_obs for the 62 Class II NHs, distributed in every element of secondary structure and representing almost 30% of the NHs in HisF, do not reveal a significant denaturant dependence indicative of a large-scale conformational change. Also, the heterogeneous distribution of multiple classes of protection in each of the eight canonical helices is not consistent with this proposal. The inappropriate application of the simple HX model is more likely to provide an explanation for the apparently high values for ΔG_HX without a corresponding sensitivity to the denaturant concentration.

Protein expression and purification

HisF was expressed from the pET11c plasmid containing the HisF gene (a gift from Dr. Reinhard Sterner) in Escherichia coli BL21 (DE 3) cells and was purified according to a published protocol.⁸⁶ An overnight culture of freshly transformed cells in LB medium supplemented with 0.1 mg/ml ampicillin was used to inoculate a 10 l culture stock. The cells were harvested, lysed by sonication and centrifuged, and the supernatant was heat shockedat 75 °C for 15 min to precipitate host proteins. The HisF in the supernatant was dialyzed against 10 mM potassium phosphate (KPi) containing 2 mMK₂EDTA and 1 mM DTT buffer and purified by column chromatography with a DEAE-Sepharose fast-flow column followed by a Sephacryl S-200 column. The purity of the HisF was confirmed by SDS-PAGE and by measuring the molecular mass with liquid chromatography–electrospray ionization mass spectrometry.

For NMR studies, the transformed E. coli BL21 (DE 3) cells were grown on minimal media supplemented with ¹⁵N ammonium chloride or ¹⁵N ammonium chloride and ¹³C glucose.⁴⁶ Cells were grown at 37 °C until an optical density at 600 nm reached 0.9, at which point the temperature was reduced to 30 °C, and the cells were allowed to grow for an additional 14 h. The purification followed the above procedures, and the purity of the HisF was confirmed by SDS-PAGE and by mass spectrometry.

Equilibrium measurements

Kinetic measurements

Native-state hydrogen–deuterium exchange

Purified HisF was dialyzed overnight into a buffer containing 10 mM KPi (pH 7.2), 50 mM KCl and 1 mM K₂EDTA. The sample was concentrated for the HX–NMR study with an Amicon Ultra-15 centrifugal filter unit with a 10 kDa membrane and then lyophilized. Exchange was initiated by dissolving the lyophilized protein in ²H₂O buffer, which had been prepared at the required pH and buffer conditions (all the components in the buffers were pre-deuterated by repeated cycles of dissolution and lyophilization). All pH values reported are corrected meter readings. Upon addition of ²H₂O buffer, the sample was immediately transferred to a 5 mm NMR tube (Wilmad LabGlass, Vineland, NJ) and placed in the spectrometer at 40 °C; the HisF concentration was ~0.4 mM. The time between the initiation of exchange, the transfer to the NMR tube, placement in the spectrometer, tuning and shimming and the beginning of data collection averaged 15 min. TROSY 2D ¹⁵N–¹H correlation spectra were recorded over a period of hours to days, and the sample remained in the spectrometer for the entire course of the exchange reactions. All NMR experiments were recorded on a Varian 600-MHz spectrometer, and the spectra were processed in NMRPipe⁸⁹ and analyzed with Sparky.⁹⁰ The temperature was calibrated using a sample of 100% methanol. Three sets of independent experiments were collected to determine the experimental error.

The backbone resonance assignment of HisF is available from the BioMagResBank‡ under accession number BMRB-15741. Due to its spectral complexity and the different experimental conditions used in our experiments, buffer and temperature, we could not unambiguously assign all the backbone amide resonances based on the published assignment on HisF. For this reason, we collected three-dimensional sensitivity-enhanced gradientselected [¹H, ¹⁵N]-TROSY triple resonance experiments [HNCA, HN(CO)CA, HNCACB, HN(COCA)CB] at 14.1 T and at 40 °C using a uniformly labeled ¹³C/¹⁵N sample of HisF. Using these experiments, we were able to assign the backbone frequencies of 222 of the 240 residues of HisF.

Analysis of hydrogen–deuterium exchange data

Exchange rate constants for main-chain NHs were obtained by fitting the decay of ¹⁵N–¹H TROSY spectrum cross-peak intensities as a function of exchange time to a single exponential, I=I₀exp(−k_obst), with the initial intensity I₀ and the observed exchange rate k_obs as free variables in the fit. The exchange time was defined as the period of time from the dissolution of lyophilized HisF in ²H₂O buffer to the end of each TROSY 2D experiment. The uncertainties for the k_obs values were taken as standard errors of the fits. Under the demonstrated EX2 exchange mechanism for the Class II NHs, ΔG°_HX was obtained from the equation: ΔG°_HX=-RTln(k_obs/k_int), where k_int is the intrinsic exchange rate calculated for amide protons in unstructured peptides. The k_int values were obtained using the program Sphere from model data.⁹¹ Residues that underwent greater than 70% exchange before the first time spectrum was collected were considered to have PF<4.12×10³ and were classified as Class I. Residues that underwent less than 35% exchange at the end of 10 days were considered to have PF>1.73×10⁷ and were classified as Class III. Further, residues that exhibited an incomplete exponential decay of proton occupancy and had a remaining amplitude >30% after 10 days were also classified as Class III.

We thank Dr. Reinhard Sterner for generously providing the pET11c expression plasmid. Thanks also to Drs. Osman Bilsel, Jill Zitzewitz, Ramakrishna Vadrevu and Can Kayatekin and to Mr. Paul Nobrega for helpful discussions. This work was supported by National Institutes of Health grants GM 23303 to C.R.M. and GM 98763 to F.M.