Hydrogen exchange in peptides and proteins using NMR spectroscopy

Hydrogen exchange in peptides and proteins using NMR spectroscopy

Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170 www.elsevier.com/locate/pnmrs Hydrogen exchange in peptides and proteins using...

399KB Sizes 19 Downloads 434 Views

Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

www.elsevier.com/locate/pnmrs

Hydrogen exchange in peptides and proteins using NMR spectroscopy Christopher E. Dempsey* Department of Biochemistry and Centre for Molecular Recognition, University of Bristol, Bristol BS8 1TD, UK Received 30 March 2001

Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Basic considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Hydrogen exchange and hydrogen bonding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Amide hydrogen exchange chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Amide hydrogen exchange mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Exchange protection factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5. Exchange protection factors and acid-catalysed amide exchange . . . . . . . . . . . . . . . . . . . . . . . . 2.6. Relayed imidic acid mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. NMR methods for measuring hydrogen exchange rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Time-resolved exchange measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Magnetization transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Pulsed ®eld gradient diffusion measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. Exchange line broadening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5. Exchange trapping methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Timescales and amplitudes for protein ¯uctuations underlying hydrogen exchange . . . . . . . . . . . . . . . 4.1. Timescales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1. Approaches to EX1 kinetics from saturation transfer/magnetization transfer/PFG diffusion measurements? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Amplitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1. Exchange under conditions that destabilise the native state . . . . . . . . . . . . . . . . . . . . . . 5. Comparison of amide exchange data with molecular dynamics simulations and 15N relaxation measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. Membrane peptides and proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

136 136 137 138 140 141 142 144 144 144 145 151 152 153 153 153 157 157 158 162 163

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 Appendix A: De®nition of rate constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Keywords: Hydrogen exchange; NMR spectroscopy; Peptides and proteins; Protein ¯uctuations; Dynamics simulation; Membrane proteins

* Tel.: 144-117-928-7427; fax: 144-117-928-8274. E-mail address: [email protected] (C.E. Dempsey). 0079-6565/01/$ - see front matter q 2001 Elsevier Science B.V. All rights reserved. PII: S 0079-656 5(01)00032-2

136

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

1. Introduction Analysis of the kinetics of hydrogen exchange in biological macromolecules is a powerful method for studying both their structural and dynamic properties. This arises from our understanding of the role of hydrogen-bonded structure and dynamics in modulating hydrogen exchange in these molecules. It is now apparent that the major factor limiting hydrogen exchange in proteins and nucleic acids is the participation of the exchangeable hydrogens within hydrogen bonds. In addition, exchange from a hydrogen bond almost certainly requires transient `opening' ¯uctuations that free the hydrogen for interaction with a solvent exchange catalyst. The analysis of hydrogen exchange has been pursued since the early studies of Linderstrom-Lang [1] but has exploded in scope with the ability to determine the exchange properties of every nitrogen-bound hydrogen in a small protein or nucleic acid using the exceptional resolving power of NMR spectroscopy. It is worth remarking on the fundamental relationship between amide hydrogen exchange (the main focus of this review) and biological macromolecular structure and dynamics that makes the technique particularly useful. At one level, the structures of biological macromolecules are de®ned by their hydrogen bonding properties. The identi®cation of hydrogen-bonded amides, and determination of their exchange properties can yield structural information (e.g. as NH¼OyC distance constraints in NMR structural analysis, and in the characterization of hydrogen-bonded secondary structure in protein folding intermediates), insight into a major class of backbone ¯uctuations that disrupt hydrogen bonds, facilitating hydrogen exchange with solvent, and thermodynamic characterization (as free energies de®ning the stability of hydrogen-bonded secondary structure and its response to environmental conditions such as pH, temperature, denaturant concentration and functional interactions). As methods evolve to improve the detection, resolution and assignment of amide signals in NMR experiments, the scope of hydrogen exchange analysis increases. Recent years have seen the applications of amide hydrogen exchange trapping and NMR spectroscopy to the analysis of hydrogen bond formation

within intermediates on protein folding pathways, and to the determination of the stability of hydrogen-bonded secondary structure in membranereconstituted polypeptides. A main aim of this review is to cover the methodological aspects of hydrogen exchange measurements using NMR spectroscopy. No attempt is made at a comprehensive account of the results of numerous experiments that characterize protein folding pathways and assess global and local folding free energies in proteins, since these continue to be reviewed elsewhere [2±7]. However, separate discussions of the timescales and amplitudes of amide hydrogen exchange-limiting backbone ¯uctuations are included, since these remain areas where information is limited, and where NMR approaches will continue to make important contributions. Recent applications of amide hydrogen exchange analysis to polypeptides and proteins in phospholipid bilayer membranes is also covered. There are expectations that molecular dynamics simulations might provide insight into the nature of some of the backbone ¯uctuations underlying amide hydrogen exchange, and recent comparisons of experimental exchange data from NMR measurements with the results of MD simulations are brie¯y described. The applications of hydrogen exchange analysis to hydrogen bonding and backbone ¯uctuations in nucleic acids are described elsewhere [8±10]. 2. Basic considerations The experimental parameter in hydrogen exchange measurements is a pseudo ®rst order exchange rate constant. Section 3 describes the numerous methods for measuring amide-selective hydrogen exchange rate constants using NMR spectroscopy. Interpretation of the exchange rate constant in terms of hydrogen-bonded structure and dynamics requires careful consideration of the chemical and kinetic factors relevant to the exchange chemistry and the ¯uctuations underlying exchange, respectively. The following sections describe the essential features required for analysis of amide hydrogen exchange data. The review by Englander and Kallenbach [8] continues to be an excellent account of the basic features of hydrogen exchange chemistry.

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

2.1. Hydrogen exchange and hydrogen bonding Despite the usefulness of amide hydrogen exchange analysis in structural biology there are surprising limitations in our understanding of some of the fundamental aspects of the phenomenon. It is clear that exchange is suppressed by the participation of amide hydrogens within hydrogen bonds in protein or nucleic acid structure, and in the original analysis of Linderstrom-Lang, exchange is proposed to require `structure-opening' ¯uctuations that remove the amide hydrogen from an exchange-protected state to an exchange competent state [1]. These may be called `closed' and `open' states whose interconversion is characterized by rate constants ko and kc, and an equilibrium constant Ko (Eq. (1); [11]) (see Appendix A). In the simplest formulation, exchange occurs from the open state with characteristics of a free amide (i.e. an amide in a so-called `random-coil' state having no residual conformational contributions to exchange; i.e. kch ˆ kint ; (see Appendix A) see Sections 2.4 and 4.1). ko

kch

NHc O NHo ! NHp o : kc

…1†

There is now abundant evidence that the major factor limiting amide hydrogen exchange with solvent is the participation of the amide in a hydrogen bond. Where crystal structures are available for analysis of detailed hydrogen bonding patterns, amide hydrogens exhibiting signi®cant exchange protection are, in the large majority of cases, found to participate in hydrogen bonds [12±15]. The sequestration of buried amides within the protein core and their limited accessibilities to solvent (and exchange catalyst) is expected to lead to protection from exchange. It is rare, however, for buried amides to be without hydrogen bonding partners. The balance of free energies in folded proteins may allow this to some extent (some possible examples are cited in Ref. [16]), but both expectation, and analysis of crystal structures, indicates that this is uncommon. The protection of amides to exchange correlates with the location in hydrogenbonded secondary structure as well as with `depth' of burial in the protein, or solvent accessibility [17±20], and there are many examples of solvent accessible amide hydrogens which are highly exchange protected, for example in surface helices [21±24].

137

Measurements of amide hydrogen exchange in isolated helical peptides in water [25,26] and methanol [27±29] indicate correlations between hydrogen bonding and exchange protection. This is not to say that suppression of an exchange rate below the rate expected for a free amide hydrogen (kint) (see Appendix A) necessarily indicates that an amide participates in a hydrogen bond. Signi®cant apparent exchange protection of surface amides, that are known from examination of crystal structures to be without hydrogen bonding partners, have been characterised in bovine pancreatic trypsin inhibitor (BPTI) [30±32]. Conformation-dependent electrostatic effects [33,34] (and possibly steric effects [35]) can signi®cantly suppress (or enhance, in the case of electrostatic effects) hydrogen exchange through their contribution to the stability of the charged exchange intermediates [25,27±32,36] (see Section 2.2). However these effects are, in principle, predictable in terms of the known contributions to amide hydrogen exchange kinetics that include local and global electrostatic effects, and less well characterized effects such as the solvent accessibility of the amide carbonyl in relation to acid catalysed exchange [31] (Section 2.2). Experimental studies to date indicate that these effects might contribute factors of up to 30±100 fold in terms of their effects on the exchange rate constant [20,25,30,31]. No consensus has been reached on the extent of amide exchange protection that should be taken to indicate hydrogen bonding, and the matter is often dealt with in a qualitative manner. In assigning hydrogen bonds in NMR-derived structures, suitably exchange-protected amides have conservatively been taken to be those observable in a two-dimensional NOESY, COSY or HOHAHA experiment; i.e. having an exchange half life of longer than around 2±3 h at pH 4±5.5. This corresponds to an exchange protection factor (at 208C) of at least 150±200 (relative to poly-D,L-alanine; see Fig. 1; Section 2.2) which is probably an acceptable value. The inclusion of hydrogen bonds in structure calculations is only made after initial structures have been calculated without hydrogen bond constraints, so that hydrogen bond partners (backbone or side chain amide carbonyls) can be unambiguously de®ned. The hydrogen bond can then be included in subsequent structure calculations as hydrogen bond (H´ ´´O and N±H´ ´ ´O) distance constraints. Interpretation of exchange

138

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

Fig. 1. Exchange rate pro®les …log kex ; 258C) as a function of measured pH for exchange of nitrogen-bound hydrogens in unstructured polypeptides, simulated using published exchange rate constants (see Table 1): (Ð) peptide amide (poly-D,L-alanine; PDLA) [52], (¼) arginine e [208] (-´-´-) arginine h [208], (X) glutamine Z [209], (X) glutamine E [209], (- - -) tryptophan indole NH [210]. Data were measured using saturation transfer (H±H exchange; Kw ˆ 10214 †† except for PDLA (H±D exchange; Kw ˆ 10215:05 † whose data is corrected for the isotope effect on the pH-electrode …pHcorr ˆ pH measured 10.4). Ref. [52] also contains H±D exchange rate constants for some of the side chain exchangeable protons.

protection factors in terms of hydrogen bonds in pulse labelling experiments to identify hydrogen-bonded folding intermediates, necessarily utilize smaller protection factors (often no more than 10±25 fold) because of the low exchange protection often found in these states (reviewed in Refs. [2±7]). Con®dence in the validity of the interpretation in terms of hydrogen-bonding is enhanced by making interpretations based on as many amides as can be identi®ed, but there may be instances where hydrogen bonds in folding intermediates have been incorrectly assigned. The well de®ned role of hydrogen bonding in suppression of amide exchange leads to the conclusion that the `structure-opening' events of Eq. (1) are a class of backbone ¯uctuations in which hydrogen bonds are transiently broken freeing the amide hydrogen for exchange with solvent. Apart from some wellde®ned circumstances, little is known about the nature of these ¯uctuations either in terms of timescales and amplitudes. Recent information on the nature of

exchange-limiting ¯uctuations from NMR studies is discussed in Section 4. The aim of many quantitative analyses of amide hydrogen exchange is the determination of an exchange protection factor (PF; see Section 2.4) which, for a given amide, is the extent to which exchange is suppressed relative to a free (non-hydrogen-bonded) amide. When all in¯uences on exchange kinetics, other than protection by hydrogen bonding, are eliminated, the protection factor is a measure of the equilibrium constant (Ko) for the exchange-limiting structural ¯uctuation (i.e. PF ˆ 1=Ko ; where Ko ˆ ko =kc in Eq. (1)). Reliable determination of exchange protection factors requires consideration of the chemistry of amide exchange. 2.2. Amide hydrogen exchange chemistry Hydrogen exchange is both acid- and base-catalysed. A pH-independent contribution to exchange

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

139

Table 1 Exchange rate constants for nitrogen-bound hydrogens of proteins in water kOH (M 21min 21)

kH (M 21min 21)

kW(M 21min 21)

pHmin e

Peptide amide (208C) a N±Ac±A±A±A±NMe Poly-D,L-Ala

2:3 £ 1010 1:1 £ 1010

110 42

0.032 0.032

3.8 3.7

Side chain amide b Asn Z E Gln Z E

5:6 £ 109 1:3 £ 109 3:3 £ 109 8:4 £ 108

4:8 £ 104 5:4 £ 104 6:8 £ 104 6:8 £ 104

4.5 4.8 4.7 5.0

Trp indole NH (278C) c

3:8 £ 108

6:0 £ 103

4.6

5:97 £ 1011 4:04 £ 1011 2:62 £ 1011 1:51 £ 1011 2:70 £ 1011

198 132 4:2 £ 104 4:5 £ 104 9:5 £ 103

2.3 2.3 3.6 3.7 3.2

d

Guanidino group (258C) e-NH (0.05 M KCl) (1 M KCl) h-NH2 (0.05 M KCl) (1 M KCl) Guanidine (0.05 M KCl)

e Data for peptide amides measured in D2O (H±D exchange); pHmin value are corrected for the deuterium isotope effect on the hydrogen electrode [52]. All other data were measured by saturation transfer NMR (H±H exchange). pHmin ˆ 0:5 ‰log Kw 2 log …kOH =kH †Š where Kw ˆ 10214 (H2O) and 10 215.05 (D2O). Values for exchange rate constants for non-peptide amide side chain NH's in D2O (H±D exchange) are given in [52]. a Data taken from Ref. [52]. Ref. [52] contains residue-speci®c corrections to kOH and kH resulting from inductive and steric contributions from amino acid side-chains, as well as activation energies for calculating kH, kOH and kW at different temperatures. Side-chain speci®c corrections for a-aminoisobutyric acid are in Ref. [197]. b Data taken from Ref. [209]. c Data taken from Ref. [210]. d Data taken from Ref. [208]

through catalysis by water has been identi®ed [37] and is included in the equation for contributions to exchange in water (Eq. (2)). The measured exchange rate constant (kex) is the sum of the acid-, base-, and water-catalysed contributions, and is pH dependent as indicated in Eqs. (2) and (3). kex ˆ ka ‰H1 Š 1 kb ‰OH2 Š 1 kw ;

…2†

kex ˆ ka 102pH 1 kb 10…pH 2 KW † 1 kw ;

…3†

log …kex † ˆ log …ka ‰H1 Š 1 kb ‰OH2 Š 1 kw †:

…4†

The measured exchange rate constant (kex) is proportional to the acid (H 1) and base (OH 2) catalyst concentrations, respectively, leading to the well-characterized curves obtained when the logarithm of the measured exchange rate constant is plotted against pH (Fig. 1), as de®ned in Eq. (4). Base catalysis of peptide amide exchange is more effective than acid catalysis

by around 8 orders of magnitude (Table 1) so that the minimum of the pH-dependent exchange curve (pHmin) in water is around pH 3 for peptide amides. The contribution to the experimental amide exchange rate constant from pH-independent (water-catalysed) exchange is signi®cant only at pH values near pHmin. Side chain exchangeable hydrogens have pHmin values generally similar (Arg h), lower (Arg e) or higher (side chain amide and Trp indole NH) than the peptide amide (Table 1; Fig. 1); the exchange rate constants for these groups are larger under most conditions than for the peptide amide (Table 1; Fig. 1), and measurement of their exchange rates in peptides generally requires use of magnetization transfer methods (Section 3). NMR spectroscopists are familiar with the observation that side chain amide groups of polypeptides in water may be observed near pH 5, whereas at higher and lower pH values the signals are lost by saturation transfer from the irradiated solvent signal. The pH-dependence of

140

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

exchange rates (Fig. 1) accounts for the observation of the Arg e proton signals at very low pH, particularly where these are involved in structure (e.g. Ref. [38]). The side chain amides and Trp indole protons may have signi®cantly suppressed exchange rates when sequestered and/or hydrogen bonded within protein structures, and under these circumstances exchange is slow enough to be measured using time-resolved methods (e.g. Ref. [39]). Signi®cant deviations of pHmin from pH 3 are observed in polypeptides due to sequence-dependent inductive contributions to exchange and generalized electrostatic contributions (see Section 2.3), and additionally in proteins from pH-dependent effects on protein structure and stability. Because most proteins undergo acid-induced destabilization, unfolding, aggregation or denaturation at pH values where acid-catalysed exchange is dominant, hydrogen exchange measurements from proteins are usually made in the base-catalysed exchange regime. Measurements of hydrogen exchange from peptides in non-aqueous solvents are often made, for example, to characterise helical conformations of membranebinding peptides or membrane proteins (or fragments) which have low solubility or are unstructured in water. The most useful solvents for these purposes are methanol, tri¯uoroethanol (or TFE:water mixtures) and chloroform:methanol mixtures. Since these are protic solvents, pH-dependent exchange data is obtainable notwithstanding complications resulting from junction potential effects at the pH electrode [27]. pH-Dependent exchange data from unstructured peptides has been characterized in methanol [29] (Table 2), allowing residue-speci®c protection factors to be measured for polypeptides in that solvent [27±29,40]. No detailed analysis of

exchange in other protic solvents has been made so that interpretations of hydrogen bond stabilities in studies in these solvents have been qualitative (e.g. Refs. [41,42]). As described in Section 6, exchange from membrane proteins and fragments has been measured in detergent micelles, and a detailed analysis of the contribution of the sodium dodecylsulphate (SDS) micelle to exchange of unstructured, micellebound hydrophobic amides has been described [43]. Acid- and base-catalysed exchange rate data in long acyl chain N-methyl amides partitioned within positively and negatively charged micelles were recently reported [44] (see also [45]). Exchange rate measurement of melittin in dimethylsulphoxide±water mixtures [46], and cytochrome c in tri¯uoroethanol (1% H2O) [47], have also been described. Earlier studies on peptide amide exchange in non-aqueous solutions and amide exchange chemistry is reviewed in [8]. 2.3. Amide hydrogen exchange mechanisms Base catalysed amide exchange occurs via the imidate anion resulting from abstraction of the amide proton by OH 2 (Eq. (5)) [48]. Acid-catalysed exchange may occur by one of two mechanisms. The ®rst of these, reversible protonation of the amide nitrogen by H 1 (Eq. (6)), is complementary to the base-catalysed mechanism so that acid- and base-catalysed exchange intermediates have, respectively, a positive or a negative charge at the amide nitrogen. RCONHR 0 1 OD2 O RCON2 R 0 O RCONDR 0 ; …5† RCONHR 0 1 D1 O RCONHD1 R 0 O RCONDR 0 : …6†

Table 2 Exchange rate constants for backbone peptide amides in (deuterio)methanol (from Ref. [29])

Ac±A±A a ±NH2 Ac±K±G±NH2 Ac±K±A±NH2 Ac±K±U b ±NH2 a b

kOH (M 21min 21)

kH (M 21min 21)

pHmin

kmin (min 21)

1:85 £ 1010 5:0 £ 1011 1:8 £ 1011 2:0 £ 1010

32 4.0 2.6 8.2

3.92 2.75 2.88 3.61

7:7 £ 1023 1:4 £ 1022 6:9 £ 1023 4:0 £ 1023

Exchange data is for the amide NH of the amino acid in bold text at 208C. U is a-aminoisobutyric acid.

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

It has been postulated that acid-catalysed exchange from peptide amides should occur via protonation at the amide oxygen (imidic acid mechanism; Eq. (7)) because of the enhanced basicity of the oxygen over the amide nitrogen [49], and kinetic studies of simple amides support this mechanism [50,51]. In terms of exchangelimiting ¯uctuations freeing hydrogen bonded amides in peptides and proteins, the imidic acid mechanism requires that hydrogen bonds involving both the amide carbonyl and amide NH must be freed for acid catalysed exchange. Evidence for the imidic acid mechanism in peptides and proteins and its consequences for interpretation

141

of acid catalysed exchange rates is described in Section 2.5. RCONHR 0 1 D1 O RC…OD† ˆ N1 HR 0 O RC…OD† D2 O

ˆ NR 0 O RCONDR 0 : …7† 2.4. Exchange protection factors The amide exchange protection factor, PF, is simply the extent to which the experimental amide exchange rate (kex) is suppressed relative to the exchange rate (kint) of a non-hydrogen bonded amide

Fig. 2. Exchange rate pro®les (in D2O) for poly-D,L-alanine (solid curve), and for the amide NH between an unstructured Gly±Asn dipeptide (dotted curve) and an Ile±Val dipeptide (dashed curve) simulated using Eq. (4) and the correction factors for the acid- and base-catalysed exchange rate constants of Bai et al. [52]. These curves cover approximately the range of variation of kmin due to sequence-dependent inductive and steric contributions to backbone peptide amide exchange. The horizontal dotted lines indicate the approximate range of experimental exchange rates accessible to several of the NMR methods described in the text (Section 3), assuming that for time-resolved methods the experimental measuring time is faster than the half-life for exchange. (a) Upper limit for two-dimensional 1H 15N HSQC (20 min) measurement; (b) upper limit for standard 1D (1 min) measurement; (c±d) sensitive range [90% (c) to 10% (d) residual intensity] for saturation transfer measurement [NH T1 ˆ 1 s (see Eq. (8))]; (e±f) sensitive range [90% (e) to 10% (f) residual intensity] for polarized INEPT in 15N labeled peptide ‰…1 JNH ˆ 95 Hz (see Eq. (9))]. The use of transfer lines for acquisition of 1D data within 6 s of introducing samples into the NMR probe has been described [63].

142

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

in an unstructured polypeptide having the same dipeptide sequence at the same pH and temperature; i.e. PF ˆ kex =kint : The protection factor yields quantitative evaluation of an apparent free energy characterising the exchange-limiting backbone ¯uctuation …DG ˆ 2RT ln Ko where Ko ˆ ko =kc ; Eq. (1)), according to Ko ˆ 1=PF: However, several factors must be considered for estimating the most accurate values of Ko from the measured protection factor. For hydrogenbonded amides, it must be established that the exchange data is measured within the EX2 exchange regime in which the structure-opening ¯uctuation is in pre-equilibrium with the chemical exchange event …kc q kch ; Eq. (1)) as described in Section 4. Since amide exchange is intrinsically slow whereas the ¯uctuations underlying amide exchange occur generally on a rather fast timescale, this assumption is usually satis®ed for exchange from proteins at pH values below pH 8.0 (Figs. 1 and 2; Section 4.1). EX2 kinetics can be validated by establishing a ®rst order relationship between the experimental exchange rate, kex, and [OH 2], although this requires that the stability of the protein is unaffected over the pH range tested (Section 4.1). Amino acid residue-speci®c inductive contributions to the stability of acid- or base-catalysed exchange intermediates, together with steric contributions that suppress both acid- and base-catalysed exchange have been characterized by Bai et al. [52](see Fig. 2). Residue-dependent correction factors to kb and ka for random coil polypeptides are listed in that paper, as well as values for the activation energies of the rate constants for normalizing exchange data obtained at different temperatures. Small corrections for isotope effects, relevant, for example, for comparing exchange data in H2O measured using magnetization transfer (H±H exchange; Section 3.2) with H±D exchange data measured in D2O, are listed in an accompanying paper [53]. Global charge effects resulting from the condensation of counterion with charged polymers, which can have large effects on hydrogen exchange from polypeptides and proteins [34], particularly at pH values far from their isoelectric points, are often neglected. These effects are largely attenuated at high ionic strengths [34], and it is advisable to collect exchange data at moderate ionic strength (e.g. 200 mM buffer concentration) for use with the corrected kint data of Bai et al. [52] in calculating apparent Ko values.

The largest remaining error in the use of protection factors for estimating apparent Ko values is the extent to which the ¯uctuation underlying amide hydrogen exchange leaves the amide in a state that differs from a random coil, resulting in residual steric and/or electrostatic contributions to exchange from the open state (NHo; Eq. (1); note that kch ˆ kint for an open state equivalent to a random coil). The extent of this contribution is related to the amplitude of the exchange limiting backbone ¯uctuation which is considered in Section 4.2. 2.5. Exchange protection factors and acid-catalysed amide exchange The complications incurred by consideration of the mechanism of acid catalysed amide exchange (Section 2.3) may seem unnecessary given the general utilization of base-catalysed exchange rate measurements to probe hydrogen-bonded structure and dynamics. However, acid-catalysed exchange rates can be measured for polypeptides and for acid-stable proteins (notably the bovine pancreatic trypsin inhibitor; BPTI), yielding potentially useful information on hydrogen bond stabilities, the accessibilities of amide carbonyls, and the cooperativity of hydrogenbond-breaking structural ¯uctuations [21,25,27±32]. Unfortunately, the situation is not straightforward because the mechanism of acid-catalysed exchange is not completely resolved. The following discussion illustrates both the dif®culties and the potential for advance once the issues are clearer. To a ®rst approximation, base-catalysed exchange and acid-catalysed exchange via N-protonation are expected to be affected in equal and opposite ways by chemical inductive and generalized electrostatic effects since the exchange intermediates have, respectively, a negative or a positive charge on the amide nitrogen (Eqs. (5) and (6)). This complementarity should lead to shifts of pH-dependent exchange curves to higher or lower pH without greatly affecting the value of kmin, the exchange rate at the minimum of the pH-dependent exchange curves (see Figs. 1 and 2). This is generally consistent with the effects of aminoacid-speci®c inductive contributions characterized by Englander and his colleagues [52] (Section 2.4). These effects act on the conformational state in which exchange occurs, i.e. the open state of Eq. (1).

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

Three consequences follow: First, if the open state is random coil (i.e. only sequence-dependent contributions to exchange remain) then pHmin values will be equal to those calculated using the data of Bai et al. [52]. Second, any shifts of pHmin to higher or lower pH report on the electrostatic contributions in the open state. Analysis of pH-dependent exchange data has been used, for example, to identify helix dipole contributions to exchange of amide hydrogens near the ends of helices [25,27±31]. Third, because kmin values are relatively unaffected by electrostatic effects, exchange protection factors can be calculated from the suppression of kmin relative to kmin for a free amide [27±29]. These protection factors should be minimally affected by electrostatic contributions in non-random open states, yielding a more accurate measure of the hydrogen bonding contribution to exchange suppression. The situation is quite different if acid-catalysed exchange occurs via the imidic acid. For hydrogen bonded amides in elements of secondary structure, the requirement for protonation of the amide carbonyl oxygen (Eq. (7)) indicates that hydrogen bonds involving the amide NH and the amide carbonyl must be broken for exchange to occur. In this case, more extensive ¯uctuations may be required for acid-catalysed exchange than those necessary to free the amide NH for base-catalysed exchange. Greater suppression of acid-catalysed exchange compared to base-catalysed exchange is expected unless the dominant ¯uctuational mode in both cases is cooperative breaking of hydrogen bonds involving the amide NH and carbonyl. The loss of the complementary nature of base- and acid-catalysed exchange removes much of the advantage relating to improved determination of exchange protection factors described in the previous paragraph, although the observation of pHmin values signi®cantly different from calculated values remains an indication that the exchangelimiting ¯uctuations do not yield random coil, open states whatever the mechanism of acid-catalysed exchange [21,27±31]. Perrin has shown that the imidic acid mechanism should operate for exchange of the peptide amide (reviewed in Ref. [51]), although the preference over N-protonation is not large (around 100-fold) and the possibility remains that N-protonation might compete with the imidic acid pathway for amides having inaccessible, or strongly hydrogen-bonded,

143

Fig. 3. N-protonation of a peptide amide NH sequestered in protein secondary structure (e.g. an a-helix whose axis is represented by the parallel lines). Perrin [54] has suggested that N-protonation is unlikely to be a productive exchange pathway in this circumstance since only the incoming solvent proton (HS) can be lost without rotation about the C±N bond required to make HE accessible. While this does not preclude N-protonation as a mechanism for acid-catalysed exchange of amides involved in secondary structure, it suggests that a conformational ¯uctuation of signi®cant amplitude is required. Based on Fig. 7 of Ref. [54].

carbonyls. Perrin has also pointed out that N-protonation is not compatible with small amplitude hydrogenbond breaking ¯uctuations of secondary structure since, without rotation about the N±C bond after Nprotonation, deprotonation necessarily involves the loss of the same (incoming) proton [54] (Fig. 3). Experimental data tends to support the imidic mechanism in polypeptides and proteins. Tuchsen and Woodward [30,31] observed a dependence of acid-catalysed exchange on the accessibility of the amide carbonyl in surface (non-hydrogen-bonded) amides of BPTI. A correlation between pHmin and the amide 1H chemical shift in apamin was suggested to result from contributions from the imidic acid mechanism [21]. Rohl and Baldwin [25] ®tted acidand base-catalysed exchange data from a poly-ala(lys) helical peptide to theoretical analysis of helix-coil transitions. They found that acid-catalysed protection factors ®tted best to theoretical values that were calculated by assuming that both the amide NH and carbonyl were freed from helical hydrogen bonds in the exchange-competent state. For the helical peptide alamethicin in methanol, acid-catalysed exchange protection factors were larger (by between 5 and 50 fold) than base-catalysed protection factors, indicating that ¯uctuational modes allowing base-catalysed exchange are more highly populated than those required for acid-catalysed exchange [29]. The issue

144

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

is not completely settled and is worth further exploration: if the imidic acid mechanism is found to be the obligatory mechanism for acid-catalysed exchange, comparison of acid- and base-catalysed protection factors may prove a useful indicator of the amplitude of exchange-limiting backbone ¯uctuations [25,29]. 2.6. Relayed imidic acid mechanism An extension of the imidic acid mechanism has been proposed that potentially allows exchange of amide NH's buried within the interior of proteins, if these amides are connected to solvent via a chain of hydrogen bonds [31]. This so-called `relayed imidic acid' mechanism (Fig. 4) allows exchange by charge delocalisation through the hydrogen-bonded network without the requirement for penetration of charged exchange catalyst or large amplitude ¯uctuations to bring the interior amide hydrogen into contact with bulk solvent. It requires diffusion of water (the source of exchangeable protons) into the exchange site, a phenomenon well characterized by NMR spectroscopy [55]. The relayed imidic acid mechanism has been used to explain the pH-dependence (high values of pHmin) of exchange of solvent-inaccessible amide NH groups at the end of hydrogen-bonding networks in several proteins [56]. A base-catalysed relayed imidic acid mechanism (see Fig. 4 legend) has been

invoked to explain the pattern of exchange-resistant amides in a membrane-reconstituted gramicidin, measured using solid-state NMR [57] (Section 6). 3. NMR methods for measuring hydrogen exchange rates 3.1. Time-resolved exchange measurements When a polypeptide or protein is dissolved in D2O the amide signals in the high resolution NMR spectrum decrease in intensity as the amide hydrogens exchange with deuterons. The measurement of individual ®rst order exchange rate constants is limited only by the resolution of the amide signals themselves, or of cross peaks in homo- or heteronuclear multidimensional NMR spectra, and the rate constants for exchange relative to the time resolution of the NMR acquisition (Fig. 2). Consideration of the exchange rate pro®les for backbone amides (Figs. 1 and 2) indicates that unprotected amide hydrogens in water can be measured by time-resolved methods only at pH values near pHmin and by using one-dimensional accumulation. However, the stabilisation of amides by hydrogen bonding within polypeptide and protein structure can bring these amides (usually the most interesting ones!) into exchange regimes accessible

Fig. 4. Acid-catalysed relayed imidic acid exchange mechanism proposed by Tuchsen and Woodward [31] (see also Ref. [56]). Note that an analogous base-catalysed relayed imidic acid mechanism may be obtained involving abstraction, by base, of a water-accessible NH proton, such as that at the bottom of the ®gure (see Ref. [57]).

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

to time-resolved sampling of exchange kinetics, and the majority of exchange measurements on proteins have been made this way. The ideal method is to acquire HSQC spectra from uniformly 15N labelled protein since these spectra have good resolution in the 15N dimension, can be acquired rapidly [around 15±30 min (or faster [58,59]) for a 2 mM sample] and give in-phase cross peaks for accurate intensity measurements. When 15N-labelled protein is not available, 2-dimensional NOESY or HOHAHA spectra can be acquired with some loss of signal resolution, sensitivity and time resolution [spectral acquisition of normally around 4±8 h for a 2 mM sample, although rapid acquisition using minimal phase cycling [60,61], 2D selective pulse acquisition [62] and/or non-phase sensitive acquisition [36,62] can reduce spectral acquisition times for extracting ®rstorder exchange rate constants to 0.5±1 h or less). Because several amide cross peaks may be assigned for any amide, limitations in cross peak resolution can often be overcome, and measurements under these circumstances can provide internal checks of rate constants by measuring from multiple cross peaks. The use of in situ stopped-¯ow mixing methods limits the delay between introduction of the sample into the magnet and spectral accumulation, and the acquisition of 1-dimensional spectra for measuring amide exchange with a time resolution of six seconds has been described [63]. Spectral simulation methods, although not widely used, should allow maximum information from poorly resolved signals of exchanging amides. A linear prediction method has been described for slow amide exchange rates sampled in a series of onedimensional spectra in which free induction decays are ®tted to a model constructed by assigning the relevant spectral parameters for each exchanging amide (i.e. its frequency, line-width, initial intensity, phase parameters and exchange rate) [64]; these procedures are more generally applicable and have been used, for example, to determine exchange rates manifest as line broadening in the indirectly detected dimension of 2-dimensional spectra collected on a timescale in which hydrogen exchange is signi®cant [65] (see Section 3.4). Extraction of exchange rates from two unresolved amides can often be made by analysing exchange data with a biexponential ®t. If the exchange rates differ by a factor of 1.5±2 (depend-

145

ing on the quality of the data) the individual rate constants are usually resolvable and can often be assigned to their respective amides using one or more 2D experiment which allow the faster and slower exchanging amides to be identi®ed (e.g. Ref. [28]). Improvement in resolution of amides for extraction of exchange rates in 1D or 2D spectra can often be made by `editing' on the basis of exchange rates themselves. After relatively long exchange-out periods, a subspectrum of the most slowly exchanging amides often remains from which accurate rate constants can be measured. Likewise, exchange rate constants for the most rapidly exchanging amides can be measured by `exchange-in' experiments in which the polypeptide or protein is fully exchange-deuterated and amide exchange (D ! H) is measured in H2O [27,30±32]. 3.2. Magnetization transfer Magnetization transfer methods (recently reviewed in Ref. [66]) are used to measure fast exchange rates on a timescale inaccessible to time-resolved methods (exchange rates of around 0.1±50 s 21), and have been applied to exchange of poorly protected amides of proteins and unstructured polypeptides. Like many aspects of protein NMR, the scope of magnetization transfer has been extended by isotopic ( 15N) labelling, and more recently by the use of pulse ®eld gradient methods (see Ref. [67]). In its classical application [68±70] applied to 15N-labeled polypeptides [71] the amide cross peak intensity in a heteronuclear HSQC or HMQC experiment is attenuated on presaturation of the water resonance when saturation transfer becomes competitive with T1 relaxation of the amide proton according to, kex ˆ …M0 2 Ma †=…M0 T1a †;

…8†

where T1a is the intrinsic relaxation rate of the proton in the absence of saturation transfer. For hydrogen bonded amides, this relationship is valid only in the limiting case (high motility limit [70]) where the conformational transition between hydrogen-bonded (closed) states and the open state(s) from which exchange occurs is much faster than the chemical exchange rate (ko, kc q kch ; Eq. (1)) as described more fully below; for non-hydrogen-bonded amides this quali®cation does not apply. Two further factors

146

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

complicate the simple use of the relationship for accurate exchange rate determination: the inability to measure T1 under solution conditions where the exchange rate measurements are made, and potential NOE contributions to amide or crosspeak signal intensities from the irradiated water, from other perturbed protein signals at frequencies close to the water, or from protons (especially side chain OH) in rapid exchange with water. Because of the logarithmic sensitivity of the amide exchange rate to pH (Figs. 1 and 2), measurement at a lower pH where saturation transfer from water makes no contribution to amide relaxation, may be used to determine T1. A value for M0 may also be determined under slow exchange conditions (low pH) with solvent presaturation, thus correcting for NOE contributions [21,71]. This approach requires that the intrinsic amide T1 and solvent-presaturated signal intensity are independent of pH; in the absence of pH-dependent changes in conformational properties (apparent, for example, from the pH-independence of chemical shifts), the pH-independence of T1 is likely. Measurement of the apparent amide exchange rate at several pH values over the range where saturation transfer is observed should yield the same value for the base catalysed exchange rate constant, again assuming insigni®cant pH-dependent changes in protein conformational properties over the relevant pH range [71]. The attenuation of the amide 15N signal intensity in heteronuclear polarization transfer experiments, that results from loss of phase coherence as the amide exchange rate approaches the timescale of the heteronuclear coupling constant ( 1J 1H 15N), has been exploited for measuring fast exchange rates in selectively 15N-labeled proteins using INEPT (insensitive nucleus enhancement by polarization transfer; [72]) [73,74]. Exchange rates may be measured over a wide window of rate constants between around 0.35±200 s 21 by combining the effects of saturation transfer (exchange rates on the order of 1=T1 † and of the loss of phase coherence in polarization transfer (exchange rates on the order of the heteronuclear coupling constant ( 1J 1H 15N). In this experiment, the recovery of the amide proton signal after saturation of both the amide and water signals is measured (rather than the attenuation of the amide signal following saturation of water as in conventional saturation

Fig. 5. (a) pH-dependence of 30.4 MHz 15N INEPT NMR spectra (1 s relaxation delay) of [ 15N]Gly-labelled M13 coat-protein in SDS micelles. (b) 15N INEPT spectra [Gly 3 near 8.8 ppm in (a)] recorded with a relaxation delay of 30 s at different pH values between pH 6.0 and 8.2. (c) pH-dependence of Gly 3 INEPT signal intensities for data from b. [30 s relaxation delay (K)] ®tted using Eq. (9). Peak heights from (a) [1 s relaxation delay (X)] were ®tted using Eq. (13). Constants required in Eqs. (9) and (13) were: T1a ˆ 0:53 s; T1w ˆ 3:3 s; JNH ˆ 94:5 Hz: Reprinted with permission from [73]. Copyright 1990 American Chemical Society.

transfer experiments), and the recovery of amide magnetization after saturation transfer is analysed as described for this special case by Forsen and Hoffman [68]. The effect of exchange on the recovery of amide magnetization is given by Eq. (9), and of saturation transfer, by Eq. (10) ([74]; derived according to Forsen and Hoffman [68]), Mz ˆ M0a exp …2kex =JNH †;

…9†

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

Mz ˆ M0a 1 C1 exp ‰…1=T1a 1 kex †tŠ 1 C2 exp…t=T1w †; …10† where t is the delay between transients, M0a is the equilibrium amide magnetization, T1a and T1w are the amide and water T1 values respectively and C1 and C2 are constants, which in the case of saturation transfer in water, simplify to: C1 ˆ M0a …1 2 T1a =T1w †=‰…T1a =T1w † 2 kex T1a 2 1Š; …11† C2 ˆ …M0a kex T1a †=‰kex T1a 2 …T1a =T1b † 1 1Š:

…12†

Combining the effects of polarization transfer and saturation transfer yields an equation (Eq. (13)) describing the pH-dependence of the amide 15N signal intensity to which experimental data may be ®t to extract the amide-speci®c, base catalysed exchange rate constant, kOH (see Eq. (14)) by solving for kex. Mz ˆ ‰M0a 1 C1 exp …1=T1a 1 kex t† 1 C2 exp…t=T1w †Šexp…2kex =JNH †; 2

kex ˆ kOH ‰OH Š ˆ kOH 10

…pH214†

:

…13† …14†

Because the timescale for kex measurements utilising exchange contributions to polarization transfer …kex < 100 s21 † is faster than for saturation transfer …kex < 2 s21 for proteins, since T1 , 0:5 s†; exchange measurements based only on Eq. (9) must be made under conditions where saturation transfer is absent (i.e. with long relaxation delays in refocused INEPT, or under conditions where water magnetization is not excited in potential HMQC/HSQC applications to fast exchange measurements) (Fig. 5). Utilization of Eq. (13), however, allows exchange measurements to be made over a broad range of kex (and pH) using short relaxation delays. By ®tting exchange data acquired across a range of pH values where magnetization transfer is manifest, independent estimates of kex can be made yielding greater con®dence in the value of kOH. Henry and Sykes [73,74] have shown that experimental exchange data for [ 15N]acetylglycine and for selectively [ 15N]-labelled M13 coat protein in detergent micelles ®t well to Eq. (13) (Fig. 5). pH-Dependent effects on either protein stability or local electrostatic contributions from titrating amino acid side chains that can intro-

147

duce a pH-dependence to kOH will result in deviations between experimental data and the pH-dependence of Mz using Eq. (13). In most cases, the attenuation of the amide signal or cross peak intensity on presaturation of solvent OH is used to determine the exchange rate according to Eq. (8) as described above. As noted above, this equation is valid only in the so-called high motility limit [70] where the conformational transitions between any exchange protected (closed) and exchange accessible (open) states are characterized by rate constants that are much larger than the chemical exchange rate constant and the amide longitudinal relaxation rate (T1). In the linear form of the Linderstrom-Lang model [1], in which exchange from the exchangeprotected state is assumed to be insigni®cant (Eq. (15)), the high motility limit requires that ko, kc q kch ; k 0ch ; Rc, Ro, where R is 1=T1 ; and Rc and Ro refer to the closed and open states, respectively [70,75]. It is not obvious whether this assumption is valid in any given case. ko

kch

kc

k ch

NHc O NHo O HOH: 0

…15†

Most backbone amides in proteins studied using high resolution NMR have T1 values of the order of 0.5±1 s, so that saturation transfer should become apparent and measurable for unprotected amides at pH values above around pH 5 where the exchange rate constant, kex, for amides in unstructured polypeptide (e.g. poly-D,L-alanine, PDLA) is near 0.1 s 21 at 208C (Fig. 2). Observable saturation transfer of hydrogen-bonded amides in the pH range where proteins are generally stable, is expected only for amides having small to medium protection factors (less than around 10 4-fold) and at relatively high pH. For an amide having 1000-fold exchange protection, for example, saturation transfer may be observed only above pH 8 where kex …ˆ kint † for PDLA is around 100 s 21. The rate constants for hydrogen-bond breaking conformational transitions, ko and kc(app), have been measured for the most stable amides in a few proteins at high pH where contributions from EX1 kinetics are signi®cant …kch , kc † [19,76±78](see Section 4.1. for methods of estimating these values). Under these conditions values of ko of 10 23 ±10 27 s 21 and of kc(app) of 0:1±5 £ 103 s21 have been found. Recent measurements on

148

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

base-stable proteins at very high pH have allowed measurement of ko for some native state ¯uctuations (see Section 4.1) of up to 200 s 21 and greater [79,80]. The rate constants for conformational transitions freeing amides from less stable hydrogen bonds in proteins are expected to occur on a fast timescale but whether these generally occur on the millisecond timescale or greater as required to maintain the high motility limit has not been established. Jardetzky and coworkers [75,81,82] have characterised exchange of marginally stabilised amides in the D and E helices of the trp repressor where the high motility limit does not apply. They propose comparison of exchange rate estimates from relaxation measurements using the McConnell equation (Eq. (16); [83]) with rates estimated from saturation transfer (using Eq. (8)) as a test of the high motility limit. Lack of agreement indicates that the high motility limit is invalid, and that the data must be analysed using the generalized LinderstromLang model, in which contributions from relaxation and exchange from the closed state (NHc; Eq. (15)) are considered. Under conditions where both the open and closed states make signi®cant contributions to the amide relaxation, estimates for the three exchange rate constants (ko, kc and km) and the average longitudinal relaxation rate, T1, can be obtained by examination of the complete longitudinal relaxation pro®les at each pH, both with and without solvent saturation. 1=T1app ˆ 1=T1 1 kex :

…16†

From a detailed analysis of amide relaxation recovery in terms of the Linderstrom-Lang model, Zheng et al. [75] describe three behaviours that can be expected to characterize amide exchange from polypeptides and proteins in the absence of saturation. (1) If ko is less than Ro, relaxation is exclusively from the closed state and pH-independent single exponential relaxation is observed. Identical relaxation curves will be obtained in the presence of presaturation of water. This behaviour is expected for all amides exchanging slowly enough to be observed in deuterium exchange experiments. (2a) Amides that are largely in the open state …ko q kc † yield single exponential relaxation pro®les which exhibit pH dependence since the contribution to relaxation from solvent exchange is pH-dependent. (2b) If ko and kc are much faster than kch (high motility limit [70]) then exchange occurs

from an averaged state and pH-dependent single exponential relaxation pro®les will again be observed. The chemical exchange rate, kch, might become faster than kc and ko at suf®ciently high pH, in which case biexponential relaxation recovery will be observed. (3) pH-Dependent biexponential relaxation recovery occurs if both open and closed states are signi®cantly populated and ko, kc are smaller than, or comparable to Rc. In case (3), the relaxation recovery is most sensitive to ko, kc, kex and Rc, and complete analysis of the pH dependence of the recovery curves allows the rate constants of the Linderstrom-Lang formulation (Eq. (15)) to be determined. These analyses [75,81] are of interest because they provide, under certain circumstances, a second means (the other being the circumstance where EX1 kinetics makes a signi®cant contribution to exchange; Section 4.1) of de®ning the rate constants for the opening ¯uctuations underlying hydrogen exchange. Hydrogen-bonded amides in the D and E helices of the trp repressor display pH-dependent biexponential relaxation kinetics (case 3 above) indicating that both the open and closed (hydrogen bonded) states contribute to relaxation through a slow opening ¯uctuation (ko, kc # Ro ; Rc). The ¯uctuation rate constants underlying exchange of these amides are 0.2±1.3 s 21 (ko), 0.1±1.0 s 21 (kc) and the open and closed states are roughly equally populated. Because the exchange protection factors (between 10 and 70) are signi®cantly larger than around 2 (expected for an exchange competent state existing around 50% of the time; i.e. ko =kc < 1 for these amides), the analysis indicates that the open state is not equivalent to a random coil state which lacks sequence-independent contributions to exchange (i.e. kch , kint †: A curious conclusion arising from the analysis of Gryk et al. [81] is that the amide 1H and 15N chemical shifts of the D and E helical amides are the same in the closed and open states; despite the fact that the equally populated states are in slow exchange on the chemical shift time scale, a single unbroadened cross peak is observed in 1H 15N HMQC spectra. This is proposed to result from a highly structured open state in which localized distortions free individual amides from helical hydrogen bonds while maintaining overall helical geometry. This interpretation may not be completely satisfactory. First, the amide chemical shift is known to be very sensitive to the polarization

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

of the N±H bond (and thus the hydrogen bond length [84]), which is unlikely to be identical in the open and closed states for all the helical amides of the D and E helices. Secondly, highly localized ¯uctuations of small amplitude are expected (at least by this reviewer) to have faster closing rate constants (kc) than those extracted from the kinetic analysis. Recent studies have focussed on direct measurement of fast exchange rates under conditions in which the water magnetization is selectively inverted [59,85], or re-aligned along the 1z axis [86±88], before a mixing time during which detectable amide magnetization recovers by exchange. These methods do not require independent determination of the amide T1, and have the advantage of yielding spectra in which only the rapidly exchanging amides are observed, resulting in good spectral resolution. As in all exchange-in rate measurements, care must be taken to determine an accurate amide intensity equivalent to an unperturbed or `fully re-exchanged' intensity. This may be made from the part of the experiment (usually an HMQC, HSQC or variant) used to detect amide signals, with corrections made for possible variations in the extent of residual equilibrium water saturation between exchange measurement and control spectra, where applicable. A source of uncertainty in exchange rate measurements using these methods remains the contribution of NOEs. Direct measurement of fast amide exchange rates in uniformly 15N- and 13C-labeled molecules without independent determination of T1, and largely free of NOE effects, has been described by Gemmecker et al. [89] using the MEXICO (Measurement of EXchange rates in Isotopically labeled COmpounds) experiment. Isotope ®lters are used to remove observable magnetization from all 15N- and 13 C-attached protons while preserving water magnetization, eliminating these sources of NOE contributions to the recovery of amide magnetization. The amide protons recover magnetization by exchange with solvent water during a mixing period, and recovered magnetization is sampled after increasing mixing periods using 2D heteronuclear spectroscopy. Amide T1 contributions to magnetization transfer at all mixing times are eliminated by subtracting alternate scans in which water magnetization is aligned along the 1z and 2z axis, respectively, since the contributions from exchange are then alternately positive and

149

negative, whereas T1 contributions are always positive. Amide±amide NOE effects between protons recovering magnetization by exchange during the mixing time cannot be eliminated. Although their effects might be minimized by adjusting the pH to sample the more rapid rates accessible where the mixing times are short compared to the NOE buildup rates, this approach is probably limited by the loss of cross peak signals in heteronuclear 1H 15N correlated spectroscopy resulting from loss of phase coherence when the amide exchange rate approaches 1 J 1H 15N (,90 s 21), which limits observation of 1 15 H N cross peaks in any heteronuclear correlated spectra to those exchanging with rate constants less than around 100 s 21 (Fig. 2). Fast exchange rates for the mannose permease domain P13, measured using the MEXICO experiment and using conventional saturation transfer HMQC [71] yielded similar results indicating that cross relaxation contributions to amide magnetization recovery in the MEXICO experiment are small. This comparison tends to validate both the classical saturation transfer method (using HMQC/ HSQC) in which T1 values and amide cross peak intensities (i.e. incorporating NOE contributions) are determined under conditions where saturation transfer is inef®cient (low pH) and the MEXICO experiment. A variation of the MEXICO experiment employing a single heteronuclear ( 15N) ®lter to eliminate amide proton magnetization has been described [90] for fast exchange rates in short 15N-labelled polypeptides in which artefacts from cross-relaxation are minimized by adjusting conditions (peptide size, temperature and spectrometer magnetic ®eld strength) to the NOE minimum …vo tc < 1†: While this approach is not applicable to proteins, it is potentially valuable for accurate determination of exchange rate constants in polypeptide fragments (where the condition vo tc < 1 can be met) for comparison with exchange rates calculated using Bai factors [52] (Section 2.4). One of the limitations of experiments that utilize selective excitation at the water frequency is radiation damping [91] which results in rapid relaxation of water magnetization that can occur during the selective pulse. A simple method of estimating amide exchange rate constants [92] makes use of radiation damping as an active element for restoring the equilibrium magnetization of water shortly (30±60 ms at a magnetic ®eld strength corresponding to 600 MHz for

150

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

protons) after a non-selective inversion (p) pulse [88,93]. A reference spectrum, obtained by gradient suppression of radiation damping during the mixing time in alternate scans, is subtracted, leaving a spectrum [detected as a two dimensional (pseudo-3D) NOESY-HSQC spectrum selective at the water frequency] in which only signals having interactions with spins at the water frequency appear. Errors in rate estimation may arise from NOEs to hydroxyl protons in rapid exchange with water, and by NOEs with CHa protons at the water frequency excited by radiation damping, although, where measurable, the latter were shown to be negligible, suggesting that the `selective pulse' due to radiation damping is highly selective. Determination of accurate exchange rate constants cannot be made due to undetermined exchange process occurring during the relatively slow recovery of equilibrium water magnetization induced by radiation damping. However measurement at higher magnetic ®eld strengths with correspondingly shorter (tenths of ms [94]) radiation damping recovery times should allow this method to be used for accurate rate measurements. Mori et al. [20,94±96] have described the measurement of fast exchange rates (0.1±100 s 21) using a `water exchange ®lter' (WEX-®lter) for selective observation of protons whose intensity builds up via exchange processes with signals at the water frequency. The WEX ®lter employs a selective 908 pulse at the water frequency followed by a short gradient phase encoding pulse. A hard 908 pulse de®nes the start of a mixing time during which phase-encoded magnetization is transferred to exchanging protons. Transverse magnetization (protein protons) resulting from the second 908 pulse is spoiled by a crusher gradient during the mixing time. The end of the mixing time is de®ned by a 908 pulse (which may be selective on the region of interest) to transform longitudinal magnetization into observables, which is rephased by a second gradient coherence selection pulse. Only signals resulting from the original water selective pulse (including protons in exchange with water) are rephased; the recovery of their intensities as a function of the mixing time, detected using WATERGATE [94,97] or fast HSQC (FHSQC [95]), can be analysed to extract exchange rates. The method allows accurate determination of exchange

rates by measurement of the initial slope of the build up curves since the start of the mixing time is well de®ned (by a hard 908 pulse). In addition, the water intensity is dephased throughout, minimizing the effects of radiation damping. By inserting a spin echo ®lter between the water selective pulse and the hard 908 pulse de®ning the start of the mixing time, the contribution of NOEs (largely from CHa protons at the water frequency) can be estimated [96]. This arises from the loss of water intensity during the spin echo period (around 40±60 ms; Fig. 6) which should result in a decrease in amide signal intensity by a wellde®ned factor (fs). If the intensity decrease is greater than fs, NOE contributions are included in the exchange rates (k 1 NOE), and the pure exchange rate, k, and the NOE contribution can be separated. These authors have described application of a phase-modulated CLEAN chemical exchange spectroscopy (CLEANEX-PM) in the mixing period of water-selective exchange experiments, in which residual NOE artefacts are suppressed [98,99]. These include NOE from water `bound' within the structure of the protein; NOE contributions from bulk water appear as negative signals. Comparison of exchange rates of rapidly exchanging amide hydrogens of staphylococcal nuclease with rates determined using the WEX-®lter experiments described in the previous paragraph, have identi®ed amides whose apparent exchange rates contain spurious NOE contributions (Fig. 6). The use of heteronuclear Hartmann-Hahn polarization transfer (rather than INEPT-type transfer) in exchange-edited HSQC spectroscopy has been shown to yield signi®cantly enhanced sensitivity in the measurement of fast exchange rates in 15Nlabelled proteins [100]. A recent method utilising the exchange-mediated decorrelation of heteronuclear two-spin order to estimate exchange rates in 15Nlabelled amides (`DECOR' [101]) has been extended to the measurement of amide exchange rates in 15Nlabelled proteins. This method, DECOREXSY [102] (incorporating elements of decor and exchange spectroscopy; `EXSY' [103]) measures the decay of the heteronuclear two-spin order term which has contributions from amide exchange and heteronuclear relaxation. The latter contribution is determined independently and subtracted to yield the absolute exchange rate in the submillisecond time scale. This

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

151

Fig. 6. NMR spectra, recorded at 11.74 T, of 15N-labeled Staphylococcal nuclease (1.5 mM, pH 6.8, 378C) measured using (a) fast HSQC (FHSQC), (b) WEX II-FHSQC …Tm ˆ 100 ms†; (c) spin-echo-®ltered WEX II-FHSQC, and (d) CLEANEX-PM-FHSQC. Signals in (b) arise from proton exchange with water, NOEs from CHa protons under the water frequency and exchange-relayed NOEs from rapidly exchanging protons. Signals in (a) that are absent from (b) exchange slower than the detection limit (around 1 s 21). The spin-echo ®lter (40 ms) identi®es signals in spectrum b in which observable magnetization arises from NOE contributions [these have greatly reduced intensity in spectrum c and are marked with solid arrows in spectrum (b)]. Residual exchange-relayed NOEs [open arrows in spectrum c] are eliminated in spectrum d Exchange rates can be measured from the Tm-dependence of amide signal intensity [20,98]. Reprinted with permission from [99].

experiment is more sensitive than the WEX-®lterFHSQC method and is apparently free of problems (particularly relayed NOEs from other protons) that may be dif®cult to eliminate from other methods. These new experimental approaches (particularly the WEX-FHSQC and CLEANEX-PM), several of which have appeared during the writing of this review, are already making valuable contributions to the study of protein dynamics on the submillisecond timescale [80,104] (see Section 4.1). A study of the accuracy of exchange rate measurements from 15N-directed WEX-FHSQC, utilizing protein fully-deuterated in the aliphatic protons to eliminate virtually all possible artifactual contributions to apparent exchange rates, is in progress [104]. Measurement of exchange rates of hydroxyl protons of carbohydrates is of interest in terms of their possible involvement in hydrogen bonds. The rapid exchange rates requires the use of magnetization transfer methods which present speci®c dif®culties since the hydroxyl protons lie close to the water signal (typically less than 1±2 ppm down®eld) which is

dif®cult to excite selectively using traditional saturation transfer methods. A number of approaches have been described to overcome this dif®culty [105±107]. Some of the methods described above, for example the active utilization of radiation damping for selective excitation of water at very high magnetic ®elds [92] may also be applicable to this problem. 3.3. Pulsed ®eld gradient diffusion measurements The physical basis of the measurement of exchange rates using pulsed ®eld gradients (the differential diffusion rates in solution of polypeptides or proteins and an exchanging `ligand') is independent of nuclear spin phenomena (e.g. Ref. [108]), affording an alternative method for measuring hydrogen exchange rates in which knowledge of the relaxation behaviour of the exchanging species is not required. Several approaches to the measurement of fast exchange rates (in the range of around 1±10 4 s 21) have been described [109±111]. The method makes use of the principle that the diffusional mobility of the

152

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

exchanging ligand (in this case an exchanging proton) re¯ects the diffusional mobilities of the two molecules between which it exchanges (i.e. polypeptide and bulk water) weighted by the respective lifetimes in each state [112]. In its basic form (a variant of 2D exchange spectroscopy; 2D EXSY [103,113]) the diffusion measurement employs two gradients either side of an evolution period during which diffusion and exchange takes place [112]. The detected signal, S, is a function of the gradient strength, G, and duration, d, the diffusion time, tdif, and the effective diffusion coef®cient, Deff, according to

3.4. Exchange line broadening

…18†

Classically, line broadening in NMR spectroscopy due to chemical exchange, is analysed quantitatively to extract exchange rate constants. While these methods are used to determine exchange rate data for small molecular weight amides (e.g. Ref. [51]) they are not practical for polypeptides and proteins. However, the time-dependent loss of amide signal intensity resulting from H±D exchange in D2O can result in linebroadening in the indirectly-detected dimension of 2D NMR spectra, which can be exploited to extract exchange rate constants from single spectra, as described in two reports published in 1993 [115,116]; see also Ref. [65]. These are particularly useful for exchange rate measurements from amides exchanging too quickly for time-resolved acquisition of 2D spectra (e.g. kex . 0:002 min21 for 2D homonuclear spectra; kex . 0:05 min21 for 15N HSQC), for poorly resolved amides exchanging on a time-scale accessible to time-resolved acquisition of 1D spectra, and for amides exchanging too slowly for measurement using magnetization transfer …kex , 10 min21 †: An application of the analysis of line broadening

If the condition tdif q tp does not apply, then a more general treatment is required to distinguish between exchanged and non-exchanged spins (the contribution of the latter will disturb the diffusion weighting) [110,112]. Andrec and Prestegard [109] described an approach that combines the advantages of selective inversion exchange measurements [59,85,86,89] and spin-echo diffusion measurements [114] using pulsed ®eld gradients. The experiment described allows the measurement of simple exchange rates with results comparable to those obtained using selective water inversion exchange measurements [59]. Although not shown experimentally, theoretical analysis of the experiment indicates that mediated exchange in which, for example, water to amide exchange occurs via an internal bound water molecule, can be studied. In a recent report, the use of the longitudinal-eddy-current-delay (LED) NMR pulse sequence with pulsed magnetic ®eld gradients have been used to determine amide NH proton lifetimes in the cyclic antibiotic peptide viomycin [111].

Fig. 7. pH-Dependence of 75.45 MHz 13C NMR spectra of partially 13 C-carbonyl-labelled M13 coat protein (0.8 mM) in SDS micelles in equimolar H2O±D2O at the pH indicated (only the signal for Gly3 13C carbonly is shown) (right hand series). The left hand series are simulated spectra from which the exchange rates of Asp-4 NH were obtained. Reprinted with permission from [118]. Copyright 1987 American Chemical Society.

S ˆ S0 exp …2g2 G2 d2 Deff tdif †:

…17†

If the diffusion time, tdif, is large compared to the lifetime of the polypeptide-bound proton, then the effective diffusion coef®cient, Deff, is the sum of the diffusion coef®cients of the polypeptide, Dp, and water, Dw, weighted according to the lifetimes of the water bound proton, tw, and polypeptide bound proton, tp, respectively: Deff tdif ˆ …Dw tw 1 Dp tp †:

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170 13

to measure exchange rates in amide carbonyl Clabelled peptides and proteins makes use of the large isotope effect on the amide 13C carbonyl signal (around 0.1 ppm) depending on the state of the amide nitrogen (i.e. NH or ND). The 13C carbonyl doublet observed in equimolar H2O±D2O broadens and collapses into a singlet when the hydrogen±deuterium exchange rate falls within the usual regime for chemical exchange effects on the 13C spectrum [i.e. k , 2pdD where d D is the chemical shift difference in hertz between the low ®eld ( 13CONH), and high ®eld ( 13COND) [117] signals (Fig. 7). This method is sensitive to exchange rates in the range kex , 2±125 s21 †; and has proved useful for rate measurements in detergent-solubilized, small membrane proteins [118], and homopolymeric helical peptides [26]. 3.5. Exchange trapping methods In many cases amide exchange measurements under non-equilibrium conditions (that preclude magnetization transfer or gradient diffusion measurement) cannot be made using time-resolved methods, because exchange occurs too quickly, or because the sample composition limits acquisition of high resolution NMR spectra. Often exchange rates may be measured in these circumstances using exchange trapping procedures in which amide exchange occurring in real time is trapped throughout an exchange timecourse, after which the sample is transferred to slow exchange conditions where high resolution NMR spectra can be measured. Exchange trapping has been used to measure exchange rates in transient intermediates on protein folding pathways using quench ¯ow methods, and in polypeptides reconstituted in phospholipid bilayer membranes (see Section 6). Since these methods do not generally require interesting or novel NMR acquisition approaches they are not described in detail here, although their contribution to characterising protein folding pathways has been remarkable [2±7]. Exchange trapping experiments should employ a zero time sample (i.e. a sample taken through the full sample preparation procedure but experiencing the smallest exchange time accessible with the experimental protocol), which accounts for any artefactual exchange occurring through the work up and during NMR spectral acquisition. This

153

sample will de®ne the set of amides for which exchange rates can be measured (i.e. those suf®ciently stable under slow exchange conditions to be observable in one- or two-dimensional spectra from which amide intensities are measured) and provides the ®rst (zero) time point for use in calculating exchange rates. Different samples are used for each exchange time point so that exchange trapping methods require considerably more sample than other forms of exchange measurement. In addition care must be taken to normalize measured amide intensities in individual spectra by ratio against signals from nonexchangeable protons. 4. Timescales and amplitudes for protein ¯uctuations underlying hydrogen exchange Within the Linderstrom-Lang model for amide hydrogen exchange from polypeptides and proteins (Eq. (19)), an amide hydrogen is released from an exchange-protected state into an exchange-competent state via structural ¯uctuations. In line with thermodynamic theory, all possible conformations are accessed by a protein, populated at equilibrium according to their relative energies under given conditions. Potentially accessible to amide exchange analysis is the class of ¯uctuations that underlie amide hydrogen exchange. Given the evidence that most amides in proteins are stabilised from exchange by hydrogen bonding (Section 2.1), the class of backbone ¯uctuations of interest are those which break hydrogen bonds and free amide hydrogens for exchange with solvent. These ¯uctuations can be characterised by their amplitudes, opening and closing rate constants (ko and kc) and an equilibrium constant …Ko ˆ ko =kc †: Measurement of (apparent) Ko values is most accessible to exchange analysis whereas values for ko and (apparent) kc are accessible only under limited circumstances as described below. ko

kch

NHc O NHo O NHp : kc

…19†

4.1. Timescales In Eq. (19) NHc is the sum of all native states which have negligible contributions to exchange. If the assumption is made that kc q kch and/or that kc q

154

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

Fig. 8. Exchange rate pro®les …log kex † as a function of measured pH (pH p) for amide exchange in D2O (208C) for selected values of ko and kc, simulated by ®tting to Eq. (22). Values of ko (in units of s 21) are in heavy type; kc values (s 21) are in small italic type. Apart from the top set of exchange curves …ko ˆ 1 s21 †; the values selected are in the range calculated from pH-dependent exchange data covering the EX1 exchange regime in BPTI [76], lysozyme [19] and OMTKY [77]. Note that exchange-limiting ¯uctuations having ko values greater than around 10 22 s 21 …,0:5 min21 † cannot be normally be measured by time resolved methods (H/D exchange; cf. Fig. 2), although exchange trapping methods can be used to extend the accessible time regime (e.g. Ref. [79]). Measurement of approaches to EX1 kinetics by saturation transfer measurements [70,71], or perhaps more reliably using water magnetization transfer [20,86,89,94] (Section 3.2), or ®eld gradient diffusion measurements [109,110] (Section 3.3) can potentially be made with ko values of up to around 100 s 21, but in the relatively few studies to date, EX1 kinetics have not been observed (e.g. Ref. [80]). Whether this is due to the lack of extensive pH-dependent exchange data so far obtained or re¯ects the absence of native state ¯uctuations having ko values in the accessible range (,0.2±200 s 21) and which satisfy the EX1 condition …kch $ kc † remains to be determined. Also the EX1 condition may be more dif®cult to achieve for native state ¯uctuations since kch may be further suppressed relative to kint [81] (see text) in the exchange-accessible (open) state.

ko ; then the ®rst order rate constant for exchange of an amide via structure opening ¯uctuations to states NHo is [11]: kex ˆ ko kch =…ko 1 kc 1 kch †:

…20†

Full kinetic analysis of the Linderstrom-Lang model without the assumptions of Eq. (20) yields the generalized rate equation describing the observed exchange rate as a function of the rate constants ko, kc and kch for any values of the rate constants (Eq. (21); [11]). kex ˆ 1=2‰…ko 1 kc 1 kch † 2 1=2‰…ko 1 kc 1 kch †2 2 4ko kch Š1=2 :

…21†

For an amide existing largely in a hydrogen bonded

state at equilibrium …kc q ko †; Eqs. (20) and (21) simplify to: kex ˆ ko kch =…kc 1 kch †:

…22†

Values for kint (equivalent to kch for an unstructured polypeptide), the exchange rate for exchange from a random coil peptide (poly-D,L-Ala; PDLA), can be calculated for any conditions of pH, temperature and solvent conditions (e.g. D2O or H2O) from the relevant (acid- or base-catalysed) exchange rate constants, and the activation energies tabulated by Englander and colleagues [52,53]. It is instructive to consider exchange from hydrogen bonded amides in proteins in the base-catalysed regime, where most exchange measurements are made, to illustrate the kinetic regimes potentially accessible to analysis. In the

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

base catalysed regime, kch …ˆ kint † ˆ kOH ‰OH2 Š ˆ kOH 10…pH2pKw †

…23†

for PDLA in D2O at 208C, kch …ˆ kint † ˆ 2 £ 108 £ 10…pH215:05† s21 :

…24†

Fitting Eq. (21) (or Eq. (22) for proteins under conditions where kc q ko † to estimated values for ko and kc [19] illustrates that the pH-dependence of exchange from hydrogen bonded amides can access two exchange regimes according to the relative values of kc and kch (Fig. 8). The two limiting cases can be extracted from Eq. (21), and correspond to the pHdependent (denoted `EX2') and pH independent (denoted `EX1') portions of the pH-dependent exchange curve (Fig. 8): kex ˆ ko kch =kc ˆ Ko kch kex ˆ ko

…kc q kch ; EX2†;

…kc p kch ; EX1†:

…25† …26†

Under most conditions of amide hydrogen exchange from the native (folded) state of peptides and proteins, EX2 conditions apply. This is understandable in terms of the relatively slow kinetics of the chemical exchange process at pH values where proteins retain native conformations (normally below pH 10), and the (apparently) rapid rate of reclosing (kc) of native state ¯uctuations underlying amide exchange. Where EX2 conditions are established from the linear relationship between log kex and pH (Fig. 8), or from mass spectrometric measurements [119], the exchange protection factor, corrected for sequence-dependent contributions to exchange (Section 2.4), is directly interpretable as an apparent equilibrium constant …Ko…app† ˆ 1=PF ˆ ko =kc †: The value of Ko is an `apparent' value because it contains the assumption that the open state of the ¯uctuation is equivalent to a random coil state, operationally de®ned for amide exchange as one in which residual conformation-dependent contributions to exchange are absent. The extent to which this assumption is valid in any circumstance is intimately related to the amplitude of the exchange-limiting ¯uctuation (see Section 4.2) which is rarely well-de®ned. Fig. 8 illustrates that direct determination of ko (EX1 kinetic regime) by time-resolved (H±D) exchange measurements under native conditions far from the denaturation midpoint is possible only for

155

highly protected amides (very small Ko), small values of ko and at high pH where the EX1 limit becomes accessible …kch , kc †: This necessarily means that a class of rare and probably large amplitude (slow kc) ¯uctuations are accessible to measurement in the EX1 limit. EX1 kinetics under folding conditions …kc q ko † have been unambiguously characterized by NMR exchange measurements for the core b-sheet protons of BPTI (that require global unfolding for exchange), at high temperatures (688C, pH , 7:5±9†; yielding ko values around 2±5 £ 1023 s21 and kc(app) around 1:5±4:5 £ 103 s21 under these conditions [76]. EX1 kinetics for these amides was established from NOE experiments de®ning correlated exchange [120]. Contributions from EX1 kinetics were characterised for slowly exchanging amides in lysozyme by ®tting the pH-dependence of experimental exchange rate constants in the base catalysed exchange regime to Eq. (21) [19]. Values for ko ranged from ,10 24 to ,10 27 s 21 at 218C (pH 4±pH 8) with kc(app) between ,0.06 and 100 s 21. Most of the amide sites analyzed exchange via native state or sub-global ¯uctuations (see Section 4.2). The expectation that these slow closing (exchange re-protection) rates are associated with large amplitude opening ¯uctuations is contradicted by activation energy measurements that indicate small amplitude exchange limiting ¯uctuations for many of these amides [121]. A similar analysis of native-state exchange in acyl-coenzyme A binding protein (ACBP) in the pH range pH 5±pH 8 yielded exchange-limiting native state ¯uctuations having open lifetimes between 35 ms and 10 s …kc…app† ˆ 0:1 2 28 s21 † [78]. Interpretation of these data is complicated by the curious observation that the apparent slow closing rates for many of the amides are slower than the rates for global protein folding (kU±F) under equivalent conditions. Thus even at pH 5, 58C, in 0.7 M denaturant (where folding is slower than under conditions of the NMR exchange measurements) ACBP folds from denaturant with a measured ®rst order rate constant of around 33 s 21 [122], a rate that is faster than kc(app) for many of the local exchanging amides determined by Kragelund et al. [78]. The global folding of disulphide-intact lysoyme is complex, having multiple (fast and slow) components and at least one folding intermediate [123,124]. However the measured fast …kU±F . 30 s21 † and

156

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

slow folding species …kU±F . 2:5 s21 † are also faster than kc(app) for many of the amides that exchange by low activation energy ¯uctuations within the native state ensemble. The calculated value of kc(app) in these analyses is subject to potentially large errors since the value of kch, the rate constant for chemical exchange in the transiently open state is not known [19]. However, since the error is expected to be in the direction of an overestimation of kch (kch is likely to be suppressed relative to kint), any correction for suppression of kch in the native-state ¯uctuation will lead to a suppression of kc(app), increasing the anomalous relationship between kc(app) for many native state exchange-limiting ¯uctuations and kU±F in these proteins (see Fig. 8). It is dif®cult to escape the conclusion that the kc(app) values in these analyses are incorrect either because contributions to exchange from EX1 kinetics were not suf®ciently accessed under the conditions of the equilibrium exchange rate measurements or because the native state ¯uctuations do not conform to the Linderstrom-Lang model. Evidence for EX1 exchange kinetics in the turkey ovomucoid third domain (OMTKY3; [125]), barnase [126] and the immunoglobulin domain, CD2 [127], has been described. In CD2, the limiting value of ko …2:6 £ 1024 s21 † for the subset of amides for which EX1 kinetics were measured, is close to the independently determined value for the rate constant, kF±I …2 £ 1024 s21 †; de®ning the transition from the native folded state (F) to a folding intermediate (I). This indicates that exchange of these amides from the native state occurs via transient formation of the intermediate, an interpretation supported by the similarities in the protection factors within the folding intermediate determined directly by stopped ¯ow kinetic measurements under folding conditions [128], and by analysis of exchange from the native state at equilibrium [127]. EX1 kinetics have been con®rmed in the case of OMTKY3 [77] where exchange measurements at pH values up to pH 10 have allowed observation of approaches of kex towards ko. For the 13 slowly exchanging amides studied (most or all of which exchange via global unfolding), ko values of 0.003 to $ 0:03 s21 ; and kc(app) values of 100±10000 s 21 were determined. Exchange data on the highly base-stable OMTKY were recently extended to pH 12.2 using stopped ¯ow measurements [79], allowing contributions from EX1 kinetics

to be measured for a large number of amides that exchange from the native-state ensemble. Values of ko of 40±200 s 21 and kc(app) values of 9000±4:3 £ 105 s21 were determined, consistent with the expectation that native state ¯uctuations are likely to have considerably faster kc values than the rate constant for global folding (kU±F). A value for kU±F for OMTKY has not, so far, been reported, but is likely to fall in the range of 10 1 ±10 5 s 21 found for similar small single domain proteins (e.g. Ref. [129]), and largely consistent with the kc(app) values for the globally exchanging amides [77,125], described above. Since kex values for amides in unfolded proteins are generally close to kint values calculated from random coil data of Bai et al. [52], the kc values for globally exchanging amides in the above analyses should have only small errors resulting from residual conformation-dependent contributions to kch in the open states. A more extensive comparison of kc values calculated for globally exchanging amides in the EX1 limit with independent kU±F kinetic data under equivalent conditions may provide insight into the nature of the reclosing that re-protects amides after global opening, for example whether reclosing involves formation of an exchange-stable folding intermediate or the native state (as found for CD2 [127]). The EX1 kinetic data for barnase [126] has not been used to determine ko and kc(app) values. In light of the previous paragraph, an interesting observation in this protein is a shift in kinetic regimes under rather subtle conditions of temperature and pH, between (partial) EX1 and EX2 kinetics [130]. This results from the presence of a folding intermediate (I) on the pathway between unfolded (U) and folded protein (N; see Eq. (27), whose stability relative to U is sensitive to temperature. k1

k2

k2 1

k2 2

UOION

…k1 ; k21 q k2 ; k22 †:

…27†

Under conditions where I is more stable than U, exchange via global unfolding has a kc value dominated by fast formation of I (in which the amide is exchange protected), resulting in EX2 kinetics. A shift in temperature destabilises I relative to U so that reclosing (exchange protection) requires reformation of N which is very much slower giving partial EX1 kinetics.

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

4.1.1. Approaches to EX1 kinetics from saturation transfer/magnetization transfer/PFG diffusion measurements? Magnetization transfer methods, described in Section 3.2 for measuring fast exchange rates, can access kex values approaching 10 3 s 21, leading to the possibility that a class of sub-global ¯uctuations might become accessible to direct measurement (of ko) in base-stable proteins via the ¯attening of pH-dependent exchange curves towards the EX1 limit (see Fig. 8). In the relatively few studies to date [59,71,75,80], contributions from EX1 kinetics have not been observed. Particularly interesting is a study of amide exchange in rubredoxin from the hyperthermophile Pyrococcus furiosis [80]. Exchange rates for exchange-stable amides within the 15N-labelled protein were measured using the WEX-FHSQC [96] or CLEANEX-PM methods [98,99] described in Section 3.2, in which spurious contributions to intensity changes arising from sources other than direct magnetization transfer from bulk water to the amide site are negligible. Magnetization transfer between water and amide hydrogen was found for all amides, indicating that hydrogen bond breaking, and access of water and exchange catalyst (OH 2) occurs faster than a second at all amide sites within the protein. The maintenance of EX2 kinetics at least up to pH 12.5 for almost all amides (except F29 NH and W3 indole NH), corresponding to measured exchange rates of up to 100 s 21, demonstrates that ko values for exchange from the native state ensemble are on the millisecond timescale or greater. If these rates for ko are combined with exchange protection factors for this protein of up to 10 6 [131], kc(app) values on the nanosecond to microsecond timescale ensue. Again, the kc(app) rates are almost certainly overestimated due to the use of kint in determining protection factors [i.e. the protection factor is an overestimation of K o …ˆ ko =kc † if the open state has an exchange rate constant (kch) below the random coil value (kint)]. However the measurements indicate that rapid, low amplitude (see Section 4.2) ¯uctuations are pathways for native state exchange in a highly stable protein, an observation which is consistent with the analysis of exchange-limiting native-state ¯uctuations in OMTKY (if not with the similar analyses of lysozyme and ACPB). These results also suggest that some classes of native state ¯uctuations might become accessible to characterization by molecular dynamics simulation (see Section 5).

157

4.2. Amplitudes Current interpretation of a vast amount of amide exchange data on numerous proteins indicates that there are three main exchange modes which differ in the amplitude of ¯uctuations from native state structure. Exchange pathways involving global unfolding and native state ¯uctuations have long been distinguished on the basis of activation energies and denaturant sensitivities (see below). More recently, detailed analyses of the denaturant sensitivity of exchange of individual amides [132] in several proteins have indicated subclasses of amides that exchange via local excursions from the native state involving exposure of a subset of residues as a cooperative unit. Amides exchanging from these partially open states, denoted Partially Unfolded Forms (PUF [133]), have DGex values below DGex for global unfolding (see following paragraphs) and intermediate denaturant sensitivities. Amide exchange-limiting ¯uctuations involving global opening can be characterised in terms of both rate constants (Section 4.1) and amplitudes at least to the extent that the globally-open state is structurally characterised. This is a state in which hydrogen bonds are broken, accompanied by a catastrophic in¯ux of water into the protein (e.g. Ref. [134]). Where measured on proteins denatured at equilibrium (e.g. thermally-unfolded BPTI [135], acid-denatured cytochrome c [136] thermally-unfolded ribonuclease [137], urea-denatured lysozyme [138]) and taking into account the rather small effect of denaturant on intrinsic amide exchange rates [139], the exchange rate constants are similar to the exchange rate constants calculated from the Bai factors [52] …kex < kint †; indicating that these states are equivalent to a random coil, operationally-de®ned for amide exchange experiments as one without residual conformation-dependent contributions to exchange. On the other hand, it cannot be assumed that globally open states participating in the ensemble of states existing at equilibrium under non-perturbing conditions are equivalent to chemically- or thermally-induced denatured states. The clearest test is from a comparison of DGex (from equilibrium amide exchange) and DGUF (from thermal- or denaturant-induced unfolding); these should be the same if the equilibrium unfolded state is thermodynamically equivalent to

158

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

the denatured state. In practice, deviations between apparent free energies for unfolding measured by amide exchange and denaturant-induced unfolding are found, resulting partly from residual stabilisation of amides in the unfolded state, often around disulphide bonds (e.g. ribonuclease A and S [132,140]; cytochrome c [141]; OMTKY [125]; peptostreptococcal protein L [142], chymotrypsin inhibitor 2, [143]; LysN [144]), differences in the stabilities of proteins in D2O (where equilibrium H/D exchange measurements are made) relative to their stabilities in H2O (where estimates of DGUF are normally made), and the contribution of cis±trans isomerism around XPro peptide bonds that affects the energetics of protein folding from denaturant- or temperature-induced unfolded states but not DGex [141]. Corrections for the latter effects (D2O vs H2O, and the contributions from X-Pro cis±trans isomerism) can be made. In a recent analysis of 20 proteins, DGex for global exchangers and DGUF were equivalent (within 1 kcal M 21) in almost every case, when the relevant corrections were made [145]. These observations indicate a thermodynamic equivalence between the transient equilibrium open state underlying global amide exchange and the temperature- or denaturant-unfolded state, and suggests that global unfolding is required for exchange of the most highly protected amides in proteins. The latter conclusion leads to the possibility for characterising folding free energies for highly stable proteins (like those from extreme thermophiles; [131]) whose denaturation conditions lie outside accessible temperatures or denaturant concentrations. It remains to be seen whether there are any proteins (or protein domains) in which all hydrogen bonded amides are exchangeable without access to the globally unfolded state. One recent example may be the SH3 domain from a-spectrin in which DGex for all amides is signi®cantly lower than DGUF [146]. Several criteria distinguish amides that exchange via sub-global ¯uctuations from those exchanging by global unfolding. The activation energy for exchange by global unfolding is approximately equal to the denaturation enthalpy when the intrinsic activation energy for chemical exchange is subtracted [19,76,147]. Activation energies for exchange from the native state are smaller [147,121,80,141]. Amides exchanging by sub-global, low activation energy ¯uctuations were early found to switch to the high activa-

tion energy, global unfolding process on increasing temperature, resulting in curved Arrhenius plots [39,147,148,149]. Exchange measurements at high pressure indicate that the activation volumes for ¯uctuations that underlie native state exchange are smaller [150] than those for global exchange [151]. The apparent free energies for the transiently open state in native state ¯uctuations is lower than DGex for global unfolding, and amides exchanging from sub-global ¯uctuations have lower denaturant sensitivity (m value; see below) than m for global exchangers. The groups of Woodward and Fersht have employed site-speci®c mutants for distinction between global and sub-global exchangers which effectively corrects for the uncertainty in kch resulting from (presently unquanti®able) contributions from residual structure in the transient exchangeaccessible open state. Assuming that no major structural effect results from the mutation, dDGex ‰ˆ DGex…mut† 2 DGex…native† Š; is equal to dDGUF ‰ˆ DGUF…mut† 2 DGUF…native† Š for global exchangers but is smaller for amides exchanging by local ¯uctuations [152±154]; see also Ref. [155]. 4.2.1. Exchange under conditions that destabilise the native state Without independent experimental access to the class of protein ¯uctuations that allow amide exchange from the native state, the nature of subglobal ¯uctuations has been controversial. Although Perrin has indicated that the amide hydrogen can be directly abstracted from an intact hydrogen bond by the hydroxide anion [54], direct abstraction is highly suppressed (as indicated by slow amide exchange in fully solvent accessible hydrogen-bonded amides [54]), and it is assumed that hydrogen bond cleavage is required for exchange to occur. The major question has concerned the amplitude of native state exchange ¯uctuations, with two models proposing exchange by diffusion of exchange catalyst into the protein interior as a result of minor structural ¯uctuations so that exchange occurs from a state similar to the native state (`penetration model' [156]), or alternatively, by local structural ¯uctuations that bring regions of the polypeptide chain into bulk solvent (`local unfolding model' [8]). The reader is referred to a paper by Miller and Dill [157] which discusses the evidence for both models existing up to around 1995. As with many

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

159

Fig. 9. Theoretical curves de®ning DGhx (the free energy de®ning the stability of hydrogen-bonded amides with respect to exchange-labile open states, determined from the exchange protection factor) as a function of denaturant concentration for amides that exchange by global opening (heavy dots), or by sub-global (local) ¯uctuations (small dots) in denaturant-free solution. Exchange-limiting sub-global openings have negligible or small denaturant sensitivities (m value) so that apparent DGhx values are relatively insensitive to denaturant at concentrations where local openings dominate exchange. The exchange rates for all amides is the sum of contributions from sub-global and global opening as de®ned in [141,161].

controversies, it seems that aspects of both models are correct: analysis of native state exchange from proteins under conditions that destabilise the native state (presence of denaturants, high temperature or high pressure) have indicated a hierarchy of exchange limiting, sub-global ¯uctuations encompassing a spectrum of amplitudes. Amide exchange in sub-denaturing concentrations of urea or guanidinium chloride generally distinguishes amides that exchange by local ¯uctuations of the native state (low denaturant sensitivity) from those exchanging by global ¯uctuations [156,158]. The latter class of amides have exchange protection factors reduced in proportion to the denaturantinduced reduction in the free energy of the native state relative to unfolded states ([132,143,158]; see Fig. 9). These studies have the potential for resolving features of native state ¯uctuations since any denaturant sensitivity in sub-global ¯uctuations will re¯ect the exposure of new denaturant-sensitive surface (e.g. Ref. [159]). This yields information on the amplitudes of sub-global ¯uctuations in a manner

analogous to the information from activation energies for exchange, which has been interpreted in terms of the numbers of hydrogen bonds broken in the ¯uctuation [121,160]. Indeed, it is expected that the denaturant sensitivity (m) and the temperature sensitivity (activation energy) of sub-global ¯uctuations should be similar, and this was shown to apply in the one study (of cytochrome c [141,160]) in which detailed residue-speci®c data is available. Since sub-global ¯uctuations have low denaturant sensitivity whereas global unfolding reactions (exposing maximal denaturant-sensitive surface) have the highest denaturant sensitivity, increasing concentrations of denaturant progressively `recruit' amides on lower lying free energy isotherms onto the isotherm for global unfolding (see Fig. 9). This can be formalized into a two process (or more strictly multiprocess) model for exchange from proteins which re¯ects the fact that, of the spectrum of ¯uctuational modes allowing exchange of an amide, those that are most populated (lowest free energy) at equilibrium under any conditions will dominate the exchange

160

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

Fig. 10. Segregation of individual amides into cooperative structural units in apocytochrome b562. Amides are grouped according to similarities in DGhx and m (the slope of the variation of DGhx with denaturant concentration; solid lines) for amides that exchange by global unfolding (cooperative unit I) in panel A. Panels B and C group amides in cooperative units II (core residues of helix 1) and III (core residues of helix 4), respectively. Variation of DG and m values for amide within helix 2 of the protein (panel D) are illustrative of `helix fraying'. Identi®cation of these cooperative structural units leads to a model describing the minimal set of (amide-exchange-limiting) partially unfolded states that comprise the native state ensemble of apocytochrome b562 (right). Note that this ®gure does not do full justice to the original published colour ®gures [164] to which the reader is referred for further details. Reprinted with permission from [164]. Copyright 1998 American Chemical Society.

process [133,161]. Thus, whereas both global and subglobal ¯uctuations are accessible to any amide in the native state, sub-global ¯uctuations are more highly populated than global unfoldings for all but the most heavily exchange protected amides, and dominate exchange from the native state. Shifting the relative equilibrium populations of sub-global and global ¯uctuations by increasing denaturant concentration progressively increases the contribution of global unfolding until it overtakes the sub-global ¯uctuation. Initially it was considered that all the isotherms de®ning sub-global exchange processes join a single isotherm for global unfolding at progressively increasing denaturant concentration (Fig. 9). In the most detailed study of residue-speci®c denaturant sensitivities reported (cytochrome c in sub-denaturing concentrations of guanidinium chloride) [133,160, 162], a series of major sub-global isotherms with progressively increasing energies and denaturant

sensitivity were identi®ed which were interpreted in terms of local unfolding of cooperative units of secondary structure. Low denaturant sensitivity isotherms for individual amides within the cooperative unfolding unit join the sub-global isotherm of the folding unit of which they are a part (see Fig. 10 for data from apocytochrome b562). Similar resolution of partially unfolded units of secondary structure from equilibrium exchange measurements were subsequently described for ribonuclease H [163], for the hyperthermostable rubredoxin protein from Pyrococcus furiosus [131] and for apocytochrome b562 [164] (Fig. 10). These intriguing observations demonstrate a hierarchical nature of ¯uctuations within the folded state of proteins that can be mapped in terms of the location of cooperative units of hydrogen-bonded secondary structure in which denaturant (and temperature) sensitivity may allow estimation of exchange-limiting amplitudes. That these cooperative

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

units have unique positions on the energy landscape of the native state ensemble is indicated by their observation independent of perturbant (i.e. chemical denaturant or temperature in the case of cytochrome c [141,160]; denaturant or pressure in the case of apocytochrome b562 [164,165]). Recent measurements on T4 lysozyme [166] and LysN [144] have not identi®ed discrete cooperative exchange units, but rather a broad hierarchy of exchange limiting ¯uctuations displaying a continuum of m values; in each case the m values are linearly correlated with DGhx (the apparent free energy for the exchange limiting backbone ¯uctuation obtained from the exchange protection factor in denaturant-free solution). The IgG-binding domain of peptostreptococcal protein L lacks detectable partially unfolded states that contribute to exchange from the native state ensemble [142]. It is probably too early to conclude that these studies indicate qualitative differences in the nature of the native state ensembles amongst these proteins; i.e. it should be remembered that the hydrogen exchange method samples only those components of the ensemble that contribute to amide exchange. The relative energies of the partially unfolded forms identi®ed in the equilibrium exchange in cytochrome c, together with independent evidence for the nature of the folding pathway for the protein led to the suggestion that the hierarchical nature of the ¯uctuational structure re¯ects the relative order of formation of structure on the folding pathway [133,162,167]. This is a development of the proposal of Woodward and colleagues [158,168] that the stable core of amides exchanging by global opening in proteins de®nes the region of structure forming early on the protein folding pathway. The latter theory has been reviewed extensively by Li and Woodward [169]. A similar conclusion (to that of Englander and his colleagues) was made by Marqusee from exchange measurements of ribonuclease H [163,170] and by Parker and colleagues for the immunoglobulin domain CD2 [127,171]. No evidence for a relationship between the folding pathway (de®ned using independent methods) and equilibrium exchange behaviour has been found for barnase [152], chymotrypsin inhibitor 2 [143] or T4 lysozyme [166], and the potential pitfalls in correlating equilibrium exchange data with kinetic protein folding pathways in the absence of independent evidence have been noted [172].

161

The importance of considering both local and global contributions to exchange [161] in observation of denaturant-induced unfolding reactions has been demonstrated in studies on ribonuclease in unfolding conditions (denaturant concentrations above the denaturation midpoint [173]). Indeed, conditions may allow a switch from EX1 exchange conditions to EX2 conditions as the pH is increased (compare with Fig. 8). This may be illustrated in a simple example based on unfolding of ribonuclease A in 4.5 M guanidinium chloride [173]. Under conditions where there are contributions from global unfolding (kF±U; which, in unfolding conditions, will exhibit EX1 kinetics), as well as residual contributions from sub-global (native state) ¯uctuations (which may be on- or off-pathway in terms of unfolding), the experimental exchange rate will be the sum of the contributions from unfolding (EX1) and native state ¯uctuations (EX2 at moderate pH values): kex ˆ kF2U …EX1† 1 Ko kint …EX2†:

…28†

Consider the situation for amides exchanging via native state ¯uctuations and having a protection factor of 103 …Ko ˆ 1023 †: If the unfolding rate constant is around 3 £ 1023 s21 (as found for ribonuclease under the conditions described [173]), then at pH 5 …kint , 0:01 s21 †; global opening (EX1) will dominate exchange since kF±U …3 £ 1023 s21 † is much greater than the contribution from native state ¯uctuations …Ko £ kint ˆ 1025 s21 †: As the pH is increased, the contribution from native state ¯uctuations …ˆ 1022 s21 at pH 8.0) will overtake the contribution from global unfolding and the measured exchange rate will become pH-dependent. The results of numerous studies indicate, therefore, that at extremes of pH and under conditions that destabilise the native state relative to globally unfolded states (high temperature or denaturant concentrations), both EX1 and EX2 contributions to exchange should be considered. Finally, it is worth remarking on the possibilities for characterising sub-global ¯uctuations in structural terms. These ¯uctuations are (at best) characterised in terms of estimates of the number of transiently broken hydrogen bonds (from exchange activation energies) or transiently exposed hydrophobic surface (from denaturant-sensitivities). The possibility that activation volumes from exchange measurements at high pressure might yield amplitudes for sub-global

162

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

¯uctuations has only been tested recently [150,165]. At present, perhaps the best opportunity for gaining a feeling for the open states of sub-global ¯uctuations comes from calculational approaches in which substates of the native state conformational ensemble are generated by combinatorial unfolding of regions of native state structure [146,174], or in which the `ability' of residues to escape the con®nes of packing within the native state are assessed [175]. The success of these approaches in reproducing experimental protection factors in most [146,174,175] (see also [176]), if not all cases [166], is promising. The possibility that molecular dynamics simulations might help in characterizing exchange limiting ¯uctuations is brie¯y described in the Section 5. 5. Comparison of amide exchange data with molecular dynamics simulations and 15N relaxation measurements Comparisons of exchange protection factors with those calculated from molecular dynamics simulations have been made to explore possible contributions of rapid timescale ¯uctuations to exchange from native state conformations. Until recently, dynamics simulations have been limited to timescales of a few nanoseconds in small solvated systems, and in the context of amide hydrogen exchange, suffer from the low structural stabilities in water of polypeptides of accessible size and from the possibility that many of the important ¯uctuational pathways for exchange may occur on longer timescales. Consideration of the nature of exchange-limiting native state ¯uctuations of proteins (Section 4.2), for example, indicates that classes of ¯uctuations having signi®cant activation energies are likely to remain inaccessible to dynamics simulation for some time; thus, simulations of cytochrome c backbone dynamics, using geometrical de®nitions (N±H´ ´ ´O distance and angle) to characterize hydrogen bond opening ¯uctuations, do not correlate with experimental hydrogen exchange data [177], almost certainly because the ¯uctuations that underlie amide exchange are not represented in the simulations. On the other hand, measurements of alanine-based helical peptides in water using laser temperature jump techniques [178,179] have shown that the helix coil transition takes place on the

,10 28 s timescale, indicating that exchange-limiting ¯uctuations in these molecules (isolated helical polypeptides) might be accessible to simulation. Two studies [180±182] have described detailed comparisons of NMR-derived exchange protection factors in helical polypeptides in methanol [27,29] or water [26] with similar data from solvated MD simulations. The proportion of the time during a dynamics trajectory that an amide hydrogen participates in an intramolecular (helical) hydrogen bond can be converted into an apparent equilibrium constant for hydrogen bond opening [%time H-bondedˆ 100%=…1 1 K o †Š allowing comparison with experimentally determined protection factors …PF , 1=K o † [183]. Dynamics simulations of alamethicin and melittin from helical starting structures in methanol [180,181] support intramolecular hydrogen bonding as the dominant contribution to amide exchange stabilisation in these peptides since each exchange-protected amide [27,29] participates in intramolecular hydrogen bonds throughout most of the simulation. Best quantitative agreement between experimental and simulated hydrogen bond lifetimes is found if hydrogen bonds are classed as broken (exchange-competent) when the Ê. NH´ ´ ´O distance is greater than around 3.5±4.0 A This suggests that low amplitude backbone ¯uctuaÊ ) excursions over tions resulting in small (1±1.5 A acceptable hydrogen bond geometry (NH´ ´ ´O distance  occurring on a timescale of a few picose# 2:5 A† conds are not pathways for exchange in these peptides. During structure-opening ¯uctuations, replacement of intramolecular (amide carbonyl) hydrogen bond partners with solvent hydrogen bonding partners, and cooperative structure-opening ¯uctuations involving reversible opening of several sequential hydrogen bonds (with open lifetimes of 40±200 ps), were observed, each of which is expected to be relevant to the exchange process. Rather close quantitative agreement between apparent exchange protection factors (% time H bonded) from MD simulations of a poly-Ala based peptide [Ac-(AAQAA)3-amide] in water and experimental protection factors measured by NMR [26] has been described [182]. The criterion for hydrogen bond `breaking' in the simulations was an excurÊ separation between amide H and sion beyond a 3 A carbonyl O. The experimental measures of exchange protection were similar to independent measures of

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

residue-speci®c helix/coil ratios from amide carbonyl 13 C chemical shifts [26]. This similarity between measured and calculated protection factors determined from direct measurement of helix-coil distributions in the experimental study [26] is compelling evidence that the exchange limiting backbone ¯uctuations in this peptide (and generally in helical peptides in water [25]) are helix-coil transitions. The quantitative agreement of simulated and experimental protection factors using a criterion of hydrogen bond Ê [182], implies that very minor deviaopening of 3 A tions from helical hydrogen bond geometry result in exchange competence, which is rather different to the conclusions from the simulations of alamethicin and melittin in methanol. It would be instructive to analyze the open states in the water simulations to determine the extent to which these deviate suf®ciently from local helical structure to be equated with the `coil' states de®ned experimentally by local backbone geometry (f , c angles) and polarization effects of intramolecular hydrogen bonding on the amide carbonyl [26]. A comparison of experimental exchange protection factors in native and partially unfolded ubiquitin (in 60% methanol) [184] with protection factors calculated from dynamics simulations in the same solvent [35] has been made. Starting structures for partially unfolded ubiquitin were obtained from a high temperature simulation (498 K), and the conformation of the `denatured' state was sampled by MD simulation at 335 K. Agreement with experimental protection factors from partially unfolded (60% methanol) ubiquitin was obtained if the criterion for exchange protection during dynamics simulation was sequestration of the amide NH from water rather than the formation of hydrogen bonds. It was suggested that care should be taken in the interpretation of rather small protection factors for amides protected within `structured' intermediates on protein folding pathways using pulse-labelling techniques, in terms of native state hydrogen bonds. On the other hand, the correlation of low protection factors with true hydrogen-bonded secondary structure in partially folded states (e.g. Ref. [185]) and the rather complex exchange protection in folding intermediates in both a-helical [186] and b-sheet proteins [128] that correlates with native state hydrogen bonds, generally supports the interpretation of small protection factors

163

in terms of hydrogen bonds. Qualitative similarities between experimental exchange protection and solvent accessibilities in molecular dynamics simulations of chymotrypsin inhibitor II (CI2) [187] and lysozyme [188] have also been noted. Recent microsecond dynamics simulations of protein folding [189] suggest that comparison of experimental amide exchange data with hydrogen bond lifetimes and static or transient solvent accessibilities from simulations on much longer timescales (likely to be more relevant to the ¯uctuations underlying amide exchange from proteins) are now possible. Residue speci®c information on the rates and amplitudes of ¯uctuations on the ps±ns timescale are obtainable from 15N relaxation time measurements [190] interpreted using model-free analyses to yield estimates of correlation times and order parameters. For well-protected amides in proteins no necessary correlation is expected between amide exchange protection and 15N relaxation behaviour since the latter is dominated by ¯uctuations among well-populated substates whereas the exchange probably re¯ects transitions into rare, high energy conformers (see Section 4.2) [14]. It is probably true to say that little useful correlation (in the sense that predictive conclusions may be made) between 15N relaxation behaviour and amide hydrogen exchange has been observed (e.g. Refs. [191±194]). For example, while high exchange protection is often associated with high order parameters (low amplitude of ps± ns ¯uctuations) and low exchange protection with low order parameters [14,191,194], many hydrogen bonded amides with low exchange protection have low order parameters, indicating that these amides are associated with regions of the polypeptide chain that have low mobility on fast timescales (ps±ns) but high mobility on the slower timescales relevant to amide exchange. Thus, dynamic properties may, or may not, be correlated across timescales [191,192]. 6. Membrane peptides and proteins The analysis of amide exchange from membranereconstituted polypeptides has been pursued with the aim of determining structural and dynamic information from these experimentally dif®cult systems

164

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

similar to that obtained for proteins in solution. Early work using bulk tritium or infra-red studies of hydrogen±deuterium exchange showed that the number of amides exchanging slowly from bacteriorhodopsin corresponds to the number expected to be hydrogen bonded within transmembrane helices [195] whereas a signi®cant number of side chain amides exchange slowly from rhodopsin in membranes [196]. These studies suggested that amide-speci®c exchange analysis might give information on the location of trans-membrane helices in membrane proteins and perhaps on side chain interactions. Detailed analysis of amide-speci®c exchange rates for membrane peptides and proteins using NMR has until recently only been obtained after solubilising in non-aqueous solution [27±29] or detergent micelles [74,118,197]. An exchange trapping method (Section 3.5) allows amide-speci®c exchange rates to be obtained from polypeptides reconstituted in fully hydrated phospholipid bilayer membranes [40,62,198,199]. The procedure makes use of the high solubility of membrane-forming phospholipids and many membrane peptides and proteins (or protein fragments) in non-aqueous solvents (e.g. methanol, ethanol, tri¯uoroethanol, dimethylsulphoxide, dichloromethane and chloroform/methanol), enabling exchange-trapped peptide:lipid complexes (obtained by rapid freezing and lyophilization of reconstituted vesicles suspended for various periods in D2O buffer) to be co-dissolved in solvent where amide exchange is intrinsically slow, so that the extent of exchange of each amide at each time point can be assessed by oneor two-dimensional NMR. The large signals from the excess membrane lipids that are also dissolved with the peptide are suppressed using shaped excitation pulses selective for the peptide amide signals [62]. Amide exchange analysis of model amides [43] membrane polypeptides [40,198,199] and proteins [74,118,197] in detergent micelles or in reconstituted membranes has shown that the micelle or membrane per se has only a small effect on amide exchange that can be rationalized in terms of known steric and electrostatic contributions to amide exchange (Section 2.1). These observations indicate that hydrogen bonding remains the dominant contribution to exchange protection in the examples studied to date. This has been characterised in detail for alamethicin in a comparison of exchange in methanol [29], in fully

hydrated bilayer membranes [199] and SDS micelles [197]. The situation may be more complex in membrane proteins composed of trans-membrane helical bundles as indicated by studies of cysteine thiol exchange (SH/SD) in single cysteine mutants of phospholamban reconstituted in hydrated bilayer membranes [200]. Using FTIR to distinguish SH and SD absorption bands, it was shown that lipid exposed thiols are more readily exchangeable than those introduced onto interhelical surfaces, suggesting that water dissolved in the membrane lipid is the major source of exchange catalyst in this case. Whether or not similar conclusions might apply to backbone amide exchange in membrane proteins remains to be determined. A similarity in analysing amide exchange from membrane-reconstituted polypeptides and soluble proteins occurs under circumstances where amide exchange can occur by partitioning of the peptide from the membrane into the aqueous phase. This is observed for alamethicin reconstituted in dioleoylphosphatidylcholine membranes and is characterised by a linear dependence of the exchange protection factor on the volume of aqueous buffer in which the vesicles are suspended [199]. Eq. (29) de®nes the scheme in which amide exchange results from partitioning of the peptide from the membrane, where exchange is negligible (NHm), into the aqueous phase (NHa) where exchange occurs; kd and ka are the rate constants for dissociation and association of peptide and membrane, respectively. It is convenient to use kint (rather than kch) in these analyses, even though exchange rates from the aqueous state (kch) may not equal the random coil rates (kint), due to residual non-random conformation in water (this may be determined by measurement in buffer without membranes [199]). kd

kint

NH m O NHa ! NDa : ka

…29†

Under conditions where ka, the rate for reassociation of the peptide with the membrane is much greater than kint, the experimental exchange rate is (by analogy with Eq. (25); Section 4.1): kex ˆ kd =ka £ kint ˆ Kpar £ kint ;

…30†

where Kpar is the equilibrium constant de®ning partitioning of the peptide between membrane and

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

165

aqueous phases. Kpar can be adjusted by altering the volume of the aqueous solution in which the vesicles are suspended. If exchange also occurs from the membrane bound state (scheme 31), then the experimental exchange rate constant is the sum of exchange contributions from each state (Eq. (32)). Again it is assumed that exchange from the membrane state follows EX2 kinetics (the exchange limiting ¯uctuation is a fast pre-equilibrium event with respect to kint) and this can be established from the pH-dependence of exchange under conditions where exchange from the membrane state dominates. kint

Ko…m†

Kpar

Ko…a†

kint

‰NDo à NHo O NHc Šm O ‰NHc O NHo ! NDo Ša ; …31† kex ˆ kint Ko…m† 1 kint Kpar Ko…a† :

…32†

The experimental exchange protection factor …PF ˆ kint =kex † is given by Eqs. (33) and (34). Fig. 11 illustrates theoretical curves de®ning the relationship between measured exchange protection factors and the relative values of Ko(m) and Kpar for a membranereconstituted peptide partitioning into the aqueous phase. When Kpar is greater than Ko(m) for all the amides, the experimental exchange rates divided by individual kint values (determined by measurement from the aqueous peptide) yield the same value (equivalent to Kpar). As individual Ko(m) values become greater than Kpar (by shifting the equilibrium to the membrane bound state by decreasing the volume of aqueous buffer) the individual protection factors separate according to the relative hydrogen bond stabilities in the membrane bound state. This situation is analogous to the sequential recruitment of amides exchanging by non-global ¯uctuations in soluble proteins, onto the curve de®ning exchange from global opening as the local opening free energies match denaturantinduced global opening free energies (Fig. 9, Section 4.2). Exchange-trapping measurements made from reconstituted vesicles hydrated in minimal excess water allows protection factors for the membranebound polypeptide to be measured (199); protection factors for peptides that bind tightly (Kpar very small [40,198]), or for membrane proteins that presumably do not partition signi®cantly, can be determined with-

Fig. 11. Theoretical curves de®ning amide-exchange protection factors for a polypeptide partitioning between membrane and aqueous phases, obtained by ®tting to Eq. (34). The solid and dotted curves are ®ts with Kop(a) set to 1 and 0.1, respectively. Data were calculated for Kop(m) values ranging from 10 25 to 10 22 in 10-fold intervals (top to bottom). The horizontal sections of the curves represent protection factors from amide sites in the membranebound polypeptide under conditions where exchange from the membrane-bound state dominates exchange.

out consideration of the volume of aqueous solution. PF ˆ kint =kex ˆ 1=…Kop…m† 1 Kop…a† Kpar †;

…33†

log PF ˆ 2log ‰Kop…m† 1 Kop…a† 10…log Kpar † Š:

…34†

Detailed exchange analysis from small membrane proteins in membranes has not yet been achieved using these methods (e.g. Ref. [201]), although it is expected that de®nition of transmembrane helices, and questions concerning amide exchange from a protein unable to undergo global opening ¯uctuations, should be accessible using in situ exchange-trapping approaches. Amide-resolved exchange analysis of membrane-reconstituted polypeptides should also be useful in conjunction with FTIR exchange analysis of membrane systems for de®ning amide IR band subcomponents [202], and for aiding assignment of

166

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170 15

deuterium and N solid state NMR spectra of deuterium-exchange-labelled polypeptides reconstituted in oriented bilayers [203]. As with exchange from proteins in solution [119], mass spectrometry is expected to make signi®cant contributions to analysis of amide exchange from membrane peptides and proteins [204]. It is also likely that developing applications of solid state NMR to membrane protein structural analysis [205,206] will allow direct measurement of amide hydrogen exchange in membrane-reconstituted peptides and proteins. Two applications have been described by Cross and colleagues using gramicidin reconstituted in stacked lipid bilayers. Acquisition of deuterium NMR spectra during an exchange time course following addition of D2O to a sample hydrated in H2O allows exchange rates to be measured directly from the increase in signal intensities in the oriented deuterium NMR spectra. Estimates of exchange rates require correction for slow diffusion

of D2O that results from the low water content necessary for good bilayer orientation in these macroscopically oriented samples [207]. It was also shown that magnetization transfer deuterium NMR can be used to measure intensity changes in the oriented deuterium NMR spectrum upon saturation of the isotropic D2O signal at high pH [207], indicating that exchange rate estimates are accessible using this method. These authors recently described the use of 15N labelled peptide to characterise exchange-labile and exchange-resistant amides in reconstituted oriented bilayers, making use of the splitting in the 15N chemical shift spectrum (obtained using cross-polarization) resulting from dipolar interactions between 15N and 2 H, that yields a triplet centred on the 15N chemical shift position. Exchange of D for H under conditions of 1H decoupling [57] collapses the triplet to a singlet (Fig. 12). The pattern of exchange protection indicated the absence of exchange of internal amides with any water within the gramicidin pore and suggested that exchange in this system involves relayed imidic acid mechanisms catalyzed either by acid or base, as described in Section 2.6. Acknowledgements I am grateful to Dr Martin Parker for valuable discussion and comments on the manuscript, and to Dr Monica Ondarroa for reading the manuscript. Appendix A

Fig. 12. 40.58 MHz 15N chemical shift spectra of 15N-Leu10-labeled gramicidin reconstituted in stacked lipid (dimyristoylphosphatidylcholine) bilayers oriented on glass plates. Spectra were obtained using cross-polarization and proton decoupling. Exchange of the 15 N±H site with deuterium results in 15N±D coupling yielding a triplet (a). Spectra (b) and (c) were obtained under two sets of conditions in which H/D exchange was inef®cient (see [57] for details). The time-dependence of triplet formation can be used to estimate exchange rates in selectively 15N-labeled sites within the membrane-reconstituted peptide. Reprinted with permission from [57]. Copyright 1999 the Biophysical Society.

The following rate constants are de®ned here (see Eq. (1)): kex, experimental (measured) exchange rate constant; ko and kc, rate constants for the `structure opening' and `structure reclosing' events, respectively, that comprise the ¯uctuation underlying amide exchange; kch, exchange rate constant for the chemical exchange event; kint, `intrinsic' (random coil) exchange rate calculated for an amide depending on solution conditions (H2O vs D2O), pH, temperature and local dipeptide sequence (i.e. incorporating the corrections tabulated in Ref. [52,53]); ka, kb, and kw are rate constants for acid-, base-, and water-catalysed amide exchange, respectively; kc(app), apparent kc value [often determined from measurements of ko and Ko(app) (calculated from the exchange protection factor:

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

Ko…app† ˆ 1=PF; see text), and using Ko…app† ˆ ko =kc…app† Š: Other rate constants are de®ned where they appear in the text. The meaning of the terms `EX1' and `EX2', and the kinetic regimes which they de®ne (EX1; kc p kch ; pH-independent exchange) (EX2; kc q kch ; pH-dependent exchange), are described in Section 4.1. References [1] A. Berger, K. Linderstrom-Lang, Arch. Biochem. Biophys. 69 (1957) 106±118. [2] S.W. Englander, L. Mayne, Annu. Rev. Biophys. Biomol. Struct. 21 (1992) 243±265. [3] C. Woodward, Trends Biochem. Sci. 18 (1993) 359±360. [4] R.L. Baldwin, Curr. Opin. Struct. Biol. 3 (1993) 84±91. [5] S.W. Englander, T.R. Sosnick, J.J. Englander, L. Mayne, Curr. Opin. Struct. Biol. 6 (1996) 18±23. [6] J. Clarke, L.S. Itzhaki, Curr. Opin. Struct. Biol. 8 (1998) 112±118. [7] S.W. Englander, Annu. Rev. Biophys. Biomol. Struct. 29 (2000) 213±238. [8] S.W. Englander, N.R. Kallenbach, Q. Rev. Biophys. 16 (1984) 521±655. [9] M. GueÂron, J.-L. Leroy, Meth. Enzymol. 261 (1995) 383±413. [10] M. Leijon, J.-L. Leroy, Biochimie 79 (1997) 775±779. [11] A. Hvidt, S.O. Neilsen, Adv. Protein Chem. 21 (1966) 287± 386. [12] G. Wagner, Q. Rev. Biophys. 16 (1983) 1±57. [13] S.N. Loh, K.E. Prehoda, J.F. Wang, J.L. Markley, Biochemistry 32 (1993) 11022±11028. [14] K.L. Constantine, M.S. Friedrichs, V. Goldfarb, P.D. Jeffrey, S. Sherrif, L. Mueller, Proteins 15 (1993) 231±290. [15] J.S. Milne, L. Mayne, A.J. Wand, S.W. Englander, Protein Sci. 7 (1998) 739±745. [16] B.B. Kragelund, J. Knudsen, F.M. Poulsen, J. Mol. Biol. 250 (1995) 695±706. [17] A.A. Kossiakoff, Nature 296 (1982) 713±721. [18] A. Wlodawer, L. Sjolin, Proc. Natl. Acad. Sci. USA 79 (1982) 1418±1422. [19] T.G. Pedersen, N.K. Thomsen, K.V. Andersen, J.C. Madsen, F.M. Poulsen, J. Mol. Biol. 230 (1993) 651±660. [20] C. Mori, J.M. Abeygunawardana, P.C.M. Berg, van Zijl, J. Am. Chem. Soc. 119 (1997) 6844±6852. [21] C.E. Dempsey, Biochemistry 25 (1986) 3904±3911. [22] A.J. Wand, H. Roder, S.W. Englander, Biochemistry 25 (1986) 1107±1114. [23] E.M. Goodman, P.S. Kim, Biochemistry 30 (1991) 11615± 11620. [24] M.F. Jeng, H.J. Dyson, Biochemistry 34 (1995) 611±619. [25] C.A. Rohl, R.L. Baldwin, Biochemistry 33 (1994) 7760± 7767. [26] W. Shalongo, L. Dugad, E. Stellwagen, J. Am. Chem. Soc. 116 (1994) 8288±8293. [27] C.E. Dempsey, Biochemistry 27 (1988) 6893±6901.

167

[28] C.E. Dempsey, Biochemistry 31 (1992) 4705±4712. [29] C.E. Dempsey, J. Am. Chem. Soc. 117 (1995) 7526±7534. [30] E. Tuchsen, C. Woodward, J. Mol. Biol. 185 (1985) 405± 419. [31] E. Tuchsen, C. Woodward, J. Mol. Biol. 185 (1985) 421± 430. [32] E. Tuchsen, C. Woodward, J. Mol. Biol. 193 (1987) 793± 802. [33] J.B. Matthew, F.M. Richards, J. Biol. Chem. 258 (1983) 3039±3044. [34] P.S. Kim, R.L. Baldwin, Biochemistry 21 (1982) 1±5. [35] D.O.V. Alonso, V. Daggett, J. Mol. Biol. 247 (1995) 501± 520. [36] M. Christoffersen, S. Bolvig, E. Tuchsen, Biochemistry 3 (1996) 2309±2315. [37] R.B. Gregory, L. Crabo, A.J. Percy, A. Rosenberg, Biochemistry 22 (1983) 910±917. [38] M.P. Crump, J. Crosby, C.E. Dempsey, J.A. Parkinson, M. Murray, D.A. Hopwood, T.J. Simpson, Biochemistry 36 (1997) 6000±6008. [39] R.E. Wedin, M. Delepierre, C.M. Dobson, F.M. Poulsen, Biochemistry 21 (1982) 1098±1113. [40] A. Halsall, C.E. Dempsey, J. Mol. Biol. 293 (1999) 901±915. [41] A. Gargaro, G.B. Bloomberg, C.E. Dempsey, M. Murray, M.J.A. Tanner, Eur. J. Biochem. 221 (1994) 445±454. [42] M.E. Girvin, V.K. Rastogi, F. Abildgaard, J.L. Markley, R.H. Fillingame, Biochemistry 37 (1998) 8817±8824. [43] L. Spyracopoulos, J.D.J. O'Neil, J. Am. Chem. Soc. 116 (1994) 1395±1402. [44] C.L. Perrin, J.-H. Chen, B.K. Ohta, J. Am. Chem. Soc. 121 (1999) 2448±2455. [45] C.A. Bunton, A.K. Yatsimirsky, Langmuir 16 (2000) 5593± 5921. [46] Y.-Z Zhang, Y. Paterson, H. Roder, Protein Sci. 4 (1995) 804±814. [47] J. Wu, D.G. Gorenstein, J. Am. Chem. Soc. 115 (1993) 6843±6850. [48] A. Berger, A. Lowenstein, S. Meiboom, J. Am. Chem. Soc. 81 (1959) 62±67. [49] R.B. Martin, W.C. Hutton, J. Am. Chem. Soc. 95 (1973) 4752±5754. [50] C.L. Perrin, E.R. Johnston, J. Am. Chem. Soc. 103 (1981) 4691±4696. [51] C.L. Perrin, Acc. Chem. Res. 22 (1989) 268±275. [52] Y. Bai, J.S. Milne, L. Mayne, S.W. Englander, Proteins 17 (1993) 75±86. [53] G.P. Connelly, Y. Bai, M.-F. Feng, S.W. Englander, Proteins 17 (1993) 87±92. [54] C.L. Perrin, T.J. Dwyer, J. Rebek, R.J. Duff, J. Am. Chem. Soc. 112 (1990) 3122±3125. [55] G. Otting, Prog. NMR. Spec. 31 (1997) 259±285. [56] M.A.L. Eriksson, T. Hard, L. Nillson, Biophys. J. 69 (1995) 329±339. [57] M. Cotton, R. Fu, T.A. Cross, Biophys. J. 76 (1999) 1179± 1189. [58] L.E. Kay, P. Keifer, T. Saarinem, J. Am. Chem. Soc. 114 (1992) 10663±10665.

168

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

[59] M. Andrec, R.B. Hill, J.H. Prestegard, Protein Sci. 4 (1995) 983±993. [60] D. Marion, M. Ikura, R. Tschudin, A. Bax, J. Magn. Reson. 85 (1989) 393±399. [61] D. Morikis, P.E. Wright, Eur. J. Biochem. 237 (1996) 212± 220. [62] C.E. Dempsey, J. Biomol. NMR 4 (1994) 879±884. [63] A.K. Bhuyan, J.B. Udgaonkar, Proteins: Struct. Func. Genet. 30 (1998) 295±308. [64] R. Moss, H. Gesmar, J.J. Led, J. Am. Chem. Soc. 116 (1994) 747±749. [65] A.M.M. Jorgensen, H.B. Olsen, P. Balschmidt, J.J. Led, J. Mol. Biol. 257 (1996) 684±699. [66] J. WoÂjcik, K. Russzczynska, J. Zhukov, A. Ejchart, Acta Biochim. Polonica 46 (1999) 651±663. [67] L.E. Kay, Prog. Biophys. Mol. Biol. 63 (1995) 277±299. [68] S. ForseÂn, R.A. Hoffman, J. Chem. Phys. 39 (1963) 2892± 2901. [69] A.G. Red®ed, S. Waelder, J. Am. Chem. Soc. 101 (1979) 6151±6162. [70] N. Rama Krishna, D.H. Huang, J.D. Glickson, R. Rowan, R. Walter, Biophys. J. 26 (1979) 345±366. [71] S. Spera, M. Ikura, A. Bax, J. Biomol. NMR 1 (1991) 155± 165. [72] S.A. Morris, R. Freeman, J. Am. Chem. Soc. 101 (1979) 760±762. [73] G.D. Henry, B.D. Sykes, Biochemistry 29 (1990) 6303± 6313. [74] G.D. Henry, B.D. Sykes, J. Magn. Reson. B 102 (1993) 193± 200. [75] Z. Zheng, M.R. Gryk, M.D. Finucane, O. Jardetzky, J. Magn. Reson. B 108 (1995) 220±234. [76] H. Roder, G. Wagner, K. Wuthrich, Biochemistry 24 (1985) 7396±7407. [77] C.B. Arrington, A.D. Robertson, Biochemistry 36 (1997) 8686±8691. [78] B.B. Kragelund, B. Heinemann, J. Knudsen, F.M. Poulsen, Protein Sci. 7 (1998) 2237±2248. [79] C.B. Arrington, A.D. Robinson, J. Mol. Biol. 296 (2000) 1307±1317. [80] G. Hernandez, F.E. Jenney, M.W.W. Adams, D.M. LeMaster, Proc. Natl. Acad. Sci. USA 9 (2000) 3166±3170. [81] M.R. Gryk, M.D. Finucane, Z. Zheng, O. Jardetzky, J. Mol. Biol. 246 (1995) 618±627. [82] M.D. Finucane, O. Jardetzky, Mol. Phys. 95 (1998) 1127± 1136. [83] H.M. McConnell, J. Chem. Phys. 28 (1958) 430±431. [84] G. Wagner, A. Pardi, K. Wuthrich, J. Am. Chem. Soc. 105 (1983) 5948±5949. [85] R.W. Kriwacki, R.B. Hill, J.M. Flanagan, J.P. Cardonna, J.H. Prestegard, J. Am. Chem. Soc. 115 (1993) 8907±8911. [86] S. Grzesiek, A. Bax, J. Biomol. NMR 3 (1993) 627±638. [87] J. Stonehouse, G.L. Shaw, J. Keeler, E.D. Laue, J. Magn. Reson. A 107 (1994) 178±184. [88] W. Jahnke, H. Kessler, J. Biomol. NMR 4 (1994) 735±740. [89] G. Gemmecker, W. Hanke, H. Kessler, J. Am. Chem. Soc. 115 (1993) 11620±11621.

[90] S. Koide, W. Jahnke, P.E. Wright, J. Biomol. NMR 6 (1995) 306±312. [91] N. Bloembergen, R.V. Pound, Phys. Rev. 95 (1954) 8±12. [92] A. Bockmann, F. Penin, E. Guittet, FEBS Lett. 383 (1996) 191±195. [93] G. Otting, E. Liepinsh, J. Biomol. NMR 5 (1995) 420±426. [94] S. Mori, M.O. Johnson, J.M. Berg, P.C.M. van Zijl, J. Am. Chem. Soc. 116 (1994) 11982±11984. [95] S. Mori, C. Abeygunawardana, M.O. Johnson, J. Berg, P.C.M. Van Zijl, J. Magn. Reson. B. 108 (1995) 94±98. [96] S. Mori, J.M. Berg, P.C.M. van Zijl, J. Biomol. NMR 7 (1996) 77±82. [97] M. Piotto, V. Saudek, V. Sklenar, J. Biomol. NMR 2 (1992) 661±665. [98] T.-L. Hwang, S. Mori, A.J. Shaka, P.C.M. van Zijl, J. Am. Chem. Soc. 119 (1997) 6203±6204. [99] T.-L. Hwang, P.C.M. van Zijl, S. Mori, J. Biomol. NMR 11 (1998) 221±226. [100] K. Zannger, I.M. Armitage, J. Magn. Reson. 135 (1998) 70± 75. [101] N.R. Skrynnikov, R.R. Ernst, J. Magn. Reson. 137 (1999) 276±280. [102] V.V. Krishnan, M. Cosman, Magn. Reson. Chem. 38 (2000) 789±794. [103] J. Jeener, B.H. Meier, P. Bachmann, R.R. Ernst, J. Chem. Phys. 71 (1979) 4546±4553. [104] V.V. Krishnan, M. Sukumar, L.M. Gierasch, M. Cosman, Biochemistry 39 (2000) 9119±9129. [105] B.R. Lee¯ang, J.F.G. Vliegenthart, J. Magn. Reson. 89 (1990) 615±6119. [106] B. Adams, L. Lerner, J. Magn. Reson. 96 (1992) 604±607. [107] E.R. Johnston, A.B. Little, J. Magn. Reson. A 114 (1995) 113±115. [108] E.O. Stejskal, J.E. Tanner, J. Chem. Phys. 42 (1965) 288± 292. [109] M. Andrec, J.H. Prestegard, J. Biomol. NMR 9 (1997) 136± 150. [110] A. Bockmann, E. Guittet, FEBS Lett. 418 (1997) 127±130. [111] M. Liu, H.C. Toms, G.E. Hawkes, J.M. Nicholson, J.C. Lindon, J. Biomol. NMR 13 (2000) 25±30. [112] C.T.W. Moonen, P. van Gelderen, G.W. Vuister, P.C.M. van Zijl, J. Magn. Reson. 97 (1992) 419±425. [113] C.M. Dobson, L.-Y. Lian, C. Red®eld, K.D. Topping, J. Magn. Reson. 69 (1986) 201±209. [114] E.L. Hahn, Phys. Rev. 80 (1950) 580±594. [115] C. Bracken, J. Baum, J. Am. Chem. Soc. 115 (1993) 6346± 6348. [116] H.B. Olsen, H. Gesmar, J.J. Led, J. Am. Chem. Soc. 115 (1993) 1456±1460. [117] J. Feeney, P. Partington, G.C.K. Roberts, J. Magn. Reson. 13 (1974) 268±274. [118] G.D. Henry, J.H. Weiner, B.D. Sykes, Biochemistry 26 (1987) 3626±3634. [119] A. Miranker, C.V. Robinson, S.E. Radford, R.T. Aplin, C.M. Dobson, Science 262 (1993) 896±900. [120] G. Wagner, Biochem. Biophys. Res. Commun. 97 (1980) 614±620.

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170 [121] N.K. Thomsen, F.M. Poulsen, J. Mol. Biol. 234 (1993) 234± 241. [122] B.B. Kragelund, P. Hùjrup, M. Jensen, C. Schjerling, E. Juul, J. Knudsen, F.M. Poulsen, J. Mol. Biol. 256 (1996) 187±200. [123] D.J. Segal, A. Bachmann, J. Hofrichter, K.O. Hodgson, S. Doniach, T. Kiefhaber, J. Mol. Biol. 288 (1999) 489±499. [124] A. Matagne, M. Jamin, E.W. Chung, C.V. Robinson, S.E. Radford, C.M. Dobson, J. Mol. Biol. 297 (2000) 193±210. [125] L. Swint-Kruse, A.D. Robertson, Biochemistry 35 (1996) 171±180. [126] S. Perrett, J. Clarke, A.M. Hounslow, A.R. Fersht, Biochemistry 34 (1995) 9288±9298. [127] M.J. Parker, C.E. Dempsey, L.L.P. Hosszu, J.P. Waltho, A.R. Clarke, Nat. Struct. Biol. 5 (1998) 194±198. [128] M.J. Parker, C.E. Dempsey, M. Lorch, A.R. Clarke, Biochemistry 36 (1997) 13396±13405. [129] K.W. Plaxco, K.T. Simons, I. Ruczinski, D. Baker, Biochemistry 39 (2000) 11177±11183. [130] P.A. Dalby, J. Clarke, C.M. Johnson, A.R. Fersht, J. Mol. Biol. 276 (1998) 647±656. [131] R. Hiller, Z.H. Zhou, M.W.W. Adams, S.W. Englander, Proc. Natl. Acad. Sci. USA 94 (1997) 11329±11332. [132] S.L. Mayo, R.L. Baldwin, Science 262 (1993) 873±876. [133] Y. Bai, T.R. Sosnick, L. Mayne, S.W. Englander, Science 269 (1995) 192±197. [134] T. Kiefhaber, R.L. Baldwin, Proc. Natl. Acad. Sci. USA 92 (1996) 2657±2661. [135] H. Roder, G. Wagner, K. Wuthrich, Biochemistry 24 (1985) 7407±7411. [136] M.F. Jeng, S.W. Englander, G.A. Elove, A.J. Wand, H. Roder, Biochemistry 29 (1990) 10433±10437. [137] A.D. Robertson, R.L. Baldwin, Biochemistry 30 (1991) 9907±9914. [138] M. Buck, S.E. Radford, C.M. Dobson, J. Mol. Biol. 237 (1994) 247±254. [139] C.K. Woodward, L. Ellis, A. Rosenberg, J. Biol. Chem. 250 (1975) 440±444. [140] J.L. Neira, P. Sevilla, M. Menendez, M. Bruix, M. Rico, J. Mol. Biol. 285 (1999) 627±643. [141] Y. Bai, J.S. Milne, L. Mayne, S.W. Englander, Proteins 20 (1994) 4±14. [142] Q. Yi, M.L. Scalley, K.T. Simons, S. Gladwin, D. Baker, Fold. Des. 2 (1997) 271±280. [143] L.S. Itzhaki, J.L. Neira, A.R. Fersht, J. Mol. Biol. 270 (1997) 88±89. [144] A.T. Alexandrescu, V.A. Jaravine, S.A. Dames, F.P. Lamour, J. Mol. Biol. 289 (1999) 1041±1054. [145] B.M.P. Huyghues-Despointes, J.M. Scholtz, C.N. Pace, Nat. Struct. Biol. 6 (1999) 910±912. [146] M. Sadqi, S. Casares, M.A. Abril, O. LoÂpez-Mayorga, F. Conejero-Lara, E. Freire, Biochemistry 38 (1999) 8899± 8906. [147] C.K. Woodward, B.D. Hilton, Biophys. J. 32 (1980) 561±575. [148] R. Richarz, P. Sehr, G. Wagner, K. Wuthrich, J. Mol. Biol. 130 (1979) 19±30. [149] B.D. Hilton, C.K. Woodward, Biochemistry 18 (1979) 5834±5841.

169

[150] T.K. Hitchins, R.G. Bryant, Biochemistry 37 (1998) 5878± 5887. [151] G. Wagner, Q. Rev. Biophys. 16 (1983) 1±57. [152] J. Clarke, A.M. Hounslow, M. Bycroft, A.R. Fersht, Proc. Natl. Acad. Sci. USA 90 (1993) 9837±9841. [153] K.-S. Kim, J.A. Fuchs, C.K. Woodward, Biochemistry 32 (1993) 9600±9608. [154] J.L. Neira, L.S. Itzhaki, D.E. Otzen, B. Davis, A.R. Fersht, J. Mol. Biol. 270 (1997) 99±110. [155] S.K. Jandu, S. Ray, L. Brooks, R.J. Leatherbarrow, Biochemistry 29 (1990) 6264±6269. [156] C. Woodward, I. Simon, E. Tuchsen, Mol. Cell. Biochem. 48 (1982) 135±160. [157] D.W. Miller, K. Dill, Protein Sci. 4 (1995) 1860±1873. [158] K.-S. Kim, C.K. Woodward, Biochemistry 32 (1993) 9609± 9613. [159] J.K. Myers, C.N. Pace, J.M. Scholtz, Protein Sci. 4 (1995) 2138±2148. [160] J.S. Milne, Y. Xu, L.C. Mayne, S.W. Englander, J. Mol. Biol. 290 (1999) 811±822. [161] H. Qian, S.L. Mayo, A. Morton, Biochemistry 33 (1994) 8167±81171. [162] Y. Xu, L. Mayne, S.W. Englander, Nat. Struct. Biol. 5 (1998) 774±779. [163] A.K. Chamberlain, T.M. Handel, S. Marqusee, Nat. Struct. Biol. 3 (1996) 782±787. [164] E.J. Fuentes, A.J. Wand, Biochemistry 37 (1998) 3687± 3698. [165] E.J. Fuentes, A.J. Wand, Biochemistry 37 (1998) 9877± 9883. [166] M. LlinaÂs, B. Gillespie, F.W. Dahlquist, S. Marqusee, Nat. Struct. Biol. 6 (1999) 1072±1078. [167] Y. Bai, S.W. Englander, Proteins 24 (1996) 145±151. [168] C. Woodward, Trends Biochem. Sci. 18 (1993) 359±360. [169] R. Li, C. Woodward, Protein Sci. 8 (1999) 1571±1591. [170] A.K. Chamberlain, S. Marqusee, Structure 5 (1997) 859±863. [171] M.J. Parker, S. Marqusee, J. Mol. Biol. 305 (2001) 593±602. [172] J. Clarke, L.S. Itzhaki, A.R. Fersht, Trends Biochem. Sci. 22 (1997) 284±287. [173] S.N. Loh, C.A. Rohl, T. Kiefhaber, R.L. Baldwin, Proc. Natl. Acad. Sci. USA 93 (1996) 1982±1987. [174] V.J. Hilser, E. Friere, J. Mol. Biol. 262 (1996) 756±772. [175] I. Bahar, A. Wallqvist, D.G. Covell, R.L. Jernigan, Biochemistry 37 (1998) 1067±1075. [176] J.O. Wooll, J.O. Wrabl, V.J. Hilser, J. Mol. Biol. 301 (2000) 247±256. [177] A.E. Garcia, G. Hummer, Proteins: Struct. Funct. Genet. 36 (1999) 175±191. [178] S. Williams, T.P. Causgrove, R. Gilmanshin, K.S. Fang, R.H. Callender, W.H. Woodruff, R.B. Dyer, Biochemistry 35 (1996) 691±697. [179] P.A. Thompson, W.A. Eaton, J. Hofrichter, Biochemistry 36 (1997) 9200±9210. [180] N. Gibbs, R.B. Sessions, P.B. Williams, C.E. Dempsey, Biophys. J. 72 (1997) 2490±2495. [181] R.B. Sessions, N. Gibbs, C.E. Dempsey, Biophys. J. 74 (1998) 138±152.

170

C.E. Dempsey / Progress in Nuclear Magnetic Resonance Spectroscopy 39 (2001) 135±170

[182] W.A. Shirley, C.L. Brooks, Proteins: Struct. Func. Genet. 28 (1997) 59±71. [183] A. Pastore, T.S. Harvey, C.E. Dempsey, I.D. Campbell, Eur. Biophys. J. 16 (1989) 363±367. [184] Y. Pan, M.S. Briggs, Biochemistry 31 (1992) 11405±11412. [185] J.I. Guijarro, M. Jackson, A.F. Chaffotte, M. Delepierre, H.H. Mantsch, M.E. Goldberg, Biochemistry 34 (1995) 2998±3008. [186] P.A. Jennings, P.E. Wright, Science 262 (1993) 892±896. [187] A.J. Li, V. Daggett, Protein Engng 8 (1995) 1117±1128. [188] S.L. Kazmirshi, V. Daggett, J. Mol. Biol. 284 (1998) 793±806. [189] Y. Duan, P.A. Kollman, Science 282 (1998) 740±744. [190] V.A. Daragan, K.H. Mayo, Prog. NMR Spec. 31 (1997) 63± 105. [191] J. Kordel, N.J. Skelton, M. Akke, A.G. Palmer, W.J. Chazin, Biochemistry 31 (1992) 4856±4866. [192] A.M. Mandel, M. Akke, A.G. Palmer, J. Mol. Biol. 246 (1995) 144±163. [193] M.F. Jeng, H.J. Dyson, Biochemistry 34 (1995) 611±619. [194] J. Habazettl, L.C. Myers, F. Yuan, G.L. Verdine, G. Wagner, Biochemistry 35 (1996) 9335±9348. [195] N.W. Downer, T.J. Bruchman, J.H. Hazzard, J. Biol. Chem. 261 (1986) 3640±3647. [196] J.J. Englander, N.W. Downer, S.W. Englander, J. Biol. Chem. 257 (1982) 7982±7986. [197] A. Yee, B. Szymczyna, J.D. O'Neil, Biochemistry 38 (1999) 6489±6498.

[198] C.E. Dempsey, G.S. Butler, Biochemistry 31 (1992) 11973± 11977. [199] C.E. Dempsey, L.J. Handcock, Biophys. J. 70 (1996) 1777± 1788. [200] I.T. Arkin, K.R. McKenzie, L. Fisher, S. Aimoto, D.M. Engelman, S.O. Smith, Nat. Struct. Biol. 3 (1996) 240±243. [201] L. Czershi, O. Vinogradova, C.R. Sanders, J. Magn. Reson. 142 (2000) 111±119. [202] E. Goormaghtigh, V. Raussens, J.M. Ruysschaert, Biochim. Biophys. Acta 1422 (1999) 105±185. [203] R.S. Prosser, S.I. Daleman, J.H. Davis, Biophys. J. 66 (1994) 1415±1428. [204] J.A.A. Demmers, J. Haverkamp, A.J.R. Heck, R.E. Koeppe, J.A. Killian, Proc. Natl. Acad. Sci. USA 97 (2000) 3189± 3194. [205] S.J. Opella, Nat. Struct. Biol. (NMR Suppl.) 4 (1997) 845± 848. [206] J.H. Davis, M. Auger, Prog. NMR Spectr. 35 (1999) 1±84. [207] S. Huo, S. Arumugam, T.A. Cross, Solid State NMR 7 (1996) 177±183. [208] G.D. Henry, B.D. Sykes, J. Biomol. NMR 5 (1995) 59±66. [209] N.R. Krishna, K.P. Sarathy, D.-H. Huang, R.L. Stephens, J.D. Glickson, C.W. Smith, R. Walter, J. Am. Chem. Soc. 104 (1982) 5051±5053. [210] S. Waelder, A.S. Red®eld, Biopolymers 16 (1977) 623±631.