Peptide conformer acidity analysis of protein flexibility monitored by hydrogen exchange

Transient conformations in the unliganded FK506 binding domain of FKBP51 correspond to two distinct inhibitor-bound states

Members of the FK506-binding protein (FKBP) family regulate a range of important physiological processes. Unfortunately, current therapeutics such as FK506 and rapamycin exhibit only modest selectivity among these functionally distinct proteins. Recent progress in developing selective inhibitors has been reported for FKBP51 and FKBP52, which act as mutual antagonists in the regulation of steroid hormone signaling. Two structurally similar inhibitors yield distinct protein conformations at the binding site. Localized conformational transition in the binding site of the unliganded FK1 domain of FKBP51 is suppressed by a K58T mutation that also suppresses the binding of these inhibitors. Here, it is shown that the changes in amide hydrogen exchange kinetics arising from this K58T substitution are largely localized to this structural region. Accurate determination of the hydroxide-catalyzed exchange rate constants in both the wildtype and K58T variant proteins impose strong constraints upon the pattern of amide exchange reactivities within either a single or a pair of transient conformations that could give rise to the differences between these two sets of measured rate constants. Poisson-Boltzmann continuum dielectric calculations provide moderately accurate predictions of the structure-dependent hydrogen exchange reactivity for solvent-exposed protein backbone amides. Applying such calculations to the local protein conformations observed in the two inhibitor-bound FKBP51 domains demonstrated that the experimentally determined exchange rate constants for the wildtype domain are robustly predicted by a population-weighted sum of the experimental hydrogen exchange reactivity of the K58T variant and the predicted exchange reactivities in model conformations derived from the two inhibitor-bound protein structures.

Computational Modeling of Molecular Structures Guided by Hydrogen-Exchange Data

Data produced by hydrogen-exchange monitoring experiments have been used in structural studies of molecules for several decades. Despite uncertainties about the structural determinants of hydrogen exchange itself, such data have successfully helped guide the structural modeling of challenging molecular systems, such as membrane proteins or large macromolecular complexes. As hydrogen-exchange monitoring provides information on the dynamics of molecules in solution, it can complement other experimental techniques in so-called integrative modeling approaches. However, hydrogen-exchange data have often only been used to qualitatively assess molecular structures produced by computational modeling tools. In this paper, we look beyond qualitative approaches and survey the various paradigms under which hydrogen-exchange data have been used to quantitatively guide the computational modeling of molecular structures. Although numerous prediction models have been proposed to link molecular structure and hydrogen exchange, none of them has been widely accepted by the structural biology community. Here, we present as many hydrogen-exchange prediction models as we could find in the literature, with the aim of providing the first exhaustive list of its kind. From purely structure-based models to so-called fractional-population models or knowledge-based models, the field is quite vast. We aspire for this paper to become a resource for practitioners to gain a broader perspective on the field and guide research toward the definition of better prediction models. This will eventually improve synergies between hydrogen-exchange monitoring and molecular modeling.

Assessing Site-Specific Enhancements Imparted by Hyperpolarized Water in Folded and Unfolded Proteins by 2D HMQC NMR

Hyperpolarized water can be a valuable aid in protein NMR, leading to amide group ¹H polarizations that are orders of magnitude larger than their thermal counterparts. Suitable procedures can exploit this to deliver 2D ¹H–¹⁵N correlations with good resolution and enhanced sensitivity. These enhancements depend on the exchange rates between the amides and the water, thereby yielding diagnostic information about solvent accessibility. This study applied this “HyperW” method to four proteins exhibiting a gamut of exchange behaviors: PhoA^(350–471), an unfolded 122-residue fragment; barstar, a fully folded ribonuclease inhibitor; R17, a 13.3 kDa system possessing folded and unfolded forms under slow interconversion; and drkN SH3, a protein domain whose folded and unfolded forms interchange rapidly and with temperature-dependent population ratios. For PhoA4^(350–471) HyperW sensitivity enhancements were ≥300×, as expected for an unfolded protein sequence. Though fully folded, barstar also exhibited substantial enhancements; these, however, were not uniform and, according to CLEANEX experiments, reflected the solvent-exposed residues. R17 showed the expected superposition of ≥100-fold enhancements for its unfolded form, coexisting with more modest enhancements for their folded counterparts. Unexpected, however, was the behavior of drkN SH3, for which HyperW enhanced the unfolded but, surprisingly, enhanced even more certain folded protein sites. These preferential enhancements were repeatedly and reproducibly observed. A number of explanations—including three-site exchange magnetization transfers between water and the unfolded and folded states; cross-correlated relaxation processes from hyperpolarized “structural” waters and labile side-chain protons; and the possibility that faster solvent exchange rates characterize certain folded sites over their unfolded counterparts—are considered to account for them.

Coarse-Grained Conformational Sampling of Protein Structure Improves the Fit to Experimental Hydrogen-Exchange Data

Monitoring hydrogen/deuterium exchange (HDX) undergone by a protein in solution produces experimental data that translates into valuable information about the protein's structure. Data produced by HDX experiments is often interpreted using a crystal structure of the protein, when available. However, it has been shown that the correspondence between experimental HDX data and crystal structures is often not satisfactory. This creates difficulties when trying to perform a structural analysis of the HDX data. In this paper, we evaluate several strategies to obtain a conformation providing a good fit to the experimental HDX data, which is a premise of an accurate structural analysis. We show that performing molecular dynamics simulations can be inadequate to obtain such conformations, and we propose a novel methodology involving a coarse-grained conformational sampling approach instead. By extensively exploring the intrinsic flexibility of a protein with this approach, we produce a conformational ensemble from which we extract a single conformation providing a good fit to the experimental HDX data. We successfully demonstrate the applicability of our method to four small and medium-sized proteins.

Ensembles of a small number of conformations with relative populations

In our previous work, we proposed a new way to represent protein native states, using ensembles of a small number of conformations with relative Populations, or ESP in short. Using Ubiquitin as an example, we showed that using a small number of conformations could greatly reduce the potential of overfitting and assigning relative populations to protein ensembles could significantly improve their quality. To demonstrate that ESP indeed is an excellent alternative to represent protein native states, in this work we compare the quality of two ESP ensembles of Ubiquitin with several well-known regular ensembles or average structure representations. Extensive amount of significant experimental data are employed to achieve a thorough assessment. Our results demonstrate that ESP ensembles, though much smaller in size comparing to regular ensembles, perform equally or even better sometimes in all four different types of experimental data used in the assessment, namely, the residual dipolar couplings, residual chemical shift anisotropy, hydrogen exchange rates, and solution scattering profiles. This work further underlines the significance of having relative populations in describing the native states.

Analysing the visible conformational substates of the FK506-binding protein FKBP12

The ¹H-¹⁵N 2D NMR correlation spectrum of the widely studied FK506-binding protein FKBP12 (FK506-binding protein of 12 kDa) contains previously unreported peak doublings for at least 31 residues that arise from a minor conformational state (12% of total) which exchanges with the major conformation with a time constant of 3.0 s at 43°C. The largest differences in chemical shift occur for the 80′s loop that forms critical recognition interactions with many of the protein partners for the FKBP family. The residues exhibiting doubling extend into the adjacent strands of the β-sheet, across the active site to the α-helix and into the 50′s loop. Each of the seven proline residues adopts a trans-peptide linkage in both the major and minor conformations, indicating that this slow transition is not the result of prolyl isomerization. Many of the residues exhibiting resonance doubling also participate in conformational line-broadening transition(s) that occur ~10⁵-fold more rapidly, proposed previously to arise from a single global process. The 1.70 Å (1 Å=0.1 nm) resolution X-ray structure of the H87V variant is strikingly similar to that of FKBP12, yet this substitution quenches the slow conformational transition throughout the protein while quenching the line-broadening transition for residues near the 80′s loop. Line-broadening was also decreased for the residues in the α-helix and 50′s loop, whereas line-broadening in the 40′s loop was unaffected. The K44V mutation selectively reduces the line-broadening in the 40′s loop, verifying that at least three distinct conformational transitions underlie the line-broadening processes of FKBP12.

Analysing the visible conformational substates of the FK506-binding protein FKBP12

The 1H-15N 2D NMR correlation spectrum of the widely studied FK506-binding protein FKBP12 (FK506-binding protein of 12 kDa) contains previously unreported peak doublings for at least 31 residues that arise from a minor conformational state (12% of total) which exchanges with the major conformation with a time constant of 3.0 s at 43°C. The largest differences in chemical shift occur for the 80's loop that forms critical recognition interactions with many of the protein partners for the FKBP family. The residues exhibiting doubling extend into the adjacent strands of the β-sheet, across the active site to the α-helix and into the 50's loop. Each of the seven proline residues adopts a trans-peptide linkage in both the major and minor conformations, indicating that this slow transition is not the result of prolyl isomerization. Many of the residues exhibiting resonance doubling also participate in conformational line-broadening transition(s) that occur ~105-fold more rapidly, proposed previously to arise from a single global process. The 1.70 Å (1 Å=0.1 nm) resolution X-ray structure of the H87V variant is strikingly similar to that of FKBP12, yet this substitution quenches the slow conformational transition throughout the protein while quenching the line-broadening transition for residues near the 80's loop. Line-broadening was also decreased for the residues in the α-helix and 50's loop, whereas line-broadening in the 40's loop was unaffected. The K44V mutation selectively reduces the line-broadening in the 40's loop, verifying that at least three distinct conformational transitions underlie the line-broadening processes of FKBP12.

Assessing the chemical accuracy of protein structures via peptide acidity

Although the protein native state is a Boltzmann conformational ensemble, practical applications often require a representative model from the most populated region of that distribution. The acidity of the backbone amides, as reflected in hydrogen exchange rates, is exquisitely sensitive to the surrounding charge and dielectric volume distribution. For each of four proteins, three independently determined X-ray structures of differing crystallographic resolution were used to predict exchange for the static solvent-exposed amide hydrogens. The average correlation coefficients range from 0.74 for ubiquitin to 0.93 for Pyrococcus furiosus rubredoxin, reflecting the larger range of experimental exchange rates exhibited by the latter protein. The exchange prediction errors modestly correlate with the crystallographic resolution. MODELLER 9v6-derived homology models at ~60% sequence identity (36% identity for chymotrypsin inhibitor CI2) yielded correlation coefficients that are ~0.1 smaller than for the cognate X-ray structures. The most recently deposited NOE-based ubiquitin structure and the original NMR structure of CI2 fail to provide statistically significant predictions of hydrogen exchange. However, the more recent RECOORD refinement study of CI2 yielded predictions comparable to the X-ray and homology model-based analyses.

Biological insights from hydrogen exchange mass spectrometry

Over the past two decades, hydrogen exchange mass spectrometry (HXMS) has achieved the status of a widespread and routine approach in the structural biology toolbox. The ability of hydrogen exchange to detect a range of protein dynamics coupled with the accessibility of mass spectrometry to mixtures and large complexes at low concentrations result in an unmatched tool for investigating proteins challenging to many other structural techniques. Recent advances in methodology and data analysis are helping HXMS deliver on its potential to uncover the connection between conformation, dynamics and the biological function of proteins and complexes. This review provides a brief overview of the HXMS method and focuses on four recent reports to highlight applications that monitor structure and dynamics of proteins and complexes, track protein folding, and map the thermodynamics and kinetics of protein unfolding at equilibrium. These case studies illustrate typical data, analysis and results for each application and demonstrate a range of biological systems for which the interpretation of HXMS in terms of structure and conformational parameters provides unique insights into function. This article is part of a Special Issue entitled: Mass spectrometry in structural biology.

Protein hydrogen exchange: testing current models

To investigate the determinants of protein hydrogen exchange (HX), HX rates of most of the backbone amide hydrogens of Staphylococcal nuclease were measured by NMR methods. A modified analysis was used to improve accuracy for the faster hydrogens. HX rates of both near surface and well buried hydrogens are spread over more than 7 orders of magnitude. These results were compared with previous hypotheses for HX rate determination. Contrary to a common assumption, proximity to the surface of the native protein does not usually produce fast exchange. The slow HX rates for unprotected surface hydrogens are not well explained by local electrostatic field. The ability of buried hydrogens to exchange is not explained by a solvent penetration mechanism. The exchange rates of structurally protected hydrogens are not well predicted by algorithms that depend only on local interactions or only on transient unfolding reactions. These observations identify some of the present difficulties of HX rate prediction and suggest the need for returning to a detailed hydrogen by hydrogen analysis to examine the bases of structure-rate relationships, as described in the companion paper (Skinner et al., Protein Sci 2012;21:996-1005).

Protein dynamics viewed by hydrogen exchange

To examine the relationship between protein structural dynamics and measurable hydrogen exchange (HX) data, the detailed exchange behavior of most of the backbone amide hydrogens of Staphylococcal nuclease was compared with that of their neighbors, with their structural environment, and with other information. Results show that H-bonded hydrogens are protected from exchange, with HX rate effectively zero, even when they are directly adjacent to solvent. The transition to exchange competence requires a dynamic structural excursion that removes H-bond protection and allows exposure to solvent HX catalyst. The detailed data often make clear the nature of the dynamic excursion required. These range from whole molecule unfolding, through smaller cooperative unfolding reactions of secondary structural elements, and down to local fluctuations that involve as little as a single peptide group or side chain or water molecule. The particular motion that dominates the exchange of any hydrogen is the one that allows the fastest HX rate. The motion and the rate it produces are determined by surrounding structure and not by nearness to solvent or the strength of the protecting H-bond itself or its acceptor type (main chain, side chain, structurally bound water). Many of these motions occur over time scales that are appropriate for biochemical function.

Experimentally assessing molecular dynamics sampling of the protein native state conformational distribution

The acute sensitivity to conformation exhibited by amide hydrogen exchange reactivity provides a valuable test for the physical accuracy of model ensembles developed to represent the Boltzmann distribution of the protein native state. A number of molecular dynamics studies of ubiquitin have predicted a well-populated transition in the tight turn immediately preceding the primary site of proteasome-directed polyubiquitylation Lys 48. Amide exchange reactivity analysis demonstrates that this transition is 10(3)-fold rarer than these predictions. More strikingly, for the most populated novel conformational basin predicted from a recent 1 ms MD simulation of bovine pancreatic trypsin inhibitor (at 13% of total), experimental hydrogen exchange data indicates a population below 10(-6). The most sophisticated efforts to directly incorporate experimental constraints into the derivation of model protein ensembles have been applied to ubiquitin, as illustrated by three recently deposited studies (PDB codes 2NR2, 2K39 and 2KOX2K392KOX). Utilizing the extensive set of experimental NOE constraints, each of these three ensembles yields a modestly more accurate prediction of the exchange rates for the highly exposed amides than does a standard unconstrained molecular simulation. However, for the less frequently exposed amide hydrogens, the 2NR2 ensemble offers no improvement in rate predictions as compared to the unconstrained MD ensemble. The other two NMR-constrained ensembles performed markedly worse, either underestimating (2KOX) or overestimating (2K39) the extent of conformational diversity.

Electrostatics of hydrogen exchange for analyzing protein flexibility

Electrostatic interactions at the protein-aqueous interface modulate the reactivity of solvent-exposed backbone amides by a factor of at least a billion fold. The brief (∼10 ps) lifetime of the peptide anion formed during the hydroxide-catalyzed exchange reaction helps enable the experimental rates to be robustly predictable by continuum dielectric methods. Since this ability to predict the structural dependence of exchange reactivity also applies to the protein amide hydrogens that are only rarely exposed to the bulk solvent phase, electrostatic analysis of the experimental exchange rates provides an effective assessment of whether a given model ensemble is consistent with the properly weighted Boltzmann conformational distribution of the protein native state.

Quantitative assessment of protein structural models by comparison of H/D exchange MS data with exchange behavior accurately predicted by DXCOREX

Peptide amide hydrogen/deuterium exchange mass spectrometry (DXMS) data are often used to qualitatively support models for protein structure. We have developed and validated a method (DXCOREX) by which exchange data can be used to quantitatively assess the accuracy of three-dimensional (3-D) models of protein structure. The method utilizes the COREX algorithm to predict a protein's amide hydrogen exchange rates by reference to a hypothesized structure, and these values are used to generate a virtual data set (deuteron incorporation per peptide) that can be quantitatively compared with the deuteration level of the peptide probes measured by hydrogen exchange experimentation. The accuracy of DXCOREX was established in studies performed with 13 proteins for which both high-resolution structures and experimental data were available. The DXCOREX-calculated and experimental data for each protein was highly correlated. We then employed correlation analysis of DXCOREX-calculated versus DXMS experimental data to assess the accuracy of a recently proposed structural model for the catalytic domain of a Ca(2+)-independent phospholipase A(2). The model's calculated exchange behavior was highly correlated with the experimental exchange results available for the protein, supporting the accuracy of the proposed model. This method of analysis will substantially increase the precision with which experimental hydrogen exchange data can help decipher challenging questions regarding protein structure and dynamics.

Strategies for the thermodynamic characterization of linked binding/local folding reactions within the native state application to the LID domain of adenylate kinase from Escherichia coli

Conformational fluctuations in proteins have emerged as an important aspect of biological function, having been linked to processes ranging from molecular recognition and catalysis to allostery and signal transduction. In spite of the realization of their importance, however, the connections between fluctuations and function have largely been empirical, even when they have been quantitative. Part of the problem in understanding the role of fluctuations in function is the fact that the mere existence of fluctuations complicates the interpretation of classic mutagenesis approaches. Namely, mutagenesis, which is typically targeted to an internal position (to elicit an effect), will change the fluctuations as well as the structure of the native state. Decoupling these effects is essential to an unambiguous understanding of the role of fluctuations in function. Here, we use a mutation strategy that targets surface-exposed sites in flexible parts of the molecule for mutation to glycine. Such mutations leave the ground-state structure unaffected. As a result, we can assess the nature of the fluctuations, develop a quantitative model relating fluctuations to function (in this case, molecular recognition), and unambiguously resolve the probabilities of the fluctuating states. We show that when this approach is applied to Escherichia coli adenylate kinase (AK), unique thermodynamic and structural insights are obtained, even when classic mutagenesis approaches targeted to the same region yield ambiguous results.

Assessing the native state conformational distribution of ubiquitin by peptide acidity

At equilibrium, every energetically feasible conformation of a protein occurs with a non-zero probability. Quantitative analysis of protein flexibility is thus synonymous with determining the proper Boltzmann-weighting of this conformational distribution. The exchange reactivity of solvent-exposed amide hydrogens greatly varies with conformation, while the short-lived peptide anion intermediate implies an insensitivity to the dynamics of conformational motion. Amides that are well-exposed in model conformational ensembles of ubiquitin vary a million-fold in exchange rates which continuum dielectric methods can predict with an rmsd of 3. However, the exchange rates for many of the more rarely exposed amides are markedly overestimated in the PDB-deposited 2K39 and 2KN5 ubiquitin ensembles, while the 2NR2 ensemble predictions are largely consistent with those of the Boltzmann-weighted conformational distribution sampled at the level of 1%. The correlation between the fraction of solvent-accessible conformations for a given amide hydrogen and the exchange rate constant for that residue provides a useful monitor of the degree of completeness with which a given ensemble has sampled the energetically accessible conformational space. These exchange predictions correlate with the degree to which each ensemble deviates from a set of 46 ubiquitin X-ray structures. Kolmogorov-Smirnov analysis for the distribution of intra- and inter-ensemble pairwise structural rmsd values assisted the identification of a subensemble of 2K39 that eliminates the overestimations of hydrogen exchange rates observed for the full ensemble. The relative merits of incorporating experimental restraints into the conformational sampling process are compared to using these restraints as filters to select subpopulations consistent with the experimental data.

Sidechain conformational dependence of hydrogen exchange in model peptides

Peptide hydrogens that are exposed to solvent in protein X-ray structures exhibit a billion-fold range in hydroxide-catalyzed exchange rates, and these rates have previously been shown to be predictable by continuum dielectric methods to within a factor of 7, based on single protein conformations. When using a protein coil library to model the Boltzmann-weighted conformational distribution for the various N-acetyl-[X-Ala]-N-methylamides and N-acetyl-[Ala-Y]-N-methylamides, the acidity of the central amide in the individual conformers of each peptide spans nearly a million-fold range. Nevertheless, population averaging of these conformer acidities predicts the standard sidechain-dependent hydrogen exchange correction factors for nonpolar model peptides to within a factor of 30% (10(0.11)) with a correlation coefficient r=0.91. Comparison with the analogous continuum dielectric calculations for the other N-acetyl-[X-Y]-N-methylamides indicates that deviations from the isolated residue hypothesis of classical polymer theory predict appreciable errors in the exchange rates for conformationally disordered peptides when the standard sidechain-dependent hydrogen exchange rate correction factors are assumed to be independently additive. Although electronic polarizability generally dominates the dielectric shielding for the approximately 10ps lifetime of peptide ionization, evidence is presented for modest contributions from rapid intrarotamer conformational reorganization of Asn and Gln sidechains.

Mechanism of 1H-14N cross-relaxation in immobilized proteins

A resonant enhancement of the water-1H relaxation rate at three distinct frequencies in the range 0.5-3 MHz has been observed in a variety of aqueous biological systems. These so-called quadrupole (Q) peaks have been linked to a dipolar flip-flop polarization transfer from 1H nuclei to rapidly relaxing amide 14N nuclei in rotationally immobilized proteins. While the Q-peak frequencies conform to the known amide 14N quadrupole coupling parameters, a molecular model that accounts for the intensity and shape of the Q peaks has not been available. Here, we present such a model and test it against an extensive set of Q-peak data from two fully hydrated crosslinked proteins under conditions of variable temperature, pH and H/D isotope composition. We propose that polarization transfer from bulk water to amide 14N occurs in three steps: from bulk water to a so-called intermediary proton via material diffusion/exchange, from intermediary to amide proton by cross-relaxation driven by exchange-mediated orientational randomization of their mutual dipole coupling, and from amide proton to 14N by resonant dipolar relaxation 'of the second kind', driven by 14N spin fluctuations, which, in turn, are induced by restricted rigid-body motions of the protein. An essentially equivalent description of the last step can be formulated in terms of coherent 1H-->14N polarization transfer followed by fast 14N relaxation. Using independent structural and kinetic information, we show that the Q peaks from these two proteins involve approximately 7 intermediary protons in internal water molecules and side-chain hydroxyl groups with residence times of order 10(-5) s. The model not only accounts quantitatively for the extensive data set, but also explains why Q peaks are hardly observed from gelatin gels.