Ligand-induced protein transition state stabilization switches the binding pathway from conformational selection to induced fit.

Protein-ligand complex formation is fundamental to biological function. A central question is whether proteins spontaneously adopt binding-competent conformations to which ligands bind conformational selection (CS) or whether ligands induce the binding-competent conformation induced fit (IF). Here, we resolve the CS and IF binding pathways by characterizing protein conformational dynamics over a wide range of ligand concentrations using NMR relaxation dispersion. We determined the relative flux through the two pathways using a four-state binding model that includes both CS and IF. Experiments conducted without ligand show that galectin-3 exchanges between the ground-state conformation and a high-energy conformation similar to the ligand-bound conformation, demonstrating that CS is a plausible pathway. Near-identical crystal structures of the apo and ligand-bound states suggest that the high-energy conformation in solution corresponds to the apo crystal structure. Stepwise additions of the ligand lactose induce progressive changes in the relaxation dispersions that we fit collectively to the four-state model, yielding all microscopic rate constants and binding affinities. The ligand affinity is higher for the bound-like conformation than for the ground state, as expected for CS. Nonetheless, the IF pathway contributes greater than 70% of the total flux even at low ligand concentrations. The higher flux through the IF pathway is explained by considerably higher rates of exchange between the two protein conformations in the ligand-associated state. Thus, the ligand acts to decrease the activation barrier between protein conformations in a manner reciprocal to enzymatic transition-state stabilization of reactions involving ligand transformation.

Transition-State Compressibility and Activation Volume of Transient Protein Conformational Fluctuations.

Proteins are dynamic entities that intermittently depart from their ground-state structures and undergo conformational transitions as a critical part of their functions. Central to understanding such transitions are the structural rearrangements along the connecting pathway, where the transition state plays a special role. Using NMR relaxation at variable temperature and pressure to measure aromatic ring flips inside a protein core, we obtain information on the structure and thermodynamics of the transition state. We show that the isothermal compressibility coefficient of the transition state is similar to that of short-chain hydrocarbon liquids, implying extensive local unfolding of the protein. Our results further indicate that the required local volume expansions of the protein can occur not only with a net positive activation volume of the protein, as expected from previous studies, but also with zero activation volume by compaction of remote void volume, when averaged over the ensemble of states.

Mapping the energy landscape of protein-ligand binding via linear free energy relationships determined by protein NMR relaxation dispersion.

Biochemical signaling is mediated by complexes between macromolecular receptors and their ligands, with the duration of the signal being directly related to the lifetime of the ligand-receptor complex. In the field of drug design, the recognition that drug efficacy in vivo depends on the lifetime of the drug-protein complex has spawned the concept of designing drugs with particular binding kinetics. To advance this field it is critical to investigate how the molecular details of designed ligands might affect the binding kinetics, as well as the equilibrium binding constant. Here we use protein NMR relaxation dispersion to determine linear free energy relationships involving the on- and off-rates and the affinity for a series of congeneric ligands targeting the carbohydrate recognition domain of galectin-3. Using this approach we determine the energy landscape and the position of the transition state along the reaction coordinate of protein-ligand binding. The results show that ligands exhibiting reduced off-rates achieve this by primarily stabilizing the bound state, but do not affect the transition state to any greater extent. The transition state forms early, that is, it is located significantly closer to the free state than to the bound state, suggesting a critical role of desolvation. Furthermore, the data suggest that different subclasses of ligands show different behavior with respect to these characteristics.

Rotamer Jumps, Proton Exchange, and Amine Inversion Dynamics of Dimethylated Lysine Residues in Proteins Resolved by pH-Dependent 1H and 13C NMR Relaxation Dispersion.

Post-translational methylation of lysine side chains is of great importance for protein regulation, including epigenetic control. Here, we present specific 13CHD2 labeling of dimethylated lysines as a sensitive probe of the structure, interactions, and dynamics of these groups, and outline a theoretical and experimental framework for analyzing their conformational dynamics using 1H and 13C CPMG relaxation dispersion experiments. Dimethylated lysine side chains in calcium-loaded calmodulin show a marked pH dependence of their Carr-Purcell-Meiboom-Gill (CPMG) dispersion profiles, indicating complex exchange behavior. Combined analysis of 1H and 13C CPMG relaxation dispersions requires consideration of 12-state correlated exchange of the two methyl groups due to circular three-state rotamer jumps around the Cε-Nζ axis combined with proton exchange and amine inversion. Taking into account a number of fundamental constraints, the exchange model can be reduced to include only three fitted parameters, namely, the geometric average of the rotamer-jump rate constants, the rate constant of deprotonation of Nζ, and the chemical shift difference between the trans and gauge positions of the 13C or 1H nuclei. The pH dependence indicates that protonation of the end group dramatically slows down rotamer exchange for some lysine residues, whereas deprotonation leads to rapid amine inversion coupled with rotamer scrambling. The observed variation among residues in their exchange behavior appears to depend on the structural environment of the side chain. Understanding this type of exchange process is critical to correctly interpreting NMR spectra of methylated lysine side chains. The exchange model presented here forms the basis for studying the structure and dynamics of epigenetically modified lysine side chains and perturbations caused by changes in pH or interactions with target proteins.

Dynamics of Aromatic Side Chains in the Active Site of FKBP12.

FKBP12, a small human enzyme, aids protein folding by catalyzing cis-trans isomerization of peptidyl-prolyl bonds, and is involved in cell signaling pathways, calcium regulation, and the immune response. The underlying molecular mechanisms are not fully understood, but it is well-known that aromatic residues in the active site and neighboring loops are important for substrate binding and catalysis. Here we report micro- to millisecond exchange dynamics of aromatic side chains in the active site region of ligand-free FKBP12, involving a minor state population of 0.5% and an exchange rate of 3600 s-1, similar to previous results for the backbone and methyl-bearing side chains. The exchange process involves tautomerization of H87. In the major state H87 is highly flexible and occupies the common HNε2 tautomer, while in the minor state it occupies the rare HNδ1 tautomer, which typically requires stabilization by specific interactions, such as hydrogen bonds. This finding suggests that the exchange process is coupled to a rearrangement of the hydrogen bond network around H87. Upon addition of the active-site inhibitor FK506 the exchange of all aromatic residues is quenched, with exception of H87. The H87 resonances are broadened beyond detection, suggesting that interconversion between tautomers prevail in the FK506-bound state. While key active-site residues undergo conformational exchange in the apo state, the exchange rate is considerably faster than the catalytic turnover, as determined herein by Michaelis-Menten type analysis of NMR line shapes and chemical shifts. We discuss alternative interpretations of this observation in terms of FKBP12 function.

Molecular insights into substrate recognition and catalytic mechanism of the chaperone and FKBP peptidyl-prolyl isomerase SlyD.

BACKGROUND: Peptidyl-prolyl isomerases (PPIases) catalyze cis/trans isomerization of peptidyl-prolyl bonds, which is often rate-limiting for protein folding. SlyD is a two-domain enzyme containing both a PPIase FK506-binding protein (FKBP) domain and an insert-in-flap (IF) chaperone domain. To date, the interactions of these domains with unfolded proteins have remained rather obscure, with structural information on binding to the FKBP domain being limited to complexes involving various inhibitor compounds or a chemically modified tetrapeptide. RESULTS: We have characterized the binding of 15-residue-long unmodified peptides to SlyD from Thermus thermophilus (TtSlyD) in terms of binding thermodynamics and enzyme kinetics through the use of isothermal titration calorimetry, nuclear magnetic resonance spectroscopy, and site-directed mutagenesis. We show that the affinities and enzymatic activity of TtSlyD towards these peptides are much higher than for the chemically modified tetrapeptides that are typically used for activity measurements on FKBPs. In addition, we present a series of crystal structures of TtSlyD with the inhibitor FK506 bound to the FKBP domain, and with 15-residue-long peptides bound to either one or both domains, which reveals that substrates bind in a highly adaptable fashion to the IF domain through ß-strand augmentation, and can bind to the FKBP domain as both types VIa1 and VIb-like cis-proline ß-turns. Our results furthermore provide important clues to the catalytic mechanism and support the notion of inter-domain cross talk. CONCLUSIONS: We found that 15-residue-long unmodified peptides can serve as better substrate mimics for the IF and FKBP domains than chemically modified tetrapeptides. We furthermore show how such peptides are recognized by each of these domains in TtSlyD, and propose a novel general model for the catalytic mechanism of FKBPs that involves C-terminal rotation around the peptidyl-prolyl bond mediated by stabilization of the twisted transition state in the hydrophobic binding site.

Ring flips revisited: (13)C relaxation dispersion measurements of aromatic side chain dynamics and activation barriers in basic pancreatic trypsin inhibitor.

Intramolecular motions of proteins are critical for biological function. Transient structural fluctuations underlie a wide range of processes, including enzyme catalysis, ligand binding to buried sites, and generic protein motions, such as 180° rotation of aromatic side chains in the protein interior, but remain poorly understood. Understanding the dynamics and molecular nature of concerted motions requires characterization of their rates and energy barriers. Here we use recently developed (13)C transverse relaxation dispersion methods to improve our current understanding of aromatic ring flips in basic pancreatic trypsin inhibitor (BPTI). We validate these methods by benchmarking ring-flip rates against the three previously characterized cases in BPTI, namely, Y23, Y35, and F45. Further, we measure conformational exchange for one additional aromatic ring, F22, which can be interpreted in terms of a flip rate of 666 s(-1) at 5 °C. Upon inclusion of our previously reported result that Y21 also flips slowly [Weininger, U., et al. (2013) J. Phys. Chem. B 117, 9241-9247], the (13)C relaxation dispersion experiments thus reveal relatively slow ring-flip rates for five of eight aromatic residues in BPTI. These results are in contrast with previous reports, which have estimated that all rings, except Y23, Y35, and F45, flip with a high rate at ambient temperature. The (13)C relaxation dispersion data result in an updated rank order of ring-flip rates in BPTI, which agrees considerably better with that estimated from a recent 1 ms molecular dynamics trajectory than do previously published NMR data. However, significant quantitative differences remain between experiment and simulation, in that the latter yields flip rates that are in many cases too fast by 1-2 orders of magnitude. By measuring flip rates across a temperature range of 5-65 °C, we determined the activation barriers of ring flips for Y23, Y35, and F45. Y23 and F45 have identical activation parameters, suggesting that the fluctuations of the protein core around these residues are similar in character. Y35 differs from the other two in its apparent activation entropy. These results might be rationalized by the fact that Y23 and F45 are located in the same region of the structure while Y35 is remote from the other two rings. As indicated by our new results for the exceptionally well-characterized protein BPTI, (13)C relaxation dispersion experiments open the possibility of studying ring flips in a range of cases wider than that previously possible.

Off-resonance rotating-frame relaxation dispersion experiment for 13C in aromatic side chains using L-optimized TROSY-selection.

Protein dynamics on the microsecond-millisecond time scales often play a critical role in biological function. NMR relaxation dispersion experiments are powerful approaches for investigating biologically relevant dynamics with site-specific resolution, as shown by a growing number of publications on enzyme catalysis, protein folding, ligand binding, and allostery. To date, the majority of studies has probed the backbone amides or side-chain methyl groups, while experiments targeting other sites have been used more sparingly. Aromatic side chains are useful probes of protein dynamics, because they are over-represented in protein binding interfaces, have important catalytic roles in enzymes, and form a sizable part of the protein interior. Here we present an off-resonance R 1ρ experiment for measuring microsecond to millisecond conformational exchange of aromatic side chains in selectively (13)C labeled proteins by means of longitudinal- and transverse-relaxation optimization. Using selective excitation and inversion of the narrow component of the (13)C doublet, the experiment achieves significant sensitivity enhancement in terms of both signal intensity and the fractional contribution from exchange to transverse relaxation; additional signal enhancement is achieved by optimizing the longitudinal relaxation recovery of the covalently attached (1)H spins. We validated the L-TROSY-selected R 1ρ experiment by measuring exchange parameters for Y23 in bovine pancreatic trypsin inhibitor at a temperature of 328 K, where the ring flip is in the fast exchange regime with a mean waiting time between flips of 320 µs. The determined chemical shift difference matches perfectly with that measured from the NMR spectrum at lower temperatures, where separate peaks are observed for the two sites. We further show that potentially complicating effects of strong scalar coupling between protons (Weininger et al. in J Phys Chem B 117: 9241-9247, 2013b) can be accounted for using a simple expression, and provide recommendations for data acquisition when the studied system exhibits this behavior. The present method extends the repertoire of relaxation methods tailored for aromatic side chains by enabling studies of faster processes and improved control over artifacts due to strong coupling.

Protein flexibility and conformational entropy in ligand design targeting the carbohydrate recognition domain of galectin-3.

Rational drug design is predicated on knowledge of the three-dimensional structure of the protein-ligand complex and the thermodynamics of ligand binding. Despite the fundamental importance of both enthalpy and entropy in driving ligand binding, the role of conformational entropy is rarely addressed in drug design. In this work, we have probed the conformational entropy and its relative contribution to the free energy of ligand binding to the carbohydrate recognition domain of galectin-3. Using a combination of NMR spectroscopy, isothermal titration calorimetry, and X-ray crystallography, we characterized the binding of three ligands with dissociation constants ranging over 2 orders of magnitude. (15)N and (2)H spin relaxation measurements showed that the protein backbone and side chains respond to ligand binding by increased conformational fluctuations, on average, that differ among the three ligand-bound states. Variability in the response to ligand binding is prominent in the hydrophobic core, where a distal cluster of methyl groups becomes more rigid, whereas methyl groups closer to the binding site become more flexible. The results reveal an intricate interplay between structure and conformational fluctuations in the different complexes that fine-tunes the affinity. The estimated change in conformational entropy is comparable in magnitude to the binding enthalpy, demonstrating that it contributes favorably and significantly to ligand binding. We speculate that the relatively weak inherent protein-carbohydrate interactions and limited hydrophobic effect associated with oligosaccharide binding might have exerted evolutionary pressure on carbohydrate-binding proteins to increase the affinity by means of conformational entropy.

High water mobility on the ice-binding surface of a hyperactive antifreeze protein.

Antifreeze proteins (AFPs) prevent uncontrolled ice formation in organisms exposed to subzero temperatures by binding irreversibly to specific planes of nascent ice crystals. To understand the thermodynamic driving forces and kinetic mechanism of AFP activity, it is necessary to characterize the hydration behavior of these proteins in solution. With this aim, we have studied the hyperactive insect AFP from Tenebrio molitor (TmAFP) with the (17)O magnetic relaxation dispersion (MRD) method, which selectively monitors the rotational motion and exchange kinetics of water molecules on picosecond-microsecond time scales. The global hydration behavior of TmAFP is found to be similar to non-antifreeze proteins, with no evidence of ice-like or long-ranged modifications of the solvent. However, two sets of structural water molecules, located within the core and on the ice-binding face in the crystal structure of TmAFP, may have functional significance. We find that 2 of the 5 internal water molecules exchange with a residence time of 8 +/- 1 micros at 300 K and a large activation energy of approximately 50 kJ mol(-1), reflecting intermittent large-scale conformational fluctuations in this exceptionally dense and rigid protein. Six water molecules arrayed with ice-like spacing in the central trough on the ice-binding face exchange with bulk water on a sub-nanosecond time scale. The combination of high order and fast exchange may allow these water molecules to contribute entropically to the ice-binding affinity without limiting the absorption rate.

Structure and Dynamics of Ribosomal Protein L12: An Ensemble Model Based on SAXS and NMR Relaxation.

Ribosomal protein L12 is a two-domain protein that forms dimers mediated by its N-terminal domains. A 20-residue linker separates the N- and C-terminal domains. This linker results in a three-lobe topology with significant flexibility, known to be critical for efficient translation. Here we present an ensemble model of spatial distributions and correlation times for the domain reorientations of L12 that reconciles experimental data from small-angle x-ray scattering and nuclear magnetic resonance. We generated an ensemble of L12 conformations in which the structure of each domain is fixed but the domain orientations are variable. The ensemble reproduces the small-angle x-ray scattering data and the optimized correlation times of its reorientational eigenmodes fit the (15)N relaxation data. The ensemble model reveals intrinsic conformational properties of L12 that help explain its function on the ribosome. The two C-terminal domains sample a large volume and extend further away from the ribosome anchor than expected for a random-chain linker, indicating that the flexible linker has residual order. Furthermore, the distances between each C-terminal domain and the anchor are anticorrelated, indicating that one of them is more retracted on average. We speculate that these properties promote the function of L12 to recruit translation factors and control their activity on the ribosome.

Conformational entropy changes upon lactose binding to the carbohydrate recognition domain of galectin-3.

The conformational entropy of proteins can make significant contributions to the free energy of ligand binding. NMR spin relaxation enables site-specific investigation of conformational entropy, via order parameters that parameterize local reorientational fluctuations of rank-2 tensors. Here we have probed the conformational entropy of lactose binding to the carbohydrate recognition domain of galectin-3 (Gal3), a protein that plays an important role in cell growth, cell differentiation, cell cycle regulation, and apoptosis, making it a potential target for therapeutic intervention in inflammation and cancer. We used (15)N spin relaxation experiments and molecular dynamics simulations to monitor the backbone amides and secondary amines of the tryptophan and arginine side chains in the ligand-free and lactose-bound states of Gal3. Overall, we observe good agreement between the experimental and computed order parameters of the ligand-free and lactose-bound states. Thus, the (15)N spin relaxation data indicate that the molecular dynamics simulations provide reliable information on the conformational entropy of the binding process. The molecular dynamics simulations reveal a correlation between the simulated order parameters and residue-specific backbone entropy, re-emphasizing that order parameters provide useful estimates of local conformational entropy. The present results show that the protein backbone exhibits an increase in conformational entropy upon binding lactose, without any accompanying structural changes.

Model-independent interpretation of NMR relaxation data for unfolded proteins: the acid-denatured state of ACBP.

We have investigated the acid-unfolded state of acyl-coenzyme A binding protein (ACBP) using 15N laboratory frame nuclear magnetic resonance (NMR) relaxation experiments at three magnetic field strengths. The data have been analyzed using standard model-free fitting and models involving distribution of correlation times. In particular, a model-independent method of analysis that does not assume any analytical form for the correlation time distribution is proposed. This method explains correlations between model-free parameters and the analytical distribution parameters found by other authors. The analysis also shows that the relaxation data are consistent with and complementary to information obtained from other parameters, especially secondary chemical shifts and residual dipolar couplings, and strengthens the conclusions of previous observations that three out of the four regions that form helices in the native structure appear to contain residual secondary structure also in the acid-denatured state.

Detection of initiation sites in protein folding of the four helix bundle ACBP by chemical shift analysis.

A simple alternative method for obtaining "random coil" chemical shifts by intrinsic referencing using the protein's own peptide sequence is presented. These intrinsic random coil backbone shifts were then used to calculate secondary chemical shifts, that provide important information on the residual secondary structure elements in the acid-denatured state of an acyl-coenzyme A binding protein. This method reveals a clear correlation between the carbon secondary chemical shifts and the amide secondary chemical shifts 3-5 residues away in the primary sequence. These findings strongly suggest transient formation of short helix-like segments, and identify unique sequence segments important for protein folding.

Dynamics of protein and peptide hydration.

Biological processes often involve the surfaces of proteins, where the structural and dynamic properties of the aqueous solvent are modified. Information about the dynamics of protein hydration can be obtained by measuring the magnetic relaxation dispersion (MRD) of the water (2)H and (17)O nuclei or by recording the nuclear Overhauser effect (NOE) between water and protein protons. Here, we use the MRD method to study the hydration of the cyclic peptide oxytocin and the globular protein BPTI in deeply supercooled solutions. The results provide a detailed characterization of water dynamics in the hydration layer at the surface of these biomolecules. More than 95% of the water molecules in contact with the biomolecular surface are found to be no more than two-fold motionally retarded as compared to bulk water. In contrast to small nonpolar molecules, the retardation factor for BPTI showed little or no temperature dependence, suggesting that the exposed nonpolar residues do not induce clathrate-like hydrophobic hydration structures. New NOE data for oxytocin and published NOE data for BPTI were analyzed, and a mutually consistent interpretation of MRD and NOE results was achieved with the aid of a new theory of intermolecular dipolar relaxation that accounts explicitly for the dynamic perturbation at the biomolecular surface. The analysis indicates that water-protein NOEs are dominated by long-range dipolar couplings to bulk water, unless the monitored protein proton is near a partly or fully buried hydration site where the water molecule has a long residence time.

Water and urea interactions with the native and unfolded forms of a beta-barrel protein.

A fundamental understanding of protein stability and the mechanism of denaturant action must ultimately rest on detailed knowledge about the structure, solvation, and energetics of the denatured state. Here, we use (17)O and (2)H magnetic relaxation dispersion (MRD) to study urea-induced denaturation of intestinal fatty acid-binding protein (I-FABP). MRD is among the few methods that can provide molecular-level information about protein solvation in native as well as denatured states, and it is used here to simultaneously monitor the interactions of urea and water with the unfolding protein. Whereas CD shows an apparently two-state transition, MRD reveals a more complex process involving at least two intermediates. At least one water molecule binds persistently (with residence time >10 nsec) to the protein even in 7.5 M urea, where the large internal binding cavity is disrupted and CD indicates a fully denatured protein. This may be the water molecule buried near the small hydrophobic folding core at the D-E turn in the native protein. The MRD data also provide insights about transient (residence time <1 nsec) interactions of urea and water with the native and denatured protein. In the denatured state, both water and urea rotation is much more retarded than for a fully solvated polypeptide. The MRD results support a picture of the denatured state where solvent penetrates relatively compact clusters of polypeptide segments.

Trifluoroethanol-induced beta --> alpha transition in beta-lactoglobulin: hydration and cosolvent binding studied by 2H, 17O, and 19F magnetic relaxation dispersion.

Alcohols, such as 2,2,2-trifluoroethanol (TFE), have been shown to induce a cooperative transition to an open helical structure in many proteins, but the underlying molecular mechanism has not been identified. Here, we employ the technique of magnetic relaxation dispersion (MRD) to study the TFE-induced beta --> alpha transition of beta-lactoglobulin at pH 2.4. Unlike traditional techniques that focus on protein secondary structure, the MRD method directly monitors the solvent, providing quantitative information about preferential solvation and solvent penetration and about the overall size and structural integrity of the protein. In this multinuclear MRD study, we use the (2)H and (17)O resonances to examine hydration and the (19)F resonance to study TFE. The transformation from the native to the helical state via an intermediate state at 300 K is found to be accompanied by a progressive expansion of the protein and loss of specific long-lived hydration sites. The observation of (17)O and (19)F dispersions from the helical state shows that water and TFE penetrate the protein. The MRD data indicate a strong accumulation of TFE at the surface as well as in the interior of the protein. At 277 K, BLG is much less affected by TFE, remaining in the native state at 16% TFE, but adopting a nonnative structure at 30% TFE. This nonnative structure is not penetrated by long-lived water molecules. The implications of these findings for the mechanism of TFE-induced structural transformations are discussed.

Water dynamics in the large cavity of three lipid-binding proteins monitored by (17)O magnetic relaxation dispersion.

Intracellular lipid-binding proteins contain a large binding cavity filled with water molecules. The role played by these water molecules in ligand binding is not well understood, but their energetic and dynamic properties must be important for protein function. Here, we use the magnetic relaxation dispersion (MRD) of the water 17O resonance to investigate the water molecules in the binding cavity of three different lipid-binding proteins: heart fatty acid-binding protein (H-FABP), ileal lipid-binding protein (I-LBP) and intestinal fatty acid-binding protein (I-FABP). Whereas about half of the crystallographically visible water molecules appear to be expelled by the ligand, we find that ligand binding actually increases the number of water molecules within the cavity. At 300 K, the water molecules in the cavity exchange positions on a time-scale of about 1ns and exchange with external water on longer time-scales (0.01-1 micros). Exchange of water molecules among hydration sites within the cavity should be strongly coupled to ligand motion. Whereas a recent MD simulation indicates that the structure of the cavity water resembles a bulk water droplet, the present MRD results show that its dynamics is more than two orders of magnitude slower than in the bulk. These findings may have significant implications for the strength, specificity and kinetics of lipid binding.

Temperature-dependent hydrogen-bond geometry in liquid water.

We have determined the hydrogen-bond geometry in liquid water from 0 to 80 degrees C by combining measurements of the proton magnetic shielding tensor with ab initio density functional calculations. The resulting moments of the distributions of hydrogen-bond length and angle are direct measures of thermal disorder in the hydrogen-bond network. These moments, and the distribution functions that can be reconstructed from them, impose quantitative constraints on structural models of liquid water.

Proton magnetic shielding tensor in liquid water.

The nuclear magnetic shielding tensor is a sensitive probe of the local electronic environment, providing information about molecular structure and intermolecular interactions. The magnetic shielding tensor of the water proton has been determined in hexagonal ice, but in liquid water, where the tensor is isotropically averaged by rapid molecular tumbling, only the trace of the tensor has been measured. We report here the first determination of the proton shielding anisotropy in liquid water, which, when combined with chemical shift data, yields the principal shielding components parallel (sigma(parallel)) and perpendicular (sigma(perpendicular)) to the O-H bond. We obtained the shielding anisotropy sigma(parallel)-sigma(perpendicular) by measuring the proton spin relaxation rate as a function of magnetic induction field in a water sample where dipole-dipole couplings are suppressed by H/D isotope dilution. The temperature dependence of the shielding components, determined from 0 to 80 degrees C, reflects vibrational averaging over a distribution of instantaneous hydrogen-bond geometries in the liquid and thus contains unique information about the temperature-dependent structure of liquid water. The temperature dependence of the shielding anisotropy is found to be 4 times stronger than that of the isotropic shielding. We analyze the liquid water shielding components in the light of previous NMR and theoretical results for vapor and ice. We show that a simple two-state model of water structure fails to give a consistent interpretation of the shielding data and we argue that a more detailed analysis is needed that quantitatively relates the shielding components to hydrogen bond geometry.