Keywords

4.1 Introduction

As a part of living systems, biomacromolecules are characterized by their various degrees of freedom with regard to internal motions, which provide them with conformational plasticity that is required during binding and reaction processes and also for conformational adaptability to their binding partners. For example, many proteins exhibit multiple conformations in the absence of their cognate ligands, which select and stabilize specific conformers of the target proteins [1, 2]. Conformational dynamics during molecular recognition processes are exemplified by intrinsically disordered proteins that are devoid of stable secondary or tertiary structures and are often folded into specific secondary structures upon interacting with their binding partners [3, 4].

Nuclear magnetic resonance (NMR) spectroscopy provides potentially useful tools for characterizing the conformational dynamics of biomolecules over a wide range of time scales (Fig. 4.1) [5, 6]. If the conformational interconversion is sufficiently slow on the NMR time scale, then NMR signals originating from individual conformers are observable and provide detailed information for each conformational state. However, under circumstances when conformational transitions occur faster, the NMR peaks converge at a position that corresponds to a population-weighted average of chemical shifts. Even under these conditions, NMR relaxation dispersion experiments can provide information on a chemical shift and a population of conformers as well as the exchange rate if the interconversion occurs on the micro-to-millisecond time scale [7–9].

Fig. 4.1
figure 1

Biomolecular NMR experiments to characterize protein dynamics on different time scales

Full size image

Conformational fluctuations on the subnanosecond time scale are characterized based on measurements of longitudinal and transverse relaxation rates (R 1 and R 2, respectively) and the heteronuclear Overhauser effect (NOE). For example, a model-free formalism is widely used to describe protein backbone dynamics with order parameters and correlation times of individual amino acid residues on the basis of observations of 15N R 1, R 2, and {1H}-15N NOE [10, 11]. However, it is challenging to characterize individual conformers that form such a dynamic ensemble of biomolecules.

In this chapter, we present several examples of strategies for obtaining detailed information on specific conformers under dynamic equilibrium, particularly minor conformational species under physiological conditions.

4.2 Trapping Conformational States

4.2.1 High-Pressure NMR Combined with a Mutational Approach

In many biomolecular NMR studies, spectral measurements are made under equilibrium solution conditions in which the molecules undergo dynamic conformational interconversions. Among pre-existing conformers, major conformers are not always biologically active. For detailed structural analyses of their minor conformational states, the equilibria are perturbed by changing solution conditions, such as the pH, salt concentration, and temperature, so as to make the originally minor species more populated.

High-pressure NMR spectroscopy is one of the most powerful methods for capturing low-population conformational species of proteins, which have higher Gibbs free energies than their basic folded state (native state). Because the conformational fluctuations of a protein in solution are closely related to the fluctuations of its partial molar volume [12, 13], applying high pressure causes a population shift of conformers within the ensemble, which results in NMR spectral changes [14]. Here we illustrate the utility of the high-pressure NMR approach using ubiquitin (Ub) as a model.

Ub is a small protein that consists of 76 amino acid residues and is a posttranslational protein modifier, which plays crucial regulatory roles in diverse cellular events [15]. A Ub modification can be made through the formation of isopeptide linkages at its C-terminus. Although Ub can be covalently attached to target proteins in a monomeric form, it can also be linked to another Ub to generate various types of polyUb chains. PolyUb chains that are mediated through a Lys48 isopeptide linkage provide a signal that is recognized by the 26S proteasome for selective protein degradation in cells. The formation of Ub isopeptide linkages is catalyzed by three enzymes, Ub-activating enzyme E1, Ub-conjugating enzyme E2, and Ub-protein ligase E3, in sequential order.

Previously reported high-pressure NMR investigations demonstrated that Ub existed in equilibrium among its natively folded state N1, a totally unfolded state U, and at least two high-energy states, an alternatively folded state N2 and a locally disordered state I [16–18]. Based on spectral data acquired at low (30 bar) and high (3 kbar) pressures, it was concluded that the N2 state accounted for 77 % of the population at high pressure, which involved reorientation of its C-terminal β5-strand and α1-helix and provided an open conformation in the proximity of its C-terminus [18]. The structural changes of pressure-stabilized N2 were supported by residual dipolar coupling (RDC) experiments [19] and 1 μs molecular dynamics (MD) simulations [20] of the protein under high pressure.

We demonstrated that the high-pressure NMR approach provided a new strategy for amplifying particular fluctuations in proteins by rational mutation designed based on the conformational states observed under different pressure conditions [21]. With this strategy, the N1 (30 bar) and N2 (3 kbar) structures of wild-type (WT) Ub were first compared to identify the key intramolecular interactions. Subsequently, the hydrogen bond or salt bridge thus identified was systematically broken by single amino acid replacements to control conformational fluctuations. A series of Ub mutants (K11A, E34A, Q41A, and Q41N) were subjected to heteronuclear single-quantum coherence (HSQC) experiment-based chemical shift comparisons (Fig. 4.2), which indicated that the Q41N mutant preferentially (70 %) adopted the N2 conformation at atmospheric pressure. The NMR structure of the N2 conformer of Q41N-mutated Ub determined at atmospheric pressure confirmed that the hydrogen bond between the Ile36 backbone carbonyl group and the Gln41 side chain amide group was critical for controlling the N1−N2 conformational interconversion in WT Ub.

Fig. 4.2
figure 2

Overlay of the HSQC spectra of WT (black) and mutated Ubs (K11A, green; E34A, blue; Q41A, purple; Q41N, red). The substituted side chains are shown in the inset Ub structure. All mutant peaks lie approximately in a straight line that connects the WT and Q41N peaks (Adapted with permission from Ref. [21])

Full size image

In the Q41N mutant, the N2 state accounted for 97 % of the population at high pressure (2.5 kbar), which enabled detailed conformational characterizations of the “pure” N2 state of Ub [22]. The N2 state exhibited an altered orientation of β5-strand as compared with that of the natively folded N1 state (Fig. 4.3). Interestingly, the conformation of N2 was quite similar to the crystal structure of Ub in complex with the Ub-activating enzyme E1 (Fig. 4.4), which indicated that Ub recognition by E1 was best explained by conformational selection rather than induced-fit motion. Similar high-energy, biologically active states are shared by NEDD8, a Ub-like posttranslational modifier that has its own E1−E2−E3 reaction cascade [23]. These findings suggest that both the major ground state and the minor active conformational state are conserved among the Ub-like modifiers.

Fig. 4.3
figure 3

(a) Superposition of HSQC spectra of WT (left) and Q41N (right) Ub at different pressures. The Q41N mutant preferentially (70 %) adopts the N2 conformation at atmospheric pressure. (b) Solution structure of WT ubiquitin at 1 bar, a model for the basic folded conformation N1 [Protein Data Bank (PDB) entry 1D3Z] (left), and the Q41N variant at 2500 bar, model for the alternatively folded conformation N2 (PDB entry 2RU6) (right) (Adapted with permissions from Refs. [21] and [22])

Full size image
Fig. 4.4
figure 4

Superposition of NMR-derived structures of WT Ub at 1 bar (green, PDB entry 1D3Z) and Q41N at 2500 bar (red, PDB entry 2RU6) on the crystal structure of the Ub−E1 complex (blue, PDB entry 3CMM). E1 is shown as a molecular surface (gray) (Adapted with permission from Ref. [22])

Full size image

Thus, high-pressure NMR measurements combined with suitably chosen point mutations could be a useful strategy for conformational trapping of natively low-populated but functionally relevant protein states.

4.2.2 Covalent Constraints of Biomolecular Conformations

Protein functions are derived primarily through the cooperative interplay between two or more domains. The plasticity in spatial arrangements of these domains is a key feature of multidomain proteins endowed with sophisticated molecular mechanisms typified by allosteric regulation. Interconversion of domain arrangements takes place over a wide range of temporal and spatial time scales. When domain rearrangements are rapid, it is again not possible to describe individual conformational states based on NMR data obtained as population-weighted averages unless specific conformers can be frozen. Here we present our results for characterizing the conformational dynamics of a Lys48-linked Ub dimer (diUb) by NMR spectroscopy in conjunction with X-ray crystallography [24].

The previously reported crystal structure of Lys48-linked diUb exhibited a compact closed conformation in which the hydrophobic surfaces in both Ub units were packed against each other and shielded from the solvent [25]. The Ub hydrophobic surface is involved in interactions with the Ub-binding proteins that function in the Ub-/proteasome-mediated protein degradation pathway, including the proteasome subunits Rpn10 and Rpn13 [26, 27]. Therefore, one may wonder how Ub-binding proteins can efficiently access the functionally important but sequestered hydrophobic surface of the Lys48-linked polyUb chains. We addressed this using Lys48-linked diUb prepared by in vitro E1–E2 (E2-25 K) enzymatic reactions [28], which generated WT Lys48-linked polyUb chains as well as their cyclic forms, in which each Ub unit was linked via two isopeptide linkages (Fig. 4.5). Chromatographically isolated WT and cyclic Lys48-linked diUb were then subjected to structural analyses.

Fig. 4.5
figure 5

Schematic diagram for preparing WT and cyclic Lys48-linked Ub chains (Adapted with permission from Ref. [24])

Full size image

Intriguingly, our resolved crystal structure of WT Lys48-linked diUb exhibited an open conformation in contrast to the previously reported crystal structures and suggested that the crystallographic data provided snapshots of the diUb molecule under dynamic equilibrium in solution (Fig. 4.6a). Furthermore, our NMR-derived 3D structure of the cyclic diUb form closely resembled the closed conformation observed in the crystals (Fig. 4.6b). Thus, the closed conformer could be successfully isolated by introducing additional isopeptide linkages. Interestingly, endogenous cyclic Lys48-linked diUb was detected in skeletal muscle and cultured mammalian cells [29], although its physiological role remains unknown.

Fig. 4.6
figure 6

(a) Superposition of the crystal structures of WT Lys48-linked diUb in the open (present structure; PDB code, 3AUL) (magenta) and closed forms (PDB code, 1AAR) (cyan). (b) Superposition of the NMR-derived model of cyclic Lys48-linked diUb (green) and the crystal structure of WT Lys48-linked diUb showing the closed conformation (PDB code, 1AAR) (cyan). The C-terminal group of Ub is linked to another Ub (designated the distal and proximal moieties, respectively) through its Lys48 side chain (Adapted with permission from Ref. [24])

Full size image

HSQC spectral data of WT Lys48-linked diUb were compared with those of the cyclic form and monomeric Ub, which mimics an open form of diUb in terms of its hydrophobic surface exposure (Fig. 4.7). In WT diUb, each amino acid residue located on the hydrophobic surface provided a single HSQC peak between the peaks that originated from the corresponding sites in monomeric Ub and cyclic diUb in the same straight line with an identical dividing ratio. These data demonstrated that Lys48-linked diUb interconverted between open and closed conformations, which were mimicked by monomeric Ub and cyclic diUb, respectively, in the fast-exchange regime. In this system, the populations of open and closed conformers were determined to be 75 % and 25 %, respectively, based simply on the division ratio of the chemical shift difference. The open conformation accounted for a greater population under lower pH conditions and almost completely predominated at pH 4.5 (Fig. 4.8). These data suggested that the intrinsic properties of Lys48-linked Ub chains to adopt the open conformation may be advantageous for their interactions with Ub-binding proteins.

Fig. 4.7
figure 7

(a) Cartoon model of the conformational equilibrium of WT Lys48-linked diUb. Cyclic Lys48-linked diUb was used as a mimic of the closed state, and monomeric Ub was used as a mimic of the open state. The hydrophobic surface is colored red. (b) 1H–15N HSQC spectra of monomeric Ub (cyan), WT Lys48-linked diUb (red), and cyclic Lys48-linked diUb (black) at pH 7.0. (c) Close-up views of the spectral regions (boxed), including the peaks for Ile13, Leu43, Leu69, and Val70, are displayed at the bottom (Adapted with permission from Ref. [24])

Full size image
Fig. 4.8
figure 8

pH-dependent chemical shift changes of the HSQC peaks for monomeric Ub (cyan), WT Lys48-linked diUb (red), and cyclic Lys48-linked diUb (black), which shows that the open conformation of WT Lys48-linked diUb becomes populated upon lowering the pH (Adapted with permission from Ref. [24])

Full size image

This study demonstrates the utility of conformational trapping of multidomain proteins by introducing covalent linkage constraints.

4.3 Exploration of Conformational Spaces of Flexible Biomolecules

In the previous sections, we described protein conformational dynamics as interchanges among a few conformers. However, in general, biomolecular dynamics involve not only conformational transitions from a low-energy minimum to another in the energy landscape but also harmonic motions in each energy minimum. This is particularly true for highly flexible biomolecules best represented by oligosaccharides and intrinsically disordered proteins. Oligosaccharides have branched structures and have numerous degrees of freedom of motion, which provide conformational spaces with a number of shallow and broad energy minima [30]. Thus, the three-dimensional (3D) structure of such a flexible oligosaccharide should be interpreted as a vast ensemble of dynamic conformations, rather than only a limited number of conformational states.

The biological functions of oligosaccharides are exerted primarily through molecular recognition processes that are mediated by carbohydrate-binding proteins collectively designated lectins [31, 32]. For example, various high-mannose-type oligosaccharides attached to proteins serve as quality indicators in the early secretory pathway, which are recognized by a series of intracellular lectins that function during protein folding, transport, and degradation processes [33–35]. On cell surfaces, various glycoconjugates mediate a number of physiological and pathological processes, as exemplified by gangliosides, glycosphingolipids containing sialyl residue(s), which are involved in cellular communications and development, cancer metastasis, viral infections, and the onset and progression of neurodegenerative disorders [36–39].

To better understand the mechanisms underlying oligosaccharide functions, their 3D structural information should be acquired with quantitative descriptions of their dynamic, rapid conformational fluctuations. NMR methods combined with computational simulation techniques provide useful tools to explore the conformational spaces of flexible biomolecules. For example, MD simulations can be used to predict the dynamic conformational motions of highly flexible oligosaccharides in solution [40, 41]. However, a frequent problem with this approach is its dependence on simulation protocols, including initial structures, computational times, and the choice of force fields. Therefore, it is critically important to validate the simulation results by comparisons with experimental data for biomolecular fluctuations on time scales ranging from subnanoseconds to microseconds. We developed a methodology to evaluate MD simulation results using paramagnetism-assisted NMR techniques to explore the conformational spaces occupied by flexible oligosaccharides in solution [42].

Introducing a paramagnetic center, such as lanthanide ions, into a target biomolecule results in significant modifications of its NMR spectra due to the generated magnetic dipole–dipole interactions between nuclei and an unpaired electron [43, 44]. For example, pseudocontact shift (PCS), a well-known paramagnetic effect observed in the presence of lanthanide ions with anisotropic magnetic susceptibility, can modulate the chemical shift of each NMR signal according to the spatial arrangements of individual nuclei with respect to the paramagnetic center [45]. Because of the large magnetic moment of unpaired electrons, through-space dipole interactions perturb not only the signals from the nuclei in proximity to the paramagnetic center but also those that are distal from it. Thus, PCS measurements have the potential to provide valuable atomic long-distance information for characterizing biomolecular conformations.

We applied this method to evaluate the conformational dynamics of a GM3 trisaccharide (NeuAcα2-3Galβ1-4Glc), the common core oligosaccharide structure shared among gangliosides [46]. Attaching a paramagnetic lanthanide tag to this trisaccharide allowed us to measure PCS effects (Fig. 4.9a). Using two-dimensional 1H–13C HSQC spectra, the PCS values of each 1H and 13C nucleus were acquired as the differences in chemical shifts between the oligosaccharide tagged with a paramagnetic ion, such as Tm3+, and that modified with a diamagnetic reference La3+ ion (Fig. 4.9b). Because the 3D structure of the oligosaccharide rapidly fluctuates in solution, the observed PCS data should be interpreted as averages over the dynamic conformational ensemble. Thus, we performed MD simulations in explicit water to create a vast conformational ensemble of this oligosaccharide. The simulation results showed wide distributions of torsion angles in this trisaccharide, particularly around a glycosidic linkage between the NeuAc and Gal residues (Fig. 4.10). From the computed trajectory of the trisaccharide over a total of 120 ns, 2000 conformers were sampled, which involved harmonic motions in a local minimum as well as transitions from one low-energy region to another in the energy landscape.

Fig. 4.9
figure 9

PCS measurements for the GM3 trisaccharide. (a) Introducing a paramagnetic lanthanide tag. (b) 1H–13C HSQC spectra of the tagged oligosaccharide in complex with Tm3+ (magenta) and La3+ (blue). Arrows indicate chemical shift differences for the anomeric CH groups of Gal and Glc residues (Adapted with permission from Ref. [46])

Full size image
Fig. 4.10
figure 10

Torsion angle density maps of the MD trajectory for the GM3 trisaccharide with a total simulation time of (a) 60 ns and (b) 120 ns and (c) the GM2 tetrasaccharide with a total simulation time of 120 ns (Adapted with permissions from Refs. [46] and [47])

Full size image

Using the atomic coordinates generated from this dynamic ensemble model, the conformational space of the GM3 trisaccharide derived from the MD simulations was quantitatively validated. With this approach, the experimentally obtained PCS values were compared with those that were back calculated from the ensemble model. The results for the back-calculated PCS values were in excellent agreement with the experimental data with a small Q value (used as criteria for agreement between experimental and calculated values), which indicated successful exploration of the conformational spaces of this oligosaccharide (Fig. 4.11). Using most single conformers or using a combination of selected low-energy conformers gave compromised Q values. Furthermore, such a low Q value was not estimated from a conformational ensemble of this trisaccharide without considering its low-populated (around 2 %) conformational cluster (Figs. 4.10 and 4.11). These results underscore the existence of the minor conformers in solution and their significant contributions to the observed NMR data.

Fig. 4.11
figure 11

Correlations between experimentally observed PCS values with Tm3+ and computationally calculated PCS data for the GM3 trisaccharide. The PCS values were back calculated from ensemble models derived from (a) the trajectory with a total simulation time of 60 ns and (b) those with a simulation time of 120 ns (Adapted with permission from Ref. [46])

Full size image

A unique structural feature of oligosaccharides is their branching with multiple modes of linkage, which provide diverse 3D structures. To examine the impact of these branches on the conformations and dynamics of oligosaccharides, paramagnetism-assisted NMR analyses in conjunction with MD simulations were also used to conformationally characterize a branched oligosaccharide, GM2 tetrasaccharide (GalNAcβ1-4[NeuAcα2-3]Galβ1-4Glc), which has an additional GalNAc branch as compared with the GM3 trisaccharide [47].

The PCS-validated conformational spaces of the GM2 and GM3 oligosaccharides showed similarities in their Gal–Glc glycosidic linkage conformations. In contrast, significant differences were observed between the GM2 and GM3 oligosaccharides in terms of the conformational spaces around the NeuAc–Gal glycosidic linkages (Fig. 4.10c). A corresponding cluster of this linkage to the most populated conformations in the GM3 trisaccharide was missing in the conformational space of the GM2 tetrasaccharide. This suggests that in the biantennary GM2 tetrasaccharide, the additional GalNAc residue controls the conformational fluctuation of another branch. The corresponding NeuAc–Gal glycosidic linkage has conformational freedom in the linear GM3 oligosaccharide.

4.4 Concluding Remarks

In this chapter, we outlined two contrasting types of NMR approaches for characterizing individual conformers of biomolecules that rapidly interchange with one another. In one approach, a specific conformer is enriched or even isolated for detailed NMR analyses. Our NMR studies of mono- and diUb illustrated this conformational trapping by applying high pressure, point mutations, pH optimization, and covalent constraints. Crystallographic snapshots can also be used to interpret NMR data for a dynamic conformational ensemble. In another more systematic approach, the dynamic behaviors of biomolecules in solution can be directly observed by NMR-validated MD simulations. This method can be applied to describe dynamic pictures of flexible biomolecules, including intrinsically disordered proteins. Thus, sophisticated NMR spectroscopic techniques combined with X-ray crystallography and MD simulations provide versatile tools for characterizing the conformational dynamics of biomolecules and exploring their conformational spaces.

In living systems, these flexible biomolecules are self-organized into highly ordered supramolecular structures and, more importantly, are autonomously disassembled to create higher functions. For these systems, NMR spectroscopy provides valuable microscopic information on the local conformational dynamics of biomolecules during the processes involved in biomolecular organization, which can be complemented by macroscopic information on the overall structures of these supramolecular complexes provided by solution scattering, electron microscopy, and other biophysical methods. In this context, NMR has important roles for exploring the dynamic ordering of biomolecular systems for the generation of integrated functions.