Advances in studying protein disorder with solid-state NMR

Abstract

Solution NMR is a key tool to study intrinsically disordered proteins (IDPs), whose importance for biological function is widely accepted. However, disordered proteins are not limited to solution and are also found in non-soluble systems such as fibrils and membrane proteins. In this Trends article, I will discuss how solid-state NMR can be used to study disorder in non-soluble proteins. Techniques based on dipolar couplings can study static protein disorder which either occurs naturally as e.g. in spider silk or can be induced by freeze trapping IDPs or unfolded proteins. In this case, structural ensembles are directly reflected by a static distribution of dihedral angels that can be determined by the distribution of chemical shifts or other methods. Techniques based on J-coupling can detect dynamic protein disorder under MAS. In this case, only average chemical shifts are measured but disorder can be characterized with a variety of data including secondary chemical shifts, relaxation rates, paramagnetic relaxation enhancements, or residual dipolar couplings. I describe both technical aspects and examples of solid-state NMR on protein disorder and end the article with a discussion of challenges and opportunities of this emerging field.

1. Introduction

The importance of intrinsic protein disorder for protein function has been widely accepted [1, 2, 3]. By one estimate at least half of the proteins in eukaryotes have long intrinsically disordered regions (IDRs) [4]. Because intrinsic disorder gives proteins the ability to form molecular interactions with high specificity, low affinity, and high promiscuity [5], it plays important roles in transcription factors [6], the nuclear pore complex [7], protein phase separation [3] etc. In addition, IDRs can function as entropic chains [8]. Intrinsically disordered proteins (IDPs) and IDRs are important in several diseases [2] but are complicated drug targets due to their undefined structure [9]. Due to their high accessibility, IDPs and IDRs are major sites of posttranslational modifications (PTMs), which can regulate their function [1]. IDRs and IDPs can only be described by an ensemble of conformations rather than a single structure. This ensemble is relatively difficult to study with traditional methods of structural biology. Consequently, IDRs and IDPs are less studied compared to their globular, well ordered counterparts. This lack of information about what is sometimes referred to as the dark proteome [10] motivates the development of new methods to characterize the structural ensemble of IDPs and their interaction with other cellular components.

Every phenomenon requires data from multiple methods to be characterized comprehensively. This is especially true for IDPs where no single method alone can reliably define the structural ensemble. There are several low resolution techniques that identify IDRs and IDPs such as circular dichroism (CD) [11] and Fourier transform infrared spectroscopy (FTIR) [12]. In addition Förster resonance energy transfer (FRET) [13, 14], small angle X-ray scattering (SAXS) [15], and size-exclusion chromatography (SEC) [16] can characterize the compactness of an IDP.

Because it is inherently difficult to study protein disorder with crystallographic methods or cryo electron microscopy (cryo EM), NMR has proven to be indispensable [17] to study IDPs with atomic resolution especially when combined with computational methods [18]. The power of NMR to study IDPs is illustrated by its application to α-synuclein a model IDP with an important role in Parkinson’s disease. The NMR resonance assignment of soluble α-synuclein and the comparison with random coil chemical shifts identified its residual secondary structure [19]. Measurements of residual dipolar couplings (RDCs) provided long-range orientational information [20]. Measurement of paramagnetic relaxation enhancements (PREs) detected transient intermolecular interactions [20, 21]. The interpretation of these data using molecular modeling led to the equivalent of a structure of soluble α-synuclein i.e. an ensemble of structures that explained all experimental data [22, 23]. Solution NMR in combination with electron paramagnetic resonance (EPR) could even characterize α-synuclein in cells [24] and its partially ordered state in micelles [25].

Intrinsic protein disorder is equally important for proteins that do not tumble freely in solution and are, therefore, not accessible to solution NMR. These include membrane proteins, large molecular complexes e.g. as part of the cytoskeleton, chromatin, viral particles, protein aggregates, and some proteins that undergo phase separation. Here α-synuclein is again a good example because it is difficult to study with solution NMR when it interacts with lipid bilayers or aggregates into cross-β (a.k.a. amyloid) fibrils. Another limitation of solution NMR is that its chemical shifts only capture the dynamic average of the conformational states that the IDP accesses during the measurement. This averaging leads to underdetermined datasets that are difficult to interpret in terms of structural ensembles.

Solid-state NMR has the potential to address all of these shortcomings and has already proven to be an ideal tool to observe and characterize protein disorder in insoluble protein systems. Similar to solution NMR, solid-state NMR can be used to study dynamic protein disorder, i.e. an ensemble of structures in which every IPD or IDR undergoes rapid conformational change in the time scale of the experiment (fast exchange). However, solid-state NMR can also be used to characterize static disorder i.e. an ensemble of conformations in which the molecules do not change conformation during the experiment (slow or no exchange). This difference of dynamic and static disorder is certainly not bi-modal and there is a continuum between these two extremes. In addition, the definition of dynamic and static disorder also depends on the time scale of the specific experiment.

But are these IDRs really worth studying? Is not the structure of the cross-β core key for understanding fibril toxicity and function? Are not selectivity filters of ion channels located in their transmembrane domains? In the case of fibrils formed by huntingtin exon-1 (HTTex1), recent findings from my lab and others suggest that conformational preference of the dynamic proline rich domain (PRD) is more important for the ability of these fibrils to seed, interact with binding partners, and for their toxicity [26, 27, 28]. For ion channels, intrinsically disordered domains are e.g. important for voltage dependent gating [29, 30]. Therefore, developing new NMR methods to study IDRs in the solid state is important.

In the following, I will discuss solid-state NMR approaches to study protein disorder. First, I will focus on techniques that study static disorder followed by approaches to study dynamic disorder. I will end this article with a discussion of current challenges and exciting opportunities for future applications of solid-state NMR on protein disorder.

2. Solid-state NMR methods to investigate protein disorder

2.1. Static protein disorder

Static protein disorder is characterized by little to no molecular reorientation on the time scale of the experiments. This is often referred to as the slow (chemical exchange) limit in which all conformations are observed by individual resonances [31]. Another consequence of static protein disorder is little motional averaging of dipolar couplings, chemical shift anisotropy (CSA), or quadrupolar couplings making these systems accessible to solid-state NMR experiments that rely on magnetization transfer based on dipolar couplings.

Static disorder is commonly encountered in synthetic polymers and approaches from material science can be adapted to study static protein disorder. A good example of such an adaptation is the work of van Beek and coworkers on the structure of dragline spider silk using static solid-state NMR experiments. They used static double-quantum/single quantum (DOQSY) experiments developed for polymers [32] to calculate the relative orientation of ¹³C carbonly CSAs in silk. They simulated DOQSY spectra for an array of dihedral angle combinations (ϕ, ψ) to fit experimental data resulting in a probability density of dihedral angles P (ϕ, ψ) (see Figure 1) [33, 34]. In addition, they applied the static DECODER experiment, originally developed to measure orientational distributions in polymers [35], to selectively ¹³C labeled spider silk. These data determined the distribution of peptide chain orientations towards the fibril as described by the probability density P (α_F ,β_F ) [34].

Figure 1:

DOQSY spectra can determine the distribution of dihedral angles in silk. a) Experimental 2D DOQSY spectra of 1-¹³C alanine-labeled N. edulis silk. b) best fitting theoretical DOQSY spectrum. c) Resulting probability density P (ϕ, ψ) describing dihedral angle distribution. (Adapted with permission from Ref. [34]. Copyright (2002) National Academy of Sciences.)

Dynamic protein disorder can be studied as static disorder in frozen solution. The fast conformational exchange an IDP undergoes in solution results in average chemical shifts (fast chemical exchange). If this conformational exchange is arrested by freezing, the sample transitions from the fast to the slow exchange limit (see Figure 2). In the slow exchange limit, every conformation gives rise to its own NMR spectrum. Averaged over the whole ensemble, this will lead to NMR spectra that reflect all structures present in the ensemble. In other words, frozen solution solid-state NMR observes a chemical shift distribution that reflects the conformational heterogeneity found in the frozen state. However, this is only true if freezing leaves the structural ensemble of the solution unaltered.

Figure 2:

Principle of studying IDPs in frozen solution. Solution NMR can only investigate the dynamic average of the rapid conformational changes inherent to IDPs (Solution). In contrast, frozen solution solid-state NMR can directly observe the conformational ensemble of an IDPs (Frozen Solution). Here, not only an average chemical shift is observed but each conformation gives rise to its own set of chemical shifts. Artwork by Erin N. Johnson.

First applications of frozen solution NMR to protein disorder studied the equilibrium conformational ensemble of small peptides. Dios and coworkers studied the conformation of peptide T in frozen glycerol/water solution at 150 K using 2D ¹³C-¹³C exchange spectroscopy [36]. Heise and coworkers applied double quantum-zero quantum-double (DQ-ZQ) spectroscopy to the neurotensin peptide at 223 K without cryoprotectant [37].

The lab of Dr. Robert Tycko further developed frozen solution NMR into a tool to study protein folding. Initially, they used chemical denaturants to compare the structure of the folded villin headpiece domain (HP35) to increasingly unfolded states of HP35 [38, 39]. For these experiments, samples were frozen relatively slowly in liquid nitrogen. Later, they went beyond studying conformational ensembles at equilibrium and developed methods to rapidly freeze trap proteins during folding. Using this approach, they studied the refolding of heat denatured HP35 [40]. Recently, they combined rapid freezing with the increased sensitivity of dynamic nuclear polarization (DNP) to study the refolding of pH denatured melittin (see Figure 3a&b) [41].

Figure 3:

Examples of frozen solution to study protein folding and protein disorder. a) Apparatus for rapid mixing and freeze-trapping based on a rotating copper plate immersed in a liquid-nitrogen. A jet of mixed solution freezes on copper after traveling a variable distance.b) DNP-enhanced, DQ-filtered ¹³C solid-state NMR spectra of frozen melittin solutions with the indicated refolding times. c) DNP enhanced 2D ¹³C-¹³C spectrum of selectively ¹³C-labeled -synuclein monomers. Intra- and interresidual crosspeaks involving extended or α-helical valine residue are highlighted in blue and orange, respectively. Panel a and b adapted with permission from Jeon et al. 2019 [41]. Panel c adapted with permission from Ref. [46], copyright (2018) Elsevier.

Frozen solution solid-state NMR has an intrinsic low signal-to-noise ratio (S/N) because of its relatively broad lines and comparably low protein concentrations. The use of DNP to overcome this low S/N is an obvious choice because DNP is most effective in frozen, glassy solutions i.e. in the same conditions most frozen protein NMR is done. The use of DNP solid-state NMR to study protein disorder and cross-β fibrils has been nicely reviewed by König and coworkers [42]. Besides studying the refolding of melittin [41], frozen solution DNP has been used to study the disorder of a signal peptide inside the ribosome [43], soluble amyloid-β (Aβ) [44], the folding intermediates of Aβ[45], and soluble α-synuclein and the disordered framing sequences of α-synuclein fibrils (see Figure 3c) [46]. While DNP solves the problem of low S/N, the relatively broad lines observed in frozen solution spectra currently limits their resolution. Therefore, most frozen solution NMR on protein disorder has been done on small peptides and selectively labeled samples. However, going to higher magnetic fields or using higher dimensional spectroscopy [47] will allow the study of fully isotope labeled samples. Another problem of frozen solution NMR is the large background signal from cryo protectants such as glycerol, especially in 1D spectra. ¹³C depleted glycerol [44] or double quantum spectroscopy can help remove natural abundance ¹³C background [36, 40, 46, 41].

Frozen solution NMR can in principal also be used to characterize small structural fluctuations in ordered, globular proteins. These small fluctuations are the origin of line broadening generally observed for non-crystallien proteins at low temperatures [48]. While this broadening creates a problem for current efforts to obtain high-resolution protein DNP spectra, it is simply a reflection of small structural fluctuations inherent to all proteins. These fluctuation are highly correlated with the dynamics of the protein hydration shell. Therefore, proteins undergo glass transitions together with their hydration shells at low temperatures [49, 48].

But how can broad NMR resonances in frozen solution be analyzed? In other words, how can the distribution of chemical shifts be translated into a conformational ensemble? Heise and coworkers [37] used Monte-Carlo (MC) simulations to create a structural ensemble from which they predicted chemical shifts using the program SHIFTX [50, 51]. The ensemble of structures was then reduced by removing structures whose predicted shifts did not match the experimental data (similar to the ASTROIDS approach in solution NMR [52]). In addition, Heise and coworkers ran molecular dynamics (MD) simulations and again used chemical shift prediction to relate the conformational ensemble from MD to the experimental data. Finally, they applied principle component analysis to compare the conformational ensembles from the MC and MD approaches [37]. A similar MD approach was used by Uluca and coworkers to analysis the frozen solution NMR spectra of α-synuclein [46]. Havlin and coworkers analyzed 2D ¹³C-¹³C spectra of partially folded HP35 by fitting them to linear combinations of 2D spectra of the folded state, the fully unfolded state, and partially denatured mixtures of HP35 fragments. Using this method, they were able to show that the unfolding of HP35 did not follow a simple two state model [38]. Similarly, Jeon and coworkers used principle components to analyze a series of 2D ¹³C-¹³C spectra of frozen mellitin solutions captured at different refolding times. The number of significant components addressed the question whether folding was a two state process or involved intermediates [41]. In addition to chemical shift analysis in 2D ¹³C-¹³C exchange spectra, Hu and coworkers also determined distributions of ϕ, ψ and angles by measuring the correlation of CSA tensors (2DEXMAS), ¹³C-¹³C dipolar couplings (CTDQFD), and CSA dependent DQ dephasing (DQCSA) of adjacent carbonyls. Simultaneous fitting of these data to several models of ϕ, ψ distributions allowed them to describe the unfolding of HP35 [39].

2.2. Dynamic protein disorder

Dynamic protein disorder is characterized by rapid conformational changes on the time scale of the experiment. In the fast exchange limit chemical shifts represent an average of all conformations that were visited during the experiment. In addition, these motions dramatically reduce anisotropic NMR interactions such as dipolar couplings and the CSAs. Consequently, methods that observe dynamic disorder in non-soluble protein samples need to combine J-coupling based magnetization transfers, common in solution NMR, with magic angle spinning (MAS), which is still required to prevent line broadening due to magnetic susceptibility effects. On the other hand, J-coupling based transfers, as e.g. a refocused INEPT, do not work in the presence of sizable dipolar coupling because these cause magnetization loss during transfer. This is especially true for techniques involving protons in fully protonated systems and at MAS frequencies below about 60 kHz where ¹H-¹H couplings lead to a rapid decay of transverse ¹H magnetization. Because of this limitation, J-coupling based transfers also function as filters for dynamic domains. In a sample that has both static domains and dynamic disorder, the former can be observed with dipolar transfers, the latter only with J-based transfers (see Figure 4). These filters can also be used to distinguish proteins in solution from proteins in the solid state [53]. Background and applications of these filters were recently reviewed by Matlahov & van der Wel [54] and Gopinath & Veglia [55].

Figure 4:

Spectral separation of static and dynamically disordered domains in solid-state NMR. Center: bottle brush model of HTTex1 fibrils as an example of a protein with a static domain (fibril core, shown in blue and orange) and a dynamic domain (Pro-rich bristles, shown in green). Left: The static fibril core can be detected with the dipolar based ¹H-¹³C CP experiment resulting in the spectrum shown in blue. Right: dynamically disordered domains can be detected with the J-based, ¹H-¹³C refocused INEPT experiment resulting in the spectrum shown in green.

Solid-state NMR spectra of IDRs in membrane proteins and cross-β fibrils were first recorded using this ¹³C detected, refocused ¹H-¹³C INEPT filter. Andronesi and coworkers did a pioneering and comprehensive study of the transmembrane polypeptide phospholamban (AFA-PLN) in DMPC bilayers using both dipolar and J-coupling based methods. They used a set of multidimensional J-based, ¹³C detected experiments to assign the dynamic domains that are located outside the membrane. They assigned the transmembrane helix using multidimensional experiments based on dipolar transfers [56] (see Figure 5b&c). Heise and coworkers showed that 1D ¹³C refocused INEPT spectra could detect the IDRs framing the core of α-synuclein fibrils [57] (see Figure 5a). When I was a graduate student, we detected similar dynamic framing sequences in the HET-s prion protein fragment 218–289 [58] (See Figure 5d). Using both ¹H and ¹³C detected methods, we showed that these regions were distinct from the fibril core assigned previously [59, 60]. To show that these disordered domains were attached to the static fibril core rather than detached protein fragments, we measured diffusion coefficients with the LED pulse sequence [58] (see Figure 6f).

Figure 5:

Examples of insoluble protein systems with dynamic disorder studied with solid-state NMR. a) ¹³C spectra of α-synuclein fibrils recorded with several methods to generate initial ¹³C magnetization as indicated. The refocused INEPT is selective to dynamically disordered domains found adjacent to the core [57]. b) ¹H-¹³C 2D INEPT HETCOR spectrum selective for the disordered domains of AFA-PLN in DMPC bilayers that are located outside the membrane. Site specific assignments are indicated. c) Structural model of AFA-PLN in DMPC bilayers illustrating the dynamic disorder of the cytoplasmic N-terminus [56]. d) ¹H-¹H TOCSY spectra detect the highly flexible regions of HET-s(218–289) fibrils. Amino acid type assignments are indicated [58]. e) AFM images of 17-mer nucleosome arrays used for the NMR spectra shown in f). f) ¹H-¹³C 2D INEPT HETCOR spectrum of 17-mer nucleosome arrays containing ¹⁵N-¹³C-labeled H3. Amino acid type assignments are indicated [61]. g) Overlay of 2D ¹³C-¹³C CP-DARR and INEPT-TOBSY spectra of mature pHP1α droplets. The CP-DARR spectra detect the more static regions in these droplets whereas the INEPT-TOBSY is selective to the dynamically disordered domains. Amino acid type assignment are indicated [62]. Panel a adapted with permission from Ref. [57]. Copyright (2005) National Academy of Sciences. Panel b and c adapted with permission from Ref. [56]. Panel d adapted with permission from Ref. [58]. Panels e and f adapted with permission from Ref. [61]. Copyright (2005, 2006, 2013) American Chemical Society. Panel g adapted with permission from Ref.[62]. Copyright (2019) John Wiley and Sons.

Figure 6:

Techniques to characterize dynamic protein disorder in the solid state. a) ¹H detection can increase S/N. HSQC spectra of HTTex1’s dynamic C-terminus. Assignments are shown and linewidths of A106 illustrated. b) Site specific assignment of IDRs using out-and-back type experiments from solution NMR. Strip plots of HNCA and HNcoCA spectra recorded of HTTex1 fibrils. Assignments are indicated. c) Site specific assignments deliver residual secondary structure via secondary chemical shifts (yellow) and degree of dynamics via peak intensity (cyan). Examples of HTTex1 fibrils are shown. d) Relaxation rates are sensitive to the timescale of dynamics. Examples of R₂ and R_1pcurves recorded on HTTex1 fibrils are shown on top and site specific R₂ rates determined by both approaches on the bottom. e) Residual dipolar couplings determine order parameters. Top: example of ¹H–¹⁵N REDOR curve recorded on HTTex1 fibrils. Bottom: site specific residual dipolar couplings. f) Diffusion measurements can confirm that signals are from an immobilized protein. Diffusion decay curves of flexible domains in HET-s(218–289) fibrils (circle) and of DSS as control (diamonds). g) PREs can detect transient long-range interactions of IDRs. PRE profiles of tau’s fuzzy coat with the spin label located in the fibril core (left). Transient interactions of the N-terminus with the fibril core result in a decrease in intensity of the spin labeled sample compared to the diamagnetic control (right). Panel a-e adapted with permission from Ref. [72]. Panel f adapted with permission from Ref. [58]. Copyright (2006, 2018) American Chemical Society. Panel g adapted with permission from Ref. [65]. Copyright (2011) John Wiley and Sons.

Dynamic domains framing a fibril core have been observed in many cross-β fibril structures such as Ure2p [63], PrP(22–144) [64], tau [65, 66], RIP1-RIP3 [67], ApCPEB [68], Sup35 [69], human parathyroid hormone [70], HTTex1 [71, 27, 72], Orb2 [73, 74], and FUS [75]. Similarly IDRs have been detected in multiple membrane proteins besides AFA-PLN [56, 76, 77]. Namely the neuropeptide Y receptor type 2 [78], NADPH-cytochrome P450 oxidoreductase [79], Anabaena sensory rhodopsin [80], cytochrome c bound to vesicles [81], and membrane bound α-synuclein [82]. Other examples of dynamic disorder in non-soluble IDRs are the N-termini of histones which are major sites of PTMs important for chromatin regulation. Gao and coworkers used 2D refocused INEPT spectra to investigate the conformational flexibility of histone H3 and H4 tails in reconstituted nucleasome arrays [61] (see Figure 5e&f). Xiang and coworkers used both dipolar and J-based methods to study the rigid and dynamic domains of histone H2A or H3 and their interaction with a nucleosome-binding peptide in sedimented nucleasomes [83]. Kashefi and coworkers applied refocused INEPT spectra to detect large dynamic domains found in the native-like functional array of a bacterial chemotaxis sensor [84].

Protein liquid–liquid phase separation is a mechanism underlying the formation of membrane-less organelles in cells. This phase separation is mediated by multiple weak interactions of often disordered protein regions that have adhesive sequence elements. Over time, phase separated protein droplets can mature i.e. transition from a state in which the proteins are mostly soluble in nature to a more vitrified state, sometimes followed by fibril formation [85]. Depending on the specific protein and on the maturation state of the protein droplet, they are amenable to either solution or solid-state NMR methods. Ackermann & Debelouchina investigated the liquid protein droplets formed by the heterochromatin protein 1α (HP1α). They used both dipolar and J-based methods that allowed them to monitor the increase of static protein domains and decrease of dynamic protein disorder with droplet maturation [62] (see Figure 5g). Damman and coworkers used both solution NMR and dipolar and J-based solid-state NMR methods to study the effects of phase separation and RNA interaction on the enhancer of decapping 3 (Edc3) protein [86]. Using the histidine-rich squid beak proteins (HBPs) as a model and solution NMR and dipolar and J-based solid-state NMR, Gabryelczyk and coworkers determined the molecular interactions that govern phase separation and vitrification [87].

Finally, J-coupling based transfers can even detect dynamically disordered protein regions in in-cell NMR under MAS [88]. Renault and coworkers used this approach to detect dynamic domains of the integral membrane protein PagL in E.coli [89]. Arnold and coworkers investigated dynamic protein regions, lipids, and sugars using INEPT filtered spectra of whole microalgea [90, 91].

From a technical perspective, dynamic protein disorder is readily detected using 1D ¹³C refocused INEPT spectra. These can be compared to cross polarization (CP) spectra detecting the more static domains (see Figure 4). Direct excitation (DE) spectra detect the entire sample irrespective of dynamics when recycle delays are in the order of the longest ¹³C T₁ in the sample. However, short recycle delays can make the DE experiment selective for dynamic regions [92, 68, 93, 94]. One dimensional refocused INEPT spectra are easily expanded to 2D ¹H-¹³C HETCOR spectra in which individual amino acids types can be identified. This approach is often sufficient to assign signals to a specific protein region or domains based on amino acid composition [58, 95, 61, 71, 73]. Kashefi & Thompson also compared the integral of INEPT HETCOR cross peaks to the amino acid composition of different domains and used a deletion mutant to locate the dynamically disordered regions of a bacterial chemotaxis sensor [84]. Site specific assignments are needed to investigate IDRs at atomic resolution. The assignment can be accomplished with a set of J-based methods that rely on ¹³C detection for which many solid-state NMR probes are optimized [56, 96]. ¹³C detected methods also have the advantage that no water suppression is required, which facilitates the detection in MAS probes that generally do not have gradient coils needed for modern solution NMR water suppression techniques. Finally, ¹³C detected methods have gained popularity in solution NMR of IDPs because of the large ¹³C chemical shift dispersion and the absence of exchange broadening compared to NH detected methods [97, 98, 99]. Nevertheless, ¹H detected methods have the advantage of higher S/N specifically if exchange broadening is not an issue and if MAS probes are optimized for ¹H detection [58, 80, 74, 77, 76]. A downside of specifically out-and-back type, ¹H detected methods that rely on amid proton is their inability to detect Pro residues and, because of the additional transfer step, their reduced sensitivity for regions of limited dynamics. Although ¹H linewidths of IDRs attached to immobilized protein are generally larger than in solution, 3D ¹H detection out and back type solution NMR experiments are still possible. This is because transverse ¹⁵N and ¹³C relaxation, which is most important for the efficiency of these methods, is less affected in these domains compared to ¹H relaxation. The assignment of IDRs can be facilitated by comparison to solution assignments. Bibow and coworkers assigned the dynamically disordered domains of tau fibrils (also known as the fuzzy coat) using HSQC and HNCA spectra and the comparison to solution assignments [100, 65]. Gath and coworkers assigned the dynamic domains framing the core of α-synuclein fibrils using the same approach [101]. Even de novo assignments of IDRs are possible in the solid-state. Using ¹H detected HNCA, HNCO, and HNcoCA spectra in combination with ¹³C detected CTUC-COSY spectra [102], we assigned the dynamically disordered regions of Orb2A and HTTex1 fibrils site-specifically [74, 72] (see Figure 6a&b).

With site-specific assignments in hand, IDRs can be characterized in more detail. Bibow and coworkers determined the variability in dynamics in tau’s fuzzy coat via the intensity of HSQC cross peaks. They also measured residue specific ¹⁵N R₂ and R_1p rates which showed that most regions experienced fast motions except for regions close to the fibril core and the N-terminus which had motions in the μs to ms range. Additionally, they measured PREs to directly probe the interaction of the tau fibril core with the fuzzy coat [65] (see Figure 6g). Using a similar approach, we determined the increase in dynamics of the PRD of HTTex1 fibrils towards the C-terminus via HSQC peak intensities and R₂ and R_1p relaxation rates [72] (see Figure 6c&d). All these studies show that ¹H-¹H couplings in IDRs can be averaged to a degree that these regions can be detected at modest MAS frequencies and without the use of perdeutation. But to which extent are dipolar couplings averaged in IDRs that are still attached to static proteins (e.g. fibril or membrane protein) and that are not tumbling freely in solution? To answer this question, we measured the residual dipolar couplings in the PRD of HTTex1 using a site-specific ¹H-¹⁵N REDOR experiment [103]. We showed that residual dipolar ¹H-¹⁵N couplings of up to about 300 Hz permitted the detection of signals in an HSQC spectrum [72] (see Figure 6e).

3. Challenges and opportunities

A challenge of studying dynamic disorder in the solid state is that some domains cannot be detected using either dipolar or J-based methods. These are regions of intermediate dynamics in which 1) dipolar couplings are only partially averaged making both dipolar based and J-based methods inefficient and 2) intermediate exchange [31] results in severe line broadening. The first issue can be resolved with several approaches. One solution is to use direct excitation in combination with long dipolar mixing (e.g. spin diffusion) to obtain transfer over weak dipolar couplings [68, 93]. Another approach is the use MAS frequencies high enough to reduce the effective ¹H-¹H in a way that J-based methods become feasible for the entire protein irrespective of dynamics [104, 105, 106]. In this scenario, dynamically disordered regions are those that can be detected with J-based but not with dipolar based methods. Alternatively, homonuclear ¹H decoupling can be used inside INEPT type transfers [107]. Likewise, perdeuteration combined with selective ¹H labeling, either through back exchange or methyl labeling, can reduce ¹H-¹H couplings and enhance INEPT transfer efficiency [108]. Still, flexible domains or loops sometimes need a separate set of methods even at high MAS frequencies or when ¹H spin dilution approaches are used. Chevelkov and coworkers applied TROSY to perdeuterated samples of crystalline α-spectrin SH3 to detect only the narrow component of a NH multiplet similar to solution NMR [109]. Linser and coworkers then applied 3D TROSY techniques to assign these flexible loops [110].

Perdeutation and fast MAS do not solve the problem of intermediate chemical exchange that will result in line broadening in some cases so severe that the corresponding regions cannot be detected. In this case, methods that are not affected by exchange broadening are good alternatives. Continuous wave (CW) EPR spectroscopy, for example, results in signals throughout a wide range of correlation times allowing the study of IDRs in the solid state [111, 112, 113]. Static ²H NMR spectroscopy is another method to measure precise dynamics in proteins [114] and has already been applied to characterize side chain dynamics in Aβ fibrils [115]. One way to observe broad or weak resonances is based on the saturation of a detectable (i.e. intense or narrow) resonance that is in exchange with this ”invisible” resonance. In solution these methods are known as dark state exchange saturation transfer (DEST) [116] and chemical exchange saturation transfer (CEST). In solid-state NMR this principle known as Z-spectroscopy [117, 118]. However, these methods rely on a detectable resonance that is in exchange with the ”invisible” state which might not always be the case.

The general challenges of studying static protein disorder are resolution and low S/N. As mentioned above, both of these can be addressed by combining DNP with multidimensional spectroscopy or selective labeling. In addition, the refinement and commercialization of freezing methods will make frozen solution protein NMR broadly applicable to study protein folding and characterize structural ensembles at equilibrium.

An important next step in the analysis of IDRs with solid-state NMR is to turn spectroscopic information into structural ensembles. For static disorder this has already been done using either specific measurements of dihedral angle distributions [33, 34, 40] or the comparison of experimental chemical shifts to those derived from MD and MC approaches [37, 46]. For dynamic disorder where many additional parameters such as relaxation rates, dipolar couplings, and PREs can be measured, MD and MC approaches analogous to those used in solution NMR of IDPs [18] will be equally successful.

IDR have been long been understudied if not ignored. However in recent decades we became aware of how important IDRs are for protein function in general and in disease. With this awareness comes an increasing need to characterize their structural ensembles. But this is only the beginning because once structural ensembles are known, the effects of PTMs and their (transient) interaction with binding partners such as ligands, antibodies, or drugs can be studied. It is, therefore, an exciting time to develop, refine, and apply NMR methods to study protein disorder in the solid state.