Introduction

Amyloid fibrils are self-assembled protein aggregates that have been associated with numerous human disorders, including Alzheimer's and Parkinson's diseases (Ke et al. 2020; Dobson et al. 2020). On the other hand, some amyloid fibrils play important roles in certain physiological functions (i.e. functional amyloids) (Loquet et al. 2018a). For instance, functional amyloids play essential roles in hormone storage (Maji et al. 2009), bacterial curli formations (Chapman et al. 2002), or in regulated cell death mechanisms in fungi (Chapman et al. 2002; Loquet and Saupe 2017). Their 3D structures are crucial for elucidating the structure–function relationship of these systems at the molecular level. For example, the triangular beta-solenoid amyloid fold uncovered by solid-state NMR (ssNMR) in the fungal amyloid fibrils of HET-s (Wasmer et al. 2008) and HELLF (Daskalov et al. 2020) was demonstrated to be essential for amyloid cross-seeding and required to trigger a regulated cell death reaction in filamentous fungi (Daskalov et al. 2020). Due to their insolubility and lack of a crystalline order, amyloid fibrils still constitute very challenging targets for structural biology. Additionally, they often exhibit a remarkable degree of local structural polymorphism (Tycko 2014). High-resolution 3D structures of amyloid fibrils can be experimentally accessed by cryo-electron microscopy (cryo-EM) (Fitzpatrick and Saibil 2019) and magic-angle spinning (MAS) NMR (Daskalov et al. 2021; Meier and Böckmann 2015; Wel 2017; Jaroniec 2019). In contrast to cryo-EM, ssNMR techniques provide access to atomic-level information on dynamics at different times scales and site-specific polymorphism (Siemer 2020). However, one of the main disadvantages of ssNMR is its inherently low sensitivity, often leading to long measurement times to perform multidimensional experiments required for extracting distance restraints required for structure determination (Loquet et al. 2018b). The sensitivity issue can be circumvented by employing dynamic nuclear polarization (DNP) (Su et al. 2015; Biedenbänder et al. 2022), which is an NMR hyperpolarization technique that mediates polarization transfer from unpaired electrons or radicals to nuclei via strategic microwave irradiation. In MAS DNP experiments, gyrotrons, high-frequency microwave sources capable of generating 20–100 W of microwaves, are employed to irradiate samples at low temperatures (≤ 100 K) for efficient performances (Matsuki et al. 2016; Thurber et al. 2010; Li et al. 2021; Thurber and Tycko 2016; Sesti et al. 2018). To cryoprotect the precious biological NMR samples, glycerol- or trehalose-water mixtures are often used to homogeneously disperse the doped radicals across the amorphous sample (Takahashi et al. 2013). This hyperpolarization technique has paved the way for ssNMR to obtain important structural information on biosolids (including amyloid fibrils) (Jaudzems et al. 2019; Zhao et al. 2020; Maciejko et al. 2015; Marin-Montesinos et al. 2019; Gauto et al. 2021; Bayro et al. 2011; Debelouchina et al. 2013; Frederick et al. 2017; Lopez del Amo et al. 2013; Nagaraj 2016; Potapov et al. 2015; Deo et al. 2021) and inorganic materials that were previously inaccessible due to poor sensitivity (Rankin et al. 2019).

Although the theoretical maximum DNP enhancement factor ε is ~ 658 for 1H nuclei, an actual experimental gain of up to 200 ~ 300 were reported in actual MAS DNP experiments performed at ~ 9 T and ≥ 80 K at 14.1 T (Mathies et al. 2015; Lilly Thankamony et al. 2017; Conroy et al. 2022; Akbey et al. 2012; Menzildjian et al. 2021) The development of higher-field DNP instruments at 18.8 T and 21.1 T is motivated by the fact that high-field NMR spectroscopy offers a better chemical-shift dispersion (Lund et al. 2020; Felch et al. 2013; Blank et al. 2016), i.e., it allows NMR peaks to be better resolved and constitutes a crucial aspect for biological NMR applications. However, it is known that the main MAS-DNP mechanism, cross effect, has decreasing DNP performance (or lower ε) with increasing magnetic field (Mathies et al. 2015; Tan et al. 2019a). Although pulsed DNP techniques with field-independent performances have been suggested to circumvent this issue (Henstra et al. 1988; Tan et al. 2019b; Wili et al. 2022; Redrouthu and Mathies 2022), their stringent microwave power requirement has limited experimental demonstrations to low fields (≤ 1.2 T). An alternate strategy is to design a new class of DNP polarizing agents (or hetero-biradicals) that maximally exploit the cross-effect matching conditions at high magnetic fields (Mathies et al. 2015; Lund et al. 2020; Wisser et al. 2018; Berruyer et al. 2020; Halbritter et al. 2022) Among these novel biradicals, we will characterize the DNP performances of M-TinyPol, NATriPol-3, and SNAPol-1 radicals, which were demonstrated to yield efficient DNP performances (vide infra) (Zhai et al. 2020; Cai et al. 2021).

Molecular motions present in biological macromolecules average NMR interactions (such as chemical shifts) over different conformations and thus may lead to narrow lines. When such motions become frozen at low temperatures required for DNP experiments, the distribution of conformational states leads to significant inhomogeneous NMR line broadening or lower peak resolution (Bauer et al. 2017). However, it was reported in the literature such deleterious features are less severe for rigid biological molecules or small peptides (Fricke et al. 2014; Bahri et al. 2022; Gauto et al. 2019; Barnes et al. 2010; Barnes et al. 2009). For molecules exhibiting motions in the intermediate timescale, such as semi-rigid side chains of amino acids or mobile loop regions in proteins, it could be difficult to efficiently detect signals at room temperature using either dipolar- or J-based techniques (Ravotti et al. 2016; Colvin et al. 2015). This is because the dipolar couplings are too small (or well-averaged) for efficient cross-polarization (CP) transfer, while they could be still large enough to interfere with the coherences generated in the J-based Insensitive Nuclei Exchanged by Polarization Transfer (INEPT) (Ravotti et al. 2016; Colvin et al. 2015; Siemer et al. 2006; Heise et al. 2008; Jirasko et al. 2021; Ni et al. 2017; Gath et al. 2014; Wiegand et al. 2016; Björklund et al. 2013). Other possibilities that could make NMR peaks invisible include the interference between the dynamics of molecules with the MAS frequency and/or 1H decoupling Rabi fields (Ni et al. 2017; Maus et al. 1996).

Observing these semi-rigid loops and side chains by MAS DNP measurements would allow a more accurate determination of their properties. For instance, it will be advantageous to extract additional distance restraints from residues that are mobile and hidden at room temperature—but become more rigid at lower temperatures—for structure calculations (Jaudzems 2018). Lowering the temperatures to observe NMR peaks of these functionally important semi-rigid parts to slow down the deleterious intermediate dynamics coincidently improves DNP performances.

In this communication, we have performed MAS DNP experiments at a high magnetic field B0 = 18.8 T (800 MHz 1H Larmor frequency) using efficient hetero-biradical, SNAPol-1, to demonstrate unprecedented 13C signal enhancement ε of ~ 50 for amyloid fibrils. These protein assemblies exhibit a sufficient spectral resolution in 13C and 15N dimensions at cryogenic temperature, enabling the observation of highly sensitive 2D 15N-13C correlation experiments using the PAIN-CP transfer scheme. Our results pave the way to develop DNP-enhanced ssNMR approaches using 15N-13C correlation experiments to reveal signals of semi-rigid side chain that are otherwise hidden at room temperature. In addition, the experiments reveal new peaks that allow the extraction of structural information on less rigid protein segments in protein assemblies.

Results and discussion

We prepared a 13C, 15N-labeled HET-s(218–289) protein sample, aggregated into amyloid fibrils with 5 mM M-TinyPol radical in DNP juice, which is a mixture of d8-glycerol/D2O/H2O in a 6:3:1 by volume ratio. A DNP experiment was performed at 10 kHz MAS using a 3.2 mm sapphire rotor, temperature T ~ 100 K, and a magnetic field B0 = 18.8 T (see Experimental Section for further details). The 1H-13C CP spectrum showed an enhancement ε ~ 10 (Fig. S1). We have doped the NWD2 fibrillar sample (Daskalov et al. 2015) with NATriPol-3 radical, which exhibited a similar performance as M-TinyPol (ε ~ 6) (see Fig. S2). Note that Bahri et al. have recently reported a DNP study on amyloid fibrils of Aβ1-42 using the same radical (M-TinyPol) at the same magnetic field but with a faster spinning frequency of 40 kHz using a 1.3 mm rotor (Bahri et al. 2022), which yielded to a higher ε ~ 22, likely due to better microwave coupling, and, hence, higher electron Rabi field (Berruyer et al. 2020; Purea et al. 2019). Figure 1a shows a 2D 13C DARR spectrum of DNP-enhanced HET-s amyloid fibrils. We observed 13C linewidths for isolated peaks in the range of ~ 200–300 Hz. Note that the low-temperature DNP spectrum has 2–3 × broadened lines relative to the room-temperature spectrum (overlayed in Fig. 1a), which has reported linewidths for the same protein system of ~ 100 Hz (Siemer et al. 2005). The relatively moderate line broadening effect indicates a high structural homogeneity (or rigidity) of HET-s amyloid fibrils even at a temperature of 100 K. So far, the most narrow 13C linewidths for amyloid fibrils at cryogenic temperatures have been reported for the Aβ40 and Aβ42 proteins (perspectives for DNP 2013; Bahri et al. 2022), implying that both of these proteins retain a sufficient homogeneity at low temperature. Note that 13C linewidths as narrow as 30 Hz and 81 Hz could be achieved in amino acids and tripeptides, respectively, at cryogenic temperatures (Barnes et al. 2010; Barnes et al. 2009).

Fig. 1
figure 1

a 2D 13C-13C 20 ms DARR spectrum of HET-s(218–289) amyloid fibrils recorded with DNP at 100 K (blue), overlayed with a 50 ms PDSD spectrum acquired at room temperature (red). b 3D structure of HET-s(218–289) amyloid fibrils (PDB entry 2rnm), with aromatic and alanine residues showed with a stick representation and colored in red, the image of protein was genererated using PyMOL. c Excerpt from Fig. 1a with a lower base contour level showing the Cβ/Cα spectral region or alanine residues of HET-s(218–289) amyloid fibrils

Full size image

In our current study, we observed that 13Cα/Cβ cross-peaks for alanine residues in the fibrils were not uniformly DNP-enhanced and do not have the same linewidths (see Fig. 1c). In particular, the Ala228 Cα/Cβ peak has weaker signal intensity, but narrower linewidth compared to the Ala237, Ala247, and Ala248 peaks. This can be explained by the fact that Ala228 is pointing inside the hydrophobic core of the triangular beta-solenoid fold of HETs (Fig. 1b), and, therefore, less accessible to the radical-doped solvent. On the contrary, the other Ala residues are in the disordered regions, which are much more exposed to the radical-doped solvent. Therefore, these alanine residues might be more efficiently enhanced while undergoing higher structural heterogeneity explaining the larger linewidths. The peaks were assigned based on assignments obtained from previously measured RT experiments. Here, we emphasize that the DNP experiments not only provide sensitivity enhancement, but are also useful in assessing solvent accessibility of local protein residues. Such an information is sometimes not directly obtainable using conventional NMR spectroscopy.

It is important to note that NMR linewidths in proteins under MAS depend on multiple factors, i.e., magnetic field strength, MAS frequency, processing parameters, radical concentration, and site-specific hydration (Bauer et al. 2017; Jaudzems 2018; David et al. 2018). To document these observations and compare our NMR data obtained on HET-s amyloid fibrils doped with NATriPol-3, we have compiled detailed 13C linewidths of previously reported amyloid fibrils at low temperatures in Table 1. We note that the linewidths obtained in our experiments on HET-s are comparable to or even better than previously reported results in the literature.

Table 1 13C linewidths of different amyloid fibrils obtained at low temperatures reported in the literature
Full size table

HET-s in its fibrillar amyloid state forms a rigid β-solenoid spanning residues 226–246 (first repeat) and residues 262–282 (second repeat) (Melckebeke et al. 2010). In the DNP experiment, we observed additional 13C–13C correlations involving aromatic atoms (120–130 ppm spectral region in Fig. 1a) that were previously absent in the experiment recorded at room temperature.

To complement the observation of 13C sites, we used multidimensional 15N–13C spectroscopy to access 15N dimension. We have observed several cross peaks in the δ(13C) ~ 140 ppm and δ(15N) ~ 130 ppm region in Fig. 2 that corresponds to His residues. These His cross peaks could either be assigned to the 6-His residues from the His-tag at C-terminus of the sample, or His242 that is in the flexible loop region that connect the two beta-solenoid repeats. None of these His residues were previously observable in RT experiment. Thus, we have established an experimental set up that could allow the detection of the His residues that are otherwise not detectable at room temperature. Further experiments (i.e., a 3D NCC correlation experiment) would be required to perform an unambiguous sequential assignment of these residues.

Fig. 2
figure 2

2D 15N–13C NCa spectrum of HET-s(218–289) amyloid fibrils recorded with DNP at 100 K (blue), overlayed with spectrum recorded at room temperature (red). The * sign denotes the MAS spinning sidebands

Full size image

Notably, the 13C linewidths for aromatic residues have broadened marginally from ~ 120 Hz to ~ 150 Hz between the room-temperature and low-temperature DNP experiments. This implies a relatively high structural homogeneity for otherwise mobile sidechains. The 2D NCa spectrum (Fig. 2) exhibited quite broad 15N linewidths (~ 300 Hz), indicating that the 15N nuclei could be more sensitive to structural inhomogeneity due to cryogenic temperatures. A similar feature has also been observed for microcrystalline proteins at low temperatures (Gauto et al. 2019; Jaudzems 2018).

Interestingly, the 15N–13C correlation DNP spectrum revealed new peaks at 15N chemical shift below 100 ppm and 13C chemical shift at ~ 107 and ~ 157 ppm (labelled in Fig. 2), which we assigned as 15N side chain nuclei. By inspecting the reported Cα chemical shifts for the HET-s(218–289) protein at RT (Melckebeke et al. 2010), we have assigned these correlations and we established new N–C proximities, i.e., 236ArgNζ-Cδ, 274ArgNζ-Cδ, and 270LysNζ-Cε in the DNP spectrum (Fig. 2) that are helpful for sidechain assignment.

Next, we examined the DNP performance of using SNAPol-1, a trityl-nitroxide hetero-biradical that has recently demonstrated an ε ~ 110 on ubiquitin using a 3.2 mm DNP system at 18.8 T (Cai et al. 2021). We added 5 mM SNAPol-1 to 13C, 15N-labeled NWD2 amyloid fibrils, which share a similar beta-solenoid fold as observed in HET-s (Daskalov et al. 2015), and found that the SNAPol-1 sample yields a higher bulk 1H enhancement (Fig. S3, ε ~ 22) than the NATriPol-3 sample (ε ~ 8). Following that, we obtained a highly promising enhancement of ε ~ 50 in a 1H–13C CP spectrum using SNAPol-1 at 11 kHz MAS frequency (Fig. 3a). Hence, we conclude that SNAPol-1 is a more efficient polarizing agent for DNP experiments on amyloid fibrils. Note that the high ε ~ 50 includes an additional 10% DNP improvement achieved by performing a freeze–pump–thaw procedure, which is a known procedure that removes dissolved paramagnetic oxygen that compromises the DNP samples (see Sample Preparation) (Delage-Laurin et al. 2021; Kubicki et al. 2014). To the best of our knowledge, ε ~ 50 is the highest DNP enhancement factor achieved for amyloid fibrils at 18.8 T reported to date. We anticipate that even more efficient DNP performance could be obtained for the 1.3 mm DNP system due to a more efficient microwave transmission system (or higher electron Rabi field) (Purea et al. 2019). Besides, the use of a smaller rotor with 4 × faster (~ 40 kHz) spinning frequency could lead to fewer spinning sidebands, less spectral overlap, and narrower linewidths. We also noted different enhancement factors were obtained between the simple 1H direct pulse (ε ~ 22) (Fig. S3) and the CP experiment (ε ~ 50) (Fig. 3A). This is likely because that some 1H nuclei in the sample are less accessible to the radical-doped DNP juice, i.e., hydrophobic residues in the core of the fibrils (vide supra) or the pools of bound water pools enclosed by the fibrils (Bauer et al. 2017; Melckebeke et al. 2011; Fitzpatrick et al. 2013). Moreover, we expect this feature to be sample and solvent dependent (Marin-Montesinos et al. 2019).

Fig. 3
figure 3

a Comparison of 1D 1H–13C CP spectra of NWD2 amyloid fibrils with 5 mM SNAPol-1 recorded at 8 kHz MAS with microwave on (blue) and off (red) condition. Enhancement factor of ε ~ 50 was measured. b 2D NCa spectrum recorded with DNP at 100 K and 8.1 kHz MAS (blue), overlayed with a similar experiment acquired at 273 K and 11 kHz MAS (red). Both spectra were recorded at 18.8 T. c DNP-enhanced 15N–13C PAIN-CP (green) show long-range cross peaks that are absent in DNP NCa experiments (blue). The arginine and lysine peaks are not unambiguously assigned.

Full size image

The excellent signal enhancement allowed us to observe new peaks in the DNP-NCa spectrum (Fig. 3b). Based on their chemical shifts, these new peaks could originate from the side chains. These peaks were not observed in conventional room-temperature ssNMR experiments, probably due to unfavorable dynamics that interfere with the timescale of dipolar-based experiments such as CP or 1H-decoupling in MAS experiments. Based on studies reported in the literature, we speculate that the dynamics exhibited by these residues are likely to be in the 103–106 s−1 (or μs) regime (Ni et al. 2017; Björklund et al. 2013; Maus et al. 1996), which unsurprisingly interferes with the ~ kHz rf fields and MAS frequencies employed in solid-state NMR experiments. Hence, when frozen at low temperatures, the side chains are sufficiently rigid to allow efficient polarization transfer through dipolar couplings—they become observable in CP-based ssNMR experiments.

To ease the assignment process of these newly observed peaks and obtain new distance restraints, we set up a 2D 15N–13C PAIN-CP experiment (Lewandowski et al. 2007; De Paëpe et al. 2011), which has been demonstrated to be efficient for obtaining the long-range heteronuclear distance restraints necessary for structure determination (Debelouchina et al. 2013; Melckebeke et al. 2010; Colvin et al. 2016; Loquet et al. 2012). Nevertheless, it is known that setting up PAR/PAIN-CP experiments is challenging due to the sensitive matching conditions, i.e., the radiofrequency (rf) nutation frequencies applied on the 1H and 13C for PAR (and additionally on the 15N for PAIN-CP) have to be precisely configured for optimum performances (Lewandowski et al. 2007; Donovan et al. 2017a, 2017b). Hence, it is common to first optimize the PAR/PAIN-CP conditions on 13C/15N-labeled small peptides or model proteins before repeating the experiments on the larger and less sensitive target biomolecules. Such an optimization strategy is not ideal because the experimental parameters might be significantly different in non-crystalline proteins such as hydrated protein assemblies, where the sample heterogeneity, rf inhomogeneity, and larger chemical shift dispersion might impede the PAR/PAIN transfer efficiency.

While it is possible to optimize the PAR/PAIN-CP condition directly on the interested proteins with a 1D spectrum, it is difficult to assess the efficiency of long-range transfers that is usually weaker and difficult to observe under the presence of stronger but less important short-range contacts. Hence, it is easier to find a robust matching condition with a 2D experiment, despite being more time-consuming. Thanks to the sensitivity enhancement of ε ~ 50 bestowed by DNP, we can acquire a 15N–13C PAIN-CP spectrum (Fig. 3c) with a modest signal-to-noise ratio in only two hours. To our knowledge, a DNP-enhanced PAIN spectrum has not been demonstrated in the literature. This could be because the relatively low biradical concentration (5 mM) used here, compared to the ≥ 10 mM biradical concentration employed in the literature (Sauvée et al. 2013; Fricke et al. 2016), leads to longer spin-locked relaxation times, T1ρ, and hence more efficient PAR/PAIN-CP transfer (Corzilius et al. 2014). The resulting 2D 15N–13C PAIN-CP spectrum (Fig. 3c) shows numerous intra-residue correlations between backbone 15N nuclei and the side chain 13C nuclei. These signals are crucial to establish the amino-acid spin-system identification and correlating the signals observed in 2D 13C–13C experiments to those observed in the conventional 2D NCa experiment. In addition to 15N backbone correlations, the frozen temperature conditions under DNP boost 15N signals arising from side-chain nitrogen, typically from Nε of Arginine and/or Nζ from lysine (Fig. 3c). These residues have long side chains that keep an important mobility at room temperature, whose NMR signals can be uncovered by low-temperature DNP experiments. Note that similar correlations have also been observed for frequency-selective TEDOR experiments for bacteriorhodopsin using DNP (Bajaj et al. 2010).

Experimental section

Sample preparation

The 13C–15N labelled HETs(218–289), and NWD2 samples were prepared according to previously published protocols (Siemer et al. 2005). Both samples were mixed with either 5 mM M-TinyPol (CortecNet) or 5 mM SNAPol-1 radicals in DNP juice (d8-glycerol: D2O: H2O in a ratio of 6:3:1 by volume) and incubated for 1 h at room temperature (RT). The samples were shortly centrifuged in a benchtop centrifuge (≤ 2000×g) before being packed into sapphire rotors. We have used Vespel MAS drive caps instead of the vendor-recommended ZrO2 drive caps because the latter design has a segmented insert, which can be easily damaged if mishandled. Nevertheless, the Vespel drive caps become loose at low temperatures due to a different heat expansion coefficient from the sapphire material. Hence, we marked the inserts of the Vespel caps with a permanent marker (Sharpie) to improve the seal between the rotor and the drive cap. The DNP sample was then degassed with at least three freeze–pump–thaw cycles using a 3-way tap, and backfilled with Argon gas (Delage-Laurin et al. 2021; Tan et al. 2022). This procedure removes the paramagnetic oxygen, which reproducibly improves the DNP enhancement factor by ~ 10%. We could not implement the insert/eject cycle degassing method demonstrated in Kubicki et al. because the rotor is likely to be blocked in our Bruker MAS DNP probe at lower temperatures (Kubicki et al. 2014).

DNP NMR spectroscopy

The spectra were acquired on a Bruker 800 MHz spectrometer, with a wide-bore 18.8 T magnet equipped with a 3.2 mm HCN DNP probe at the temperature T = 100 K and at 8–10 kHz MAS frequencies. Note that the M-TinyPol and SNAPol-1 samples coincidentally yield the maximum positive cross-effect DNP performances at the same static magnetic field B0 for a fixed gyrotron microwave frequency. Hence, all DNP experiments were recorded at B0 = 18.796 T (1H Larmor frequency of 800.28 MHz) with a gyrotron microwave frequency of 526.965 ± 0.01 GHz. The gyrotron output power of 15 W (with 130 mA collector current in the gyrotron) was used in all DNP experiments. Although the commercial probe is rated for operation at ~ 10–12 kHz at ~ 100 K, we have performed most DNP experiments at 8 kHz MAS frequency, which is significantly less likely to have rotor crash events than the experiments at 10 kHz. The spectra were processed with TopSpin 3.1 (QSine = 3 apodization functions were applied for both dimensions in all 2D spectra) and analyzed with the CcpNmr 2.1 program (Vranken et al. 2005).

Conclusions

In summary, we have demonstrated the potential of DNP-MAS NMR in studying amyloid fibrils, particularly the semi-rigid protein segments and mobile side chains that are otherwise inaccessible in conventional room-temperature ssNMR experiments. We have obtained an enhancement factor of ε ~ 50 using SNAPol-1, which is the highest enhancement reported so far at 18.8 T for amyloid fibrils. Contrary to DNP studies reported on biomolecules, the 13C nuclei in HET-s(218–289) fibrils retain a modest NMR resolution at cryogenic temperatures (T ~ 100 K) in DNP experiments. The boost of sensitivity and the low-temperature environment helps elucidate new peaks usually unobservable at room temperature due to mobility in an unfavorable regime. Additionally, we demonstrated that a 2D 15N-13C PAIN-CP spectrum with excellent sensitivity could be obtained in less than 2 h. This approach allowed us to observe new correlations involving 15N side chain nuclei. The promising DNP experiments and results will be combined with selective labelling schemes to obtain unambiguous distance restraints of previously unassigned peaks in the future. This will be crucial for refining the 3D structures of biological macromolecules and gaining insights into important biological functions.