Content

To overcome the inherent insensitivity of NMR in the solid state by hyperpolarization, dynamic nuclear polarization (DNP) has become the most promising approach to date (Barnes et al. 2008; Lee et al. 2015). It has been shown that DNP can drastically increase the sensitivity in solid-state NMR (ssNMR) experiments (Ni et al. 2013; Sauvee et al. 2013; Song et al. 2006) and it has been successfully applied to several biological systems (Becker-Baldus et al. 2015; Debelouchina et al. 2013; Fricke et al. 2015; Gupta et al. 2016; Kaplan et al. 2015; Koers et al. 2014; Potapov et al. 2015). Unfortunately, those experiments need to be carried out at cryogenic temperatures as the enhancement factor decreases with increasing temperature (Rosay et al. 2010; Sauvee et al. 2013; Zagdoun et al. 2013). At ambient temperatures, at which standard protein ssNMR experiments are conducted, the enhancement is negligible.

At usual DNP temperatures (~100 K) different conformations of the protein are frozen out. This aspect can result in severe line-broadening due to the overlapping peaks of the individual conformers which exhibit different chemical shifts (inhomogeneous line-broadening) (Linden et al. 2011; Siemer et al. 2012). Indeed, previous work has found a remarkable correlation between experimentally observed 13Ca line-width and protein backbone fluctuations as predicted from molecular dynamics (MD) simulations (Koers et al. 2014). Additionally, there is homogeneous line-broadening, which, for example, is caused by the proximity of the biradicals (Koers et al. 2014; van der Cruijsen et al. 2015). In the crowded NMR spectra of proteins (Shi et al. 2015), these effects are highly undesired and make the analysis of such spectra extremely difficult. Furthermore, broad lines lead to a loss of sensitivity, as the signal intensity at a given peak volume reduces with the line-width and thus the sensitivity advantage provided by DNP is lowered (Takahashi et al. 2012). This effect becomes more severe when recording higher-dimensional spectra, where the line-broadening “penalty” has to be paid for each dimension.

Improving the resolution is therefore a principal goal to fully exploit the advantages of the increased sensitivity conveyed by DNP. This can be achieved by searching for different sample preparation protocols rather than using the established glycerol matrix (Liao et al. 2016; Takahashi et al. 2013) and by searching for biradicals that allow reasonable enhancement factors at higher temperatures (Lelli et al. 2015; Perras et al. 2016).

Another option is to increase the magnetic field strength (Koers et al. 2014). A higher magnetic field would in principle not be expected to considerably reduce the line-width in case it is dominated by inhomogeneous broadening. For example, such a situation would occur in the presence of overlapping conformer peaks as they are distributed in terms of their chemical shift, which is directly proportional to the external magnetic field. Only the homogeneously broadened part scales with Hz and thus results in improved resolution at higher fields.

In samples where structural heterogeneity is not the limiting factor, it has been shown (Lopez del Amo et al. 2013) that the resolution at cryogenic temperatures improves considerably when increasing the external magnetic field strength. Here, we observed a considerable improvement in resolution when comparing two DNP 13C-13C proton-driven spin diffusion (PDSD) 2D NMR spectra of the same protein acquired at magnetic fields of 600 and 800 MHz (Fig. 1). The 13C-13C correlation via supercycled POST-C5 (SPC-5) (Hohwy et al. 1999) is shown in Fig. S1 (supplementary material).

Fig. 1
figure 1

Overlay of 13C-13C PDSD 2D NMR spectra of MxiH needles using DNP at 800 MHz (red, 8 kHz MAS, 95 K, 30 ms mixing time) and at 600 MHz (blue, 11 kHz MAS, 104 K, 20 ms mixing time). 45° phase-shifted squared sine bells were used as window functions for both spectra. Full line-widths at half height in the direct dimension (F2) are annotated for some resolved peaks illustrating the increase in resolution at 800 MHz. The peaks between 26 and 32 ppm (F2) and 50 ppm (F1) are present in the 800 MHz spectrum (see mirror peaks across diagonal), but are not visible here as they have lower intensity at this side of the diagonal

Full size image

The studied sample consists of T3SS needles formed by the supramolecular assembly of the 83-amino acid protein MxiH (Demers et al. 2014, 2013). The preparation is described and illustrated (Fig. S2) in the supplementary material.

In a previous study conducted at 600 MHz (Fricke et al. 2014), this system exhibited one of the best resolutions ever achieved for a protein sample under DNP conditions (Can et al. 2015; Su et al. 2015). Increasing the field strength to 800 MHz and using the biradical AMUPol (Sauvee et al. 2013), the observed line-widths at half height are about 22 % smaller, improving the overall resolution of the spectrum sizably. In the previous study based on the biradical TOTAPOL (Song et al. 2006), an enhancement of 23 was observed at 600 MHz. In the new study using AMUPol at 800 MHz, we obtained an enhancement factor of 21 (Fig. S3 in the supplementary material).

The fact that this protein assembly shows such well-resolved DNP spectra is most likely caused by the overall stiffness of the needle structure and the resulting reduced dynamics: A structurally nearly identical and homologous (>60 % sequence identity) needle protein (Demers et al. 2013) is PrgI which also forms T3SS needles. Due to this similarity, its dynamics will be used here to make deductions for the dynamics of MxiH needles. For such a PrgI needle sample (sparsely 13C, uniformly 15N-labeled) (Loquet et al. 2012) the 15N T 1 times were measured to be on average 106 s at 800 MHz 1H Larmor frequency (see Fig. 2 for T 1 times per residue, supplementary material for experimental details). This value is in contrast to other protein systems: For microcrystalline SH3 (uniformly 2H, 13C, 15N-labeled, 10 % back-protonated sample), the corresponding value is 26 s (900 MHz) (Chevelkov et al. 2007; Pauli et al. 2001). For microcrystalline ubiquitin (uniformly 2H, 13C, 15N-labeled, 30 % back-protonated sample), the reported value is 35 s at 850 MHz (average over all residues) (Fasshuber et al. 2015; Schanda et al. 2010). As can be seen, the T 1 relaxation times of the T3SS needles are about four times longer and therefore prove the remarkably low flexibility of the system. This results in a much smaller contribution of inhomogeneous line-broadening under DNP conditions and can serve as an explanation on why the T3SS sample yields good resolution at room temperature and at cryogenic temperatures alike.

Fig. 2
figure 2

15N T 1 times in the homologous T3SS needle with PrgI subunits. The corresponding secondary structure elements are illustrated on top. The averages over all residues are indicated for PrgI needles (average over shown bars), ubiquitin and SH3 (bars not shown) as dashed lines. Note that the residues facing inward (towards the needle lumen) have higher T 1 times (are less flexible) than the ones facing outward

Full size image

It has been shown that ssNMR is a useful tool for the investigation of protein dynamics (Ivanir-Dabora et al. 2015; Lamley et al. 2015; Lewandowski et al. 2015; Ma et al. 2015). In our previous study (Fricke et al. 2014), we found that the peaks of the residues located on the outer surface of the needle are more difficult to identify than those that face towards the needle lumen. This had been attributed to a lower flexibility of the inner residues due to the constraining effect of the needle structure. In contrast, the outer part of the needle has more freedom to move resulting in different conformations that interconvert at room temperature (and thus lead to a single chemical shift). They freeze out at low temperature leading to such severe line-broadening that those residues become unobservable. Another possible explanation of this observation could be that the biradical molecules do not enter the needle so that only the outer residues experience quenching by the proximity of the biradicals (Koers et al. 2014; van der Cruijsen et al. 2015).

To rule out the latter option and to lend further support to our hypothesis that this effect indeed stems from the differing degrees of dynamics at room temperature, ambient temperature 13C-13C PDSD 2D NMR spectra of the pure protein and of the DNP sample (containing the DNP matrix and the biradical) have been compared (Fig. 3). It can be seen that for all peaks the line-widths at half height are generally slightly higher in the spectrum of the AMUPol-containing sample. Nevertheless, it does not become apparent that the peaks of the outer amino acids (residues 1-39, the N-terminal helix) suffer from more severe line-broadening than the inner amino acids (residues 45-83, the C-terminal helix), as the comparison of the selected 1D-slices illustrates.

Fig. 3
figure 3

Overlay of 13C-13C PDSD 2D NMR spectra of MxiH needles at ambient temperature at 800 MHz and 11 kHz MAS. The blue spectrum represents the pure protein sample without any biradical (total measurement time: 13 h 30 min); the red spectrum is from the sample used for the DNP measurements (same total measurement time) and contains the biradical AMUPol (same sample as the one used to record the red spectrum in Fig. 1). 1D traces through the annotated peaks (black) are shown on the right. They compare the different peak widths at half height (FWHH) for the two NMR spectra. The green annotations represent peaks which have severely diminished intensity in the AMUPol spectrum in comparison with the biradical-free spectrum. The two valine peaks in the green box are shifted in terms of the Cβ chemical shift

Full size image

There are some peaks that have severely reduced intensity in the AMUPol sample (some annotated in green). These cases are located in the inner and outer parts of the assembly and do not follow any identifiable preference for residue location. The concerned amino acid types are hydrophobic so that it can be assumed that there is a certain interaction with the also slightly hydrophobic biradical. Shifted peaks could be found only for the βγ-correlations of V3 (green box). This residue may possibly form a direct interaction with AMUPol. A more flexible outer part of the needle is further corroborated by the shorter 15N T 1 times of the outward-facing residues in comparison with the inward-facing-residues in the homologous protein PrgI (Fig. 2).

In summary, we found that the already excellent spectral resolution of our macromolecular protein assembly increases significantly going from DNP at 600 MHz to DNP at 800 MHz. This is attributed to the overall very low flexibility of the structure which leaves homogeneous line-broadening as the main cause for the broad peaks at cryogenic temperatures—an effect that decreases with increasing field strength. For crowded protein NMR spectra, the resolution is a determining factor, so this increase is a substantial benefit. The lower DNP enhancement factor at higher magnetic fields (Mance et al. 2015) can therefore be a reasonable compromise as the information content in the spectra increases and the sensitivity loss due to line-broadening is reduced—an effect that becomes important when recording 3D spectra required for resonance assignment.

We also found that at ambient temperature the line-broadening due to the presence of the biradical is not significantly different for inner and outer helix residues. This supports the notion that the spectral differences between inner and outer parts of the needle observed using DNP at low temperature are not the result of the absence of biradical inside the needles, but rather represent true differences in conformational plasticity at room temperature.