Optimal RAP sample

The RAP labeling scheme yields randomly protonated samples in a deuterated matrix. The de-gree of protonation can be adjusted by the relative amount of H2O in the M9 growth medium.

To determine the spectroscopic optimal protonation level, we presented a systematic analysis of sensitivity and resolution of different randomly protonated microcrystallineα-spectrin SH3 samples as a function of the MAS frequency. Not surprisingly, fast MAS spinning is most bene-ficial for sensitivity and resolution in¹H-detected INEPT based¹H,¹³C correlation experiments, due to the elongation of the effectiveT₂times for¹H and¹³C, respectively.

We found, that aα-spectrin SH3 RAP sample expressed from a M9 minimal medium with a H2O content of 15% to 25%, rotated at 60 kHz MAS, yielded the best compromise in terms of spectroscopic performance. For the 25% RAP sample, the relative sensitivity gain at 60 kHz MAS is on average ≈4.5 and ≈4 fold for backbone and methyl resonances, respectively (cf.

Figure 3.7A, page 42). The ratio of absolute peak intensities for a fully-packed 3.2 mm rotor at 20 kHz MAS (700 MHz) to a 1.3 mm rotor at 60 kHz (850 MHz) amounts to≈1.1-1.3, whereas

the ratio of the active sample volumes is on the order of≈20, indicating a significant saving of isotopically enriched protein material, using 1.3 mm rotors. The 3.2 mm rotor was packed by a benchtop centrifuge (≈40×10³ g), while the 1.3 mm rotor was packed by ultracentrifugation (≈135×10³g). This might account for a factor of 1-2 in the amount of material in the 1.3 mm rotor. In addition to reduced¹H dipolar dephasing at high MAS frequencies, this enhancement is also attributed to a higher efficiency of the 1.3 mm probe [Hoult and Richards,1976].

We performed 2D¹H,¹³C HMQC experiments at a MAS rotation frequency of 40 kHz, and determined residue-specific¹H line widths for a 5% and a 25% RAP sample (cf. Figure 3.12B, page 50). For the 25% RAP sample, the average ¹H line width amounts to 44-49 Hz. The line width for the 5% RAP sample is on the same order, which is supported by the¹HT₂ echo experiments, carried out at 60 kHz MAS (cf. Figure 3.7B, page 42).

The highest sensitivity was obtained for the 25% RAP sample rotated at 60 kHz (cf. Figure 3.7A, page 42). Under these conditions, dipole mediated line broadening was not yet outper-forming sensitivity and resolution. However, the use of a higher relative concentration of H₂O in the bacterial growth medium seems unfavorable due to an increase of the¹³CDH2 isotopomer, which results in additional resonances and, thus, in a decrease of resolution (cf. Figure 3.21, page 61).

In the course of the experimental series with increasing MAS frequencies, the samples became dehydrated. However, a sealing method was suggested, which enabled long-term measurements at spinning frequencies≥50 kHz. In total, we expect that, enhanced coherence lifetimes at fast rotation will facilitate solution-state like multi-bond experiments in the future, and allow scalar transfers even for weakly coupled spin systems [Linser et al.,2008,Schanda et al.,2009].

Recently, high-resolution DQ-²Hα,¹³Cα correlation spectra using perdeuterated α-spectrin SH3 were reported [Agarwal et al., 2009]. The observed resonances in the here presented

1Hα,¹³Cαcorrelation experiment usingα-spectrin SH3 RAP samples (cf. Figure 3.6B, page 41) match rather well the²Hα,¹³Cαcorrelation spectra. The experimental²H line width for the

2Hαresonances was (30±9) Hz. The¹Hαline width in the presented¹H,¹³C HMQC spectra of a 5% RAP sample amounted to (50±9) Hz at a MAS frequency of 20 kHz and an external mag-netic field of 14.1 T (600 MHz). Taking the 6.5 times smaller gyromagmag-netic ratio of deuterium

γ( H) to protonsγ( H) into account, the resolution in the H-detected spectrum is improved by a factor of about two. At the same time, the signal-to-noise is increased by a factor of

0.05× γ(¹H)^5/2

γ(²H)γ(¹³C)^3/2 ≈0.05×52≈3

(to first approximation, the prefactor of 0.05 takes account of the proton concentration of the 5%

RAP sample).

Figure 4.1:Combination of the RAP approach with back-exchanging of labile protons to determine¹H-detected high-resolution¹H,¹³C and¹H,¹⁵N correlation spectra, using the same sample. Here, for the¹H,¹³C spectrum (left, 14.1 T), however, a 5% RAP labeledα-spectrin SH3 sample was employed, while the¹H,¹⁵N spectrum (right, 16.4 T) was obtained, using a 25% RAP labeled sample, which was back-exchanged in a 30%/70% H2O/D2O buffer, respectively. For both samples the MAS frequency was adjusted to 20 kHz.

Sensitivity and resolution of resonances in the aliphatic region of¹H,¹³C HMQC spectra us-ing RAP samples were not compromised, even though the amount of protons had been increased in comparison to the previously presented approaches [Agarwal et al.,2006,Agarwal and Reif, 2008,Agarwal et al.,2008]. In contrast to an approach designed for the determination of high-resolution¹H,¹⁵N correlations, in which the exchangeable protons have to be partially replaced

with deuterons [Chevelkov et al.,2006,Akbey et al.,2010], the presented RAP labeling scheme does not require an¹H/²H exchange step. This will be of particular importance for the inves-tigation of membrane proteins, which have very stable amide protons, that might not exchange within months.

However, the RAP labeling can be combined with back-exchanging of labile protons, as stochastical incorporation of protons in both approaches reduces the probability of high pro-ton concentrations for a single protein molecule. In Figure 4.1 the¹H,¹³C HMQC spectrum of a 5% RAP sample was plotted (left) next to the¹H,¹⁵N correlation spectrum of a 25% RAP sample, back-exchanged in a 30%/70% H₂O/D₂O buffer (right). Even though both labeling schemes were employed at the same time, using relatively high proton concentrations, high-resolution spectra were obtained already at moderate MAS frequencies. In principle, the here presented analysis of sensitivity and resolution as a function of the RAP proton concentration and MAS frequency has to be extended by an additional dimension, accounting for the degree of back-exchanged protons. However, a 15% RAP labeled sample, back-exchanged in a 20%/80%

H₂O/D₂O buffer seems to be a reasonable compromise.

In conclusion, we could demonstrate that the presented RAP labeling scheme facilitates¹H detection of aliphatic resonances and opens a new avenue for biomolecular MAS solid-state NMR spectroscopy. The presented RAP approach is easy to implement and allows to bypass more complicated labeling schemes, which rely on selectively labeled precursors, such as

[1,3]-13C or [2]-¹³C glycerol [LeMaster and Kushlan,1996,Hong and Jakes,1999,Castellani et al., 2002], or which rely on dilution of exchangeable protons [Chevelkov et al.,2006].

However, we also investigated the possibility of RAP labeling, combined with sparse isotopi-cal¹³C enrichment using [u-²H, 2-¹³C]-glycerol as the carbon source, instead of u-[²H,¹³ C]-glucose. That way, a 10% RAP-glycerol sample ofα-spectrin SH3 was prepared, employing 10% H2O and 90% D2O in the M9 expression buffer. The recorded¹H,¹³C correlation spec-tra yielded excellent signal dispersions (Figure 4.2) due to the elimination of homonuclear J-couplings and the reduction of spectral crowding, as only every second carbon was isotopically

13C-labeled. In principle, for a full characterization of RAP-glycerol samples, the site-specific

1H concentrations remain to be determined, since glycerol enters the amino acid metabolism at

a different level than glucose [Lemaster and Cronan,1982]. These can be obtained by solution-state NMR experiments, which were already carried out here for the RAP-glucose samples.

To first order approximation, half of the¹³C resonances are attenuated, using [u-²H, 2-¹³ C]-glycerol as the carbon source. However, to detect the missing resonances, a complementary RAP sample could be prepared, employing [u-²H, 1,3-¹³C]-glycerol in the M9 expression medium.

An 1:1 mixed sample could be employed to obtain the full spectrum, but with improved resolu-tion, as ¹³C,¹³C scalar couplings are eliminated, which could be used for the determination of intermolecular contacts [Castellani et al.,2002,Loquet et al.,2010,2012].

To spectroscopically improve the ¹³C resolution by removing ¹³C,¹³C scalar couplings in uniformly ¹³C-labeled RAP-glucose samples, two approaches were probed. As known from solution-state NMR, constant-time experiments can be employed, in which¹³C,¹³CJ-couplings are refocussed during a constant evolution period [Vuister and Bax,1992]. We showed, that high MAS frequencies enable¹H,¹³C constant-time HSQC experiments in the solid-state using RAP-glucose samples, while employing 2-3 kHz low-power decoupling (cf. Figure 3.17B, page 55).

The experiment requires long¹³CT₂times, which are, however, limited for¹³Cαbackbone res-onances. For this case, we implemented a homonuclear¹³C’+¹³Cβ-selective scalar decoupling sequence, which can be employed in “real”-time during the¹³Cαevolution period. That way,

1Hα,¹³Cαcorrelations were obtained by a cross polarization (CP) based sequence at a moderate spinning frequency of 24 kHz. Spectra with a significantly improved backbone resolution were yielded (cf. Figure 3.16B, page 54).