Introduction

Solid-state Nuclear Magnetic Resonance (ssNMR) combined with Magic Angle Spinning (MAS, (Andrew et al. 1958)) has in the last years made significant progress to study complex biomolecular systems including membrane proteins (Lange et al. 2006a; Ader et al. 2008; Bajaj et al. 2009; Ahuja et al. 2009; Shi et al. 2009; Etzkorn et al. 2007, 2010; Cady et al. 2010) or protein assemblies (Heise et al. 2005; Andronesi et al. 2008; Wasmer et al. 2008; Poyraz et al. 2010; Sun et al. 2009; Kumar et al. 2010; Jehle et al. 2010). In parallel, methods have been devised to determine entire three-dimensional structures from a single (Nomura et al. 1999; Rienstra et al. 2002; Lange et al. 2005; Manolikas et al. 2008; Bertini et al. 2010a) or a few (Castellani et al. 2002) protein preparations. With increasing molecular size, spectral resolution becomes critical for several aspects of the structure determination process. To deal with these challenges, multi-dimensional correlation experiments have been proposed and more elaborate isotope labeling schemes have been used (See Renault et al. 2010 for a recent overview). Some of the latter approaches simplify the spectral analysis to detect certain protein resonances but the essential process of structure determination, i.e., polarization transfer via C–C, C/N-HH-C (Lange et al. 2002)), or C/N–H-C (Seidel et al. 2005; Paepe et al. 2008; De Paepe et al. 2011)) spin moieties remains largely unaffected. At the same time, protein deuteration that has long been recognized as a powerful tool for macromolecular structural analysis by solution-state NMR (Englander et al. 1996; Gardner and Kay 1998) has been introduced in ssNMR for resolution enhancement of 1H solid-state NMR (Pines et al. 1976; McDermott et al. 1992; Zheng et al. 1993). In the last years, such approaches have been optimized to further reduce 1H line width (Chevelkov et al. 2006; Zhou et al. 2007; Linser et al. 2011), establish structural constraints (Reif et al. 2001; Paulson et al. 2003; Reif et al. 2003; Zhou et al. 2007; Huber et al. 2011; Varga et al. 2007) and to characterize protein-water interactions (Bockmann et al. 2005; Lesage et al. 2006). However, increasing levels of deuteration compromise the prospects to probe structurally relevant proton–proton distance constraints, affect relaxation times and may be prohibited by reduced protein expression levels in complex biomolecules such as membrane proteins.

In the following, we show that fractional deuteration (Rosen et al. 1996; Shekhtman et al. 2002; Otten et al. 2010) which can be readily obtained during protein expression by the combined use of protonated precursors and D2O provides a route to reduce spectral crowding and enhances the prospects to detect long-range correlations in standard ssNMR correlation experiments on complex biomolecules. For our study, we produced a fractionally deuterated version of a chimeric potassium channel (KcsA-Kv1.3) for which ssNMR resonance assignments have been reported (Schneider et al. 2008) and which has previously been characterized by ssNMR in different functional states in a lipid bilayer environment (Ader et al. 2008; Ader et al. 2009a, b). We analyzed the residual level of protein protonation and the effect of fractional deuteration on (13C,13C) as well as (15N,13C) and (1H,13C) ssNMR experiments. We show that a substantial reduction in proton density in [1H/2H,13C,15N] labeled KcsA-Kv1.3 influences the polarization transfer dynamics in the context of chemical-shift selective transfer. Because the residual protonation level favours peripheral amino-acid positions, establishing side-chain resonance assignments as well as the detection of long-range contacts, in particular between aromatic and methyl side-chains, is facilitated.

Materials and methods

Preparation of [1H/2H,13C,15N] KcsA-Kv1.3

Following the work of Legros et al. (Legros et al. 2000), the pQE32 expression construct (Lange et al. 2006a) was transformed into E.coli strain M15 prep4. For protein production, E.coli cells were grown on a medium containing protonated glucose and D2O. Cultures were adapted from initially 33–99% D2O over 3 days on small scale shaker flasks containing M9 minimal medium. The final culture was tenfold diluted into the expression culture. Protein expression was induced at 25°C by adding 0.5 mM IPTG at OD600 = 0.9. Cells were harvested as soon as the stationary phase was reached (5–6 h after induction). The protein was purified from 10 L of expression culture as described before (Lange et al. 2006a, b). Reconstitution into Asolectin liposomes was performed as described (Lange et al. 2006a, b), with minor changes. Briefly, the buffer of the purified channel protein was exchanged against 50 mM Sodium Phosphate, pH 7.4, 100 mM Sodium chloride, 4 mM n-decyl-ß-Dmaltopyranoside (DM) (Calbiochem) using a HiPrep 26/10 desalting column (GE Healthcare). Asolectin from soybean (Fluka) was resuspended in the same buffer and added at a 100/1 Asolection/KcsA-Kv1.3 molar ratio. This suspension was incubated for 2 h at room temperature. Subsequently, detergent was removed with Calbiosorb Adsorbent (Calbiochem). Liposomes containing [1H/2H,13C,15N]-labeled KcsA-Kv1.3 protein were pelleted by centrifugation at 134,000 × g for 2 h at 4°C.

Solid-state NMR

All experiments were recorded using a 3.2 mm triple-resonance (1H,13C,15N) probe head at a static magnetic field of 16.4 T corresponding to 700 MHz proton resonance frequency (Bruker Biospin, Karlsruhe, Germany). 1H field strengths of 83.3 kHz were employed for 90° pulses and SPINAL64 (Ref. (Fung et al. 2000)) decoupling during evolution and detection periods. Initial (1H,13C) cross-polarization (CP) times were set to 80 μs to largely restrict polarization transfer to directly bonded 1H-13C spin pairs in (1H,13C) and (13C,13C) correlation experiments. For (15N,13C) correlation experiments, an initial (1H,15N) CP time of 800 μs was used. 15N-13C transfers were performed with a SPECIFIC-CP (Baldus et al. 1998) time of 3 ms, followed by homonuclear 13C-13C DARR (Takegoshi et al. 2001) mixing of 100 ms under 10.92 kHz MAS. (1H,13C) correlation experiments were conducted using frequency-switched homonuclear Lee-Goldburg (FSLG, Bielecki et al. 1989) decoupling at 83.3 kHz 1H field strength during the indirect 1H evolution period and the proton dimension was calibrated using regular (1H,13C) HETCOR spectra. For 1H-1H double-quantum (DQ) mixing, we employed the SPC5 recoupling scheme (Hohwy et al. 2002) with DQ excitation and reconversion times of 285 μs and a CP time as in the FSLG HETCOR experiments. To enhance rotational resonance recoupling for carboxyls and backbone Cα’s, MAS rates were set to 10.92 kHz or 8.5 kHz. Transfer among aromatic and aliphatic methyl side chains was enhanced using an MAS rate of 10 kHz. Mixing times ranged from 20 to 500 ms in (13C,13C) proton-driven spin-diffusion (PDSD) experiments. All experiments were performed with an effective sample temperatures ranging from +2°C to +7°C. Spectra were processed in Topspin (Bruker Biospin) and analyzed with Sparky (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco).

Assignment and structural analysis

SsNMR resonance assignments for KcsA-Kv1.3 in lipid bilayers were taken from Ref. (Schneider et al. 2008). Since KcsA-Kv1.3 only differs by 11 turret residues from the 4 × 160 amino acid KcsA channel (Schneider et al. 2008) and in line with previous ssNMR work (Ader et al. 2008; Schneider et al. 2008; Ader et al. 2009b), the structure of the closed-conductive state of KcsA-Kv1.3 must share essential structural features with crystalline KcsA. Therefore, we created a structural homologue of the KcsA-Kv1.3 channel in the closed conductive state using the crystal structure of full length KcsA (PDB ID 3EFF, Uysal et al. 2009). Intra- and intermolecular 13C-13C correlations were then predicted using the KcsA-Kv1.3 model with an upper distance cutoff of 5 Å & 6 Å and, at the same time, taking into account the residual protonation pattern identified from ssNMR experiments. With these cutoff parameters (which were varied between 4 and 8Å) we observed the best overall agreement between experimental data sets and predicted cross peak patterns.

Results

Identification of residual protonation pattern

To investigate the residual level of protonation of [1H/2H,13C,15N] KcsA-Kv1.3 in lipid bilayers, we compared a series of two-dimensional ssNMR experiments with previous solution-state NMR work (Rosen et al. 1996; Shekhtman et al. 2002; Otten et al. 2010) and amino-acid biosynthetic pathways (Nelson and Cox 2008). Firstly, we conducted a conventional (13C,13C) proton-driven spin diffusion experiment using a mixing time of 20 ms using short (Fig. 1a) and longer CP (Fig. 1b, black) times.

Fig. 1
figure 1

a (13C,13C) PDSD correlation spectrum recorded on [1H/2H,13C,15N] KcsA-Kv1.3 with a mixing time of 20 ms. b Overlay of (13C,13C) PDSD correlation spectra recorded on [1H/2H,13C,15N] (black, in Asolectin lipids) and [1H,13C,15N] (green, in PC/PI lipids) KcsA-Kv1.3 at pH 7.4 acquired under similar experimental conditions (MAS: 10.92 kHz, T: 7°C, 700 MHz) but with a CP of 900 μs. c Cut-out of the aliphatic region of an NCACB-type correlation spectrum recorded with DARR mixing for 100 ms on [1H/2H,13C,15N] KcsA-Kv1.3. N–C cross peaks suppressed by fractional deuteration are indicated in red in several spectral regions. Amino acids are given in single letter notation

Full size image

The aliphatic region of the resulting spectrum is largely devoid of Cα-Cβ correlations (such as relating to Ile, Lys, Phe, Tyr or Asp residues), except for amino acids in which only one of the 13C positions is deuterated (Fig. 1a, red). For such protein residues (Ser, Thr, Cys, etc.) we observe, as expected for the short CP time (used in Fig. 1a), asymmetric correlation peaks. In line with earlier solution-state NMR studies (Rosen et al. 1996; Shekhtman et al. 2002; Otten et al. 2010), Cα positions of all amino acids are largely deuterated because the respective keto-acid undergoes a transamination reaction during synthesis (Nelson and Cox 2008). To further explore the residual 1H pattern, we conducted a conventional NCACB correlation experiment (Baldus 2002) using (13C,13C) DARR (Takegoshi et al. 2001) mixing (Fig. 1c). Similar to the results of the (13C,13C) correlation experiments, the NCA part of the spectrum largely agrees with data obtained on a protonated version (Supporting Figure 1) of the channel but the aliphatic region of the spectrum lacks several of the correlations that involve deuterated Cβ or Cγ positions. Indeed, Cβ positions of Glu, Gln, Pro and Arg that relate to α-Ketoglutarate as precursor in the biosynthetic pathway (Ref. (Nelson and Cox 2008), see also supporting table 1) are largely removed compared to the protonated case (Fig. 1c, red and Fig. 1b, green). Additional missing intensities relate to Cβ positions of Val and Ile, the aromatic amino acids of Phe, Tyr and His as well as the Cγ1 positions of Leu and Ile residues. On the other hand, pyruvate serves as a precursor to alkyl containing residues by direct incorporation (Ala, Val, Ile, Leu, Lys, etc.) or to aromatic amino acids and amino acids derived from Serine via other metabolites such as phosphoenol pyruvate and 3 phosphoglycerate (Supporting table 1). Hence, side chains of several amino acids containing alkyl groups are expected to exhibit sizable levels of protonation in line with our data. The protonation pattern at the remaining positions of amino acids is subject to residual protons from glucose itself and various intermediary steps that include cyclization, hydration, transamination or decarboxylation (Nelson and Cox 2008).

To directly infer the residual level of protonation, we conducted a (1H,13C) HETCOR experiment using FSLG decoupling (Bielecki et al. 1989) in the t1 dimension (Fig. 2). Compared to the case of the protonated channel (Lange et al. 2006b), the 1H-13C dispersion is remarkably improved. Firstly, all Hα-Cα correlations are largely eliminated and only some residual Ala, Leu, Glu Hα protonation remains. Because of the strong suppression of Hα protonation, the 1H-13C polarization transfer dynamics are determined by the residual NH and side-chain protonation level (Fig. 2, insert). Note that a similar transfer profile would require significantly longer mixing time in the case of soluble molecules where transfer occurs via through bond interactions. For amino acids such as Lys, Ile, Phe or Tyr, we expect dominant 1H-13C correlations within the NH resonance regime (dashed boxes in Fig. 2). On the other hand, Hβ-Cβ correlations can be readily identified for Thr, Cys, Ser residues in the spectrum (green box) in full accordance with our CC/NC data. Finally, a considerable reduction in spectral crowding is also visible in the methyl region of the (1H,13C) spectrum. Here, the spectrum is a result of the superposition of different methyl isotopomers that contribute to the residual protonation pattern of Ala, Thr, Val, Ile and Leu (Rosen et al. 1996; Shekhtman et al. 2002; Otten et al. 2010). Indeed, additional 13C-edited double quantum-single quantum 1H filtered experiments (Fig. 3) revealed a variety of correlations involving methyl proton pairs. Correlations between amide protons and aliphatic protons confirmed our FSLG HETCOR results. Furthermore, correlations involving Thr, Ser and Cys residues that appeared at 10 ppm in the 1H DQ dimension suggest that the Hβ positions of these residues are fully protonated. From our 2D data, we estimate 1H line widths ranging between 0.26 (Ile Hδ) and 0.4 ppm (Thr Hβ).

Fig. 2
figure 2

(1H,13C) FSLG-HETCOR spectra recorded on [1H/2H,13C,15N] KcsA-Kv1.3 (MAS: 13 kHz, T: 7°C, CP contact time: 200 μs). A schematic representation of magnetization transfer is shown in the inset. Deuterated sites are given in red

Full size image
Fig. 3
figure 3

a Pulse scheme for obtaining HH-C DQ-SQ correlations in two spectral dimensions. The SPC5 sequence is used for double quantum excitation and reconversion among dipolar coupled 1H spins. After isotropic chemical shift evolution during t1 and DQ reconversion, a short CP (90 μs) step ensures the transfer of magnetization of dipolar coupled protons to their directly attached carbon atoms which is read out during t2. b (1H1H-13C) DQ-SQ correlation spectrum recorded on [1H/2H,13C,15N] labeled KcsA-Kv1.3 at 700 MHz 1H resonance frequency with a 1H-1H DQ mixing time of 285 μs employing the SPC5 sequence (MAS: 14 kHz, T: 7°C, CP contact time: 90 μs). A schematic representation of magnetization transfer of Hβ protons to surrounding carbons is shown in the inset. Deuterated sites are labeled in red

Full size image

A more detailed analysis will be possible using topology schemes developed for solid-state NMR (Sakellariou et al. 2001) or using three-dimensional 1H-1H-13C or 1H-13C-13C experiments. A summary of the residual protonation pattern at the carbon sites identified from our CC/NCACX and DQ-SQ (1H,1H)-13C experiments is given in supporting table 1.

Assignments and structural constraints

Compared to the protonated case (Fig. 1b, green), fractional deuteration significantly reduces spectral complexity in complex biomolecules such the KcsA-Kv1.3 channel. We hence explored the use of such data for spectral assignment as well as for the structural analysis. Firstly, knowledge of the protonation pattern and the unique amino acid sequence of KcsA-Kv1.3 readily allowed us to obtain resonance assignments for Cys90 (Fig. 1a) not reported previously (Ader et al. 2008; Ader et al. 2009a, b; Schneider et al. 2008).

Additional sequential as well as medium to long-range correlations became accessible by recording (13C,13C) correlation experiments at mixing times beyond 100 ms. Firstly, we directed our attention to correlations involving aromatic side-chains. Interestingly, we observed intense aromatic–aromatic side-chain correlations that were otherwise not visible in the fully protonated version of the channel (Supporting Figure 2). In Fig. 4, many of the observed correlations can be readily explained by intra-residue correlations within in the aromatic side-chains of Trp, Tyr and His. Apart from these, there are only 4 residue pairs that would give rise to sequential correlations, i.e., (H25, W26), (W67, W68), (W113, F114) and (H124, F125). Our analysis of these correlations with the structural model (see “Materials and methods”) suggested that contacts in the range of 3.5–5 Å are only expected for the aromatic side-chains of the (H25, W26) pair. The (13C,13C) data at a 200 ms mixing and MAS rate of 10.92 kHz were fully consistent with such sequential correlations (green lines in Fig. 4) leading to tentative aromatic side-chain assignments for H25 and W26.

Fig. 4
figure 4

Aromatic side-chain contacts identified in a 2D (13C-13C) PDSD correlation spectrum recorded on [1H/2H,13C,15N] KcsA-Kv1.3 with a spin diffusion time of 200 ms and a CP time of 80 μs. A sequential walk for residues H25 and W26 is indicated by green lines. Unequivocally cross-peak assignments were made for correlations for which other spectral predictions are separated by at least 0.6 ppm in one or two spectral dimensions

Full size image

Subsequently, we investigated the use of (13C,13C) PDSD data with longer mixing times. Compared to the protonated case, the resulting spectral congestion can be reduced by choosing short CP times of 80 μs that select for protonated 13C sites as evolution and detection spins in t1 and t2. Spectral cut outs correlating aliphatic and aromatic-aliphatic regions for a 13C-13C mixing time of 500 ms (MAS rate: 10.92 kHz) are shown in Fig. 5a and b, respectively. Even without residue-specific assignments, the observed correlations between SCβ-ACβ, SCβ-LCβ and SCβ-VCγ1/2 spin pairs (indicated by dashed boxes in Fig. 5a) mostly encode medium to long-range correlations. Moreover, our structural model is compatible with the detection of a set of resolved long range and medium range intramolecular contacts. For example, starting with the chemical shift assignment on T140Cβ from our previous studies (Schneider et al. 2008) (Fig. 5a), we now identified a unique medium range correlation with the A143Cα position. Other correlations relate to T56Cβ-V84Cγ2 or A52Cα-D54Cα in Fig. 5a. The assignment of A143Cα not only correlated to E146Cγ (Fig. 5a) but suggested polarization transfer to H145Cε1 (Fig. 5b). Notably, the identified H145Cε1-A143Cα correlation can only be an intermolecular contact (4.3 Å) as the predicted intramolecular distance is significantly longer (8.6 Å). Additional contacts were also identified that connect L146C’ and F148Cδ2 to H145Cε1. Finally, intermolecular contacts were also observed between residues Y82Cδ1 and D80Cβ (highlighted in red, Fig. 5c).

Fig. 5
figure 5

a, b 2D (13C,13C) PDSD spectra of [1H/2H,13C,15N] KcsA-Kv1.3 recorded at 10.92 kHz MAS and a CC mixing time of 500 ms employing a CP time of 80 μs. Potential intramolecular (black labels) and intermolecular (red labels) contacts identified have been indicated. c A selected set of contacts identified from ssNMR experiments (a, b) that are compatible with the structural model of KcsA-Kv1.3

Full size image

The transfer efficiency of proton-mediated longitudinal transfer schemes such as PDSD is dependent on the proton density and the MAS rate (see, e.g., Suter and Ernst 1985; Kubo and McDowell 1988; Oas et al. 1988; Colombo et al. 1988; Lange et al. 2003). The reduced proton density enhances the influence of rotational resonance (RR) effects that depend on the setting of the MAS rate relative to the chemical-shift difference among all spins exchanging polarization. We thus performed a series of 2D (13C-13C) PDSD experiments with different MAS rates and (13C-13C) mixing times ranging between 20 and 500 ms to examine the MAS-dependence of the polarization transfer. For example, in Fig. 6 we compare the transfer profile between aromatic side-chains and methyls for two different MAS rates (left: 10 kHz, right: 10.92 kHz). In line with expectations based on the RR recoupling conditions for Cα-C’, Carom-Cα/Cmethyl/C’ pairs, polarization transfer becomes band-selective and is enhanced at 10 kHz MAS.

Fig. 6
figure 6

a Cutouts from the (13C,13C) PDSD spectra (using a mixing time of 500 ms) employing two different MAS frequencies (10 kHz & 10.92 kHz) to enhance the rotational resonance recoupling between aromatic and methyl side-chain groups. The width of the red boxes is given by the size of chemical dispersion among aromatic 13C frequencies. b Enlarged spectral regions with annotated intramolecular (black labels) and inter-molecular (red labels) contacts

Full size image

Remaining intensities at 10.92 kHz in the spectral regime [40, 25 ppm] can be well explained by transfer across short, mostly intra-residue distances. The identified aromatic-aliphatic side-chain correlations are compatible with the detection of sequential side-chain contacts between L86 and W87 (green box) and long range contacts between W68 (Cζ2, Cζ3, Cη2) and A47Cβ. Furthermore, we could identify resolved intermonomer contacts between the aromatic-aliphatic side-chains (highlighted in red) as well as cross peaks between specific amino acid pairs (F–A Cβ, F–T/V Cγ1/2) that solely encode intermolecular contacts (red boxes).

Following the analysis of other spectral regions of our spectra, we could in total trace 62 assignments mostly involving side-chain atoms (see Supporting Information, Table 2). Furthermore, the data were, in conjunction with the structural model, compatible with the observation of 42 medium range, 36 long-range constraints (Supporting Information, Table 3) and 23 intermonomer contacts (Supporting Information, Table 4). In Fig. 7, these residue–residue contacts are displayed along the sequence and with the number of occurrences per residue.

Fig. 7
figure 7

Correlation plot containing resolved medium to long-range contacts and their occurrence identified in our ssNMR experiments on [1H/2H,13C,15N] KcsA-Kv1.3. A pictorial scheme describing polarization transfer mediated through protonated side-chains (green cloud) in the presence of a largely deuterated backbone (red) for the case of [1H/2H,13C,15N] KcsA-Kv1.3 is shown as an insert

Full size image

Conclusions

Overall, our results on a fractionally deuterated ion channel embedded in lipid bilayers suggest a reduction in molecular protonation by approximately 50% in line with earlier solution-state NMR work on globular proteins (Rosen et al. 1996; Shekhtman et al. 2002; Otten et al. 2010). Reminiscent of effects seen using specifically-labeled glycerol precursors during bacterial growth (LeMaster and Kushlan 1996; Hong 1999; Castellani et al. 2002), the exact deuteration level can vary in protein side-chains and methyl positions are expected to contain several isotopomers. In 1H correlation experiments, we observed an improved spectral dispersion and a reduction in line width compared to the protonated case. In standard PDSD experiments on [1H/2H,13C,15N] KcsA-Kv1.3, one-bond (13C,13C) correlations appeared after 20 ms mixing similar to fully protonated KcsA-Kv1.3 suggesting that polarization transfer dynamics are not strongly affected by the reduction in proton density. However, additional studies will be needed to dissect in detail the polarization transfer rates as a function of MAS rate, B0 field and local proton density. In general, our experiments were conducted at modest MAS rates (10-14 kHz), it seems likely that further improvements in 1H line width and spectral resolution may be possible using a combination of dedicated pulse schemes (Sakellariou et al. 2000), ultra-fast MAS (Wickramasinghe et al. 2009; Bertini et al. 2010b) and ultra-high field NMR systems.

The reduced protonation level also simplifies the spectral analysis of 13C/15N correlation experiments. Because residual protons are preferably found at the amino-acid side-chains (Fig. 7 insert, Supporting Information table 1), polarization transfer among and spectroscopic assignments of protein side-chain positions is facilitated. For long mixing times and longitudinal mixing schemes, proton-mediated transfer becomes band-selective around the rotational resonance conditions among aliphatic, aromatic and carboxyl carbons. Experimental results shown here suggest that these conditions can aid the detection of medium to long-range correlations occurring in a particular spectral window. Notably, such measurements also revealed intermolecular contacts in our tetrameric [1H/2H,13C,15N] ion channel for which the combined application of dedicated ssNMR schemes and mixed labelling approaches that previously allowed detecting such constraints (see, e.g., Etzkorn et al. 2004; Wasmer et al. 2008; Etzkorn et al. 2010) is precluded. It seems likely that fractional deuteration will also facilitate the determination of longer internuclear distances using rotational-resonance recoupling (Spencer et al. 1991; Costa et al. 1997) or rotating-frame (Nomura et al. 1999; Sonnenberg et al. 2004) and MAS-modulated variants (Verel et al. 1997; Ramachandran et al. 2003) thereof. In addition, coherent transfer schemes that mediate (13C,15N) transfer via proton spins such as CHC (Seidel et al. 2005), PAR (Paepe et al. 2008) or PAIN-CP (De Paepe et al. 2011) experiments may be readily combined with fractional deuteration to suppress chemical-shift offset affects or to enhance transfer efficiencies.

Compared to schemes involving (partially) deuterated precursors, fractional deuteration reduces the influence of isotope effects on ssNMR chemical shifts (Hansen 1988) and offers a cost efficient way to sizably reduce protonation levels in complex biomolecules. These considerations and our results suggest that fractional deuteration can provide a powerful means to aid structural studies of complex biomolecules by high-resolution ssNMR.