Introduction

The NMR spectra of larger proteins are characterized by numerous overlapped signals, which are severely broadened by dipolar interactions between nearby protons. Therefore, it is practically impossible to observe all of the individual NMR signals for quite large proteins. Conceptually, however, it may not be necessary to gather all of the NMR signals for studying protein structures and dynamics, since the amino acid side-chains contain somewhat redundant structural information. For example, if we could deuterate either one of the prochiral methyls or methylene protons stereo-specifically, then the missing information could be compensated more than adequately by the remaining geminal counterpart signals, which have unambiguous stereospecific assignments. The stereo-array isotope labeling (SAIL) method systematically creates such situations for all of the amino acid residues in proteins, by trimming off the redundant information by isotope-labeling. The original SAIL amino acids preserve the through-bond 13C–13C and 13C–15N connectivity paths for the sake of the backbone and side-chain sequential assignments, and completely eliminate the ambiguities of the stereo-specific assignments for all of the prochiral groups (Kainosho et al. 2006, 2018; Kainosho and Güntert 2009). Although the overall non-exchangeable proton density in a SAIL protein, which is exclusively composed of SAIL amino acids, becomes reduced to 50–60% of that of the fully protonated protein, the remaining protonated sites are perfectly untouched. It is quite important to emphasize that, even though the level of deuteration is as high as 60%, only a single isotopomer exists for a SAIL protein. The terminology “stereo-array isotope labeling” actually seeks to highlight this striking feature of this new isotope-labeling approach.

The significantly decreased overall proton density for a SAIL protein mitigates the spin diffusion without sacrificing the signal intensities of the remaining protonated sites, and thus facilitates the acquisition of accurate inter-proton NOEs for larger proteins. By virtue of these distinctive features of the SAIL method, facile and accurate structure determinations of proteins as large as 40–50 kDa have become feasible. Obviously, there is a practical size limit for protein structural determinations even by the SAIL method, since spectral analyses become incorrigibly cumbersome for extraordinarily large proteins. Fortunately, the other methods, such as X-ray crystallography and cryo-electron microscopy, have significantly advanced during the past decades. These methods are more efficient and widely applicable for proteins regardless of their sizes; therefore, the structure determinations of quite large proteins may no longer be the primary role for NMR spectroscopy. Instead, in the contemporary paradigm of integrative structural biology, NMR spectroscopy is envisaged as the main technique to study protein structures and dynamics. This information would be most efficiently obtained through the NMR probe signals belonging to the functionally important regions of proteins, and thus is crucial to understand their biological functions.

Although the structural probes are presently limited to the TROSY methyl and backbone amide signals for larger proteins (Pervushin 2000; Tugarinov et al. 2003; Tugarinov and Kay 2003), in this article we show that any type of NMR signals other than those from methyl and amide groups; i.e., aromatic and aliphatic CH groups, can also be observed equally well even for the 82 kDa single domain protein, malate synthase G (MSG), by optimizing their transverse relaxation properties by appropriate isotope labeling. We purport that these newly accessible NMR signals will provide unprecedented opportunities for studying the structures and dynamics of larger proteins and protein complexes.

Transverse relaxation optimized SAIL (SAIL TROSY) approach

Tugarinov and Kay successfully observed narrow 1H–13C HMQC signals for the methyl groups of the Ile, Leu and Val (ILV) residues in deuterated MSG, in which the ILV methyls were selectively labeled with 13CH3 by using metabolic precursors (Tugarinov et al. 2003; Tugarinov and Kay 2004). Although the ILV residues represent only 22% of the 723 residues of MSG, a well-defined structural model was obtained by combining all of the information obtained by the ILV methyls and the backbone 15NH signals. Since the methyl signals of the Ala, Thr and Met residues could also be observed, the methyl-containing amino acid residues thus represented as much as 40% of the MSG protein (Isaacson et al. 2007; Godoy-Ruiz et al. 2010). A commonly held belief is that the rapid rotation around the pseudo threefold axis is a prerequisite to generate narrow 1H–13C HMQC signals, as derived from the theoretical consideration (Tugarinov et al. 2003). This breakthrough technology, which is often referred to as “methyl TROSY”, has been widely used for studying large proteins together with 15NH TROSY (Pervushin et al. 1997; Pervushin 2000). If similar considerations are also valid for the methylene and methine groups of the amino acid side-chains, then they would rarely provide narrow NMR signals. Fortunately, this is not the case for the methylene and methine groups. Although they cannot rotate rapidly on their own, the aliphatic 13CH groups generate narrow line-widths by improving their transverse relaxation properties by the optimal isotope labeling strategy; i.e., the “TROSY by isotope labeling”. We also observed the narrow aromatic ring 13CH signals for large proteins by the aromatic TROSY method (Pervushin et al. 1998), with further optimization of the transverse relaxation properties of the original SAIL aromatic amino acids by isotope labeling. Although the aromatic amino acid residues, Phe, Tyr, Trp and His, represent only 10% of the total residues in MSG, the structural information about the bulky side-chain rings is extremely important for filling in the gaps between the methyl groups and the backbone amides.

Observation of the aromatic ring 13CH signals

Pervushin et al. showed that the aromatic ring 13C signals exhibit the TROSY effect, as a consequence of the cross relaxation effect between the 13C chemical shift anisotropy (CSA) and the dipolar–dipolar (DD) relaxation due to the directly bonded proton (Pervushin et al. 1998). Since they used [U–13C,15N]-proteins, the DD-CSA cross-relaxation effect was far from the optimum, due to the complex scalar and dipolar spin coupling networks. It is interesting to mention that according to the theoretical prediction, the TROSY effect on the 13C T2 is most efficient at around 600 MHz (Pervushin et al. 1998). This issue was recently revisited by Takeuchi et al. in the context of the optimal field strengths for direct observations of 15N and 13C NMR (Takeuchi et al. 2016). Takeuchi et al. claimed that the maximum 1H-detected 15NH TROSY and 13C-detected aromatic 13CH TROSY should be around 1.5 GHz and 900 MHz, respectively. We actually found that the 1H-detected aromatic 2D 13CH TROSY sensitivity increases substantially, by increasing the field strengths in the range of 600–900 MHz. This may depend on the performance of the NMR equipment, but it is also affected by the drastically simplified spin-networks implemented by the SAIL aromatic amino acids (Torizawa et al. 2005; Takeda et al. 2010; Miyanoiri et al. 2011, 2012; Yang et al. 2015).

We prepared MSGs labeled with either ζ-SAIL Phe or [δ1, ε3, η2]-SAIL Trp by the conventional Escherichia coli cellular overexpression system, using deuterated M9 medium, and found that the labeling rates were higher than 90% for both amino acids, even with small amounts of the SAIL Phe or SAIL Trp (~ 10 mg/L). The 2D aromatic 13CH correlation spectra were measured at 800 MHz for a 0.3 mM solution under various conditions, to define the optimal acquisition parameters. The F1-decoupled HSQC showed virtually no signals even after ~ 7 h, since the anti-TROSY components broaden the F1-decoupled signals beyond detection (Fig. 1a). However, the 1H–13C TROSY HSQC, using the S3CT pulse sequence (Sørensen et al. 1997), gave 36 beautiful cross peaks for the δ1, ε3, an η2 CHs for the 12 Trp residues of MSG within the same duration (Fig. 1b). All of the signals are well dispersed within the large chemical shift ranges for 13C (~ 12 ppm) and 1H (~ 2.5 ppm). The assignments of the signals were completed as follows. The δ1-CH signals were identified by correlating the δ1-13CH to the ε3-15NH through the 1H–15Nε–13Cδ–1H connectivity, and the ε3-signals were identified by labeling MSG with the ε3-dueterated [δ1, ε3, η2]-SAIL Trp. The sequential assignments for the cross peaks were firmly established by a series of single amino acid replacements of each Trp residue with Phe (Fig. 1c).

Fig. 1
figure 1

Aromatic regions of the 800 MHz 2D 1H–13C correlation spectra of the MSG selectively labeled with [δ1,ε3,η2]-SAIL Trp, and otherwise uniformly labeled with deuteron and 15N. Sample: 0.3 mM in 20 mM sodium phosphate buffer (pH 7.1), containing 20 mM MgCl2, 5 mM DTT, and 5% D2O. NMR: NS = 128, d1 = 1.5 s, TD = 1024 × 256, 310  K. The total experimental time for each spectrum was ~ 7 h with a cryogenic probe. a F1-decoupled HSQC; b Aromatic 13CH TROSY; c Aromatic 13CH TROSY with signal assignment

Full size image

We also illustrate the 1H–13C TROSY HSQC spectrum obtained for the MSG selectively labeled with ζ-SAIL Phe. Although we used a 0.15 mM H2O solution in this case, 19 well resolved signals for the 19 Phe residues were obtained at 900 MHz within ~ 9 h (Fig. 2). The sequential assignment shown in the figure was initially completed by a series of single amino acid mutations for each residue. The assignment strategy using a series of single amino acid replacements has been successfully used for the assignments of methyl signals in large proteins (Amero et al. 2011). In the case of bulky aromatic rings, we should be especially careful not to cause serious structural perturbations by mutations, and thus we used either Tyr or Leu to replace each Phe residue. The other aromatic 13CH signals of Tyr and His in MSG were also observed using transverse relaxation optimized SAIL amino acids. Many of the assignments for the aromatic protons by the amino acid replacement method have been further confirmed by the 1H–1H NOEs between the ILV methyl protons, for which the sequential assignments were firmly established (Tugarinov et al. 2003).

Fig. 2
figure 2

Aromatic region of the 900 MHz 2D aromatic 13CH TROSY spectrum of the MSG selectively labeled with ζ-SAIL Phe, and otherwise uniformly labeled with deuteron and 15N. Sample: 0.15 mM in 20 mM sodium phosphate buffer (pH 7.1), containing 20 mM MgCl2, 5 mM DTT, and 5% D2O. NMR: NS = 64, d1 = 2 s, TD = 1024 × 128, 310 K. The total experimental time was ~ 9 h with a cryogenic probe

Full size image

Observation of aliphatic 13CH signals other than methyls

Since the backbone amides, side-chain methyls, and aromatic rings comprise the considerable portion of a protein, their narrow NMR signals provide valuable information about the structures and dynamics of larger proteins. However, this information is not completely sufficient for the detailed analysis of the conformational dynamics of the biologically important side-chain moieties, which could be manifested most directly by their aliphatic methylene and methine signals. If we recollect the theoretical considerations for methyl TROSY (Tugarinov et al. 2003), the aliphatic methylene and methine groups are not likely to generate narrow NMR signals, since they are not amenable to the line-narrowing effects by rapid rotation, DD-CSA interference, or interference between auto- and cross-correlated 1H–1H dipolar relaxation. In contrast to these pessimistic expectations, we proved that very narrow methylene and methine 13CH signals can be obtained for the isolated 13CH pairs, implemented by suppressing the major 1H–1H and 13C–13C scalar and dipolar coupling pathways using appropriate isotope labeling.

We synthesized two prototypes of tryptophan customized for observing the β3H and β2H signals, in which the 13Cβ has no directly bonded 13C and either one of the prochiral protons is stereo-specifically deuterated (Fig. 3). Even though the indole ring remains protonated, we successfully observed all 12 of the 13CH pairs for each of the MSG proteins expressed in deuterated M9 medium containing a small amount (10 mg/L) of either [β-13C; β2-D]-Trp (Fig. 3a) or [β-13C; β3-D]-Trp (Fig. 3b). The sequential assignments of the 13Cβ were readily established by adapting the 13Cβ chemical shift data obtained by the 4D NMR analysis of [U–13C,15N, D]-MSG (Tugarinov et al. 2002). Their data fit almost completely with our 13Cβ data, by accounting for the small differences due to the isotope shifts. These assignments were further confirmed by the intra- and inter-residue NOEs and by the single amino acid replacements, as described above.

Fig. 3
figure 3

2D Aliphatic 13CH HSQC spectra of MSGs selectively labeled with a [β-13C; β2-D]-Trp or b [β-13C; β3-D]-Trp, and otherwise fully deuterated. Sample: 0.12 mM in 20 mM sodium phosphate buffer (pH 7.1), containing 20 mM MgCl2, 5 mM DTT, and 5% D2O. NMR: d1 = 4 s, TD = 1024 × 128, 310 K

Full size image

Structure information obtained by the aromatic and aliphatic 13CH signals

It is now apparent that all of the aromatic and aliphatic 13CH signals could be observed and assigned for proteins as large as 80 kDa. In a preliminary effort to estimate the molecular weight limit for observing aromatic and aliphatic 13CH signals by “the SAIL by isotope labeling” method, we successfully observed the aromatic TROSY 13CH signals for an 880 kDa Thermus thermophilus chaperonin (GroES–EL complex). Therefore, these newly accessible NMR probes provide unprecedented opportunities to apply a variety of NMR techniques, which have been exclusively used for small and medium size proteins, to extremely large proteins and protein complexes. Although some of the local structure information acquired from the aromatic and aliphatic 13CH signals may seem too trivial for studying supramolecular complexes, it is quite important to emphasize that such information had never been obtained for extremely large proteins. In the following, we show two examples.

The side-chain conformation of Trp residues

The side-chain conformation of a Trp residue in a protein can be defined by two rotation angles, χ1 and χ2. The χ2 angle, which represents the rotation angle around the Cβ–Cγ bond, can be estimated by the intra-residue NOEs between the side-chain β-H and the indole ring δ1- and ε3-H. The SAIL Trp residue was quite useful to observe the intra-residue NOEs relevant to the χ2 angles, as shown for the 7 Trp residues in the 12 kDa cMyb-R2R3 domain (Miyanoiri et al. 2011). In this small protein, we could determine the χ1 and χ2 rotamers using the relevant intra-residue NOEs, without the deuteration of the residues other than SAIL Trp. The same approach could be applied to determine the three χ2 rotamers; i.e., ± 90° and 0°, for the 12 Trp residues in the 82 kDa MSG, although we had to fully deuterate MSG in order to maximize the 13CH TROSY effect. In Fig. 4, we show two strips of the 3D 13C-edited NOESY-HSQC for the MSG selectively labeled with the “β2-deuterated” SAIL-Trp, to illustrate the intra-residue NOEs for the W67 and W509 residues. According to the crystalline state, the χ2 angles of W67 and W509 are + 103° and − 102°, respectively. The χ2 angle difference between these two residues results in the inter-proton distances between Hβ3 and δ1- and ε3-H. The NOEs were observed only for the β3-ε3 (2.7 Å) and β3-δ1 (2.6 Å), for W67 and W509, respectively, indicating that the χ2 angles determined in the solution and the crystal are the same. Although the χ2 angles of W67 and W509 can be roughly estimated as + 90° (± 30º) and − 90° (± 30º) by NMR, the side-chain conformations of the aromatic amino acids play a crucial role in the local structures around them.

Fig. 4
figure 4

900 MHz 3D 13C-edited NOESY-HSQC spectrum of the MSG selectively labeled with β2 deuterated SAIL-Trp, and otherwise fully deuterated. The two strips for the W67 and W509 residues are shown. The χ2 angles and the β3-δ1, ε3 hydrogen distances in the crystal (PDB 1D8C): W67, χ2 = + 103º, β3-δ1 3.8 Å, β3-ε3 2.7 Å; W509, χ2= − 102º, β3-δ1 2.6 Å, β3-ε3 4.3 Å. Sample: 0.25 mM, in 20 mM sodium phosphate buffer (pH 7.1), containing 20 mM MgCl2, 5 mM DTT, and 1% D2O. NMR: NS = 32, mixing time = 200 ms, TD = 2048 × 20 × 128, d1 = 4 s, 310 K

Full size image

The relative orientation between aromatic rings

The inter-residue NOEs associated with the aromatic ring protons are especially important, to acquire the local structural information around the bulky aromatic rings. We illustrate the most obvious case for determining the relative orientations of two spatially close aromatic rings; namely, the aromatic rings of F35 and W36. As shown in Fig. 5a, the relative orientation between the two aromatic rings cannot be defined in the NMR structures calculated based on the NOEs for the ILV methyls and backbone amides. Obviously, the NOE data set has no distance constraints for the relative orientation between the two rings. However, the overall NMR structures calculated based on the NOE data sets including the aromatic ring protons of 19 Phe, 18 Tyr and 12 Trp residues (a total of 49 residues) converged much better, especially in terms of the relative orientation between the aromatic rings, as illustrated in Fig. 5b, as did the structures around the ILV methyls and aromatic rings. We are still collecting NOEs and hope to prove that the NMR structures of MSG ultimately converged as robustly as those for small proteins.

Fig. 5
figure 5

Effect of the implementation of the additional NOE constraints acquired for the aromatic and aliphatic methylene protons on the NMR structure of MSG. Structural calculations including NOE constraints for a ILV methyls and backbone amides; b ILV methyls, backbone amides, and aromatic ring protons of Phe, Tyr, and Trp residues. Only the relative orientation between F35 and W36 is illustrated, although the overall structure was substantially refined by including the NOEs involving aromatic and aliphatic protons (details will be published elsewhere)

Full size image

A next generation labeling strategy: Low-dimensional 1H-direct observation spectroscopy at ultra-high fields over 1 GHz

We described above a breakthrough isotope labeling strategy to observe aromatic and aliphatic 13CH signals for extremely large proteins. The idea of transverse relaxation spectroscopy (TROSY) by isotope labeling, which is the heart of this new technology, emerged from the original SAIL method, which was developed more than a decade ago (Kainosho et al. 2006, 2018). During the past decade, however, the major role of NMR spectroscopy in structural biology has dramatically changed from structure determinations to structure dynamics studies of biologically interesting targets, such as membrane proteins and supramolecular complexes. Therefore, we hope that newly accessible NMR probes, besides methyls and backbone amides, for difficult targets will contribute to maintain NMR spectroscopy as the major player in the coming era of integrative structural biology. In the future, ultrahigh-field NMR spectrometers will be available to the biological NMR community and scientists will keenly desire access to them. Obviously, such machines will be unaffordable for individual laboratories, and thus they should be shared. The machine time allocated to each laboratory will be quite limited, and therefore it should be used most efficiently. This is somewhat an entirely new situation for the NMR community, since the major NMR laboratories always have their own high-field equipment. Therefore, we definitely need much more sensitive NMR methods than ever explored.

The advanced isotope-aided methods developed so far are based on the indirect 1H-detection via spin-coupled heteronuclei, such as 13C and 15N. Although there are many innovative techniques to shorten the measurement time, unavoidably most of the 1H-NMR magnetization is lost during the coherent transfer processes in heteronuclear multidimensional spectroscopy. Recently, Takeuchi et al. mentioned the same problem and suggested that amide 15N-detected TROSY and aromatic 13C-detected TROSY would be most sensitive at 1.15 GHz and 900 MHz, respectively (Takeuchi et al. 2016). These nuclei have much lower gyromagnetic ratios (γ) as compared to 1H, and thus their inherent NMR sensitivities are much lower than 1H. On the other hand, due to their low γ values, they give narrower line-widths, because of that they are less sensitive to the dipolar broadening from the nearby 1H signals. The effect may be largely offset by the remaining CSA for the amide 15N and aromatic 13C, and thus it is where the direct 1H-detection method comes in.

Proton (1H) is the most sensitive nucleus other than tritium (3T), and has very little CSA. Therefore, if there is no other 1H nearby, the isolated 1H attached to 12C should have the highest sensitivity. Actually, the method, called “selective deuteration”, was developed a half century ago, as the first isotope-aided NMR method to observe the isolated 1H of aromatic amino acids embedded a deuterated background (Markley et al. 1968). We revisited this old method to observe the 1H12Cζ signals of Phe residues in the MSG labeled with [δ,ε-D4]-Phe, with an otherwise fully deuterated background (Fig. 6). Surprisingly, but somewhat expectedly, extremely narrow and well dispersed 1Hζ signals were observed at 900 MHz. All of the signals were readily assigned just by transferring the 13CH TROSY signal assignment, shown in Fig. 2. There were no other signals in this region, since the amide protons were fully deuterated. The 1Hζ of the Phe residue in MSG has virtually no other protons within a few Å, and the observed line-widths are about 5–6 Hz, a value about one order of magnitude smaller than the 1H line-widths estimated from the TROSY 13CH signals at 900 MHz. The sensitivity is reasonably high, since the spectrum was obtained with a 0.2 mM solution with 512 transients. A long T1 value for the isolated 1H12C was an anticipated problem, and thus we used a relatively long relaxation delay time (d1) of 30 s to acquire the 512 transients, and the total experimental time was ~ 4.5 h. The long T1 problem would be largely improved by incorporating the ILV methyls simultaneously, as they served as an efficient heat sink.

Fig. 6
figure 6

900 MHz 1D 1H-NMR spectrum of [U-D;δ,ε-D4-Phe]-MSG. Sample: 0.2 mM in 20 mM deuterated sodium phosphate buffer (pH ~ 7), containing 20 mM MgCl2 and 5 mM DTT. NMR: d1 = 30 s, NS = 512, 310 K. The total experimental time was ~ 4.5 h. The signal assignments were transferred directly from the 2D aromatic 13CH TROSY spectrum shown in Fig. 2

Full size image

The 1D 1H-NMR is especially useful for the aromatic ring protons, since there are only amide protons in the same chemical shift region, which can be completely eliminated by deuterium substitution in D2O. However, low-dimensional 1H-direct detection NMR spectroscopy may not be restricted to the aromatic ring protons, and should be explored as the next generation isotope-aided method in the ultrahigh-field NMR era. Since any type of proton has a relatively small CSA, the present approach of “selective protonation” should work better at higher field strengths, if we could overcome the T1 versus T2 problem by controlling the relaxation paths.