Triple resonance experiments for aligned sample solid-state NMR of ¹³C and ¹⁵N labeled proteins

Abstract

Initial steps in the development of a suite of triple-resonance ¹H/¹³C/¹⁵N solid-state NMR experiments applicable to aligned samples of ¹³C and ¹⁵N labeled proteins are described. The experiments take advantage of the opportunities for ¹³C detection without the need for homonuclear ¹³C/¹³C decoupling presented by samples with two different patterns of isotopic labeling. In one type of sample, the proteins are ~20% randomly labeled with ¹³C in all backbone and side chain carbon sites and ~100% uniformly ¹⁵N labeled in all nitrogen sites; in the second type of sample, the peptides and proteins are ¹³C labeled at only the α-carbon and ¹⁵N labeled at the amide nitrogen of a few residues. The requirement for homonuclear ¹³C/¹³C decoupling while detecting ¹³C signals is avoided in the first case because of the low probability of any two ¹³C nuclei being bonded to each other; in the second case, the labeled ¹³C_α sites are separated by at least three bonds in the polypeptide chain. The experiments enable the measurement of the ¹³C chemical shift and ¹H–¹³C and ¹⁵N–¹³C heteronuclear dipolar coupling frequencies associated with the ¹³Cα and ¹³C′ backbone sites, which provide orientation constraints complementary to those derived from the ¹⁵N labeled amide backbone sites. ¹³C/¹³C spin-exchange experiments identify proximate carbon sites. The ability to measure ¹³C^–15N dipolar coupling frequencies and correlate ¹³C and ¹⁵N resonances provides a mechanism for making backbone resonance assignments. Three-dimensional combinations of these experiments ensure that the resolution, assignment, and measurement of orientationally dependent frequencies can be extended to larger proteins. Moreover, measurements of the ¹³C chemical shift and ¹H–¹³C heteronuclear dipolar coupling frequencies for nearly all side chain sites enable the complete three-dimensional structures of proteins to be determined with this approach.

1. Introduction

The vast majority of applications of aligned-sample solid-state NMR spectroscopy to proteins have utilized ¹H/¹⁵N double resonance experiments [1–3]. This reflects the relative ease of ¹⁵N labeling of proteins by expression in bacteria, the strategic locations of the amide nitrogens within the planes of the polypeptide backbone, and the favorable spectroscopic properties of the directly bonded pair of ¹H and ¹⁵N nuclei in each amide site [4]. Even when all nitrogen sites in a protein are labeled with ¹⁵N, these nuclei behave as if they are “dilute” because of the combination of their low gyromagnetic ratio and spatial isolation. Three bonds separate the backbone amide nitrogens, and there are no directly bonded pairs of nitrogen atoms in polypeptides. Consequently, it is unnecessary to incorporate homonuclear ¹⁵N/¹⁵N decoupling at any stage of the SLF (separated local field) [5] and HETCOR (heteronuclear correlation) experiments that yield well resolved two- and three-dimensional spectra of proteins, and are used to measure the orientationally dependent ¹H chemical shift, ¹H–¹⁵N heteronuclear dipolar coupling, and ¹⁵N chemical shift frequencies that are input for the structure calculations. However, it is desirable to measure frequencies from additional spin interactions, especially those associated with carbon atoms, in order to include the side chains in the structure determinations, and to improve the precision and accuracy of the structure calculations.

¹H/¹³C/¹⁵N triple-resonance experiments are widely employed in both solution-state NMR [6] and MAS (magic angle spinning) solid-state NMR of proteins [7,8]. Initial steps in the development of a suite of triple-resonance solid-state NMR experiments applicable to aligned samples of ¹³C and ¹⁵N labeled proteins are described in this article. These experiments enable the measurement of the ¹³C chemical shift and ¹H–¹³C and ¹⁵N–¹³C heteronuclear dipolar coupling frequencies associated with the ¹³C_α and ¹³C′ backbone sites. These frequencies provide orientation constraints complementary to those derived from the ¹⁵N labeled amide backbone sites, and will improve the quality of structure determinations. In particular, with a sufficient number of heteronuclear dipolar couplings measurements, there is the possibility of improving the precision of the calculated protein structure calculations by avoiding the use of the chemical shift frequencies that are the largest source of uncertainty due to residue-to-residue variations of the chemical shift tensors. Moreover, measurements of the ¹³C chemical shift and ¹H–¹³C heteronuclear dipolar coupling frequencies for nearly all side chain sites will enable the complete three-dimensional structures of proteins to be determined with this approach.

Despite these well-established benefits, only a few triple-resonance experimental results relevant to aligned-sample solid-state NMR have been reported. This is largely because of the difficulty in implementing C detection on uniformly ¹³C labeled molecules with their dense network of homonuclear ¹³C/¹³C couplings among the directly bonded carbons. The triple-resonance solid-state NMR experiments described in this article take advantage of the opportunities for ¹³C detection without the need for homonuclear ¹³C/¹³C decoupling presented by samples with two different patterns of isotopic labeling. In one type of sample, the proteins are ~20% randomly labeled with ¹³C in all backbone and side chain carbon sites and ~100% uniformly ¹⁵N labeled in all nitrogen sites. Similar labeling schemes have been used in solution NMR studies of biomolecules [9,10]. In the second type of sample, the peptides and proteins are ¹³C labeled at only the α-car-bon and ¹⁵N labeled at the amide nitrogen of a few residues. The corresponding experiments for the complementary labeling scheme at only the C′ sites can be readily envisaged. The requirement for homonuclear ¹³C/¹³C decoupling while detecting ¹³C signals is avoided in the first case because of the low probability (~4%) of any two ¹³C nuclei being bonded to each other. In the second case, the labeled ¹³C_α sites are separated by at least three bonds in the polypeptide chain, so in the “worst case”, where two adjacent residues in an α-helix are labeled, the C_α sites are separated by 3.8 Å, which results in a maximum ¹³C/¹³C homonuclear dipolar coupling of 135 Hz when the inter-nuclear vector is parallel to the direction of the magnetic field. This corresponds to about 0.5 ppm in a 21T magnet, and will generally not contribute to the linewidths of the observed resonances; in magnetically aligned bilayer samples (perpendicular bicelles), the corresponding values are 54 Hz and 0.2 ppm, substantially less than the typical 1–2 ppm linewidths of the protein resonances [11]. Complementary approaches that result in alternate sites labeled with ¹³C have been used in MAS solid-state NMR experiments [7,12], and may also prove useful in aligned sample solid-state NMR with experiments similar to those described here.

We have previously described applications of both ¹⁵N- and ¹³C-detected versions of triple-resonance experiments to single-crystal samples of model peptides, and aligned samples of filamentous bacteriophage particles and membrane proteins in phospholipid bilayers [11,13–17]. Recently, we have reviewed [18] these results and extended them somewhat by constructing a new generation of triple-resonance probes and implementing SPINAL-16 decoupling [19] for greater bandwidth [20]. However, these experiments are fundamentally limited in scope: in order to observe ¹³C chemical shift frequencies in the spectra, it is necessary to apply homonuclear ¹³C/¹³C decoupling in an indirect dimension, and utilize ¹⁵N detection in the presence of heteronuclear ¹H and ¹³C decoupling at the cost of the sensitivity advantage of ¹³C detection. An interesting approach to ¹³C detection of uniformly ¹³C and ¹⁵N labeled samples using band-selective homonuclear decoupling has been described by Ishii and Tycko [21]; however, this does not allow the entire ¹³C NMR spectrum to be observed in a single experiment. Thus, there is a need for the development of additional triple-resonances experiments applicable to aligned sample solid-state NMR of ¹³C and ¹⁵N labeled proteins [22], especially those that utilize ¹³C detection for higher sensitivity.

2. Results

Timing diagrams for several two- and three-dimensional ¹H/¹³C/¹⁵N triple-resonance solid-state NMR experiments are shown in Fig. 1. Experimental spectra demonstrating applications of these pulse sequences to ¹³C and ¹⁵N labeled peptides and proteins are presented in Figs. 2–13. Four different samples are used for the demonstrations of pulse sequences:

Fig. 1.

Timing diagrams for triple-resonance solid-state NMR experiments. (a) Two-dimensional ¹H–¹³C or ¹H–¹⁵N PISEMA. The pulse sequence for the triple resonance version of the PISEMA is same as its double resonance version except that low power irradiation (mismatched to the irradiations in the other two channels) is applied during t₁ period at the resonance frequency of the undetected nucleus. (b) Two-dimensional ¹³C–¹⁵N dipolar coupling measurement. The phase of the 180° pulse is same as for the cross-polarization irradiation. The receiver phase is cycled as (x, —x) and the 90° pulse on the ¹H channel is phase cycled as (x, −x). (c) Twoα-dimensional ¹³C/¹³C spin exchange. The 90° pulse after the t₁ period is phase incremented by 90° for quadrature detection. (d) Two-dimensional ¹H–¹⁵N PISEMA with ¹³C detection. (e) Two-dimensional ¹³C/¹⁵N HETCOR. (f) Three-dimensional ¹H–¹³C dipolar coupling/¹³C–¹⁵N dipolar coupling/¹³C chemical shift or ¹H–¹⁵N coupling/¹³C–¹⁵N dipolar coupling/¹⁵N chemical shift correlation.

Fig. 2.

Comparison of ¹³C and ¹⁵N detected one-dimensional NMR spectra. (a–c) are from the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL), and resulted from signal averaging 8 scans. The four resonances in the spectrum correspond to the four molecules in the unit cell of the crystal. (d) and (e) are from the Pf1 coat protein in magnetically aligned phospholipid bilayers where the protein is labeled with ¹³C_α and ¹⁵N at all valine, leucine and isoleucine residues (¹³C_α, ¹⁵N VLI Pf1), and resulted from signal averaging 256 scans. The ¹H and ¹⁵N (or ¹³C) decoupling fields applied during data acquisition were 50 kHz. All of the cross-polarization mix times were 1 msec and the recycle delays 6 s. (a) ¹³C detection following cross-polarization. (b) ¹⁵N detection following cross-polarization. (c) ¹³C detection following double-cross-polarization. (d) ¹⁵N detection following cross-polarization. (e) ¹³C detection following cross-polarization.

Fig. 13.

Results from the three-dimensional experiment diagrammed in Fig. 1f that correlates ¹⁵N chemical shift/¹H–¹⁵N dipolar coupling/¹³C–¹⁵N dipolar coupling frequencies performed on the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL). (a) Three-dimensional cube displaying all four resonances. (b)–(e) Two-dimensional planes at ¹⁵N chemical shift frequencies containing resonance intensity. (f) Two-dimensional projection. 32 points in the ¹H–¹⁵N dimensions (spectral width = 40 kHz) and 21 points in the ¹³C–¹⁵N dimensions (spectral width = 5.0 kHz) were acquired with 4 scans.

1. ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL);

2. ¹⁵N labeled single crystal of N-acetyl valine (NAV);

3. Pf1 coat protein in magnetically aligned phospholipid bilayers where the protein is labeled with ¹³C_α and ¹⁵N at all valine, leucine and isoleucine residues (¹³C_α, ¹⁵N VLI Pf1);

4. Pf1 coat protein in magnetically aligned bacteriophage particles where the protein is labeled ~20% randomly with ¹³C and ~100% uniformly with ¹⁵N Pf1 (¹³C, ¹⁵N Pf1).

The one-dimensional spectra in Fig. 2 demonstrate the improvement (approximately fourfold) in signal to noise ratio that results from ¹³C detection compared to ¹⁵N detection with as many other factors as possible kept constant. The spectra in Fig. 2a – c were obtained from the single crystal sample of ¹³C_α and ¹⁵N labeled N-acetyl-leucine (¹³C_α, ¹⁵N NAL). The four lines in the spectra reflect the four unique molecular orientations in the unit cell of the crystal. The fully decoupled one-dimensional ¹³C Fig. 2a) and ¹⁵N (Fig. 2b) NMR spectra were obtained by conventional single-contact spin-lock cross-polarization [23]. The ¹³C NMR spectrum in Fig. 2c was obtained by double-cross-polarization [24] with transfers of magnetization from ¹H to ¹⁵N to ¹³C in successive steps. The double-cross-polarization procedure is not as effective as direct cross-polarization in generating ¹³C magnetization, nonetheless, the signal to noise ratio of the spectrum in Fig. 2c is substantially higher than that observed with simple ¹⁵N cross-polarization and detection (Fig. 2b). The ¹³C resonance at 62 ppm in the double-cross-polarization spectrum in Fig. 2c has noticeably lower intensity than the other resonances because its very small ¹³C–¹⁵N dipolar coupling (as observed with other experiments, e.g. Fig. 8a) reduces the relative efficiency of the ¹⁵N to ¹³C transfer step. To a lesser extent, this is the case for the ¹³C resonance at 35 ppm, since it too has reduced intensity in experiments with transfers that rely on ¹³C–¹⁵N dipolar couplings. The extent of magnetization transfer between ¹⁵N and ¹³C depends mainly upon the magnitude of the ¹³C–¹⁵N dipolar coupling and the length of the mix time for cross-polarization. The experiments were individually optimized for maximum signal intensity by varying the mix time for the transfer; for relatively weak ¹³C–¹⁵N couplings (<200 Hz), maximum transfer occurs with a 5.0 ms mix time; in contrast, for larger C- N couplings (~1 kHz) a mix time of 1.0 ms yields the largest signals. The comparison of the N- (Fig. 2d) and ¹³C- (Fig. 2e) detected cross-polarization spectra demonstrates similar improvements for a double-labeled membrane protein in magnetically aligned bilayers. In this example, the 46-residue coat protein of the filamentous bacteriophage Pf1 is selectively labeled in the valine, leucine, and isoleucine residues with ¹³C_α and ¹⁵N (¹³C_α, ¹⁵N VLI Pf1). Although a portion of the signal intensity in Fig. 2e comes from the natural abundance ¹³C background from the lipids and the protein, most of the increase in the amplitude relative to that in Fig. 2d is due to signals from the labeled residues.

Fig. 8.

Two-dimensional ¹³C–¹⁵N SLF spectra obtained using the pulse sequence in various mix times in the pulse sequenceFig. 1b. (a) and (b) are from the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL). (c) Is from the ¹⁵N labeled single crystal of N-acetyl valine (NAV). (a) and (c) ¹³C detected. (b) ¹⁵N detected. For spectra a and b, four scans were collected for 48 points in the indirect dimension with 100 µs dwell time and 16 scans were collected for spectrum c.

2.1. PISEMA

The pulse sequence for two-dimensional PISEMA (polarization inversion spin exchange at the magic angle) applicable to ¹³C and ¹⁵N labeled samples shown in Fig. 1a is compatible with either ¹³C or ¹⁵N detection. This triple-resonance version of PISEMA differs from the standard double-resonance experiment [25] in that low power irradiation is applied to the unobserved nucleus during the heteronuclear spin-exchange interval (t₁), and high power irradiation is applied to both ¹H and the unobserved nucleus during data acquisition (t₂). It is intended for situations where homonuclear decoupling of the observed nuclei is unnecessary because of the isotopic or spatial isolation of the labeled sites. In the experiments diagrammed in Fig. 1, the irradiations responsible for heteronuclear decoupling are modulated by SPINAL-16 [19] to increase their bandwidth for high field applications [20] with the exception of the low level continuous wave irradiations applied to the unobserved nuclei in PISEMA-based pulse sequences.

The two-dimensional ¹H–¹³C PISEMA spectra in Fig. 3 demonstrate that all ¹³C sites in peptides, including those without directly bonded hydrogens, e.g. carbonyl and carboxyl carbons, have measurable ¹H–¹³C heteronuclear dipolar couplings. The two-dimensional spectrum in Fig. 3a was obtained on the single crystal sample of ¹³C_α, ¹⁵N NAL and each of the ¹³C_α resonances has a different ¹H–¹³C dipolar coupling frequency between about 3 and 17 kHz. All of the dipolar coupling frequencies are reported as half the total splitting, since one half of the spectrum is displayed, which is symmetric about the zero frequency. For comparison, the spectrum in Fig. 3b was obtained on a single crystal sample of ¹⁵N labeled N-acetylvaline (NAV) with all of the carbon sites in natural isotopic abundance. There are eight distinct resonances from the carbonyl and carboxyl carbons (one each in each of four unique molecules in the unit cell) between 95 and 240 ppm in the spectrum in Fig. 3b, and each of them has discernable ¹H–¹³C heteronuclear dipolar couplings varying between 2 and 4 kHz. The ¹³C_α resonances are recognizable as single peaks between 30 and 55 ppm. Other aliphatic carbons contribute more complex patterns [26] to the spectrum between 10 and 40 ppm. The ¹³C chemical shift and ¹H–¹³C dipolar coupling frequencies and spectral patterns demonstrate the potential for using long-range heteronuclear dipolar couplings as distance and angular constraints for the carbonyl and carboxyl carbons as well as for obtaining angular constraints on the aliphatic side chains.

Fig. 3.

Two-dimensional ¹H–¹³C PISEMA spectra obtained using the pulse sequence in Fig. 1a. (a) Spectrum from the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL) obtained with 4 scans for each of 128 t1 points. (b) Spectrum from the ¹⁵N labeled single crystal of N-acetyl valine (NAV) obtained with 8 scans for each of 128 t1 points. During the heteronuclear spin exchange period (t1), the radiofrequency field strength of 55 kHz power level on the ¹H channel enabled the maximum ¹H–¹³C dipolar couplings to be observed without aliasing. The recycle delay was 6 s. The two-dimensional data sets were zero filled to 1K data points and multiplied by a sine bell window function in both dimensions before double Fourier transformation.

The two-dimensional ¹H–¹³C PISEMA spectrum of a selectively ¹³C_α, ¹⁵N labeled membrane protein in magnetically aligned bilayers is shown in Fig. 4. Pf1 coat protein has a total of fifteen valine, leucine, and isoleucine residues, nine of which are in the transmembrane helix and contribute resonances (within the box) to the spectrum in Fig. 4; their ¹H–¹³C dipolar couplings range between 2.7 and 5.8 kHz with ¹³C chemical shift frequencies between 54 and 70 ppm. Nearly all of the other signals in the spectrum are from the natural abundance background of the lipids, as verified by comparison with a spectrum of a similar sample without ¹³C labels in the protein. These data provide access to two orientationally dependent frequencies at the α carbon sites, which are valuable complements to the previously available frequencies associated with the amide nitrogen sites.

Fig. 4.

Two-dimensional ¹H–¹³ C PISEMA spectrum of the membrane-bound form of the 46-residue Pf1 coat protein in magnetically aligned phospholipid bilayers where the protein is labeled with ¹³C_α and N at all valine, leucine and isoleucine residues (¹³C_α, N VLI Pf1). The spectrum was obtained using the pulse sequence in Fig. 1a. All of the signals in the box are from ¹H–¹³C_α dipolar couplings. Nearly all of the other intensity observed in the spectrum in from the natural abundance signals from the phospholipids in the sample. Ninety six scans were acquired for each of 128 points in the indirect dimension. The recycle delay was 5 s and the temperature 40 °C. The two-dimensional data set was zero filled to 1K data points and multiplied by a sine bell window function in both dimensions before double Fourier transformation.

The two-dimensional NMR spectrum in Fig. 5 was obtained from the structural form of the Pf1 coat protein in magnetically aligned bacteriophage particles [27]. In this sample, the protein is randomly labeled with ~20% ¹³C and ~100% uniformly labeled with N. For economy of presentation, the portion of the C NMR spectrum between 80 and 160 ppm that does not contain signal intensity is not shown. The two-dimensional NMR spectrum in Fig. 5 shows that resonances from carbonyl and carboxyl carbons (160–220 ppm), C_α (55–75 ppm) and aliphatic side chain carbons (5–55 ppm) have substantial heteronuclear ¹H- C dipolar couplings. The several hundred discernable resonances in this spectrum indicate the potential for analyzing essentially all carbon sites in a protein with this approach.

Fig. 5.

Two-dimensional ¹H–¹³ C PISEMA spectrum of the structural form of the 46-residue Pf1 coat protein in magnetically aligned virus particles where the protein is labeled ~20% randomly with ³C and ~100% uniformly with ¹⁵N Pf1 (¹³C and ¹⁵N Pf1). (a) Carbonyl and carboxyl carbon resonances. (b) Aliphatic carbon resonances. Thirty two scans were acquired for each of 128 points in the indirect dimension. The recycle delay was 6 s, and the temperature 0 °C. The central ¹H carrier frequency during t₁ was +2 ppm relative to the water resonance during the PISEMA portion of the pulse sequence. The two-dimensional data set was zero filled to 1K data points and multiplied by a sine bell window function in both dimensions before double Fourier transformation.

Two-dimensional ¹H–¹⁵N PISEMA experiments can be performed with ¹³C detection using the pulse sequence shown in Fig. 1d. After the t₁ period, during which the spin exchange between ¹H and ¹⁵N nuclei takes place, the ¹⁵N magnetization is transferred to ¹³C by cross-polarization. The ¹³C signals modulated by the ¹H–¹⁵N heteronuclear dipolar couplings are detected during t₂ in the presence of both ¹H and ¹⁵N heteronuclear decoupling. A short (1.5 ms) ¹⁵N to ¹³C cross-polarization mix time ensures selective one bond polarization transfers. The resulting two-dimensional spectrum correlates ¹H–¹⁵N dipolar coupling and ¹³C chemical shift frequencies.

Fig. 6 contains ¹H–¹⁵N PISEMA spectra of the single crystal sample of ¹³C_α, ¹⁵N NAL. The spectrum in Fig. 6a was obtained using the pulse sequence in Fig. 1a with ¹⁵N detection. The four resonances have ¹H–¹⁵N dipolar couplings ranging from 300 Hz to 6.0 kHz at this orientation of the crystal. Fig. 6b is a ¹³C-detected ¹H–¹⁵N PISEMA spectrum acquired using the pulse sequence shown in Fig. 1d. The double-cross-polarization mix time for the transfer of the magnetization from ¹⁵N to ¹³C was tuned for one bond correlation (1.0 ms). However, the signal at 62 ppm does not yield a dipolar coupling frequency due to the effects of its very small ¹³C–¹⁵N dipolar coupling on cross-polarization efficiency. The signal associated with the ¹³C resonance at 35 ppm is observable at lower contour levels, consistent with its relatively weak intensity.

Fig. 6.

Two-dimensional ¹H–¹⁵N PISEMA spectra of the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL). (a) ¹⁵N detected. (b) ¹³C detected. Sixteen scans were acquired for each of 128 points in the indirect dimension for spectrum a and 32 scans for spectrum b. The contact time was 1.0 ms during the double cross-polarization from ¹⁵N to ¹³C, and the recycle delay was 5 s.

2.2. ¹³C/¹³C homonuclear spin-exchange

A pulse sequence for two-dimensional ¹³C/¹³C homonuclear spin-exchange [28] spectroscopy is diagrammed in Fig. 1c. Following cross-polarization, the ¹³C magnetization undergoes chemical shift evolution during t₁ in the presence of both ¹⁵N and ¹H heteronuclear decoupling; it is then positioned along the z-axis through application of a π/2 pulse with the appropriate phase. After the fixed-length mixing period t_mix during which the z component of ¹³C magnetization evolves in the presence of homonuclear ¹³C/¹³C dipolar couplings, the ¹³C magnetization is returned to the transverse plane where it is detected in the presence of H and N hetero-nuclear decoupling.

Fig. 7 contains two-dimensional ¹³C/ ¹³C homo-nuclear spin-exchange spectra of the single crystal sample of ¹³C_α, ¹⁵N NAL. These spectra were obtained at four different mix times (t_mix) using the pulse sequence diagrammed in Fig. 1c. The four peaks along the diagonal correspond to the ¹³C_α resonances from the four molecules in the unit cell of the crystal. The off-diagonal cross-peaks result from homonuclear ¹³C/¹³C spin exchange between α carbons located on different molecules in the crystal. The internu-clear distances between ¹³C_α sites on different molecules in the crystal vary from 5.7 Å (~40 Hz) to 8.4 A (~13 Hz), and the corresponding splittings are too small to be observed directly in the experimental spectra. For the shorter mix times (1.0–2.0 s) only a limited amount of correlation intensity is observed in the spectra. For the longest mix time (4.5 s), complete correlation among all four molecules in the unit cell can be observed. Since the cross-peaks in these spectra arise from through-space interactions, the experiment strongly discriminates towards proximate pairs of nuclei, and thus provides information about distances between alpha carbons useful for making resonance assignments and as input for structure calculations.

Fig. 7.

Two-dimensional homonuclear ¹³C/¹³C spin-exchange spectra obtained from the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL) at various mix times in the pulse sequence in Fig. 1c. (a) 0.5 s. (b) 1.0 s. (c) 2.0 s. (d) 4.5 s. Eight scans for 90 t1 points were acquired with a 50 µs dwell for each spectrum. The ¹H and ¹⁵N decoupling fields applied during the experiment were 50 kHz.

2.3. ¹³C–¹⁵ N heteronuclear dipolar couplings

The pulse sequence in Fig. 1b can be used to measure ¹³C–¹⁵N hetero–nuclear dipolar couplings. The ¹³C (or ¹⁵N) magnetization generated by cross-polarization from ¹H evolves during t₁ in the presence of 1H decoupling. The chemical shift evolution that occurs during the t₁ period is refocused and the ¹³C–¹⁵N dipolar coupling preserved by the simultaneous application of πpulses in the middle of the period at both the ¹³C and ¹⁵N resonance frequencies. During t₂, the ¹³C (or ¹⁵N) magnetization modulated by the ¹³C–¹⁵N coupling is observed in the presence of ¹H and ¹⁵N (or ¹³C) heteronuclear decoupling. The resulting two-dimensional spectra have the chemical shift frequencies of the detected (¹³C or ¹⁵N) nuclei along the frequency axis associated with the direct dimension, and the ¹³C–¹⁵N heteronuclear dipolar coupling frequencies symmetric about the zero frequency of the indirect dimension. A version of this experiment with multiple π pulse during t₁ on the unobserved nuclei with XY-16 [29] phase cycling is more tolerant of imperfections in the π pulses.

The spectra in Fig. 8 demonstrate the measurement of resolved ¹³C–¹⁵N couplings using the pulse sequence in Fig. 1b. The ¹³C-detected (Fig. 8a) and ¹⁵N-detected (Fig. 8b) spectra were obtained on the single crystal sample of ¹³C_α, ¹⁵N NAL. In both spectra the same ¹³C–¹⁵N coupling frequencies are observed, ranging from very small, with no detectable splitting, to 900 Hz. Comparison to the ¹⁵N-detected spectrum in Fig. 8b provides a method for cross-assigning ¹³C and ¹⁵N resonances associated with the same residue in a protein.

The data in Fig. 8c were obtained from the single crystal sample of ¹⁵N labeled N-acetyl-valine (NAV). The one-dimensional ¹³C NMR spectrum obtained by cross-polarization is aligned along the ¹³C chemical shift frequency axis. Four of the carbonyl peaks have vanishingly small ¹³C–¹⁵N splitting, as expected for carboxyl carbons, and four have measureable ¹³C–¹⁵N splittings between 200 Hz and 1 kHz, as expected for the groups in the peptide linkages. The a carbon signals appear in the range between 40 and 70 ppm, and the other aliphatic carbons give overlapping resonance patterns, as expected, between about 0 and 40 ppm.

As illustrated with the spectra in Fig. 9, it is also possible to observe the ¹³C–¹⁵N dipolar couplings in two-dimensional PISEMA spectra through the simple expedient of acquiring the data in the absence of heteronuclear decoupling of the unobserved nucleus. The same four splittings, one of which is very small, along the ¹³C chemical shift axis (Fig. 9a) and the ¹⁵N chemical shift axis (Fig. 9b) are the same. The spectra are more crowded than those in Fig. 8, but clearly provide complementary opportunities for measuring ¹³C–¹⁵N dipolar couplings and compiling data that will be useful for making resonance assignments.

Fig. 9.

Two-dimensional PISEMA spectra of the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL). (a) ¹H–¹³C PISEMA spectrum obtained without ¹⁵N decoupling during acquisition (t₂). (b) ¹H–¹⁵N PISEMA spectrum obtained without ¹³C decoupling during acquisition (t₂). For the 2D spectrums, 4 scans were added for each t₁ points (128). During spin exchange period, 55 kHz power level on the ¹H channel were used. The two-dimensional data sets were zero filled to 1K data points and multiplied by sine bell window function before double Fourier transformation. The one-dimensional ¹³C (a) and ¹⁵N (b) NMR spectra obtained without decoupling of the unobserved nucleus are aligned along the corresponding two-dimensional spectra.

2.4. ¹³C/¹⁵N HETCOR

The pulse sequence in Fig. 1e correlates the chemical shift frequencies of directly bonded ¹⁵N and ¹³C nuclei, and yields two-dimensional ¹³C/¹⁵N HETCOR spectra with ¹³C and ¹⁵N chemical shift frequencies along separate axes. The ¹⁵N magnetization generated by cross-polarization from ¹H undergoes chemical shift evolution during the incremented t₁ interval in the presence of ¹H and ¹³C heteronuclear decoupling. The ¹⁵N magnetization is then transferred to ¹³C by selective cross-polarization, and the ¹³C signals are detected in the presence of 1H and ¹⁵N heteronuclear decoupling. The phases of the 90° pulses labeled “a” and “b” are chosen to provide phase sensitive quadrature detection in the indirect dimension using the method of States et al. [30].

The presence of both ¹³C and ¹⁵N in the labeled molecules provides opportunities for ¹³C/¹⁵N HETCOR experiments. Fig. 10 contains a two-dimensional ¹³C/¹⁵N HETCOR spectrum of the single crystal sample of C_α, N NAL. As indicated by the one-dimensional spectra aligned along the ¹³C chemical shift and ¹⁵N chemical shift frequency axes, four resonances are expected. Only three resonances are observed due to the very small ¹³C–¹⁵N dipolar coupling at the site with a C chemical shift frequency of 62 ppm. Fig. 11 contains a two-dimensional ¹³C¹⁵N HETCOR spectrum of the same ~20% ¹³C, and ~100% ¹⁵N labeled protein sample used to obtain the data in Fig. 5. The ¹³C/¹⁵N heteronuclear correlation resonances in Fig. 11a for the carbonyl carbons and in Fig. 11b for the a carbons reflect the proximity of these sites to the ¹⁵N labeled amide sites in the peptide linkages in the backbone of the protein. Notably, there are no ¹³C/¹⁵N correlation resonances associated with the side chain carbons with ¹³C chemical shifts between 0 and 40 ppm. This demonstrates the selectivity of the experiments for sites where ¹³C and ¹⁵N are directly bonded.

Fig. 10.

Two-dimensional ¹⁵N/¹³C HETCOR spectrum of the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL) obtained using the pulse sequence in Fig. 1e. Four scans for each of 90 t₁ points were acquired in the indirect dimension with a dwell of 50 µs. The ¹H decoupling field during the t₁ and t₂ intervals was 50 kHz. The recycle delay used for the experiment was 6 s.

Fig. 11.

Two-dimensional ¹⁵N/ ¹³C HETCOR spectrum of the magnetically aligned sample of Pf1 bacteriophage with the coat protein labeled with both ¹³C and ¹⁵ N. The spectrum was acquired at 0 °C. Thirty two scans of 80 complex points were acquired in the indirect dimension with a spectral width of 30 kHz. The recycle delay was 6 s. The data were zero filled to 1K points and multiplied by a sine bell window function in both dimensions prior to double Fourier transformation. The carrier frequency for ¹³C irradiation was in the center of the spectrum, between the aliphatic and carbonyl resonance regions, and the double cross-polarization was optimized for both types of resonances.

2.5. Three-dimensional correlation

Three-dimensional experiments will be needed to take full advantage of the possibilities provided by the triple-resonance experiments on proteins with many ¹³C and ¹⁵N labeled sites. The increase in sensitivity that accompanies the use of ¹³C detection will contribute to the feasibility of routinely applying three-dimensional experiments to aligned protein samples. A three-dimensional triple-resonance correlation experiment is diagrammed in Fig. 1f. The t₁ period of this pulse sequence is the same as that of the PISEMA experiment, and results in the measurement of the ¹H–¹³C (or ¹H–¹⁵N) dipolar couplings. During the t₂ period, the transverse ¹³C (or ¹⁵N) magnetization evolves under the influence of the heteronuclear ¹³C–¹⁵N dipolar coupling, as in the pulse sequence shown in Fig. 1b. Finally, during t₃, ¹³C (or ¹⁵N) magnetization is detected in the presence of two-channel heteronuclear decoupling. The resulting three-dimensional spectra have the ¹³C chemical shift correlated with ¹H–¹³C and ¹³C–¹⁵N dipolar couplings. The complementary ¹⁵N detected version of the experiment yields three-dimensional spectra where the ¹⁵N chemical shifts are correlated with ¹H–¹⁵N and ¹³C–¹⁵N dipolar couplings.

Three-dimensional spectra of the single crystal of ¹³C_α, ¹⁵N NAL are shown in Figs. 12 and 13. The data are presented as three-dimensional cubes, two-dimensional projections, and two-dimensional planes at frequencies corresponding to the four resonances resolved in the chemical shift dimension. Fig. 12 contains results from the ¹³C detected version of the three-dimensional experiment, and Fig. 13 the ¹⁵N-detected version. The cross-sections along the chemical shift planes have ¹H–¹³C (or ¹H–¹⁵N) and ¹³C–¹⁵N dipolar coupling frequency axes. The peaks are labeled in the figures to indicate the cross-assignments between the two experiments. It can be seen that the cross-section of the three-dimensional spectrum correlates two dipolar couplings, ¹H–¹³C/¹³C–¹⁵N dipolar coupling in ¹³C detected versions and ¹H–¹⁵N/¹³C–¹⁵N dipolar couplings in ¹⁵N detected versions. Since the cross-sections from the two versions of the three-dimensional experiments have a common axis of ¹³C–¹⁵N dipolar couplings, this experiment can also complement the chemical shift correlation between ¹³C and ¹⁵N resonance.

Fig. 12.

Results from the three-dimensional experiment diagrammed in Fig. 1f that correlates ¹³C chemical shift/¹H–¹³C dipolar coupling/¹³C–¹⁵N dipolar coupling frequencies performed on the ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α, ¹⁵N NAL). (a) Three-dimensional cube displaying all four resonances. (b)–(e) Two-dimensional planes at ¹³C chemical shift frequencies containing resonance intensity. (f) Two-dimensional projection. 32 points in the ¹H–¹³C dimensions (spectral width = 40 kHz) and 21 points in the ¹³C–¹⁵N dimensions (spectral width = 5.0 kHz) were acquired with 4 scans.

3. Discussion

In aligned sample solid-state NMR, the structure and topology of a protein are mapped onto the spectra by the orientation-dependent frequencies resulting from the anisotropic chemical shift and dipole–dipole interactions. Using ¹H/¹⁵N double resonance experiments it has been possible to determine the structures of a number of proteins in virus particles and phospholipid bilayers. However, much more structural information can be gained from the use of ¹³C and ¹⁵N labeled proteins.

The principal advantage of using samples labeled with ¹³C is the increase in sensitivity associated with the detection of ¹³C signals. In the peptide and protein examples in Fig. 2, approximately a factor of four increase in signal to noise ratio is observed for ¹³C detection compared to ¹⁵N detection. This can significantly decrease the amount of time required for signal averaging.

The triple-resonance experiments feasible with ¹³C and ¹⁵N labeled samples also provide access to additional orientationally dependent frequencies from the ¹³C chemical shift and the ¹H–¹³C and ¹³C–¹⁵N heteronuclear dipolar couplings. By themselves, these three frequencies are sufficient to determine the orientation of a peptide plane with respect to the direction of the magnetic field and bilayer normal. Moreover, combinations of ¹H–¹⁵N, ¹H–¹³C, and ¹³C–¹⁵N dipolar couplings are sufficient to calculate protein structures, and eliminating chemical shift frequencies from the structure calculations has the potential to improve the quality of the resulting structures by eliminating the uncertainties associated with the residue-to-residue variations of the chemical shift tensors.

These experiments provide opportunities for making resonance assignments. ¹³C/¹³C spin-exchange experiments identify nearby carbon sites by allowing for the broad classification of carbon resonances based on homonuclear ¹³C/¹³C dipolar coupling strengths. The ability to measure ¹³C–¹⁵N dipolar coupling frequencies and correlate ¹³C and ¹⁵N resonances provides a mechanism for making backbone resonance assignments. Three-dimensional combinations of these experiments ensure that the resolution, assignment, and the measurement of orientationally dependent frequencies can be extended to larger proteins.

These and other opportunities for improvements in resolution, sensitivity, and spectroscopic analysis arising from applications of ¹³C-detected triple-resonance spectroscopy to aligned protein samples have been reviewed by Vosegaard and Nielsen [22]. Our earlier implementations of triple-resonance spectroscopy were compatible with ¹³C detection only in a few favorable situations [13–18]. Ishii and Tycko [21] successfully implemented an approach that is generally compatible with uniformly ¹³C labeled proteins, although it is limited in the range of multi-dimensional experiments it can be applied to by its band-selective characteristics. By combining isotopic labeling schemes with the pulse sequences, we demonstrate the feasibility of triple-resonance spectroscopic experiments capable of simultaneous interrogation of all sites in a protein.

In summary, triple-resonance solid-state NMR experiments provide improved sensitivity through ¹³C detection, enhance resolution through additional frequency dimensions, enable measurements of orientation-dependent frequencies for all backbone and side chains sites, and provide mechanisms for making resonance assignments.

4. Experimental

4.1. Samples

The samples used to demonstrate the two- and three-dimensional triple-resonance experiments are single crystals of ¹³C_α, ¹⁵N labeled single crystal of N-acetyl leucine (¹³C_α,¹⁵N NAL) and ¹⁵N labeled N-acetyl valine (NAV); Pf1 coat protein in magnetically aligned phospholipid bilayers where the protein is labeled with _α and ¹⁵N at all valine, leucine and isoleucine residues (¹³C_α, ¹⁵N VLI Pf1); and Pf1 coat protein in magnetically aligned bacteriophage particles where the protein is labeled ~20% randomly with ³C and ~100% uniformly with ¹⁵N Pf1 (¹³C, ¹⁵N Pf1). The single crystal of ¹³C_α, ¹⁵N NAL was prepared by slow evaporation of aqueous solution after acetylating ¹³C_α and ¹⁵N labeled leucine (Cambridge Isotope Laboratories, Andover, MA). Bacteriophage Pf1 was prepared as described previously [27]. Uniformly ~20% C and ~100% ¹⁵N labeled bacteriophage was prepared using growth media obtained from Cambridge Isotope Laboratories, Inc. and selectively (Val, Leu, and Ile) ¹³C_α and ¹⁵N labeled Pf1 was prepared using SLS media to avoid isotopic scrambling [31].

The sample of Pf1 coat protein in magnetically aligned bilayers was prepared as described previously [32,33]. The phospholipids were 1,2-dimyristoyl-sn-glycero-3-phospho-choline (DMPC) and 1,2-di-O-hexyl-sn-glycero-3-phospho-choline (6-O-PC) from Avanti Polar Lipids, Inc. (Alabaster, AL). Five milligrams of the lyophilized protein was dissolved in 200 µl of 100 mM 6-O-PC solution. The protein/6-O-PC solution was transferred to 43.4 mg of lyophilized DMPC. The resulting solution was vortexed and freeze-heated several times until it becomes a clear, nonviscous solution on ice. The molar ratio between long-chain lipids and short-chain lipids (q) was 3.2 and a lipid concentration was 26% (w/v). The pH of the sample was adjusted to 6.8 on ice. A flat-bottom NMR tube with a 5-mm outer diameter (New Era Enterprises, Vineland, NJ) was filled with 160 µl of the solution.

4.2. NMR instrumentation

The experimental spectra were obtained on a Varian Ino-va spectrometer with a Magnex 500/89 AS magnet using a homebuilt probe with a single 5 mm ID coil triple tuned to the ¹H, ¹³C and ¹⁵N resonance frequencies of 500.09, 125.76 and 50.68 MHz [18]. The ¹³C and ¹⁵N channels were isolated from each other using band-stop (parallel resonance) filters, which provide ~35 dB of isolation. This design is robust enough to withstand 25 ms of continuous irradiation on all three RF channels at power levels of 500, 200 and 100 W on the ¹⁵N, ¹³C and ¹H channels, respectively. For all the experiments, B₁ field strengths of 50 kHz were used unless otherwise indicated. ¹³C chemical shifts are referenced by defining the high field peak of external adamantane to be 38.4 ppm and ¹⁵N chemical shifts by defining the resonance of external ammonium sulphate to be 26.8 ppm [34].