Two-Dimensional ¹⁹F–¹³C Correlation NMR for ¹⁹F Resonance Assignment of Fluorinated Proteins

Abstract

¹⁹F solid-state NMR is an excellent approach for measuring long-range distances for structure determination and for studying molecular motion. For multi-fluorinated proteins, assignment of ¹⁹F chemical shifts has been traditionally carried out using mutagenesis. Here we show 2D ¹⁹F-¹³C correlation experiments that allow efficient assignment of the ¹⁹F chemical shifts. We have compared several rotational-echo double-resonance (REDOR)-based pulse sequences and ¹⁹F-¹³C cross polarization (CP) for 2D ¹⁹F-¹³C correlation. We found that direct transferred-echo double-resonance (TEDOR) transfer from ¹⁹F to ¹³C and vice versa outperforms out-and-back coherence transfer schemes. ¹⁹F detection gives 2-fold higher sensitivity over ¹³C detection for the 2D correlation experiment. At MAS frequencies of 25–35 kHz, double-quantum ¹⁹F-¹³C CP has higher coherence transfer efficiencies than zero-quantum CP. The most efficient TEDOR transfer experiment has higher sensitivity than the most efficient double-quantum CP experiment. We demonstrate these 2D ¹⁹F-¹³C correlation experiments on the model compounds t-Boc-4F-phenylalanine and GB1. Application of the 2D ¹⁹F-¹³C TEDOR correlation experiment to the tetrameric influenza BM2 transmembrane peptide show intermolecular ¹³C-¹⁹F cross peaks that indicate that the BM2 tetramers cluster in the lipid bilayer in an antiparallel fashion. This clustering may be relevant for the virus budding function of this protein.

Introduction

The application of solid-state NMR spectroscopy to molecular structure determination chiefly relies on the chemical shift interaction and the nuclear spin dipole-dipole interaction. The dipole-dipole interaction is proportional to the gyromagnetic ratios of the two spins and the internuclear distance r according to 1/r³. The larger the γ, the stronger the dipolar coupling, and the longer the measurable distances. Although ¹H has the highest γ among stable spin-1/2 isotopes, the high ¹H density in biological and organic molecules causes truncation of the structurally informative weak dipolar couplings by the strong dipolar couplings (Bayro et al. 2009). In comparison, the second highest-γ nucleus, ¹⁹F, has numerous benefits for structure determination (Danielson and Falke 1996; Kitevski-LeBlanc and Prosser 2012; Sharaf and Gronenborn 2015). Fluorine is 100% abundant, requiring no enrichment. It does not naturally occur in biological molecules but can be readily and sparsely introduced into protein sidechains with minimal perturbation to structure (Salwiczek et al. 2012). Fluorine is also already incorporated into many organic compounds and ~30% of pharmaceutical compounds (Gee et al. 2016; Wang et al. 2014). These properties make ¹⁹F an excellent nuclear spin for studying the structure and dynamics in biological macromolecules and pharmaceutical compounds, as well as for characterizing interactions between small molecules and biological macromolecules.

Ideally, ¹⁹F-based dipolar coupling measurements should be conducted at high magnetic fields where most ¹³C and ¹⁵N-based structure determination NMR experiments are carried out. However, ¹⁹F has a large isotropic and anisotropic chemical shift range (Dürr et al. 2008; Lu et al. 2018), which complicates dipolar coupling measurements. Recently, we explored the capability of high-field ¹⁹F solid-state NMR for distance measurements by conducting 2D ¹⁹F–¹⁹F correlation experiments at 14.1 Tesla under 20–55 kHz MAS. Both spin diffusion (Roos et al. 2018b) and homonuclear dipolar recoupling (Gilchrist Jr et al. 2001; Roos et al. 2018a) approaches were found to permit efficient ¹⁹F-¹⁹F polarization transfer, and cross peaks for internuclear distances up to ~2 nm can be observed readily using mixing times of ~100 ms for spin diffusion and ~5 ms for radiofrequency-driven dipolar recoupling (RFDR). The spin diffusion buildup rates give semi-quantitative distances after the ¹⁹F isotropic chemical shift difference is taken into account (Roos et al. 2018b), whereas RFDR buildup rates can be simulated to give distance restraints independent of the ¹⁹F chemical shift tensor orientation (Roos et al. 2018a). The 2D ¹⁹F-¹⁹F correlation experiment was also shown to give cross peaks corresponding to ~2 nm distances in the HIV capsid protein (Wang et al. 2018).

In addition to ¹⁹F-¹⁹F distances, 2D-resolved heteronuclear ¹³C-¹⁹F, ¹⁵N-¹⁹F, and ¹H-¹⁹F distance techniques have also been recently described. Using 2D ¹³C-¹³C correlation spectra (Shcherbakov and Hong 2018), ¹H-¹⁵N correlation spectra (Shcherbakov et al. 2019) and ¹³C-¹⁵N correlation spectra (Graesser et al. 2007) to resolve the resonances, one can measure many ¹⁹F-X distances as peak intensity decays between a pair of REDOR control and dephased spectra (Gullion and Schaefer 1989). The multiplexing by 2D correlation spectra allows many nanometer distances to be obtained rapidly from each pair of 2D spectra. These ¹³C-¹⁹F, ¹⁵N-¹⁹F, and ¹H-¹⁹F distances supplement chemical-shift-derived torsion angles, and have been shown to restrain the structure of the model protein GB1 to high resolution (Shcherbakov and Hong 2018; Shcherbakov et al. 2019). High-sensitivity 2D and 3D ¹H-detected ¹H-¹⁹F correlation experiments at 12 – 65 kHz MAS have also been demonstrated on pharmaceutical compounds for structure determination (Lu et al. 2019c).

Despite the capability of ¹⁹F SSNMR to provide structurally informative long-range distance constraints, the presence of multiple fluorinated residues of the same type in a protein requires assignment of the ¹⁹F spectra. In the absence of single-site mutations, such assignment is most readily achieved by correlating the ¹⁹F chemical shift with the ¹³C chemical shifts of neighboring residues. Since production of simultaneously ¹³C-labeled and ¹⁹F-tagged amino acids is still at an early stage of development (Boeszoermenyi et al. 2019), we consider the more accessible situation of fluorinated residues with natural abundance in ¹³C. Under this condition, the closest ¹³C spins in an otherwise uniformly ¹³C-labeled protein come from residues sequential to the fluorinated residue, with ¹³C-¹⁹F distances in the 4 – 6 Å range. In this distance range, the ¹⁹F spin polarization can in principle be transferred to ¹³C by cross polarization (CP) or by REDOR-based pulse sequences (Gullion and Schaefer 1989), and which method is more efficient has not been investigated.

In this study, we compare the sensitivities of various ¹⁹F–¹³C REDOR pulse sequences for 2D ¹⁹F-¹³C correlation under fast MAS. We show that “out-and-back” (OaB) pulse sequences have very low sensitivity, indicating fast ¹⁹F relaxation processes, whereas direct ¹³C→¹⁹F or ¹⁹F→¹³C TEDOR (Hing et al. 1992) have significantly higher sensitivity. For small molecules for which ¹³C labeling is difficult, the full molecular structure can be determined from ¹⁹F–¹³C distance constraints alone. We also compare double-quantum (DQ) and zero-quantum (ZQ) CP for ¹⁹F-¹³C polarization transfer, and show that at MAS frequencies higher than 25 kHz, DQ CP outperforms ZQ CP. The relative efficiencies of the most sensitive TEDOR pulse sequence and DQ CP depend on the T₂ and T_1ρ relaxation times of the biomolecule of interest.

Materials and Methods

Fluorinated compounds

Three fluorinated compounds were used in this study, t-Boc-4F-Phenylalanine (t-Boc-4F-Phe), GB1, and the influenza B M2 (BM2) transmembrane (TM) peptide. t-Boc-4F-Phe is natural abundance in ¹³C, and was packed into a 4 mm MAS rotor (~45 mg) and a 1.9 mm rotor (~11 mg) for the experiments. The model protein GB1 is uniformly ¹³C, ¹⁵N-labeled and fluorinated at the three Tyr residues as 3F-Tyr. The protein was expressed in E. coli. (Shcherbakov and Hong 2018) and crystallized in pH 5.5 phosphate buffer using 2-methyl-2, 4-pentanediol (MPD) and isopropanol 2 : 1 v/v (Franks et al. 2005). A total of 8.5 mg GB1 crystals, containing organic solvent in addition to the protein, were packed into a 1.9 mm MAS rotor. The BM2-TM peptide (residues 1–33) contains ¹³C, ¹⁵N-labeled residues at S12, I14, H19, and A22, and fluorinated 4-¹⁹F-Phe5 and 5-¹⁹F-Trp23. The peptide contains a H27A mutation compared to the wild-type sequence, and was synthesized using Fmoc solid-phase peptide synthesis and reconstituted into a virus-mimetic lipid mixture at pH 5.5 (Kwon et al. 2019). A total of 8.5 mg of hydrated membrane sample containing 1.2 mg of ¹³C, ¹⁵N-labeled peptide was packed into a 1.9 mm MAS rotor.

Solid-state NMR experiments

Solid-state NMR experiments were carried out on Bruker Avance III HD 400 MHz (9.4 Tesla) and Avance III HD 600 MHz (14.1 Tesla) spectrometers. The ¹⁹F Larmor frequencies are 375.72 MHz at 9.4 Tesla and 564.66 MHz at 14.1 Tesla. A 4 mm HFX MAS probe with a maximum MAS frequency of 14 kHz was used for the 400 MHz experiments, and a 1.9 mm HFX MAS probe with a maximum MAS frequency of 42 kHz was used for the 600 MHz experiments. Chemical shifts were externally referenced to the 38.48 ppm CH₂ peak of adamantane on the tetramethylsilane scale for ¹³C and the −122.1 ppm peak of 5-¹⁹F-Trp on the CF₃Cl scale for ¹⁹F (Dürr et al. 2008). All spectra were processed using the Bruker TopSpin software.

The direct TEDOR and OaB REDOR spectra were measured on the 400 MHz spectrometer under 12 kHz MAS. The ¹³C magnetization was established using 1 ms ¹H-¹³C ramped CP, with a constant ¹H spin-lock field strength of 50 kHz and a 90–100% ¹³C spin-lock field strength centered at 62 kHz. ¹H-¹⁹F CP was achieved using a contact time of 0.5 ms. During the ¹³C-¹⁹F TEDOR period, the ¹³C and ¹⁹F radiofrequency (rf) field strengths were 40 kHz while the ¹H TPPM decoupling field strength was 83 kHz. During acquisition the ¹H decoupling field was 71 kHz. Each REDOR spectrum was signal-averaged with 4096 scans and the TEDOR mixing times varied from 0.33 ms to 7 ms to detect the buildup curves.

We compared the performance of TEDOR and OaB REDOR experiments on the 600 MHz spectrometer under 35 kHz MAS. A set temperature of 253 K was used, corresponding to an estimated sample temperature of 298 K. 1D ¹⁹F→¹³C TEDOR experiments as a function of mixing times were initiated using ¹H-¹⁹F CP, where the ¹⁹F spin-lock field strength was 71 kHz at the midpoint of a 80–100% ramp and the ¹H spin-lock field strength was 106 kHz. The ¹⁹F→¹³C TEDOR pulse train used 6 μs ¹⁹F and ¹³C 180˚ pulses while ¹H TPPM decoupling used an rf field strength of 130 kHz. During ¹³C detection, weak 10 kHz WALTZ ¹H decoupling was applied and was found to give sufficient spectral resolution. The TEDOR mixing times varied from 0.11 ms to 7.2 ms. The ¹³C→¹⁹F→¹³C OaB REDOR experiments were conducted using similar rf field strengths.

We also carried out ZQ and DQ ¹⁹F→¹³C CP experiments to compare with the TEDOR coherence transfer. These ¹⁹F-¹³C CP spectra were measured under 25 and 35 kHz MAS. The initial ¹⁹F magnetization was established by 1 ms ¹H-¹⁹F CP with a ¹⁹F spin-lock field strength of 80 kHz and ¹H spin-lock field strength of 45 or 55 kHz. For ¹⁹F→¹³C ZQ CP, we tested rf field strengths (ν_1H, ν_19F, ν_13C) of (130 kHz, 62 kHz, 27 kHz), (130 kHz, 80 kHz, 45 kHz), (10 kHz, 80 kHz, 45 kHz), and (130 kHz, 45 kHz, 80 kHz). The ¹⁹F spin lock used a 70–100% ramp where the indicated field strength referred to the middle of the ramp. The ¹⁹F→¹³C CP contact times varied from 0.05 ms to 3 ms. For ¹⁹F→¹³C DQ CP under 35 kHz MAS, we tested rf field strengths of (10 kHz, 15 kHz, 20 kHz), (120 kHz, 15 kHz, 20 kHz), (120 kHz, 20 kHz, 15 kHz), (120 kHz, 10 kHz, 25 kHz), (118 kHz, 10 kHz, 15 kHz), (118 kHz, 15 kHz, 10 kHz). The DQ CP contact times ranged from 0.5 to 7 ms. We also conducted the ¹⁹F→¹³C DQ CP experiments on 3F-Tyr, ¹³C, ¹⁵N-labeled GB1. For these experiments the (ν_1H, ν_19F, ν_13C) rf field strengths were (110 kHz, 10 kHz, 25 kHz) and the contact times were 1 ms, 5 ms, and 8 ms .

We measured the 2D ¹⁹F-¹³C TEDOR correlation spectra of t-Boc-4F-Phe using both ¹³C and ¹⁹F detection to compare the sensitivities. These spectra were measured at 600 MHz under 35 kHz MAS. The ¹³C-detected ¹⁹F-¹³C correlation experiment started with ¹H-¹⁹F CP, signal-averaged 256 scans per t₁ slice and collected 128 indirect ¹⁹F t₁ slices. The maximum ¹⁹F evolution time was 7.3 ms while the ¹³C acquisition time was 20 ms. For the 0.46 ms TEDOR mixing time, 10 kHz WALTZ decoupling was applied on both the ¹H and ¹⁹F channels during ¹³C detection to suppress the one-bond ¹³C-¹⁹F J-coupling. At longer TEDOR mixing times, the 2D spectra did not exhibit the one-bond correlation peak, thus no ¹⁹F decoupling was applied during ¹³C detection.

The ¹⁹F-detected 2D ¹³C-¹⁹F TEDOR correlation spectra were measured using 56 scans per t₁ slice and 650 ¹³C t₁ slices. The maximum ¹³C evolution time was 9.3 ms while the ¹⁹F acquisition time was 9.2 ms. For fair comparison of the 2D sensitivity, we measured the 3.89 ms mixing ¹⁹F-detected and ¹³C-detected 2D correlation spectra both in 18 hours. The spectra were processed to have identical ¹³C and ¹⁹F acquisition times.

We also conducted 1D and 2D ¹⁹F-¹³C TEDOR experiments on GB1. The 1D spectra were measured using TEDOR mixing times of 0.46, 1.37, 1.83, and 2.29 ms. Each 1D spectrum was signal averaged to 30,000 – 55,000 scans. The 2D correlation spectrum was measured using a mixing time of 1.37 ms, 3584 scans per t₁ slice, and 48 t₁ slices. The maximum ¹⁹F evolution time was 4.8 ms while the ¹³C acquisition time was 15 ms.

2D ¹³C-¹⁹F TEDOR correlation spectra of H27A BM2-TM were measured under 35 kHz MAS at an estimated sample temperature of ~275 K. The 2D spectra were measured with initial ¹H-¹³C CP and ¹⁹F detection. The TEDOR mixing times were 1.49 ms and 2.97 ms. The 1.49 ms spectra were acquired with 10 kHz WALTZ ¹H decoupling during ¹⁹F detection, 1024 scans per t₁ slice, 216 t₁ slices, and a maximum ¹³C evolution time of 3.6 ms. The 2.97 ms TEDOR spectrum was measured with 2048 scans per t₁ slice and 60 t₁ slices, corresponding to a maximum ¹³C evolution time of 0.85 ms and a ¹⁹F acquisition time of 3 ms.

TEDOR simulations

Measured TEDOR buildup curves were simulated in Mathematica using numerical implementation of the powder averaging integral for a dipolar coupled two-spin system (Hing et al. 1993). The simulated TEDOR intensity depends on the recoupling time nt_r, the internuclear distance r, a relaxation time constant$T_{rlx}$, and an intensity scaling factor, A. Specifically, the simulated TEDOR intensity can be written as

$$I\left(n{t}_r;r,T_{rlx},A\right)=A·f\left(n{t}_r;r\right)·g\left(n{t}_r,T_{rlx}\right)$$

where f(nt_r; r) is the powder averaging integral (Hing et al. 1993), and $g\left(n{t}_r,T_{rlx}\right)$is the exponential relaxation decay. For the fitting, we first guessed a value of r, and specified a range of possible A and $T_{rlx}$ values. The values of $I\left(n{t}_r;r,T_{rlx},A\right)$ were then calculated for all the experimental mixing times across all possible A and $T_{rlx}$ values, and the best-fit simulation for a given r was determined by the minimum χ² value. The internuclear distance was then varied until the χ² value converged to a minimum. The best-fit parameters r, $T_{rlx}$ and A were used to calculate the final buildup curve as a continuous function of TEDOR mixing time. The final best fits show A values ranging from 0.34 to 13.8, and $T_{rlx}$ values ranging from 0.67 to 4.8 ms.

Results and Discussion

Fig. 1 shows six variants of ¹³C-¹⁹F REDOR-based pulse sequences and the ¹⁹F-¹³C CP pulse sequence. These approaches differ mainly in the relaxation mechanisms during the mixing periods. During REDOR, T₂ relaxation of in-phase and anti-phase ¹³C and ¹⁹F coherences occur, while during CP, ¹⁹F and ¹³C T_1ρ relaxation are the main relaxation mechanisms. In addition, for REDOR-based pulse sequences under fast MAS, the 180˚ pulses occupy increasing fractions of each rotor period, causing ¹³C and ¹⁹F T_1ρ relaxation to contribute to intensity losses (Jaroniec et al. 2000). Within the suite of REDOR-based pulse sequences, the experiments differ in whether coherences are transferred from ¹³C to ¹⁹F or vice versa, and whether the coherence transfer pathway is direct (Hing et al. 1992) (Fig. 1A, B) or follows an out-and-back pathway (Fig. 1C–F). For clarity, we refer to the direct transfer experiments as direct TEDOR (Fig. 1A, B) while the out-and-back experiments as OaB REDOR (Fig. 1C–F). These experiments can be detected on either ¹⁹F or ¹³C spins, and the 180˚ pulses may be mainly placed on the ¹³C channel or the ¹⁹F channel. To investigate which pulse sequence gives the most efficient coherence transfer and the highest-sensitivity spectra, we measured 1D spectra as a function of TEDOR mixing time. We use ¹³C natural abundance t-Boc-4F-Phe and ¹³C, ¹⁵N-labeled GB1 containing three 3F-Tyr residues as the model compounds to compare these experiments.

Fig. 1.

Pulse sequences for 2D ¹⁹F-¹³C correlation experiments. (A) ¹³C-detected ¹⁹F-¹³C TEDOR. (B) ¹⁹F-detected ¹³C-¹⁹F TEDOR. (C) Out-and-back ¹³C-¹⁹F correlation with ¹³C detection and most pulses on ¹⁹F. (D) Out-and-back ¹³C-¹⁹F correlation with ¹³C detection and most pulses on ¹³C. (E) Out-and-back ¹³C-¹⁹F correlation with ¹⁹F detection and most pulses on ¹³C. (F) Out-and-back ¹³C-¹⁹F correlation with ¹⁹F detection and most pulses on ¹⁹F. (G) ¹⁹F-¹³C CP, with the initial ¹⁹F magnetization established by ¹H-¹⁹F CP.

Direct TEDOR outperforms Out-and-Back REDOR and CP in coherence transfer efficiency

Fig. 2 shows 1D ¹³C and ¹⁹F spectra of t-Boc-4F-Phe obtained from TEDOR and CP experiments. The TEDOR spectra were measured using mixing times of 0.33 ms to 7 ms. Both ¹⁹F-to-¹³C and ¹³C-to-¹⁹F direct TEDOR spectra (Fig. 2A, B) show much higher intensities than the four OaB spectra (Fig. 2C–F). These OaB spectra give high intensities only at short mixing times but exhibit rapid T₂ relaxation at longer mixing times. Between ¹³C and ¹⁹F detection, the ¹³C-detected spectra have lower intensities, regardless of whether more pulses are placed on ¹⁹F or ¹³C channels. The ¹⁹F-detected OaB spectra have higher sensitivity than the ¹³C-detected spectra, but the ¹⁹F intensities still decrease rapidly with the mixing time.

Fig. 2.

Comparison of the mixing-time dependent ¹³C-¹⁹F coherence transfer in t-boc-4F-Phe. The spectra are shown in the same order as the pulse sequences in Fig. 1. The TEDOR mixing times range from 0.33 ms to 7 ms. Spectra in (A-F) were measured under 12 kHz MAS on the 400 MHz spectrometer, while spectra in (G) were measured under 25 kHz or 35 kHz MAS on the 600 MHz spectrometer. The red spectrum in each mixing-time series guides the eye for comparison between different pulse sequences. (A) ¹⁹F-to-¹³C TEDOR. Relatively high intensities are observed. (B) ¹³C-to-¹⁹F TEDOR, also showing high coherence transfer intensities. (C) Out-and-back transfer with ¹³C detection and most pulses on ¹⁹F, showing very low transferred intensities. (D) Out-and-back transfer with ¹³C detection and most pulses on ¹³C. The transferred intensities are still low. (E) Out-and-back transfer with ¹⁹F detection and most pulses on ¹³C. Transfer efficiencies are high at short mixing times but low at long mixing times. (F) Out-and-back transfer with ¹⁹F detection and most pulses on ¹⁹F. Transfer efficiencies are low at short mixing times and high at long mixing times. (G) ¹⁹F-¹³C CP spectra compared to direct ¹⁹F- ¹³C TEDOR spectra. DQ CP is much more efficient than ZQ CP, but the DQ CP spectra still have lower intensities than the TEDOR spectra. The ¹H/¹⁹F/¹³C rf fields during the ¹⁹F-¹³C CP periods are indicated.

The ¹⁹F-detected OaB spectra with most recoupling pulses on ¹⁹F (Fig. 2F) show high intensities in the initial regime, loss of intensity at intermediate mixing times, and reappearance of intensities at longer mixing times. We attribute this observation to pulse imperfections on the ¹⁹F channel. A similar trend is seen in the ¹³C-detected OaB spectra with most recoupling pulses on the ¹³C channel. Thus, the application of many pulses on the observed channel is detrimental to the sensitivity and coherence transfer efficiency.

Compared to the OaB spectra, the direct TEDOR spectra, measured with an equal number of 180˚ pulses on the ¹³C and ¹⁹F channels, show high intensities in the full mixing time range. The signals of carbons that are one and two bonds away from ¹⁹F show fast relaxation, suggesting the ¹⁹F chemical shift anisotropy (CSA)-driven relaxation as the primary relaxation mechanism, whereas the signals of ¹³C sites further from the ¹⁹F spin show slower relaxation and maximal intensities at longer mixing times.

We measured ¹⁹F-¹³C CP spectra (Fig. 2G) under 25 and 35 kHz MAS, where high-power ¹H decoupling is necessary during the ¹⁹F-¹³C CP period. We found that the polarization transfer efficiency of ZQ CP improved with stronger ¹⁹F and ¹³C spin-lock field strengths. Using moderate spin-lock field strengths of 46 kHz for ¹⁹F and 81 kHz for ¹³C, ZQ CP only gave one-bond transfer at 35 kHz MAS. In comparison, DQ CP gave much higher polarization transfer over multiple bonds. However, the best DQ-CP spectra still give lower intensities than the best direct TEDOR spectra for all ¹³C signals except for the one-bond ¹³C peak. Taken together, these comparisons indicate that TEDOR is more efficient for long-range coherence transfer and should be the method of choice if long-range ¹⁹F-¹³C correlation signals are desired in 2D spectra.

Recently, ¹⁹F-¹³C ZQ CP was used under DNP conditions to measure correlations over ~8 Å in the HIV capsid protein (Lu et al. 2019b). In that experiment, strong spin-lock field strengths of 108 kHz and 84 kHz were applied to ¹⁹F and ¹³C channels, respectively, which may have improved the performance of ZQ CP. The use of a strong ¹³C spin-lock also eliminated the need for ¹H decoupling. We find that TEDOR experiments at intermediate MAS frequencies of 25–35 kHz still require ¹H decoupling, thus requiring a triple-resonance probe that simultaneously tunes ¹H, ¹⁹F and ¹³C. At moderate MAS frequencies with separate ¹H and ¹⁹F channels, it is important to test both ¹⁹F-¹³C CP and TEDOR experiments, since the relative sensitivity crucially depends on the T₂ (TEDOR) and T_1ρ (CP) relaxation times of ¹⁹F and ¹³C spins in the protein of interest. Fluorinated and ¹³C, ¹⁵N-labeled proteins may have longer ¹⁹F T_1ρ relaxation times than ¹⁹F T₂, in which case DQ CP under fast MAS may give higher sensitivity 2D correlation spectra.

TEDOR intensities can be quantified for extracting distances

In addition to providing qualitative proximity information based on cross peak intensities in 2D spectra, the ¹⁹F-¹³C TEDOR experiment is expected to give quantitative ¹⁹F-¹³C distances. Fig. 3 shows representative TEDOR buildup curves for t-Boc-4F-Phe. The ¹³C chemical shifts are assigned based on the published shifts in AIST SDBS (https://sdbs.db.aist.go.jp/sdbs/cgi-bin/cre_index.cgi) (Fig. 3B). The ¹³C intensities increase with mixing time till an intermediate mixing time, then decay due to T₂ relaxation (Jaroniec et al. 1999). The carbons that are the closest to the ¹⁹F spin such as C2 manifest the fastest intensity buildup, while carbons that are further away from ¹⁹F such as C8 exhibit slower buildup (Fig. 3C). For comparison, we also measured the REDOR difference spectra, and found similar buildup behavior, except that most signals do not exhibit relaxation during the mixing times used. We attribute this difference to the fact that signals disappear due to ¹³C T₂ relaxation in the REDOR experiment whereas ¹⁹F T₂ relaxation, which is usually faster than ¹³C, may dominate the TEDOR experiment.

Fig. 3.

¹⁹F-¹³C TEDOR buildup curves for t-Boc-4-¹⁹F-Phe. (A) Chemical structure of t-Boc-4-¹⁹F-Phe with carbons numbered in the order of decreasing chemical shifts. (B) ¹⁹F-¹³C TEDOR spectra as a function of mixing time (black). A ¹³C CP spectrum (green) is shown for comparison. REDOR difference spectra for the same mixing times are shown on the right for comparison. (C) TEDOR buildup curves for three representative carbons. The best-fit distances are in good agreement with the expected distances. (D) χ² values between the experiment and simulation as a function of the ¹⁹F-¹³C distance r and the effective relaxation time, T_rlx, for C4 and C8. An intensity scaling factor of 0.6 and 4.8 was used for these maps. The minimum χ² values (cross) give the best-fit ¹⁹F-¹³C distances of 3.9 Å and 6.9 Å shown in (C).

We fit the TEDOR buildup curves in Mathematica using three parameters: the ¹³C-¹⁹F distance r of interest, an intensity scaling factor A, and an effective relaxation time constant, T_rlx (Fig. 3C). The fitted distances agree well with the expected intramolecular distances within experimental uncertainty. The fitted effective relaxation times range between 0.67 and 4.8 ms, with an average of 1.8 ms for t-Boc-4F-Phe. Fig. 3D shows representative 2D χ² maps as a function of r and T_rlx for the deviation between the experimental data and simulations for C4 and C8. For C4, the χ² map shows a deep minimum at the best-fit distance of 3.9 Å with T_rlx = 2.6 ms, and the χ² values are much more sensitive to r than to T_rlx, indicating low uncertainty in the best-fit distance. For C8, the χ² map shows a rapid decent to a shallow minimum at r = 6.9 Å and T_rlx = 1.0 ms, indicating a well constrained lower-bound distance, but the upper-bound distance has larger uncertainty. The χ² contour lines are no longer completely parallel to the T_rlx axis, indicating that the relaxation time has a moderate effect on the best-fit distance.

2D ¹⁹F–¹³C correlation spectra allow ¹⁹F chemical shift assignment

With the higher sensitivity of ¹⁹F-¹³C direct TEDOR coherence transfer, we extended the 1D experiments to 2D to assign the ¹⁹F chemical shifts. Fig. 4 shows ¹³C-detected 2D ¹⁹F-¹³C correlation spectra of t-Boc-4F-Phe. At a short mixing time of 0.46 ms, only one- and two-bond correlation signals are observed. Applying ¹⁹F decoupling during ¹³C detection collapsed ¹J_CF doublet into a singlet (Fig. 4C), further increasing the sensitivity. We did not observe significant lineshape differences between the two peaks of the ¹J_CF doublet, indicating no significant TROSY effect in the solid state (Boeszoermenyi et al. 2019). At a longer mixing time of 3.89 ms, multi-bond correlations appear, giving most of the ¹³C signals in the CP spectra.

Fig. 4.

2D ¹⁹F-¹³C correlation spectra of t-Boc-4-¹⁹F-Phe, measured under 35 kHz MAS. (A) Spectrum with 0.46 ms TEDOR mixing. (B) Spectrum with 3.89 ms TEDOR mixing. 1D ¹⁹F and ¹³C CP spectra are shown for comparison. (C) Zoomed-in views of selected regions of the 2D spectra, showing the ¹⁹F-¹³C one-bond J-splitting of 250 Hz, which is removed by ¹⁹F decoupling (blue).

It is of interest to compare the sensitivity of the ¹³C-detected and ¹⁹F-detected 2D correlation spectra. ¹⁹F detection has the advantage of higher sensitivity due to the higher gyromagnetic ratio of ¹⁹F, but the indirect ¹³C dimension requires a large spectral window and thus dense time-domain sampling. In comparison, ¹³C detection has lower detection sensitivity but benefits from a small ¹⁹F spectral window if similar functional groups are present in the protein, thus would allow sparse sampling of the indirect dimension. In addition, the coherence transfer efficiency from ¹⁹F to ¹³C may not be identical to the reverse ¹³C to ¹⁹F transfer efficiency. Therefore, we compared ¹⁹F-detected (red) and ¹³C-detected (black) 2D TEDOR correlation spectra of t-Boc-4F-Phe (Fig. 5), measured using identical total experimental times, line broadening parameters, and effective ¹³C and ¹⁹F acquisition times. Both spectra were acquired with a mixing time of 3.9 ms using the same number of rotor periods. We found the same correlation peaks in the two spectra, except that the ¹⁹F-detected spectrum shows an additional peak at (−119, 36) ppm, which corresponds to the ¹⁹F-Cβ cross peak. In the ¹³C cross sections, we found that the ¹⁹F-detected 2D spectrum (red) has higher signal-to-noise ratios (SNR) than the ¹³C-detected spectrum by 1.2 to 5.7 fold, with an average sensitivity enhancement of 2.5-fold. The only exception is the 81-ppm peak of the quaternary C7, whose ¹³C-detected intensity is similar to the ¹⁹F-detected intensity.

Fig. 5.

Sensitivity of ¹⁹F-detected and ¹³C-detected 2D TEDOR correlation spectra. (A) ¹³C-detected ¹⁹F-¹³C 2D spectrum. (B) ¹⁹F-detected ¹³C-¹⁹F 2D spectrum. Both spectra were measured using a 3.89 ms mixing time under 35 kHz MAS. (C) ¹³C cross sections at ¹⁹F chemical shifts of −117.6 ppm and −119.5 ppm from the two 2D spectra. The ¹⁹F-detected 2D spectrum shows 2-fold higher sensitivity than the ¹³C-detected spectrum. High-intensity ¹³C peaks are observed for C10, C8, C7, and C1, some of which are due to intermolecular magnetization transfer.

The non-uniform sensitivity enhancement between ¹³C-detected and ¹⁹F-detected spectra may result from the initial magnetization difference. For this ¹³C natural abundance compound, the initial ¹³C magnetization from ¹H-¹³C CP depends on the local ¹H density. Aliphatic carbons in the ¹H-rich environment have high ¹³C initial magnetization, and thus should give rise to high-sensitivity transfer to ¹⁹F. In comparison, the ¹⁹F spin lies on a ¹H-poor aromatic ring, far from the aliphatic methyl groups, thus it is expected to give low polarization transfer to the remote carbons. In uniformly ¹³C-labeled proteins, the relayed ¹³C-¹³C dipolar coherence transfer is likely to increase the coherence transfer efficiency.

To investigate how ¹³C-¹⁹F TEDOR behaves in uniformly ¹³C-labeled proteins, we applied the experiment to 3F-Tyr, ¹³C, ¹⁵N-labeled GB1. Among the three Tyr residues in GB1, Y3, Y33, and Y45, only two are well resolved in the 1D ¹⁹F spectra, at −136 ppm and around −133 ppm. The 2D ¹³C-¹⁹F TEDOR correlation spectrum (Fig. 6A) shows that the −132.8 ppm ¹⁹F peak correlates with M1 ¹³CO, Cα and Q2 Cβ peaks, thus this ¹⁹F chemical shift can be assigned to Y3. The −133.2 ppm ¹⁹F peak correlates with T51 Cα and K50 Cβ, thus it can be assigned to Y45. This leaves the −136 ppm peak to Y33. Interestingly, fewer cross peaks are observed for Y33 at −136 ppm, consistent with the fact that Y33 is more dynamic than the other two Tyr residues, as shown by its smaller CSA (Roos et al. 2018a). The 2D spectrum shows that some of the ¹³C peaks correlate with multiple ¹⁹F spins. For example, the A23 Cβ peak correlates with both Y3 and Y45, indicating that this residue lies at similar distances to the two Tyr residues.

Fig. 6.

2D ¹⁹F-¹³C TEDOR correlation spectrum of ¹³C, ¹⁵N-GB1 containing 3F-Tyr. (A) 2D spectrum measured with a 1.37 ms mixing time under 35 kHz MAS. (B) TEDOR buildup curves for representative resolved signals. (C) High-resolution structure of GB1 (PDB: 2LGI), showing the three 3F-Tyr residues (cyan) and ¹³C-labeled residues that exhibit significant dipolar dephasing (green). The two possible fluorinated positions (3- and 5-) of Tyr are colored in orange and pink.

1D ¹³C-¹⁹F TEDOR buildup curves of GB1 (Fig. 6B) show faster T₂ relaxation than the natural abundance t-Boc-4F-Phe, suggesting the influence of ¹J_CC couplings and protein dynamics (Lu et al. 2019a). Thus, for uniformly ¹³C-labeled proteins, the 2D ¹⁹F-¹³C TEDOR experiment is best conducted at short mixing times of less than ~3 ms to ensure sufficient sensitivity. This restricts the distance range of interest to 4–6 Å. Fig. 6C highlights the carbons observed in the 2D ¹⁹F-¹³C correlation spectra, showing residues in the vicinity of the three Tyr sidechains.

2D ¹³C–¹⁹F correlation spectra of a membrane protein reveal clustering

We applied the 2D ¹⁹F-¹³C TEDOR correlation experiment to membrane-bound BM2-TM to test the sensitivity of this technique on a membrane protein and the utility of the resonance assignment and distance information. The peptide contains two fluorinated native residues, 4F-F5 and 5F-W23, which can couple to ¹³C-labeled S12, I14, H19, and A22 (Kwon et al. 2019). Two ¹⁹F chemical shifts at −126 ppm and −116 ppm are resolved. With a short mixing time of 1.49 ms, cross peaks with A22 Cβ and Cα readily allowed the −126 ppm ¹⁹F peak to be assigned to W23 (Fig. 7A). Surprisingly, W23 also exhibits weak cross peaks with I14 Cδ2 and Cγ1. These correlations cannot be explained by the parallel four-helix bundle structure of M2 (Hong and DeGrado 2012; Kwon et al. 2019; Williams et al. 2017; Williams et al. 2016), since both the intrahelical W23-I14 distances and interhelical distances within a tetramer are longer than 15 Å, whose dipolar couplings are too weak to be observed (Mandala et al. 2019; Mandala et al. 2020). Thus, these W23-I14 cross peaks must be attributed to multiple BM2 tetramers packed closely together in the membrane, but in opposite orientations, so that W23 in one tetramer can have a measurably short distance to I14 in another tetramer (Fig. 7B).

Fig. 7.

2D ¹³C-¹⁹F TEDOR correlation spectra of BM2-TM peptide in VM+ membranes. The peptide is fluorinated at 4F-Phe5 and 5F-Trp23 and contains ¹³C, ¹⁵N-labeled S12, I14, H19, A22. (A) 2D ¹³C-¹⁹F correlation spectra with mixing times of 1.49 ms and 2.97 ms. 1D ¹⁹F and ¹³C spectra are overlaid for reference. (B) Structural model of BM2 TM. Adjacent tetramers are positioned in opposite orientations to account for the observed I14-W23 ¹³C-¹⁹F cross peaks. The inter-tetramer distance of ~8 Å (orange lines) is not experimentally measured but is based on model building.

Clustering of influenza A virus M2 tetramers in lipid bilayers has been recently reported (Andreas et al. 2015; Paulino et al. 2019). At a high protein/lipid molar ratio of 1 : 7.5 in DOPC/DOPE bilayers, ²H NMR spectra of a lipid-facing Ala residue indicate immobilization of the methyl sidechain, suggesting tetramer clustering. Coarse-grained simulations support this notion, and show that both parallel and antiparallel arrangements of different tetramers are possible. The present observation of W23-I14 contacts in BM2 gives unambiguous evidence that BM2 tetramers also cluster in lipid bilayers (Mandala et al. 2019). At our protein : lipid molar ratio of 1 : 14.5, the mass ratio is 1 : 3.3, thus clustering is not surprising. The antiparallel arrangement of two neighboring tetramers is energetically favorable because the asymmetric “wedge” shape of each tetramer may be compensated by a neighboring tetramer inserted in the opposite orientation into the membrane (Elkins et al. 2018; Elkins et al. 2017). Thus, the 2D ¹³C-¹⁹F correlation spectrum provides experimental evidence that BM2 tetramers cluster in lipid membranes.

Conclusions

These data show that 2D ¹⁹F-¹³C correlation spectra can be measured with high sensitivity using TEDOR and DQ CP. At short mixing times, the correlation peaks allow assignment of the ¹⁹F chemical shifts without the need for single-site mutants. At long mixing times, the correlation peaks provide structurally informative long-range distance constraints between fluorinated and ¹³C-labeled residues. We find that the TEDOR coherence transfer efficiency is higher with ¹⁹F detection than with ¹³C detection. For the BM2 TM peptide, the 2D ¹³C-¹⁹F TEDOR spectrum revealed cross peaks indicative of tetramer clustering in the membrane. This 2D ¹⁹F-¹³C correlation approach complements the recently reported ¹³C-¹⁹F and ¹H-¹⁹F REDOR technique (Shcherbakov and Hong 2018; Shcherbakov et al. 2019) for nanometer distance measurements.