Solid-State NMR ¹⁹F-¹H-¹⁵N Correlation Experiments for Resonance Assignment and Distance Measurements of Multi-fluorinated Proteins

Abstract

Several solid-state NMR techniques have been introduced recently to measure nanometer distances involving ¹⁹F, whose high gyromagnetic ratio makes it a potent nuclear spin for structural investigation. These solid-state NMR techniques either use ¹⁹F correlation with ¹H or ¹³C to obtain qualitative inter-atomic contacts or use the rotational-echo double-resonance pulse sequence to measure quantitative distances. However, no NMR technique is available for disambiguating ¹H-¹⁹F distances in multiply fluorinated proteins and protein-ligand complexes. Here, we introduce a new 3D ¹⁹F-¹⁵N-¹H correlation experiment that resolves the distances of multiple fluorines to their adjacent amide protons. We show that optimal polarization transfer between ¹H and ¹⁹F spins is achieved using an out-and-back ¹H-¹⁹F REDOR sequence. We demonstrate this 3D correlation experiment on the model protein GB1, and apply it to the multidrug-resistance transporter, EmrE, complexed to a tetra-fluorinated substrate. This technique will be useful for quickly deriving distance constraints around aromatic sidechains in multiply fluorinated proteins, leading to significant savings of time and of precious samples compared to using one fluorine at a time. Moreover, the method enables structural determination of protein-ligand complexes that contain multiple fluorines.

Graphical Abstract

Introduction

Fluorinated small molecules and fluorinated proteins are ubiquitous in the pharmaceutical industry, medical imaging and structural biological research. In 2020, ~25% of all drugs approved by the FDA contain fluorine atoms¹, and this number is estimated to increase to approximately 30%². The incorporation of fluorine in pharmaceutical compounds can improve the metabolism and bioavailability of the drug³. Fluorine is also widely incorporated in positron emission tomography (PET) tracers to diagnose cancer, cardiovascular diseases, neurodegenerative disorders and other diseases^4–6. While fluorinated small molecules are excellent probes for studying ligand binding to macromolecules, fluorinated proteins provide opportunities for studying protein structure and dynamics⁷. Although fluorine does not occur naturally in biological macromolecules, it can be readily introduced into proteins biosynthetically or synthetically^8–10. When sparsely incorporated, fluorine usually causes minimal perturbation to protein structure and function, as can be assessed by ¹³C, ¹⁵N and ¹H NMR and other biophysical techniques^11–12.

The stable isotope of fluorine, ¹⁹F, has many attractive properties for NMR spectroscopy. ¹⁹F is a 100% abundant spin-1/2 nucleus with a large gyromagnetic ratio (γ), which is 94% of the γ of ¹H. Thus, ¹⁹F NMR has intrinsically high detection sensitivity. ¹⁹F chemical shift is extremely sensitive to its electronic environment, hence it reports on subtle changes in the chemical and conformational structure of the molecule^13–16. The large ¹⁹F γ increases the strength of dipole-dipole interaction between ¹⁹F and other nuclear spins, thus ¹⁹F allows inter-atomic distances to be measured to longer ranges than possible using low-γ nuclei such as ¹³C and ¹⁵N. As a result, ¹⁹F has been exploited for distance measurements in magic-angle-spinning (MAS) solid-state NMR spectroscopy for many years^17–18. However, until recently, the most common approach for this purpose has been one-dimensional (1D) rotational-echo double-resonance (REDOR), which measures one distance at a time, giving low throughput¹⁹. To accelerate ¹⁹F-based distance measurements, multidimensional solid-state NMR techniques that achieve ¹³C-¹⁹F, ¹H-¹⁹F and ¹⁹F-¹⁹F correlation and distance measurement have been introduced in the last few years⁷. These techniques operate at relatively high magnetic fields of 11.7 Tesla or above and under relatively fast MAS frequencies of 25 to 110 kHz. Thus they have higher sensitivity and resolution than traditional low-field slow-MAS ¹⁹F NMR experiments. Two main approaches, heteronuclear correlation (HETCOR) and REDOR, have been explored under these high-field fast-MAS conditions. The HETCOR experiments typically involve cross polarization (CP) between ¹⁹F and other nuclei to assign the resonances and extract qualitative distance information^20–24. Similarly, homonuclear ¹⁹F-¹⁹F correlation experiments using either spin diffusion^25–26 or dipolar recoupling for polarization transfer have been used, with semi-quantitative distances extracted from cross-peak intensities^27–29. These 2D heteronuclear and homonuclear correlation experiments have been demonstrated on fluorinated small molecules, pharmaceutical compounds^30–32, model proteins, and the HIV-1 capsid protein^{23, 26}.

Compared to correlation experiments that use cross peak intensities to derive semi-quantitative distance information, REDOR relies on time-dependent dipolar dephasing to provide quantitative distance constraints between ¹⁹F and a heteronuclear spin. To increase the throughput of REDOR NMR, we recently demonstrated a 2D ¹³C-¹³C spectrally resolved ¹³C-¹⁹F REDOR experiment²¹. Similarly, ¹H-¹⁹F REDOR can be conducted in a 2D ¹H-¹⁵N spectrally resolved fashion with ¹H detection, thus giving high sensitivity as well as excellent site-specific resolution³³. These 2D resolved REDOR distance experiments have been applied to several membrane proteins and amyloid proteins for structural studies. These include the transmembrane (TM) domains of the SARS-CoV-2 envelope (E) protein³⁴, the influenza BM2 protein^35–36, the multidrug-resistance transporter EmrE^37–38, and the Alzheimer’s Aβ40 fibril bound to a PET tracer³⁹.

For ¹⁹F REDOR experiments on singly fluorinated proteins or small molecules, no explicit ¹⁹F chemical shift encoding is necessary. However, for multiply fluorinated systems, ¹⁹F chemical shift encoding and correlation with other nuclei become important for assigning distance constraints to specific fluorine atoms. The ¹⁹F-¹³C and ¹⁹F-¹H HETCOR experiments are often not sufficient to resolve the signals of the residues that are close to each fluorine. Therefore, to better resolve the signals of fluorine-proximal residues, two chemical shift dimensions in addition to ¹⁹F are desirable. Since 2D ¹³C-¹³C correlation experiments have low sensitivity while ¹H-¹H 2D correlation experiments have low chemical shift dispersion, correlating two different nuclei with ¹⁹F is expected to give the highest information content.

In this study, we introduce 2D and 3D ¹⁹F, ¹H and ¹⁵N correlation experiments to resolve the signals of fluorine-proximal protein residues. We choose ¹⁵N as the third nucleus because ¹H-¹⁵N correlation is now the standard fingerprint in ¹H-detected solid-state NMR experiments. We compare several polarization transfer methods to achieve triple-resonance ¹H-¹⁹F-¹⁵N correlation. We show that an out-and-back (OaB) REDOR-CP experiment and a Lee-Goldberg cross-polarization (LG-CP) experiment both have adequate sensitivity. We demonstrate these techniques on the model protein GB1, and show that the ¹H^N-¹⁹F correlation spectra can be disambiguated without the use of multiple CDN-labeled protein samples. We also apply the OaB REDOR-CP experiment to the multidrug-resistance transporter, EmrE, bound to a multi-fluorinated substrate. The 3D correlation experiment resolves, for the first time, the specific protein sidechains that are in close contact to each fluorine of this tetra-fluorinated ligand.

Methods

Preparation of deuterated and fluorinated microcrystalline GB1

Uniformly ¹³C, ²H, ¹⁵N (CDN)-labeled GB1 containing one or two fluorinated residues was expressed, purified and crystallized using a modified protocol from the literature^{21, 40}. One sample contains a single 5-¹⁹F-Trp43 label (W-GB1), and the second sample contains 4-¹⁹F-labeled Phe30 and Phe52 (FF-GB1).

All isotopically labeled reagents were obtained from Cambridge Isotope Laboratories. CDN-labeled and fluorinated GB1 was expressed in M9 minimal media by stepwise training of the bacteria from protonated culture to deuterated culture, and by using glyphosate to introduce the fluorinated amino acids. All growth and expression media contain 100 mg/mL ampicillin.

To express CDN-labeled FF-GB1, a 15 mL LB starter culture in H₂O was inoculated with ampicillin-resistant E. coli stored in a glycerol stock, and the cells were grown at 37°C for ~14 hours, reaching an OD₆₀₀ of ~4. About 0.5 mL of this starter culture was added to 12.5 mL filter-sterilized LB media in D₂O and allowed to grow to an OD₆₀₀ of ~1.3 in 3 hours. This 12.5 mL LB/D₂O culture was then added to 50 mL of filter-sterilized M9 media, which contains 2 g/L d₇-¹³C-glucose and 1 g/L ¹⁵NH₄Cl in 99% D₂O. The cells were allowed to grow in the M9 media at 37°C to reach an OD₆₀₀ of ~1.0. The culture was then added to 150 mL of M9/D₂O media to a volume of 220 mL and grew for another hour. At this point, 25 mg each of L-tyrosine, L-tryptophan and 4-¹⁹F-phenylanaline were dissolved in 5 mL D₂O as the aromatic amino acid solution. When the M9/D₂O media reached an OD₆₀₀ of ~0.7, glyphosate was added to a final concentration of 1 g/L to suspend the aromatic amino acid synthesis in the cells. After 5 minutes of incubation at room temperature, the aromatic amino acid solution was added to the culture. The cells were grown for another 2 hours to an OD₆₀₀ of ~1.0, then another 1 g/L of d₇-¹³C-glucose was added together with 30 mg IPTG to start protein expression. The total concentration of the ¹³C-labeled glucose was 3 g / L. GB1 expression proceeded for 4 hours at 37°C, then the cells were harvested by centrifugation at 5,000 g. The cell pellet was resuspended in 40 mL lysis buffer containing 50 mM potassium phosphate and 200 mL sodium chloride at pH 7. The cells were lysed using sonication on ice for 10 minutes. The lysate was centrifuged at 16,000 g for one hour, and the protein in the supernatant was purified by size-exclusion chromatography²¹. CDN-labeled W-GB1 was expressed and purified similarly, except that L-phenylalanine, L-tyrosine, and 5-¹⁹F-tryptophan were used in the aromatic amino acid solution.

To assess the purity and ¹⁹F incorporation levels of the protein, MALDI-TOF mass spectra were measured for CDN W-GB1, CDN FF-GB1, natural abundance GB1, natural abundance W-GB1, and natural abundance lysozyme C (Fig. S1). The molecular weight difference between the unlabeled GB1 and unlabeled 5-¹⁹F-Trp43 GB1, and between W-GB1 and FF-GB1, suggests that both fluorinated proteins have >90% ¹⁹F incorporation.

To produce microcrystals, W-GB1 and FF-GB1 were dialyzed against 50 mM potassium phosphate at pH 5.5 in 100% H₂O for 32 hours (outer solution changed every 8 hours) and concentrated to 30 mg/mL, as estimated by A₂₈₀. The protein was mixed with isopropanol and 2-methyl-2,4-pentanediol at a 1 : 1 : 2 volume ratio and incubated at 4°C overnight⁴¹. About 8 mg (dry mass) of CDN FF-GB1 and 4 mg of CDN W-GB1 were crystallized. The hydrated microcrystals were packed into 1.9 mm Bruker MAS rotors by centrifugation (3,000 g) using a Beckman Coulter Allegra X-15R with a swinging bucket rotor. In addition, microcrystals containing ~2 mg of CDN FF-GB1 were centrifuged (311,000 g) into a 1.3 mm Bruker rotor using a Beckman Optima XL-80 with a SW60 Ti rotor.

CDN-labeled S64V-EmrE was expressed and purified as described previously³⁸. The protein was bound to d₅₄-DMPC bilayers at pH 8.0. The sample was incubated with an excess amount of the fluorinated substrate 4-¹⁹F-tetraphenylphosphonium (F₄-TPP⁺) at room temperature with end-to-end rocking for >16 hours. Excess F₄-TPP⁺ was removed using microcentrifugation (7,500 g, 5 min).

Solid-state NMR experiments

All solid-state NMR experiments were conducted on a 14.1 Tesla Bruker Avance III HD NMR spectrometer operating at ¹H, ¹⁹F and ¹⁵N Larmor frequencies of 600.10 MHz, 564.66 MHz, and 60.81 MHz. The ¹⁹F-containing pulse sequences were implemented on a Bruker 1.9 mm HFX MAS probe. The samples were spun at 38 kHz at a thermocouple-recorded temperature of 273 K. At this spinning rate, frictional heating increases the sample temperature by ~25 K, giving a sample temperature of ~298 K. The temperature differential was estimated by measuring the water ¹H chemical shift of hydrated protein samples spinning under similar conditions using the equation T_eff (K) = 96.9 × (7.83−δ_H2O)⁴². ¹H chemical shift was externally referenced to sodium trimethylsilyl-propanesulfonate (DSS) at 0 ppm. ¹⁵N chemical shift was externally referenced to the Phe amide signal of formyl-MLF at 110.09 ppm on the liquid ammonia scale⁴³. ¹⁹F chemical shift was referenced to the ¹⁹F peak of crystalline 5-¹⁹F-trptophan at −122.10 ppm on the CF₃Cl scale. ¹H and ¹⁵N chemical shifts of FF-GB1 were assigned using the 3D hCANH experiment under 55 kHz MAS on a 1.3 mm HXY probe. The thermocouple temperature was 253 K and the actual sample temperature was ~290 K. The ¹³C chemical shift was referenced to the 14.0 ppm methyl ¹³C peak of Met in formyl-MLF on the tetramethylsilane (TMS) scale.

Typical radiofrequency (RF) field strengths for excitation and refocusing were 83.3 kHz on ¹H, 50 kHz on ¹⁵N, 62.5 kHz on ¹³C and 71.4 kHz on ¹⁹F. WALTZ-16 decoupling at an RF field strength of 10 kHz was applied on the ¹H, ¹⁵N and ¹³C channels for all experiments shown in Fig. 1. Solvent suppression in the ¹H-detected hNH, FNH and hCANH experiments was achieved using the MISSISSIPPI sequence at an RF field strength of 15 kHz on the ¹H channel⁴⁴. More detailed experimental conditions are listed in Table S1.

Figure 1.

Pulse diagrams for correlating ¹H, ¹⁵N, and ¹⁹F chemical shifts and for ¹H-¹⁹F distance measurements. The 3D experiments are named in the order of chemical shift encoding. (a) The out-and-back (OaB) REDOR-CP FNH experiment. (b) The LG-CP NHF experiment. The crucial spin-diffusion free LG spin lock on ¹H is colored in red. (c) The CP-TEDOR NHF experiment. This scheme does not work well due to fast relaxation of the multi-spin ¹⁹F-¹H coherence. (d) The 2D hNH-resolved ¹H-¹⁹F REDOR experiment (Adapted with permission from³³. Copyright 2019 American Chemical Society.) Phase cycles (ϕ_i) for the pulse sequences in (a) and (b) are given in the experimental section. Full Bruker pulse programs for (a) and (b) are provided in the Supporting Information.

The pulse sequences for the 3D OaB REDOR-CP FNH and the 3D LG-CP NHF are provided in the Supporting Information. Phase cycles for the 3D OaB REDOR-CP FNH experiment (Fig. 1a) are Φ₁= 13, Φ₂= 8 × (0) 8 × (2), Φ₃= 0, Φ₄= 2, Φ₅= 22 00, Φ₆= 1, Φ₇= 0, Φ₈= 0000 2222, and Φ_rec’r= 1331 3113 3113 1331. Here, 0, 1, 2, 3 correspond to +x, +y, -x, and -y, respectively. The ¹⁹F 180° pulses in the REDOR pulse train were phase-incremented using XY-16⁴⁵. It is worth noting that the two-step phase cycling (Φ₂) of the 90° ¹H storage pulse is necessary to prevent the ¹H diagonal artifacts due to imperfect ¹H inversion pulses. Phase cycles for the 3D LG-CP NHF experiment (Fig. 1b) are Φ₁= 13, Φ₂=Φ₃=1, Φ₄= 11 33, Φ₅= 0, Φ₆= 1111 3333, Φ₇= 1, and Φ_rec’r= 1331 3113. No phase cycling is implemented on the ¹H 180° pulses in the REDOR period.

All NMR spectra were processed in the Bruker Topspin software, using versions 3.2, 3.5 and 4.1. Version 3.2 allows visualization of 1D cross sections of the 2D planes of 3D correlation spectra and allows direct overlay of 2D planes of 3D spectra with measured 2D spectra. Thus, we have found it to be superior to the newer Topspin versions for spectral analysis.

Results

Design of ¹H-¹⁵N-¹⁹F correlation NMR experiments

To design a high-sensitivity 3D experiment that correlates ¹H, ¹⁹F and ¹⁵N chemical shifts, we consider two main factors: the detection nucleus and the polarization transfer method between ¹H and ¹⁹F spins. Either ¹H or ¹⁹F can serve as a high-sensitivity detection spin, giving two possibilities for pulse sequence design. Polarization transfer between ¹H and ¹⁵N can be readily achieved by CP. Thus, the only remaining design variable is the ¹H-¹⁹F polarization transfer method, which can be either a REDOR-based pulse sequence or CP. Within the various implementations of REDOR, we can use OaB REDOR⁴³, in which antiphase magnetization is created, rotated to antiphase coherence of the second spin to encode chemical shift evolution, and then rotated back and refocused on the initial spin. Alternatively, one can implement TEDOR⁴⁶, in which the antiphase coherence is rotated to and then refocused on the second spin, achieving complete magnetization transfer. Due to the sparseness of the ¹⁹F dimension and the ability to reduce the number of CP steps, the OaB REDOR is best detected using proton, therefore we appended the hNH sequence after it to form a 3D FNH experiment. Both CP and TEDOR work with either ¹⁹F or ¹H detection. To avoid water suppression and to minimize polarization transfer steps, we implemented both as ¹⁹F-detected experiments. Fig. 1a–c show three of the four pulse diagrams tested based on these considerations. The experiments are named in the order of the chemical shift encoding of the three frequency dimensions, preceded by the method of coherence transfer. The OaB REDOR-CP FNH experiment (Fig. 1a) is the only pulse sequence with ¹H detection. Among the three ¹⁹F-detected experiments, the CP-TEDOR NHF experiment uses CP for ¹⁵N–¹H polarization transfer and TEDOR for ¹H–¹⁹F polarization transfer (Fig. 1c). The LG-CP NHF experiment uses LG spin lock on the ¹H channel⁴⁷ to achieve spin-diffusion free polarization transfer from ¹⁵N to ¹H, then uses regular CP for ¹H to ¹⁹F polarization transfer (Fig. 1b). The CP NHF experiment differs from the LG-CP NHF experiment only in that regular CP is used for ¹⁵N–¹H polarization transfer, thus its pulse diagram is not shown. The presence or absence of ¹H spin diffusion during the ¹⁵N-¹H polarization transfer is a crucial detail and CP NHF is not tenable for ¹⁹F-¹⁵N-¹H correlation. For comparison, we also show the previously published 2D hNH resolved ¹H-¹⁹F pulse sequence (Fig. 1d) for long-range distance measurements³³.

We first compare the OaB REDOR-CP FNH and CP-TEDOR NHF experiments. The former evolves antiphase ¹⁹F–¹H magnetization during the ¹⁹F chemical shift evolution period, then converts it to ¹H single-quantum magnetization for CP to ¹⁵N. ¹⁵N chemical shift evolution during t₂ is followed by a final reverse CP to amide protons for detection. The OaB ¹H-¹⁹F polarization transfer block is similar to the transfer element in the ZF-TEDOR experiment for ¹³C-¹⁵N correlation⁴⁸, however it is important to note that no z-filters are used here in order to avoid ¹H spin diffusion. In comparison, the CP-TEDOR NHF experiment transfers the ¹⁵N-encoded amide ¹H magnetization to ¹⁹F using the TEDOR element. We show below that these two methods of ¹H–¹⁹F coherence transfer have different spin dynamics when multiple protons are coupled to each fluorine.

We consider a three-spin system containing two protons, ¹H₁, ¹H₂, and a single ¹⁹F. We assume the ¹H₁-¹⁹F distance is much shorter than the ¹H₂-¹⁹F distance, so that the ¹H₁-¹⁹F dipolar coupling ${\omega }_{d,1}$ is much stronger than the ¹H₂-¹⁹F dipolar coupling ${\omega }_{d,2}$. This situation is expected for most samples of interest, whether the fluorine is incorporated into protein sidechains or in a small molecule. The closest protons usually occur in the fluorine-containing residue or small molecule, which is usually undeuterated, while the more remote protons can be an amide proton in a CDN-labeled protein.

In the OaB REDOR-CP experiment, the transverse magnetization $H_{1x}+H_{2x}$ of the two protons is converted to ¹H antiphase magnetization with ¹⁹F under the average REDOR Hamiltonian ${\overline{\omega }}_{d,1}2H_{1z}{F}_z+{\overline{\omega }}_{d,2}2H_{2z}{F}_z$ during the REDOR mixing time t_m:

$$H_{1x}+H_{2x}\stackrel{t_m}{\to }{H}_{1x}\text{cos}{\overline{\omega }}_{d,1}{t}_m-2H_{1y}{F}_z\text{sin}{\overline{\omega }}_{d,1}{t}_m+H_{2x}\text{cos}{\overline{\omega }}_{d,2}{t}_m-2H_{2y}{F}_z\text{sin}{\overline{\omega }}_{d,2}{t}_m$$ (1)

Here ${\overline{\omega }}_{d,1}$ and ${\overline{\omega }}_{d,2}$ are the time-averaged dipolar couplings under the REDOR pulse sequence^{19, 49}. We neglect the cosine terms as they lack ¹⁹F correlation and are filtered out by phase cycling. The pair of 90° pulses on ¹H and ¹⁹F converts the sine terms from ¹H antiphase magnetization to ¹⁹F antiphase magnetization $2H_{1z}{F}_y\text{sin}{\overline{\omega }}_{d,1}{t}_m+2H_{2z}{F}_y\text{sin}{\overline{\omega }}_{d,2}{t}_m$. The ensuing ¹⁹F chemical shift evolution modulates this ¹⁹F antiphase magnetization by a factor $e^{-i{\Omega }_F{t}_1}$, after which the second pair of 90° ¹H and ¹⁹F pulses reconverts the ¹⁹F antiphase magnetization back to ¹H antiphase magnetization:

$$\stackrel{90° \mathit{\text{pulses}}}{\to }\left(2H_{1y}{F}_z\text{sin}{\overline{\omega }}_{d,1}{t}_m+2H_{2y}{F}_z\text{sin}{\overline{\omega }}_{d,2}{t}_m\right)e^{-i{\Omega }_F{t}_1}.$$ (2)

During the second half of the REDOR mixing time, each term of the ¹H antiphase magnetization evolves under its respective dipolar coupling, ${\overline{\omega }}_{d,1}2H_{1z}{F}_z$ or ${\overline{\omega }}_{d,2}2H_{2z}{F}_z$, into observable ¹H single-quantum coherence (again neglecting cosine terms that are unusable and removed by phase-cycling):

$$\stackrel{t_m}{\to }\left(H_{1x}{\text{sin}}^2{\overline{\omega }}_{d,1}{t}_m+H_{2x}{\text{sin}}^2{\overline{\omega }}_{d,2}{t}_m\right)e^{-i{\Omega }_F{t}_1}.$$ (3)

Therefore, each proton’s magnetization is modulated by fluorine chemical shift, as desired, and is scaled by each proton’s effective transfer efficiency, ${\text{sin}}^2{\overline{\omega }}_{d,i}{t}_m$. This transfer efficiency is independent of the other proton’s interactions with the fluorine.

The TEDOR sequence begins similarly, with an initial REDOR block followed by a pair of 90° pulses on the ¹H and ¹⁹F channels. Again, these steps convert ¹H magnetization to ¹⁹F antiphase magnetization. The conversion can be written:

$$\begin{gathered} H_{1x}+H_{2x}\stackrel{t_m}{\to }-2H_{1y}{F}_z\text{sin}{\overline{\omega }}_{d,1}{t}_m-2H_{2y}{F}_z\text{sin}{\overline{\omega }}_{d,2}{t}_m \\ \stackrel{90° \mathit{\text{pulses}}}{\to }2H_{1z}{F}_y\text{sin}{\overline{\omega }}_{d,1}{t}_m+2H_{2z}{F}_y\text{sin}{\overline{\omega }}_{d,2}{t}_m \end{gathered}$$ (4)

However, unlike OaB REDOR, in the second half of the TEDOR mixing period, each of the ¹⁹F antiphase magnetization terms will be influenced by dipolar couplings to both ¹H spins, ${\overline{\omega }}_{d,1}2H_{1z}{F}_z+{\overline{\omega }}_{d,2}2H_{2z}{F}_z$. Sequential evolution by the two commuting dipolar couplings gives rise to observable ¹⁹F magnetization that is modulated by the product of sine and cosine terms of the two dipolar phases:

$$\stackrel{t_m}{\to }{F}_x{\text{sin}}^2{\overline{\omega }}_{d,1}{t}_m\text{cos}{\overline{\omega }}_{d,2}{t}_m+F_x{\text{sin}}^2{\overline{\omega }}_{d,2}{t}_m\text{cos}{\overline{\omega }}_{d,1}{t}_m$$ (5)

Importantly, the transfer efficiency of each proton spin i to the fluorine not only depends on its own coupling to the fluorine ${\text{sin}}^2{\overline{\omega }}_{d, i}{t}_m$, but also depends on a factor $\text{cos}{\overline{\omega }}_{d,j}{t}_m$ for every other proton j coupled to the fluorine. The cosine terms reduce the magnitude of the observable ¹⁹F magnetization, especially because mixing times that maximize the ${\text{sin}}^2{\overline{\omega }}_{d,i}{t}_m$ terms will generally produce low values of $\text{cos}{\overline{\omega }}_{d,j}{t}_m$. If the second mixing period t_m2 is chosen to be different from the first mixing period t_m1 before the 90° pulses, then the modulation terms become $\text{sin}{\overline{\omega }}_{d,1}{t}_{m1}\text{sin}{\overline{\omega }}_{d,1}{t}_{m2}\text{cos}{\overline{\omega }}_{d,2}{t}_{m2}$ and $\text{sin}{\overline{\omega }}_{d,2}{t}_{m1}\text{sin}{\overline{\omega }}_{d,2}{t}_{m2}\text{cos}{\overline{\omega }}_{d,1}{t}_{m2}$, but these are still smaller than the optimal efficiency of ${\text{sin}}^2{\overline{\omega }}_{d, i}{t}_m$ for the OaB REDOR experiment.

This density operator analysis indicates that for all realistic situations where multiple protons are coupled to each fluorine, the OaB REDOR-CP experiment should have higher sensitivity than the CP-TEDOR experiment. Indeed, this is confirmed experimentally by the GB1 data below (Fig. S3a). Based on similar arguments, we expect that OaB REDOR from ¹⁹F to ¹H would have even worse sensitivity, as both REDOR periods would experience multiple couplings.

Cross polarization between ¹H and ¹⁹F is used in the LG-CP NHF experiment (Fig. 1b) as well as the CP NHF experiment (pulse diagram not shown). Between these two, the LG-CP experiment is expected to higher sensitivity because the fluorinated residues or small molecules are usually undeuterated. The resulting ¹H spin diffusion, when not suppressed during CP, is expected to lead to ¹⁵N-¹H correlations not only for directly bonded amides but also between aromatic protons and the ¹⁵N, complicating spectral analysis (Fig. S3b). Lee-Goldberg CP suppresses this ¹H spin diffusion^{47, 50}, thus ensuring the detection of one-bond ¹⁵N–¹H cross peaks for those amide protons that are in close proximity to the fluorines.

¹H, ¹⁵N and ¹⁹F correlation spectra of fluorinated GB1

We assessed the ¹⁹F incorporation and ¹⁹F chemical shifts of fluorinated GB1 and substrate-bound EmrE using MALDI-MS and 1D ¹⁹F direct-polarization (DP) experiments. The mass spectra of unlabeled GB1, fluorinated GB1, and CDN-labeled and fluorinated GB1 samples show a dominant peak whose masses differ in accordance with the presence of one or two fluorines at a greater than 90% level in the protein (Fig. S1a–f). The two fluorinated GB1 samples are highly pure, as assessed by the size-exclusion chromatographs (Fig. S1g).

1D ¹⁹F DP spectra were measured under 38 kHz MAS to give the isotropic chemical shifts and under 7 kHz to give spinning sideband intensities (Fig. 2). The singly fluorinated W-GB1 exhibits a narrow ¹⁹F peak at −122.7 ppm (Fig. 2a), consistent with previous data²¹. The 2D hNH fingerprint spectrum of this W-GB1 sample shows a single set of ¹⁵N-H^N correlation peaks, indicating high structural homogeneity. The doubly fluorinated FF-GB1 sample exhibits three peaks in the quantitative ¹⁹F DP spectra: a pair of peaks at −109.1 ppm and −109.5 ppm and an isolated peak at −111.1 ppm. These three peaks have an integrated intensity ratios of 1 : 1 : 2 (Fig. 2b). The downfield pair of peaks have a ¹⁹F T₁ relaxation time of 3.7 s while the upfield peak has a distinct ¹⁹F T₁ of 10.7 s. Based on these intensities and T₁ values, we assign the two downfield peaks to one Phe residue and the upfield peak to the other Phe. Which peak corresponds to which Phe is obtained from the 3D correlation spectra shown below. At 7 kHz MAS, we observed high sideband intensities (Fig. 2e, f), which were analyzed using the Herzfeld-Berger method⁵¹ to give a ¹⁹F anisotropy parameter of 58.1 and 59.4 ppm for the two Phe residues and 46.1 ppm for the Trp43. These values are near the rigid limit of ¹⁹F chemical shift anisotropy (CSA)⁵², indicating that these aromatic sidechains are largely immobilized. The ¹⁹F DP spectrum of F₄-TPP⁺ bound to EmrE in the lipid membrane shows four peaks at isotropic chemical shifts of −96.5 ppm, −98.8 ppm, −100.6 ppm and −101.4 ppm. This distribution indicates that the four chemically equivalent fluorines of the ligand are magnetically inequivalent due to their interactions with different protein sidechains³⁸.

Figure 2.

One-dimensional ¹⁹F direct polarization (DP) spectra of fluorinated GB1 and F₄-TPP⁺ bound EmrE in lipid bilayers. Spectra in (a-c) were measured under 38 kHz MAS while spectra in (e, f) were measured under 7 kHz MAS. (a) ¹⁹F spectrum of 5-¹⁹F-Trp43 labeled GB1, measured with a recycle delay of 5.5 s. The Trp43 ¹⁹F T₁ relaxation time is 4.0 s. (b) ¹⁹F spectrum of 4-¹⁹F-Phe labeled GB1, measured with a recycle delay of 25 s. The ¹⁹F T₁ relaxation times are 3.7 s for F52 and 10.7 s for F30. (c) ¹⁹F spectrum of F₄-TPP⁺ bound to EmrE, measure with a recycle delay of 2 s. (d) Structure of GB1, showing the positions of 4-¹⁹F-Phe30, 4-¹⁹F-Phe52, and 5-¹⁹F-Trp43. (e) ¹⁹F DP spectrum FF-GB1 under 7 kHz MAS. Fitting the spinning sideband intensities yielded the ¹⁹F anisotropy parameter δ and asymmetry parameter η. (b) ¹⁹F DP spectrum of W-GB1 under 7 kHz MAS. Fitting the spinning sideband intensities yielded the ¹⁹F CSA parameters.

The two CDN-labeled GB1 samples allowed us to test the ability of the 3D FNH experiment to assign fluorines based on their correlations with amide protons. We first conducted the 2D hNH-resolved ¹H-¹⁹F REDOR experiment on the two GB1 samples. The REDOR difference (ΔS) spectrum between a control 2D spectrum (S₀) measured without ¹⁹F pulses and a dephased spectrum (S) measured with the ¹⁹F pulses yielded the signals of amide protons in close proximity to the fluorines. The difference spectrum of W-GB1 after 1.89 ms ¹H-¹⁹F REDOR mixing (Fig. 3a) shows G41, T53, V54, I6 and N8 signals, consistent with previous results²¹. In comparison, the REDOR difference spectrum of FF-GB1 (Fig. 3b) shows a different set of peaks that chiefly involve residues in the N-terminal half of the protein. Resonance assignment using 3D hCANH (Fig. 3d) indicates that FF-GB1 has slightly different ¹H and H^N chemical shifts from W-GB1 (Fig. S2a, b),^53–54. For example, E27, T53 and V54 in FF-GB1 are perturbed compared to W-GB1, and Q2, T17, A24, T25 and A26 in FF-GB1 show peak splitting. Since the singly fluorinated W-GB1 has similar ¹H and ¹⁵N chemical shifts as those of hydrogenated GB1³³, the chemical shift changes of FF-GB1 might reflect a small degree of conformational perturbation due to the incorporation of two fluorines.

Figure 3.

Comparison of 2D hNH-resolved ¹H-¹⁹F REDOR difference spectrum (red) and 2D OaB REDOR-CP (F)NH correlation spectrum (black) of fluorinated GB1. (a) Spectra of FF-GB1. The REDOR ΔS spectrum was measured with a mixing time of 1.89 ms, while the OaB REDOR-CP (F)NH spectrum was measured with a ¹H-¹⁹F mixing time of 2 × 1.1 ms. All strong difference signals in the REDOR ΔS spectrum are also detected in the (F)NH spectrum. (b) Spectra of W-GB1, measured using the same REDOR mixing times as in (a). ¹H and ¹⁵N chemical shift assignment was taken from a previous study³³. (c) Representative 1D cross sections of the REDOR ΔS spectra and (F)NH spectra to compare the signal-to-noise ratios (SNRs) of the two experiments. The SNRs of V54 and E19 are indicated. For this sensitivity comparison, the 1D cross sections are processed with the same Gaussian window function (LB = −15, GB = 0.07) for the REDOR and (F)NH spectra, while the 2D spectra shown in (a, b) are processed using slightly different window functions. (d) Representative ¹³CA-¹H^N planes of the 3D hCANH spectrum of CDN-labeled FF-GB1. Positive and negative intensities are represented by blue and orange contours. Cross peaks within each residue are connected by vertical dashed lines in each strip, while sequential cross peaks are connected by horizontal dashed lines between strips. Blue dashed lines at ~40 ppm in the ¹³C dimension mark the boundary of the ¹³C dimension above which peaks are aliased. Chemical shifts assignment is guided by literature values^53–54.

The different REDOR ΔS spectra between W-GB1 and FF-GB1 are not surprising. F30 lies near the N-terminus of the protein, surrounded by the first and second β-strands, and F52 points to the α-helix after the β2 strand (Fig. 2d). In comparison, W43 resides on the β3 strand and is surrounded by residues in the C-terminal half of the protein. Thus, the three aromatic fluorines are surrounded by distinct residues, which should give rise to distinct REDOR difference spectra. For structurally unknown proteins, which fluorine atom causes dipolar dephasing to which H^N cannot be concluded from the REDOR difference spectra, and explicit correlation of the ¹⁹F chemical shifts with the ¹H and/or ¹⁵N chemical shifts is required.

To assign the fluorine-proximal H^N signals to each Phe sidechain in the doubly fluorinated FF-GB1, we first conducted a 2D (F)NH version of the 3D OaB REDOR-CP FNH experiment. The omission of the ¹⁹F chemical shift evolution allows us to evaluate the sensitivity and feasibility of this experiment. The 2D (F)NH correlation spectra of W-GB1 and FF-GB1 (Fig. 3a, b), measured with a ¹H-¹⁹F REDOR mixing time of 2 × 1.1 ms, show good agreement with the hNH-resolved REDOR difference spectra. All residues that exhibit S/S₀ values of less than 0.8 in the 1.89 ms hNH-resolved ¹H-¹⁹F REDOR spectra exhibit cross peaks in the (F)NH spectrum. Those residues that have less substantial REDOR dephasing, such as I6 and L7 in FF-GB1, which have S/S₀ values greater than 0.9, do not show strong cross peaks in the 2D (F)NH spectra after 7 hours of signal averaging. These 2D (F)NH correlation spectra were measured with 2-fold longer experimental time than the hNH-resolved REDOR difference spectra. But the signal-to-noise ratios of the (F)NH cross peaks are still 2–3 fold lower than the REDOR difference spectra (Fig. 3c). Thus, the 2D (F)NH experiment has 20–40% of the sensitivity of the hNH-resolved REDOR experiment (Fig. 3c).

A 2D F(N)H experiment that correlates the ¹⁹F and ¹H^N chemical shifts while omitting the ¹⁵N chemical shift evolution is another approach for ¹⁹F chemical shift assignment. The 2D F(N)H spectrum of FF-GB1 (Fig. 4a) shows three ¹H cross peaks for the upfield ¹⁹F signal at −111.1 ppm and two strong ¹H cross peaks for the downfield ¹⁹F peak at −109.2 ppm. Comparison with the known H^N chemical shifts of GB1 allows us to assign the ¹H cross peaks in the −111.1 ppm cross section to T16, T17 and T18 or E19, while the two ¹H signals in the −109.2 ppm ¹⁹F cross section can be assigned to A23 or E27 and A26, respectively.

Figure 4.

Comparison of 2D ¹⁹F-¹H correlation spectra of FF-GB1 measured using the OaB REDOR-CP FNH experiment and the LG-CP NHF experiment. The spectra were measured under 38 kHz MAS. (a) 2D OaB REDOR-CP F(N)H spectrum measured using a ¹H-¹⁹F REDOR mixing time of 2 × 1.1 ms. The ¹⁹F peaks at −111.1 ppm and −109.2 ppm can be assigned to F30 and F52, respectively, based on the H^N cross peaks. ¹H cross sections are shown on the right. (b) 2D LG-CP (N)HF spectrum, measured with a ¹H-¹⁹F CP contact time of 1.4 ms. Note the two frequency dimensions are rotated from the spectrum in (a). Only one H^N cross peak is observed in each ¹⁹F slice, indicating dipolar truncation of the weak ¹H-¹⁹F couplings by the strong ¹H-¹⁹F coupling.

The 2D LG-CP (N)HF experiment is another way to correlate ¹⁹F and H^N chemical shifts (Fig. 4b). Using a ¹⁵N-¹H LG-CP contact time of 0.8 ms and a ¹H-¹⁹F CP contact time of 1.4 ms, we obtained a 2D spectrum that shows one main ¹H cross peak for each ¹⁹F signal. The ¹H resonances in the −109.2 ppm ¹⁹F cross section can be tentatively assigned to A23 and E27, while the ¹H cross peak in the −111.1 ppm cross section can be assigned to T17. Compared to the OaB REDOR-CP F(N)H spectrum, the LG-CP (N)HF spectrum shows only one ¹H cross peak for each ¹⁹F. We attribute this to dipolar truncation of weak ¹H-¹⁹F coupling by the strong ¹H-¹⁹F dipolar coupling during the ¹H-¹⁹F CP step⁵⁵. In addition, spin diffusion between the amide protons and aromatic protons during ¹H-¹⁹F CP could preferentially enhance the intensities of certain amide protons over others.

In addition to the different numbers of H^N-F cross peaks, the OaB REDOR-CP F(N)H experiment and LG-CP (N)HF experiment differ in the ¹H chemical shift resolution. The former gives high ¹H chemical shift resolution due to ¹H detection, while the sparse ¹⁹F spectrum is encoded in the indirect dimension. The LG-CP (N)HF experiment detects the sparse ¹⁹F spectrum while encoding the ¹H chemical shifts in the indirect dimension, thus giving inferior ¹H spectral resolution. For these reasons, we chose to conduct the 3D ¹⁹F, ¹⁵N and ¹H correlation experiment using the OaB REDOR-CP pulse sequence.

Fig. 5a shows the 3D OaB REDOR-CP FNH spectrum of FF-GB1, measured in 39 hours. The ¹H-¹⁵N plane for the −111.1 ppm ¹⁹F peak exhibits T16, T17 and E19 cross peaks, whereas the −109.2 ppm ¹⁹F cross section shows ¹H-¹⁵N cross peaks for A23, A24, A26 and E27. The sum of the two cross sections matches the 2D (F)NH spectrum (Fig. 5b), as expected. Based on the GB1 structure (Fig. 5c), we can assign the −111.1 ppm ¹⁹F peak to F30, whose Hζ atom, replaced by ¹⁹F here, has short distances of 5.7 – 6.1 Å to the three amide protons resolved in the plane (Table 1). The −109.2 ppm peak can be assigned to F52, whose Hζ is 3.4 – 5.2 Å away from the four resolved H^N sites. Among these four amide protons, the close contact of E27 H^N to F52 Hζ had not been detected in the 2D (F)NH spectrum (Fig. 5b), the F(N)H spectrum (Fig. 4a), and ¹H-¹⁹F REDOR difference spectrum (Fig. 3a). Thus, 3D ¹⁹F-¹H-¹⁵N correlation allowed full resolution of the short distances between the fluorines and their neighboring amide protons. The 3.4 Å distance between F52 Hζ and E27 H^N is much shorter than the 5.0 Å distances between F52 Hζ and the three Ala’s amide protons. This suggests that the single H^N peak in the F52 ¹⁹F cross section in the 2D LG-CP (N)HF spectrum (Fig. 4b) may arise from E27.

Figure 5.

OaB REDOR-CP 3D FNH spectrum of FF-GB1 to demonstrate the resolution of H^N-F contacts. The spectra were measured under 38 kHz MAS. (a) 3D FNH spectrum, with the two 4-¹⁹F Phe cross sections shown separately for F30 and F52. Resonance assignment was made based on the 2D F(N)H and (F)NH spectra. (b) 2D (F)NH spectrum, measured using a ¹H-¹⁹F REDOR mixing time of 2 × 1.1 ms. (c) Crystal structure of GB1 (PDB: 2LGI). The residues whose amide protons are close to the two Phe residues and that are detected in the 3D FNH spectra are indicated. The residues closest to 5-¹⁹F-Trp43 are identified by the 2D (F)NH spectrum in Fig. 3b.

Table 1.

Distances between Hζ of the two Phe residues and neighboring amide protons of GB1 (PDB: 2LGI). The listed amide protons correspond to the signals observed in the ¹⁹F, ¹H, ¹⁵N correlation spectra. Distances for observed 4-¹⁹F-F30 to H^N and 4-¹⁹F-F52 to H^N cross peaks are bolded.

Residue	Hζ F30	Hζ F52
T16 HN	6.5 Å	12.1 Å
T17 HN	5.7 Å	12.8 Å
E19 HN	6.1 Å	11.2 Å
A23 HN	10.1 Å	5.1 Å
A24 HN	10.8 Å	5.2 Å
A26 HN	7.3 Å	4.9 Å
E27 HN	7.4 Å	3.4 Å

3D OaB REDOR-CP FNH experiment of F₄-TPP⁺ bound EmrE

With this demonstration of the 3D FNH experiment on GB1, we next applied the technique to the bacterial transporter EmrE. EmrE is a dimeric membrane protein that effluxes polyaromatic cationic substrates across the inner membrane of gram-negative bacteria to cause multidrug resistance^56–58. Substrate export against the concentration gradient is driven by coupling to proton import from the acidic periplasm to the neutral cytoplasm. Using ¹⁹F-¹H REDOR NMR, we recently determined two high-resolution structures of EmrE bound to a tetra-fluorinated substrate, F₄-TPP^{+ 37–38}. The two structures were solved at pH 5.8 and pH 8.0 to understand how the protonation state of the proton-selective residue E14 affects the substrate-bound structures of the protein.

Interestingly, although the ligand TPP⁺ has tetrahedral symmetry around the central phosphorous, the four fluorines at the corners of the ligand are not structurally equivalent after binding. At both low and high pH, the 1D ¹⁹F NMR spectra resolve multiple chemical shifts (Fig. 2c). The high-pH protein-substrate complex exhibits three resolved ¹⁹F signals, numbered as 4 to 1 from the downfield to the upfield chemical shifts. This chemical shift distribution indicates that the four fluorines of the substrate experience different chemical and conformational environments, most likely due to their contact with different protein residues. Although we measured 2D hNH resolved ¹H-¹⁹F REDOR difference spectra, without ¹⁹F correlation to ¹H or ¹⁵N chemical shifts, we did not directly assign which of the four fluorines caused dipolar dephasing to specific protein amide protons. Instead, the disambiguation of the short-distance ¹⁹F-H^N spin pairs was carried out computationally during structure calculation.

The 3D OaB REDOR-CP FNH experiment allowed us to assign the ¹⁹F peaks with respect to their neighboring protein amide protons. We first measured a 2D F(N)H spectrum to correlate the three resolved ¹⁹F signals of the ligand with the H^N chemical shifts (Fig. 6a). The most downfield ¹⁹F peak at −97 ppm (site 4) has the narrowest linewidth and was previously shown to have the strongest ¹³C-¹⁹F dipolar coupling with the protein³⁸. Consistently, the ¹⁹F-¹H cross peak intensities are the highest for site 4, followed by site 3, the ¹⁹F signal at −99 ppm. No ¹H assignment can be made from this 2D F(N)H spectrum due to substantial resonance overlap in the ¹H dimension. By introducing a ¹⁵N chemical shift dimension, it became possible to resolve the amide protons that are correlated to the three fluorines. In the −97 ppm ¹⁹F cross section for site 4, three ¹H-¹⁵N cross peaks are resolved and can be assigned to S43A (of subunit A), F44A and E14A (Fig. 6b). In the −99 ppm ¹⁹F cross section for site 3, we resolved four peaks that can be assigned to Y60A/V64A, A61A and I68A. Inspection of the high-pH EmrE structural model (Fig. 6c) indicates that the ¹⁹F atom shown in red is in close proximity to the S43A (5.0 Å), F44A (4.5 Å) and E14A (5.2 Å) amide protons, thus it can be assigned to site 4. The ¹⁹F atom colored in magenta is in close proximity to V64A (4.0 Å), Y60A (6.9 Å), A61A (7.1 Å) and I68A (7.5 Å), and thus can be assigned to site 3.

Figure 6.

2D and 3D OaB REDOR-CP FNH spectrum of F₄-TPP⁺ complexed to EmrE at pH 8. (a) 2D F(N)H spectrum, measured under 38 kHz MAS using a ¹H-¹⁹F REDOR mixing time of 2 × 1.1 ms. (b) 3D FNH spectrum of substrate-bound EmrE. Three cross sections at ¹⁹F chemical shifts of −101 ppm, −99 ppm and −97 ppm are shown. Assignment of the ¹H-¹⁵N cross peaks is based on previously reported chemical shifts³⁸. (c) Structure of F₄-TPP complexed EmrE in lipid bilayers at high pH. The four ligand fluorines are surrounded by different protein residues. Site-4 fluorine (red) is in close proximity to S43A, F44A and E14A, while site-3 fluorine (magenta) is in close proximity to Y60A, V64A, A61A and I68A. These are consistent with the ¹H-¹⁵N cross peaks seen in the respective ¹⁹F cross sections of the 3D spectrum.

Discussion and Conclusions

The 2D and 3D ¹⁹F-¹H-¹⁵N correlation spectra of GB1 (Fig. 4, 5) and EmrE (Fig. 6) above demonstrate that the OaB REDOR-CP technique is effective for revealing which protein amides are close to each fluorine in a multi-fluorinated system. The correlation of the three frequency dimensions allows structurally based assignment in two contexts. First, if the ¹⁹F chemical shifts of individual residues are already known based on mutagenesis and single-site fluorination⁵⁹, then the FNH spectrum provides information about which fluorine is close to which amide protons in the hNH-resolved REDOR difference spectra. For structurally unknown proteins that contain n fluorines, the lack of ¹⁹F correlation gives rise to an n-fold ambiguity, which is removed by the FNH correlation experiment. Second, if the ¹⁹F chemical shifts are not known a priori, which is expected to be the case for many small molecules and when single-site fluorination of a protein is difficult to accomplish, then the FNH correlation spectrum allows the association of each fluorine to a collection of its nearest amide protons. Thus, even if the fluorine identity is not known, the grouping all amide protons that are dephased by the same fluorine dramatically reduces ambiguity in structure calculation.

It is of interest to compare the present OaB REDOR-CP FNH experiment with previously reported 2D and 3D correlation experiments that involve a ¹⁹F dimension. 3D ¹⁹F-¹H-¹H (FHH) and ¹⁹F-¹⁹F-¹H (FFH) experiments have been demonstrated on pharmaceutical compounds^{24, 30}. The ¹⁹F–¹H correlations are established by CP while the ¹H-¹H and ¹⁹F-¹⁹F correlations are established by radiofrequency-driven recoupling (RFDR) under ~65 kHz MAS. These 3D experiments take advantage of the 100% natural abundance of ¹⁹F and ¹H. However, they cannot be easily extended to macromolecules because the number of protons that need to be spectrally resolved is much larger than in small molecules even with protein perdeuteration. Therefore, multidimensional correlation involving another heteronuclear spin beside ¹H is necessary for spectral assignment and distance analysis.

Interestingly, the OaB REDOR pulse sequence was recently shown to have inferior efficiency and sensitivity compared to TEDOR sequences for ¹³C-¹⁹F correlation²². We attribute this opposite behavior of ¹³C-¹⁹F and ¹H-¹⁹F correlation to the lack of ¹³C spins in the fluorinated aromatic residues. ¹⁹F T₁ and T₂ relaxation are well known to be more rapid than for other nuclei due to its large CSA and strong couplings to protons. Thus, for ¹³C-¹⁹F correlation, the experiment that entails the least ¹⁹F relaxation will have the highest sensitivity, while for ¹H-¹⁹F correlation, the experiment that suffers the least dipolar truncation will outperform other experiments.

The 3D ¹⁹F-¹⁵N-¹H correlation technique demonstrated here is compatible with the commonly used 2D hNH experiment for measuring ¹H-detected solid-state NMR spectra under fast MAS. Since the 3D FNH correlation experiment has lower sensitivity than the 2D implementations, only relatively short ¹⁹F-¹H REDOR mixing times should be used. This restricts the distance range that can be measured in the 3D experiment to less than 1 nm. However, the main purpose of the 3D experiment is to assign each fluorine to its spatially proximal amide hydrogens. Thus, we anticipate that a divide-and-conquer approach of conducting the 3D FNH experiment with short mixing times for resonance assignment and the 2D hNH resolved ¹H-¹⁹F REDOR experiments (Fig. 1d) for measuring nanometer distances to be the most fruitful. The 3D FNH experiment is also complementary to ¹³C-¹⁹F correlation, which allows the use of conformation-dependent ¹³C chemical shifts to resolve ¹⁹F chemical shifts and ¹⁹F-based distances. Therefore, these two experiments can be used in combination to measure nanometer distances between fluorinated ligands and their protein targets, or between sparsely fluorinated aromatic sidechains and backbone amide protons. Finally, the FNH experiment can be extended in two related directions. First, ¹³C instead of ¹⁵N correlation can be implemented, giving an FCH experiment that should be useful for measuring distances to protein sidechains. Second, the FNH or FCH experiment can be implemented on fully protonated samples by spinning at ~100 kHz using 0.7 mm or smaller rotors. The use of protonated samples would simplify protein expression and purification, especially for challenging systems such as membrane proteins. Under ~100 kHz MAS, we expect the ¹H-¹H dipolar couplings to not affect ¹H-¹⁹F polarization transfer beyond a moderate change of the ¹H T₂ relaxation time. A 2D Hα-Cα correlation spectrum can be potentially sufficiently resolved to augment the 2D ¹H-¹⁵N fingerprint to give site-resolved distance information. One potential drawback of faster MAS is that the ¹⁹F refocusing pulses will take up a larger fraction of the rotor period. However, previous work on REDOR with finite pulses has shown that this is not a significant limitation^{21, 60}.