Introduction

Membrane proteins do not stay isolated; they interact with other biomolecules, such as other membrane proteins, soluble proteins, lipids and small molecules, to carry out their biological functions^[1-3]. Understanding the functions of membrane proteins requires a detailed characterization of their dynamic structures.

Among numerous biophysical tools, solid-state nuclear magnetic resonance (ssNMR) is an excellent choice for the study of membrane proteins in membranous environments^[4-8]. While the traditional ¹H, ¹⁵N and ¹³C ssNMR provide plenty of information about membrane protein structure, ¹⁹F ssNMR is an attractive technique in probing interactions within the membranes because of the excellent NMR properties of the ¹⁹F nucleus. The ¹⁹F nucleus with spin of 1/2 has a high gyromagnetic ratio. These two properties lead to a simple spectral line shape and high NMR sensitivity, which allows the detection of ¹⁹F compounds with only 0.5 mg materials. The ¹⁹F nucleus also possesses a wide chemical shift range of δ~400^[9], which means that subtle changes in the local chemical environment might lead to chemical shift variations that are detectable with ¹⁹F ssNMR^{[10, 11]}. Importantly, the total absence of fluorine in nature facilitates the easy identification of target ¹⁹F compounds in complex cellular environments, making ¹⁹F ssNMR suitable for in vivo studies.

Recently, ¹⁹F ssNMR has been used to elucidate protein assembles including parainfluenza virus fusion protein^{[12, 13]}, alamethicin dimers^[14] and amyloid-like beta-sheet peptide^[15], and to characterize orientation of peptides such as cyclotide kalata B1^[16], antimicrobial peptide PGLa^[17] and SB056^[18]. ¹⁹F ssNMR has also been used for in vivo ssNMR. In this review, we summarize the recent progress using ¹⁹F ssNMR to probe interactions among proteins and other biomolecules in bio-membranes. We introduce labeling strategies, ¹⁹F ssNMR approaches and the impressive studies that have used ¹⁹F ssNMR to characterize protein interactions. We also briefly discuss in vivo applications of ¹⁹F ssNMR. Other recent reviews^[19-25] have summarized the successful use of solution ¹⁹F NMR to characterize the structure and dynamics of soluble proteins, both in vitro and in vivo.

1 Preparation of ¹⁹ F-labeled proteins

The ¹⁹F nucleus can be introduced into proteins using fluorinated amino acids or covalently attached to proteins by chemical modification of the thiol group of cysteine residues. The ¹⁹F-probes used for protein studies always contain either a mono F- or a CF₃-reporter. Single fluorine is the simplest way of detecting change in the chemical environment, and the ¹⁹F-¹⁹F dipolar coupling within the CF₃-triad could provide information about the orientation and dynamics of the target protein. Three strategies have typically been used to prepare fluorinated proteins: chemical synthesis, biosynthesis and chemical modifications, as shown in Fig. 1.

1.1 Chemical synthesis

Based on the solid-phase peptide synthesis (SPPS) method pioneered by Merrifield^[26], peptide synthesis can be achieved by chemically coupling amino acids sequentially onto the solid supporter (resin). The SPPS process can now be performed using an automated synthesizer machine. With the automated SPPS method, fluorinated unnatural amino acids can be incorporated in opted positions. Commercially available fluorinated amino acids are mostly aromatic amino acids, such as 3F-Phe^{[27, 28]}, 4F-Phe^{[29, 30]}, 5F-Trp^[31-33], 6F-Trp^{[34, 35]} and 3F-Tyr^[36-40]. Further separation and identification are necessary to characterize the racemization. It is worth noting that the SPPS method is practical for peptides that consist of less than 40 amino acid residues. For longer peptides, the coupling efficiency is dramatically reduced.

1.2 Biosynthesis

Biosynthesis is potential to produce large proteins, and E.coli strains are often used as expression hosts. The DNA sequence for the protein or peptide of interest is usually carried by a plasmid and transformed into the hosts. The fluorine labels can be introduced into target proteins by feeding ¹⁹F-labeled unnatural amino acids, such as 5F-Trp, into the growth media^[31-33]. The ¹⁹F-labeled unnatural amino acids are then absorbed by the host cells and incorporated into the expressed proteins at the positions of the native amino acids of the same type. This biosynthetic approach is convenient for generating residue-type specific ¹⁹F-labeling but faces challenges. The method can produce low yields due to the toxicity of fluorine and the inefficient uptake by the corresponding tRNA-synthetase, which results from the high competition among native amino acids from cellular environments.

Another biotechnical advance that is suitable for ¹⁹F-labeling is the use of cell-free systems to enable protein expression in cell extract systems^{[41, 42]}. The fluorinated amino acids can be easily added into the open system and captured by the synthetase^{[43, 44]}.

1.3 Chemical modification

Fluorination at aliphatic sites can be achieved in a protein by covalent chemical modification on the thiol group of cysteine residues. The fluorinated tags usually contain a trifluoromethyl-group and attach to cysteine by a disulfide or a S-C bond. For example, 2, 2, 2-trifluoroethanethiol (TFET)^{[11, 45]} and 3-bromo-1, 1, 1-trifluoroacetone (BTFA) ^{[46, 47]} are two of the most commonly used ¹⁹F NMR probes due to their relatively small size. The chemical modification method results in straightforward labeling and might be applicable for in vivo studies.

2 ¹⁹F ssNMR technique for membrane proteins study

Bio-membranes provide a hydrophobic environment for proteins to fold into a specific structure. Membranes also offer an enclosed place for various biological interactions, which involve different types of proteins, i.e., A. antimicrobial peptides (AMPs), B. cell-penetrating peptides (CPPs), C. membrane channel protein, D. membrane fusion peptides (FP), E. membrane receptor protein, as shown in Fig. 2. Membrane-active proteins interact with other proteins in membrane environments, and ssNMR is an outstanding tool to study membrane-active proteins. The technique can determine the structure or topology of small peptides and can also address the local structure of large proteins. For example, ssNMR can determine the distances between selectively labeled sites on a protein, which facilitates the understanding of mechanisms for protein-protein or protein-ligand interactions. Also, the orientation and topology of membrane proteins and their interactions with lipids can be studied with ssNMR. A number of ¹⁹F ssNMR techniques have emerged to study protein structure, orientation and interactions.

2.1 Orientation determination

The orientation of membrane proteins in lipid bilayers is critical to understanding the topology, structure and function of membrane proteins. The orientation and dynamic behavior of membrane proteins in bio-membranes can be described by a set of constraints (Fig. 3). Using the α-helix as an example, those constraints include the tilt angle (τ) with respect to the membrane normal (N), the azimuthal rotational angle (ρ) that describes the tilt direction of the peptides and the order parameter (S_mol) that represents the wobbling motion.

The orientation information is commonly derived from static NMR experiments using uniaxially aligned samples, rather than the dipolar recoupling techniques under magic angle spinning (MAS) ssNMR. The CF₃-group is an effective probe in terms of minimum material consume, maximum sensitivity and experimental ease. To obtain information on membrane alignment and motional dynamics, the CF₃-probe needs to be rigidly attached to the peptides in a well-defined geometry, e.g., CF₃-phenylglycine (CF₃-Phg). A single pulse ¹⁹F NMR experiment allows the simultaneous detection of both the anisotropic chemical shifts and the homonuclear dipolar coupling within the CF₃-group. The fast rotation of the CF₃-group around the C-CF₃ axis makes the three spins chemically and magnetically equivalent. The homonuclear dipolar interaction within the CF₃-triad results in a triplet in the static ¹H-decoupled ¹⁹F NMR spectrum. The splitting of the triplet depends on the angle θ between the C-CF₃ axis and the external magnetic field B₀, where θ is given by

$ \mathit{\Delta }_{{\rm{FF}}}^0 = \mathit{\Delta }_{{\rm{FF}}}^{\max }(3{\cos ^2}\theta - 1)/2 $

(1)

When the CF₃-axis aligns parallel to the membrane normal N, it theoretically produces the maximum intra-CF₃ dipolar coupling, given by

$ \mathit{\Delta }_{{\rm{FF}}}^{\max } = \frac{1}{2}\frac{{3{\gamma ^2}\hbar }}{{r_{{\rm{FF}}}^3}} = 8.{\rm{7 kHz}} $

(2)

The Δ and the r represent the dipolar couplings and the distance between the two ¹⁹F nuclei. Any additional motion to the CF₃-rotation, such as axial rotation in the crystalline lipid membrane around the membrane normal, molecular wobble or diffusion, further scales the dipolar splitting by an order parameter, given by

$ \mathit{\Delta }_{{\rm{FF}}}^{} = {S_{mol}}\mathit{\Delta }_{{\rm{FF}}}^0 $

(3)

and can be used to describe the molecular motional dynamics.

Obtaining the orientational constraints τ, ρ and ${S_{mol}}$ for the whole peptide requires the collection of the segmental information of at least four positions around the peptide helix. The optimum solution for the peptide orientation and molecular motional dynamics can be obtained by comparing the experimental data with simulations that consider all possible orientations of the peptides, with an assumption of an ideal helix. The simulations require prior knowledge about the side chain geometry of the CF₃-label.

2.2 Distance measurements

The distance between ¹⁹F nuclei can be measured based on ¹⁹F-¹⁹F dipolar interactions. The introduction of ¹⁹F-labels prolongs the detection distance to the maximum extent. The longest distance up to 17.5 Å (1 Å=0.1 nm), corresponding to a weak dipolar coupling of 30 Hz, can be obtained between ¹⁹F-¹⁹F, just less than ¹H-¹H. Approaches using ssNMR for distance measurements have recently been developed, and the basic concept underlying these ssNMR methods is the detection of distance-dependent mononuclear and heteronuclear dipolar couplings (~1/r³), spin diffusion (~1/r⁶) or the paramagnetic relaxation effect (PRE; ~1/r⁶). Typical NMR approaches for measuring distance are shown in Fig. 4.

2.2.1 Car-Purcell-Meiboom-Gill (CPMG) experiment

The CPMG sequence is of great interest, as it is the unique approach for studies of protein interactions in fluid membranes under physiological conditions. The CPMG sequence utilizes a train of echo pulses after the first π/2 pulse, suppressing all interactions except homonuclear dipolar coupling^[48-52]. Between every two successive π pulses, the echoes are recorded with a new dwell time (dw) as a function of the acquisition time. Fourier transformation gives the dipolar spectrum with pure homonuclear dipolar coupling.

2.2.2 Rotational-echo double-resonance (REDOR) experiment

As a representative method for detecting heteronuclear dipolar coupling, REDOR is a versatile sequence that can be adapted to measure all types of spin pairs. It is routinely used to measure ¹⁵N-¹³C spin pairs^[53] and more recently ¹⁹F-involved spin pairs such as ³¹P-¹⁹F^[54], ²H-¹⁹F^[55], ¹³C-¹⁹F^[56] and ¹⁵N-¹⁹F^[57]. A REDOR experiment utilizes a train of rotor-synchronized 180° pulses on the dephasing channel (for an isolated X-¹⁹F spin pair, X to be obtained and ¹⁹F to be dephased), which can prevent the averaging of dipolar coupling by MAS and result in dephasing of the X magnetization transversed from the ¹H by cross polarization (CP). The obtained spectrum is termed S and is followed by a reference spectrum S₀ without the dephasing pulse trains. The plot of S/S₀ as a function of the dephasing time can be fitted to obtain the inter-spin distances.

2.2.3 Centerband-only detection of exchange (CODEX) experiment

CODEX experiment is based on the concept of ¹H-driven spin diffusion under MAS strategy. It provides an attractive distance measurement approach because it measures the precise distance between selectively labeled sites and characterizes the oligomeric state^[58]. The CODEX sequence contains a pair of rotor-synchronized π-pulse trains spaced by a mixing time (T_m), which is an integer number of the rotor period. The π-pulse trains recouple and enhance the chemical shift anisotropy (CSA) interactions. After the second 90˚ pulse, the magnetization flips back to the z-direction. During the T_m, the ¹⁹F-spins dephase, and this process is driven only by the ¹H spin diffusion between the chemically equivalent but orientationally inequivalent spins. After the second π-pulse train but before acquisition, a short T_z is added. For each desired T_m, spectrum S and reference spectrum S₀ are acquired with and interchanged T_m and T_z, and spectrum S will show reduced intensities. The CODEX curves are plotted as the value of S/S₀ against the T_m. If an oligomer bundle exists, the curve decays to an equilibrium value of 1/m, where m indicates the oligomer number, and the rate of decay can reveal the intermolecular distances. CODEX experiments have been carried out at extremely low temperatures (e.g., 240 K), at which all the molecular motions are frozen and only distance-dependent spin diffusion occurs. The unwanted aggregation or denaturation of proteins at such low temperatures needs to be avoided.

2.2.4 Paramagnetic relaxation enhancement (PRE) experiments

NMR experiments using PRE are a feasible alternative to obtain distances (12~24 Å) that exceed the threshold of the other NMR approaches^[59]. The T₂ relaxation rate of ¹⁹F has been recorded to extract the average distances between ¹⁹F nuclei and the paramagnetic center [e.g., a S-(1-oxyl-2, 2, 5, 5-tetramethyl-2, 5-dihydro-1H-pyrrol-3-yl)methyl methanesulfonothioate (MTSL) spin label].

3 Structural insights into membrane proteins using ¹⁹F ssNMR 3.1 Probing peptide-membrane interactions by orientational determination: antimicrobial peptide PGLa

Antimicrobial peptides (AMP) are one of the most concerning membrane-active polypeptides and can be found in almost all kinds of living organisms. AMPs kill microorganisms by physical disruption of the membrane integrity. These peptides could be a new type of antimicrobial drug with a low risk of drug resistance. The peptide PGLa belongs to the family of AMPs and is a distinct member of the magainin family, which has broad-spectrum antimicrobial activity. The structure and biological function of PGLa have been extensively studied. The antimicrobial mechanism has been suggested to permeabilize the microbial membranes by pore-formation^[60-65], as shown in Fig. 2, and recent advances in understanding the action mechanism have been included in the review by Bechinger^[66].

An accurate orientational and dynamic analysis of PGLa has been carried out with only four single CF₃-Phg labeled PGLa mutants^[67]. The orientational constraints can be extracted from the spectra using best fits, and the experimental data can be compared to simulated values, which are obtained by rotating the peptide molecule through all possible orientations [Fig. 5(a)~(c)]. The obtained orientational constraints (τ = 89˚, ρ= 106˚ and S_mol = 0.6) agreed with previous ¹⁵N NMR data on PGLa, with the expected surface-bound "S-state" alignment at a peptide to lipid ratio of 1:200^[68]. By increasing the peptide concentration, the peptide can tilt with the C-terminus into the membrane as a "T-state" by self-assembly as dimers, or it can insert into the membrane in a nearly standing "I-state"^{[67, 68]}. The orientation of the peptide PGLa is also temperature dependent^[69]; it was found to adopt an S-state above 45 ℃, a T-state with reduced temperature and even an I-state below 25 ℃ [Fig. 5(d)~(g)].

3.2 Probing protein-protein and/or protein-peptide interactions with distance measurements: M2 proton channel

Protein-protein and protein-peptide interactions include mono- and hetero-oligomerization through hydrophobic effects or electrostatic interactions that often happen within the lipid membrane or on the membrane surface. The determination of the local structure parameters in complex biological assemblies can be achieved by distance measurements. The ¹⁹F-labels incorporated into postulated oligomer interfaces allow for the exploration of possible contacts by detecting the spin diffusion of the ¹⁹F-isotopes or dipolar couplings between fluorine and other isotopes.

The M2 proton channel is an essential component in the envelope of the influenza A virus, which forms highly selective, proton-conducting channels across the membrane. It forms homotetramers in vivo^[70] and in micelles^[71]. Direct evidence about the oligomer state was first provided by ¹⁹F CODEX experiments. The M2 transmembrane segment was labeled with 4F-Phe in position Phe30 and then reconstituted into DMPC lipid bilayers at a peptide to lipid ratio of 1:15. The CODEX curve was obtained by fitting the experimental points to exponential curves. The curves decay to an equilibrium value of 1/4, which corresponds to an oligomer number of 4. The distances between distinct ¹⁹F-residues can be then obtained by analyzing the decay rate. The CODEX data revealed that the transmembrane segment of the M2 protein of the influenza A virus forms tetramers in DMPC bilayers with a nearest interhelical distance of 7.9~9.5 Å^[58] (Fig. 6). Such an interhelical distance of 7.9~9.5 Å between the Phe30 residues suggests that the side chain conformation of Phe30 restricts the tilt angle of the helix to be over 20˚ and also defines the diameter of the tetramer channel.

3.3 In vivo applications

Proteins might behave differently in native cell environments than in solution or artificial phospholipids^{[72, 73]}. As reported by Ulrich using the antimicrobial peptides PGLa and Gramicidin S, no distinct membrane tilted or inserted state was detected in native cells even though these states were found in artificial model membranes^[74]. Findings like this one indicate that studying proteins inside living cells is becoming more attractive for biophysicists. One prerequisite for cellular NMR experiments is uniform-labeling with ¹³C and ¹⁵N or selective labeling with ¹⁹F, which has been successfully explored in E. coli overexpression systems and also eukaryotic systems^[75-78]. ¹³C or ¹⁵N enrichment joined with solution NMR or high-resolution MAS ssNMR has been used for obtaining protein structures in the cellular context^[79-82].

Routine ¹³C or ¹⁵N enrichment is often insufficient for the detection of globular proteins, probably due to the high viscosity and weak nonspecific intracellular interactions of the proteins^[83]. In addition, the detection could be hampered by the intrinsic low sensitivity and high background of the ¹³C and ¹⁵N isotopes. These problems make the ¹⁹F to be a better choice in studying the globular and intrinsically disordered proteins, taking advantages of its total absence from biology and high sensitivity. The first application of ¹⁹F labeling for in-cell NMR experiments was in 1989. It used fluorinated phosphoglycerate kinase that was obtained by feeding 5F-Trp during enzyme synthesis in yeast. Two signals were detected and attributed to the two Trp residues in the protein^[84]. ¹⁹F in-cell NMR methods have also been developed to monitor cellular biological processes. For example, Okamura and colleagues observed three states of the ¹⁹F-labeled peptide R8 in living cells and confirmed the spontaneous penetration of cell-penetrating peptides (CPPs) into cell membranes^[85]. Tian and colleagues monitored the Tyr phosphorylation of Etk in lyophilized E.coli using the ¹⁹F ssNMR probe of 3, 5-difluorotyrosine incorporated at the position Tyr574 in the α-loop^[86].

4 Limitations and future perspectives

There are many fluorinated amino acids that have been developed for structural and functional studies of proteins^[87-91]. Besides the NMR approaches mentioned above, ¹⁹F ssNMR also provides simple and straightforward way monitoring membrane protein interactions by detecting chemical shifts^[92]. Despite the successes of applications of ¹⁹F in biological systems, it has an important potential limitation: the structural perturbations induced by the introduction of the unnatural ¹⁹F amino acid residues create a steric effect. The fluorination is always performed by the substitution of hydrogen, as the ¹⁹F nucleus is often considered to be isosteric with hydrogen because of their comparable van der Waal's radius. However, fluorine has the highest electronegativity of all the elements, at 3.98 on the Pauling scale^[93]. On one hand, the fluorine substitution results in a C-F bond that is stronger than the C-H bond, which is stable throughout peptide synthesis or protein bio-expression. On the other hand, substitution of hydrogen for fluorine affects the electronic properties of the protein. The fluorine atom can donate electrons, while the hydrogen atom works as electron-acceptor. While many studies have shown that the introduction of a limited number of ¹⁹F-labels do not perturb the global structure, however, the unwanted steric and electronic effects cannot be ignored, especially in cases involving fluorine and hydroxyl interactions or the substituted sites locating in tightly packed protein cores. In such cases, even one single fluorine could cause significant changes to the protein structure^[94]. On the other hand, novel fluorine labels need to be developed to meet a wide range of experimental requirements.