1 Introduction

Integral membrane proteins are encoded by approximately 30% genes of any genome, and these proteins are involved in a wide range of functional activities that are essential for the cell (1, 2). Membrane proteins, especially the membrane-bound receptors and ion channels, are the known targets for more than half of all currently used drugs (3). Successful determination of almost 250 structures of membrane proteins has been achieved in various forms such as crystals, micellar solutions, and membranes, by the use of X-ray crystallography, solution, and solid-state NMR spectroscopy, respectively (4). There still is a stark contrast between the number of protein structures studied and the functional significance of the membrane proteins in general. Determination of structures using NMR methods, in particular the high-resolution solution NMR methods, is very important because the technique provides information not only about the atomic coordinates of the structures but also dynamic motions of the backbone and side chains of membrane proteins solubilized in a large excess of mild detergents, such as DPC or DHPC, that structurally closely resemble membrane lipids and thus preserve the native structures and functions of membrane proteins in a biologically relevant environment. Membrane proteins fall into two families. Membrane proteins of the eukaryotic cell or plasma membrane and the inner (cytoplasmic) membrane of prokaryotes belong to the α-helical bundle family. Membrane proteins of the outer membranes of gram-negative bacteria, and of chloroplast and mitochondria, belong to the β-barrel family (5, 6). In recent times, solution NMR spectroscopy methods have been very successful for determining the solution structures of several large α-helical and β-barrel membrane proteins like OmpG, VDAC-1, UCP2, DsbB, DAGK, pSRII, KcsA, and proteorhodopsin (7–14). In addition to structure and dynamics, NMR technique also offers unique opportunities to study the interactions of the membrane proteins with several other interacting partners. This chapter aims to present the recent advances in the methodology of sample labelling and preparation and the development of newer NMR-based strategies and their successful applications. Needless to say that pushing the boundaries of existing techniques for structure determination of membrane proteins is an ongoing and continuous process.

2 Samples

2.1 Preparation of Isotopically Labeled Membrane Proteins

NMR solution studies on membrane proteins are based on multidimensional multinuclear NMR experiments on 0.5–1.5 mM proteins that are isotopically labeled with 15N, or 13C, 15N, or 2H, 13C, 15N isotopes. Based on the molecular weight, this can translate into 5–15 mg of membrane protein per sample at a concentration of 15–30 mg/ml. In most of the cases, the production of membrane protein has been achieved by bacterial expression using a suitable E. coli strain (15), although other available cellular expression systems are gram-positive bacteria (Lactococcus lactis), monocellular eukaryotes (Pichia pastoris and Saccharomyces cerevisiae), and cell lines from higher eukaryotes (insect cells and mammalian cells). The gene encoding the desired protein is cloned in expression vectors and transformed into a bacterial host. Vectors containing the T7 and T5 viral promoters, especially the pET and pQE series, are quite commonly used. The common bacterial hosts are E. coli strains BL21(DE3), tuner, C43(DE3), and Rossetta(DE3). It is fairly common to express the membrane protein together with a cleavable poly-histidine tag at the N- or the C-terminal. In many cases, the cleavage site selected has been the one recognized by TEV protease. In the case of phospholamban, the 50 residue membrane peptide was expressed as MBP fusion protein, cleaved, and purified by reverse-phase gel chromatography (16).

2.1.1 Isotopic Labelling of Membrane Proteins

Assuming that bacterial expression and purification of unlabeled protein have been well established, the 2H, 13C, 15N labelling process starts by making minimal media plates in D2O in which deuterated glucose may be used if needed for acclimatization of the expression host. The competent cells of the selected host are freshly transformed with the plasmid construct, and the transformed cell suspension is plated on the MM-D2O plates. The plates are incubated over 24- to 36-h period at set temperatures, and the colonies obtained are tested for contamination. A single colony is picked and grown in a small volume in a conical flask for 24–36 h. If the cell growth is healthy and no clumping is observed, then the overnight culture is used to inoculate a larger size (1:40–1:100 dilution factor) modified M9 medium prepared in D2O containing 2–4 g/l 2H (98%)- and 13C (99%)-labeled d-glucose, and 1 g/l 15N (98%)-labeled (NH4)2SO4. Commercially available 10× media for various levels of deuterations (e.g., Bioexpress media from CIL) are available and can be prepared by simply diluting with D2O under sterile conditions. The resulting media, although prohibitively costly, gives a good growth and yield of the expressed protein. A cheaper alternative is to spike the modified M9 minimal media described above with 1% of the 10× solution of CDN-Bioexpress-1000 medium for 2H, 13C, 15N labelling. The growth of the cells is carefully monitored, and the cells are most commonly induced with standardized minimal amounts of isopropyl β-d-1-thiogalactopyranoside (IPTG). After letting the induced cells grow further for several hours, they are harvested and processed further (17).

For the production of uniformly 2H, 13C, 15N protein samples with methyl protonation of Ile-δ1-[13CH3] and Val, Leu-[13CH3, 13CH3], the CDN-Bioexpress-1000 is not added to the deuterated minimal medium described above, while 50 mg/l of 2-keto-3-d2-1,2,3,4-13C-butyrate and 100 mg/l of 2-keto-3-methyl-13C-3-d1-1,2,3,4-13C-butyrate (18) are added 1 h before induction.

In addition to stable isotope labelling, it has become quite necessary for obtaining long-range constraints to strategically introduce spin-labels covalently bound at specific positions in the sequence. Structural restraints are derived from the paramagnetic relaxation enhancement (PRE) effect of the spin-label. For this strategy, normally a cys-less mutant of protein is prepared, and then several specific single-cystine mutants are generated and labeled with S-(2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrol-3-yl)methyl methanesulfonothioate (MTSL) (19).

2.2 Choice of Detergent, Temperature, Additives, and Other Considerations

Solution NMR studies are carried out by solubilizing the proteins in detergent micelles. Since this environment is topologically and structurally different from a typical lipid bilayer of a membrane, extensive validation of the detergent-solubilized membrane protein through biochemical and biophysical characterizations in detergent micelles is a prerequisite for initiating any serious structural work. For various proteins, this has to be taken up on a case-by-case basis, although, there are some general considerations which are described here.

For purification, it is often necessary to solubilize membrane proteins either in detergent micelles or detergent–phospholipid mixed micelles. β-Barrel membrane proteins are often purified from inclusion bodies that are dissolved in denaturing agents such as 8 M urea or 6 M guanidine hydrochloride (7, 8, 17, 20). For α-helical membrane proteins, more elaborate procedures are needed starting from solubilization and purification to final reconstitution. Shorter peptides corresponding to single to two transmembrane helices can be extracted in mixtures of organic solvents like methanol and chloroform. Solubilization in detergents, especially the nonionic detergents, is the mildest in terms of perturbation of protein structure in comparison to the alternatives mentioned above.

Wherever membrane proteins are reconstituted in detergent micelles for the purpose of solution NMR studies, the molar ratio of protein to detergent is always kept somewhere in between 1:100 and 1:600. This is done to minimize protein–protein contacts that would broaden and diminish the number of NMR peaks in the NMR spectra. Out of hundreds of seemingly suitable detergents for NMR studies, there are only handfuls that have been repeatedly used for most of the determined structures. Among these are dodecylphosphocholine (DPC), dihexanoyl-phosphatidylcholine (DHPC), lauryl dimethylamine oxide (LDAO), lyso-phosphatidylglycerol (LPPG), N-octyl-β-d-glucopyranoside (OG), n-dodecyl-β-d-maltoside (DDM), and sodium dodecyl sulfate (SDS).

DPC is most commonly used as a detergent for structural studies of membrane proteins by solution NMR (7, 10, 11, 16, 17, 20, 21). The CMC of DPC is 1.5 mM and typically in 50 mM NaCl, a micelle of DPC consists of 70–80 monomers at 25 C. This corresponds to an apparent molecular mass of ~26 kDa for the DPC micelle (22). Therefore, monomeric or oligomeric protein of total molecular mass up to 60–75 kDa can be reconstituted in DPC micelles for structural studies. The other detergents that have been used are DHPC (23, 30), β-OG (20), SDS, lyso-myristoyl-PG (24), and LDAO (8, 25). Schematic representations of chemical structures of some commonly used detergents are shown in Fig. 1. For NMR structural studies of small α-helical membrane proteins of MW 4–10 kDa and having 2–4 transmembrane segments, LPPG has been found to be most suitable for obtaining samples that have a long lifetime and yield high-quality NMR spectra (26).

Fig. 1.
figure 00171

Schematic representation of various detergents used for solubilization of membrane proteins for structure determination using NMR spectroscopy: (a) LDAO, (b) SDS, (c) DPC, (d) LPPG, (e) CYFOS-7, (f) β-OG, (g) DHPC.

Full size image

Detergent concentration needs to be standardized carefully by counting the number and also line widths at half height of the peaks in the HSQC spectra of proteins. For DPC, detergent concentrations from 150 to 600 mM have been used. In many of the reported studies, use of perdeuterated detergent has been made. However, based on the fast dynamics and exchange of individual detergent molecule, it would appear that perdeuteration of detergent is not necessary. The advantage of using non-deuterated over deuterated detergent is the huge difference in terms of the higher cost of the latter. However, where there is interference, such as in case of critical characterization of methyl–methyl NOEs, use of deuterated detergent is preferred. Most of the NMR studies have been conducted at temperatures between 30 and 50 C, as higher temperature improves the sensitivity of 13C by reducing the line widths. However, this could also lead to higher conformational exchange, and this is sorted by recording some preliminary 2D and 3D experiments at two or more temperatures. In the presence of certain detergents, the NMR sample can solidify at ambient temperature, leading sometimes to irreversible changes in the sample state (26).

Detergents cooperatively assemble above their CMC in aqueous solutions to form micelles. Typically, detergents have CMCs in low millimolar range. The CMCs for SDS and DHPC are 2.2 and 15.2 mM, respectively. The aggregation numbers for SDS, DPC, and DHPC are ~70, ~80, and ~35, respectively, corresponding to aggregate molecular masses of ~25, ~28, and ~16 kDa, respectively. Detergent micelles are usually depicted as idealized spherical structures. Indeed, molecular dynamics (MD) simulations of pure DPC micelles with either 56 or 65 detergent molecules have revealed almost spherical structures with a ratio of the principal moments of inertia of 1.2:1.1:1 (27). However, when membrane proteins are solubilized in detergent micelles, the number of detergent molecules associated with the protein would be determined by the hydrophobic surface of the protein that needs to be coated with the hydrophobic alkyl chains of the detergent molecules (15) and the length of the alkyl chain and the head-group of the detergent. A variety of techniques, for example analytical ultracentrifugation, dynamic light scattering, gel exclusion chromatography, and NMR gradient spin-echo diffusion, have been used to determine the sizes of the detergent–protein aggregates. Indirect estimate of the size of the detergent–protein aggregate can also be derived from the rotational correlation time and global diffusion tensor by analysis of the amide 15N T1, T2, and heteronuclear NOE relaxation parameters. The 8-stranded β-barrel proteins have all been found to have rotational correlation time of 20–22 ns corresponding to an apparent molecular weight of the detergent–protein aggregate of around 50–60 kDa (17, 20, 23, 28, 30). These values are in close agreement with the dynamic light scattering measurements and usually significantly lower than the molecular mass estimated from NMR gradient diffusion experiments (29).

There is a general consensus based on several experimental measurements like internuclear NOEs between membrane protein and detergent (OmpX/DHPC), rotational dynamics (OmpA/DPC), and molecular dynamics simulations (29) that the detergent molecules form an annular ring that coats the hydrophobic surface area of membrane proteins. The detergent/protein interface also appears to be slippery, such that the protein can rotate within the confines of the micelle. The lifetimes of detergent molecules in contact with the membrane protein could be of the order of 0.3 ns, which is based on the observation of negative intermolecular NOEs between the protein backbone amide protons and the protons on detergent molecules. Furthermore, the association of detergent molecules and protein is dynamic, and detergent molecules exchange between the protein–detergent complex, free monomeric detergent, and detergent-only micelles on a sub-millisecond timescale (29).

2.3 Sample Preparation with Some Specific Examples

2.3.1 β-Barrel Membrane Proteins

The β-barrel proteins for which NMR structures have been determined are all expressed into inclusion bodies. The inclusion bodies are solubilized into either 8 M urea (for OmpA) or 6 M guanidinium chloride (OmpX, PagP, and VDAC-1). For NMR studies, the proteins are refolded by diluting the urea or guanidinium chloride solution into buffer containing detergent micelles. Some of the specific cases are described below.

For the OmpA TM domain, refolding was carried out by slowly diluting protein solution in 15 mM Tris buffer, pH 8.5, containing 8 M urea into more than tenfold excess of 20 mM borate buffer, pH 10.0, containing 150 mM NaCl, 1 mM EDTA, and 20 mM DPC. After overnight refolding, the solution was concentrated through Amicon PM10 or YM1 membranes. The dilution–concentration step was repeated twice with NMR buffer (20 mM sodium phosphate, pH 6.3, 50 mM NaCl, 0.1% NaN3) to give a final concentration of 1 mM protein and ~600 mM DPC (17).

PagP was refolded from 6 M GdmCl solution into phosphate buffer containing DPC, pH 6.0, using procedure similar to that described above for OmpA. For studying the interaction of PagP with Cyfos-7, a sample of PagP refolded in β-octyl glucoside (OG) was also prepared by dissolving precipitated PagP in 1 ml of 5% perdeuterated SDS and dialyzing for 5 days through membranes with MWCO of 3.5 kDa, against 50 mM sodium phosphate buffer, pH 6.0. Thirty-two milligram of OG was slowly dissolved, and then d5-ethanol was added to 1%. The sample was concentrated down to yield a 1 mM PagP NMR sample in a ~200 mM OG. Both SDS and ethanol were required for 100% refolding (20).

OmpX was solubilized in 6 M GdmCl in 20 mM Tris buffer (20 mM Tris and 5 mM EDTA), pH 8.5. Solubilization and reconstitution of OmpX were performed by slowly diluting the protein solution in 6 M GdmCl into a sixfold excess of refolding buffer containing 3% DHPC, 20 mM Tris-HCl, 5 mM EDTA, and 0.6 M arginine, pH 8.5, at 4 C. The OmpX/DHPC solution was then dialyzed against 2.5 l of 20 mM Tris buffer pH 8.5, having 5 mM EDTA and 100 mM NaCl, at 4 C for 20 min to remove residual GdmCl and l-arginine. The sample was concentrated to a volume of 300 μl. The final protein concentration was about 2 mM, and the detergent concentration was adjusted to 300 mM by adding solid DHPC to the solution (30).

The refolding of OmpG sample in 8 M urea was standardized in various detergents (7) like DHPC, Fos-cholines, β-OG, and DDM at various concentrations, pH values, and temperatures, and in presence of various additives like arginine, ethanol, glycerol, and glucose. Finally, the most suitable method for folding and preparation of NMR sample that was stable for a long time involved first refolding the protein in 70 mM β-OG at pH 9 at room temperature, followed by two steps of dilution and concentration, with 10× of 15 mM DPC in 25 mM bis-Tris buffer containing 50 mM NaCl and 0.05% NaN3. The final NMR sample had OmpG concentration of 1.0–1.3 mM (7).

VDAC-1 was purified from inclusion bodies solubilized in 8 M urea, precipitated, and resolubilized in 6 M GdmHCl buffer (100 mM NaPO4 pH 7.0, 100 mM NaCl, 6 M GdmHCl, 5 mM DTT, 1 mM EDTA). Refolding of VDAC-1 was carried out at 4 C by tenfold dilution in 1% LDAO-containing buffer, followed by further purification through cation-exchange chromatography, and a series of dialysis and dilution/concentration steps. The LDAO detergent concentration was monitored by 1D 1H-NMR and adjusted by a series of dilution and concentration steps. The final sample conditions were 25 mM NaPO4, 5 mM DTT, pH 6.8, 300–500 mM LDAO, and 0.5–1 mM VDAC-1 (8).

OprH from Pseudomonas aeruginosa was also purified from inclusion bodies and refolded from 8 M urea in DHPC micelles. Briefly, a sample containing 0.4 mM OprH in elution buffer was diluted tenfold into 20 mM Tris–HCl at pH 8.5, 5 mM EDTA, and 0.6 M l-arginine (refolding buffer) with 3% DHPC. The refolding solution was incubated for 72 h at 37 C before being dialyzed (20 min, room temperature) against 2.5 l of 20 mM Tris–HCl at pH 8.5, containing 5 mM EDTA, and 50 mM KCl. The solution was concentrated, and the buffer was exchanged against 25 mM NaPO4 at pH 6.0, containing 50 mM KCl, 0.05% NaN3, with 5% D2O, by dilution/concentration. Final NMR samples were concentrated to 1.0–1.3 mM OprH and contained 150–175 mM DHPC as determined by 1H NMR spectroscopy (23).

2.3.2 α-Helical Membrane Proteins

α-Helical membrane proteins differ in the number of transmembrane domains which can vary from a single transmembrane spanning α-helix to multiple monomeric or oligomeric α-helical bundles. Sample preparation protocols vary substantially, often depending on the number of transmembrane α-helices. Simpler membrane-spanning peptides can be synthesized through solid-phase peptide synthesis while more complex systems are overexpressed and purified in native form by solubilization into detergents. Several membrane peptides synthesized through solid-phase synthesis can be lyophilized and dissolved in buffer having the appropriate detergent. Lyophilized Pardaxin (33 residues) and phospholamban (50 residues) were dissolved in 20 mM phosphate-buffered saline, pH 6.5, containing 300 and 600 mM DPC, respectively (31, 32). Several small membrane peptides, for example, E. coli and B. pseudofirmus OF4 ATP synthase c subunits, have been expressed in bacterial cells and extracted with organic solvents. The peptides were reconstituted by mixing detergent and peptide in 3:1 chloroform:methanol and drying to a thin, clear film under a stream of argon for up to 12 h. Detergent/protein films were then dissolved in 1–50 mM aqueous potassium phosphate buffer, pH 4.5–8 for NMR experiments (26). For more complex α-helical membrane proteins that have to be purified in fully folded or partially unfolded form, more complex procedures have been followed as can be judged from the specific examples described below.

A specialized protocol of reconstitutive refolding was followed for the preparation of NMR sample for DAGK. The protein that was purified in DPC micelles was inactive. In order to get active protein, DAGK/DPC micelles were mixed with POPC to generate DAGK/DPC/POPC mixed micelles. DPC was removed by extensive dialysis to yield POPC lipid vesicles containing DAGK at 1:100 DAGK:POPC:mol:mol ratio, in refolding buffer (pH 7.8 buffer containing 10 mM imidazole and 0.05 mM EDTA). The membrane preparation was resolubilized in DPC micelles, and POPC was removed to give DAGK/DPC micelles having DAGK in active form (33).

Recently structure of a bacterial plasma membrane protein which plays a key role in disulfide bond formation, DsbB, was determined. Out of the six possible intermediate states representing steps of DsbB-catalyzed DsbA oxidation, the structure was determined for a Cys → Ser double mutant (DsbB(CSSC), C41S, C104S) termed as inter-loop disulfide bond state. DsbB(CSSC) was expressed in Rosetta(DE3) cells and purified by solubilizing the membranes with Tris-HCl/NaCl buffer containing 10 g/l DPC, followed by purification over Ni-NTA superflow column with Tris-HCl/NaCl buffer containing 1 g/l of DPC. For NMR experiments, the protein sample was concentrated and exchanged in 25 mM PBS, pH 6.2, containing 50 mM KCl, and 0.47 g/l DPC. The NMR experiments were carried out at 40 C. The DsbB[CSSC] sample was found to retain activity and displayed long-term stability under these conditions. The protein was found to have an apparent size in DPC micelles of 40 kDa, corresponding to a monomer (10).

Mouse UCP2 (residues 14–309 with a C-terminal His6 tag) was expressed using a pET-21 vector in E. coli Rosetta DE3 cells. Prior to purification, the lipid composition of the membrane fraction was adjusted with suitable amounts of DMPC, cardiolipin, and trace amounts of phytanoyl lipid, and these were pre-extracted with 10% OG at 4 C. UCP2 was extracted in 40 mM potassium phosphate (pH 8.0), 250 mM NaCl, 50 mM BME, 10 mM GDP, and 0.2% DPC and was purified by following a series of metal-affinity, ion-exchange, nucleotide-affinity, and gel-filtration chromatographic procedures. In the final step, UCP2 was eluted in 50 mM potassium phosphate (pH 6.5), 100 mM NaCl, and 5 mM DPC. The eluted UCP2 sample was supplemented with GDP, detergent, and lipids such that the final NMR sample contained 0.8 mM UCP2, 5 mM GDP, 150 mM DPC, 2 mM DMPC, 1 mM cardiolipin, 5 mM BME, 30 mM potassium phosphate (pH 6.5), and 80 mM NaCl (9).

Seven-helix transmembrane receptor sensory rhodopsin II (pSRII, 1–241) from Natronomonas pharaonis plus a sequence encoding a C-terminal hexahistidine-tag was expressed from pET-28b(+) vector into the E. coli expression strain BL21 Tuner (DE3). Protein expression was carried out in the presence of 10 μM all-trans-retinal. Crude membranes were extracted with 1.5% (w/v) of detergent DDM, and solubilized pSRII was purified with nickel-nitrilotriacetic affinity beads. The protein was eluted in 50 mM Tris–HCl, pH 7.0, 300 mM NaCl, and 0.1% DDM containing 150 mM imidazole, and the solution was exchanged into phosphate buffer by repeated rounds of concentration and dilution with a final volume of 0.5 ml. For the preparation of NMR samples, the protein was mixed with nickel-nitrilotriacetic acid beads as before and, after removal of the flow-through, washed with 50 column volumes of 20 mM sodium phosphate buffer, pH 7.0, 50 mM NaCl containing 0.6% (w/v) 1,2-diheptanoyl-sn-glycero-3-phosphocholine (diC7PC). The protein was eluted in the same buffer containing 300 mM imidazole, and the eluted solution was repeatedly concentrated and diluted in a concentrator (10-kDa cutoff) with 60 ml sodium phosphate, pH 5.9, and 50 mM NaCl containing 0.06% (w/v) DHPC to a final volume of 0.45 ml (12).

Proteorhodopsin (PR) has been expressed and isotopically labeled through the cell-free expression system. The green-absorbing variant of PR was cloned without the signal peptide into the pIVEX2.3d vector optimized for cell-free expression. The construct yielded 229 residues from PR (amino acids 21–249), a five residue linker, and a hexa-histidine tag, resulting in a final size of 26 kDa corresponding to 241 amino acids. For assignment of the characteristic H75 a Strep-tag (WSHPQFEK) was cloned at the C-terminal end in place of the His-tag. Continuous-exchange cell-free expression was based on an S30 extract and followed standard protocols. PR was expressed in the presence of 0.6 mM all-trans retinal in the detergent mode (DCF) using 0.4% digitonin mixed with diC7PC in a 4:1 molar ratio. Ni-affinity purification was conducted with 0.05% DDM in the loading and washing buffer, and the detergent was exchanged into 0.1% diC7PC prior to elution. Strep-tagged PR was purified via a Strep-Tactin matrix with 0.1% diC7PC in the washing and elution buffer. Buffer exchange was achieved by either centrifugal concentration devices (10,000 MWCO) or gravity flow PD10 Desalting columns. NMR samples were buffered with 25 mM NaOAc, pH 5, and 2 mM DTT. The diC7PC concentration was brought up to 2%. For side chain assignment and 13C-edited NOESY experiments, deuterated acetate and diC7PC were used in the sample. For preparation of deuterated protein samples, cell-free expression was conducted in completely deuterated solutions (except RNAse inhibitor and T7 Polymerase). Selective labelling was achieved by replacing single types of amino acid in the unlabeled amino acid mixture used for cell-free expression with the labeled equivalent. In addition to the overlap and relaxation-optimized (oro)-SAIL amino acid, the remaining 19 amino acids were 2H labeled. Side chain assignment of oro-SAIL Leu and oro-SAIL Val was achieved with information from 1H-methyl-detected out-and-back experiments, 4D 15N,13C-separated NOESY, as well as 3D 13C-separated NOESY. The 11, 20-13C-labeled retinal was co-translationally bound to K231 by replacing the unlabeled cofactor in the cell-free expression system (14).

3 NMR Methodology

3.1 NMR Methods

Structure determination of membrane proteins has benefited tremendously from the advent of newer NMR techniques, especially transverse relaxation-optimized NMR spectroscopy (TROSY (34)). TROSY results in a substantial reduction of the 15N relaxation rates of the 15N–1H moieties during coherence transfer steps by constructive use of the interference between the 15N and 1H dipolar–dipolar coupling and 15N-chemical shift anisotropy. The TROSY pulse scheme selects the narrowest of the four component signals of the doubly decoupled 15N–1H correlations in a 15N–1H-HSQC spectrum. The TROSY pulse scheme has already been implemented in many of the standard and specialized 3D and 4D triple-resonance experiments for sequential assignments of proteins, in 3D 15N-edited NOESY experiments, and in 15N T1, T2, and {1H}15N-NOE relaxation experiments. In all such cases, the TROSY-type selection improves the sensitivity and the resolution of experiments starting on 1HN and 15N of the amide residues. Further, the TROSY effect is most efficient for deuterated proteins and at high magnetic fields (between 900 and 1,100 MHz), both of which are increasingly amenable (34–36).

3.1.1 NMR Assignments

Sequential assignment is easier to obtain for the β-barrel membrane proteins because of the relatively better chemical shift dispersion of resonances in β-sheets and the alternating nature of hydrophilic and hydrophobic residues in this class of proteins. In addition, a good number of long-range NOEs are available which, together with theoretical hydrogen bonds, allow calculation of the backbone fold to high accuracy. The sequential assignments are harder for α-helical membrane proteins because of the smaller dispersion of the shifts of many consecutive aliphatic and other apolar residues that typically occur in this class of proteins. Although sequential HN–HN NOEs in the α-helix can be helpful to some extent, it is ordinarily difficult to find enough long-range NOEs that are required to pack helices against each other and thus fold the tertiary structure.

The triple-resonance suite of experiments that have been utilized for the sequential backbone assignments include the 3D TROSY versions of HNCA, 3D ct-TROSY-HNCA, HN(CO)CA, HN(CA)CB, HNCACB, HN(CO)CACB, HN(COCA)CB, HN(CA)CO, and HNCO (35). In case of OmpG, a new, just-in-time (JIT) version of HN(CA)CO (7) was found to be critical for obtaining the backbone CO assignments. Assignments of NOEs, while being critical for generating distance constraints for structure calculations, are also useful for aiding and removing ambiguities in the backbone assignments. NOE assignments are made by recording 3D and 4D 15N-edited 1H–1H NOESY experiments in combination with HMQC, HSQC, and TROSY. For OmpG, NOEs were determined from 3D 15N–15N–1H TROSY–NOESY–TROSY and [1H, 1H]-NOESY-15N-TROSY experiments (7). For UCP2, the 3D (1HN, 1HN)-HMQC-NOESY-TROSY experiment was utilized (9). For human VDAC-1, up to 600 NOEs were identified from 3D [1H, 1H]-NOESY-15N-TROSY and [1H, 1H]-NOESY-13C-HMQC experiments in combination with a set of 4D experiments employing multidimensional decomposition (MDD) NOESY with nonuniform sampling (NUS), viz., 4D NUS-MDD-13C-HMQC-[1H, 1H]-NOESY-13C-HMQC and 4D NUS-MDD-15N-HMQC-[1H, 1H]-NOESY-13C-HMQC, at a field strength of 900 MHz (8).

For DsbB, the Ile, Leu, and Val methyl assignments were achieved by using a HMCM[CG]CBCA experiment and 3D [13C-F1, 13C-F2]-edited NOESY (200 ms mixing time) (18) on a 1.5 mM U-[2H]-, Ile-δ1-[13CH3]-, and Val, Leu-[13CH3, 13CH3]-labeled DsbB[CSSC] sample in deuterated DPC micelles. Using these experiments, methyl groups in all 8 Ile(δ1), 14 Val, and 14 out of 28 Leu side chains of DsbB[CSSC] were assigned. In general, for α-helical membrane proteins, the assignment of the chemical shifts of methyl protons can often not be completed due to the high degeneracy of chemical shifts of 13Cα/13Cβ pairs and 13CH3 proton and carbon chemical shifts of the methyl groups. However, even a few unambiguous methyl–methyl long-range NOEs are sufficient to determine the tertiary fold as in the case of DsbB in DPC micelles (10).

3.1.2 NMR Constraints

Backbone dihedral angle constraints are usually determined from chemical shift-based predictions by TALOS or TALOS+ software program which utilizes database mining against fragments of known structure (37). For a deuterated protein, the Cα, Cβ, and CO chemical shifts are corrected for deuteration effects, and the HN and N chemical shifts are corrected for the TROSY–HSQC difference (38). The mainstay of structure determination by solution-state NMR spectroscopy is to assign short-, medium-, and long-range 1H–1H NOEs which may fall into various categories like intra-residue, sequential, inter-residue, intra- and inter-β-strand, intra- and inter-α-helical, and also inter-monomer NOEs in case of oligomeric proteins. These are often combined with hydrogen bonds which are particularly useful in determining the 3D-fold of closed β-barrel proteins. For β-barrel proteins, measurement of HN–HN NOEs is amenable through 15N-edited [15N,1H]-TROSY-[1H-1H]-NOESY-[15N-1H]-TROSY or its HSQC versions (39). For smaller β-barrel systems, HN–Hα NOEs can be assigned through partial deuteration. These NOEs in combination with theoretical hydrogen bonds and dihedral angle constraints are sufficient to give good RMSD for the β-strands of small closed β-barrels. Additional NOE information can be obtained through I,L,V-methyl protonation and measurement of HN-methyl and methyl–methyl NOEs. This approach has contributed towards structure determination and refinement of several membrane proteins (8, 10, 40–43). The inclusion of even modest number of inter-helical NOEs in conjunction with additional structural restraints, such as RDCs and PREs as described below, offers the possibility of significantly improving the quality of membrane protein structure.

3.1.3 Residual Dipolar Couplings

Residual dipolar couplings (RDCs) in the range of roughly −27 to +25 Hz for single-bond-coupled heteronuclear spin pairs have offered an alternative route for structure calculation and refinement of membrane proteins (44–47). RDCs measured for membrane proteins involve directly bonded spin pairs: HN–N, Hα–Cα, N–CO, and CO–Cα. These are obtained from IPAP-HSQC and TROSY-HNCO experiments. RDCs provide angular constraints based on the orientation of internuclear vectors relative to the magnetic field, which can be used for improving structural accuracy. However, for the measurement of RDCs, weak alignment of the protein molecules is required. Weak alignment in membrane protein samples has been achieved through either steric or electrostatic interactions with an alignment medium such as bicelles, filamentous phage, strained or charged polyacrylamide gel matrices, and DNA nanotubes. Usually, DMPC/DHPC or DMPC/CHAPS bicelles are not suitable for alignment of integral membrane proteins. Filamentous phage pf1 is also not stable in the presence of detergents especially at the high experiment temperature. The rehydrated compressed or stretched polyacrylamide gels for alignment have been successfully used for OmpA, phospholamban, and PR (14, 48, 49). Useful degree of alignment was achieved for OmpA in DPC micelles in negatively and positively charged copolymer gels (48). Negatively charged copolymer gels were prepared from a 1:1 mixture of acrylamide and either acrylic acid or 2-acrylamido-2-methyl-1-propanesulfonic acid (AMPS). Positively charged copolymer gel was prepared from a mixture of 1:1 acrylamide and N-(2-acryloamidoethyl)triethylammonium chloride (APTMAC) (50, 51). Compressed gels in Shigemi tubes having copolymer concentration of 3–4% gave alignment of OmpA characterized by 1DHN in the range between −22 and +25 Hz, −10 and +20 Hz, and −10 and +25 Hz, for AMPS, acrylic acid, and APTMAC copolymers, respectively. In all cases, very good agreement, with a quality factor of ~24%, was obtained between the experimental RDCs and those calculated from the 1.65 Å crystal structure of OmpA TM domain (48, 52).

For PR, best spectral quality and sufficiently large dipolar couplings were obtained at an acrylamide:N,N′-methylene-bisacrylamide ratio of 150:1. For sample preparation, a commercially available gel system (New Era Enterprises, Inc.) was used. Gels were cast with a diameter of 6 mm, dialyzed against an excessive amount of H2O overnight, and dried at room temperature. Dried gels were incubated with the PR NMR samples and allowed to swell to its original diameter within 24 h. Afterwards, the gels were transferred into NMR tubes with an inner diameter of 4.2 mm. To reduce peak overlap, three separate combination of selectively labeled PR were prepared as follows: sample 1—Ala, Gly, Leu, Phe, and Ser; sample 2—Ala, Met, Ile; and sample 3—Phe, Thr, Trp, Tyr, and Val (14).

For alignment of UCP2, use of DNA nanotubes was made. 260 μl of 0.5 mM (2H, 13C, 15N) UCP2 sample containing 5 mM GDP, 100 mM DPC, 2 mM DMPC, 1 mM cardiolipin, 5 mM BME, 30 mM potassium phosphate (pH 6.5), and 80 mM NaCl was mixed with 260 μl of 20 mg/ml preformed DNA nanotubes. The mixture was then concentrated down to 260 μl using a Centricon. The final aligned sample gave 2H quadrupolar splitting of 5.0 Hz. A homogeneous suspension of DNA nanotubes was prepared by mixing folded monomers of staple-strand oligodeoxyribonucleotides, followed by precipitation with PEG8000 and centrifugation. The recovered nanotubes were resuspended and centrifuged once again, before concentrating to 30 mg/ml with Centricon-100 concentrator by centrifugation (9).

3.1.4 Paramagnetic Relaxation Enhancement Measurements

The PRE approach has been quite useful in enabling long-range distance constraints that are very important for getting the correct tertiary structure. It exploits the distance-dependent line broadening of NMR resonances caused by the interaction of an NMR-active nucleus with an unpaired electron. This approach has been successfully used to complement the 1H–1H NOEs long-range distance information and the other experimental constraints for determining the structures of several α-helix and β-barrel membrane proteins. The starting point for PRE is preparation of cysteine-less protein and then rationally introducing single Cys mutations usually at rigid positions where spin-labelling is not expected to spatially or functionally affect the protein being studied. The single Cys residues are most commonly covalently linked with the spin-label MTSL. Due care must, however, be taken to check for incomplete tagging. Once a probe has been successfully incorporated, the effects of the PRE are typically measured by comparing peak line widths and intensities between spectra acquired under paramagnetic and diamagnetic conditions. The PREs induced by the spin-label can be measured by comparing the labeled sample either with the unlabeled samples or with the same sample after reduction of the spin-label with excess of ascorbic acid. A specific protocol is described below for UCP2 (9).

To introduce a single paramagnetic site for PRE measurement, the five cysteines of UCP2 were all mutated to alanine or serine (Cys 25 Ala, Cys 191 Ser, Cys 217 Ser, Cys 227 Ser, Cys 256 Ser); the cysteine-free UCP2 had GDP binding properties similar to those of wild-type protein. A single cysteine was introduced into the protein on the basis of the known secondary structures from molecular fragment replacement (MFR) segments. MTSL was then attached at the cysteine position by adding sevenfold excess label to 10 mM UCP2 in the NMR buffer at pH 8.0 and incubating at 25 C for 4 h. Excess label was removed to avoid nonspecific broadening. The pH was changed back to 6.5 for NMR measurements. To quantify residue-specific broadening of backbone 1HN, two TROSY-HNCO spectra were recorded, one after nitroxide labelling and the other after reduction of the nitroxide-free electron with 5 M excess of ascorbic acid. On the basis of observable PREs, distance constraints were created as follows: For resonances that were strongly affected by the spin-label, distances of less than 13 Å were assumed. For resonances showing measurable change in intensities and line widths, distances ranging from 13 to 20 Å were converted into upper and lower limits with a ±4 Å error margin. While for resonances that did not display a detectable PRE effect, a lower distance limit of 20 Å was used (9).

3.2 Structure Calculations and Refinement

The recipe for structure calculations and refinement by most of the common programs like CYANA, CNS, and XPLOR (53–55) is to analyze, calculate, and iteratively refine both the calculated structures and the experimental information to a high level of consistency. The input parameters for structure calculations are the dihedral angle restraints, NOE-derived distance restraints, hydrogen bonds, PRE-derived long-distance restraints, RDC restraints, radius of gyration restraint, database potentials, and restraints from biochemical experiments and/or those involving ligand/cofactor. Some specific examples are described below. Overall, the methodologies described have proven to be of great value in calculation and refinement of structures of large membrane proteins and in particular α-helical proteins having many transmembrane segments.

In the simplest of cases as for PagP, backbone dihedral angles were calculated with TALOS using chemical shift assignments corrected for deuterium isotope shifts. 1H–1H NOEs were obtained from 4D 15N-edited NOESY spectra and were separated into strong, medium, and weak, corresponding to distance restraints with upper limits of 3.5, 5.0, and 6.0 Å, respectively. Hydrogen bonding partners were selected based on NOE patterns, and two distance restraints were used to simulate a hydrogen bond. NOE- and hydrogen bond-based distance restraints and backbone dihedral angle restraints were used as input for the simulated annealing protocol of CNS v. 1.0 with 2,000 high-temperature steps, 6,000 torsion angle dynamics cooling steps, and 5,000 Cartesian coordinate dynamics cooling steps. Two hundred and fifty structures were generated, and 20 lowest energy structures were selected for representing the structures for PagP-DPC and PagP-β-OG. The selected structures had a backbone RMSD for well-defined regions of 0.90 ± 0.19 and 0.80 ± 0.13 Å, respectively (20).

For OprH, distance constraints were calibrated using Cyana version 2.1, and backbone dihedral angle constraints were determined using TALOS. Hydrogen bond constraints derived from 1H/2H exchange experiments were set at 2.5 and 3.5 Å for HN–O and N–O distances, respectively. Structure calculations were performed using CNS version 1.2 with 4,000 high-temperature, 8,000 torsion slow-cool, and 8,000 Cartesian slow-cool annealing steps. A total of 200 structures were calculated, and the 20 lowest energy structures were selected for ensemble analysis. The mean global backbone RMSD over the 20 structures for β-sheet residues was 0.85 ± 0.20 Å (23).

Structure of OmpX was calculated in a similar way and was refined/recalculated by including additional hydrogen bonds and NOEs between the protons of the amide group and the methyl groups of I,L,V-methyl protonated and otherwise uniformly 2H-, 13C-, 15N-labeled OmpX (40), which resulted in significant improvement in the quality over the structures that were determined initially (30).

For high-resolution structure determination of a four-helix bundle membrane integrating protein Mistic (110 residues, 13 kDa) in LDAO micelles, long-range distance constraints, necessary for the determination of three-dimensional fold, were obtained from PREs by incorporating site-directed spin-labels individually at five different positions along the sequence of Mistic. The long-range distance constraints aided in the determination of a preliminary scaffold, which was refined through an iterative process by collecting long-range and medium-range NOEs and refining calibration of the spin-label restraints. The final structure calculation was performed with 573 NOE distance restraints, 346 angle restraints from chemical shifts and NOEs, and 478 distance restraints from the spin-label experiments (25).

The 3D structure of pSRII was calculated with the program ARIA27 interfaced to CNS on the basis of 5,564 distance restraints (4,055 unambiguous, 1,509 ambiguous), 190 pairs of backbone torsion angle restraints, and 132 hydrogen bond restraints. In this case, the large number of NOEs assigned is quite noteworthy. The final structures were well defined, with average pair-wise RMSD of 0.48 Å for the backbone residues 1–221 over 30 structures. The overall structure was in excellent agreement with the previously determined X-ray structures (14).

For calculation of PR structures, PRE was assigned into three categories of long-range distance restraints. The experimental RDCs, covering the range from −26.9 to −2.4 Hz, were used to determine the initial alignment tensor magnitude and rhombicity via the histogram method incorporated in CYANA, which were subsequently refined by fitting against the structure. CYANA structure calculations with torsion angle dynamics were performed with 200 conformers. The 20 conformers with the lowest target function values were superimposed with an RMSD to the mean coordinates of 0.81 Å within the transmembrane region. To validate the structure, complete sets of PRE restraints for each cysteine mutant were excluded one by one from the structure calculation. Comparison of [15N–1H]-correlation spectra was made with deuterated PR in deuterated diC7PC/DMPC bicelle which further established that the structure in micelle faithfully represented that in the bilayer membrane environment (14).

A novel approach has been used very recently for the determination of structure of mouse UCP2. This approach utilized the RDC-based MFR that had been validated for structure determination of ubiquitin using molecular fragments fit to RDCs. RDCs were measured from protein samples that were weakly aligned in a DNA nanotube liquid crystal as described above. The local and secondary structures of the protein were determined by constructing a database containing 320,000 seven-residue fragments from structures. For each seven-residue stretch along UCP2, the corresponding RDCs were fitted to all fragments using singular value decomposition incorporated in the program PALES (56). The candidate fragments were selected on the basis of quality of fit and were used to determine the local backbone structure by following a three-step protocol of fragment assignment, gap filling, and end extension. The RDCs also determined the relative orientation of the secondary structural segments. In addition to RDCs, semiquantitative distance restraints were obtained from PRE measurements, and these restraints provided the spatial arrangement of the transmembrane helices in the tertiary fold. Structures were calculated using XPLOR-NIH with the assigned structured segments, RDCs and PREs. The φ and ψ values of the segments were strongly enforced by a harmonic potential with force constant ramped from 10 to 1,000 kcal/mol/rad2. All RDCs used for determining the segments were applied, and the RDC force constant was ramped from 0.01 to 1.5 kcal/mol/Hz2. PRE restraints were enforced with flat-well harmonic potentials, with the force constant ramped from 1 to 40 kcal/mol/A2. In addition to experimental restraints, a weak database-derived ‘Rama’ potential function was ramped from 0.02 to 0.2 for the general treatment of side-chain rotamers. A total of 30 monomer structures were calculated using a simulated annealing protocol in which the bath temperature was cooled from 2,000 to 200 K. Fifteen low-energy structures were selected as the structural ensemble. The average backbone pair-wise RMSD for well-structured regions over 15 structures was 1.32 Å (9).

The structures determined by solution NMR spectroscopy of selected β-barrel and α-helical bundle membrane proteins are shown in Figs. 2 and 3, respectively. The experimental restraints and structural statistics for the final ensemble of structures for selected membrane proteins are given in Table 1.

Fig. 2.
figure 00172

Solution NMR structures of selected β-barrel membrane proteins: (a) OmpG, PDB ID: 2JQY ; (b) OprH, PDB ID: 2LHF; (c) outer membrane protein OmpA transmembrane domain, PDB ID: 1G90; (d) VDAC-1, PDB ID: 2K4T; (e) PagP, PDB ID: 1MM4.

Full size image
Fig. 3.
figure 00173

Solution NMR structures of selected α-helical bundle membrane proteins: (a) UCP2, PDB ID: 2LCK; (b) proteorhodopsin, PDB ID: 2L6X; (c) pSRII, PDB ID: 2KSY; (d) DsbB, PDB ID: 2K74; (e) DAGK, PDB ID: 2KDC.

Full size image
Table 1 Experimental restraints and structural statistics for the final ensemble of solution structures of selected \( \beta \)-barrel and \( \alpha \)-helical bundle membrane proteins determined with NMR spectroscopy
Full size table

3.3 Protein–Detergent Interactions Monitored by NMR Spectroscopy

The interaction between the membrane protein and the detergents in the surrounding micelle can be studied by a variety of NMR methods and is quite useful for structural validation. Exposure of backbone amides to solvent can simply be studied by solvent exchange experiments like the SEA-TROSY (57). Based on these experiments, the picture that usually emerges is that the detergent is “wrapped” as a belt around the hydrophobic core of the membrane protein with the loops and turns extending into the solvent. Internuclear NOEs between distinct group of protons of the detergent and the assigned backbone and side-chain protons of the protein in protein–detergent micelle can be measured directly or through designing suitable saturation transfer difference (STD) NMR experiments. In many cases, internuclear NOEs have been measured between the detergent and the protonated methyl groups of an I,L,V-methyl protonated 2H-, 13C-, 15N-labeled protein. Generally, intermolecular NOEs between the I,L,V-methyl groups and detergent alkyl groups are strongest for side chains in the hydrophobic core that are pointing outwards, while for regions exposed to the aqueous medium, like the extracellular loops and turns, such NOEs are weaker. This also suggests that the orientation of detergent molecules is in the form of an annular ring which is generally perpendicular to the long axis of the solubilized membrane protein (29).

The organization of detergent molecules around membrane proteins can also be studied by monitoring the differential relaxation effects of water-soluble and lipophilic paramagnetic relaxation agents on the resonances of the detergent and the protein. The paramagnetic relaxation agents that have been used are Gd(DOTA) (Gd3+ chelated with 1,4,7,10-tetraazocyclododecane-N,N′,N″,N′″-tetraaceticacid), 16-doxylstearc acid (16-DSA), and 5-doxylstearic acid (5-DSA). Gd(DOTA) is water soluble and does not penetrate into the micelle interior. On the other hand, 16-DSA and 5-DSA are incorporated into the detergent micelles. The latter two spin-labels can help to further differentiate between the residues in the central hydrophobic belt and those in the interfacial region. Such studies have been used for characterization of protein–detergent micelle as well as for validation of protein structure for several membrane proteins (8, 12, 14, 25, 57).

3.4 Conformational Transitions of Membrane Proteins in Detergent Micelles

Experimental measurement of molecular dynamics is a major advantage that NMR has over X-ray crystallography. Detailed NMR dynamics measurements are based on the analysis of amide 15N longitudinal (T1) and transverse (T2) relaxation times and heteronuclear nuclear Overhauser enhancement ({1H}15N-NOE) which can characterize motions of the protein on timescales ranging from picoseconds to milliseconds. The relaxation parameters measured at one or more fields can be analyzed using model-free formalisms to yield dynamics information on picosecond to millisecond timescales (58, 59). Further direct characterization of exchange processes, in which a protein locally exchanges between two or multiple states, can be carried out through relaxation–dispersion experiments. The relaxation– dispersion experiments yield kinetic rate constants for the exchange, and these can be coupled to functional aspects of the protein (59–61).

As a general feature, membrane protein dynamics are characterized by a relatively rigid transmembrane region combined with flexible loops and turns that connect the TM β-strands of α-helices. However, some finer dynamic elements have also been observed in many cases. For example, existence of a flexibility gradient was first determined from {1H}15N-NOE measurements on OmpA in DPC micelles (17). The central portion of the β-barrel was found to be very rigid, with flexibility increasing towards the two ends. Similar results were obtained for OmpX and PagP in detergent micelles, suggesting that this may be a common feature of membrane proteins.

Backbone amide 15N NMR dynamics measurements of OmpA have been performed at three magnetic fields. A total of nine relaxation data sets were globally analyzed using an extended model-free formalism. The diffusion tensor was found to be prolate axially symmetric with an axial ratio of 5.75, indicating that the protein rotated free from the micelle along its long axis, while along the perpendicular axis, it rotated along with the annular ring of the micelle. The generalized order parameters gradually decreased from the mid-plane towards the two ends of the barrel establishing the flexibility gradient described above. This protein dynamics gradient possibly counteracts the dynamics gradient of the lipids in a bilayer. The extracellular loops displayed large-scale internal motions on the ns timescale, and it was suggested that they were likely to undergo concerted (“sea anemone”-like) motions emanating from their anchoring points on the barrel (62).

For PagP, very interesting dynamics measurements have been made in detergent CYFOS-7. PagP is an 8-stranded β-barrel outer membrane enzyme that catalyzes the transfer of a palmitate chain from phospholipids to the lipid A moiety of lipopolysaccharide. The putative catalytic site of PagP is located on the highly mobile extracellular L1 loop that connects strands A and B. PagP is functionally inactive in LDAO, DPC, and β-OG, perhaps because the detergent molecules occupy the binding site and compete with the phospholipid substrate. PagP is active in CYFOS-7 which has a cyclohexyl ring at the end of its alkyl chain and therefore cannot penetrate into the β-barrel. Two distinct set of peaks for the amide residues are seen in the HSQC spectrum of PagP in CYFOS-7 detergent at 25 C. Based on the chemical shifts of the L1 loop region, the major conformer has been designated to represent the relaxed or “R” form, while the minor conformer has been assigned to represent the tense or “T” form. Structural differences between the two forms are localized mainly to the L1 loop, the β-bulge of strand A, and the adjacent region on strand H. The R → T and T → R conversion rates at 25 C has been found to be 2.8 ± 0.5/s and 6.5 ± 0.9/s, respectively, from 15NZZ exchange spectroscopy, based on which the proportion of the minor conformer was determined to be ~30%. The choice of detergent and temperature are both important for the observation of the minor component. Existence of the two conformers and interchange between them in CYFOS-7 are thought to reflect the functioning of this protein in its native membrane environment with the R form facilitating substrate entry and the T form facilitating substrate catalysis (63).

Role of local dynamics in the channel function (13) of the voltage-gated potassium channel from Streptomyces lividans (KcsA) has been studied by dynamics measurements in DPC micelles at pH 4 (open state) and pH 7 (closed state), which roughly corresponds to the pH-dependent gating. The functional form of KcsA is a tetramer. Each monomer features two transmembrane helices and an additional, shorter helix in the transmembrane region that forms a selectivity filter for ions, which in terms of size corresponds to the hydration shell of K+ ions in water. Several recent dynamics studies have indicated that two or more conformational states can exist for the selectivity filter and the C-terminal of second transmembrane helix, with rapid interconversion between the two states in ion-conducting open state of the channel. The structural changes observed at different pH values are correlated with the pH-dependent open probability profile of the channel measured by functional tests.

The membrane protein dynamics of the seven transmembrane helical protein pSRII have also been investigated. Amide 15N relaxation rate constants indicate that the helices of this protein are well ordered and pack to a regular bundle. Some of the connecting loop regions are highly flexible and experience motions on the picosecond-to-nanosecond timescale. In the protein core, a limited amount of motion on the microsecond-to-millisecond timescale was observed (12).

4 Conclusion

The belief in high potential of NMR spectroscopy for structural and functional characterization of membrane proteins is being consolidated with every newer and more challenging problem that is being addressed by the courageous investigators in this field. Advanced expression and labelling methods, methyl-protonation of I, L, V, use of RDC and PRE, molecular fragment searches, use of SAXS and SANS data, and functional measurements are likely to become the norm for studying membrane proteins with this technique. The value of NMR spectroscopy would lie in high-fidelity correlation of protein dynamics and mapping of structural interactions to obtain mechanistic information for hundreds of functionally relevant membrane proteins that are being discovered continuously. As all aspects associated with NMR spectroscopy, like the high-field magnets, newer pulse techniques, high-sensitivity multichannel cryogenically cooled probes, and labelling strategies, are undergoing spontaneous development, the technique, therefore, holds a great promise for studying membrane proteins of far greater complexity in the future.