Membrane proteins in their native habitat as seen by solid-state NMR spectroscopy

Abstract

Membrane proteins play many critical roles in cells, mediating flow of material and information across cell membranes. They have evolved to perform these functions in the environment of a cell membrane, whose physicochemical properties are often different from those of common cell membrane mimetics used for structure determination. As a result, membrane proteins are difficult to study by traditional methods of structural biology, and they are significantly underrepresented in the protein structure databank. Solid-state Nuclear Magnetic Resonance (SSNMR) has long been considered as an attractive alternative because it allows for studies of membrane proteins in both native-like membranes composed of synthetic lipids and in cell membranes. Over the past decade, SSNMR has been rapidly developing into a major structural method, and a growing number of membrane protein structures obtained by this technique highlights its potential. Here we discuss membrane protein sample requirements, review recent progress in SSNMR methodologies, and describe recent advances in characterizing membrane proteins in the environment of a cellular membrane.

Introduction

Membrane proteins play many important roles in biology of the cell, from regulation of the membrane transport to signal transduction, both between the extracellular and intracellular media and between different intracellular compartments. Although it is generally accepted that their structure and function are sensitive to the environment, the majority of high-resolution structural methods require membrane proteins to be prepared in the form either completely devoid of, or containing a small amount of lipids. Amongst high-resolution structure determination methods, X-ray crystallography has been the most prolific to date, providing the most detailed view of molecular structure at the same time. Since the first crystallographic structure of myoglobin,¹ ∼90,000 protein structures have been determined by this method out of a total of more than 100,000 structures, and impressive progress has been achieved in crystallography of membrane proteins.^2–5 Solution NMR approaches have also continued to mature,^6–9 and provided a few structures of polytopic α-helical membrane proteins in detergent micelles.^10–14 Despite this impressive progress in membrane protein structure determination, sample preparation remains the main challenge for both X-ray crystallography and solution NMR, and the number of membrane protein structures in the protein data bank is less than 2% of the total number of structures.

Membrane proteins have evolved to perform their functions in the environment of a cell membrane. Being able to study them in the native-like environment of the lipid bilayer ensures that the sample represents a biologically relevant protein conformation, that is, as close as possible to the one present in living cells. Solid-state NMR (SSNMR) has a long history of applications to the studies of structure and dynamics of membrane proteins in phospholipids. While earlier studies focused primarily on samples with selectively isotopically labeled sites, or small ligands and cofactors,^15–18 the emergence of high-field magnets, new probe technologies,^19,20 methods for sample preparation,²¹ and new SSNMR methodologies over the past decade have greatly expanded the range of membrane proteins amenable to structural investigations, and allowed detailed studies of the samples with uniform incorporation of isotopic ¹³C and ¹⁵N labels. Similar to solution NMR, multiple individual sites can now be resolved and uniquely assigned in the solid-state, paving the way for in-depth structural, dynamic and functional analyses. The potential of SSNMR has initially been demonstrated on soluble model proteins prepared in a microcrystalline state, such as Bovine Pancreatic Trypsin Inhibitor (BPTI),²² α-spectrin SH3 domain,^23,24 ubiquitin,^25,26 the β1 immunoglobulin binding domain of protein G (GB1),²⁷ and thioredoxin,²⁸ and recently has been expanded to amyloid fibril forming proteins,^29–32 which are not readily accessible to X-ray crystallography and solution NMR. SSNMR has also been successfully used to determine structures of several membrane proteins including influenza A M2 proton channel³³ and its complex with the drug amantadine,³⁴ phospholamban,³⁵ mycobacterial cell division regulatory protein CrgA,³⁶ transmembrane domain of Yersinia enterocolitica adhesin A (YadA),³⁷ light-sensitive receptor Anabaena Sensory Rhodopsin (ASR)³⁸ and G-protein coupled chemokine CXCR1 receptor.³⁹ In addition, X-ray structures of disulfide bond formation protein B (DsbB),⁴⁰ and its complex with DsbA⁴¹ have been refined to higher resolution by combining electron densities obtained from X-ray diffraction and solid-state NMR restraints. SSNMR offers even more exciting possibility of studying membrane protein structure in situ, either in cell membranes or even in whole cells.^42–48 The main focus of this review is on the recent progress in solid-state NMR membrane protein structure determination, primarily concentrating on the magic angle spinning solid-state NMR, as well as on the prospects of studying membrane proteins in cell membrane environment.

Protein Expression and Isotope Labeling

Preparation of samples of stable functional membrane proteins in sufficient quantities is the main challenge for all structural methods. For solid-state NMR studies, one must recombinantly produce a few milligrams of isotopically ¹³C,¹⁵N-enriched (and, sometimes, deuterated) protein. The most frequently used are E. coli based heterologous expression systems, as they offer flexibility in the choice of constructs and labeling strategies, and are relatively inexpensive. Much effort has been directed at the development of the bacterial expression, refolding and functional purification protocols for the production of eukaryotic membrane proteins for NMR, most notably G-protein coupled receptors, including neuropeptide Y2 receptor,⁴⁹ human cannabinoid Type 2 receptor,⁵⁰ and human chemokine CXCR1 receptor.⁵¹

The use of lower eukaryotes such as Pichia pastoris is another attractive route for the production of isotopically labeled eukaryotic proteins. The utility of this approach for membrane proteins has been demonstrated using eukaryotic (fungal) rhodopsin from Leptosphaeria maculans,⁵² and human water channel aquaporin 1.⁵³ Both proteins gave solid-state NMR spectra of excellent quality, suitable for high-resolution structural and dynamic characterization. On the other hand, Pichia uses methanol as the main carbon source, and offers a limited flexibility in the choice of ¹³C-labeling schemes, e.g., not allowing sparse ¹³C labeling (e.g., by using 2-¹³C or 1,3-¹³C labeled glycerol, 1-¹³C, 2-¹³C labeled glucose, etc., as carbon sources),^54–57 which is easily achieved in bacterial E.coli based systems. Production of sparsely labeled samples is essential part of SSNMR methodology, as it is widely used in SSNMR to both improve spectral resolution for spectroscopic assignments,^58,59 and to facilitate distance measurements for structure determination.²⁴ Other species of yeast, such as Kluyveromyces lactis, which can use sugars as carbon source, are being explored to correct this deficiency.^60,61

More advanced eukaryotic expression systems, for example, transfected mammalian cells and baculovirus-infected insect cells, have their own challenges such as relatively low expression yields, high cost of production, incomplete labeling, as well as lack of reliable deuteration and uniform labeling protocols, even though these parameters keep improving.^62,63 Nevertheless, there are a number of successful SSNMR studies of selectively labeled GPCRs (most notably, visual rhodopsin) produced in these systems.^64,65 New eukaryotic expression systems are being developed. For example, a recent study demonstrated successful isotope labeling of bovine visual rhodopsin in a roundworm Caenorhabditis elegans.⁶⁶ One should also mention significant progress in the cell-free expression methods, which have been used to produce isotopically labeled membrane proteins^14,67–69 with several applications in solid-state NMR.^70,71

Solid-State NMR Approaches and Sample Formulation

The term “solid-state NMR” implies application of NMR to molecular systems that lack significant motion. In contrast to solution NMR, which requires rapid tumbling of molecules, and is therefore limited by size of the molecule, SSNMR can be applied to protein and/or protein complexes of arbitrary molecular weight. Because of lack of fast molecular tumbling SSNMR spectra are broadened by anisotropic dipolar interactions and chemical shift anisotropies, which need to be removed for site-specific resolution. There are two principal approaches to accomplish this, either by uniformly orienting molecules relative to a magnetic field (oriented sample NMR, OS-NMR), or by applying magic angle spinning (MAS) technique (Fig. 1). The OS-NMR approach relies on the uniaxial alignment of the sample, and the strong angular dependence of the ¹⁵N chemical shift anisotropy and ¹⁵N-¹H dipolar couplings on the molecular orientation with respect to the magnetic field is used to determine the orientation of individual sites relative to the common axis. Most commonly, amino acid selectively ¹⁵N-labeled proteins, either incorporated into the bilayers mechanically aligned on glass plates or reconstituted in bicelles, are used to acquire NMR spectra correlating ¹⁵N-¹H dipolar couplings with ¹⁵N chemical shift. In addition, magnetically aligned bicelles^72,73 provide high quality of alignment, and they have been used to study membrane proteins, including CXCR1,⁷⁴ OmpX,⁷⁵ and cytochrome b_5.⁷⁶

Figure 1.

Solid-state NMR approaches to membrane protein structure determination. (A) Protein molecules can be uniformly oriented with respect to the magnetic field B_o NMR spectra correlating the ¹H-¹⁵N dipolar coupling to the ¹⁵N chemical shift show a circular pattern, which reports on the topology of the protein andits backbone structure. An example on the left is a spectrum of the transmembrane domain of virus protein “u” comprising residues 1–32.⁸⁰ Reprinted from Journal of Molecular Biology, 333, S.H. Park, A.A. Mrse, A.A. Nevzorov M. F. Mesleh, M. Oblatt-Montal, M. Montal and S.J. Opella, “Three-dimensional Structure of the Channel-forming Trans-membrane Domain of Virus Protein “u” (Vpu) from HIV-1” (2003) 409–424. Copyright 2003, with permission from Elsevier. (B) Magic angle spinning (MAS) NMR is applied to unoriented samples. MAS averages out anisotropic interactions to their isotropic values, and results in spectra that resemble those from samples in solution phase. An example on the right is a 2D NCA spectrum of Anabaena Sensory Rhodopsin correlating isotropic chemical shifts of ¹⁵N and ¹³C_α atoms.⁸¹ Reproduced with permission from Angewandte Chemie Int Ed., ‘‘Conformation of a seven-helical transmembrane photosensor n the lipid environment’’ 50 (2011) 1302–1305. Copyright 2011 John Wiley & Sons, Inc. Additionally, in the case when a protein undergoes fast (<10⁻⁶ s) rotational diffusion, correlating the ¹H-¹⁵N dipolar couplings to the ¹⁵N chemical shift would result in circular patterns similar to that shown on the left (Rotational Alignment), and can be used as an independent way to obtain backbone structure, or combined with distance measurements under MAS conditions.

In oriented samples, the topology and structure of a membrane protein can be determined from the circular spectral patterns, called PISA wheels, or polarity index slant angle (Fig. 1),^77,78 which are directly related to the periodicity of an α-helix and to its orientation with respect to the external field (and, consequently, to the membrane plane). The majority of solid-state NMR structures of membrane proteins available so far have been determined using this approach (http://www.drorlist.com/nmr/MPNMR.html). While the initial applications of OS-NMR have been mostly limited to small proteins and peptides because of the spectral complexity and low sensitivity of heteronuclear ¹⁵N detection, very promising SSNMR data for larger proteins have been published in recent years, including well-resolved spectra of the multidrug resistance transporter EmrE⁷⁹ and a structure of cell division regulatory protein CrgA.³⁶

Magic Angle Spinning (MAS) refers to a technique in which unoriented samples are spun around an axis oriented at 54.7° with respect to the external magnetic field (Fig. 1). MAS coherently averages anisotropic dipolar couplings and chemical shift anisotropies to their isotropic values. A typical carbon or nitrogen MAS spectrum of extensively isotopically labeled membrane protein contains hundreds to thousands of lines, and the spectral widths of individual lines become the most important parameter that essentially determines whether a protein is amenable to high-resolution structural analysis. The spectral resolution and the line widths depend on sample homogeneity, which in turn is strongly affected by the sample preparation. Even slight structural heterogeneity may cause large inhomogeneous line broadening, and render high-resolution structure determination impossible.

One of the major advantages of MAS solid-state NMR is that it allows examining many sample forms ranging from membrane proteins in a precipitated and/or microcrystalline state, to proteins reconstituted in model membranes, or even in more complex and chemically heterogeneous cellular membranes. In Figure 2 we show a few selected examples of typical two-dimensional carbon-carbon correlation spectra of membrane proteins prepared under different conditions. They all have comparably high resolution with tens of cross peaks resolved, indicating high level of structural homogeneity, while some difference in the appearance can be attributed to different methods employed to obtain spectral correlations.

Figure 2.

Magic angle spinning carbon-carbon correlation spectra of various membrane proteins. (A) 900 MHz spectrum of trimeric YadA.⁸² Reprinted by permission from Macmillan Publishers Ltd: Scientific Reports, 2, S.A. Shahid, S. Markovic, D. Linke, B.J. van Rossum, “Assignment and secondary structure of the YadA membrane protein by solid-state MAS NMR”, (2012) 803. Copyright 2012. (B) 800 MHz spectrum of chimeric potassium channel KcsA-Kv1.3.⁸⁵ Adapted with permission from J. Am. Chem. Soc., 130, R. Schneider, C. Ader, A. Lange, K. Giller, S. Hornig, O. Pongs, S. Becker, M. Baldus, “Solid-state NMR spectroscopy applied to a chimeric potassium channel in lipid bilayers” (2008) 7427–7435. Copyright 2008 American Chemical Society. (C) 750 MHz spectrum of ⁴⁰DsbB. Reprinted from Journal of Molecular Biology, 425, M. Tang, A.E. Nesbitt, L.J. Sperling, D.A. Berthold, C.D. Schwieters, R.B. Gennis, C.M. Rienstra, “Structure of the disulfide bond generating membrane protein DsbB in the lipid bilayer” (2013) 1670–1682. Copyright 2013, with permission from Elsevier. (D) 800 MHz spectrum of proteorhodopsin (unpublished data). (E) 900 MHz spectrum of VDAC1 sample prepared as 2D crystals.⁸⁶ Adapted with permission from J. Am. Chem. Soc., 134, M.T. Eddy, T.C. Ong, L. Clark, O. Tejido, P.C.van der Wel, R. Garces, G. Wagner, T.K. Rostovtseva, R.G. Griffin, “Lipid dynamics and protein-lipid interactions in 2D crystals formed with the ß-barrel integral membrane protein VDAC1” Copyright 2012 American Chemical Society. (F) 800 MHz spectrum of water channel human aquaporin 1.⁵³ Reprinted from Journal of Biomolecular NMR, S. Emami, Y. Fan, R. Munro, V. Ladizhansky, L.S. Brown, “Yeast-expressed human membrane protein aquaporin-1 yields excellent resolution of solid-state MAS NMR spectra” 55, (2013) 147155. Reproduced with kind permission from Springer Science and Business Media.

An important question in determining the sample preparation strategy is to what degree the resolution of SSNMR spectra is determined by the local order. For example, samples of YadA⁸² [Fig. 2(A)] and DsbB⁸³ [Fig. 2(B)] proteins were prepared in precipitated/microcrystalline states. While small, poorly diffracting microcrystals are considerably easier to prepare than the high-quality crystals suitable for X-ray crystallographic studies, the resolution of NMR spectra is comparable between the two, as was demonstrated on polycrystalline and nanocrystalline ubiquitin.⁸⁴ Dense packing achieved in crystals maximizes the amount of protein that can be packed in the active volume of the MAS NMR rotor, and offers a significant sensitivity advantage. Although proteins in small crystals are likely to be prone to the artifacts resulting from crystal contacts, restricted dynamics, etc., same as in large crystals used for protein crystallography, a variety of biochemical parameters can be varied with relative ease, allowing for the studies of protein function. Both structural and functional information can thus be deduced from such samples even in cases when the production of large crystals for X-ray analysis is problematic.

The environment of the lipid bilayer allows for a variety of native-like biochemical and biophysical conditions to be applied, including protein-to-lipid ratio, pH, temperature, bilayer fluidity and hydrophobic thickness, and is often the preparation of choice for SSNMR studies. Early publications argued that 2D crystalline preparation promotes local conformational homogeneity,⁸⁷ and indeed a few proteins prepared in the form of 2D crystals yielded spectra of excellent quality, such as those of BR,⁸⁸ OmpG,⁸⁷ ASR,⁸¹ VDAC1⁸⁶ [Fig. 2(E)], and human aquaporin 1⁵³ [Fig. 2(F)]. It is important however for the 2D crystals to be macroscopically homogeneous. For example, 7TM light-driven proton pump proteorhodopsin (PR) forms 2D crystals of different symmetries under various lipid reconstitution conditions.^89–91 In particular, PR embedded in the DOPC lipid matrix at high protein-to-lipid ratio (w/w 4:1) forms 2D assemblies of crystalline domains of hexagonal (and occasionally pentagonal) symmetry, which coexist with non-crystalline densely packed domains.⁹¹ While SSNMR samples prepared under these conditions yielded spectra of acceptable quality,⁹² even better resolution was obtained in non-crystalline PR samples produced under alternative reconstitution protocol, which used the DMPC/DMPA lipid mixture (9:1 w/w), and a lower protein-to-lipid ratio (2:1 w/w) [Fig. 2(D)].⁹³ Thus, homogeneity of 2D crystal formation, as well as the chemistry of lipids, protein local mobility, and uniform state of oligomerization are also important factors for spectral resolution.

Overall, MAS SSNMR has yielded well-resolved spectra for a large number of membrane proteins of different classes, secondary structure and architecture. While samples need to be optimized individually for each protein, the examples given above demonstrate that the preparation of structurally homogeneous samples does not appear to be a limitation for the structural characterization of membrane proteins by solid-state NMR, at least in the cases where intrinsic conformational heterogeneity is not present or can be minimized.

While traditional applications of MAS SSNMR to membrane proteins mainly relied on carbon detection in ¹⁵N, ¹³C-labeled samples, there is a growing number of studies where proton detection in back-H/D-exchanged perdeuterated samples is employed.^94–97 Proton detection is attractive because it provides both the improvement in resolution, and higher sensitivity due to the large gyromagnetic ratio of protons (four times larger than that of carbons). The utility of this approach has principally been demonstrated in the initial applications to α-spectrin SH3 domain, where excellent spectral resolution rivaling that of solution NMR was obtained.^98,99 Although excellent resolution was also achieved in the first application to a membrane protein bacteriorhodopsin, only about a third of the expected resonances could be observed because of the incomplete back-exchange.⁹⁴ Similarly, only solvent-accessible sites could be observed in another 7TM membrane protein proteorhodopsin.⁹⁵ Thus, application of proton detection in membrane proteins would require the use of unfolding-refolding protocols, or alternative methods for the introduction of the detectable protons in the perdeuterated protein.^100,101 Overall, the use of proton detection is an attractive route and will continue to be the area of active development. Not only it gives access to proton chemical shifts, which are sensitive probes of protein structure and interactions, but it also permits using approximately an order of magnitude smaller amounts of sample, which is important for hard-to-express proteins.

Structure determination based on distance restraints

MAS solid-state NMR structure determination strategies follow in the footsteps of solution NMR spectroscopy. The majority of studies of membrane proteins to date used carbon detection, and site-specific assignments can be obtained using established suites of ¹³C-detected 3D or 4D chemical shift correlation experiments^83,102,103 assisted by simplifying isotope labeling schemes.^58,59,104 The analysis of experimental data and extraction of chemical shifts is assisted by the available software tools, which reduce the possibility of error.^105–107 Chemical shifts of individual residues can be used to determine the backbone structure using either the established correlations between chemical shifts and secondary structure,¹⁰⁸ or an empirical relation between the chemical shifts and backbone dihedral angles.¹⁰⁹ Additionally, torsion angles defining relative orientation of chemical bonds can be determined from correlation experiments, which relate relative orientations of dipolar and/or chemical shift anisotropy interactions, which are fixed with respect to the protein backbone.^110,111

Structure determination in the solid state is based on determining a large number of structurally constraining internuclear distances (typically up to ∼6 Å between carbon atoms, or between carbon and nitrogen atoms), which are normally obtained from the dipolar recoupling experiments combined with multidimensional chemical shift correlation experiments for optimal spectral resolution.^24,112,113 Such approach turned out to be useful even for proteins of relatively large size, such as 7TM alpha-helical microbial rhodopsin ASR. Long-range distance restraints (approximately up to 20 Å) can also been obtained in the form of Paramagnetic Relaxation Enhancements (PRE)^114,115 from samples with paramagnetic tags attached to cysteines introduced through site-directed mutagenesis. Owing to their longer-range order, PREs are ideally suited for the characterization of protein−protein interactions, and have been used to determine the oligomerization interface of ASR, resulting in a refined SSNMR structure of its 81 kDa trimer.¹¹⁶

Structure determination using rotational alignment

As pointed out above, the term “solid-state NMR” generally applies to the systems that lack rapid overall tumbling. There are, however, other types of global motions that may have a profound effect on the NMR spectra. In particular, it has been observed in the pioneering experiment in BR that rapid submicrosecond rotational diffusion in the bilayer results in partial averaging of dipolar couplings and chemical shift anisotropies, and produces NMR line shapes that are dependent on the orientation of an internuclear vector with respect to the magnetic field axis.¹¹⁷ Based on this observation, Opella and co-workers have proposed a method for membrane protein structure determination termed Rotational Alignment (RA), which relies on MAS¹¹⁸ performed on samples of membrane proteins that undergo rotational diffusion. While OS-NMR is primarily restricted to low sensitivity nitrogen detection, RA MAS NMR can be combined with higher sensitivity of carbon detection, and can be further combined with more general (e.g., not requiring rapid rotational diffusion) MAS approaches discussed above.

RA relies on fast rotational diffusion of a protein about the bilayer normal, and this imposes some restrictions on the sample condition. Any of the factors that limit the rate of diffusion, for example, large molecular weight and/or oligomerization state, interactions with other proteins, fluidity of the lipid bilayer, and so forth, will render the RA approach inapplicable. In the aforementioned study of BR,¹¹⁷ the protein had to be reconstituted in DMPC lipids in the fluid phase, where it exists as a monomer,¹¹⁹ while the rotational diffusion of BR in the purple membrane was much slower, likely limited by interprotein contacts in a trimer, and between trimers within the hexagonal lattice. Still, many important classes of membrane proteins fall into the regime of fast rotational diffusion, as was demonstrated on the monomeric G-protein coupled chemokine CXCR1 receptor, whose structure was determined using the RA technique.³⁹

Solid-state NMR of membrane proteins in cellular membranes

While artificial phospholipid bilayer represents a good initial approximation of a realistic cell membrane, true cell membranes are much more complex and contain many more components, e.g., other integral and peripheral membrane proteins, various lipids, carbohydrates, etc., which can transiently or stably interact with a protein of interest and affect its structure. Solid-state NMR offers a unique capability to examine membrane protein structure site-specifically in the environment of a cell membrane. Principal limitations of SSNMR studies of proteins in cellular membranes are as follows: (i) limited content of the protein of interest and reduced relative and absolute amplitude of its NMR signals; (ii) abundance of other proteins in the membrane results in high spectral overlap; (iii) highly heterogeneous environment may result in structural heterogeneity and inhomogeneous line broadening of the spectra, further reducing both amplitude and resolution.

For a long time, bacteriorhodopsin had been the only protein subjected to detailed structural investigations in its native membrane.^120–123 BR, however, represents a special case because it is localized in the 2D crystalline purple membrane (PM) and is highly abundant, constituting ∼75% of the PM content by weight. Several recent studies have considerably expanded the list of peptides and proteins studied in cell membranes, and demonstrated the feasibility of obtaining structural data in cell membranes or even in whole cells.^42–48 The first case of in situ SSNMR was presented by Tian et al. who have studied a TM peptide LR11 in E. coli.⁴² LR11 interacts with Amyloid Precursor Protein and regulates its localization in cells. LR11 was expressed as MBP fusion construct, cleaved in the E. coli membrane, and constituted ∼80% of the total labeled protein in the sample. This greatly facilitated SSNMR assignments of ∼50% of residues of the protein and allowed characterization of its secondary structure.

Low concentration of protein of interest in the “in-membrane” samples currently precludes the de novo structure determination. NMR can however be used to detect chemical shift perturbations and use them for comparison with the respective structures obtained in the liposome environment. Such an approach was recently used to validate the structure of the influenza A M2 proton channel which had been previously obtained in proteoliposomes.³³ From the direct comparison of chemical shifts of the conducting part of the channel it was concluded that the channel conformation in the E. coli membrane is very similar to that in synthetic liposomes.⁴⁵

The presence of background proteins and the associated spectral crowdedness constitutes another significant complication for in-cell and in-membrane SSNMR. To reduce the background signal, Baldus et al. have implemented a new sample preparation strategy, which enabled a study of a much larger 150-residue outer membrane enzyme PagL, which is responsible for the removal of fatty acid chains from lipopolysaccharides in Pseudomonas aeruginosa.⁴⁷ PagL was expressed in the E. coli strain lacking genes for the two most abundant proteins of the outer membrane, OmpA and OmpF. They further used a labeling protocol, in which the target protein was preferentially labeled, thus reducing the overall background signal in the NMR spectra (Fig. 3).

Figure 3.

Sample preparation strategy for cellular NMR of PagL employed in Ref. 47. A: Schematic representation of the E. coli K-12 cell envelope. B: Preparation of whole cell, cell envelope, and proteoliposome samples. In the initial step, cells are first grown on the unlabeled medium, and then transferred to isotopically labeled medium for induction. Adapted with permission from Ref. 47, The National Academy of Sciences of the USA, 2012.

These improvements have allowed them to simplify spectra and to enhance the relative signal from PagL protein to a degree sufficient to transfer spectroscopic assignments obtained for PagL in proteoliposomes. Of the 13 residues that could be observed and reassigned in PagL in the environments of both cell envelopes and in whole cells, the majority belonged to the extramembranous water-exposed fragments, which were found to be in conformations similar to those in the proteoliposome environment (Fig. 4).

Figure 4.

¹³C-¹³C correlation spectra of whole cell (A), cell envelope (B) and proteoliposome (C) samples containing PagL. Encircled regions demonstrate characteristic alanines, serines and threonines of PagL. Adapted with permission from Ref. (47), The National Academy of Sciences of the USA, 2012.

One of the important observations in the aforementioned study was surprisingly high resolution of PagL spectra, despite the apparent heterogeneity of the environment. However, because of the presence of the background E. coli proteins and associated spectral complexity only a handful of resonances belonging to PagL could be identified.

In our investigations of ASR in E. coli membrane, we have implemented additional purification steps commonly used in membrane biochemistry, aimed at increasing the relative content of ASR (Fig. 5). As ASR preferentially partitions into the inner membrane, we applied sucrose gradient to isolate the ASR-containing inner membrane from the outer membrane,¹²⁴ and to increase the relative content of the protein in the NMR sample. Further sample simplification was achieved by using metal affinity two-phase system,¹²⁵ which allowed isolation of ASR-containing vesicles through preferential binding of the His-tagged ASR to the Ni²⁺ resin. This led to increase of the relative content of the protein of interest to ∼10% of the total protein, and was sufficient to record three-dimensional chemical shift correlation experiments, and transfer spectroscopic assignments previously obtained in proteoliposomes for 40% of residues, located both in the TM and solvent-exposed regions.⁴⁸

Figure 5.

Sample preparation of ASR in E. coli membranes. Following growth in unlabeled media to the logarithmic phase (Step 1), cells are resuspended in isotopically enriched media at a high concentration, and protein expression is induced (Step 2).⁴³ Cells are broken in Step 3, subjected to sucrose gradient to separate inner and outer membranes (Step 4)¹²⁴ and to 2 phase purification system to isolate membrane fractions containing the His-tagged ASR (Step 5).¹²⁵ Proteoliposomes are represented as green circles, ASR is shown as red insertions.

The detected chemical shifts of ASR in the E.coli membrane match very well those measured in the proteoliposomes, indicating, overall, a rigid structure of ASR. ASR remains trimeric in the cellular membrane, which is evident from the minimal perturbations of chemical shifts of residues at the intermonomer interface, as well as from complementary visible range CD spectra, which show a characteristic bilobe shape consistent with the trimer structure.⁴⁸ Similarly, the all-trans conformation of retinal as well as the conformation of the retinal binding pocket are well conserved. At the same time, we could identify a number of small changes in chemical environment and mobility of certain regions, showing a subtle adaptation to a different environment of the E. coli membrane (Fig. 6).

Figure 6.

ASR in E. coli membrane. A: 16 peaks could be assigned from comparison of 2D NCA spectra of the proteoliposome and E. coli sample (the latter is shown in A). B−D: Much higher resolution was obtained using 3D CANCO and NCACB spectroscopy, which permitted assignments for ∼40% of the protein residues. Red and black contours in the representative 2D planes represent peaks detected in the proteoliposome and E. coli sample, respectively. Blue contours are from the control sample containing the background proteins, but not ASR. E,F: Structural conservation in ASR in E. coli vs. proteoliposomes. In E, residues surrounding retinal are well conserved and shown in green. F: Residues at the intermonomer interface are well conserved (shown in green). Small chemical shift perturbations (up to 1 ppm for ¹⁵N and up to 0.5 ppm for ¹³C) are detected in some of the residues in the helices forming oligomerization interface but facing interior of the protein or lipids (shown in yellow).⁴⁸ Reprinted from Biophysical Journal, 108, 7, 1683–1696, M. Ward, S. Wang, R. Munro, E. Ritz, I. Hung, P. Gor'kov, Y. Jiang, H. Liang, L. Brown, V. Ladizhansky, “In situ Structural Studies of Anabaena Sensory Rhodopsin in the E. coli Membrane”, with permission from Elsevier.

Current status and future prospects of SSNMR structure determination of membrane proteins

The research described in this review illustrates the potential of solid-state NMR approaches for the characterization of membrane protein structure. High spectral resolution has been demonstrated for several classes of membrane proteins, and currently available sensitivity is already sufficient for providing detailed insights into structure and function of proteins and protein complexes with molecular weights of tens of kilodaltons. So far, SSNMR has resulted in about 70 SSNMR structures including those of large oligomeric membrane protein assemblies, including those of influenza A M2 proton channel,³³ phospholamban,³⁵ Yersinia adhesin A (YadA),³⁷ Anabaena Sensory Rhodopsin,³⁸ the chemokine receptor CXCR1,³⁹ DsbB, and of DsbB-DsbA complex^41,40 (Fig. 7).

Figure 7.

Selected structures of membrane proteins determined or refined using solid-state NMR restraints. Protein name, molecular weight, lipid mimetic used, method and PDB code are given for each structure. Different colors are used to differentiate individual subunits.

Most importantly, the initial studies of membrane proteins in the cellular environments discussed above indicate that the resolution detected under cellular conditions is comparable to that in proteoliposomes, while application of three-dimensional spectroscopy provides additional resolution sufficient to resolve many resonances. Sensitivity remains the most critical practical limitation. Further success of SSNMR strongly depends on progress in the development of high-field magnets, probe technologies, and proton detection methods. The introduction of ultra-high magnetic fields and cryogenic MAS probes, and the associated sensitivity and resolution improvement will allow an expansion of the list of molecular systems amenable to structural analysis by SSNMR.

An important development in this area is the introduction of Dynamic Nuclear Polarization (DNP) MAS NMR at high magnetic fields, which provides enhancements of SSNMR signals in the range of 10−60 fold. DNP makes use of paramagnetic agents, whose electronic spin polarization is transferred to nuclear spins in the NMR sample, resulting in enhanced NMR signal amplitudes. Pioneering DNP studies on BR allowed observing protein conformational substates not accessible to traditional SSNMR,^121,122 and a number of other membrane proteins are being studied by the technique^46,126 While this new technique is still under development, and search for new paramagnetic agents and better methods is ongoing,^127–130 the most serious limitation for application of DNP for detailed structural studies of membrane proteins is a significant loss of spectral resolution, which complicates DNP studies of uniformly labeled samples. Nevertheless, it promises to be very useful with selectively labeled samples, and may be applied not only to purified proteins with low expression yields but also to in-cell and in-membrane samples.^46,131

Glossary

ASR

Anabaena Sensory Rhodopsin

BPTI

Bovine Pancreatic Trypsin Inhibitor

BR

bacteriorhodopsin

CrgA

mycobacterial cell division regulatory protein

CD

circular dichroism

DNP

Dynamic Nuclear Polarization

DMPA

1,2-dimyristoyl-sn-glycero-3-phosphate

DMPC

1,2-dimyristoyl-sn-glycero-3-phosphocholine

DsbA

Disulfide bond formation protein A

DsbB

Disulfide bond formation protein B

MAS

magic angle spinning

NMR

nuclear magnetic resonance

OmpA

outer membrane protein A

OmpF

outer membrane protein F

OmpG

outer membrane protein G

OmpX

outer membrane protein X

OS-NMR

oriented sample NMR

PISA

polarity index slant angle

PR

proteorhodopsin

PM

purple membrane

PRE

paramagnetic relaxation enhancement

RA

rotational alignment

SSNMR

solid-state NMR

VDAC1

voltage-dependent anion channel 1

YadA

Yersinia enterocolitica adhesin A.