Determination of Structural Topology of a Membrane Protein in Lipid -Bilayers using Polarization Optimized Experiments (POE) for Static and MAS Solid State NMR Spectroscopy

Abstract

The low sensitivity inherent to both the static and magic angle spinning techniques of solid-state NMR (ssNMR) spectroscopy has thus far limited the routine application of multidimensional experiments to determine the structure of membrane proteins in lipid bilayers. Here, we demonstrate the advantage of using a recently developed class of experiments, polarization optimized experiments (POE), for both static and MAS spectroscopy to achieve higher sensitivity and substantial time-savings for 2D and 3D experiments. We used sarcolipin, a single pass membrane protein, reconstituted in oriented bicelles (for oriented ssNMR) and multilamellar vesicles (for MAS ssNMR) as a benchmark. The restraints derived by these experiments are then combined into a hybrid energy function to allow simultaneous determination of structure and topology. The resulting structural ensemble converged to a helical conformation with a backbone RMSD ∼ 0.44 Å, a tilt angle of 24° ± 1°, and an azimuthal angle of 55° ± 6°. This work represents a crucial first step toward obtaining high-resolution structures of large membrane proteins using combined multidimensional O-ssNMR and MAS-ssNMR.

Introduction

Membrane protein function is modulated by lipid membranes (Tamm 2005; White 2009). Therefore, understanding the structure-function relationship of these proteins requires their characterization in fully hydrated lipid bilayers. Multidimensional solid-state NMR, i.e. oriented solid-state NMR (O-ssNMR) and magic angle spinning solid-state NMR (MAS-ssNMR), allows for such structural characterization in the presence of fluid lipid bilayers – the most faithful mimetic of natural membranes. O-ssNMR provides information on dipolar couplings and anisotropic chemical shifts that enable the determination of topological parameters, such as the tilt and rotation angles of transmembrane and membrane anchored domains (Veglia et al. 2012). On the other hand, MAS-ssNMR provides chemical shift information on both main chain and side chains that can be converted into torsion angles for assessing secondary structure elements. MAS-ssNMR can also be utilized to measure residual dipolar couplings in aligned bicelles (Canlas et al. 2008). In the case of fast uniaxial rotation around membrane normal, MAS-ssNMR can provide orientational restraints that can further refine the structure of small peptides (Hu et al. 2010) as well as large membrane proteins (Park et al. 2012; Das et al. 2012).

As we demonstrated for phospholamban (Traaseth et al. 2009; Verardi et al. 2011), the ab initio determination of the structural topology of a membrane protein in the absence of homology models entails the combination of all of the above information. Lack of information or even relying on only one approach to determine the structural topology of membrane proteins leads to misinterpretation of the experimental results (Traaseth et al. 2007; Shi et al. 2011b).

A major advantage of combining MAS-ssNMR and O-ssNMR data is to obtain structural and topological data using similar lipid compositions in the sample preparations, circumventing the necessity for cross-validation of the structural information obtained from different membrane mimetic systems. Nonetheless, the ssNMR experiments applied to membrane proteins are inherently insensitive. This is due in part to the need for high lipid-to-protein ratios that, while maintaining protein's functional integrity, reduce the effective protein concentrations in the sample. In addition, recurring secondary structure elements (multiple transmembrane α helices) as well as sample heterogeneity tend to give poorly resolved spectra for these proteins. As a result only a few structures of membrane proteins have been determined by ssNMR techniques (Ketchem et al. 1993; Traaseth et al. 2009; Cady et al. 2009; Sharma et al. 2010; Park et al. 2010; Verardi et al. 2011). Our work is motivated by the necessity of improving these techniques for membrane protein samples. Toward this goal, we recently designed several new pulse sequences to improve the sensitivity of both O-ssNMR and MAS-ssNMR experiments. An underlining concept for these new experiments is the optimization of nuclear spin polarization (polarization optimized experiments, POE) to augment the NMR signals or deconvolute multiple excitation pathways to obtain several 2D or 3D experiments simultaneously. The latter translates into a substantial decrease in experimental time.

For O-ssNMR, the polarization is optimized by recovering the previously discarded dipolar (Gopinath and Veglia 2009) and chemical shift coherences (Gopinath et al. 2010a). We demonstrated that these experiments improve the sensitivity of traditional rotating frame SLF experiments such as PISEMA (Wu et al. 1994), HIMSELF (Dvinskikh et al. 2006), SAMPI-4 (Nevzorov and Opella 2003) and HETCOR (Maudsley and Ernst 1977) by ∼40%, thereby cutting experimental time in half (Gopinath et al. 2010a; Gopinath et al. 2010b). Such a strategy has also been demonstrated for the original PELF experiment (Schmidt-Rohr et al. 1994), and enhances sensitivity by > 100% for highly mobile segments in proteins (Gopinath et al. 2011). The gain in sensitivity has allowed us to record comparatively insensitive constant-time experiments, such as the SECT-PISEMA (Gopinath and Veglia 2010) for increasing resolution in 2-dimensional experiments (Mote et al. 2011). In the 3D SE-PISEMAI-HETCOR experiment (Gopinath et al. 2010a), we designed a pulse scheme to recover chemical shift coherences in addition to retaining the sensitivity gain obtained by the polarization inversion. The resulting boost in sensitivity is 80-180%, which enables the implementation of 3D spectroscopy to correlate ¹⁵N chemical shifts, ¹⁵N-¹H dipolar couplings, and ¹H chemical shifts. The three- to seven-fold savings in time alone bring a number of membrane protein systems within reach of this technique. This experiment is used here to obtain the complete set of backbone orientation restraints for sarcolipin in oriented bicelles (Dürr et al. 2012).

In contrast to O-ssNMR, MAS-ssNMR experiments have reached a much higher level of sophistication. In the past decade, many groups have used MAS to for high-resolution structural studies of macromolecular systems, including microcrystalline proteins (Castellani et al. 2002; Wylie et al. 2011; Knight et al. 2012), fibrillar systems (Loquet et al. 2012), membrane proteins (Shi et al. 2011a; Tang et al. 2011), protein complexes (Nieuwkoop and Rienstra 2010), cell extracts (Miao et al. 2012) and whole cell (Renault et al. 2012b). Nonetheless, membrane protein structural biology still offers the biggest challenge for MAS techniques, due to dynamic and static disorder that broaden the resonances of ¹³C and ¹⁵N spectra. Various approaches have been devised to overcome these hurdles: faster spinning rates (Demers et al. 2011), extensive (Reif 2012) and partial deuteration (Asami et al. 2012) coupled with ¹H detection (Reif et al. 2001; Zhou et al. 2012; Marchetti et al. 2012), sparse ¹³C labeling to reduce spectral complexity (Sperling et al. 2010) and minimize ¹³C-¹³C couplings as a source of line-broadening (Loquet et al. 2011) as well as the use of techniques such as RELOAD (Lopez et al. 2009) to gain sensitivity . One can boost sensitivity to an even higher degree by Dynamic Nuclear Polarization (DNP) (Hall et al. 1997), opening up this technique to even bigger systems (Renault et al. 2012a).

All of the above experiments, however, do not take full advantage of the sparse nuclear spin polarization. To address this issue, we introduced the Dual-Acquisition MAS strategy (DUMAS) (Gopinath and Veglia 2012a). This novel strategy allows the concatenation of 2D or 3D pulse sequences into a single experiment and thus potentially combines all the aforementioned advances, boosting the capacity of any NMR spectrometer at least two-fold without the need for sample manipulation or additional hardware. At the heart of this approach is simultaneous cross-polarization from ¹H to ¹³C and ¹⁵N, which has also been used in time-shared experiments (Linser et al. 2011; Nielsen et al. 2012), and long lived ¹⁵N longitudinal magnetization (Giraud et al. 2005) which can be stored and recalled to perform additional experiments. The DUMAS scheme enables one to acquire two experiments: the first with 100% of the sensitivity with respect to classical pulse sequences, and a second experiment generated from the ¹⁵N polarization with typically 80% of the sensitivity (Gopinath and Veglia 2012a). A well-designed combination of these experiments allows sequential assignment as well as side chain assignment in a single 3D experiment (Gopinath and Veglia 2012b). Using the initial scheme of the DUMAS experiments, it is possible to decode the coherences in multidimensional NMR experiments to recover the orphan spin operators that are discarded from the classical experiments. This approach that we called MEIOSIS (Multiple Experiments via Orphan SpIn operatorS) enables one to collect up to four multidimensional NMR experiments with a time saving greater than 50% (Gopinath and Veglia 2013). See Supplementary Tables 1-3 and Fig S5 and S6 for a detailed comparison of time savings using different sequences on SLN.

In this article, we use POE both in O-ssNMR and MAS spectroscopy to determine the structural topology of sarcolipin, a 3.8 kDa single transmembrane regulator of the sarco(endo)plasmic reticulum Ca²⁺-ATPase (SERCA) (Odermatt et al. 1998; Traaseth et al. 2008). The traditional approach to solve membrane protein structure by ssNMR would involve the following experimental steps before structure calculation: a) Obtaining preliminary 2D spectra to optimize elements in the pulse sequence such as DARR mixing times, indirect evolution times as well as the determining the relative sensitivity and resolution attainable in each dimension, b) Obtaining 3D experiments based on the knowledge gained from the 2D experiments to sequentially assign the protein and c) complementing these secondary structure restraints obtained from MAS-ssNMR with orientation restraints on the backbone amide residues from O-ssNMR. We show the utility of DUMAS and MEIOSIS in combining multiple experiments to reduce experimental time, so one can simultaneously optimize mixing times, indirect evolution times as well as make decisions about which 3D experiments have the best chance of giving a complete sequential assignment. We also show the use of complementary 3D experiments simultaneously acquired to sequentially assign the protein. Finally we show the tremendous time savings that can be obtained by using the sensitivity enhanced versions of 3D O-ssNMR pulse sequences which give the complete set of orientation restraints on the protein backbone, viz. ¹⁵N and ¹H anisotropic chemical shifts and ¹⁵N-¹H dipolar couplings. The individual restraints (anisotropic and isotropic) obtained exclusively by these solid state NMR approaches are then combined in a manner analogous to the previously described hybrid approach (Shi et al. 2009b; Veglia et al. 2012) using an improved algorithm to converge to a structure ensemble with sub-angstrom precision.

Experimental Methods

Protein expression, purification and sample preparation

SLN was over-expressed in E. coli bacteria fused to a maltose binding protein (MBP), cleaved, and purified as described earlier (Buck et al. 2003; Veglia et al. 2010). Recombinant SLN was purified using reverse-phase HPLC. The fractions containing the protein were lyophilized and used for all solid state NMR experiments. The protein concentration was assessed using UV spectrophotometry using absorption at 280 nm (ε=10,300 M^-1cm^-1) as well as SDS-PAGE gels using densitometry. For the bicelle preparations, approximately 3.5mg SLN was dissolved in 6.7mg D6PC in 20mM HEPES/100mM NaCl buffer at pH 7.0 and then added to 31mg 4:1 DMPC/POPC vesicles suspension. After several freeze thaw cycles, the bicellar phases were formed and transferred to a cylindrical sample holder. For MAS preparations, 2.2mg of uniformly-[¹³C,¹⁵N]-SLN was reconstituted in DPC micelles and added to 15mg deuterated DMPC lipids dissolved in DPC micelles. The detergent was removed by incubating the solution with Biobeads SM2 (30:1 ratio with the detergent) for 3h at room temperature. The protein-lipid vesicles were collected as a pellet by centrifugation at 200,000xg at 4°C for 1 hour and packed in a 3.2mm MAS rotor for NMR experiments without further treatment.

NMR Spectroscopy

O-ssNMR Experiments

Experiments on bicelle preparations were carried out on a 16.85T VNMRS Spectrometer with a low-E bicelle probe built by the RF Program at NHMFL, Tallahassee, FL (Gor'kov et al. 2007). Temperature was maintained constant at 298.15 K and bicelle alignment was confirmed by a ³¹P spectrum. After initial optimization, the 3D-SE-PISEMAI-HETCOR experiment (Fig 1A) was performed with 80 transients for each of the 20 t1 (¹H-¹⁵N-DC dimension, evolution time of 2.4ms) and 15 t₂ (¹H-CS dimension, evolution time of 1.13ms). A cross polarization time of 1000 μs, direct dimension acquisition time of 5ms, and a recycle delay of 3s were used in all experiments. A 50-70Hz Lorentz-to-Gauss apodization was used in the direct dimension and a 100Hz Gaussian apodization was used in the indirect dimensions. The matrix size before Fourier transformation was 8192 × 256 × 256.

Fig. 1.

Pulse sequences O-ssNMR and MAS-ssNMR experiments.

(A) SE-PISEMAI-HETCOR, (B) 2D-MEIOSIS-(DARR/N(CA)CX/CA(N)CO/NCO) with the following phases: Φ; 1={y y y y −y −y −y −y}, Φ₂={x x −x −x}₂, Φ₃={y −y y −y}₂, Φ4={y y −y −y}, Φrec={x −x −x x −x x x −x} and (C) 3D-DUMAS(CANCO/NCACX) with the following phases: Φ1 = (y, −y)₄, Φ:2 = (y, y, y, y, −y, −y, −y, −y), Φ3 = (x, x, −x, −x)₂, Φrec = (x, −x, −x, x, −x, x, x, −x). For the pulse sequence in (B), the ¹³C carrier frequency is placed in the center of the spectral window for DARR (100ppm). SPECIFIC-CP is achieved by a phase modulated pulse on ¹³C that effectively changes the offset to ¹³Cα (64ppm) for the first transfer and to C′ (178ppm) for the second transfer.

MAS-ssNMR Experiments

MAS experiments on SLN in lipid vesicles were carried out on either a 14.09T or a 16.85T VNMRS Spectrometer with a 3.2mm BioMAS™ probe (Agilent Technologies) at a spinning speed of 10.0kHz and a temperature of 277.15 K. Two-dimensional MAS-ssNMR experiments were carried out by combining ¹³C-¹³C DARR, CA(N)CO, NCO and N(CA)CX experiments into a single 2D experiment using the MEIOSIS (Multiple Experiments via Orphan Spin Operators) strategy (Fig 1B) (Gopinath and Veglia 2013). Simultaneous cross-polarization from 1H to ¹³C and ¹⁵N was obtained using a contact time of 300μs, with the RF power of ¹⁵N and ¹³C channels set to 35kHz and ¹H RF power optimized to achieve the n=2 side-band matching condition (59kHz). The phase of ¹⁵N spinlock during NCA transfer is altered between +x and −x and the resultant data is stored in separate files. Sum and difference of the two data sets with phases x and −x respectively gives DARR and N(CA)CX for the first acquisition, and NCO and CA(N)CO for the second acquisition. Fig S3 and Fig S4 show the dependence of the contact time for the first SPECIFIC-CP step on the distribution of polarization amongst different pathways. It was found that 3000μs was the optimal contact time, as it gave the maximum NCα transfer for both N(CA)CX and CA(N)CO experiments. As with a microcrystalline preparation of ubiquitin (Gopinath and Veglia 2013), we found that at this contact time, 54% residual polarization is left on ¹³C (to be used for a ¹³C-¹³C DARR experiment) and 34% residual polarization is left on ¹⁵N (to be used for a NCO experiment).

The ¹³C-edited experiments (DARR and CA(N)CO) were acquired with a dwell time of 32μs for 128 indirect points (t1 evolution of 4.096ms) while the ¹⁵N-edited experiments were acquired with a dwell time of 320μs for 16 indirect points (t1 evolution of 4.8ms). Acquisition time of 15ms and recycle delay of 2s was used. 64 transients were used for each t1 increment. Due to the reduced spectral width in the ¹⁵N dimension, the ¹⁵N-edited experiments can be acquired eight times along with a single acquisition of ¹³C-edited experiments. The number of transients for the NCO and N(CA)CX experiments thus becomes 512 (64 × 8). We acquired two independent experiments at the mixing times of 10ms and 50ms during the DARR transfer. As this transfer does not affect the NCO and CA(N)CO experiments, a further boost in sensitivity is obtained by adding these experiments, giving the total number of transients for NCO and CA(N)CO as 1024 and 128, respectively. As these are the least sensitive amongst the four concatenated experiments (Fig S3), this design allowed us to boost its sensitivity by repetition, even while optimizing the other experiments. Thus, in the end, we had a total of 6 spectra from 2 separate experiments − ¹³C-¹³C DARR with 2 different mixing times (with 64 transients each), N(CA)CX with 2 different mixing times (with 512 transients each), CA(N)CO (128 transients) and NCO (1024 transients). Note that the traditional approach would have allowed the recording of only 2 of these 6 experiments in the same time.

3D-DUMAS-ssNMR

To assign the backbone and side chain atoms, we performed the 3D-DUMAS-NCACX-CANCO experiment (Fig 1C) at 277.15 K and 10.0 kHz MAS spinning. The t1 dimensions for both experiments were co-evolved with a dwell time of 150μs for the ¹³C-dimension (CANCO) and 300μs for the ¹⁵N-dimension (NCACX). 16 t1 points were obtained, giving a total of 2.4ms evolution for ¹³C and 4.8ms evolution for ¹⁵N respectively. Bidirectional cross polarization (Gopinath and Veglia 2012b) from ¹³Cα to ¹⁵N (CANCO) and ¹⁵N to ¹³Cα (NCACX) was achieved with a contact time of 3ms at RF field strength of 25 kHz (5.ω_r/2) on ¹⁵N, and a ¹³C RF amplitude of 15 (3.ω_r/2) and 35 (7.ω_r/2) kHz for Cα and CO transfer respectively. ¹⁵N magnetization was stored along the z-direction with a 90° pulse after the t1 evolution periods. ¹³C-dimension (NCACX) was then evolved, followed by a 10ms DARR mixing period and finally detected (NCACX) with a 20ms acquisition time. After this acquisition, the ¹⁵N magnetization was brought to the transverse plane with another 90° pulse and evolved (CANCO). Magnetization was then transferred to ¹³CO with a 3ms contact time. The second detection of the ¹³C dimension (CANCO) was again carried out with an acquisition time of 20ms. The t2 dimensions were thus evolved independently for both experiments with 16 points each, giving a ¹⁵N (CANCO) evolution time of 4.8ms and ¹³C (NCACX) evolution time of 2.4ms. 512 transients were co-added for each t1 and t2 points. A recycle delay of 2 sec was used.

The ¹³C-dimension was referenced externally with adamantane at 40.48ppm and the ¹⁵N dimension was referenced indirectly from the ¹³C-dimension. All spectra were processed with NMRPipe (Delaglio et al. 1995) and analyzed with Sparky (Goddard and Kneller 2008). A Lorentz-to-Gauss apodization of 60-80Hz was used in the direct and indirect dimensions. The matrix size before Fourier transformation was 16384 × 512 × 512.

Structure Calculations

As the experiments were performed in lipid bicelles with the bicelle-director perpendicular to the external magnetic field, the ¹⁵N average isotropic chemical shift of 120ppm and ¹H isotropic chemical shift of 8.1ppm were subtracted from each observed anisotropic chemical shift and scaled by 2.5 to obtain the input restraints to obtain the input for structure calculations. Observed dipolar couplings were also scaled by 2.5 to obtain the necessary restraints. This factor of 2.5 is a combination of the scaling of anisotropic parameters by 2 in the parallel orientation and a scaling factor of 1.25 to account for bicelle motions (Nevzorov 2011). Restraints were incorporated with an associated error of ±5ppm for ¹⁵N chemical shifts, ±1.5ppm for ¹H chemical shifts and ±0.5kHz for dipolar couplings. Dihedral angle restraints from MAS-ssNMR were calculated using the isotropic chemical shifts of Cα, Cβ, C and N (amide) atoms as an input for TALOS+ (Shen et al. 2009) and implemented with an error of ±30°.

All oriented ssNMR restraints (anisotropic chemical shifts for amide ¹⁵N and amide ¹H and ¹⁵N-¹H dipolar couplings) and MAS-ssNMR restraints (dihedral restraints from TALOS) were combined in a single energy function along with potentials corresponding to standard peptide geometry (E_chem = E_bond + E_angle + E_improper + E_vdW) (Schwieters et al. 2003). In addition, hydrogen bonds between the carbonyl oxygen and amide proton of {i} and {i+4} residues were implemented for the transmembrane segment. Orientation restraints were implemented with flat-well harmonic potentials using csaPot and rdcPot modules from XPLOR-NIH (version 2.29). These inbuilt C++ based modules allow for a faster calculation of structure than the PYTHON based modules (Bertram et al. 2000) we previously used (Shi et al. 2009b) for the structure determination. The ¹⁵N tensor was defined with the requisite parameters in XPLOR-NIH with the following traceless CSA tensor components: σ11=55.3ppm, σ22=-97.6ppm, σ33=43.3ppm, β= -17° and γ=0°. Similarly, 1H traceless tensor components were defined with σ11 = 6.3 ppm, σ22 = 1.3ppm, σ33 = -7.6ppm, β = 97° and γ = 0°(Yao et al. 2010).

The hybrid energy function (E_O-ssnmr, E_Mas-ssnmr and E_chem) was minimized using a simulated annealing protocol in the torsion angle space. Starting from an extended conformation of SLN, we carried out simulated annealing after an initial equilibration step at 6000K for 10ps. The temperature was decreased in steps of 5 K with a 0.5 ps time step to reach a final temperature of 10 K. A knowledge based E_z-potential (Senes et al. 2007) was implemented as this stage to insert SLN into a virtual membrane bilayer. The structure was then equilibrated at 10 K for 20 ps and subjected to Powell Minimization in Cartesian space. Of 200 structures generated, 10 lowest energy structures were selected for simulated annealing in the Cartesian space to relax local geometries. ¹H-CSA restraints were included at this stage. After a 10ps equilibration at 300K, the temperature was reduced to 5K in steps of 2.5K, with 0.5ps simulation at each step. E_z-potential was again implemented and a final equilibration at 5K for 20ps followed by Powell minimization was carried out to give the final structure. One hundred structures were generated and the ten lowest energy conformers were selected for further analysis. The convergence of the lowest energy structures was determined by aligning the projections of each of the structures on the XY plane (Shi et al. 2009b). The RMSD of each conformer from the average structure was calculated without further manipulation of the coordinates in order to maintain the topology calculated from anisotropic NMR parameters. The tilt angle (with respect to the bilayer normal) and rotation angle (with respect to the Cα atom of residue R6) were calculated by aligning a model 20-residue alpha helix with the residues 14-28. Quality of the final structure ensemble was validated by ProCheck (Laskowski et al. 1996) and MolProbity (Davis et al. 2007), as well as the convergence of R-values for individual orientation restraints.

Results

Magnetically aligned bicelle samples give resolved spectra with a sensitivity and resolution higher than mechanically aligned samples. The latter combined with the ease of sample preparation, better hydration, pH control and sample stability makes bicelles an excellent membrane mimetic for O-ssNMR experiments (Veglia et al. 2012; Dürr et al. 2012). This allows highly sensitive samples to be prepared for 3-dimensional spectroscopy. Fig S1 shows a typical 1D spectrum after initial optimization of CP conditions. For SLN, these preparations give 2D SLF spectra with remarkable resolution and a distinct PISA-wheel pattern (Page and Cross 2008; Mote et al. 2011). All these improvements make it possible to acquire 3D experiments with the sensitivity enhancement pulse sequences. Figure 2A shows strip plots from the 3D-SE-PISEMAI-HETCOR experiment. All of the resonances belonging to the transmembrane domain of SLN are present and well resolved, with characteristic oscillatory patterns (Mesleh et al. 2002; Mascioni and Veglia 2003) evident for the chemical shifts and the dipolar couplings. We have previously assigned the 2D SE-PISEMA spectrum for SLN in these lipids using a 3D-SE-PISEMA-PDSD experiment that allowed for a sequential assignment using ¹⁵N-¹⁵N correlations established by proton driven spin diffusion (Mote et al. 2011). This allowed a straightforward transfer of the assignment to the resonances observed in the 3D SE-PISEMAI-HETCOR experiment. The high quality of the spectrum, with linewidths for ¹⁵N-CS of ∼150Hz, ¹H-CS of ∼800-1000 Hz and the ¹⁵N-¹H DC of ∼100Hz allowed us to obtain the set of orientation restraints that completely describe the backbone topology.

Fig. 2.

Strip plot of the 3D-SE-PISEMAI-HETCOR spectrum. All resonances from N11 to S28 are completely resolved and assigned based on our previous assignment for the 2D spectrum. Average signal is 55σ (range of 24σ-124σ). Spectrum is plotted starting at 18σ with contours progressing by factor of 1.2.

For a rapid screen to determine the relative dispersion attainable in each dimension, as well as optimizing mixing times for the DARR element (Takegoshi et al. 2001) in the pulse sequence, we used the 2D-MEIOSIS experiment. Figure 3 shows a variant of this experiment, i.e. MEIOSIS-(DARR-CA(N)CO-NCO-N(CA)CX), in which the four 2D spectra are obtained simultaneously within a single experiment. Although several peaks are resolved and residue type assignments are easily obtained from these experiments, they give only an ambiguous assignment. The comparatively low resolution in the 2D spectra is a direct result of fast T2 relaxation, which is also observed for other small membrane embedded proteins (Su et al. 2011), and underscores the importance of 3D experiments to assign these proteins. In the N(CA)CX obtained at 10ms mixing time, it is possible to identify several intense correlations, indicating that the mixing period is sufficient to transfer the polarization from the backbone to the side chains. At a 10ms mixing time, the N(CA)CX spectrum shows several Cβ and C' resonances. At a 50ms mixing time, the Cβ and possibly the other side chain peaks increase in intensity at the expense of both C' and Cα. Our assignment strategy relies on obtaining information on residue type using statistically predicted chemical shifts and connecting these residues via a ‘backbone-walk’. Based on this screening, we concluded that the DUMAS-CANCO/NCACX pulse sequence with a 10ms mixing time during the DARR transfer in NCACX is optimal for assigning SLN, as it would give information on the residue type in the NCACX spectrum and also allow a simultaneous main-chain walk via the common carbonyl resonance of the {i} and {i+1} spin systems. A mixing time of 10ms was deemed optimal as longer mixing times decreased the intensity of the C' resonances. Therefore, we acquired two 3D experiments simultaneously using the 3D-DUMAS-CANCO/NCACX pulse sequence, with a 10ms mixing time for the NCACX spectrum during ¹³C-¹³C DARR transfer. After assignment of a majority of residue types in the transmembrane segment based on the statistical Cα and Cβ chemical shifts obtained from the 2D and the 3D spectra, we sequentially assigned the transmembrane segment from residue E7 to V26 by the classical ‘main chain walk’, aligning the carbonyl chemical shifts for {i} and {i+1} residues. This assignment was aided by relatively high sequence variability in the SLN transmembrane domain, which has threonine, methionine and tryptophan residues in addition to the hydrophobic leucine, isoleucine and valine residues frequently found in TM domains. The leucine region of the spectrum (∼56ppm) suffers from severe overlap of resonance. However, with resolution in the carbonyl region of the spectrum, we were also able to assign these residues. Residues N4-R6 which are known to undergo conformational dynamics, were not assigned, possibly due to inefficient cross-polarization for these residues. The terminal residues M1-I3 and Y31 were assigned in a refocused-INEPT experiment at 298.15K (Supplementary Fig 2).

Fig. 3.

Spectra obtained with MEIOSIS-(DARR/CA(N)CO/N(CA)CX/NCO) pulse sequence (A) NCO Correlation, (B) CA(N)CO Correlation. (C) ¹³C-¹³C DARR Correlation with 10ms mixing time, (D) ¹³C-¹³C DARR Correlation with 10ms mixing time, (E,F) N(CA)CX correlations with 10ms mixing time and (G,H) N(CA)CX correlation with 50ms mixing time. (C) and (D) are plotted starting at 7σ and contours progressing by a factor of 1.1. All other spectra are plotted starting at 6σ with contours progressing by a factor of 1.1.

Figure 5A shows the ensembles of the SLN conformers obtained. The high precision of the ensemble (backbone, incl. H^N + Cβ RMSD ∼0.44 Å) is also reflected in low energies for all structural restraints. The global as well as the individual R-values for all orientation restraints are converged below 1 (Figure 5 E-G), indicating a good agreement between the calculated structures and the experimental O-ssNMR data. Table 1 summarizes the structural statistics for these calculations. The hybrid NMR ensemble shows disordered N- and C- termini in agreement with the solution NMR ensemble obtained in DPC micelles (Mascioni et al. 2002; Buffy et al. 2006a). In fact, experiments utilizing cross polarization both in bicelles and lipid bilayers are unable to detect the resonances corresponding to residues 1-3 and 31. These were assigned using the refocused-INEPT experiment at 298K (Fig S3). The resonances corresponding to residues 29-31 detected by SE-PISEMA experiments in aligned bicelles show dipolar couplings that are averaged by the substantial motions in these segments and hence were not included in the final structure calculations. For the most structured residues (Residues 6-28), hybrid NMR ensemble shows a higher precision than the corresponding solution NMR ensemble (PDB Code: 1JDM). The topology is converged to a well-defined minimum in the conformational space. The tilt angle for SLN with respect to the bilayer normal was found to be 24±1° and the rotation angle was 57±5°. The indole ring of W23 is oriented with its plane parallel to the bilayer normal, with the N-H bond vector oriented in the direction of the C-terminal bilayer interface, possibly making contact with the phosphate head groups. Along with W23, the conserved residues R6, located near the cytosolic N-terminus, and R27, near the lumenal C-terminus, anchor SLN to the lipid membrane, modulating the tilt and rotation angles which seem to be only minimally affected in the different membrane mimetic systems (Mote et al. 2011). The NMR structural ensembles determined by the hybrid method are in good agreement with molecular dynamics simulations (Shi et al. 2009a) which show a similar orientation of the indole side chain of W23, and a previous oriented solid state NMR study (Buffy et al. 2006b) carried out mechanically aligned DOPC/DOPE bilayers, which shows a similar the tilt angle of ∼23° with respect to the bilayer.

Fig. 5.

(A) Structure ensemble obtained for SLN from the hybrid MAS/O-ssNMR structure calculations (10 lowest energy structures) (B-D) Correlation of experimental O-ssNMR restraints with the back-calculated restraints (E-F) Oscillation patterns for all orientation restraints and the R-values of back-calculated restraints

Table 1. Summary of Structure Ensemble Statistics (10 structures).

RMSD from MAS NMR Restraints
NH-C(O) hydrogen bond, Å	0.052 ± 0.004
Dihedral Angles (TALOS)	0.000 ± 0.000
PISEMA R-factors
15N CSA	0.5 ± 0.4
15N-1H DC	0.7 ± 0.3
1H CSA	0.3 ± 0.3
RMSD from Idealized Covalent Geometry
Bond, Å	0.004 ± 0.000
Angle, °	0.60 ± 0.01
Improper angle, °	0.13 ± 0.01
Measure of Structure Quality, %
Most Favored Region	99 ± 1
Additionally Allowed Region	1 ± 1
Generously Allowed Region	0
Disallowed Region MolProbity® Score	098 ± 2 percentile
Precision of Structures (RMSD for atoms NH+HN+C'+0+Cα+Cβ), Å
Domain Ib+II (Res 6-28)	0.43
Domain Ib (Res 6-13)	0.44
Domain II (Res 14-28)	0.40
Protein Topology
Tilt Angle (Res 6-28),°	24 ± 1
Rotation Angle (from R6-Cα), °	55 ± 5
Depth of Insertion (R6-Cα), Å	16.1 ± 0.2

Discussion

While multidimensional MAS-ssNMR techniques are mature for structure determination of large systems, their application to membrane proteins is limited by the sensitivity and resolution that can be attained in these systems. Obtaining complete resolution is often not possible without 3D experiments, which take a long time due to the poor sensitivity of these systems. The DUMAS strategy allowed us to substantially reduce experimental time as compared with traditional MAS-ssNMR pulse sequences.

The initial experiments with MEIOSIS allowed the determination of parameters, such as the appropriate mixing time, that were important to set up the 3-dimensional experiments for assignments. These initial experiments also allowed us to determine the best possible experimental approach to assign the protein based on relative resolution attainable in each dimension. The DUMAS-CANCO/NCACX experiment allowed us to assign a majority of the protein sequentially as the experiment combines the complementary pulse sequence into a single experiment. Moreover, the simultaneous acquisition of both the experiments allowed the data to be fairly compared without the need for concern regarding sample stability or correction factors.

Compared to MAS-ssNMR, multidimensional O-ssNMR spectroscopy is still in its infancy, in spite of the fact that the number of membrane protein structures being solved by this technique is greater. Sensitivity gains obtained from the 3D-SE-PISEMAI-HETCOR are a substantial leap forward in obtaining complete resolution for these studies. It must be noted here that the resolution of the structure depends primarily on the O-ssNMR data, as the secondary structure obtained from MAS-ssNMR experiments contains no information on the topological parameters of tilt and rotation angle. As such, the O-ssNMR experiments described herein are indispensable in order to characterize the structure of membrane proteins precisely. TALOS based dihedral angle restraints obtained from MAS-ssNMR are a perfect complement to these orientation restraints from O-ssNMR and help convergence of the structure ensemble in spite of inherent degeneracy in the observed values of dipolar couplings and anisotropic chemical shifts.

In the past these MAS-ssNMR and O-ssNMR techniques were thought to be mutually exclusive and most of the researchers used either MAS or O-ssNMR to obtain de novo structure determination. Recent developments of structural biology emphasize the use of multiple techniques to substantiate biological and structural data (Traaseth et al. 2007; Shi et al. 2011b). In this spirit, we combined MAS-ssNMR and O-ssNMR restraints into a unique refinement protocol. The idea to combine these techniques is similar to solution NMR, where distance and torsion angle restraints are energy minimized together with residual dipolar couplings to obtain high-resolution structures and reciprocal orientations of protein domains. While some concerns can be raised regarding the compatibility of solution NMR restraints in micelles, which can also serve as complements to O-ssNMR restraints, the agreement between MAS and O-ssNMR data stems from the similarity between the two membrane mimicking environments.

Conclusions

We show here that the complete set of high resolution restraints for both MAS-ssNMR and O-ssNMR can be conveniently obtained for membrane proteins using POE for both oriented and MAS-ssNMR spectroscopy. The combination of structural and topological restraints from these techniques into a hybrid NMR calculation approach makes it possible to determine the high resolution structure and topology of membrane proteins. Although we anticipate that POE experiments will have the biggest impact on data acquisition of membrane proteins, they are universally applicable to other biomacromolecular systems.

Fig. 4.

Strip plots from the 3D-DUMAS-NCACX (red)/CANCO (blue) experiment. NCACX spectrum is plotted starting at 12σ (Cα-Cβ region) and 8.5σ (C region) and contours progressing by a factor of 1.1. Average signal per residue for NCACX is 28σ (Cα-Cβ region, range 12σ-63σ) and 15σ (carbonyl region, range 9σ to 25σ). CANCO spectrum is plotted starting at 6σ and contours progressing by a factor 1.1. The average signal for per residue is 10σ (range 7σ to 20σ). The average linewidth in each spectrum is ∼1.1 ppm for ¹³Cα (indirect dimension), 1.0ppm ¹³C (direct dimension) and 2.5ppm for ¹⁵N (indirect dimension)

Abbreviations

SLN

Sarcolipin

SERCA

sarcoplasmic reticulum Ca²⁺-ATPase

DMPC

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

D6PC

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

POPC

1-palmitoyl, 2-oleyl-sn-glycero-3-phosphocholine

PISA

Polarity Index Slant Angle

DUMAS

Dual Acquisition Magic Angle Spinning

MEIOSIS

Multiple Experiments via Orphan Spin Operators