Nanotube Array Method for Studying Lipid-Induced Conformational Changes of a Membrane Protein by Solid-State NMR

Abstract

Anodic aluminum oxide substrates with macroscopically aligned homogeneous nanopores of 80 nm in diameter enable two-dimensional, solid-state nuclear magnetic resonance studies of lipid-induced conformational changes of uniformly ¹⁵N-labeled Pf1 coat protein in native-like bilayers. The Pf1 helix tilt angles in bilayers composed of two different lipids are not entirely governed by the membrane thickness but could be rationalized by hydrophobic interactions of lysines at the bilayer interface. The anodic aluminum oxide alignment method is applicable to a broader repertoire of lipids versus bicelle bilayer mimetics currently employed in solid-state nuclear magnetic resonance of oriented samples, thus allowing for elucidation of the role played by lipids in shaping membrane proteins.

Main Text

During the last decade, solid-state nuclear magnetic resonance (ssNMR) has emerged as a unique biophysical tool that is capable of determining structures of membrane proteins in nativelike lipid bilayer environments without the need for crystallization (1). Tremendous progress in solving structures of membrane proteins has been achieved by further developing both magic-angle spinning (2) and oriented-sample (OS) NMR methods (3). OS NMR is particularly attractive for biophysical studies, inasmuch as this method could be employed at nearly physiological conditions and without the fast (typically ∼10–40 kHz or higher) sample spinning required for magic-angle spinning NMR. The remaining road-blocks for the OS ssNMR methods appear to be in the following areas: 1) the laborious task of preparing lipid bilayer samples aligned between thin glass plates (4) and 2) the need for development on new efficient lipid bilayer alignments methods to yield exceptionally narrow NMR resonances and acceptable spectral resolution for a broad repertoire of lipid/and surfactant compositions.

While magnetically aligned DMPC/DHPC bicelles (5–7) represent an attractive alternative to the glass-supported bilayers, the bicelles systems could be extended to only a relatively narrow repertoire of lipids. In addition, the shorter-chain DHPC or various surfactants employed in bicelles may act as detergents for the bicelle-embedded proteins and, thus, able to interfere with the protein native conformations.

An acceptable degree of lipid macroscopic alignment, albeit in cylindrical geometry, has been demonstrated by self-assembling nanotubular bilayers within ordered nanochannels of anodic aluminum oxide (AAO, Fig. 1 A) (8–11). Subsequent studies by electron paramagnetic resonance and differential scanning calorimetry have demonstrated that lipid hydration and thermodynamic properties of these nanopore-confined bilayers are remarkably similar to those exhibited by unsupported multilamellar liposomes (12,13). The principal advantages of the AAO nanopore alignment method are in its applicability to bilayers of essentially any lipid composition over a broad range of temperature, pH, and ionic strength—parameters that could be controlled easily and precisely. These features warranted further development of this method for studying lipid-induced conformations of membrane proteins by OS ssNMR, even though the initial studies produced resonance peaks with the widths comparable to those obtained by the glass plate method.

Figure 1.

Nanoporous AAO substrate and macroscopic alignment of lipid bilayers inside the nanopores for OS ssNMR. (A) Cartoon of nanotubular bilayer with inserted Pf1 protein inside a nanopore. (B) Histograms of the pore diameters obtained from SEM images of Whatman Anodisc (blue, magnified fivefold; d = 227 ± 53 nm) and in-house AAO (green; d = 81 ± 4 nm). (Solid lines) Best fits to Gaussian distributions. SEM images of Whatman Anodisc from the side of narrower pores (C) and in-house homogeneous AAO (D).

Previous OS ssNMR studies employed commercial Whatman Anodiscs (GE Healthcare, Pittsburgh, PA) that have somewhat disordered pore morphology (Fig. 1, B and C). Furthermore, multinuclear ssNMR data for lipids in Whatman AAO were consistent with formation of wavy bilayer tubules ∼0.4 μm in length, which have contacts with the pore walls at the end of the tubules (14), presumably at surface defect sites. The effect of lipid-rhodopsin hydrophobic mismatch on one-dimensional ²H NMR spectra of perdeuterated lipids in AAO has also been investigated (15), but no two-dimensional spectra of uniformly labeled protein have been reported, to the best of the authors’ knowledge.

Here we describe a dramatic improvement of lipid alignment that was achieved by employing AAO substrates with a uniform nanoporous structure (Fig. 1, B and D) prepared in-house by a two-step oxalic acid anodization process (16) from a high-purity aluminum foil (99.997%; Strem, Newburyport, MA). Additional details on AAO preparation are given in the Supporting Material. SEM images of in-house made AAO (Fig. 1 D) revealed significantly improved pore morphology over Whatman Anodiscs (Fig. 1 C). The image analysis by the software MATLAB (The MathWorks, Natick, MA) scripts yielded 81 ± 4 and 227 ± 53 nm pore diameters for the in-house and Whatman AAO respectively (Fig. 1 B). Larger pieces of homogeneous AAO substrates ∼50 μm thick were cut into smaller rectangular 3.8 × 8.0 mm strips to fit into 5.0 mm round NMR glass tube (New Era Enterprises, Vineland, NJ).

Macroscopic alignment and high rotational mobility of the phospholipids within fully hydrated nanochannels of AAO are two determining factors for the line widths of OS NMR resonances. When pure DOPC lipids were deposited within the nanopores of Whatman AAO and mounted inside a rectangular NMR tube that typically is used for OS NMR of glass plate samples, the ³¹P NMR spectrum demonstrated the phosphate line-width of ∼600 Hz (compare to Fig. S1 in the Supporting Material), which was similar to the results of the previous studies (7,8). For lipids deposited into in-house homogeneous AAO, improved alignment resulted in a significant decrease of the line-width of the ³¹P resonance to ∼240 Hz using a flat tube, and to ∼140 Hz when using a round tube (see Fig. S1), mainly due to a decrease in magnetic susceptibility broadening and improved radiofrequency field homogeneity.

Fig. 2 shows ¹H-decoupled ³¹P (Fig. 2 A) and ¹⁵N NMR (Fig. 2 B) spectra of Pf1 coat protein reconstituted in DOPC/DOPG (9:1 molar ratio) AAO-confined nanotubular bilayers. Least-squares fitting of the spectrum (Fig. 2 A) to a two-peak model yielded ∼10:1 DOPC/DOPG molar ratio in the actual sample (see Fig. S2). The full width at half-height of the rightmost (DOPC) peak is ∼170 Hz, suggesting that the lipid bilayers within the AAO nanopores are highly aligned even in the presence of the protein. The resolution of the single peaks of ¹H-decoupled ¹⁵N spectrum of the same sample (Fig. 2 B), especially those at ∼60 ppm, suggests a good alignment and refolding of the Pf1 coat protein within the DOPC:DOPG lipid bilayers.

Figure 2.

One-dimensional ¹H-decoupled ³¹P (A) and ¹⁵N (B) NMR spectra of Pf1 coat protein reconstituted in DOPC/DOPG (9:1) nanotubular lipid bilayers formed within the homogeneous AAO nanopores. A total of 1024 and 8192 scans were averaged for the ³¹P and ¹⁵N experiments, respectively.

The improved morphology of AAO nanopores enabled us to obtain well-resolved two-dimensional NMR SAMPI4 spectra of Pf1 in DMPC and DOPC/DOPG bilayers with B₀ aligned along the lipid nanotubes (Fig. 3). Spectroscopic assignment of the SAMPI4 spectra has been taken from Opella et al. (17). Similar but measurably different spectral patterns of Fig. 3 suggest a change in the Pf1 conformation caused by bilayer lipid composition. The helix tilt angle α and order parameter S₀ were determined by minimizing the distances between the centers of the spectral peaks (Fig. 3, crosses) and the theoretical PISA wheels (18,19) using an in-house-written MATLAB script. ¹⁵N chemical shift tensors for glycines were taken as σ = (41, 64, 215) ppm and σ = (64, 77, 222) ppm for all other residues. In order to fit the static PISA wheels to the perpendicular orientation of the lipid nanotube director, the NMR frequencies were transformed as (20) ${\nu }_{\text{nanotube}}={\nu }_{\text{isotropic}}-S_0\left({\nu }_{\text{static}}-{\nu }_{\text{isotropic}}\right)/2$.

Figure 3.

Two-dimensional SAMPI4 spectra of Pf1 coat protein reconstituted in DMPC (top) and DOPC/DOPG at 9:1 molar ratio (bottom) and aligned by AAO are overlaid with least-squares PISA wheels. Positions of the centers of NMR intensities for the individual peaks corresponding to nonglycine residues (shown as red crosses) were least-squares fitted to a PISA wheel (continuous red lines); fits to glycines (shown as blue crosses) are also given (dashed blue lines). Spectra were recorded at T = 45°C and B₁= 51 kHz with 1024 scans used for each of the 64 t₁ increments. To see this figure in color, go online.

Further details of the fitting procedure are given in the Supporting Material. The best fit for the PISA wheel for Pf1 in DMPC (Fig. 3 A) corresponds to an order parameter S₀ = 0.85 with a helix tilt angle, α = 23.1°, whereas the fit for DOPC/DOPG yielded a slightly higher S₀ = 0.87 with a measurably smaller tilt angle of α = 20.2°. Errors in S₀ and α are estimated to be <1.4% and <3.3%, respectively, and arise from ±5 ppm uncertainties in the chemical shift tensors (see Fig. S3). Note that S_o and α reported here were calculated for the same set of chemical shift tensors and, therefore, when comparing these parameters for Pf1 for the two bilayer compositions, these errors contribute in a systematic way. The observed trends in S₀ and α are expected because the length of the hydrocarbon chain in DOPC is ∼2 Å longer than that of DMPC (21). As a result, the larger thickness of the hydrophobic core of DOPC/DOPG leads to a smaller tilt angle of the Pf1 coat protein with regard to the bilayer normal (22) versus those in DMPC. A larger order parameter S₀ for Pf1 in DOPC/DOPG is likely due to a lesser degree of dynamic fluctuations of the bilayer surface for these longer lipids. Notably, previous ssNMR studies (23) of Pf1 aligned in DMPC:DHPC bicelles demonstrated that the TM helix adopts a significantly higher tilt of α ≈ 30°. We attribute this to a lower DMPC lipid-packing density in bicelles versus lipid nanotubes. The order parameter for DMPC/DHPC bicelles S₀ = 0.8 is also lower versus DMPC nanotubes. This is expected inasmuch as the lipid nanotubes are constrained by the rigid nanopores of AAO, whereas highly anisotropic tumbling of bicelles is controlled only by the magnetic forces.

Our results also indicate that the tilt of the Pf1 transmembrane (TM) helix is not entirely governed by the hydrophobic mismatch (22). While the helix tilt angles values are close to those determined for Pf1 at the parallel to B₀ orientation of the bilayer normal in glass plates, biphenyl bicelles, and flipped bicelles (24), recent molecular dynamics simulations for DOPC/DOPG bilayers (25) reported a greater tilt angle α = 29° that is comparable to 30° observed for the magnetically aligned bicelles formed by 14-O-PC lipids (23). Had the hydrophobic match alone been indeed responsible for the tilt angle, the ratio of the hydrophobic thicknesses in the two alignment media would have been proportional to the ratio of the cosines of the corresponding helical tilt angles (22). By assuming a constant length of the TM helix, the ratio of the cosines of the two experimental angles is 0.98, whereas the ratio of the hydrophobic thicknesses for DMPC (21) (25.0 Å) and DOPC (26) (27.2 Å) bilayers is 0.92. If one assumes the classical values of Lewis and Engelman (27) for the DMPC and DOPC bilayer thickness of 23.0 Å and 29.5 Å, respectively, an even-lower ratio of 0.78 is obtained.

A possible reason for this discrepancy for Pf1 is the presence of the extracellular helix and residues Lys 20 and Lys 45 that are flanking the TM domain. These positively charged residues are expected to interact electrostatically with the negatively charged DOPG headgroups more strongly than with the zwitterionic DMPC. Such interactions have been observed in recent molecular dynamics simulations of Pf1 bilayer system (25). An independent study has also found lysines to be responsible for stabilizing the TM helix conformation (28) in addition to the hydrophobic mismatch.

In conclusion, we have demonstrated the feasibility of high-resolution solid-state NMR spectroscopy of the uniformly ¹⁵N-labeled Pf1 coat protein reconstituted in nanoporous AAO substrate-confined lipid bilayers. Two different lipid compositions have revealed some marked deviations in the positions of the spectral peaks attributed to structural changes. This is consistent with the hypothesis that folding of the membrane-spanning proteins is lipid-dependent (29). Thus, the improved AAO alignment technique provides a general method for studying lipid-induced structural conformations of membrane proteins under physiologically relevant conditions by OS ssNMR.