Magic Angle Spinning and Oriented Sample Solid State NMR Structural Restraints Combine for Influenza A M2 Protein Functional Insights

Abstract

As a small tetrameric helical membrane protein, the M2 proton channel structure is highly sensitive to its environment. As a result, structural data from a lipid bilayer environment has proven to be essential for describing the conductance mechanism. While oriented sample solid state NMR has provided a high resolution backbone structure in lipid bilayers, quaternary packing of the helices and many of the sidechain conformations have been poorly restrained. Furthermore, an understanding of the quaternary structural stability has remained a mystery. Here, the isotropic chemical shift data and interhelical cross peaks from magic angle spinning solid state NMR of a liposomal preparation strongly support the quaternary structure of the transmembrane helical bundle as a dimer of dimers structure. The data also explains how the tetrameric stability is enhanced once two charges are absorbed by the His37 tetrad prior to activation of this proton channel. The combination of these two solid state NMR techniques appear to be a powerful approach for characterizing helical membrane protein structure.

M2 protein from Influenza A is an important drug target and has been the subject of numerous structural efforts. The protein has shown a high degree of structural sensitivity to the sample environment and hence the need to pursue studies in an environment similar to that of the native membrane.¹ As a tetramer, M2 forms a proton channel that shuttles protons across the membrane. The channel has a primary gate formed by four Trp41 residues and a secondary gate formed by four Val27 residues near the entrance to the pore.^2,3 The shuttling is achieved by a tetrad of His37 residues once activated by binding a third proton and hence a third charge.⁴ An amphipathic helix at the C-terminus of the transmembrane (TM) pore is associated with viral budding resulting from its insertion into the lipid interface.⁵ This helix is also necessary for achieving native conductance properties of the full length protein.⁶ It is important to note that M2 functions at the neck of the viral bud⁵ and consequently does not appear to be solubilized and functioning in the high cholesterol/sphingomyelin membrane environment that dominates the viral envelope. Consequently, we continue to study M2 in lipids that form liquid crystalline environments. Here, magic angle spinning solid state NMR (MAS ssNMR) provides numerous structural restraints in the same lipid environment used for oriented sample (OS) ssNMR spectroscopy that provided a high resolution backbone structure of the M2 conductance domain (residues 22–62) including both the TM and amphipathic helices.⁷ The MAS chemical shift data and distance restraints provide unique insights into the stability of this tetrameric structure and the functionality of the histidine tetrad.

Early backbone structures of the M2 TM domain (residues 22–46) in lipid bilayers were achieved by OS ssNMR without (PDB #1NYJ) and with (PDB #2H95) the antiviral drug amantadine.^8,9 This latter structure was refined recently with MAS ssNMR data by Hong and coworkers (PDB #2KQT).¹⁰ Also crystal structures of the TM domain in detergent environments have been achieved in recent years with and without amantadine (PDB #3BKD, 3C9J, 3LBW).^11,12 In this same timeframe solution NMR structures of the conductance domain (residues 18–60) in detergent micelles were achieved (PDB 2RLF, 2KIH, 2KWX)^13,14 and the recent structure by OS ssNMR structure of the conductance domain in liquid crystalline bilayers (PDB #2L0J).⁷

The backbone structure of the TM domain has progressed towards a consensus. Functionally important details of the conserved HxxxW and the amphipathic helices elucidated in lipid bilayers however, have not achieved such consensus. In 2006 the pKas of the histidine tetrad were determined identifying a very high proton and charge affinity by this channel with two pKas at 8.2.⁴ The much higher than normal proton affinity suggests that the structure is stabilized by these charges and yet His37 is in the low dielectric environment near the center of the membrane suggesting substantial chargecharge repulsion for these closely packed charges. Such repulsion would destabilize the tetrameric structure and yet it has been shown that at pH 6.5 where the histidine tetrad carries two charges the tetramer is much more stable.⁶ It has been suggested that the charges may be extensively delocalized over a dimer of dimer structure.^7,15 Two identical pKas further suggest cooperative proton binding and hence a conformational change following the binding of the first proton to the histidine tetrad, but the conformational change associated with cooperative binding has not been previously identified.

None of the solution NMR or crystal structures in a detergent environment, nor the ssNMR data obtained from M2 in a cholesterol and sphingomyelin rich environment^16,17 have observed this dimer of dimer configuration with imdazole-imidazolium hydrogen bonding as described in the OS ssNMR structure. However, Griffin and coworkers recently published MAS ssNMR data obtained from frozen and specifically labeled samples that show multiple backbone sites generating resonance pairs including His37 and Trp41.¹⁵ In so doing the dimer of dimer conformation for the HxxxW tetrad was confirmed and they further suggest that the dimerization extends throughout the TM domain.

NCOCX and NCACX 3D data sets were recorded for the M2 conductance domain in a pH 7.5 preparation of DOPC/DOPE (4:1 molar ratio) liposomes representing the same model membrane environment as that used for the OS ssNMR (sample and spectroscopic details in the SI). Sample assignment strips are shown in Fig. S1 and the resonance assignments for the TM helix are presented in Table S1. These assignments are consistent with the selective ¹⁵N and ¹³C labeling results from the Griffin and Hong labs when lipid bilayers capable of forming liquid crystalline environments have been used. The resonances in the TM helix are significantly narrower than those in the amphipathic helix and hence we have focused the data analysis effort here on the TM helix.

TALOS+¹⁸ was used to calculate the backbone torsion angles from the backbone chemical shifts for an 18 residue stretch from Leu26 through Leu43. The predictions (Table S2) generated average ϕ, ψ angles of −62.8° and −41.7° that are remarkably close to the average ϕ, ψ angles in 2L0J, which for the same residues is −60.9° and −43.5°. The OS ssNMR values are closer to those expected for TM helices (ϕ = −60°, ψ = 45°) for which intrahelical hydrogen bonding is stronger.¹⁹

Figure 1 displays the aliphatic cross peaks to carbonyl, aromatic and aliphatic sites at various mixing times for ¹³C-¹³C correlation spectra of M2 22–62 (see SI for details). These spectra immediately identify a major challenge for MAS spectroscopy of α-helical membrane proteins. The chemical shifts of the hydrophobic sidechains have remarkably little dispersion. This coupled with the fact that of the 21 residues in the M2 TM helix (residues 26–46), 15 are either alanine (2), valine (2), isoleucine (5) or leucine (6) residues making it very challenging to uniquely interpret cross peaks involving these resonance clusters. The 1 ppm average ¹³C linewidths are excellent for liposome preparations. Overall, it would be very difficult to achieve a well-restrained structure from these spectra alone in this native-like liposome environment. While the backbone structure of the conductance domain as well as the helical tilt and rotation angles were determined to high resolution by orientational restraints, limited data was available for the sidechain conformations and the quaternary structure. Fortunately, there are multiple unique residues in the TM helix: Ser31, Gly34, His37, Trp41, Asp44, and Arg45 that are resolved, uniquely assigned, and can be used for obtaining distance restraints from the MAS ¹³C-¹³C correlation spectra.

Figure 1.

¹³C-¹³C correlation spectral regions obtained from a liposomal preparation of M2 22–62. All spectra were obtained at 600MHz using a ¹³C optimized ¹H/¹³C/¹⁵N 3.2 mm biosolids MAS probe with Low-E coil technology.^20,21 A) Carbonyl-aliphatic region, 9ms (blue) and 50 ms (red) mixing times; B) Aromatic-aliphatic region, 9ms (blue) and 100 ms (red) mixing times; C) Aliphatic-aliphatic region, 20 ms (blue) and 50 ms (red) mixing times.

Six unique inter-helical restraints were obtained from mixing times between 20 and 200 ms. These restraints and their multiplicity in this tetrameric structure are tabulated in Table S3 and illustrated in Fig. 2. In 2L0J the Ser31 and Ala30 Cβ groups are in van der Waals contact via inter-helical packing while considerably separated in their intra-helical interactions and therefore, the cross peak (18.8 and 63.2 ppm) that is visible in the 20 and 50 ms spectra (Fig. 1C) is assigned to the four-fold symmetric inter-helical distance. The remaining unique cross peaks leading to distance restraints were observed through His37 and mostly from HisB/D (chains B and D). The two sets of His37 sidechain resonances (forming a dimer of dimer conformation) have a greater (HisB/D: Cα 59.5, Cβ 31.1, Cγ 137.3, Cδ2 119.4, Cε1 136.4 ppm) and lesser (HisA/C: Cα 62.4, Cβ 32.1, Cγ 131.0, Cδ2 115.6, Cε1 136.7 ppm) fractional charge consistent with the histidine titration data.²² These chemical shifts provide an opportunity to observe cross peaks between the His37 residues as well as unique cross peaks from His37 to other sidechains.

Figure 2.

Distance restraints (red dashed lines) in the TM region of the M2 protein. Carbon colors: Ala30 (brown); Ser31 (cyan); Ile33 (magenta); His37 (green); and Leu38 (yellow). A) Side view of the structure; B) View down the pore axis from the C-terminal end of the TM helix.

The dynamics of the two imidazoles in each pair appear to be quite different. For HisB/D, the Nε2 donates a hydrogen bond to HisA/C Nδ1 and in addition the Nδ1H of HisB/D forms an intra-residue electrostatic interaction with the backbone carbonyl, thereby significantly stabilizing the HisB/D sidechain. The HisA/C Nε2H forms an interaction with the π electron density of the Trp41 indole rings, but this interaction does not strongly restrain either χ1 or χ2 because of the expanse of indole π electron density. In the ¹³C-¹³C correlation spectra the spectral intensities for HisB/D are significantly stronger and are also observed at longer mixing times than those for HisA/C. In particular, the Cδ2 of HisA/C displays only weak signals in these samples. Similarly, all of the Trp41 indole cross peak intensities are very weak and in the aromatic cross peak region the resonances mostly disappear for mixing times of 100 ms or greater. The implied dynamics for the Trp41 indoles further corroborate the previous conclusion about the dynamics of the His37A/C that interacts with Trp41.⁷

In addition to the cross peaks between Ala30 and Ser31, and between imidazole (HisA/C) and imidazolium (HisB/D), HisB/D displays cross peaks with Leu38A/C and Ile 33B/D. Each of these distance restraints is twofold symmetric resulting in a total of 16 restraints with 14 of them being inter-helical for the quaternary structure. Three of the unique distance restraints are from His 37B/D Cγ, Cδ2, and Cε1 to a Leu Cγ. While the Leu resonance assignments have some ambiguity the distances in 2L0J show that Leu38A/C provides the shortest distances (by ~2Å) to His37B/D and hence the specific assignment for these distances. In addition to the helical structure and its orientation relative to its environment from orientational restraints, these distance restraints provide additional restraints to precisely define the quaternary structure of the TM domain (Fig. 2). These restraints are consistent with both the OS ssNMR structure (2L0J) as well as the recent x-ray crystal structure of the TM domain (3LBW), however the Ala30-Ser31 distance is inconsistent with the solution NMR structures (2RLF, 2KWX)^13,14 that have somewhat different helical packing near the N-terminus.

The two sets of resonances observed for the His37 residues support the unique imidazole-imidazolium dimer arrangement based on the pH titration of Nδ1 and Nε2 labeled M2 TM domain that defined a strong Nδ1-H-Nε2 hydrogen bond between these sidechains.⁴ The difference in chemical shifts between the charged histidine amino acid and neutral ε2 protonated tautomer for the Cδ2 (5.8 ppm), Cγ (−9.0 ppm), and Cε1 (1.0 ppm) sites²² are significantly larger than the chemical shift differences observed between HisB/D and HisA/C in M2 22–62, Cδ2 (3.8 ppm), Cγ (−6.3 ppm), and Cε1 (−0.3 ppm) sites reported here. This reduction in the chemical shift differences suggests that the Coulombic charge is not localized to a single imidazole but is dispersed as hypothesized based on the need for minimizing the electrostatic repulsion.^4,7 At the same time the unique chemical shifts for these two His37 sidechains rule out four fold symmetric structures, such as the recent crystal structure (3LBW). However, the M2 solution NMR structures represent time-averaged structures at 303K and consequently may reflect an average of the two conformations. Figure 3 shows a comparison of the ¹³C-¹³C correlation spectra for M2 22–62 at pH 7.5 and 6.0 illustrating that there is no significant change in the Cα and Cβ chemical shifts of His37 and Trp41 over this pH range. Consequently, there does not appear to be a change in the ionization state over this pH range. Previously the third pKa was characterized from the TM domain (residues 22–46) samples as being 6.3. It has been documented by analytical ultracentrifugation that the TM domain has much lower structural stability (via K^app) than the full length protein.⁶ Consequently, we anticipate that for the conductance domain and full length protein the third pKa will be lower than pH 6.3.

Figure 3.

¹³C-¹³C correlation spectra from liposomal preparations of M2 22–62. A) pH 6.0, 9 ms mixing; B) pH 7.5, 20 ms mixing. Otherwise as described in Figure 1.

The assigned chemical shifts for the His37 residues and distance restraints involving these residues support the unique histidine tetrad structure in 2L0J. The formation of imidazole-imidazolium hydrogen bonds explains analytical ultracentrifugation data showing that the tetrameric stability increases as the pH is lowered from 9 to 6.5 by three orders of magnitude for the full length protein.⁶ Without these h-bonds it is not possible to explain the increased stability while charge repulsion is increased due to the addition of two charges in the histidine tetrad.

As suggested previously⁷ the dimer of dimer conformation formed by the imidazole-imidazolium h-bonds may undergo conformational exchange in which the pairs of imidazoles sharing a h-bond, exchange with those that do not share a h-bond. The recent four-fold symmetric crystal structure has Cα carbons separated by 8.6 and 8.7Å, while the separation of Cα carbons in a QM/MM refined imidazole-imidazolium pair for the 2L0J structure is 9.0 to 9.1Å. It appears that the formation of the imidazole-imidazolium h-bond forces the backbone apart and in so doing facilitates the binding of the second proton, hence the cooperativity in binding the first two protons to the histidine tetrad. The structure suggests that the separation of His37 Cα carbons that do not form imidazole-imidazolium pairs could be significantly less than 9.0Å resulting in a rectangular placement of the Cα carbons. This could explain the data supporting a dimer of dimer structure that is observed for many sites in the polypeptide backbone.¹⁵

The unique histidine tetrad forms a dimer of dimer conformation so that two charges can be stabilized in the middle of the lipid bilayer in this small helical bundle. Charge repulsion is minimized by an Nδ1-H Nε2 hydrogen moving back and forth over a low energy barrier between two imidazoles. Here, it is also apparent that the charge is not equally distributed over the two rings, that the proton resides preferentially with the Nε2 site, consistent with the known propensity for Nε2 to bind protons preferentially over Nδ1.²²

MAS and OS ssNMR spectroscopies are highly complementary. Here, the same sample preparation used previously for the OS ssNMR samples has been used for the MAS samples. The combination of orientational and distance restraints take advantage of both approaches and eliminates the need to completely define the structure using one approach. Considering the lack of MAS spectral dispersion for α-helical membrane proteins and the difficulty in characterizing the aliphatic sidechains with OS ssNMR, this combination represents a powerful approach. Here, numerous inter-helical restraints as well as isotropic chemical shift data from MAS ssNMR spectroscopy were obtained that constrain the tetrameric structure. These restraints have allowed us to significantly advance our knowledge of the structure, dynamics and functionality of this important drug target.

ABBREVIATIONS

OS ssNMR

Oriented Sample solid-state NMR

MAS ssNMR

Magic Angle Spinning solid-state NMR

QM/MM

Quantum mechanical/ Molecular Mechanics

Low-E

Low-Electric Field

DOPC

dioleoylphosphatidylcholine

DOPE

dioleoylphosphatidylethanolamine