INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS ANALYSIS OF 2D SPECTRA RESULTS AND DISCUSSION REFERENCES

The efficient utilization of structural constraints and incorporation of all structural information into structural modeling is an ongoing challenge. The solid-state NMR method, polarization inversion spin exchange at magic angle (PISEMA), which correlates ¹⁵N-¹H dipolar interactions with ¹⁵N chemical shifts (Wu et al., 1994), provides high-resolution orientational constraints. Here these constraints are obtained for the transmembrane peptide of M2 protein from influenza A virus, M2-TMP. Through assignments made by single-site labels, these high-resolution constraints are combined with the known -helical structure of M2-TMP to enhance a structural model and to determine specific torsion angles.

While distance and torsional constraints can be obtained by solid-state NMR from unoriented samples, orientational constraints in this paper are obtained from samples uniformly aligned with respect to the magnetic field axis of the NMR spectrometer. Such constraints have numerous advantages for assembling structures and for obtaining a structure that is oriented with respect to its environmentin this molecular system, a lipid bilayer (Fu and Cross, 1999; Quine, 1999).

The two-dimensional PISEMA pulse sequence, developed by Opella and co-workers, has been successfully used to obtain anisotropic ¹⁵N chemical shifts and directly bonded ¹⁵N-¹H dipolar interactions (Wu et al., 1994; Marassi et al., 1997; Tian et al., 1998). This spectroscopy of uniformly aligned samples yields much improved dipolar resolution and a favorable dipolar scaling factor (0.816) compared to earlier separated local field (SLF) spectroscopy (Hester et al., 1976; Waugh, 1976). The scheme uses conventional I-S cross-polarization (CP) followed by frequency-switched Lee-Goldburg (LG) cycles (Bielecki et al., 1990; Lee and Goldburg, 1965). During the LG cycles, the I-spins are locked along the magic angle, 54.74° relative to the magnetic field, and matched by a phase-alternated spin-lock field applied to the S-spins. The Fourier transform of the NMR signal against acquisition time, t₂, and evolution time, t₁, yields a 2D spectrum with chemical shifts in the ₂ dimension and dipolar splittings in the ₁ dimension.

The M2 protein (97 amino acids) from the influenza A virus is an integral membrane protein with proton channel activity in the viral coat (Lamb et al., 1994; Wang et al., 1994). Its single transmembrane -helix is contained within the 25-amino acid peptide M2-TMP (residues 22-46), which has been reported to show channel activity similar to that of native M2 protein (Duff and Ashley, 1992). However, recent studies of various truncated proteins have brought this result into question (Tobler et al., 1999). The transmembrane portion of both M2 and M2-TMP primarily adopts an -helical structure, and M2 forms a homotetramer in lipid environments, as characterized by several studies (Duff et al., 1992; Holsinger and Lamb, 1991; Sakguchi et al., 1997; Kovacs and Cross, 1997). M2-TMP also appears to form an oligomer. The helix tilt of a monomer should be sensitive to the thickness of the lipid bilayer, but M2-TMP has been shown to be quite insensitive to a change in bilayer thickness (Kovacs et al., 2000). Therefore, the helix tilt appears to be an intrinsic propensity of an oligomer. In addition, the oligomer is not a heterogeneous aggregate, inasmuch as single site labels give rise to a single identical resonance from each monomer. M2-TMP is therefore presumed to be a tetrameric state, similar to M2 protein, but the oligomeric state of M2-TMP has not been specifically defined.

In previous investigations of isotopically labeled M2-TMP, orientational constraints derived from NMR and infrared spectroscopy were interpreted to show that the helices are tilted by 32°-38° with respect to the lipid bilayer normal (Kovacs and Cross, 1997; Kukol et al., 1999; Kovacs et al., 2000). Moreover, model building of the tetramer suggested that the -helical bundle is left-handed (Kovacs and Cross, 1997). The channel function of M2 can be blocked by amantadine, potentially through binding to a specific sequence (residues 27-31) in its transmembrane domain (Skehel, 1992; Hay, 1992; Wang et al., 1993). Here a preliminary binding experiment of amantadine with M2-TMP is described.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS ANALYSIS OF 2D SPECTRA RESULTS AND DISCUSSION REFERENCES

Sample preparation

Several ¹⁵N-labeled amino acids were purchased from Cambridge Isotope Lab (Cambridge, MA). M2-TMP (NH₂-Ser²²-Ser²³-Asp²⁴-Pro²⁵-Leu²⁶- Val²⁷-Val²⁸-Ala²⁹-Ala³⁰-Ser³¹-Ile³²-Ile³³-Gly³⁴-Ile³⁵-Leu³⁶-His³⁷-Leu³⁸- Ile³⁹-Leu⁴⁰-Trp⁴¹-Ile⁴²-Leu⁴³-Asp⁴⁴-Arg⁴⁵-Leu⁴⁶-CO₂H) was obtained by solid-phase synthesis, using fluorenoylmethoxy carbonyl chemistry on an Applied Biosystems 430A Synthesizer. Amino acid blocking, purification, and characterization of the peptides were performed as previously described (Kovacs et al., 2000).

To orient a lipid bilayer preparation, M2-TMP and dimyristoylphosphatidylcholine (DMPC) or dioleoylphosphatidylcholine (DOPC) were codissolved in trifluoroethanol (TFE) with a peptide-to-lipid molar ratio of 1:8 and 1:30, respectively. Then the sample was spread onto ~60 thin glass plates (75 µm × 10.5 mm × 10.5 mm). After vacuum drying to remove TFE, the glass plates were stacked into a glass tube (11 mm × 11 mm), and the samples were hydrated with high-performance liquid chromatography-grade water (~50% by weight). Finally, the samples were sealed and incubated at ~45°C for several days.

NMR experiments

The NMR measurements were performed on a spectrometer with a 9.4-T magnet, operating at a ¹⁵N Larmor frequency of 40.585 MHz. The probe was constructed in house and has a rectangular coil suited to the sample size. The oriented M2-TMP samples were placed with the order axis (i.e., the lipid bilayer normal) parallel to the magnetic field axis. All ¹⁵N chemical shifts are relative to the resonance for a saturated solution of ¹⁵NH₄NO₃ at 0 ppm.

For the PISEMA experiment, the cross-polarization (CP) period was 1 ms. The rf field strengths were typically 31.4 kHz for the CP match and 38.5 kHz for the Lee-Goldburg (LG) condition, corresponding to a LG time interval of t_m = 26 µs. A delay of 1 µs was used at the onset of each ±LG cycle to compensate for the frequency synthesizer (Programmed Test Sources type) switch time, which was found to be critical for achieving the theoretical dipolar scaling factor (0.816). The t₁ value was incremented from 0 to 24 Lee-Goldburg cycles, and the refocused ¹⁵N-signal was acquired with ~2000 transients for each t₁ increment. For data processing, 512 and 256 points in the t₂ and t₁ dimensions were used, respectively, and exponential line broadening of 100 Hz was used in the t₂ dimension, but not in the t₁ dimension. The spectral symmetry in the dipolar dimension was achieved by setting the imaginary portion of the data points to zero before the Fourier transform against t₁.

	ANALYSIS OF 2D SPECTRA

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS ANALYSIS OF 2D SPECTRA RESULTS AND DISCUSSION REFERENCES

For spectral analysis, several relevant orientation reference frames are defined (Fig. 1) as follows: P, the ¹⁵N chemical shift anisotropy (CSA) principal axis frame with the principal tensor elements ₁₁, ₂₂, ₃₃; Nh, the frame of the ¹⁵N-¹H dipolar interaction tensor with its unique z axis along the ¹⁵N-¹H bond; D, the director frame with its z axis along the lipid bilayer normal. Any two of these coordinate systems are related by Euler rotations (, , ).

View larger version (13K): [in this window] [in a new window]   FIGURE 1   Definition of the 15N chemical shift principal axis frame (PAS) and the 15N-1H dipolar interaction tensor with respect to the sample director axis. This latter axis is the bilayer normal in the samples studied here, and it is arranged to be parallel to B0.

In our oriented samples, the z axis of the director frame is parallel to the magnetic field, B₀, and the observed chemical shift can be written in terms of tensor magnitudes and orientations as

(1)

where the Euler angles _PD, _PD relate the CSA principal axis frame to the sample director frame. Notice that _PD = 0.

The ¹⁵N-¹H dipolar splitting can be described by

(2)

where  = (µ₀/4) (_HNh/2r_NH³) is the dipolar coupling constant; _H and _N are the gyromagnetic ratios of ¹H and ¹⁵N, respectively; h is Planck's constant; and r_NH is the N-H bond length. The angle  is the polar angle relating the N-H internuclear vector to B₀. Correlating with the CSA frame yields

(3)

where the Euler angles (_PD, _PD) and (_PNh, _PNh) refer to the transformation from the CSA principal axis frame to the director frame and to the dipolar tensor frame, respectively.

Taking the ¹⁵N-¹H dipolar coupling constant to be  = 10.735 kHz, based on an N-H bond length of 1.041 Å, and assuming that the ¹⁵N amide tensor orientations relative to the ¹⁵N-¹H bond are typically _PNh = 0°, _PNh = 17° (Mai et al., 1993), Eqs. 1-3 can be solved for each site, using pairs of (_obs, ) values obtained from 2D PISEMA and the corresponding CSA tensor values ₁₁, ₂₂, ₃₃. The resulting  and (_PD, _PD) values relate the N-H bond and the CSA principal axis frame, respectively, to the director z axis. These angles give important orientational constraints for structural analysis. In general, four possible _PD, _PD solutions, (± _PD, _PD), ( ± _PD,   _PD), may be derived from the combination of Eqs. 1-3, while two solutions, (,   ), may be derived from a positive dipolar splitting value according to Eq. 2. For two adjacent residues, such as Val^27-28 and Ile^32-33, the orientations of the principal axis frames for each residue can be used to obtain a list of possible ,  torsion angles that correspond to the data for those residues.

The PAF₁ is defined as the CSA principal axis frame for the first residue (either Val²⁷ or Ile³²) with the B₀ orientation (_PD1, _PD1), and the PAF₂ is the CSA principal axis frame for the second residue (Val²⁸ or Ile³³) with the B₀ orientation (_PD2, _PD2). For the peptide plane geometry in Fig. 2, the supplements of the bond angles C-C₁-N and C₁-N-C are 65° and 59°, respectively, and the tetrahedral geometry at C is such that the supplement of the angle N-C-C is 70°. Unit vectors in the direction of the C₁-N and N-C bonds in PAF₁ are denoted by u₁ and u₂, and the unit vectors in the direction of the C-C₁ and C₁-N bonds in PAF₂ are denoted by u₃ and u₄.

View larger version (10K): [in this window] [in a new window]   FIGURE 2   Bond direction unit vectors of a peptide plane and bond angles used for the calculations of backbone torsion angles.

Based on four possible _PD, _PD solutions for each pair of residues, the torsion angles can now be computed by using orientational constraints and a torsion angle identity:

(4)

where

(5)

for unit vectors v₁, v₂, and v₃ (Quine, 1999). Here, arg(x, y) denotes the complex argument of (x, y), which is the angle between 180° and 180° formed by the x axis and the point (x, y) in the plane. The key for minimizing ambiguities is to compute the torsion angles by using only dot products of vectors, to avoid the need for changing frames between PAF₁ and PAF₂. The necessary dot products are

(6)

The scalar triple product in Eq. 5 does, however, give an ambiguity, known as a chirality ambiguity (Quine, 1999; Quine et al., 1997), when it is computed using dot products:

(7)

where
Thus these equations yield 16 possible torsion angle pairs for every choice of (_PD1, _PD1) and (_PD2, _PD2). Here we will minimize this set based on the known -helical secondary structure.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS ANALYSIS OF 2D SPECTRA RESULTS AND DISCUSSION REFERENCES

The chemical shift tensor element magnitudes have been determined by using single site labeled samples of M2-TMP. While the tensors were not determined in a hydrated lipid bilayer environment, they were determined as a powder prepared from a trifluoroethanol solution in which the samples are -helical. Therefore, the tensors are characterized for the sites of interest in the conformation of interest (Table 1).

The PISEMA spectra of oriented M2-TMP samples with selective ¹⁵N-labeling are displayed in Figs. 3, 5, and 6. The 2D contour plots display ¹⁵N chemical shifts in the horizontal dimension and ¹⁵N-¹H dipolar splittings in the vertical dimension, with the scale expanded by 1.22 to account for the experimental scaling of the dipolar dimension.

View larger version (14K): [in this window] [in a new window]   FIGURE 3   PISEMA spectrum of oriented 15N-Ile32,33,35,39,42-labeled M2-TMP in hydrated DMPC lipid bilayers. A conventional 1D spectrum is shown above. The slices on the left side represent the 15N-1H dipolar splittings corresponding to the chemical shifts indicated by the arrows in the contour plot. Chemical shifts are referenced to saturated 15NH4NO3 solution at 0 ppm.

Fig. 3 shows the PISEMA spectrum of ¹⁵N-Ile^{32,33,35,39,42}-labeled M2-TMP in hydrated DMPC lipid bilayers. For comparison, the conventional 1D chemical shift spectrum for this sample is shown at the top. While the peaks of four Ile sites between 118 and 129 ppm cannot be resolved in the 1D spectrum, they are resolved in this 2D spectrum and in the corresponding dipolar slices on the left side.

The spectral results are compared in Table 2 to the chemical shift data from single site labeled M2-TMP in similar bilayer preparations (spectra not shown). The assignments for Ile³³, Ile³⁵, and Ile⁴² and Val²⁷ and Val²⁸ are achieved from the single site labels. However, without the single site dipolar information, the assignments for Ile³² and Ile³⁹ are tentative, based on a novel assignment strategy that recognizes a helical wheel pattern of resonances in the PISEMA spectra (Wang et al., 2000). Note that, while the chemical shift for Ile⁴² is within the error bar for Ile³⁹ and Ile³², the dipolar splitting from the single site label for Ile⁴² of 13.8 kHz confirms this assignment.

Normally, the sign of the dipolar splitting is considered to be ambiguous if is less than half-maximum (see Eq. 2). Fig. 4, however, shows that these peptide samples undergo an axial motion about the bilayer normal. Consequently, when this peptide sample is oriented with the bilayer normal perpendicular to the field, a single sharp line is observed rather than a powder pattern (a powder pattern would be seen if axial motion had not occurred). This means that the motionally averaged ¹⁵N chemical shift and ¹⁵N-¹H dipolar tensors are collinear. Because all of the _obs obtained are greater than their corresponding _iso values (Table 1), the values are all positive (Tian et al., 1998). This substantially reduces the potential structural ambiguities in the following analysis. Note that such a motion about an axis parallel to B₀ for all of our samples, except Fig. 4 B, has no effect on the NMR observables.

View larger version (14K): [in this window] [in a new window]   FIGURE 4   1D spectra of oriented 15N-Leu43-labeled M2-TMP in DMPC with the bilayer normal parallel (a) and perpendicular (b) to the magnetic field direction. The isotropic frequency is ~95 ppm. If global rotation about the bilayer normal had not occurred, a broad powder pattern would be observed for the perpendicular orientation. Furthermore, note that this global rotation has no impact on the observed orientational constraints obtained from samples aligned with the bilayer normal parallel to the magnetic field direction. The signal at 25 ppm is likely to be natural abundance 15N from the choline headgroup. Because of the extensive dynamics of this choline site, the reproducibility of this intensity is poor, as it is very difficult to cross-polarize.

Based on these data, a variety of qualitative results can be identified. First, as previously noted (Kovacs and Cross, 1997; Kukol et al., 1999; Kovacs et al., 2000), the chemical shift and dipolar splitting values are not consistent with either a helix parallel to the bilayer normal or one perpendicular to it. Because the N-H axis and ₃₃ tensor elements are approximately parallel to the helix axis, helices that are parallel to the bilayer normal give rise to chemical shifts near ₃₃ and to values that are nearly maximal. In contrast, helices that are perpendicular give rise to chemical shifts near ₁₁/₂₂ and dipolar splittings that are at half-maximum and negative. The observed orientational constraints are variable and far from these extreme values; therefore, the helix has a considerable tilt.

Second, the substantial variation in dipolar splitting for nearly the same chemical shift (e.g., Ile⁴² versus Ile³⁹) illustrates the noncollinearity of the static ¹⁵N chemical shift and ¹⁵N-¹H dipolar tensors. The ₃₃ element is typically 17° from the N-H axis (Harbison et al., 1984; Oas et al., 1987; Mai et al., 1993). Third, the similarity of the orientational constraints for residues i and i + 7 (e.g., Ile³² and Ile³⁹, Ile³⁵ and Ile⁴²) suggests considerable uniformity for this -helix. Even differences such as 8.8 kHz versus 4.4 kHz in dipolar splitting suggest a difference as small as 8° in the orientation of the N-H axis.

To investigate the structural influence of the channel inhibitor amantadine, the PISEMA spectra of oriented ¹⁵N-Val^27,28-labeled M2-TMP in hydrated DMPC bilayers with and without amantadine were acquired. For comparison, they are shown in Fig. 5 as overlaid spectra. The two splittings from Val²⁷ and Val²⁸ are well resolved in both chemical shift and dipolar dimensions. From conductivity measurements (Wang et al., 1993), the binding constant for amantadine has been established to be 3 × 10⁶ M¹. Therefore, the addition of an equal molar amount of M2-TMP and amantadine should result in a predominance of the complex. It has been suggested that amantadine binds in the vicinity of the Val^27,28 residues along the channel pore (Skehel, 1992; Hay, 1992; Wang et al., 1993). From our previous results and model (Kovacs and Cross, 1997), Val²⁷ should line the pore of the channel, while Val²⁸ is oriented toward an adjacent helix. The observed differences in the dipolar splittings and in the chemical shifts are within their error bars for these two preparations, as clearly seen in the dipolar and chemical shift slices through the PISEMA spectra. Consequently, there is no evidence for a significant structural or orientational change for either site upon amantadine binding. Therefore, no evidence is presented here for amantadine interacting at these sites and directly blocking the channel pore. However, the lack of a change in the backbone structure and orientation does not refute this possibility, because amantadine is likely to be interacting directly with the side chains and not the backbone.

View larger version (22K): [in this window] [in a new window]   FIGURE 5   Overlay of PISEMA spectra of oriented 15N-Val27,28-labeled M2-TMP in hydrated DMPC lipid bilayers with (b) and without (a) amantadine bound. Dipolar and chemical shift slices are displayed for each resonance with and without amantadine bound. The negative contour artifacts near zero frequency in the dipolar dimension are deleted for clarity.

The improvement in spectral resolution shown in the chemical shift slices of the PISEMA spectra compared to the one-dimensional (1D) chemical shift spectra is nearly a factor of 2. Several different lipid environments and molar ratios have been tried to improve the 1D spectral resolution; however, little effect has been observed. PISEMA spectra of ¹⁵N-Val²⁷-labeled M2-TMP in DOPC and DMPC bilayers are shown in Fig. 6. The slight difference in chemical shift and dipolar splitting possibly reflects a very small change in helix tilt due to the increased hydrophobic thickness of the DOPC bilayers. No difference, however, is seen in the linewidths, although they are substantially better than the 1D chemical shift spectra. Data from a T₂ measurement on this sample in DOPC yields a T₂ value of 750 µs, which suggests an intrinsic line width of 10 ppm (Fig. 7).

View larger version (21K): [in this window] [in a new window]   FIGURE 6   Overlay of PISEMA spectra of oriented 15N-Val27-labeled M2-TMP in hydrated DMPC and DOPC lipid bilayers. Dipolar and chemical shift slices are presented for the resonances in the two lipid environments (1,3 for DMPC; 2,4 for DMPC). The inserted peak corresponds to that for DMPC, plotted at a lower contour level and showing clearly the diagonally distributed resonance broadening.

View larger version (20K): [in this window] [in a new window]   FIGURE 7   T2 measurements of oriented 15N-Val27-labeled M2-TMP in DOPC bilayers at room temperature, obtained with cross-polarization and a variable delay Hahn echo.

The line width in 1D spectra or in a projection of the PISEMA spectra in the chemical shift dimension displays a resonance width of ~20 ppm. In the 2D PISEMA spectra the heterogeneous broadening appears along a diagonal axis through 0 kHz at _iso = 96 ppm. Heterogeneous broadening in these 2D spectra will be constrained within the powder pattern of an unoriented sample, and because of the global motion this powder pattern lies on these diagonal axes. Thus the improved resolution in the chemical shift slices of the 2D PISEMA spectra (Fig. 6) results from a minimization of the heterogeneous broadening.

In Table 2, the  angles derived from dipolar splittings are presented with the _PD and _PD angles determined with Eqs. 1-3. In Fig. 8, the dipolar splitting (a) and chemical shift (b) are plotted as a function of orientation angles _PD, _PD (ranging from 0 to 360° and 0 to 180° respectively), using the , _PNh, _PNh values mentioned above and typical ₁₁, ₂₂, ₃₃ values from the powder pattern of ¹⁵N-Val²⁷-labeled M2-TMP. Both and _obs change more dramatically as a function of _PD than as a function of _PD. For _obs, this reflects the small asymmetry parameter of the amide ¹⁵N CSA (typically   0.2). If  equaled 0, ₁₁ would equal ₂₂ and _PD would have no effect on _obs. For , it reflects the effect of both  and _PNh. If _PNh were also zero, then essentially would be independent of _PD. Because _PNh is significant (typically 17°), the influence of _PD on generally is more pronounced. This partially explains why the five Ile sites, particularly Ile³², Ile³⁹, and Ile⁴², could not be resolved by their chemical shifts alone, but were resolvable by their dipolar splittings.

View larger version (54K): [in this window] [in a new window]   FIGURE 8   Orientation-dependent 15N-1H dipolar splitting (a) and chemical shift (b) as function of Euler angles PD, PD in M2-TMP. The CSA tensor values used in b are from the 15N Val27-labeled M2-TMP powder sample. The dots represent all possible solutions from Eqs. 1-3 before the elimination of ambiguities for the site in M2-TMP.

Because there is a limited data set from which it is not yet possible to calculate a unique molecular structure, we will take advantage of the knowledge that this polypeptide is predominantly -helical (Duff et al., 1992; Kovacs and Cross, 1997), that the helices have an average tilt of 37° ± 3° (Kovacs et al., 2000), and that the -helical geometry is well defined (Quine, 1999; Kovacs et al., 2000). The ambiguity in  presented in Table 2 places the N-H orientation between 0° and 54.7° or 125.7° and 180°. For an -helix the  values must all be in one range or the other, because the N-H bond makes an angle in the range of ±17° with respect to the helix axis, which in turn makes an angle of 37° with the magnetic field axis. Moreover, this same argument eliminates all _PD values greater than 90°. Therefore, only a single sign ambiguity remains in _PD for each of the peptide planes, as shown in Table 3.

Many of the possible torsion angles calculated from Eqs. 4-7 can be eliminated based on steric hindrance information for Val and Ile residues in the -helical region. Based on Ramachandran plots, -branched amino acids have a more limited -helical region than do other amino acids. In this analysis, the allowed -helical region is taken from Creighten (1984). In Table 3, torsion angles within these dimensions are in boldface; borderline values within 10° of this region are in plain type and outlying values are italicized. The most probable torsion angles, thus, are limited to (83°, 60°) and (136°, 60°) for Val²⁷ and (81°, 30°) for Ile³².

The transformation from one peptide plane to the next is defined by a (, ) torsion angle pair. From this transformation, the local helix axis direction vector for two adjacent peptide planes can be computed (Quine, 1999). Using _PD and _PD to define the magnetic field direction vector B, the tilt angle, , can be obtained from the dot product of the helix axis direction vector and B. Comparing computed tilts with the accepted tilt of 37° ± 3° provides another way to filter the possible orientations of B. In Table 3, tilts within the range of 37° ± 10° are in boldface; borderline values (37° ± 15°) are in plain type; and outlying values are italicized. Tilt calculations for Val²⁷ support the validity of the (, ) solutions and _PD, _PD solutions while solving the chirality ambiguity. For Ile³², the unique torsion angle solution is supported and the chirality is solved.

Enhancement of spectral resolution has been achieved through 2D PISEMA of oriented M2-TMP samples. Information about ¹⁵N chemical shift and ¹⁵N-¹H dipolar interactions can be analyzed to derive specific site orientations leading to specific peptide plane orientations. In turn, these orientations lead to the structural solution for backbone peptide ,  torsion angles in the transmembrane domain of M2 transmembrane peptide. From this solution set, helix axis orientations consistent with those previously described were found. Furthermore, the M2-TMP structure at residues Val^27,28 appears not to be distorted by amantadine binding.