Structure of the Influenza C Virus CM2 Protein Transmembrane Domain Obtained by Site-specific Infrared Dichroism and Global Molecular Dynamics Searching*

The 115-residue protein CM2 fromInfluenza C virus has been recently characterized as a tetrameric integral membrane glycoprotein. Infrared spectroscopy and site-directed infrared dichroism were utilized here to determine its transmembrane structure. The transmembrane domain of CM2 is α-helical, and the helices are tilted by β = (14.6 ± 3.0)° from the membrane normal. The rotational pitch angle about the helix axis ω for the 1-13C-labeled residues Gly59 and Leu66 is ω = (218 ± 17)°, where ω is defined as zero for a residue pointing in the direction of the helix tilt. A detailed structure was obtained from a global molecular dynamics search utilizing the orientational data as an energy refinement term. The structure consists of a left-handed coiled-coil with a helix crossing angle of Ω = 16°. The putative transmembrane pore is occluded by the residue Met65. In addition hydrogen/deuterium exchange experiments show that the core is not accessible to water.

The 115-residue protein CM2 from Influenza C virus has been recently characterized as a tetrameric integral membrane glycoprotein. Infrared spectroscopy and sitedirected infrared dichroism were utilized here to determine its transmembrane structure. The transmembrane domain of CM2 is ␣-helical, and the helices are tilted by ␤ ‫؍‬ (14.6 ؎ 3.0)°from the membrane normal. The rotational pitch angle about the helix axis for the 1-13 Clabeled residues Gly 59 and Leu 66 is ‫؍‬ (218 ؎ 17)°, where is defined as zero for a residue pointing in the direction of the helix tilt. A detailed structure was obtained from a global molecular dynamics search utilizing the orientational data as an energy refinement term. The structure consists of a left-handed coiled-coil with a helix crossing angle of ⍀ ‫؍‬ 16°. The putative transmembrane pore is occluded by the residue Met 65 . In addition hydrogen/deuterium exchange experiments show that the core is not accessible to water.
The virus Influenza is known from the past till today as the causative agent of the most deadly disease outbreaks. Recently, the CM2 protein of Influenza C virus (1) has been characterized as an integral membrane glycoprotein that forms disulfide linked dimers and tetramers. Based on the overall topology containing a 23-residue extracellular part, a 23-residue membrane spanning part, and a 69-residue cytoplasmic tail, CM2 is assumed to be structurally similar to the Influenza A M2 protein and the Influenza B NB protein (2,3). The transmembrane domains of M2 and NB both form ion channels in lipid membranes (4,5), and the M2 H ϩ channel is blocked by the anti-Influenza drug amantadine. Because these small viral membrane proteins form possible targets for drugs, structural data may facilitate the development of new antiviral drugs.
Because of the lack of high resolution x-ray data, molecular modeling is commonly used to obtain structural models of transmembrane proteins (6,7). The M2 transmembrane domain has been a popular choice for molecular modeling efforts (8,9). We have applied the recently developed method of sitedirected infrared dichroism (10) to the M2 transmembrane domain resulting in orientational data (11), which is in accordance with results obtained using solid state NMR (12). Further, the orientational data were used in a global molecular dynamics search to obtain a detailed structure that is in agreement with all functional and mutagenesis data (11).
Here we present the first structural data about the CM2 transmembrane domain. Initially, we characterized its secondary structure by infrared spectroscopy and showed that it is mostly ␣-helical. Subsequently, the approach of site-directed infrared dichroism and molecular modeling was used to determine the helix tilt, the relative orientation of the helix within the helix bundle, and a detailed structure based on this orientation.

MATERIALS AND METHODS
Peptide Purification and Reconstitution-Synthetic peptides corresponding to the predicted transmembrane domain of the Influenza C/Ann Arbor/1/150 virus CM2 sequence were made by solid-phase Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry, cleaved from the resin with trifluoroacetic acid, and lyophilized. The sequences corresponding to residues 47-76 (numbering includes a 24 residue signal peptide) are respectively. The peptides were further purified as described elsewhere (11) for analogous transmembrane peptides. Briefly, the peptide was dissolved in trifluoroacetic acid and purified by reversed phase high pressure liquid chromatography (Jupiter 5C4-300 Å column, Phenomenex, Cheshire, United Kingdom). Peptide elution was achieved with a linear gradient to a final solvent composition of 5% H 2 O, 20% trifluoroethanol, 28% acetonitrile, and 47% 2-propanol (Biocad Sprint, Perceptive Biosystems, Cambridge, MA). All solvents contained 0.1% (v/v) trifluoroacetic acid. After lyophilization of pooled fractions, the peptide was reconstituted into lipid vesicles by dialysis from a solution of 5% (w/v) ␤-octylglycopyranoside (Melford Laboratories, Ipswich, UK) and 12.5 mg/ml dimyristoylphosphocholine (Sigma) against 0.1 mM phosphate buffer, pH 7 (11).
Infrared Spectroscopy-Fourier transform infrared (FTIR) 1 spectra were recorded on a Nicolet Magna 560 spectrometer (Nicolet, Madison, WI) equipped with a high sensitivity liquid nitrogen-cooled MCT/A detector. Attenuated total reflection-FTIR spectra were measured with a 25 reflections attenuated total reflection accessory from Graseby Specac (Kent, UK) and a wire grid polarizer (0.25 M, Graseby Specac). 200 l of sample (ϳ2.5 mg/ml protein and 12.5 mg/ml lipid) were dried onto a germanium trapezoidal internal reflection element (50 ϫ 2 ϫ 20 mm) under a stream of nitrogen. After extensively purging the instrument with dry nitrogen, 1000 interferograms were averaged for every sample and processed with 1 point zero filling Happ-Genzel apodization. Transmission spectra were obtained by drying 50 l of sample on a CaF 2 window with a 15-mm diameter.
Fourier self-deconvolution (13) was applied to the spectra in the amide I region to separate the overlapping 12 C and isotope-shifted 13 C (14) amide I peaks. The enhancement factor used in Fourier selfdeconvolution was 2.0, and the half-height bandwidth was 13 cm Ϫ1 , as reported previously (15). The dichroic ratio, R, was calculated as the ratio between the integrated absorption of parallel and perpendicular polarized light of the absorption bands (between 1670 and 1645 cm Ϫ1 for the helix, 12 C, and 1610 and 1630 cm Ϫ1 for the site-specific label, 13 C. The site-specific dichroic ratio, R site , was corrected for the 1.1% natural abundance 13 C as described (11). D 2 O exchange was performed by * This work was supported by the Wellcome Trust and BBSRC. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Data Analysis-The data were analyzed according to the theory of site-specific dichroism presented in detail elsewhere (10) with the extensions described (11). Briefly, the measured dichroic ratio, the absorption ratio between parallel and perpendicular polarized light R ϭ A ʈ /A Ќ of a particular transition dipole moment is a function of its spatial orientation. For the amide I mode (mainly the CAO bond vibration) of an ␣-helical protein the geometric relation between the transition dipole moment and the helix is known from fiber diffraction studies (16). Therefore, by measuring the orientation of the amide I transition dipole one can determine the helix tilt angle ␤ and the rotational pitch angle of the specific dipole moment about the helix axis. The rotational pitch angle is arbitrarily defined as 0°when the transition dipole moment, the helix director, and the z axis all reside in a single plane. Thus, measuring the site-specific dichroic ratio R site of the 13 C amide I mode from a particular label and the helix dichroic ratio R helix allows calcu-lation of the helix tilt ␤ and the rotational pitch angle of a particular label as detailed previously (10), if measurements from two samples with labels at different are analyzed together. Note that the difference of between two consecutive residues is assumed to be 100°as in a canonical ␣-helix (17). To enhance the 13 C amide I mode intensity we introduced two labels at position i and i ϩ 7 assuming that they posses the same as described elsewhere (18).
Molecular Modeling-A global search with respect to rotation about the helix axis, assuming tetrameric symmetry, was carried out as described elsewhere in detail (11,19). In brief, all calculations were performed with the parallel processing version of the Crystallography and NMR System (Version 0.3) (20) using the OPLS parameter set with a united atom topology that explicitly represents the polar hydrogen and aromatic side chain atoms (21). All calculations were carried out in vacuo with the initial coordinates of a canonical ␣-helix (3.6 residues/ turn). Symmetric tetramers were generated from the sequence YMLT-LASLGLGIITMLYLLV, acetylated at the N terminus, and methylaminated at the C terminus by replicating the helix and rotating it by 360°/4 around the center of the tetramer. An initial crossing angle of 25°f or left-handed and Ϫ25°for right-handed structures was introduced by rotating the long helix axis with respect to the long bundle axis. The symmetric search was carried out by applying a rotation to all helices simultaneously between ϭ 0°and ϭ 360°in 10°steps. Four trials were carried out for each starting structure, using different initial random atom velocities in each case at both right-and left-handed crossing angles yielding 36 ϫ 4 ϫ 2 equaling 288 structures. Each structure was subjected to a simulated annealing and energy minimization protocol. Clusters of similar structures were defined such that the root mean square deviation of the coordinates between all structures within a cluster was not larger than 1.2 Å; a cluster was formed by a minimum of 12 structures. For each cluster an average structure was calculated, energy was minimized, and then the structure was subjected to the same simulated annealing protocol used in the systematic search.
Orientation Refinement-To take into account the results obtained

TABLE I Measured dichroic ratios
Dichroic ratios for the helix R helix , the 13 C-labeled site R site , and lipid asymmetric stretching (CH 2 ) modes R lipid for the two CM2 peptides. Also shown are the calculated lipid order parameters S lipid ϭ (3cos 2 ␤ Ϫ 1)/2 (24), the helix tilt angle ␤, and the rotational pitch angle about the helix axis. from the site-directed dichroism analysis, we have incorporated an orientation refinement energy term in all molecular dynamics and energy minimization calculations as described elsewhere (11). According to the experimental data, a helix tilt restraint was applied. To account for the rotational pitch angle, four site-specific dichroism restraints were applied by setting the angles between the 13 CAO bonds of Gly 59 , Leu 66 , Gly 61 , and Leu 68 and the z axis to those obtained from the experiment. These angles are a function of the helix tilt ␤, the rotational pitch angle (10). This restraint causes structures that deviate from the experiment to be higher in energy. The rotational pitch angles of the resulting structures were determined from a geometric analysis with a self-written program.

RESULTS AND DISCUSSION
FTIR Measurements-The transmission FTIR spectra of the CM2 peptide reconstituted in lipid vesicles (Fig. 1, A and B) are indicative of an ␣-helical structure. FTIR spectroscopy has become a standard tool to determine the secondary structure of proteins (15), and even subtle differences in helical conformations of bacteriorhodopsin have been monitored (22,23). Quantitative estimates of the secondary structure may be obtained by curve-fitting individual components contributing to the amide I absorption and assigning each component to a certain type of secondary structure (15). The analysis yields a total of 68% ␣-helix, 28% unordered, and 3.4% ␤ structure, the latter of which we attribute to not properly reconstituted peptide. This compares well with the 20-residue predicted transmembrane part (2) comprising 69% of the synthesized 29-residue peptide.
To estimate the amount of protein that is not accessible to water and therefore embedded in the membrane, the hydrogen/ deuterium exchange was measured using the amide II absorption band (Fig. 1C). The amount of exchange after a 4-h incubation in D 2 O is 34% of the total amide II absorption. This indicates that the transmembrane part consists of 19 -20 residues that are not accessible to water.
The attenuated total reflection-FTIR spectra of the [ 13 C 1 ]Gly 59 /Leu 66 -and [ 13 C 1 ]Gly 61 /Leu 68 -labeled peptide shown in Fig. 2 are indicative of a perpendicular orientation of the ␣-helix relative to the membrane plane, because the absorbance obtained with parallel polarized light is more intense than the absorbance obtained with perpendicular polarized light. The dichroic ratios for the helix and the 13 C-labeled sites averaged from seven different measurements for each peptide are shown in Table I. The order parameter S, used extensively in conventional infrared dichroism analysis (24), is a measure  of the average orientation of molecules. An order parameter of S ϭ 1 is indicative of an orientation perpendicular to the membrane plane, whereas S ϭ Ϫ0.5 indicates an orientation parallel to the membrane plane. The order parameters S calculated for the helix and the lipid bilayer (Table I) are indicative of a well ordered lipid bilayer with helices possessing a net trans-bilayer orientation. The quantitative analysis of the helix and site-specific dichroic ratios yielded a helix tilt angle ␤ ϭ (14.6 Ϯ 3.0)°and a rotational pitch angle Gly 59 /Leu 66 ϭ (218 Ϯ 17)°and Gly 61 /Leu 68 ϭ (58 Ϯ 17)°. Note that Gly 61 /Leu 68 ϭ Gly 59 /Leu 66 ϩ 200°, as in canonical ␣-helix (17). Global Molecular Dynamics Search-A model built using the transmembrane sequence of CM2, YMLTLASLGLGIITM-LYLLV, was subjected to a molecular dynamics search protocol assuming a tetrameric symmetrical helix bundle. The molecular dynamics calculations were carried out applying the experimental constraints. The distribution of structures as a function of the energy and the helix rotation parameter is shown as a polar plot in Fig. 3. The arcs represent the individual structures that are clustered into groups of similar structures.
From these clusters average structures were calculated and the energy was minimized. They are shown as numbered circles. The negative energy is given by the distance from the origin. In Fig. 3 and Table II it can be seen that structure number 2 has the lowest energy. Geometric analysis of the structures reveals that structure 2 has a rotational pitch angle ϭ 224° (Table  II), which is in excellent agreement with the experimental value ϭ (218 Ϯ 17)°. It can be seen from the helix crossing angle ⍀, that structure 2 forms a left-handed coiled helical bundle. Structure 11, which is right-handed, with ϭ 197°, is the next closest to the experimental value but is ruled out not only because of the fact that it is outside the experimental error range, but it also has a much higher energy than the lefthanded structure 2. Accordingly, structure 2 was chosen as the correct model for the CM2 transmembrane domain.
Structure Description and Biological Implications-The structure in Fig. 4 reveals a left-handed coiled-coil tetramer where the residues Leu 55 , Leu 58 , Met 65 , and Leu 68 are pointing toward the pore. The structure is tilted by 15°from the membrane normal. The pore is occluded mainly by Met 65 and partially by Leu 55 as it can be seen from the space filling model in Fig. 5. This is in agreement with the data from the deuterium exchange experiments, which revealed that the transmembrane part is not accessible to water. By analogy to the Influenza M2 virus proton channel it has been speculated that CM2 has ion channel function as well (2). Based on our structural model, this hypothesis can be supported if we assume a closed conformation of this ion channel with Met 65 forming part of the gate. The closed conformation is very likely to be present in a partly dehydrated lipid membrane, where it is assumed that the lipids are in a gel phase compared with the more fluid liquid crystalline phase present under in vivo conditions. To observe a conformational change to the open conformation, a simulation time in the order of milliseconds would be necessary possibly including the lipid bilayer as well (25), which is not yet feasible with the current speed of computers.
It is interesting to note that the hydrophilic residues Thr 54 , Ser 57 , and Thr 64 are not lining the channel pore but are located at the outer surface. This is in accordance with other ion channel structures forming mostly hydrophobic pores, e.g. the K ϩ channel (26) or the HIV-I virus vpu protein (18).
The N-terminal extracellular part forms a wide entrance with the first constriction at Leu 55 (Fig. 5). This entrance could be a possible target for antiviral drugs blocking the channel as it is assumed for amantadine in the Influenza M2 virus proton channel (27).