Analysis of the C-Terminal Membrane Anchor Domains of Hepatitis C Virus Glycoproteins E1 and E2: toward a Topological Model

Materials.

Calculations were performed on RAMSES (Rapid Analysis Master/Slaves Extensible System) clusters of 320 Pentium processor PCs in parallel connected with a 100Mbits/s Ethernet network. The calculation softwares were developed in our laboratory (5, 18). Molecular views and hydrophobicity potentials were drawn with WinMGM 1.0 software (47) from Ab Initio Technology (Obernai, France).

HCV sequences.

The E1 and E2 sequences of the following 25 HCV strains were extracted from PIR, SwissProt, EMBL, and GenBank sequence databases: HCV-BEINNX11 (GenBank accession no. A48711), HCV-BK (SwissProt accession no. P26663), HCV-J (SwissProt accession no. P26662), HCV-J33 (GenBank accession no. D14484), HCV-J4/83 (GenBank accession no. D13558), HCV-J4/91 (GenBank accession no. D10750), HCV-JK1 (PIR accession no. S18030), HCV-JT (GenBank accession no. D11168), HCV-JT" (GenBank accession no. D11355), HCV-L2 (GenBank accession no. U01214), HCV-N (GenBank accession no. S62220), HCV-T (SwissProt accession no. P29846), HCV-Unkcds (EMBL accession no. M96362), HCV-K1-R1 (GenBank accession no. D50480), HCV-K1-R2 (GenBank accession no. D50481), HCV-K1-R3 (GenBank accession no. D50482), HCV-K1-S1 (GenBank accession no. D50483), HCV-K1-S2 (GenBank accession no. D50485), HCV-K1-S3 (GenBank accession no. D50484), HCV-1 (SwissProt accession no. P26664), HCV-H (SwissProt accession no. P27958), HCV-J1 (EMBL accession no. D10749), HCV-G9 (GenBank accession no. D14853), HCV-J6 (SwissProt accession no. P26660), and HCV-J8 (SwissProt accession no. P26661). Note that amino acid numbers always refer to the mature HCV-BEINNX11 glycoproteins. E1 extends from amino acid 1 (= amino acid 192 in the polyprotein) to amino acid 192 (= amino acid 383 in the polyprotein), and E2 extends from amino acid 1 (= amino acid 384 in the polyprotein) to amino acid 363 (= amino acid 746 in the polyprotein).

Kyte and Doolittle's method.

The mean hydrophobicity of a stretch of sequence is calculated by moving a seven-residue window along the sequence (30). The value is attributed to the central amino acid of the window, and the hydrophobicity of each amino acid is defined in the original work (30).

HCA.

The hydrophobic cluster analysis (HCA) plot (21) relies upon the 2-D helical representation of the sequence: the sequence is written on a classical α-helix (3.6 residues per turn) smoothed on a cylinder. To make the view easier to handle, the cylinder is then cut parallel to its axis and unrolled. As some adjacent amino acids are separated by the flattening of the cylinder, the plot is duplicated to restore all the connections of residues. The hydrophobic residues (F, V, L, I, M, Y, and W) are circled and hatched; proline is represented by a star, glycine is represented by a diamond, serine is represented by a square with a dot in it, and threonine is represented by an empty square. This method allows β-strands to be distinguished from α-helices, as the shape and morphology of the hydrophobic clusters are related to these secondary structures: long horizontal hydrophobic clusters denote hydrophobic or amphipathic α-helices, while short vertical or “mosaic” clusters are indicative of β-strands. This representation is also useful in predicting membrane or transmembrane domains.

Peptide structure and conformational search.

Helical peptides were constructed with standard values of φ and ψ angles (−60° and −40°) and subjected to a Simplex minimization (44). For the study of the putative β-pins, conformational searches of the turn were performed using the stereoalphabet procedure (14, 36). The analysis was carried out on 8 residues and in two steps. During the first step, the (φ, ψ) couples of the stereoalphabet were systematically attributed to the first five amino acids of the putative turn while the other residues were fixed as the β-strand. The couples of φ and ψ were as follows: −60°, −40° (right α-helix); −160°, 160° (β-sheet); −140°, 80° (intermediate structure between β-sheet and 3¹⁰ helix); −80°, 160° (helix of the polyproline); −80°, 80° (3¹⁰ helix); and 60°, 60° (left α-helix). From the 7,776 conformations tested, the 20 lowest-energy structures were subjected to a second analysis: the φ and ψ angles of the last five amino acids of the putative turn were tested using all couple-of-angle values of the stereoalphabet (the conformation of the fourth and the fifth amino acids were thus studied twice). For each of the 20 first-step structures, the 10 lowest-energy structures were conserved in the second step (200 structures out of a total of 155,520 calculated conformations). The energy of these 200 structures was minimized using the Simplex method (44). The structure with the lowest energy was considered to be the most-probable structure. The energy of all structures was calculated as the sum of the torsion potential and the Van der Waals, electrostatic, and hydrophobic interactions (35). All calculations take into account the presence of a hydrophobic/hydrophilic interface with the concomitant variation of the dielectric constant (5).

MHP.

The molecular hydrophobicity potentials (MHP) were used to describe the hydrophobic or hydrophilic environment of peptides as previously described (6). For the MHP calculation, the assumption is made that the hydrophobic potential decreases exponentially with the distance as follows:

where N is the number of atoms; Etr_i and r_i are the transfer energy and the Van der Waals radius of the i atom, respectively; and d_i is the distance between the i atom and the point where the potential is calculated.

Orientation towards a hydrophobic-hydrophilic interface.

A hydrophilic center is calculated using the following equation (5):

where r⃗_i and Etr_i are the coordinate and the transfer energy of atom i when only the hydrophilic atoms of the molecule are taken into consideration (phi). A hydrophobic center (C⃗_pho) is calculated using a similar equation taking into account only the hydrophobic atoms. The Cdelta⃗ = Cpho⃗ − Cphi⃗ vector is stated as perpendicular to the interface. The interface position (I⃗) is given by the equation:

IMPALA.

In the IMPALA procedure (18), an empirical function, C, is used to describe the water-lipid interface. Considering that the properties of membranes are constant in the plane of the bilayer, C varies only along the z axis, perpendicular to that plane. The z axis has its origin at the center of the membrane, and the empirical function C is described by the following equation:

where α and z₀ are defined so that C ≈ 1 for z < −18.5 and z > 18.5 and C ≈ 0 for −13.5 < z < 13.5. The interactions between the peptide and the bilayer are described by the sum of two restraint functions as follows:

E_int simulates the hydrophobic effect by taking the accessible surface (S_i), the free energy of transfer per accessible surface unit (ASP_i), and the position (z_i) of each atom i. The general evolution of this equation is an increase when accessible hydrophilic atoms (i.e., Etr > 0) penetrate the membrane and a decrease when accessible hydrophobic atoms do so. The more accessible the atoms are, the greater the effect is. The second function mimics the solvophobic effect in lipids:

where a_lip is an empirical constant equal to 0.018. E_lip tends to minimize the interactions between the peptide and the bilayer. Assuming a constant structure for all peptides, the position and orientation in the membrane were tested by running 10 Monte Carlo procedures of 10⁵ steps at 310 K each. Maximal rotations of 5° and translations of 2 Å per step were allowed. The algorithm used to orient the structures with respect to a hydrophobic-hydrophilic interface (see above) was used to determine the starting position of the peptides. The position of minimal energy was retained.

Sequence analysis.

The hydropathy of E1 and E2 sequences was analyzed to determine which fragments of their C-terminal domains were likely to interact with membranes. From the HCA and the Kyte and Doolittle plots of the E1 protein (Fig. 1), we pointed out that a C-terminal hydrophobic domain extends from proline 137 to glycine 192 (position in the mature protein). For E2, a hydrophobic domain was detected from proline 281 to alanine 363 (position in the mature protein) (Fig. 2).

Alignment of the E1 and E2 C-terminal sequences from 25 HCV strains was performed in order to identify potent transmembrane stretches and/or other structural entities able to interact with membrane. In these alignments, conservation of residues, as well as conservation of their hydrophobicity, charges, etc., was studied.

In the E1 C-terminal domain, several stretches with conserved properties were identified from the C- to the N-terminal extremities (Fig. 3A) as follows. (i) The Val180-Ala188 stretch contains only hydrophobic residues and ends with an aromatic residue (Phe187). In addition, hydrophilic, charged, and coil-favorable residues are located on both sides (Gly175, Asn176, Lys179, Gly189, Asp191, and Gly192). (ii) The Trp162-Tyr171 region has no or few charged and polar residues (a lysine is present in one sequence). It is edged by aromatic residues on both sides (Trp162, Tyr170, and Tyr171) and preceded by a hydrophilic stretch containing a coil-favorable residue (Gly159). (iii) The Ala140-Ala158 region is rather amphiphilic. It contains hydrophobic and charged or polar residues and is preceded by a stretch containing polar residues that are often present in coil or turn structures (Pro137-Thr139).

From the C- to the N-terminal extremities, the following properties were found in the C-terminal domain of E2 (Fig. 3B). (i) The Leu352-Ile358 region contains only hydrophobic residues and an aromatic residue is present at the N-extremity (Trp353). This stretch is surrounded by polar and charged residues (Ala344-Cys351 and Ala359-Ala363). (ii) The Tyr335-Leu343 region contains only hydrophobic residues and starts with an aromatic residue (Tyr335). It is preceded by a charged stretch (Lys332-Glu334). (iii) The Ile313-Ile331 stretch is rather amphiphilic, since it contains hydrophobic and charged or polar residues. (iv) The N-terminal stretch (Pro281-Asn312) is also amphiphilic and is moreover rich in coil- or turn-favorable amino acids (Pro293 and -300; Ser285, -295, and -303; Thr286, -287, -297, -298, and -304; Gly305).

Based on these analyses, we suggest that the C-terminal domains of the E1 and E2 glycoproteins share a similar pattern (Fig. 3 and 4): a long amphiphilic region (18 or 19 residues) followed by two short hydrophobic stretches (7 to 10 residues) surrounded by hydrophilic residues.

An even number of transmembrane stretches is expected to anchor E1 and E2 in the membrane, because the N- and C-terminal extremities are in the lumen of the ER. The two hydrophobic stretches are the best candidates for these transmembrane stretches. It is usually thought that 18 to 20 hydrophobic residues are required to constitute transmembrane α-helices. The hydrophobic stretches of E1 and E2 are 7 to 10 residues long. This is inappropriate for classical transmembrane α-helices. In other words, since the mean distance between two consecutive residues of an α-helix is equal to 1.5 Å in the direction of the helix axis (13), these stretches should be 10.5 to 15 Å as α-helices. This is only half the width of a lipid bilayer. In a β-strand, the distance between two adjacent amino acids is 3.4 Å (13). Therefore, the E1 and E2 stretches would be 23.8 Å long for the shortest strand and 34 Å for the longest strand. This type of configuration seems, at least arithmetically, to be the most likely to cross the membrane (Fig. 4C).

In the amphiphilic region before the putative transmembrane β-strands, the HCA profile shows long horizontal hydrophobic clusters suggestive of amphipathic α-helices (21) (Fig. 4).

In order to test the hypotheses, several 3-D models were built for the HCV-BEINNX11 strain. These models were tested for membrane interaction.

3-D model of amphipathic α-helices.

Helices were built for every stretch of a minimal length of 13 amino acids between Ala140 and Ala158 for E1 and between Ile313 and Ile331 for E2. The energy of each α-helix was minimized using the Simplex method. Each structure was then oriented towards a hydrophobic-hydrophilic interface. Helices eliciting a tilt of less than 30° between their axis and this interface were considered to be amphipathic and were selected for a thorough study using the IMPALA procedure. The 10 IMPALA Monte Carlo simulations gave similar results for each peptide. The MHP were also calculated to check the hydropathy of each structure. The best candidate as amphipathic α-helix was selected as the stretch with a maximal length, an insertion angle close to 0°, and a maximal depth of penetration in the bilayer.

Table 1 summarizes the screening for putative amphipathic α-helices in E1. Most peptides behave as amphipathic α-helices, and Val142-Ala158 was selected as the best candidate (Fig. 5). This 17-residues stretch makes a 10° angle with respect to its hydrophobic-hydrophilic interface and with the IMPALA membrane surface (Fig. 5C). In this position, the mass center of the structure was located at 13.5 Å from the membrane center, indicating that the peptide is able to adsorb at the membrane surface by inserting between the phospholipid heads. The MHP confirm the amphipathy of the helix since the hydrophobic and hydrophilic isopotential envelopes are well segregated at the helix surface (Fig. 5B).

Table 2 summarizes the screening for the E2 glycoprotein. As for E1, most α-helices are amphipathic. The 16-amino-acid Val316-Ile331 peptide was selected as the best candidate, as its angle with respect to its hydrophobic-hydrophilic interface or with the IMPALA modeled lipid bilayer surface is close to 0° (Fig. 6). The mass center of the helix in the membrane is at 12.5 Å from the membrane center. The MHP confirm the segregation of the hydrophobic and hydrophilic faces separately (Fig. 6B).

3-D model of transmembrane β-strands.

To test the hypothesis of a transmembrane β-pin, a β structure was imposed on the residues Trp162 to Tyr171 and Val180 to Ala188 of E1 and on Tyr335 to Leu343 and Leu352 to Ala359 of E2. The conformation of the eight residues located between the two β-strands was then calculated using the stereoalphabet method. For each protein, 155,520 conformations were tested.

The lowest energy structure obtained for E1 is shown in Fig. 7A. The analysis shows that the best conformation of the loop corresponds to an antiparallel pair of β-strands (Trp162 to Met173 and Trp177 to Ala188). Interstrand hydrogen bonds are implicated in the energy stabilization of the structure. MHP indicate that the final structure is highly hydrophobic except at the backbone contact (due to accessible main-chain C=O and N-H groups), in the turn, and at the peptide N- and C- terminal extremities (Fig. 7B). The hydrophobic part is longer than 35 Å, and this should be sufficient to cross a membrane. The results of the 10 IMPALA simulations confirmed this hypothesis. The most favorable position of this structure is transmembranous (Fig. 7C), with the soluble and adsorbed states being unlikely.

Figure 8Ashows the E2 transmembrane β-pins after the stereoalphabet procedure, and once again, the β-strands are antiparallel with interstrand hydrogen bonds (Tyr335 to Asp345 and Cys349 to Ile358). As for E1, the structure is highly hydrophobic, with hydrophilic patches in the turn, at the backbone contact, and at the N- and C-terminal ends, as shown with the MHP (Fig. 8B). The hydrophobic part of the structure is 35 Å long, and IMPALA simulations show that the position of minimal energy is transmembranous (Fig. 8C).

Conclusions.

In addition to their role in membrane association, the E1 and E2 transmembrane domains (the 31 and 29 last amino acids, respectively) are responsible for heterodimerization and retention of the two HCV envelope glycoproteins in the ER compartment. This is crucial for the HCV life cycle. We put forward the hypothesis that the E1 and E2 transmembrane domains adopt, at least transiently, a transmembrane β-pin conformation downstream of an amphipathic α-helix. The β-structure, the multimerization, and the ER retention properties of the transmembrane stretches would be correlated to their length. To our knowledge, this model is in agreement with results from the literature. Further experimental results will be needed to validate or invalidate our model.