Determinants of Hepatitis C Virus p7 Ion Channel Function and Drug Sensitivity Identified In Vitro^▿

DNA constructs.

pGEX6P-1 (Pharmacia)-based bacterial expression constructs for FLAG-tagged J4 and JFH-1 p7 fused to glutathione-S-transferase have been described previously (4, 12). Chimeric J4/JFH-1 constructs and J4 p7 containing point mutations were generated by overlap/fusion PCR using appropriately designed primers (sequences available on request) and ligation of amplimers into pGEX6P-1 digested with EcoRI/NotI. All constructs were confirmed by double-stranded DNA sequencing.

Protein expression and purification.

Glutathione S-transferase-FLAG-p7 proteins were expressed in Escherichia coli, purified from inclusion bodies, cleaved, and FLAG-p7 purified by reverse-phase high-performance liquid chromatography (HPLC) as previously described (4, 36). Protein molecular weights were confirmed by mass spectrometry.

Detection of recombinant p7.

Western blot analysis of purified p7 was carried out as described previously (4, 36) using a commercial anti-FLAG mouse monoclonal antibody (M2; Sigma) or rabbit polyclonal serum raised to the carboxy terminus of J4 p7 (PPRAYA; no. 1055) (11), the JFH-1 p7 amino terminus (ALEKLVVLHAAS; no. 2715) (12), or JFH-1 p7 carboxy terminus (PRQAYA; no. 2717). Rabbit sera were peptide affinity purified, and both antibodies 2715 and 2717 were further concentrated ∼50-fold by ammonium sulfate precipitation. Secondary antibodies conjugated to horseradish peroxidase were purchased from Sigma.

Native gel analysis of recombinant p7.

Ten micrograms of lyophilized p7 from HPLC fractions was reconstituted in either 50 μM liposomes (see “Liposome dye release assay and drug inhibition studies” below) or 300 mM detergent (1,2-diheptanoyl-sn-glycero-3-phosphocholine [DHPC], dodecylphosphocholine [DPC], 1-myristoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)] [LMPG], or 1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)] [LPPG]) and then adjusted to 150 mM Tris-Cl (pH 7.0), 30% glycerol, and 0.05% bromophenol blue prior to native polyacrylamide gel electrophoresis (PAGE) analysis on a 4 to 20% Tris-glycine gel (Invitrogen). Samples were not heated prior to analysis. Protein was visualized by staining with Coomassie brilliant blue.

CD of recombinant p7 in methanol.

Circular dichroism (CD) spectra were collected at 1-nm intervals using a Jasco J175 spectropolarimeter (Jasco International Co., Ltd.) equipped with a Peltier thermostat, with a 1-mm path length quartz cuvette over the wavelength range of 190 to 260 nm. The protein samples were between 0.25 and 0.3 mg/ml in methanol. Values for methanol alone were subtracted for averaging and baseline correction.

Liposome flotation on discontinuous Ficoll gradients.

Liposome flotations were carried out as previously described (36). Briefly, duplicate 90-μl batches of a protein-liposome mixture containing 5 μg of protein and 50 μM rhodamine-labeled liposomes were incubated at room temperature for 30 min. Ten microliters of assay buffer (36) or 1 M Na₂CO₃ (pH 11.4) was then added prior to a further 5-min incubation at room temperature. The mixture was then added to an equal volume of 40% Ficoll-assay buffer (wt/vol) and layered at the bottom of a 2.2-ml ultracentrifuge tube. This was then overlaid with 1.6 ml of 10% Ficoll-assay buffer (wt/vol) and then with 0.3 ml of assay buffer and spun at 100,000 × g for 1 h at room temperature. Gradients were fractionated, liposomes were detected by rhodamine fluorescence, and 20 μl from each fraction was subjected to Western blot analysis as described previously (36).

Liposome dye release assay and drug inhibition studies.

Liposome dye release assays were performed as described previously (36). Assays are based on the real-time release of carboxyfluorescein from self-quenching concentrations within liposomes, resulting in dilution and subsequent fluorescence gain. Briefly, standard 100-μl assay mixtures contained 5 μg of purified protein dissolved in MeOH (giving ∼5 μM final concentration) mixed with 50 μM carboxyfluorescein (CF)-liposomes (dioleoyl phosphatidic acid-phosphatidyl choline from Avanti polar lipids at 1:1 as described previously [36]). Inhibitors were added to reaction mixtures prior to temperature elevation. Baselines were determined using liposome-only and solvent controls (5% MeOH for addition of protein and/or 0.2% dimethyl sulfoxide for inhibitors). Release of CF was measured using a FLUOstar Optima plate reader (BMG Labtech) set to 37°C. Excitation and emission wavelengths were set at 485 nm and 520 nm, respectively. Fluorescence values were read every 30 s over a 30-min period in triplicate, with each experiment repeated at least twice on separate protein batches. Initial rates were calculated from the first 5 min of the reaction and expressed as change in relative fluorescent units per min (ΔRFU/min) as a percentage of the untreated control. The bee venom peptide mellitin (Sigma-Aldrich) at 10 nM and 1% Triton X-100 were used as positive controls for CF release. Amantadine-HCl was purchased from Sigma, and rimantadine-HCl was a kind gift from Helen Bright (formerly of GlaxoSmithKline, Stevenage, United Kingdom).

Molecular modeling of p7 ion channel complex.

Models of the J4 p7 heptameric channel were constructed using Maestro (Schrödinger Inc.). Monomers were built amino acid by amino acid with energy minimization (magnetomotive force in a simulated water dielectric), docked into a heptameric structure, and then subjected to further energy minimization. Images of resultant structures were obtained by either Maestro or Pymol, version 0.9 (Delano Scientific).

Contribution of the p7 trans-membrane helices and loop regions to drug sensitivity.

Recently, we found that p7 from different HCV isolates displayed discrete sensitivity patterns to inhibitors in vitro, which directly correlated with drug effects on particle secretion in culture (12). Models of the p7 channel structure heavily implicate the amino-terminal trans-membrane helix or the basic loop as sites for potential drug interactions as it is thought that their hydrophilic surfaces form the channel lumen (4, 13, 25). We reasoned, therefore, that chimeric proteins where these domains were exchanged would confer drug sensitivity relating to the genotype from which the amino-terminal helix/loop was derived.

To test this hypothesis, we constructed chimeric proteins using the J4 (genotype 1b) (43) and JFH-1 (genotype 2a) (20, 40) sequences as these display differential sensitivity to adamantane compounds: rimantadine blocks the activity of both proteins, whereas amantadine is only effective against J4 p7 (12). Importantly, both proteins are equally active in our dye release assay, and the chemistry of the two adamantane compounds is extremely similar. As both the helices and the cytosolic loops of J4 and JFH-1 differed in primary amino acid sequences, we generated a panel of four chimeric proteins (C1 to C4) (Fig. 1A). These combined the amino-terminal helix of one isolate with the carboxy-terminal helix of the other, in the context of either the J4 or JFH-1 cytosolic loop. Each protein was expressed to similar levels in bacteria, and purified chimeric p7 was confirmed by Western blotting using genotype-specific antibodies to the p7 amino or carboxy termini (Fig. 1A) (see Materials and Methods).

The compound sensitivity of each chimera was then tested in a liposome-based fluorescent dye release assay, a convenient alternative to labor-intensive suspended bilayer systems (36) (Fig. 1C). Amantadine and rimantadine were used at 1 and 5 μM, based on the ability of amantadine to effect an ∼80 and ∼100% inhibition of the J4 protein at these concentrations (Fig. 1B). Chimeric proteins had reproducibly similar activities overall to parental proteins, as tested for at least three separate preparations of each (data not shown), but to accurately compare the inhibitory effects of increasing inhibitor concentration on each protein, activity is expressed as the percent initial rate of dye release compared to untreated samples in replicate experiments (Fig. 1C). Baseline levels of dye release were determined from solvent-only controls. As shown previously, J4 and JFH-1 p7 proteins displayed differential sensitivity to amantadine, with the JFH-1 protein remaining insensitive to the drug up to a 5 μM concentration (Fig. 1C, top panels) (12). Unexpectedly, chimeric proteins did not acquire the characteristics of one or other of the parental proteins and instead displayed intermediate sensitivity to both compounds (Fig. 1C, middle and bottom panels). While each chimera was more resistant to amantadine at 1 μM than J4 p7, the chimeras were also less sensitive to rimantadine than either parental protein. Improved inhibition was seen by increasing the drug concentration to 5 μM for all but the C4 chimera, which retained up to 50% activity in the presence of both compounds at this concentration. In addition, the C1 chimera showed increased resistance to both compounds present at 1 μM yet was strongly inhibited at higher drug concentrations. In each case, a degree of inhibition was seen for both compounds, as is the case for the J4 protein, yet no clear pattern emerged as to which elements were responsible for resistance to one or the other drug. Taken together, these data suggest that residues responsible for p7-drug interactions may act in a context-dependent fashion, dictated by the overall structure of the channel.

Design and expression of p7 point mutants in vitro.

As data from chimeric proteins suggested that the contribution of individual residues within p7 may be context dependent, we designed a series of alanine substitutions targeting conserved residues in the J4 background thought to play a role in channel function and so, possibly, therefore drug interactions (Fig. 2A, top panel). The mutations fell into two categories: those changing charged/hydrophilic residues thought to line the channel lumen and those targeting residues predicted to affect channel stability. Within the first group, the most obvious targets were the basic charges on the cytosolic loop (K33A/R35A), which are required for channel activity in cells (14) as well as particle assembly (19, 34). We also changed three conserved hydrophilic residues within the amino-terminal helix, including a highly conserved histidine residue at position 17 (H17A, S21A, and C27A). Helical wheel models predict that these form one face of the amphipathic helix, thus lining the lumen of an oligomeric p7 channel (1, 4, 25). The second group of mutations comprised residues thought to stabilize the oligomeric channel structure: three phenylalanine residues in the amino-terminal helix (F22A/F25A/F26A) and a conserved glycine in the carboxy-terminal helix (G39A) thought to interact with adjacent protomers, a conserved proline predicted to form a “kink” in the carboxy-terminal helix (P49A), and, lastly, a leucine-rich region implicated in oligomerization (L50A to L55A [L(50-55)A]). All proteins were successfully expressed and purified from E. coli, verified using a J4-specific anti-p7 antibody (no. 1055) (see Materials and Methods) as well as the anti-FLAG monoclonal antibody (Fig. 2A). Western analysis revealed the sodium dodecyl sulfate (SDS)-induced formation of higher-molecular-weight forms of each protein as we have observed previously (Fig. 2A) (4).

The basic loop and a conserved histidine within the amino-terminal helix are required for p7 channel activity.

We previously demonstrated that p7 is capable of functionally replacing the influenza A virus M2 protein in a surrogate cell system (14) and observed increased activity at lower pH for the J4 p7 protein in vitro (36). These observations strongly imply that p7 may act as a proton channel. In accordance with our findings in the M2 surrogate assay, mutation of the highly conserved basic charges on the cytosolic loop of J4 p7 resulted in a protein unable to mediate dye release from liposomes (Fig. 2B). Although a role for these residues in channel opening is likely, additional effects of this mutation are described in greater detail below.

The gating of M2 involves a conserved histidine at position 37, the protonation of which is critical for channel opening (30, 41). The p7 amino-terminal helix contains three hydrophilic residues (or clusters of residues in some isolates such as genotype 2a JFH-1), which are predicted to line the lumen and so are implicated in channel gating (Fig. 1 and 2). We mutated these residues to alanine in the J4 p7 sequence and tested the effect on channel opening in the liposome assay. Interestingly, of these, only the histidine mutation at position 17 caused a decrease in p7 channel activity (Fig. 2B). Neither of the other two hydrophilic residues (S21A or C27A mutations) had any effect on channel activity, nor did they alter susceptibility to amantadine (Table 1). This combined with our previous observations indicates that the pH-dependent mechanism governing the opening of p7 channels is similar to that of M2.

Mutations predicted to affect p7 stability affect both channel function and drug susceptibility.

The hydrophobic interactions postulated to stabilize p7 oligomers were targeted by mutating key residues (Fig. 2A). One potential mechanism for p7 oligomerization is thought to proceed via the extended side chains of phenylalanine residues projecting into pockets in the hydrophobic surface of trans-membrane helices resulting from the small side chains of glycine residues (1). We therefore mutated three conserved phenylalanine residues in the p7 amino-terminal helix, as well as a conserved glycine in the carboxy-terminal helix, to alanine. Mutating the phenylalanine residues individually had no effect on channel activity in the liposome assay (data not shown); however, the triple mutation had the surprising effect of increasing p7-dependent dye release (Fig. 2B). This was reproducible for multiple protein preparations, yet no obvious increase in the amount of protein incorporated into liposomes was evident (data not shown), suggesting that the specific activity of the channel itself had been altered. In contrast, the G39A mutation significantly reduced p7 channel activity. Interestingly, neither of these mutations affected the ability of p7 to form higher-molecular-weight structures in the presence of SDS (Fig. 2A).

An additional level of stability may be conferred to p7 channel structures by a predicted kink in the carboxy-terminal helix caused by a highly conserved proline residue at position 49. This kink is thought to allow the carboxy terminus of the helix to bend around adjacent monomers and form close-knit hydrophobic interactions (13, 25). The precise type of interaction this brings about is unknown although many HCV isolates possess a leucine-rich region toward the end of the carboxy-terminal helix which may act as a hydrophobic “zipper,” as increasingly documented for oligomeric channel proteins (8, 44). We therefore mutated both motifs and assessed their effects on channel activity. Surprisingly, whereas the P49A markedly reduced channel activity, the L(50-55)A mutation retained wild-type activity yet showed an increased resistance to 1 μM amantadine (Fig. 2B).

Impaired mutant channel activity is not due to defects in protein folding or oligomerization.

It was necessary to demonstrate that the effects of the alanine substitutions with impaired activity were due to the formation of inactive channels rather than an inability of the protein to fold or self-associate. Although SDS-PAGE “laddering” correlates well with the formation of heptameric complexes induced by SDS in solution (4), the denaturing aspects of SDS-PAGE (SDS itself along with heating of the sample) disrupt channel complexes to a degree, thereby reducing the amount of resolvable heptameric species (Fig. 2A). We therefore tested a number of less-denaturing detergents for their ability to induce the formation of p7 oligomers by native PAGE (Fig. 3A, left-hand panel). We found that the phospholipid-like DHPC and DPC were capable of inducing the formation of heptameric J4 FLAG-p7 complexes, whereas neither LMPG nor LPPG did so. Equivalent assemblages were observed when protein was mixed with liposomes, as described for the dye release assay (Fig. 3A, middle panel), making it likely that these species represented a relevant oligomeric form of the channel. Reconstitution of null mutants in DHPC also resulted in heptameric complexes (Fig. 3A, right-hand panel) although the K33A/R35A mutant protein gave a less-well-defined smear, raising the possibility of some heterogeneity or a less efficient self-association.

In addition to their ability to oligomerize, the secondary structure of mutant proteins was also assessed by CD (Fig. 3B). Results are shown for protein dissolved in methanol; this was deemed valid as protein solubilized in this way is functional in liposome dye release assays and so presumably is correctly folded or is, at the very least, in a conformation capable of assuming a native structure upon lipid/detergent incorporation. Indeed, wild-type J4 p7 protein reconstituted in liposomes or DHPC gave identical profiles (data not shown). All mutated proteins gave the same alpha-helical profile as seen for wild-type p7, indicating that no gross defects in secondary structure had arisen.

Dominant negative effects of basic loop and H17A mutants support preassembly of channel complexes prior to membrane insertion in vitro.

Given that all mutated p7 proteins were correctly folded and could oligomerize in DHPC, we next assessed whether their ability to associate with membranes was intact. Protein-liposome mixes were subjected to flotation on a discontinuous Ficoll gradient in the presence or absence of a high-pH wash to strip nonintegrated protein. Fractions were tested for liposome content by rhodamine fluorescence, which indicated that liposomes had migrated to the 10% Ficoll-aqueous interface (Fig. 4A, fraction 11). The presence of high pH did not affect liposome flotation (Fig. 4A, histogram in right-hand panel). Fractions were subjected to Western blotting to assess the amount of liposome-associated protein. As seen previously, a significant proportion of the wild-type J4 p7 associated with liposomes (36). Furthermore, this association was resistant to high pH, indicating that the protein had fully inserted into the liposome membrane. This was also true for p7 mutants with the exception of the K33A/R35A mutant protein, which dissociated from the liposomes following high-pH treatment. This mutation represents the most severe change to the overall charge/hydrophobicity of p7, making this observation perhaps unsurprising. It does, however, make it impossible to assign a role for this conserved motif in channel opening using this system.

We next asked whether inactive mutant proteins could exert a dominant negative effect over wild-type p7 function via the formation of nonfunctional mosaic channels. As this phenotype should manifest if even one protomer of a heptameric channel structure contained the mutation, we performed liposome assays titrating mutant proteins into a wild-type reaction mixture at increments of one-seventh of the total protein content and compared activity to the equivalent amount of wild-type protein alone. We tested both the K33A/R35A and H17A mutants to assess whether any effect could occur in the absence of membrane insertion. Surprisingly, both mutant proteins were able to reduce wild-type dye release activity to nearly baseline levels even when present at only one-seventh of the total protein concentration (Fig. 4B). As the K33A/R35A mutant exerted the same effect as the H17A protein, this strongly supports the idea that, at least in this system, p7 oligomers form at the membrane surface prior to the insertion of intact channel complexes (Fig. 4A).

Titration of p7-null mutants reveals a key role for conserved proline 49 in channel oligomerization.

It was important to demonstrate that the null phenotypes of mutants were not due to concentration-dependent effects. We therefore performed titration experiments with protein inputs from 20 to 200% (1 to 10 μg of protein) of that in a standard reaction mixture (5 μg) and measured activity relative to the wild type. Importantly, liposomes were still present in excess where 10 μg of protein was used since dye release for wild-type p7 did not exceed 40% of either mellitin- or Triton X-100-positive controls (data not shown).

For the K33A/R35A mutant, no activity was observed even at 200% input, implying that this protein is totally incapable of membrane insertion. For the G39A mutation, again no increase in activity was observed over the lowest input, indicating that this mutant protein formed truly nonfunctional channels (Fig. 5A). For the H17A mutant, however, a modest increase in CF release was observed with increasing concentration (Fig. 5A) although this difference was not statistically significant compared to standard reaction conditions.

Unlike the other mutations, the P49A mutation displayed a concentration-dependent switch in activity, shifting from 40% activity under standard conditions to almost threefold wild-type activity at 150% and 200% input (Fig. 5A). To investigate the cause of this switch, we subjected protein from P49A and G39A titration experiments to SDS-PAGE and Western analysis to visualize oligomeric forms under partially denaturing conditions (Fig. 5B). In agreement with the sudden switch in activity for the P49A mutant, a marked increase in the amount of higher-molecular-weight species was observed upon increasing protein input from 5 to 7.5 μg (Fig. 5B, right panel). This was not observed for either the wild-type or G39A mutant proteins (Fig. 5B, left and central panels), implying that Pro 49 may serve to regulate efficient oligomerization of p7 channel complexes.

A carboxy-terminal polyleucine motif contributes to amantadine sensitivity of genotype 1b p7.

As described above, the L(50-55)A mutation appeared to display increased resistance to treatment with 1 μM amantadine. To determine whether this effect was specific, we tested this protein in drug titration assays against increasing concentrations of both amantadine and rimantadine (Fig. 6). While this mutation conferred resistance to 1 μM amantadine, increasing the drug concentration effectively blocked its activity, whereas the protein was equally sensitive to both low and high concentrations of rimantadine (Fig. 6). This pattern of drug sensitivity is reminiscent of the JFH-1 protein (Fig. 1 and 6), albeit to a much lesser degree as higher amantadine concentrations effectively block activity. This may be evidence that p7 possesses distinct sensitivity determinants for these two related compounds or that rimantadine binds with much higher affinity than amantadine.