Unusual architecture of the p7 channel from hepatitis C virus

Many viruses have developed integral membrane proteins to transport ions and other molecules across the membrane barrier to aid various steps of viral entry and maturation^9,10. These membrane structures, known as viroporins, usually adopt minimalist architectures that are significantly different from those of bacterial or eukaryotic ion channels. Therefore, understanding the structural basis of how viroporins function broadens our knowledge of channels and transporters while generating new opportunities for therapeutic intervention.

The viroporin formed by the HCV p7 protein has been sought after as potential anti-HCV drug target^5,11. p7 is a 63-residue membrane protein that oligomerizes to form ion channels with cation selectivity, for Ca²⁺ over K⁺ and Na^+2,3,12,13, and more recent studies also reported p7-mediated H⁺ intracellular conductance¹⁴. The p7 channel is required for viral replication1⁵; it has been shown to facilitate efficient assembly and release of infectious virions^16,17, though the precise mechanism of these functions remains unclear. The channel activity can be inhibited by adamantane and long alkylchain iminosugar derivatives and hexamethylene amiloride in vitro, with varying reported efficacies^2,3,12,13. In addition to ion conduction, p7 has been shown to specifically interact with the non-structural HCV protein NS2, suggesting that its channel activity could be regulated^18,19.

There is not yet a detailed structure of the p7 channel, though a number of pioneering NMR studies showed that the p7 monomer has three helical segments: two in the N-terminal half of the sequence and one near the C-terminus^12,20. A single-particle electron microscopy (EM) study obtained a 16 Å resolution electron density map of the p7 oligomer using the random conical tilting approach⁴. The map shows that the p7 channel is a 42 kDa hexamer and adopts a flower-like shape that does not resemble any of the known ion channel structures in the database.

How does the small p7 polypeptide assemble into what appears to be a complex channel structure? Has the viroporin adopted novel structural elements for cation selectivity and channel gating? Amantadine or rimantadine blocks the influenza M2 channel by binding to the small pore formed by four transmembrane helices^21-23, but the pore of the p7 hexamer is expected to be much bigger and it is thus unclear how would these small molecules fit. We sought to address these important questions by determining detailed structures of the p7 hexamer and its drug-binding site.

We systematically tested p7 amino acid sequences from various HCV genotypes and found that the sequence from genotype 5a (EUH1480 strain) generated samples that were sufficiently soluble for structure determination (Supplementary Fig. 1). This p7 construct, designated here as p7(5a), could be efficiently reconstituted in dodecylphosphocholine (DPC) micelles at near physiological pH and generated high quality NMR spectra (Supplementary Fig. 2). Negative-stain EM of the DPC-reconstituted p7(5a) in NMR buffer showed hexameric, flower-shaped particles that are similar to those in the electron micrographs of the p7 (JFH-1 strain, genotype 2a) hexamer in dihexanoyl-phosphatidyl-choline (DHPC) micelles used earlier for single-particle reconstruction⁴ (Supplementary Fig. 3). Moreover, isothermal titration calorimetry and NMR chemical shift perturbation analyses of p7(5a)-rimantadine interaction showed that the drug binds specifically to the reconstituted protein with a binding constant (K_d) from 50-100 μM at 3 mM detergent concentration (Supplementary Fig. 4&5). The above results together indicate that the p7(5a) polypeptides reconstituted in DPC micelles form structurally relevant hexamers.

Structure determination of the p7(5a) hexamer by NMR employed approach taken earlier for oligomeric membrane proteins^24-26, which involves 1) determination of local structures of the monomers and 2) assembly of the oligomer with intermonomer distance restraints and orientation constraints. The NMR-derived restraints define an ensemble of structures with backbone r.m.s. deviation of 0.74 Å (Fig. 1a). Each monomer consists of an N-terminal helix (H1) from residues 5-16, a middle helical segment (H2), with a kink at Gly34, from residues 20-41, and a C-terminal helix (H3) from residues 48-58. These secondary structures are consistent with earlier NMR studies of p7 monomers in DHPC detergent and organic solvent^12,20. There are no intramonomer contacts (Fig. 1a). The monomers are intertwined to form a tightly packed channel, where H1 and H2 form the channel interior and H3 is lipid-facing and packs against H2 of the i+2 and H1 of the i+3 monomer (Fig. 1a&b). The intermonomer association between H3 and H2 appears to be stabilized by interaction involving conserved residues such as Trp30, Tyr42, and Leu52, and the contacts between H3 and H1 are mostly between the alanine rich region of H1 (residues 10-15) and Ala61 and Ala63 of H3 (Fig. 1c). The overall structure of the p7(5a) hexamer has a flower-like shape that agrees with the EM map (EM database ID:1661), fitting to the map with a correlation coefficient of 0.94 (Fig. 1d).

The channel cavity has a funnel profile that resembles a champagne flute and is largely hydrophilic (Fig. 2a). The H2 helices form the wide cylindrical region (internal diameter, I.D. ~12 Å) by packing with each other at large angles (angle between adjacent helices ~ -47°), and the H1 helices assemble at smaller packing angles (~ -34°) to form the narrow conical region of the funnel (smallest I.D. at 6.8 Å). Residues 17-19 constitute the flexible joint between H1 and H2; their NMR resonances are significantly broader than other regions of the protein, suggesting the presence of conformational exchange.

The channel architecture described above represents a novel topology and exemplifies how HCV has optimized the short p7 polypeptide to achieve a rather complex channel structure. What are then the elements for cation conduction and gating? An in-depth examination of the channel interior found two strongly conserved polar residues with salient structural features (Fig. 2b). One is Asn9, which forms a ring of carboxamide that constricts the conical region of the channel (Fig. 2c). Residue 9 is asparagine in all strains except being substituted with histidine in genotype 2 viruses. Both asparagines and histidines have affinity for monovalent and divalent cations. We hypothesize that the Asn9 ring serves as a broad selectivity filter that dehydrates cations, allowing them to pass the hydrophobic ring formed by Ile6. The Ile6 ring defines the narrowest point of the channel and likely serves as a hydrophobic gate. Another feature is the Arg35 ring that defines the wider, C-terminal end of the channel (Fig. 2b). Placement of a positively charged ring on the other end of the pore was incomprehensible to us initially because it can repel cations. But the recent structure of an Orai Ca²⁺ channel also revealed a stretch of basic residues in the ion conducting pore²⁷. We hypothesize that one of the Arg35 roles is to bind and obstruct anions at the pore entrance while allowing cations to diffuse into the pore. In this model, cation conduction is unidirectional from the C- to N-terminal end of the channel.

To test the above hypotheses, we established an assay that uses the two-electrode voltage-clamp technique to record p7-mediated current in Xenopus oocytes (METHODS). Due to the poor stability of oocytes that overexpress p7(5a), p7 (JFH-1 strain, genotype 2a) was used instead for these experiments. As expected of the proposed role of residue 9 in selectively dehydrating cations, replacing His9 of p7(2a) with alanine caused ~70% reduction in channel conductance at +80 mV (Fig. 2d). The proposed role of Arg35 infers that placing negatively charged residues at the channel entrance would bind cations and hinder their diffusion into the pore, and indeed the R35D mutation also reduced conductance by ~70% (Fig. 2d).

We next investigated the mechanism of amantadine binding to the p7 channel using proteins that are ¹⁵N-labeled and deuterated so that Nuclear Overhauser Enhancement (NOE) between the protein backbone amide protons and drug protons could be measured unambiguously. At 10 mM amantadine (not corrected for drug partitioning to detergent micelles), the ¹⁵N-edited NOESY spectrum showed NOE crosspeaks between the adamantane protons and the amide protons of Val26, Leu55, Leu56, and Arg57 (Fig. 3a). We then identified contacts between the drug and protein sidechains using protein that is (¹H/¹³C)-labeled at the methyl positions of alanines, valines and leucines but is otherwise deuterated. In this case, the ¹³C-edited NOESY showed several methyl-drug NOEs (Fig. 3b).

These NOEs were used to dock amantadine into the structure determined in the absence of drug. In doing so, we emphasize that the relevance of the p7-amantadine complex is confined to only the drug binding region because we do not know how and to what degree does drug binding alter the global conformation of the channel. The relatively poor stability of the protein-drug complex at the current stage of our study precludes full-scale structure determination. Nonetheless the available NMR data show that the drug adamantane binds to six equivalent hydrophobic pockets between the pore-forming and peripheral helices (Fig. 3c). The pocket consists of Leu52, Val53, and Leu56 from H3, and Phe20, Val25, and Val26 from H2. The amantadine amino group on average points to the channel lumen. The same NOESY spectrum as above recorded using a sample with 5 mM rimantadine indicates that rimantadine binds to the same pocket with the methyl and amino groups pointing to the lumen (Supplementary Fig. 6).

The binding site is overall consistent with mutational study showing that mutations in residues 50-55 significantly reduced drug sensitivity of the channel²⁸. It is also consistent with a L20F mutation in genotype 1b virus originally identified in clinical trials that confers amantadine resistance^8,29. In the p7(5a) structure, residue 20 is an integral part of the drug pocket and is in direct contact with the drug adamantane. Therefore, replacing Leu20 in p7(1b) with phenylalanine is expected to reduce hydrophobic interaction with the drug. Elucidation of previous functional data in the context of the structure suggests that the binding site shown in Fig. 3c is relevant to drug inhibition and that interactions between the drug adamantane and protein hydrophobic residues are critical for inhibition. Variations in the hydrophobicity of the binding pocket among the p7 variants (Supplementary Fig. 7) thus explain the large differences in drug efficacies observed between different HCV genotypes.

We have learnt from KcsA and other channels that a gated ion channel generally adopts two essential features: pore elements that provide ion selectivity and gating mechanism that can transiently open the channel to allow ion permeation. By virtue of being a funnel, the p7 structure suggests that the tip of the funnel represented by the Ile6 and Asn9 rings is the key region for channel gating (Fig. 4). The role of the Asn9 ring is to provide ion selectivity by recruiting and dehydrating cations near the funnel exit, whereas the Ile6 ring is a hydrophobic constriction that would prevent water from freely passing through. Channel activation may involve reorientation of the H1 helices that widens the funnel tip, analogous to the dynamic C-terminal helix of KcsA³⁰, and such structural rearrangement can be afforded by the flexible hinge between H1 and H2, the intervening loop between H2 and H3, and the C-terminal tail that “latches” onto H1. We thus propose that binding of adamantane derivatives inhibit channel activity by restricting the structural rearrangement. Our NMR titration data (Supplementary Fig. 5c) is consistent with this proposal, which showed that in the absence of rimantadine, the Ile6 methyl resonance is split into an intense and weak peak, possibly corresponding to the open and closed state, respectively, and that increasing the drug concentration shifted the equilibrium that made the weak peak stronger. Although rigorous testing of the model is needed, the preliminary observation suggests the existence of multiple states of the p7 channel.