Single-molecule-binding studies of antivirals targeting the hepatitis C virus core protein

All solutions and reagents were prepared using ultrapure (18.2 MΩ cm) water. Na₂HPO₄, NaH₂PO₄, K₂HPO₄, KH₂PO₄, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), ethylenediaminetetraacetic acid (EDTA), tris(2-carboxyethyl) phosphine (TCEP), trifluoroacetic acid (TFA), dimethyl sulfoxide (DMSO), isopropyl β-D-1-thiogalactopyranoside (IPTG), TWEEN20, and guanidinium chloride (GdmCl) were purchased from Sigma-Aldrich (St. Louis, MO). NaCl, tris(hydroxymethyl)aminomethane (Tris), and NaOH were purchased from Thermo Fisher Scientific (Fair Lawn, NJ). 200 proof Ethyl alcohol (EtOH), acetonitrile (ACN), and dithiothreitol (DTT) were purchased from Decon Labs (King of Prussia, PA), Supelco (Bellefonte, PA), and bioWORLD (Dublin, OH), respectively.

To explore the degree of sequence conservation among the 10 HCV proteins, all 220,667 amino acid sequence entries related to Hepacivirus C (Taxon ID: 11103) were downloaded from the UniProt Knowledgebase. Then, a suite of command line tools (65) was used to align the protein sequences using a local installation of NCBI-BLAST. First, a local database was created using the 220,667 protein sequences. Next, the 3,017 amino acid polyprotein derived from a genotype 6a (ZS202) HCV isolate (Accession: AHH29575.1) was used as a query sequence to generate an alignment of nearly every other entry in our local database. Finally, the sequence conservation for each protein (Fig. 1A) was calculated by averaging the conservation rate for each amino acid associated with each of the 10 distinct proteins of HCV.

The ^NTDHCVcp amino acid sequences used in this study (Fig. S1) are minimally modified, dual-cysteine variants associated with either a genotype 1a (PE8) HCV isolate [Accession: CAE46584.1 (66)] or a genotype 2a (J6/JFH-1) HCV isolate [Accession: BAB32872.1 (67)], where the two cystine residues are used for the bioconjugation of donor and acceptor fluorophores for FRET-based studies. The 136 amino acid sequence of the genotype 1a variant used in our study (cp1a) differed from the first 133 amino acids of the PE8 viral isolate at three positions (S2C, C98M, S130C) and contains three additional residues (GPG) at N-terminus resulting from recombinant expression. The 131 amino acid sequence of the genotype 2a construct used in our study (cp2a) also contained three additional residues at the N-terminus, with the only other difference from the first 128 amino acids of the J6/JFH-1 viral isolate being an S2C mutation, as there is naturally a cysteine at position 128. Importantly, the locations of the two cysteine residues in cp2a (i.e., C2 and C128) closely mimic the location of those in cp1a (i.e., C2 and C130), allowing us to make meaningful comparisons between the two in our FRET-based studies.

The expression, purification, and labeling of cp1a have all been described in detail elsewhere (18). Those procedures were used to guide the expression, purification, and labeling of cp2a, which are described in detail here. Competent BL21(DE3) cells (Thermo Scientific, Waltham, MA) were transfected with a pET47b(+) plasmid (Celtek Genes, Franklin, TN) whose codons were optimized for the expression of a recombinant protein consisting of an N-terminal hexahistidine affinity tag adjacent to a HRV 3C protease cleavage site followed by the ^NTDHCVcp amino acid sequence of interest. An Erlenmeyer flask containing 2 L of lysogeny broth and Kanamycin at a concentration of 100 µg mL⁻¹ was inoculated with a 30-mL overnight culture of transfected cells. After incubating for 2 hours at 37°C, protein expression was induced by adding IPTG into the broth at a final concentration of 1 mM. Cells were grown for 3 hours to allow for the expression of the recombinant protein. Next, the broth was centrifuged at 4°C for 15 minutes at 4,000 rpm to pellet the cells. The pellet was resuspended in a lysis buffer (50 mM Tris-HCl, 10 mM imidazole, 6 M GdmCl, 10 mM NaH₂PO₄, 5 mM TCEP, pH 8.0). The resulting milieu was centrifuged at 19,000 rpm for 30 minutes to pellet insoluble cellular debris. The supernatant, which contained hexahistidine-tagged recombinant protein, was allowed to flow through the Probond nickel-chelating resin (Novex, Carlsbad, CA) of an immobilized-metal ion affinity chromatography (IMAC) column pre-equilibrated with lysis buffer. The nickel-bound recombinant protein was then eluted from the column with elution buffer (50 mM Tris-HCl, 250 mM imidazole, 6 M GdmCl, 260 mM NaH₂PO₄, 5 mM TCEP, pH 8.0). The eluate containing the hexahistidine-tagged recombinant protein was dialyzed against 2 L cleavage buffer (50 mM Tris-HCl, 150 mM NaCl, pH 7.5) at 4°C. After overnight dialysis, urea was added to a final concentration of 400 mM to solubilize the protein precipitate, and then, HRV 3C protease (Millipore Sigma, Burlington, MA) was added to a final concentration of 0.6 µM. This solution was again dialyzed against a fresh 2 L of cleavage buffer overnight at 4°C to allow ample time for the enzyme to cleave the N-terminal hexahistidine tag from the recombinant protein and the solutions to osmotically equilibrate. The cleavage product, cp2a, was purified from the hexahistidine tag using second round of IMAC. The mass of the purified cp2a (14,655.80 Da) was confirmed with the assistance of the mass spectrometry core facility at the University of Kansas (Fig. S2).

The strategy for labeling cp2a was based on a general approach for single-molecule FRET studies (68). First, cp2a was incubated in 10 mM DTT to ensure that all cystines were reduced. Progress of the reduction reaction was monitored via a high-performance liquid chromatography (HPLC) system (Infinity II 1260; Agilent Technologies, Inc., Santa Clara, CA) equipped with an analytical reversed-phase column (Reprosil Gold C18; Dr. Maisch, Ammerbuch-Entringen, Germany) using a 0.1% TFA water-acetonitrile gradient. The reduced form of cp2a was separated from DTT and purified using HPLC, and then lyophilized to remove the solvent. The lyophilized cp2a was then resuspended with a denaturing labeling buffer (6 M GdmCl, 0.1 M potassium phosphate, pH 7.3) to achieve a protein concentration of 50 µM to minimize disulfide bond formation. The thiol-reactive donor (Cy3B-maleimide) and acceptor (CF660R-maleimide) fluorophores were suspended in DMSO, sonicated for 15 minutes to break up aggregates, and added to the resuspended cp2a to achieve a final concentration of 50 µM, resulting in slight excess of the two fluorophores. The reaction mixture was vortexed and incubated overnight at 4°C in the dark. Afterward, unreacted fluorophores were removed by passing the labeling reaction through a desalting column (Zeba Spin; Thermo Scientific) pre-equilibrated with the denaturing labeling buffer. The FRET-labeled permutations of the labeled proteins (i.e., donor-acceptor and acceptor-donor) were partially purified using HPLC. However, due to small differences in elution times, complete separation from the other labeling permutations (e.g., donor-donor and acceptor-acceptor) was not possible. Nevertheless, downstream data analyses (see below) are used to filter out these unwanted species. Finally, the resulting fluorescently labeled proteins were lyophilized and stored at −80°C.

The Cnew antiviral aptamer and a library of random 44-nucleotide long sequences (Fig. S1) were purchased from Integrated DNA Technologies (Coralville, IA). A monoclonal antibody (ab2740) with an epitope that reportedly recognizes amino acids 21–40 in ^NTDHCVcp was purchased from Abcam (Cambridge, United Kingdom). A set of four antiviral peptides (SL173, SL174, SL175, and SL571) with amino acid sequences (Fig. S1) derived from the ^NTDHCVcp of a genotype 1a (H77) HCV isolate were chemically synthesized by the Synthetic Chemical Biology core facility at the University of Kansas. The same core facility also chemically synthesized C20 (Fig. S1), a small-molecule antiviral compound that reportedly inhibits oligomerization of ^NTDHCVcp (43). The molecular mass of Cnew, SL173, SL174, SL175, SL571, and C20 was confirmed using mass spectrometry.

All single-molecule measurements were conducted on samples consisting of 50 µL of aqueous solution within a single well of a 15-well 3D μ-slide (ibidi GmBH, Grafelfing, Germany). These solutions consisted of labeled ^NTDHCVcp at a concentration of roughly 100 pM (to ensure single-molecule detection) in measurement buffer (25 mM HEPES, 12.5 mM NaOH, 0.5 mM TWEEN20, and 2 nM unlabeled ^NTDHCVcp) containing varying concentrations of NaCl and antivirals depending on the particular measurement. The use of unlabeled ^NTDHCVcp and TWEEN20 in the measurement buffer helps minimize unwanted interactions between labeled ^NTDHCVcp and the wells of the µ-slide.

All single-molecule measurements were performed on a single-molecule instrument (Fig. 2A) consisting of a modified confocal fluorescence microscope (MicroTime200; PicoQuant, Berlin, Germany) equipped with two dual-mode diode lasers with output wavelengths of 515 nm (Quixx 515–80PS; Omicron, Rodgau, Germany) and 642 nm (Quixx 642–140PS; Omicron). The output from the two excitation sources was each adjusted to 90 µW while being digitally modulated by a function generator (577–4C; Berkeley Nucleonics Corporation, San Rafael, CA). Most experiments were conducted with the lasers in continuous-wave mode and the function generator set to a repetition rate of 20 kHz to achieve the so-called alternating laser excitation (ALEX) configuration (69). However, measurements of antibody-containing samples were made with the lasers in pulsed mode and the function generator set to a repetition rate of 20 MHz to achieve the so-called pulsed interleaved excitation (PIE) configuration (70), enabling accurate measurements of the sub-millisecond translational diffusion of ^NTDHCVcp. In either configuration, the excitation light passed through a multi-band dichroic mirror (ZT532/640rpc; Chroma, Bellows Falls, VT) prior to being focused into the sample via a 1.2 NA, 60×, water immersion objective (UPlanSAPO; Olympus, Tokyo, Japan).

Given extremely dilute concentrations of fluorescent species in the sample, the focus is expected to be devoid of fluorescent molecules >99% of the time. However, when a fluorescently labeled ^NTDHCVcp molecule in the aqueous sample diffuses through the diffraction-limited focus (Fig. 2B), the fluorophores attached to it will repeatedly absorb photons from the alternating excitation sources and emit a burst of fluorescence photons. Dozens of these photons are collected via the microscope objective before being directed back through the multi-band dichroic mirror toward both a 100-µm pinhole to remove out-of-plane photons and toward a polarizing beam splitter (BS cube polarizer; PicoQuant), which spatially separate the emitted photons based on polarization. Next, the two streams of photons each impinge on a long-pass dichroic (635LPXR; Chroma) to spatially separate them again, this time based on wavelength. The resulting four photon streams are then focused on to one-of-four single-photon avalanche photodiode (APD) detectors (SPCM-AQRH-14-TR; Excelitas Technologies, Waltham, MA). Finally, the arrival time of each detected photon is recorded and processed via a time-correlated single-photon counting (TCSPC) module (HydraHarp 400; PicoQuant) with a temporal resolution of either 16 ps (PIE) or 16,384 ps (ALEX) and saved to a .ptu file. All single-molecule measurements were performed at 23°C for either 10 minutes (ALEX) or 20 minutes (PIE).

Photon detection events saved to the .ptu data files were analyzed in Mathematica using Fretica (71), a user-extendable toolbox for analyzing single-molecule fluorescence data.

First, it is of course critical to establish a baseline for ^NTDHCVcp in the absence of any antivirals. To do this, we prepared aqueous samples containing fluorescently labeled variants derived from either genotype 1a (cp1a) or 2a (cp2a) HCV isolates at a concentration of roughly 100 pM under our standard experimental conditions (measurement buffer with 100 mM NaCl). When single FRET-labeled ^NTDHCVcp molecules pass through the confocal volume, a burst of several dozen fluorescence photons is generated that lasts about 1 ms. The photons associated with each burst are then used to calculate the transfer efficiency, $E$ , of each burst, which, after the duration of the entire measurement, are compiled together in a so-called transfer efficiency histogram and analyzed to determine the mean transfer efficiency, $⟨E⟩$ (Fig. S4). Using this approach and averaging over several repeated measurements, we found $⟨E⟩$ = 0.33 ± 0.02 and $⟨E⟩$ = 0.38 ± 0.02 for cp1a and cp2a, respectively.

Next, to demonstrate our ability to resolve conformational changes in ^NTDHCVcp that occur after adjusting experimental conditions, we first adjusted the NaCl concentration of the samples. Increasing the NaCl concentration systematically shifts the transfer efficiency histograms to higher $E$ values (Fig. S5), whereas the opposite occurs in solution with lower concentrations of NaCl. Given the high net positive charge of cp1a (+22) and cp2a (+23), it is not surprising that both proteins become more compact at high NaCl concentrations due to the additional charge screening provided by the counterions in solution. Indeed, similar observations have been made in studies of other positively charged, intrinsically disordered proteins (78 – 83) including HCVcp (18, 76). Furthermore, these findings demonstrate that single-molecule transfer efficiency measurements can be used to detect solute-induced conformational changes within ^NTDHCVcp.

Previously, SELEX was used to identify several DNA aptamers that bind to HCVcp and exhibit antiviral activity in cell-culture systems (22). Here, we studied the interactions between ^NTDHCVcp and one such aptamer, Cnew, which is 44-nucleotides long (Fig. S1). This was accomplished by performing single-molecule transfer efficiency measurements of fluorescently labeled cp1a under standard experimental conditions with increasing concentrations of unlabeled Cnew aptamer. Analysis of this Cnew titration revealed a systematic decrease of the unbound population at $⟨E⟩$ = 0.324 ± 0.004 and a corresponding increase of a bound population at $⟨E⟩$ = 0.551 ± 0.003, suggesting that the intrinsically disordered cp1a becomes significantly more compact after binding to Cnew at low-nanomolar concentrations (Fig. S6A).

Given that Cnew is highly negatively charged and ^NTDHCVcp is highly positively charged, we expected the interactions between these two binding partners to strongly depend on electrostatics. Therefore, we carried out analogous Cnew titrations at higher NaCl concentrations to increase the extent of charge screening between these oppositely charged polymers (Fig. 4; Fig. S6). All of these Cnew titrations showed the same qualitative behavior regardless of NaCl concentration—the $⟨E⟩$ value increases at lower concentrations of Cnew before it asymptotically plateaus at higher Cnew concentrations (Fig. 4A).

Consistent with our observations in the absence of antivirals (Fig. S5), we see that increasing NaCl concentration in the absence of Cnew increases the $⟨E⟩$ of free ^NTDHCVcp (Fig. 4A), indicating that the molecular dimensions of this intrinsically disordered protein are likely modulated by repulsive intramolecular electrostatic interactions. Conversely, we observed that at saturating concentrations of Cnew, the $⟨E⟩$ of the bound cp1a-Cnew complex decreases with increasing NaCl concentration (Fig. 4A), suggesting that the molecular dimensions of cp1a within the nucleoprotein complex are highly dependent on attractive intermolecular electrostatic interactions.

To quantify the binding affinity of the ^NTDHCVcp-aptamer interaction, we calculated the apparent dissociation constant ( ^appK_d ) for each of the complete Cnew titrations (Fig. 4.A) by fitting our $⟨E⟩$ vs Cnew concentration data to a quadratic form of the single-site binding equation (equation 4). Under standard experimental conditions, the ^appK_d for the cp1a-Cnew complex was much lower than the total ^NTDHCVcp concentration in the solution ( ^appK_d ≪ 2 nM), and therefore, it could not be accurately quantified using our fitting routine. However, as the NaCl concentration increased, the ^appK_d for the cp1a-Cnew complex also increased from ^appK_d = ${\displaystyle {\text{0.16}}_{-\text{0.06}}^{+\text{0.10}}}$ nM at 150 mM NaCl to ^appK_d = ${\displaystyle {\text{160}}_{-\text{100}}^{+\text{260}}}$ nM at 400 mM NaCl. These findings suggest that attractive intermolecular electrostatic interactions play a major role in the binding of positively charged cp1a to the negatively charged antiviral aptamer, Cnew. From an electrostatic perspective, one can imagine that the decrease in binding affinity at elevated NaCl concentration could arise from counterions in solution screening and weakening of the attractive electrostatic forces between oppositely charged functional groups in cp1a and Cnew. A quantitative linkage analysis of the NaCl-dependent binding affinity data (84) suggests that the association of these two biomolecules gives rise to the release of approximately eight counterions, as indicated by the slope, Γ, of a double logarithmic plot of ^appK_d vs NaCl concentration (Fig. 4C, solid line). From a thermodynamic perspective, this indicates that releasing counterions to the bulk solution becomes increasingly less favorable as the bulk counterion concentration increases, yielding a positive increase in the binding free energy change.

Next, we sought to assess the specificity of the binding interaction between cp1a and Cnew. Therefore, we performed an entire collection of single-molecule measurements wherein the Cnew aptamer was replaced with Rand, a chemically synthesized library of 44-nucleotide long oligos with random sequences. The results of the measurements with Rand (Fig. 4B) are quite similar to what was observed with Cnew (Fig. 4A), except that the transition midpoints and the corresponding ^appK_d values were systematically shifted to higher concentrations. This observation was also present at higher NaCl concentrations with more electrostatic screening. Furthermore, quantification of the NaCl-dependent binding affinity data suggests that the cp1a-Rand binding interactions also release approximately eight ions (Fig. 4C, dashed line). Taken together, the above findings indicate that cp1a can promiscuously bind to a range of nucleic acid sequences via electrostatic interactions but that the SELEX-derived sequence associated with the Cnew aptamer binds more than an order of magnitude more tightly under conditions with near-physiological concentrations of monovalent ions.

Finally, to assess the applicability of these findings to other viral genotypes, an analogous suite of experiments was conducted using a ^NTDHCVcp variant (cp2a) derived from a genotype 2a HCV isolate. Under standard experimental conditions, we were unable to generate a complete series of single-molecule transfer efficiency histograms for cp2a because the addition of low-nanomolar concentrations of Cnew resulted in the formation of large, slowly diffusing, fluorescent particles and a corresponding reduction in the relative number of high-quality bursts, $B_N$ (Fig. 5). Nevertheless, at higher Cnew concentrations, values of $B_N$ returned to typical values near unity, and we were again able to generate transfer efficiency histograms, this time with an $⟨E⟩$ = 0.616 ± 0.014, suggesting that, like cp1a, cp2a also becomes compact after binding to Cnew.

Fortunately, complications arising from the formation of large, slowly diffusing, fluorescent particles were less severe at higher NaCl concentrations, resulting in larger values of $B_N$ (Fig. 5). A collective analysis of all measurements where $B_N$ > 0.4 generated well-populated transfer efficiency histograms for cp2a in both the Cnew- and Rand-binding titrations with NaCl concentrations ranging from 225 mM to 400 mM (Fig. S7). For all binding titrations carried out over this range, the $⟨E⟩$ values for cp2a first increase and then plateau asymptotically at high oligo concentrations (Fig. S7). A quantitative analysis of the NaCl-dependent binding affinity data reveals that as the NaCl concentration increases, the ^appK_d values for both Cnew and Rand systematically increase in a manner that is consistent with the release of approximately 11 counterions and 6 counterions, respectively (Fig. S7). All told, this suite of experimental evidence suggests the ^NTDHCVcp-aptamer binding interactions involving cp2a are incredibly similar to those involving cp1a. Nevertheless, we observed one stark difference between the two genotypic variants of ^NTDHCVcp, which is that cp2a and Cnew can form higher-order structures resulting in the appearance of large, slowly diffusing, fluorescent particles and an overall decrease in $B_N$ (Fig. 5).

Next, we shift our attention to a second class of prospective antiviral therapeutics that can bind to HCVcp and impair viral replication, specifically, antigen-binding proteins (e.g., scFv) (24). To do so, we again carried out single-molecule fluorescence studies of the binding interactions between ^NTDHCVcp and a simple, commercially available, mouse monoclonal IgG1 antibody (ab2740) raised against the core protein of a genotype 1b HCV isolate.

Under standard experimental conditions, the addition of ab2740 resulted in little-to-no change in the $⟨E⟩$ of fluorescently labeled cp1a across the entire binding titration (Fig. 6A), suggesting that either ab2740 doesn’t bind to cp1a under our experimental condition or that binding does not appreciably change the conformation of cp1a. Given that the IgG1 antibody (150 kDa) is much larger than ^NTDHCVcp (15 kDa), the diffusive properties of fluorescently labeled cp1a should change significantly when bound to ab2740. Therefore, we use fluorescence correlation spectroscopy (FCS) analyses to calculate the average diffusion time, $〈{\tau }_D〉$ , of cp1a through our sub-femtoliter diffraction-limited confocal volume. In the absence of ab2740 $〈{\tau }_D〉$ = 449 ± 6 µs. Upon increasing the concentration of ab2740 in solution, the values of $〈{\tau }_D〉$ first increase and then plateau asymptotically at $〈{\tau }_D〉$ = 856 ± 6 µs (Fig. 6B). The binding affinity of cp1a for ab2740 was then quantified by fitting the antibody-dependent $〈{\tau }_D〉$ of cp1a (Fig. 6C) to our single-site binding equation (equation 4), which resulted in an ^appK_d = ${\displaystyle {\text{3.7}}_{-\text{0.6}}^{+\text{0.7}}}$ nM. This result indicates that cp1a does indeed bind tightly to ab2740 and that binding must not substantially alter the conformational ensemble of the intrinsically disordered ^NTDHCVcp.

Motivated by the pronounced NaCl dependence observed with the antiviral aptamers, we again conducted additional experiments under conditions with elevated NaCl concentrations. Like our antibody-binding titration under standard experimental conditions, we found that even in the presence of elevated NaCl concentrations, the $⟨E⟩$ of cp1a was effectively unaltered across the entire range of ab2740 concentrations we surveyed (Fig. 6A). However, we were again able to detect binding between cp1a and ab2740 at higher NaCl concentrations using our FCS analyses to calculate the ${\displaystyle ⟨{\tau }_D⟩}$ of cp1a (Fig. 6B). Moreover, we observed the binding affinity between cp1a and ab2740 slightly, but systematically, increased upon increasing the NaCl concentration (Fig. 6C), resulting in an ^appK_d = ${\displaystyle {\text{0.09}}_{-\text{0.05}}^{+\text{0.09}}}$ nM at 800 mM. These findings indicate that this intermolecular interaction is hindered by repulsive intermolecular electrostatic interactions between the epitope of ab2740 and the highly conserved stretch of amino acids in ^NTDHCVcp it is reported to recognize. Similarly, a quantitative linkage analysis of the NaCl-dependent binding affinity data (84) shows that approximately one additional counterion is taken up from the bulk solution during binding (Fig. 6E, cyan), which presumably helps to screen this slightly repulsive binding interaction.

Given that the epitope of ab2740 consists of the highly conserved amino acids at positions 21–40 of ^NTDHCVcp (Fig. S1), we expected that this antibody would also efficiently bind cp2a. Indeed, a qualitative analysis of analogous sets of experiments conducted using cp2a supported this expectation. First, the $⟨E⟩$ of cp2a was largely insensitive to the addition of ab2740 (Fig. 6D), whereas the $〈{\tau }_D〉$ was dependent on the concentration of ab2740 in a manner that was consistent with a single-digit nanomolar-binding affinity (Fig. 6E). Curiously, the ^appK_d was substantially less dependent on NaCl concentration (Fig. 6C), suggesting that neither attractive nor repulsive interactions dominate, and the formation of this antibody-antigen complex does not change the number of counterions associated with the two binding partners.

Past research efforts have shown that short peptides derived from regions of ^NTDHCVcp believed to be involved in homotypic interactions can inhibit oligomerization of ^NTDHCVcp in vitro and impair viral assembly in cell culture (25, 44, 45, 85). We selected a set of four such antiviral peptides whose amino acid sequences were derived from the ^NTDHCVcp of a genotype 1a (H77) HCV isolate (Fig. S1) and studied their ability to interact with ^NTDHCVcp using single-molecule fluorescence spectroscopy.

We chose to study the peptide SL173 first due to its previously reported high binding affinity for ^NTDHCVcp and its antiviral potency in cell culture (45). Single-molecule measurements of cp1a were conducted under standard experimental conditions in the presence of increasing concentrations of SL173. During these antiviral peptide-binding titrations, we noted that our ability to generate transfer efficiency histograms dramatically decreased at SL173 concentrations above 1 µM due to a reduction in the relative number of high-quality bursts ( $B_N$ ). Nevertheless, for those histograms that could be generated, the corresponding $⟨E⟩$ values were experimentally indistinguishable from measurements conducted in the complete absence of SL173 (Fig. 7A), suggesting that this antiviral peptide doesn’t stably interact with single molecules of ^NTDHCVcp. Instead, the reduction of $B_N$ , which coincided with the emergence of large, slowly diffusing, fluorescent particles, suggests that interactions between cp1a and SL173 lead to rapid formation of higher-order multimeric assembles or aggregates. To quantify the concentration of SL173 needed to produce a half-maximal effect ( ${EC}_{50}$ ), we fit our $B_N$ data (Fig. 7B) to a cooperative ligand-binding model, which resulted in an ${EC}_{50}$ = 3.2 ± 0.3 µM (Fig. 7B).

Although cp1a contains many positive charges, the SL173 antiviral peptide, whose sequence is derived from the amino acid residues between position 85 and 110 in the ^NTDHCVcp of an H77 HCV isolate, is surprisingly uncharged. As such, we expected that the mechanism of oligomerization would be largely insensitive to NaCl concentration. Indeed, the results from analogous peptide-binding titrations conducted at elevated NaCl concentrations support this notion and show only a modest increase in across the entire range of experimental conditions (Fig. 7C).

Again, the rationale for the use of these peptides stems from their perceived potential to competitively form homotypic interactions with ^NTDHCVcp. Because SL173 was derived from an amino acid sequence in a more variable region of ^NTDHCVcp, we sought to determine if SL173 would also cause cp2a to oligomerize or aggregate. Given the minimal sequence similarity between cp1a and cp2a over the region of amino acids that correspond to SL173 (Fig. S1), we were surprised to see that SL173 was nearly as effective at inducing oligomerization of cp2a as it was with cp1a and that this effect was also quite insensitive to NaCl concentration (Fig. 7C). Together, these findings suggest that the formation of higher-order oligomeric structures containing ^NTDHCVcp might be driven by the physicochemical properties of SL173 (e.g., hydrophobicity) rather than its ability to form sequence-specific homotypic interactions with ^NTDHCVcp.

Finally, we set out to determine if three other antiviral peptides had similar interactions with our two ^NTDHCVcp variants. We found that the of cp1a and cp2a remained largely unaltered across the entire binding titration for the SL174 antiviral peptide and values of only decreased at the highest concentrations, resulting in an > 40 µM for both high (800 mM) and low (100 mM) NaCl concentrations (Fig. 7C). Like SL173, titrations of the antiviral peptide SL175, did give rise to a decrease in corresponding to the formation of oligomeric species for both cp1a ( = 6.6 ± 0.4 µM) and cp2a ( = 17 ± 1 µM), but only at 100 mM NaCl. However, unlike SL173, which has no net charge, SL175 is charged (+2), which likely makes this binding interaction more susceptible to electrostatic contributions that can be screened at high NaCl concentrations, thus explaining why no detectable binding of SL175 was observed at 800 mM NaCl. The amino acid sequence of the antiviral peptide SL571 is the reverse of SL175 (Fig. S1), and yet, both are characterized by low-micromolar values at 100 mM NaCl and little-to-no binding at 800 mM NaCl. This finding further supports the notion that the molecular interactions between ^NTDHCVcp and these antiviral peptides are not necessarily governed by primary sequence and instead depend simply on the general physicochemical properties of the two binding partners.

Finally, we wanted to characterize the interactions between ^NTDHCVcp and a small-molecule compound to which it had been reported to bind (43). Previously, during an initial small-molecule screen of a indoline alkaloid library, four candidate compounds were shown to greatly disrupt oligomerization of ^NTDHCVcp (56). Further efforts to improve these compounds ultimately resulted in a more potent and less cytotoxic compound referred to as C20 (Fig. S1), which was derived from the same indoline alkaloid scaffold associated with the initial four candidates. Whole-cell studies of C20 and its derivatives demonstrated that these compounds were able to reduce viral replication in Huh-7.5 cells infected with a genotype 2a (J6/JFH) HCV isolate (43, 56). However, the biochemical details of the molecular interaction between ^NTDHCVcp and these indoline alkaloid compounds have yet to be resolved. Therefore, we used single-molecule FRET spectroscopy to monitor the conformational dimensions of our two fluorescently labeled variants of ^NTDHCVcp in the presence of increasing concentrations of C20.

Initially, we measured samples containing cp1a in the presence of increasing concentrations of C20 under standard experimental conditions (Fig. S8). We observed that $⟨E⟩$ remained unchanged (Fig. S7A) over the entire range of concentrations (0–25 µM) where C20 was soluble in solution. The same was true for analogous measurements under non-standard experimental conditions with elevated concentrations of NaCl (800 mM). These results suggest that the addition C20 does not alter the average end-to-end distance of cp1a under our experimental conditions. Next, to determine if the presence of C20 alters the diffusive behavior of cp1a, we calculated the $〈{\tau }_D〉$ of cp1a under these conditions, and it also remains unchanged at both 100 and 800 mM NaCl, regardless of the C20 concentration (Fig. S8). Then, to determine if C20 causes oligomerization of cp1a in a manner that was similar to what was observed for the antiviral peptides, we looked at $B_N$ and found that it too remains unchanged under all experimental conditions (Fig. S8). Finally, to determine if these negative findings were genotype-specific, we conducted an analogous set of measurements using fluorescently labeled cp2a. The results associated with cp2a were nearly identical to those associated with cp1a—the presence of C20 does not alter the end-to-end distance, diffusivity, or oligomerization tendency of cp2a.

Global eradication of hepatitis C virus by 2030 is a key target set by the World Health Organization (86). To achieve this goal, cost-effective pangenotypic antiviral therapeutics for HCV remain crucial. Although current medications are highly effective (87, 88), resistance-associated mutations remain a concern (89, 90), and therefore, combination therapies are often required, further increasing the financial burden associated with HCV treatment plans (91). One approach for the development of novel pangenotypic antivirals is to inhibit critical processes within the viral replication cycle that are highly conserved across all genotypes. Viral assembly is one such process and is mediated by HCVcp (19, 21, 92), which is the most conserved protein across the different HCV genotypes (15). Although HCVcp has the potential to be a promising new target for therapeutic intervention, more work is needed to assess the merits of this approach. Here, we used single-molecule spectroscopy to study fluorescently labeled variants of the N-terminal domain of HCVcp (^NTDHCVcp) from either genotype 1a or 2a HCV isolates and determine the biophysical principles that govern their ability to bind to and interact with four different classes of prospective antiviral therapeutics, namely, aptamers, antibodies, peptides, and small molecules.

All of our antiviral-binding titrations contained measurements of fluorescently labeled ^NTDHCVcp conducted in the complete absence of antivirals. In these reference measurements, we observed that both ^NTDHCVcp variants (cp1a and cp2a) became increasingly compact at higher NaCl concentrations, which is a well-known phenomenon in polyelectrolytic IDPs (18, 76, 78 – 83) and implies that the nearly polycationic ^NTDHCVcp adopts a heterogenous ensemble of conformations that is dictated by repulsive intramolecular electrostatic interactions capable of being screened at high ionic strength. However, given the nature of our single-molecule experiment, one cannot completely rule out the possibility of slight structural (or functional) alterations arising from the fluorogenic probes. Nevertheless, our findings are consistent with previous studies that show the monomeric form of ^NTDHCVcp is intrinsically disordered under a wide range of experimental conditions (16 – 18, 93); some of these studies used alternative fluorophores conjugated to different positions within ^NTDHCVcp. Furthermore, AlphaFold (10) predictions yield structures of ^NTDHCVcp with low to very low confidence, which others have noted can be an effective predictor of protein disorder (13, 14). Finally, previous in vivo studies of tetra-cysteine-modified core proteins labeled with biarsenical fluorophores showed only minor changes to virus titers (94), suggesting that the protein remains functional and seems to tolerate minor molecular perturbations quite well.

A critical component of viral assembly involves formation of the nucleocapsid, which resides at the core of the virus (19, 92). In HCV, the nucleocapsid is formed when many positively charged core proteins condense around the negatively charged viral genome (19, 21, 92). As such, it is not at all surprising that ^NTDHCVcp has a high affinity for both RNA (18) and DNA (95), including the previously identified antiviral DNA aptamer, Cnew (22). In fact, previous studies have reported that other antiviral aptamers bind to HCV core proteins from different genotypes with nanomolar affinities under physiologically relevant monovalent ion concentrations (22, 23, 76, 77). Some of these studies showed that antiviral aptamers also decrease virus titers in naïve Huh-7.5 cells (22, 23). In one of these studies, four DNA and four RNA aptamers were generated using SELEX with core proteins from HCV isolates with different genotypes as the target (23). Both the RNA aptamers and the DNA aptamers had nanomolar affinities for core proteins from six different genotypes (genotypes 1 through genotype 6) (23). Our results with the DNA aptamer Cnew yield binding affinities that are comparable to those in the studies discussed above. However, our studies shed new light on the molecular mechanisms governing the interaction between antiviral aptamers and HCVcp. First, we show that ^NTDHCVcp undergoes a conformational change upon binding to Cnew (Fig. 4A). The compaction of ^NTDHCVcp that occurs after binding to nucleic acids has been observed previously in other biochemical studies of nucleocapsid protein-nucleic acid interactions (95), including interactions between ^NTDHCVcp and short RNAs excised from the HCV genome (18). Furthermore, we show that the binding affinity of ^NTDHCVcp for Cnew depends strongly on ionic strength, indicating that binding is dominated by attractive electrostatic interactions and results in the formation of what is likely to be an intrinsically disordered polyelectrolyte complex (96). Finally, we demonstrated that despite involving strong electrostatic interactions, which are often thought of as promiscuous, ^NTDHCVcp does have some structural specificity for Cnew (Fig. 4C), either at the level of nucleic acid primary or secondary structure. Again, such behavior was also observed previously in studies of ^NTDHCVcp and its preferential ability to form nucleocapsid-like particles in the presence of short RNA sequences excised from the HCV genome (18). Together, these findings paint a picture where antiviral aptamers function as particularly effective competitive inhibitors that sequester ^NTDHCVcp in RNA-protein complexes, thus preventing ^NTDHCVcp from properly condensing on the viral genome to form the nucleocapsid (76).

Our results show that the monoclonal antibody for ^NTDHCVcp, ab2740, binds to both of our ^NTDHCVcp variants with low-nanomolar affinity. In both cases, antibody binding has no detectable effect on the conformation of ^NTDHCVcp (Fig. 6A and D), suggesting that amino acids far outside of the reported epitope likely remain conformationally heterogeneous. The lack of any pronounced differences between cp1a and cp2a is consistent with ^NTDHCVcp being extremely conserved across all of the viral genotypes. Indeed, the amino acid residues associated with the epitope (21–40) are identical between the two variants. Nevertheless, we did observe that the cp1a-ab2740 interaction was slightly dependent on NaCl concentration, whereas the cp2a-ab2740 interaction was not (Fig. 6C). Curiously, cp1a has a neutral amino acid (Q) at position 20, whereas cp2a has an acidic residue (E) at this position (Fig. S1). Ionization of this residue may help to counterbalance attractive and repulsive interactions between the two binding partners. With a neutral residue at this position, the electrostatic interactions between the hypervariable loops of ab2740 and the epitope of cp1a would be repulsive and thus could be effectively screened at elevated NaCl concentrations giving rise to slightly higher binding affinities. Although ab2740 is well suited to bind to ^NTDHCVcp, it is not known if the bound ^NTDHCVcp is functional. Even if antigen recognition were to disrupt nucleocapsid formation, ab2740 is simply too large to readily enter cells. However, the variable domains of the heavy and the light chains can be connected via a polypeptide linker to form a single-chain variable fragment (scFv) antibody, which is a much smaller biomolecule that can more readily enter cells (97, 98). Indeed, a scFv antibody that binds to a less conserved epitope on ^NTDHCVcp (residues 82–88) has previously been shown to inhibit viral replication (24). Therefore, engineering a scFv based on ab2740 may be an attractive starting point for novel antibody-based antiviral therapeutics for HCV.

Peptides are an attractive choice for therapeutic applications because of their ease of synthesis and biocompatibility. As such, previous research efforts have identified a handful of peptides derived from ^NTDHCVcp that can compete in the formation of homotypic interactions. Micromolar concentrations of these peptides have been shown to both inhibit oligomerization of ^NTDHCVcp (as determined by fluorescence polarization and surface plasmon resonance) and impair viral replication in cell culture (as determined by immunohistochemistry and RT-PCR) (45). In our single-molecule-binding studies of antiviral peptides, none of the four appeared to stably bind to and alter the end-to-end distance (Fig. 7A) of fluorescently labeled ^NTDHCVcp. However, at mid-micromolar concentrations, three of these peptides (i.e., SL173, SL175, and SL571) formed large, slowly diffusing, oligomeric complexes with ^NTDHCVcp (Fig. 7B). The concentrations of SL175 and SL571 required for half-maximal oligomerization ( ${EC}_{50}$ ) were dependent on NaCl concentration (Fig. 7C), indicating that electrostatic forces mediate the formation of these binding interactions. Curiously, the oligomerization-inducing behavior of these two antiviral peptides was quite similar, which seems to indicate that their ability to oligomerize with ^NTDHCVcp is more strongly dependent on their physicochemical properties rather than their primary amino acid sequence. It is certainly conceivable that the ability of these peptides to oligomerize with ^NTDHCVcp in a manner that is independent of viral RNA is related to their antiviral activity. Similar to the antiviral DNA aptamers, the peptide-induced oligomerization of ^NTDHCVcp could sequester core proteins away from the viral RNA ultimately preventing assembly of viral nucleocapsids—although this has yet to be properly verified.

Small molecules are the most popular therapeutics to date and constitute 90% of the overall market (46). They can be administered orally and penetrate cells to inhibit biomolecular functions. Previous work has shown that several indoline alkaloid-type compounds are capable of inhibiting the oligomerization of ^NTDHCVcp and impairing the viral replication in cell culture (43, 56). We studied one of these molecules (C20) to determine how it binds to and interacts with ^NTDHCVcp. Our investigations showed that C20 does not change any of the experimental observables we are able to monitor with our single-molecule fluorescence approach (Fig. S8). Previously, a study using an ALPHA screen assay reported that C20 inhibits the oligomerization of ^NTDHCVcp with an ${EC}_{50}$ = 0.092 µM (43). If this small-molecule antiviral does indeed bind at micromolar concentrations, our results would suggest that binding does not substantially alter the conformational properties of ^NTDHCVcp.

Interactions between molecules are essential for life, and for this reason, they have been exploited for both diagnostic and therapeutic purposes. Here, we studied binding interactions between fluorescently labeled variants of the N-terminal domain of the hepatitis C virus core protein (^NTDHCVcp) and four different classes of antiviral therapeutics utilizing single-molecule methods. Our findings highlight several distinguishing features associated with these interactions. First, we observed that the low-nanomolar equilibrium dissociation constants of antiviral DNA aptamers arise from strong attractive electrostatic interactions that compact the intrinsically disordered ^NTDHCVcp. Conversely, the binding of high-affinity antibodies does not perturb the end-to-end distance of ^NTDHCVcp and is weakly impeded by electrostatic repulsions between similarly charged residues in the two binding partners. Although both of these antivirals have exceptionally high binding affinities for ^NTDHCVcp, they also have noteworthy limitations. For example, the antibody has a molecular weight of approximately 150 kDa which limits its ability to enter the cell. Although the aptamers are much smaller (13.7 kDa), their binding mechanism is strongly dependent on electrostatic interactions, and as such, they may tightly associate with other highly positively charged biomolecules in the cellular milieu. Next, we observed that low-affinity antiviral peptides bind to ^NTDHCVcp, causing it to oligomerize/aggregate in the absence of viral RNA. While these peptides do associate with ^NTDHCVcp, they are only effective at mid-micromolar concentrations, which ultimately limits their therapeutic utility. Finally, the previously identified small-molecule antiviral compound does not appear to affect any of our experimental observables, suggesting that binding may not alter the conformation properties of HCVcp.