Unique Features of Hepatitis C Virus Capsid Formation Revealed by De Novo Cell-Free Assembly

Abstract

The assembly of hepatitis C virus (HCV) is poorly understood, largely due to the lack of mammalian cell culture systems that are easily manipulated and produce high titers of virus. This problem is highlighted by the inability of the recently established HCV replicon systems to support HCV capsid assembly despite high levels of structural protein synthesis. Here we demonstrate that up to 80% of HCV core protein synthesized de novo in cell-free systems containing rabbit reticulocyte lysate or wheat germ extracts assembles into HCV capsids. This contrasts with standard primate cell culture systems, in which almost no core assembles into capsids. Cell-free HCV capsids, which have a sedimentation value of ≈100S, have a buoyant density (1.28 g/ml) on cesium chloride similar to that of HCV capsids from other systems. Capsids produced in cell-free systems are also indistinguishable from capsids isolated from HCV-infected patient serum when analyzed by transmission electron microscopy. Using these cell-free systems, we show that HCV capsid assembly is independent of signal sequence cleavage, is dependent on the N terminus but not the C terminus of HCV core, proceeds at very low nascent chain concentrations, is independent of intact membrane surfaces, and is partially inhibited by cultured liver cell lysates. By allowing reproducible and quantitative assessment of viral and cellular requirements for capsid formation, these cell-free systems make a mechanistic dissection of HCV capsid assembly possible.

Hepatitis C virus (HCV) affects about 2% of the world's population (53) and is the major cause of non-A, non-B hepatitis, which often leads to cirrhosis of the liver or hepatocellular carcinoma (44). Acute viremia develops into chronic infections in as many as 85% of cases (25), and current treatments are still poorly tolerated and not completely effective (34). Development of better drugs and vaccines has been hampered by the lack of an infectious culture system. Infectious HCV cannot be produced in routine cell culture systems, and the chimpanzee is the only animal model capable of being infected and yielding high HCV titers (18). As a result, the requirements for critical steps in virion formation, including capsid assembly, genome encapsidation, budding, and release, remain largely unknown.

HCV is an enveloped, single-stranded, positive-sense RNA virus in the Flaviviridae family (4). The HCV genome has a single open reading frame that codes for a ≈3,000-amino-acid polyprotein. The core protein is the N-terminal cleavage product from the polyprotein. The polyprotein is targeted to the endoplasmic reticulum (ER) by an internal signal sequence (SS) that is cleaved by signal peptidase and signal peptide peptidase, releasing core into the cytoplasm (37, 45, 57). After release from the polyprotein, core assembles into capsids at the cytoplasmic face of the ER (7, 8, 48). Core is known to interact with the HCV envelope glycoprotein E1 at the ER (38), and assembled capsids are thought to acquire their envelopes by budding into the ER (5, 7, 8, 38). However, the specific details of HCV budding and release have not been elucidated because no standard cell culture models recapitulate these steps.

The development of HCV replicon systems which support autonomous replication of HCV RNA (3, 10, 39, 40) has allowed the specific requirements of HCV RNA replication to be studied. However, HCV capsids are not produced even in full-length replicon systems that express HCV structural proteins at high levels (11, 26). Rather, HCV core appears to localize to lipid droplets and does not colocalize with E1 and E2 at the ER (11). Consistent with this observation, infectious particles are not released from these cells (26). The reason that HCV core fails to assemble into capsids in these systems is not known.

Because cells expressing HCV replicons do not appear to support HCV capsid assembly, a variety of other systems have been used to study the assembly of HCV capsids, including systems that use purified recombinant core, baculovirus-insect cell expression, and mammalian cell culture. In the simplest of these systems, recombinant HCV core is purified, renatured, and assembled in vitro. In the presence of structured RNA, wild-type recombinant core assembles into particles with irregular shapes, while purified C-terminal truncation mutants assemble into regularly shaped capsids that more closely resemble HCV capsids from infected individuals (30). Similar results were obtained by assembling truncated core constructs in Escherichia coli (41). Together, these systems show that HCV core can assemble into capsid-like structures in the presence of RNA (30) and are useful for structural studies (31). However, because they do not contain eukaryotic cellular factors or organelles, they are of limited utility for understanding assembly in mammalian cells.

Some cellular systems have also been used to study capsid assembly. HCV core, when overexpressed with baculovirus vectors in insect cells, assembles into 30- to 60-nm particles at the ER (5, 6, 42). When the envelope proteins E1 and E2 are also expressed, capsids can be seen budding into the ER and cytoplasmic vesicles (5); however, no virus-like particles are released (5, 6, 42). While mammalian cell lines in general do not support capsid assembly, some success has been achieved in mammalian cells with Semliki Forest virus replicons and vesicular stomatitis virus expression vectors (7, 8, 16). Electron microscopic studies reveal that HCV core expressed from these viral vectors forms capsids, but the yield and reproducibility of assembly in these systems have not been assessed quantitatively. Moreover, a very recent study shows that in cultured hepatocytes grown in a radial-flow bioreactor, HCV is able to replicate to low titers (1). While this system holds promise, its ability to produce high titers and be used for a biochemical analysis of capsid assembly remains unclear. To date, no eukaryotic system has been used to systematically and quantitatively identify domains of HCV core important for capsid formation or other requirements of HCV capsid assembly.

The linkage of de novo translation to posttranslational events makes cell-free systems excellent tools for mechanistic studies of cellular processes. In these systems, cellular events are faithfully reproduced in eukaryotic cell extracts that can be manipulated readily, allowing dissection of complex processes. They have resulted in identification of critical machinery involved in protein trafficking (12, 62, 63) and transient events in protein biogenesis (13, 20, 21) and have been used to study assembly of viral capsids (14, 35, 36, 49, 54-56, 64, 66). Furthermore, cell-free systems have resulted in identification of cellular proteins that play important roles during capsid assembly of hepatitis B virus (HBV) and primate lentiviruses, including human immunodeficiency virus type 1 (HIV-1) (14, 23, 24, 66). These systems have even successfully produced infectious poliovirus de novo (49), demonstrating the ability of cell-free system to reconstitute the entire complex process of virion formation. Recently, analogous systems have also been established to study the replication of HCV RNA (2, 19, 32). Here we demonstrate that cell-free systems faithfully reconstitute HCV capsid assembly and use these systems to identify determinants of HCV capsid formation that have not been examined previously.

MATERIALS AND METHODS

DNA plasmids.

Cell-free plasmid vectors were derived from the SP64 vector (Promega) into which the 5′ untranslated region of Xenopus laevis globin had been inserted at the HindIII site (46). The HCV core coding region (C191) from an HCV 1b isolate (obtained from T. Wright, University of California-San Francisco) was amplified by PCR with primers containing BglII and EcoRI sites and inserted into the BglII and EcoRI sites of the vector. Wild-type C173, the C-terminal truncation mutants, and the N-terminal truncation mutants were constructed with an analogous method. For mammalian cellular expression, the same C191 coding region was amplified by PCR and inserted into the NheI and EcoRI sites of the pCDNA 3.1Zeo vector (Invitrogen). All coding regions were verified by sequencing. The prolactin plasmid was obtained from V. Lingappa, University of California-San Francisco.

In vitro transcription and cell-free translation and assembly.

In vitro transcription was performed with the SP6 polymerase (New England Biolabs) and the SP64 expression plasmids described above or H₂O for mock transcripts, as indicated, and followed by cell-free translation and assembly for 120 min at either 26°C with wheat germ extracts or 37°C with rabbit reticulocyte lysate and [³⁵S]methionine (ICN Biochemicals), as described previously (15, 36, 51). When noted in the text, Nikkol was added to a final concentration of 0.1%. Where indicated, rough microsomes from dog pancreas (RMDs) (obtained from V. Lingappa) (60) or Tris buffer (20 mM, pH 7.4) as a control were added to cell-free reactions 3 min after the start of the reaction to 15% of the final reaction volume. For pulse-chase analysis, cell-free reactions were performed as stated above with the exception that [³⁵S]cysteine (ICN Biochemicals) was added instead of [³⁵S]methionine. After a 3-min pulse, excess unlabeled cysteine (30 mM) was added 1:10 to a final concentration of 3 mM. For cell-free reactions in the presence of HepG2 lysate, wheat germ extract was added to 10% of the final reaction volume and HepG2 lysate (or a buffer control) to 30% of the final volume. The reaction buffers were compensated to maintain the same salt concentrations as a typical reaction. The amount of transcript programmed into the buffer control was 1% of the amount programmed into reactions containing HepG2 lysate to maintain equivalent amounts of translation.

Quantitation of core translated in the cell-free system.

Twofold serial dilutions of purified core fused to β-galactosidase (ViroGen) were used as standards in a dot blot. Cell-free reactions with wheat germ extract were programmed with C191 and C173 transcripts with unlabeled methionine. Aliquots (1 μl) of the reactions were analyzed by dot blot in parallel with the standards. Dot intensities from a single film were analyzed by densitometry, and the amount of core for each reaction (corrected to account for the fusion protein) was interpolated with a best-fit line made from the standards. The bands analyzed were within the linear range for signal intensity.

Gradient analysis of cell-free reactions and cellular lysate.

Calibration of gradients to determine S-value positions has been described previously (35, 36); 200 μl of either cell lysate or cell-free assembly reactions, diluted into a 200-μl final volume containing 0.625% NP-40 detergent, 10 mM Tris-acetate (pH 7.4), 50 mM potassium acetate, 100 mM NaCl, and 4 mM magnesium acetate, were layered onto gradients containing sucrose prepared in the same buffer. Except in Fig. 6, velocity sedimentation was performed on step gradients containing 400 μl each of 10, 20, 30, and 40% sucrose with a 200-μl 50% sucrose cushion. Centrifugation of velocity sedimentation gradients was performed at 201,000 × g for 55 min at 4°C in a TLS55 rotor (Beckman Coulter Optima Max centrifuge), and 200-μl fractions were collected serially from the top.

FIG. 6.

High-resolution velocity sedimentation gradients reveal two HCV capsid subpopulations. Higher-resolution 10 to 50% linear sucrose gradients were used to separate wheat germ cell-free reactions programmed with C173, C115, and HBV core transcript. Fractions were analyzed by SDS-PAGE and autoradiography. The graph shows the amount of core in each fraction (quantitated by densitometry) for C173 (solid diamonds), C115 (solid squares), and HBV core (open circles). Arrows indicate the migration of sedimentation markers (80S and 114S); 80S ribosomes were identified by Coomassie staining, and 114S was identified by titering bacteriophage φX174 (open triangles). Shown are representative results from three independent experiments.

Equivalent aliquots of each gradient fraction were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by autoradiography or immunoblotting. Higher-resolution gradients (see Fig. 6) were performed as above but by layering 25-μl sample volumes on linear 10 to 50% sucrose gradients and collecting 100-μl fractions. For equilibrium centrifugation, 100 μl of fraction 6 and 100 μl of fraction 7 (containing ≈100S complexes) from sucrose gradients were pooled and layered onto 2-ml CsCl gradients (337 mg/ml). Equilibrium centrifugation was performed at 166,000 × g for 24 h at 20°C in a TLS55 rotor (Beckman Coulter Optima Max-E centrifuge), and 100-μl aliquots were taken serially from the top of the gradient. Odd fractions were analyzed by refractometry, and even fractions were analyzed by SDS-PAGE and autoradiography.

Serum samples.

Serum samples were obtained from the Molecular Virology Laboratory at the University of Washington, following Institutional Review Board-approved guidelines. Frozen samples were from random, nonidentified patients with high viral loads, quantified by Taqman real time reverse transcription-PCR. Serum samples were thawed and processed with buffer and detergent (1:10) to obtain the same final concentration as our gradient samples and to remove the viral envelope.

RNA quantitation.

RNA was isolated from each fraction with the RNeasy kit (Qiagen) and eluted in 12 μl of water. The number of HCV genomic RNA copies in each gradient fraction was quantitated from 8 μl of the eluate with the TaqMan EZ reverse transcription-PCR kit (Applied Biosystems) and a Taqman machine per the manufacturer's protocol. BB7 plasmid DNA, which contains a subgenomic HCV replicon (9), was used as a standard. BB7 DNA was precisely quantitated by fluorimetry (PicoGreen dsDNA quantitation kit; Molecular Probes). For each run, a standard curve was generated with known concentrations of BB7 (0 to 10⁷ copies). SDS version 2.1 software (ABI) was used to derive an equation describing the standard curve, from which the number of RNA copies per experimental sample was derived (L. Cook, K.-W. Ng, A. Bagabag, L. Corey, and K. R. Jerome, submitted for publication).

Sedimentation markers.

Approximately 10⁷ bacteriophage φX174 particles (obtained from P. Gouldfarb, New England Biolabs) were subjected to velocity sedimentation as described above. A standard plaque assay was used to determine the phage titer in each fraction. Briefly, 100 μl of gradient fractions (or dilutions) was incubated with mid-log-phase E. coli strain H4714 (obtained from P. Gouldfarb, New England Biolabs) for 10 min at 37°C; 3 ml of Luria broth (LB) top agar was added to the bacteria-phage mixture, layered onto LB plates, and incubated overnight at 37°C. Plaques were counted the following day. Ribosomes were isolated from wheat germ extract by collecting the supernatant from two serial centrifugations. The first centrifugation was performed at 174,000 × g for 8 min with a Beckman TLA100.2 rotor. The supernatant from the first step was layered onto a 1.8 M sucrose cushion and centrifuged for 40 min at 430,000 × g in the same rotor. The pellet, containing ribosomes, was resuspended in 0.5 M sucrose-100 mM KCl-40 mM HEPES-5 mM magnesium acetate. The ribosomal preparations were processed in parallel with cell-free reactions on velocity sedimentation gradients.

Electron microscopy.

Cell-free reactions were subjected to velocity sedimentation centrifugation, and fractions 6, 7, and 8 were pooled and dialyzed (55,000-molecular-weight cutoff) for 1 h against phosphate-buffered saline at room temperature. Dialyzed samples were then settled onto carbon-coated grids (Ted Pella) for 2 to 3 min, stained with 1% uranyl acetate for 30 s, and visualized with a Jeol 1010 or Jeol 100SX transmission electron microscope. An experienced electron microscopist identified and examined reactions and controls in single-blinded fashion in three separate experiments. Histograms were prepared by measuring the diameters of all capsids in three to five fields (equivalent to 80 to 300 capsids). Criteria for excluding particles were determined by analyzing particles in the unassembled fractions (i.e., fractions 3, 4, and 5) of patient serum. These particles were all less than or equal to 25 nm (data not shown). Thus, particles of this size were excluded from diameter measurements.

Protease digestions.

Digestion reactions were programmed with different concentrations of proteinase K (Roche) diluted in 20 mM Tris (pH 7.4) with 150 mM NaCl. Cell-free reactions (0.5 μl) were added to the digestion reactions (final volume, 10 μl). Digestions were incubated at room temperature for 15 min, inactivated with 10 μl of SDS protein loading buffer, immediately boiled for 10 min, and analyzed by SDS-PAGE and autoradiography.

Transfections and cell harvests.

Cells were transfected in 60-mm dishes with 4 μg of pCDNA-C191 and either 24 μl of Lipofectamine (Invitrogen; COS-1 cells), 12 μl of Lipofectamine and 8 μl of Plus reagent (Invitrogen; 293T cells), or 15 μl of Lipofectamine 2000 (Invitrogen; HepG2 and Huh-7 cells) with standard protocols. Cells were harvested 30 to 40 h posttransfection. Cells were washed with phosphate-buffered saline and harvested on ice in 300 μl of buffer containing 0.625% NP-40, 10 mM Tris-acetate (pH 7.4), 50 mM potassium acetate, 100 mM NaCl, and 4 mM magnesium acetate in addition to 1 mM phenylmethylsulfonyl fluoride. Lysates were sheared by 25 passes through a 20-gauge needle and clarified by centrifugation at 365 × g for 5 min at 4°C in a GH-3.8 rotor (Beckman Coulter Allegra 6R centrifuge) and 18,000 × g for 20 s in a conventional microcentrifuge.

HepG2 cellular lysate preparation.

Cells in 60-mm dishes were washed with phosphate-buffered saline, scraped up, and pelleted at 50 × g for 2 min at 4°C in a GH-3.8 rotor (Beckman Coulter Allegra 6R centrifuge). The phosphate-buffered saline was removed, and the cells were resuspended in an equal volume (compared to cell volume) of buffer containing 40 mM HEPES, 100 mM potassium acetate, 5 mM magnesium acetate (pH 7.4), and 0.01% Nikkol. Cells were then sheared by 25 passes through a 20-gauge needle.

Immunoblotting.

Samples were subjected to SDS-PAGE and transferred onto nitrocellulose membranes (Osmonics, Inc.). Immunoblotting was performed with anti-HCV core (Affinity Bioreagents) at a concentration of 1 μg/ml. Anti-mouse immunoglobulin G coupled to horseradish peroxidase (Santa Cruz) was used as the secondary antibody, and enhanced chemiluminescence was performed (Pierce).

Quantitation.

Autoradiographs and Western blots were digitized with an AGFA Duoscan T1200 scanner and Adobe Photoshop 5.5 software (Adobe Systems Incorporated). Mean band densities were determined and adjusted for band size and background. Assembly profiles were generated by plotting the amount of total core present in each fraction as a percentage of total core in all gradient fractions.

Statistics.

P values were obtained with a paired Student t test.

RESULTS

Cell-free translations produce HCV core proteins of the expected sizes.

HCV core is the N-terminal cleavage product of the HCV polyprotein. Two major forms of core exist in vivo. The first results from cleavage by signal peptidase at amino acid 191, which releases the N-terminal 191 amino acids of the HCV polyprotein. Signal peptide peptidase then cleaves the E1 signal sequence (SS) from HCV core between amino acids 173 and 182, releasing the mature form of core that is the primary species of core in the virion (65). Plasmids coding for both core containing the downstream SS (C191) and the mature form of core (C173) were constructed, as diagramed in Fig. 1A. Translations programmed with transcript encoding either C191 or C173 produced radiolabeled HCV core of the expected sizes (Fig. 1B), corresponding to what has been seen by others (57). We next determined how much HCV core protein was made in a typical cell-free reaction. A standard curve was generated by immunoblotting purified HCV core protein. With the standard curve, we demonstrated that typical reactions containing wheat germ extracts programmed with C173 and C191 transcript produced about 35 and 21 ng of newly synthesized core protein per microliter of translation reaction mixture, respectively (Fig. 1C).

FIG. 1.

Quantitation of the amount of core produced in the cell-free system. (A) Schematic representation of the HCV genome and the wild-type core constructs used here. C191 contains the E1 signal sequence (SS) at its C terminus, and C173 encodes the mature form of core lacking the SS. (B) Radiolabeled cell-free reactions containing wheat germ extracts were programmed with transcripts encoding C173 (lane 1) and C191 (lane 2). Aliquots were analyzed by SDS-PAGE and autoradiography, which shows the expected difference in migration. (C) Cell-free reactions were performed as above but with unlabeled methionine. Reactions were analyzed by dot blot and Western blotting. Dilutions of purified core were analyzed in parallel. Shown is a graph of the band intensities for the HCV core dilutions (gray diamonds) in the linear range of the Western blot. The graph was used to determine the amounts of core produced in cell-free reactions, which are shown superimposed on the graph (black squares) and in the table.

Wheat germ and rabbit reticulocyte lysate cell-free systems produce HCV capsids.

We next determined whether the HCV core protein, independent of other structural proteins, is sufficient for assembly of HCV capsids in cell extracts. As a positive control for migration, we used authentic HCV capsids from infected patient serum treated with detergent to remove the viral envelope. Detergent-treated serum was subjected to velocity sedimentation on sucrose gradients, and fractions from parallel gradients were analyzed by reverse transcription-PCR (Fig. 2A, graph) and Western blotting with an antibody specific for HCV core (Fig. 2A, blot). A large peak of HCV RNA and a small peak of HCV core protein were seen in fractions 7 and 8, corresponding to a sedimentation value of approximately 100S. Given that HBV capsids are similar in size and have a sedimentation coefficient of approximately 100S (36), this was an expected sedimentation coefficient for an HCV capsid. Additionally, HCV core but not HCV RNA was present at the top of the gradient, most likely reflecting soluble HCV core present in the serum or capsids that had not stayed intact during processing. However, these analyses of both HCV RNA and core protein suggest that intact HCV capsids migrate in fractions 7 and 8 on our sucrose step gradients.

FIG. 2.

Sedimentation coefficients and buoyant densities of HCV capsids from cell-free reactions and infected patient serum. (A) HCV-infected patient serum was treated with detergent to remove the viral envelope and analyzed by velocity sedimentation. RNA was extracted from each fraction, and HCV genomic RNA was quantitated. The graph shows the average amount of HCV genomic RNA in each fraction from three independent gradients; error bars represent the standard error of the mean. The Western blot (below) of a parallel gradient shows the amount of HCV core protein in each corresponding fraction. Fraction 1 is from the top of the gradient. (B to D) Cell-free reactions containing wheat germ extracts (WG) or rabbit reticulocyte lysate (RRL) were programmed with C191 mRNA transcribed in vitro. Translation and assembly reaction products were analyzed after a 2-h incubation. (B) Sucrose gradient velocity sedimentation profile of a typical cell-free reaction containing wheat germ extract. Fractions were analyzed by SDS-PAGE and autoradiography. The graph shows the amount of core in each fraction (as a percentage of total core) determined by densitometry. The bar represents fractions containing ≈100S particles (i.e., fractions 6, 7, and 8). Approximate sedimentation coefficients are indicated with arrows. Below the graph is a representative autoradiograph. Radiolabeled bands that migrated faster than full-length core represent early termination and late initiation products frequently seen in cell-free translations. (C) Fractions 6 and 7 from a velocity sedimentation gradient containing ≈100S particles (see panel B) were pooled and subjected to buoyant density analysis on CsCl gradients, and core in even fractions was analyzed as in panel B. The density of odd fractions was determined by refractometry and is graphed as grams per milliliter, demonstrating that core peaks in fraction 12 at a density of 1.28 g/ml. Below the graph is a representative autoradiograph. (D) Velocity sedimentation profile of a typical cell-free reaction containing rabbit reticulocyte lysate. Reactions were analyzed by velocity sedimentation and graphed as in panel B. Below the graph is a representative autoradiograph. Shown are representative results from three independent experiments.

We next asked whether newly synthesized core in our cell-free reactions containing wheat germ extracts and in vitro transcript encoding C191 assembled into HCV capsids. Cell-free reactions were analyzed by velocity sedimentation on sucrose gradients, as above. Velocity sedimentation revealed that newly synthesized C191 assembled into particles that peaked in fraction 7 of our gradients (Fig. 2B), the same fraction that contained a peak of HCV core from the serum samples. To confirm that the ≈100S particles produced in the cell-free system were in fact capsids, we subjected them to equilibrium centrifugation on cesium chloride gradients. Fractions 6 and 7 from velocity sedimentation gradients containing the ≈100S particles was layered onto CsCl gradients. Figure 2C shows that ≈100S HCV particles from the cell-free system had a density of 1.28 g/ml, which is equivalent to the density of capsids assembled from purified core in isolation (30) and similar to the density reported for HCV capsids from insect cells as well as infected patients and chimpanzees (5, 27, 47, 48). Thus, by two biochemical analyses, expression of C191 in a cell-free system programmed with wheat germ extract resulted in the formation of capsids. Similar results were obtained upon expression of C173 in the cell-free system (data not shown; see Fig. 5).

FIG. 5.

The N terminus of core appears to be important for capsid assembly. (A) Schematic of the truncation mutants that were constructed, including wild-type C191 containing the E1 signal sequence (SS). (B) Wheat germ cell-free reactions were programmed with the indicated constructs, and assembly for each construct was calculated as in Fig. 2B. The amount of assembled core as a percentage of total core synthesized for each reaction is graphed. The total amount of translation for each reaction was also measured by SDS-PAGE, autoradiography, and densitometry and is shown as a line graph. Error bars represent standard error of the mean from three independent experiments. Constructs with statistically different amounts of assembly (* = P < 0.05 and ** = P < 0.01) relative to both C173 and the assembly-incompetent construct ΔN68 are indicated. (C) Representative autoradiographs from velocity sedimentation analyses of wild-type C173, an assembly-competent mutant (C115), a partial assembler (ΔN20), and an assembly-incompetent mutant (ΔN68) are shown. (D) An assembly-competent mutant (C115) and an assembly-incompetent mutant (ΔN68) were analyzed by TEM, as in Fig. 3. Bars, 100 nm. (E) Distribution of capsid diameters, shown as a histogram, for C115.

To determine whether HCV assembly can occur in other extracts, we tested whether core could assemble in rabbit reticulocyte lysate, a mammalian cell extract. Rabbit reticulocyte lysate was programmed with the C191 transcript, and reaction products were analyzed by velocity sedimentation. C191 assembled into ≈100S capsids very efficiently, with about 70% of newly synthesized core chains present in the ≈100S fractions (Fig. 2D). Thus, like the wheat germ extracts, the rabbit reticulocyte lysate supported efficient HCV capsid assembly, and because of the similar velocity sedimentation profiles, the studies below were done with wheat germ extracts unless indicated. Note that one striking feature of both extracts is the relative absence of HCV core in the top fractions that contained monomeric protein and small complexes.

Capsids from the cell-free system closely resemble capsids from patient sera.

To confirm the biochemical data suggesting that newly synthesized HCV core forms capsids in the cell-free system presented in Fig. 2, ≈100S particles were isolated by velocity sedimentation and analyzed by negative staining and transmission electron microscopy (TEM). Frozen sera from infected patients were treated with nonionic detergent to remove viral envelopes and analyzed in parallel. TEM revealed that ≈100S particles from the cell-free system were 33 to 46 nm in diameter, with a size range, size heterogeneity, and morphological appearance similar to those isolated from patient sera (Fig. 3, compare panels A and B to C). High-magnification views showed more ultrastructural detail in both sets of capsids (Fig. 3A and C, insets). Reactions programmed with mock transcript and analyzed in parallel did not contain any capsids (data not shown).

FIG. 3.

The ≈100S HCV capsids from the cell-free system (CFS) are morphologically similar to authentic capsids. (A and B) Cell-free reactions programmed with C191 transcript were separated by velocity sedimentation, and fractions 6, 7, and 8 (containing ≈100S capsids) were pooled, dialyzed against phosphate-buffered saline, and subjected to negative-stain TEM. (C) Serum from an infected patient, treated with detergent to remove envelopes, was processed in parallel. Bars, 100 nm. The insets in A and C contain 6× magnifications of individual capsids relative to the rest of the panel. Shown are representative results from three independent experiments. (D) Distribution of capsid diameters, shown as histograms, for capsids from the CFS and from patient serum.

Individual capsid diameters were measured, and their size distributions are shown in Fig. 3D. The histograms show that capsids produced in the cell-free system and in infected patients had a similar bimodal frequency, with one peak at a diameter size of 33 nm. However, the second peaks seen for the two types of capsids were different, causing the average capsid size of serum samples to be slightly smaller than that of cell-free HCV capsids. This discrepancy may reflect differences in handling of the samples, since cell-free capsids were processed fresh, while patient serum had been repeatedly frozen. Nevertheless, the appearance, heterogeneity in size, and structure of HCV capsids from both patient sera and the cell-free system are consistent with what others have reported for HCV capsids isolated from insect cells (5, 6, 33), chimpanzees (58), and infected humans (17, 27, 42). Taken together, these findings show that HCV core translated in the cell-free system assembles into capsids that are nearly indistinguishable from capsids produced in HCV-infected patients by biochemical and morphological criteria. Therefore, we used this system to examine various requirements of HCV capsid assembly.

HCV capsid assembly is independent of SS cleavage.

As described above, HCV core requires two cleavage events to release it from the HCV polyprotein. The first cleavage produces C191, in which the E1 SS is still attached to core, and the second produces C173, in which the SS is completely removed (Fig. 1A). As expected, there were no cleavage events in the reactions presented in Fig. 2. To reconstitute SS cleavage and test whether SS cleavage alters the ability of core to assemble, we programmed cell-free reactions with C191 transcript in the presence or absence of ER-derived rough microsomes from dog pancreas (RMDs), which contain the peptidases necessary for HCV SS cleavage (37, 45, 57). Total cell-free reactions were analyzed by SDS-PAGE and autoradiography to assess SS cleavage (Fig. 4A) and by velocity sedimentation to assess assembly (Fig. 4B). As expected, only in the presence of RMDs was C191 efficiently cleaved into the processed form (Fig. 4A, lanes 1 and 2). When analyzed by velocity sedimentation, the profiles of both reactions were very similar, with the majority of core migrating as ≈100S capsids (Fig. 4B). When the cell-free capsids produced in the presence of RMDs were examined by TEM, they appeared morphologically identical to the capsids produced in the absence of RMDs (data not shown). Taken together, these data suggest that the cell-free system supports SS cleavage but that the SS cleavage event does not alter the ability of core to assemble into capsids.

FIG. 4.

HCV capsid assembly is not affected by SS cleavage or membranes. Cell-free reactions containing wheat germ (WG) were programmed with transcripts encoding C191 or the secretory protein prolactin containing its N-terminal SS (SS-PRL), in the presence or absence of ER-derived microsomes (RMDs), as noted. (A) An autoradiograph of aliquots of total translation showing SS cleavage in the presence of RMDs, converting C191 to C173 (compare lanes 1 and 2). C191 and C173 translated in wheat germ without RMDs are also shown as a reference for migration (lanes 3 and 4). Conversion of SS-PRL to mature prolactin (PRL) is shown as a positive control for SS cleavage (lanes 5 and 6). (B) The HCV cell-free reactions from A were analyzed by velocity sedimentation on sucrose gradients, and the autoradiographs show velocity sedimentation profiles for each reaction. (C) Aliquots of cell-free reactions from A were digested with increasing concentrations of proteinase K (PK), as indicated. An autoradiograph shows the susceptibility of the translated products to proteolytic degradation. (D) Cell-free reactions were programmed with C191 transcript in the presence or absence of the nonionic detergent Nikkol. Reactions were analyzed by velocity sedimentation, and the amount of assembled core as a percentage of total core synthesized for each reaction is graphed. An autoradiograph of total translations also shows that Nikkol did not affect the amount of core translation. Shown are representative results from three independent experiments.

To determine whether the capsids formed in the cell-free system programmed with HCV core alone bud into ER-derived vesicles, cell-free reactions performed in the absence and presence of RMDs were treated with various concentrations of proteinase K and analyzed by SDS-PAGE and autoradiography. If the capsids are completely enveloped by a lipid bilayer, they should be protected from protease digestion. C191 translated in the presence or absence of RMDs was equally susceptible to proteinase K digestion (Fig. 4C, top and middle panels). In a parallel reaction, the SS of prolactin, a secretory protein commonly used as a control, was cleaved only in the presence of RMDs (Fig. 4A, lanes 5 and 6). ER-derived membranes protected cleaved prolactin from proteolysis, while uncleaved PRL that had not been translocated remained susceptible to protease digestion (Fig. 4C, bottom panel, and data not shown). Thus, while ER-derived membranes protected cleaved prolactin, cleaved HCV core remained protease sensitive. Taken together, these data suggest that although HCV core is cleaved by ER-associated peptidases in the cell-free system, the capsids composed of core were not fully enveloped and protected by ER-derived membranes.

The above data suggest that RMDs, which efficiently mediate cleavage of the SS, are not required for HCV capsid assembly; however, they do not address the question of whether capsid assembly is dependent on endogenous membranes present in the cellular extracts. Such a dependence of capsid assembly on membranes present in cellular extracts has been seen for HIV-1 capsid assembly (35). Therefore, cell-free reactions were performed after treating cell extracts with the nonionic detergent Nikkol to solubilize endogenous membranes. Nikkol had no effect on the total amount of translation (Fig. 4D) (35, 61). These reactions were then analyzed by velocity sedimentation to determine the amount of assembled core. The addition of Nikkol (0.1%) at the beginning of the reactions did not affect HCV capsid assembly in either wheat germ extract or rabbit reticulocyte lysate (Fig. 4D). This is in contrast to cell-free HIV-1 capsid assembly, which was dramatically reduced by Nikkol treatment (data not shown) (35). These data further suggest that although ER-derived membranes are required for SS cleavage, the presence of intact membrane surfaces is not required for assembly of HCV core into capsids per se.

The N terminus of core is important for capsid assembly.

Many HCV core mutants are degraded when expressed in mammalian cells (50, 59). In contrast, when cell-free reactions are programmed with transcripts encoding mutations, the amount of translation is typically similar to that of wild-type reactions, making the cell-free system an excellent tool for determining the effects of certain domains on capsid assembly in a cellular context. To identify domains in core that are important for capsid assembly, a panel of HCV core expression plasmids encoding serial truncations in the N and C termini of HCV core were constructed (Fig. 5A). The ability of these truncation mutants to assemble in the cell-free system with wheat germ extracts was assessed by velocity sedimentation. The amount of translation of N- and C-terminal truncations of HCV core constructs was roughly equivalent to that of wild-type core (Fig. 5B, line graph). Deleting the C terminus of core had little or no effect on capsid assembly, while serial truncations of the N terminus progressively reduced capsid assembly (Fig. 5B, bar graph, and 5C). Deleting the first 10 N-terminal residues (ΔN10) had no effect on capsid assembly; however, deleting the N-terminal 20 or 30 residues (ΔN20 and ΔN30) decreased assembly to 30 to 40% of the wild-type level. Deleting the N-terminal 42 or 68 residues (ΔN42 and ΔN68) completely abolished capsid assembly. The intermediate level of assembly of ΔN20 and ΔN30 was statistically different from the amount of assembly of both C173 and ΔN68 (P < 0.05 and P < 0.01, respectively; Fig. 5B). As shown in Fig. 5C, as the amount of assembled core decreased, a proportionate increase in core was observed in the top fractions. This shift was most striking for the mutant ΔN68 but was also apparent for all constructs with decreased assembly.

To verify that the ≈100S particles from the assembly-competent truncation mutants were indeed capsids, samples from cell-free reactions programmed with assembly-competent and assembly-incompetent mutants were analyzed morphologically. TEM revealed that the assembly-competent C115 mutant formed particles that were very similar in morphology but somewhat larger in size than capsids produced from wild-type core, while the assembly-incompetent ΔN68 mutant failed to form any capsid-like structures (Fig. 5D; compare to Fig. 3). The size distribution of capsids composed of C115 is presented in Fig. 5E, which shows that C115 formed capsids that were 34 to 75 nm in diameter. The finding that C-terminal truncation mutants formed larger capsids is consistent with what was observed in studies of recombinant core encoding truncation mutations (30, 41). Thus, these data indicate that the C terminus of core beyond residue 115 is not absolutely required for assembly but does affect the size of the capsids formed. In contrast, biochemical and morphological studies suggest that a region of the N terminus may play a critical role in capsid formation.

Higher-resolution gradients shows two distinct subpopulations of cell-free HCV capsids.

Our initial studies were analyzed on sucrose step gradients, which yield relatively crude measurements on velocity sedimentation coefficients. To better resolve the velocity sedimentation patterns of our wild-type capsids and the mutant C115 capsids, cell-free reactions programmed with both C173 and C115 transcript were analyzed on linear 10 to 50% sucrose gradients (Fig. 6). Multiple sedimentation markers were analyzed on parallel gradients, including ribosomes as an 80S marker and the bacteriophage φX174 as a 114S marker (43). The ribosomes and φX174 peaked in fractions 11 and 13, respectively. C173 had two distinct peaks, one in fraction 10 and another broader peak in fraction 13, while C115 only peaked in fraction 11. The two peaks seen with C173 may be consistent with the bimodal distribution seen for capsid diameters in Fig. 3D and by others for capsids in patient serum (42). However, we cannot rule out that only one of the HCV peaks contains HCV capsids of both sizes, since TEM studies were performed with pooled samples that contained both peaks. In contrast to HCV, HBV capsids synthesized in the cell-free system, which have a uniform diameter of ≈30 nm by TEM (36), had only a single peak in fraction 10, consistent with the homogenous size of HBV capsids. Thus, the ≈100S capsids on our step gradients consist of distinct subpopulations of just under 80S and about 114S when analyzed on higher-resolution linear sucrose gradients.

Capsid assembly proceeds at extremely low concentrations and at a high rate.

In systems that use purified, recombinant proteins, millimolar concentrations of capsid protein are typically required for assembly to occur. Given the striking efficiency of capsid assembly in our cell-free systems under standard conditions where core is present at a concentration of 1 to 2 μM, we examined whether HCV core would assemble at lower concentrations in the cell-free system. Cell-free reactions were programmed with dilutions of HCV core transcript over a 3-log range, and assembly was analyzed by velocity sedimentation. The total amount of core translated decreased in a roughly linear manner with transcript dilution (Fig. 7A). When these reactions were analyzed by velocity sedimentation, we found that a significant amount of core had assembled into capsids, even at an HCV core concentration of ≈5 nM (Fig. 7B). Note that the amount of core produced in the most dilute reaction is equivalent to ≈25 pg of core/mg of total cellular protein (≈5 nM), which is less than the 75 pg of core/mg of total cellular protein that has been reported in cellular systems (1). Even at 25 pg of core/mg of protein, ≈30% of HCV core assembled into 100S capsids (0.05% transcript in Fig. 7B). Thus, reduction of core synthesis in the cell-free system by 200-fold resulted in only a 2.3-fold decrease in assembly (Fig. 7A and B). These data indicate that assembly of HCV core in a cytoplasmic environment can occur at nanomolar core concentrations and is only modestly affected by core concentration.

FIG. 7.

Effect of core concentration on HCV assembly. (A) HCV cell-free assembly reactions were programmed with different percentages of C173 transcript (as indicated). Mock in vitro transcriptions were also added to compensate for volume. Equivalent aliquots were analyzed by SDS-PAGE and autoradiography. The total amount of C173 translated for each reaction was determined by densitometry. Each reaction is graphed as a percentage of maximum translation (i.e., reaction with 100% transcript), and values next to the points indicate the exact percentage of C173 transcript that was used. Below the graph is a representative autoradiograph of the total translations. The lower autoradiograph is a longer exposure of the same gel, showing the amount of core in the most dilute reactions. (B) The above reactions were analyzed by velocity sedimentation, as in Fig. 2B. The amount of assembled core as a percentage of total core synthesized for each reaction is graphed. Error bars represent standard error of the mean from three independent experiments. (C) A cell-free reaction was programmed with C173 transcript and pulsed with [³⁵S]cysteine for 3 min. Reactions were then chased for the indicated times with excess unlabeled cysteine. Reaction products were analyzed by velocity sedimentation. The total amount of radiolabeled core is represented as a line graph, and the amount of assembled core shown as a percentage of total core synthesized for each reaction is represented as a bar graph. The experiment was performed three independent times, and a representative experiment is shown.

To elucidate the rate of capsid assembly, pulse-chase analyses were performed. Wheat germ cell-free reactions were programmed with C173 transcript in the presence of radiolabeled [³⁵S]cysteine for 3 min and then chased with excess unlabeled cysteine. At various times during the chase with unlabeled cysteine, reaction products were analyzed by velocity sedimentation to determine what percentage of the initial radiolabeled cohort of core polypeptides had assembled into ≈100S capsids. The majority of radiolabeled core (50%) was assembled into ≈100S capsids at 7 min, the earliest time point at which full-length core could be readily detected (Fig. 7C, bar graph). While the total amount of radiolabeled core continued to increase for the first 15 min (Fig. 7C, line graph), reflecting continued incorporation of [³⁵S]cysteine after addition of unlabeled cysteine, there was little change in the percentage of core that was fully assembled. These data are consistent with an extremely rapid rate of assembly for newly synthesized core polypeptides. Similar results were seen with lower concentrations of core and at different time points (data not shown). These findings indicate that in the cell-free system, not only does core assemble efficiently even at very low concentrations, it does so at a very rapid rate.

Biochemical analysis reveals that ≈100S HCV capsids are not produced in mammalian cells.

Others have found that in most standard mammalian cell culture systems, HCV capsids are formed inefficiently, if at all (7, 8, 11, 16, 52). However, these studies have largely been performed with electron microscopy, which does not lend itself readily to quantitative analysis. The finding that assembly occurs when HCV core is present at very low concentrations in cellular extracts led us to ask whether we could detect HCV assembly in cell culture with biochemical analyses. We transfected various mammalian cell lines, including human liver cell lines, with a C191 expression vector and analyzed detergent lysates of transfected cells by velocity sedimentation and Western blotting for the presence ≈100S capsids. Although core was expressed in all mammalian cell types examined, little or no core was detected in ≈100S fractions compared to cell-free reactions, where the majority of core was present in these fractions (Fig. 8A and B). Instead, core appeared to be present in the top of the gradient and in the pellet. It is unclear why core sediments to the bottom of the gradient in mammalian cells; however, similar complexes can be seen in the cell-free system (for example, see Fig. 2B). These findings are consistent with the morphological reports of others that capsid assembly is inefficient, at best, in cells, in marked contrast to the cell-free systems described here.

FIG. 8.

Cultured mammalian cells inhibit capsid assembly. (A and B) A core expression plasmid was transfected into mammalian cell lines Cos-1, 293T, HepG2, and Huh-7. HCV core-expressing cell lysates were analyzed by velocity sedimentation. Fractions were subjected to SDS-PAGE followed by immunoblotting with an antibody specific for HCV core. Cell-free reactions containing either wheat germ or rabbit reticulocyte lysate were programmed with C191 transcript and analyzed by velocity sedimentation, SDS-PAGE, and autoradiography. Bands were quantitated by densitometry. (A) Amount of assembled core (as a percentage of total core) in the cell-free systems and in transfected mammalian cells is graphed. (B) Autoradiographs and Western blots from panel A, showing velocity sedimentation analysis of radiolabeled core translated in cell-free systems and core expressed in mammalian cells. Shown are representative results from three independent experiments. (C) Buffer or lysate from HepG2 cells (HepG2 Lys) was added to wheat germ cell-free reactions programmed with C191 transcript. Reaction products were analyzed by velocity sedimentation as in Fig. 2B. Shown are representative autoradiographs of the velocity sedimentation gradients. The amount of core present in fractions corresponding to the top (fractions 1 to 3), assembled region (fractions 6 to 8), and the pellet (fraction 10) of the gradient is graphed as a percentage of total core. The inset contains an autoradiograph of aliquots of total reaction products (lane 1, CFS plus buffer; lane 2, CFS plus HepG2 lysate), showing comparable core translation. Error bars represent the standard error of the mean from three independent experiments. Statistically significant differences (** = P < 0.005) relative to control cell-free reactions (+ buffer) are indicated.

Therefore, the cell-free system is highly permissive for capsid assembly, while mammalian cell culture systems are restrictive. This led us to examine whether lysates made from mammalian cells could inhibit capsid assembly in the cell-free system. Towards this end, we made HepG2 cellular lysates in buffer containing 0.01% Nikkol detergent and added them to wheat germ cell-free reactions programmed with C191 transcript. To compensate for decreased translation efficiency in the presence of HepG2 lysate, we programmed our control reaction with reduced levels of transcript, which had a minimal effect on capsid assembly (see Fig. 7). This normalized the total amount of core synthesis in the control reaction and the reaction containing HepG2 lysate to comparable levels (Fig. 8C, inset). When we tested the ability of newly synthesized HCV core to assemble in both of these reactions, we found that the amount of assembled core was significantly reduced by about one-third in the presence of HepG2 lysate (Fig. 8C). In addition, a corresponding increase was seen in the amount of core present in the top fractions, along with a smaller increase in the pellet (Fig. 8C), which is where core migrates when expressed in mammalian cells (Fig. 8B).

Both the shift in core migration and the inhibition of assembly were significant (P < 0.005) and reproducible (see error bars in Fig. 8C) and were not due to excess buffer or detergent, which were present at equal amounts in both reactions. The addition of extra wheat germ to the reactions did not alter the assembly profile (data not shown), suggesting that these effects are specific to the HepG2 lysate. Furthermore, the observed effects were dose dependent, with less HepG2 lysate causing less inhibition (data not shown). Thus, it appears that in addition to being restrictive to HCV capsid assembly, mammalian cells contain factors that can inhibit capsid assembly in the cell-free system, albeit incompletely. Moreover, the appearance of core in the top and pellet fractions of the gradient in the presence of mammalian extracts suggests that the cellular lysate is inhibiting HCV assembly in a manner similar to that seen in cells.

DISCUSSION

Here we demonstrate for the first time that cell-free systems containing either wheat germ extract or rabbit reticulocyte lysate faithfully and robustly reconstitute HCV capsid assembly, a key step in virion formation. The fact that both of these cellular extracts support efficient capsid assembly suggests that any cellular factors required for HCV assembly are most likely conserved across the plant and animal kingdoms. Conversely, these findings also raise the possibility that potential inhibitors of HCV capsid assembly may only be present in specific cell types or under specific physiologic conditions. We used these cell-free systems to define the features of capsid assembly that have not been addressed previously because experimental systems for quantitative analysis of HCV capsid formation have been lacking.

We found that HCV capsid assembly in these systems is independent of the presence of the HCV envelope glycoproteins and HCV nonstructural proteins. In contrast to others, who have proposed that the E1 SS may be important for capsid formation (7), we demonstrate that HCV capsid assembly occurs similarly in the presence and absence of the E1 SS. In addition, assembly of HCV capsids occurs in the presence and absence of ER-derived membranes that support SS cleavage, does not require other intact membrane vesicles, and occurs in the absence of complete envelopment. Together, these findings support a model in which assembly of preformed HCV capsids most likely occurs cytoplasmically, possibly in close proximity to the ER given efficient cleavage of the SS, and can be dissociated from the subsequent events of budding and envelopment. This is consistent with data from others indicating that capsids produced in the absence of envelope proteins are present in the cytoplasm of insect cells (42) and that large numbers of unenveloped nucleocapsids are present in the cytoplasm of hepatocytes from infected patients (17). Additionally, we found that HCV core assembles into capsids efficiently even at nanomolar concentrations in the cell-free system. This is surprising in light of the failure of HCV to assemble efficiently in cell culture systems. Finally, we found that HCV capsid assembly is markedly dependent on the N terminus of core. Thus, using cell extracts, we present a first view of early events in HCV capsid formation that fail to occur efficiently in cultured mammalian cells.

It should be noted that the buoyant density of cell-free HCV capsids corresponds to one of two different buoyant densities that have been reported for HCV capsids. Maillard et al. (35) found that HCV nucleocapsids isolated by detergent treatment from insect cells expressing HCV core migrated in two peaks on CsCl, a major peak at a density of 1.32 to 1.34 g/ml and a minor peak at ≈1.25 g/ml. Capsids in the lower-density fraction were very similar by electron microscopy to those in the higher-density fraction and those from patient sera except that they were associated with membrane fragments (42). Nucleocapsids isolated by detergent treatment from virions present in patient serum have been reported to have a density of 1.35 g/ml on cesium (42), and 1.23 to 1.27 g/ml in sucrose (27, 28). Possible explanations for the different capsid densities include association of the highly lipophilic HCV core protein (22, 45) with lipids in some situations and packaging of different amounts of RNA. While the issue of HCV capsid density requires further investigation, the 1.28-g/ml density value obtained for cell-free capsids corresponds closely to the 1.25-g/ml density seen for at least one population of HCV capsids produced in cells (42).

In addition to faithfully reconstituting capsid assembly, the cell-free systems described here offer other important experimental advantages. In a typical reaction, about 60 to 80% of newly translated HCV core assembles into capsids, making it much more efficient than other cell-free capsid assembly systems, in which 15 to 30% of newly synthesized polypeptides assembles into capsids when synthesized at similar concentrations (35, 36). The robustness and reproducibility of cell-free HCV capsid assembly allow even subtle differences in capsid assembly to be easily quantified. This is illustrated by our ability to demonstrate statistically significant differences between truncation mutants that result in intermediate levels of assembly (ΔN20 and ΔN30) versus assembly-competent constructs (for example, wild-type C173) or assembly-incompetent mutants (for example, ΔN68), as shown in Fig. 5. An assembly assay of such sensitivity will be very useful in the future for distinguishing whether mutations or biochemical manipulations have subtle or dramatic effects on assembly.

Another advantage of these cell-free systems is that they contain little protease or proteasome activity, resulting in stable expression of both wild-type and mutant HCV core proteins, in striking contrast to cellular systems. When C-terminal truncation mutants are expressed in mammalian cells, they are transported to the nucleus and degraded by the proteasome (30, 50, 59). While proteasome inhibitors reduce this problem to some extent (30, 50, 59), the toxicity of these inhibitors to cells limits their ability to be used. Thus, even as better cellular systems for capsid assembly are developed, mutational analyses of HCV core could continue to be hampered by the propensity of core mutants to be degraded in mammalian cells. Thus, the stability of core mutants in cell-free systems will permit a much-needed systematic and quantitative analysis of the effect of deletions, point mutations, and charge substitutions in various domains of HCV core on the assembly of HCV capsids in a cellular context.

The ability of HCV core to assemble into capsids in the cell-free system is in stark contrast to what happens when HCV core is expressed in standard mammalian cell culture systems. Upon expression of HCV core in current replicon systems (52), there appears to be no capsid assembly even when structural proteins are expressed in human liver cell lines. In other mammalian expression systems, HCV capsids can be visualized by electron microscopy but do not appear to be stable enough to isolate, quantify, or study biochemically (7, 8, 16). Our biochemical analyses of HCV core in a variety of mammalian cells are in agreement with these findings (Fig. 8). Consistent with this, we did not observe HCV capsids when we used TEM to analyze Cos-1 cells expressing HCV core (data not shown). Together, these findings raise the possibility that cultured cells contain an inhibitor of capsid assembly.

The possibility of such inhibitory factors is supported by our finding that capsid assembly can be inhibited by addition of HepG2 lysate to our highly permissive cell-free assembly reaction (Fig. 8C and data not shown). Addition of HepG2 lysate resulted in a reproducible and significant partial inhibition in the amount of assembled core, and this effect was dose dependent. Addition of HepG2 lysate caused the appearance of unassembled core in low-molecular-weight complexes, which was not seen at all in standard cell-free reactions. Furthermore, we observed HCV core in low-molecular-weight complexes migrating at the top of the gradient in all situations in which there was negligible or partial assembly, including in mammalian cell lines, upon expression of assembly-defective mutants in the cell-free system and when HepG2 lysate was added to cell-free reactions. The fact that addition of HepG2 cell extract caused the pattern that is characteristic of assembly defective conditions suggests that it is reproducing the same process of inhibition that occurs in cultured cells.

There are many possibilities that could explain why the inhibition that we observed is only partial. First, there might be a balance between proassembly factors and inhibitory factors present in both extracts (wheat germ and HepG2, respectively), and by combining the two, an intermediate assembly phenotype is seen. Alternatively, an inhibitory factor in HepG2 cells may be required in a stoichiometric amount or may be saturable and, in these experiments, might not be present at sufficient quantities to achieve complete inhibition. Regardless of the mechanism, the partial inhibition seen here is significant and can be used as a readout for identification of negative regulators of capsid assembly. One possible negative regulator that others have found is lipid droplets, which appear to sequester wild-type core in mammalian cells (22, 45, 52). It is possible that cellular factors govern whether HCV core assembles, gets sequestered in lipid droplets, or gets targeted for degradation. If so, then the cell-free assembly system and mammalian cell extract complementation experiments described here could be used to identify cellular factors that are either positive or negative regulators of assembly. Indeed, precedents exist for using cell-free systems to identify critical cellular factors important for virion formation, including Hsp90, which is critical for activation of the HBV polymerase (23, 25), and HP68, a cellular factor that is important for assembly of HIV-1 (66).

Cell-free systems also allow sensitive detection of morphological as well as biochemical differences seen during assembly of diverse viral capsids. For example, HBV capsids produced in a cell-free system are relatively homogeneous (Fig. 6) (36), but when the identical extracts are programmed with HCV core transcript, capsids of heterogenous sizes are produced (Fig. 3 and 6), thereby reproducing morphological differences seen in vivo for HBV and HCV capsids. Pulse-chase analyses in cell-free systems show that newly translated HCV core assembles extremely quickly (Fig. 7C), in contrast to both HIV-1 and HBV capsid assembly in cell-free systems, in which a prolonged posttranslational phase is needed to complete capsid assembly (35, 36). Furthermore, the assembly of HCV capsids appears to be membrane-independent, in contrast to cell-free HIV-1 capsid assembly, which shows a strict dependence on both membrane-targeting domains and the presence of intact membrane vesicles (35). This underscores the ability of cell-free systems to elucidate many morphological and biochemical differences intrinsic to different viruses when their capsid proteins are used to program the system.