A Cell-Free Assembly System for Generating Infectious Human Papillomavirus 16 Capsids Implicates a Size Discrimination Mechanism for Preferential Viral Genome Packaging

ABSTRACT

We have established a cell-free in vitro system to study human papillomavirus type 16 (HPV16) assembly, a poorly understood process. L1/L2 capsomers, obtained from the disassembly of virus-like particles (VLPs), were incubated with nuclear extracts to provide access to the range of cellular proteins that would be available during assembly within the host cell. Incorporation of a reporter plasmid “pseudogenome” was dependent on the presence of both nuclear extract and ATP. Unexpectedly, L1/L2 VLPs that were not disassembled prior to incubation with a reassembly mixture containing nuclear extract also encapsidated a reporter plasmid. As with HPV pseudoviruses (PsV) generated intracellularly, infection by cell-free particles assembled in vitro required the presence of L2 and was susceptible to the same biochemical inhibitors, implying the cell-free assembled particles use the infectious pathway previously described for HPV16 produced in cell culture. Using biochemical and electron microscopy analyses, we observed that, in the presence of nuclear extract, intact VLPs partially disassemble, providing a mechanistic explanation to how the exogenous plasmid was packaged by these particles. Further, we provide evidence that capsids containing an <8-kb pseudogenome are resistant to the disassembly/reassembly reaction. Our results suggest a novel size discrimination mechanism for papillomavirus genome packaging in which particles undergo iterative rounds of disassembly/reassembly, seemingly sampling DNA until a suitably sized DNA is encountered, resulting in the formation of a stable virion structure.

IMPORTANCE Little is known about papillomavirus assembly biology due to the difficulties in propagating virus in vitro. The cell-free assembly method established in this paper reveals a new mechanism for viral genome packaging and will provide a tractable system for further dissecting papillomavirus assembly. The knowledge gained will increase our understanding of virus-host interactions, help to identify new targets for antiviral therapy, and allow for the development of new gene delivery systems based on in vitro-generated papillomavirus vectors.

INTRODUCTION

Papillomaviruses (PVs) are a group of nonenveloped viruses with small circular double-stranded DNA genomes that infect the skin and mucosa. PV replication is tightly coupled to the differentiation of the epithelium (1). Human papillomavirus 16 (HPV16) is medically the most important HPV type, as it is responsible for about 50% of cervical cancers and for the vast majority of noncervical HPV-positive anogenital and oropharyngeal cancers (2). The expression of viral early genes occurs in the lower layers of the epithelium, while vegetative replication of the viral genome, late gene expression, and virus assembly occur in more differentiated suprabasal cells (1). The strict dependency of viral replication on epithelial differentiation has made it difficult to produce virus in reasonable amounts and quality in cultured cells, although organotypic epithelial cell “raft” culture systems have been useful for the study of some aspects of PV biology. An alternative method for producing infectious PV capsids is the PV pseudovirus (PsV) production system (3, 4). This system is based on the cotransfection of 293TT cells with a plasmid encoding the two PV capsid proteins and a reporter plasmid. Simian virus 40 (SV40) T-antigen-driven overreplication of the reporter plasmid mimics the high copy number of viral genomes present during the productive stage of the PV life cycle in vivo. The high copy number drives capsid assembly around the reporter DNA plasmid that substitutes as a pseudogenome containing a reporter gene, such as that encoding green fluorescent protein (GFP), which can be easily quantified. The PsV system allows the generation of large amounts of homogenous particles whose capsids are indistinguishable from authentic PV biochemically and by electron microscopy (EM) (4, 5). Surprisingly, the encapsidation of the pseudogenome is largely sequence independent, in that several distinct plasmids, entirely lacking in PV sequences, can be efficiently packaged (4). This observation raises the question of how pseudogenomes and the authentic viral genome are selectively packaged in the presence of an excess of cellular DNA.

With authentic virus, the 60-nm PV capsid protects a chromatinized circular double-stranded DNA (dsDNA) genome of ∼8 kb. The viral capsid is composed of the major capsid protein, L1, and the minor capsid protein, L2. The icosahedral (T=7) PV capsid is formed by 72 L1 pentamers (6–8) linked by the C-terminal arm of L1, which invades the neighboring pentamers and folds back to the original pentamer (9). L2 is located predominantly inside the capsid, within the central internal cavity of the pentamers, and is present at up to 72 molecules per virion (5, 10). Most of the interactions between the two viral structural proteins are hydrophobic (11, 12). The PV capsid undergoes a maturation process that results in the formation of disulfide bonds between L1 pentamers that stabilize the viral capsid (5, 13).

L1 alone can self-assemble into VLPs that morphologically resemble virions produced during natural infection. However, L2 may assist in capsid formation and stabilization (14–17). L2 also has been shown to play important roles in DNA packaging for some PV types (18, 19), and it is required for the efficient infection of the target cell (18, 20–22).

PV assembly occurs in the nucleus (23), and L2 nuclear localization is required for its incorporation into capsids (24). Studies indicate that L2 is synthesized and localized in the nucleus of suprabasal cells prior to L1 for alpha genus HPVs (25). Since L1 is transported into the nucleus as a pentamer (25–28), it has been suggested that L1 assembles into pentamers in the cytoplasm and then is transported to the nucleus. Taken together, these observations suggest a mechanism for PV assembly where viral DNA is first replicated, followed by nuclear import of L2 and L1 pentamers. This sequence allows all viral components to be present in the nucleus, leading to efficient assembly. Karyopherins and Hsp70 family members have been implicated in the nuclear transport of both L1 and L2 (26–31). Karyopherins also have been implicated in the inhibition of premature capsid assembly from L1 pentamers (26). This mechanism of regulating capsid assembly, while facilitating nuclear transport, may inhibit the premature assembly of HPV in the cytoplasm, ensuring that it will occur only when all viral components are present in the nucleus. Once assembled, the naturally occurring redox gradient in terminally differentiated squamous epidermal tissue would allow capsid maturation (32).

Despite these advances, PV assembly remains a poorly understood process. In this study, we report the development of a cell-free in vitro papillomavirus assembly system, using HPV16 as the model virus. This reductionist system allowed the assembly of HPV16 capsids from capsomers around a reporter plasmid in a nuclear extract milieu. Exposure to nuclear extract also induced the destabilization of assembled intact particles that then could reassemble around a provided reporter plasmid. Based on the data presented here, we propose a mechanism for HPV16 assembly where the capsids are preassembled and package the correct DNA by an iterative process that is sequence independent and size restricted.

MATERIALS AND METHODS

Cell lines, virus, and plasmids.

293TT (Christopher B. Buck, NCI), 293H (Invitrogen), HeLa (ATCC), and HaCaT (Norbert E. Fusenig, German Cancer Research Center) cells were grown in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (DMEM-10).

HPV16 VLPs and PsV were produced and purified on OptiPrep medium as previously described (33). HPV16 L1-only and L1/L2 VLPs were produced using p16L1only and p16sheLL plasmid, respectively. Both plasmids are >8 kb and cannot be packaged efficiently. Green fluorescent protein (GFP)- or red fluorescent protein (RFP)-expressing HPV16 PsV (GFP-PsV or RFP-PsV) were produced using p16sheLL and pfwB (GFP) or pRwB (RFP) as reporter plasmids.

HPV16 C428S (Cap/L2) was produced as described for HPV16 VLPs and PsV by cotransfecting 293TT cells using Lipofectamine 2000 with pumL1B and p16L2h. Cap/L2 was purified by ultracentrifugation through a 46/30/20/15% OptiPrep step gradient. Ultracentrifugation was performed on an NVT65 rotor for 1 h at 65,000 rpm.

All plasmids used for virus production and reassembly experiments were produced in competent Escherichia coli DH5α (Bioline) and purified using Qiagen Midiprep kits. Plasmids are described at http://home.ccr.cancer.gov/LCO/plasmids.asp.

Antibodies and inhibitors.

Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and histone H2B antibodies were obtained from Cell Signaling Technologies (no. 2118 and no. 12364, respectively). CAMVIR-1 antibody against HPV16 L1 was from Santa Cruz (sc-47699).

Decanoyl-RVKR-chloromethilketone (dec-RVKR-cmk) and compound XXI {(S,S)-2-[2-(3,5-difluorophenyl)-acetylamino]-N-(1-methyl-2-oxo-5-phenyl-2,3-dihydro-1H-benzo[e][1,4]diazepin-3-yl)-propionamide} were from Calbiochem (no. 344930 and no. 565790, respectively). NH₄Cl and apyrase from potatoes were purchased from Sigma (A0171 and A7646, respectively).

VLP and PsV disassembly.

Disassembly was based on the protocol described by Mukherjee and colleagues (34). HPV16 VLPs or PsV were incubated for 3 h at 37°C in buffer containing 50 mM NaCl, 20 mM Tris, pH 8.2, 2 mM dithiothreitol (DTT), and 0.01% Tween 80.

Nuclear extract preparation.

Nuclear extract was prepared according to a protocol adapted from Abmayr and colleagues (35). Cells were displaced from the culture flask by vigorous pipetting (293TT and 293H cells) or by scraping (HaCaT cells). Cells were spun at 1,850 × g for 10 min, resuspended in 5× pcv (packed cell volume) hypotonic buffer (10 mM HEPES, pH 7.9, 1.5 mM MgCl₂, 10 mM KCl, 0.5 M DTT, protease inhibitor cocktail [cOmplete Mini, EDTA-free; from Roche]), and pelleted again at 1,850 × g for 10 min. Cells then were resuspended in 3× pcv hypotonic buffer and incubated on ice for 10 min. Cells were homogenized in a glass Dounce homogenizer with pestle B until 80 to 90% of the cells were lysed. Nuclei were pelleted at 3,300 × g for 15 min. Supernatant (cytosolic fraction) was stored for further analysis, and pellet (nuclear fraction) was washed twice with phosphate-buffered saline (PBS). Nuclei were suspended in PBS at a density of 2 × 10⁸ cells (or nuclei)/ml, which was the number of starting cells used for this calculation. All procedures were done on ice, and buffers and material were ice cold. The nuclei were subjected to freeze/thaw cycles to liberate their soluble contents prior to use in the assembly reaction.

Reassembly.

For reassembly, 1 μg (based on L1 amounts quantified by Coomassie staining) of intact or dissembled VLPs or PsV was mixed with 3 × 10⁶ nuclei (15 μl), 150 ng GFP-reporter plasmid (pfwB), 150 mM NaCl, 100 mM Tris, pH 7.2, 0.02% Tween 80, 10 mM CaCl₂ in a total volume of 200 μl. Samples were incubated for 20 h at 37°C. Samples then were treated for 6 h at 37°C with nucleases. For this treatment, 200 μl of buffer containing 0.2% benzonase (E1014; Sigma), 0.2% BAL-31 (M0213; New England BioLabs), and 20 mM MgCl₂ was added to the mixture.

To study the effect of ATP on reassembly, 5 U of apyrase was added to the reassembly mix.

For kinetic analyses, reassembly was performed as described above; however, at the indicated time points, samples were treated for only 10 min with nucleases and directly frozen at −80°C. All samples were frozen for at least 12 h before being used in an infectivity assay.

Infection.

HeLa cells (4,500/well) were seeded on 96-well plates 24 h prior to infection. Cells were infected with 20 μl or, in the case of the competition assay, 10 μl of the reassembly mixture in a total volume of 200 μl in DMEM-10. This resulted in 20 to 40% infection for nondisassembled L1/L2 VLPs with nuclear extract. GFP expression, as a measure of infection, was scored 72 h postinfection (p.i.) by flow cytometry (FACSCanto II; BD). Data analysis was performed with FlowJo v10 software (TreeStar).

For infection inhibition studies, cells were preincubated for 30 min with 10 μM dec-RVKR-cmk, 300 nM compound XXI, or 20 mM NH₄Cl prior to infection. Inhibitors were present during the 72-h infection.

When control infection was performed with PsV, reassembly mix containing inhibitors but no viral proteins (mock) was added to the cell culture immediately prior to the addition of purified PsV.

Trypsin treatment.

Reassembled samples were incubated for 10 min at 37°C with 0.125% Trypsin. Samples were diluted 1:40 with 5× SDS sample buffer (10% SDS, 50% glycerol, 0.025% bromophenol blue, 250 mM Tris-HCl, pH 6.8) and boiled for 10 min at 95°C. Samples were analyzed by Western blotting with CAMVIR-1 antibody against HPV16 L1 protein.

OptiPrep gradient.

Reassembled particles were centrifuged for 3.5 h at 50,000 rpm at 16°C in an SW55Ti rotor (Beckman Coulter) on a 27/33/39% OptiPrep gradient. Samples were collected from the bottom (fraction 1) to top (fraction 10). Fraction 1 was 750 μl, fractions 2 to 8 were 250 μl each, and fractions 9 and 10 were 1 ml each. Ten microliters of each fraction was analyzed by SDS-PAGE and Western blotting with CAMVIR-1. For quantification, densitometric analysis using ImageJ 1.47v was performed.

Electron microscopy.

Samples were adsorbed to carbon-coated copper grids and negatively stained with either 1% uranyl acetate for 15 s for analysis of disassembly (Fig. 1B) or with 1% phosphotungstic acid for 4 min and 1% uranyl acetate for 1 s for analysis of nuclear extract-mediated reassembly (see Fig. 4D). The double staining gave less background and sharper staining when using reassembly buffer. Samples were examined with an FEI Tecnai T12 transmission electron microscope.

FIG 1.

Cell-free assembly. (A) Depiction of the experimental procedure. (B) HPV16 L1L2 VLPs were incubated for 3 h at 37°C in disassembly buffer (pH 8.2, 50 mM NaCl, 2 mM DTT). Samples taken before or after disassembly were analyzed by electron microscopy. Scale bars represent 100 nm. (C) Nuclear and cytoplasmic fractions were analyzed by Western blotting for the presence of GAPDH (cytoplasmic marker) or histone 2B (H2B [nuclear marker]). The same numbers of cell equivalents were loaded in each lane.

FIG 4.

Intact particles disassemble in the presence of nuclear extract. (A) Disassembled or intact HPV16 L1/L2 VLPs or HPV16 C428S+L2 (Cap/L2) were incubated with reassembly mix in the presence or absence of nuclear extract. After 20 h in reassembly mix, samples were treated for 10 min with 0.125% trypsin at 37°C. Samples were analyzed by SDS-PAGE and Western blotting using the CAMVIR-1 antibody. S represents the protein standards; from top to bottom, the bands correspond to 60, 50, 40, and 30 kDa. (B) Particles reassembled as described for panel A were run on a 27/33/39% OptiPrep gradient. Fractions were collected from bottom (fraction 1) to top (fraction 10) and analyzed by SDS-PAGE and Western blotting with the CAMVIR-1 antibody. The arrow indicates full-length L1 proteins. (C) Densitometric quantification of fractions 4 and 5 (intact particles) and fractions 9 and 10 (capsomers/disassembled particles) from panel B. The percentage of L1 found in the indicated fractions is represented. (D) Samples reassembled as described for panel A were analyzed by electron microscopy. Scale bars represent 100 nm.

RESULTS

Characterization of assembly reaction components.

To better understand the molecular mechanism of papillomavirus virion assembly, we aimed to establish and characterize a cell-free assembly system. Since various host proteins might serve to chaperone assembly reactions, we aimed to establish an in vitro method that would more closely resemble the process that occurs in the host cell nucleus by initiating the assembly with pentamers in the presence of nuclear proteins. We used HPV16 as a model PV type for designing the protocol outlined in Fig. 1.

Capsomers were prepared by disassembling HPV16 L1 or L1/L2 VLPs at pH 8.2 in 50 mM NaCl and 2 mM DTT for 3 h at 37°C, as previously described (34) (Fig. 1A). As monitored by electron microscopy, this treatment disassembled the HPV16 capsids into capsomers and disordered aggregates of capsomers (Fig. 1B, compare images 1 and 3 to images 2 and 4).

In the next step, we generated a crude nuclear extract (here referred to as nuclear extract) (Fig. 1A) whose efficiency in the extraction process was evaluated by Western blot measurement of the relative content of marker proteins in the cytosolic and nuclear fractions corresponding to the same number of cells. The nuclear extract contained minimal amounts of cytosolic contaminants, as determined by the low reactivity with the antibody recognizing the cytoplasmic marker GAPDH, while, as expected, it was enriched for the nuclear marker histone H2B (Fig. 1C). When the extracts were normalized by protein amount, the purity of the nuclear fraction was even more pronounced (data not shown). The nuclear extract was freeze-thawed before use in the reassembly reaction mix in order to release nuclear components.

To trigger the reassembly of the viral particles under conditions similar to those of a cellular environment, we mixed the capsomers with the nuclear extract under isotonic conditions (pH 7.2, 150 mM NaCl, 37°C) and supplemented this mixture with an ∼6-kb GFP reporter plasmid to act as a marker for the efficiency of viral DNA packaging and infection (Fig. 1A). The reassembly mixture, composed of nuclear extract, plasmid, and viral proteins (Fig. 1A, reassembly mix), then was incubated for 20 h at 37°C to allow ample time for the assembly process to occur. After this step, the mixture was nuclease treated for 6 h to degrade unencapsidated DNA (3, 4, 36). To assess successful PsV assembly, HeLa cells were infected with the resultant reaction products, and PsV infection was measured by flow cytometry at 72 h p.i.

Reassembly experiments also were performed in a reassembly mix in the absence of nuclear extract. Additionally, to ensure that GFP expression upon infection was attributable to the delivery of the packaged reporter plasmid and not to capsomer-mediated delivery, we examined an HPV16 L1 mutant, C428S, that is only capable of assembling into loosely associated pentamers, as measured by sucrose gradient centrifugation and electron microscopy (37). This mutant was prepared in the presence of L2 to generate L2-associated capsomers (Cap/L2). Cap/L2 was used in the reassembly mixture instead of the disassembled HPV16. Since these mutant capsomers cannot assemble into stable pseudovirions, their use tests the possibility that the capsomers per se are able to transduce the GFP reporter plasmid in the described system.

Cell-free in vitro-assembled viruses are infectious.

Initially, we used nuclear extracts prepared from 293TT cells (Fig. 2A), since this has been the main cell line used for PsV production. We examined the process with both L1-only VLPs and L1/L2 VLPs, since L2 is required for infection by intracellularly generated PsV and may affect DNA packaging and capsid stability (14–16, 18–21). When we disassembled the VLPs and mixed them with the nuclear extract and the reporter plasmid at near physiological conditions, the L1/L2 VLPs, but not the L1-only VLPs, produced infectious particles (Fig. 2A, + nuclear extract). The reassembly of infectious PsV also was dependent on the nuclear extract, since there was no infection without it (Fig. 2A, − nuclear extract). There also was no infection with the mutant Cap/L2 after incubation with reassembly mix (Fig. 2A) in the presence of nuclear extract, indicating that unassembled capsomers cannot transduce the reporter plasmid in this system.

FIG 2.

Infectious HPV16 PsV are formed in the presence of nuclear extract. The indicated type of VLP preparation was incubated for 20 h in reassembly mix with a GFP reporter plasmid with (+) or without (−) nuclear extract from 293TT (A), 293H (B), or HaCaT (C) cells. After treatment with nucleases, HeLa cells were infected with the product of the different reassembly reactions. The number of GFP-transduced HeLa cells was analyzed 72 h p.i. by flow cytometry. The mean values for at least three independent experiments ± standard deviations (SD) normalized for intact L1/L2 particles (highest level of infection) are shown.

We also tested HPV16 VLPs that had not been subjected to the disassembly process (i.e., intact particles) in the reassembly reaction. We expected that this condition would serve as an additional negative control, as we assumed a unidirectional progression from capsomers to capsids. Surprisingly, under these conditions, L1/L2 VLPs produced infectious particles even more efficiently than if the particles had been deliberately disassembled before adding the reassembly mix (Fig. 2A, intact). As with the deliberately disassembled particles, productive encapsidation of the reporter plasmid into intact PsV and infection was dependent on the nuclear extract (Fig. 2A, compare + nuclear extract to − nuclear extract) and did not occur if L1-only VLPs were used.

Since the reporter plasmid contains an SV40 origin of replication (ori) and 293TT cells express the SV40 replicase (large T antigen, or LT), it is conceivable that LT played a role in the packaging/infectivity that we observed. Therefore, we replicated the assembly conditions using nuclear extract from 293H cells, which do not express LT (Fig. 2B). The results obtained were very similar to the ones obtained with nuclear extract from 293TT cells: infection occurred with L1/L2 VLPs but not with L1-only VLPs or capsomers (Fig. 2B). Importantly, the observation that intact particles had higher rates of infectivity also was replicated (Fig. 2B). Similar results also were observed when the reassembly mix used nuclear extracts from HaCaT cells, a human keratinocyte cell line (Fig. 2C). We also compared the infectivity of the cell-free in vitro-assembled virus with PsV obtained from cell culture. For the three nuclear extracts used, intact particles had about 50× to 100× less infectivity than the same L1 amounts of cell-derived PsV (data not shown).

Taken together, these results indicate that cell-free assembly of infectious HPV16 capsids is dependent on one or more nuclear factors. Unexpectedly, infectious yields were better for particles that were intact prior to the reassembly procedure, which suggests that nuclear factors also can trigger the destabilization of assembled capsids to allow packaging of exogenous plasmid DNA.

Cell-free assembled PsV use the pathway established for HPV16 assembled in cultured cells.

To determine if cell-free in vitro-reassembled particles had characteristics similar to those of intracellularly assembled HPV16 PsV, we evaluated if they utilized well-characterized components of the HPV16 entry pathway. HeLa cells were infected with the reassembled particles in the presence of an inhibitor of furin (dec-RVKR-cmk), an inhibitor of γ secretase (compound XXI), or with NH₄Cl. The first two inhibitors inhibit cellular enzymes known to be required for HPV16 PsV infection (38–40), while NH₄Cl is a lysosomotropic inhibitor that neutralizes the pH of endosomes and is known to inhibit HPV16 PsV infection (41, 42). All inhibitors prevented infection of HeLa cells with particles reassembled in nuclear extracts prepared from 293TT, 293H, or HaCaT cells (Fig. 3A, B, and C). When the reassembled particles were incubated with an L1 neutralizing antibody prior to infection of HeLa cells, it also inhibited the infectivity of each type of cell-free in vitro-assembled PsV (Fig. 3A, B, and C). These findings indicate that the observed gene transfer was the result of transduction by infectious HPV16 PsV via the previously established entry pathway.

FIG 3.

Cell-free in vitro-assembled virus uses the previously described HPV16 entry pathway. Disassembled or intact L1/L2 HPV16 VLPs were mixed with a reporter plasmid and nuclear extracts prepared from 293TT (A), 293H (B), or HaCaT (C) cells. HeLa cells were infected with the PsV in the presence or absence (untreated) of 10 μM furin inhibitor (dec-RVKR-cmk), 300 nM gamma secretase inhibitor (compound XXI), or 20 mM NH₄Cl after reassembly and nuclease treatment. For the neutralizing antibody, reassembled viruses were incubated for 1 h at room temperature with the L1 neutralizing antibody before addition to HeLa cells. The number of GFP-positive cells (infected) was analyzed 72 h p.i. by flow cytometry. The mean values for at least three independent experiments ± SD normalized for untreated cells (no inhibitor, no antibody) are shown.

Because the nuclear extracts from the three cell lines produced similar results, all subsequent experiments were performed with 293H cell extracts. These cells have the advantage of being T-antigen free (in contrast to 293TT cells) and yielding a greater quantity of nuclear extract than HaCaT cells, mostly because it is easier to displace the 293H cells from the cell culture flask during nuclear extract preparations (see Materials and Methods).

HPV16 capsids are destabilized by the nuclear extract.

Since infectious PsV were generated using VLP preparations that were intact prior to the procedure, we hypothesized that components in the nuclear extract promoted at least partial disassembly of the particles, which allowed the association and subsequent packaging of the reporter plasmid. For encapsidation to occur, the particles would need to be destabilized by the nuclear extract, or the higher efficiency of the intact VLPs in generating infectious capsids might be attributable to the presence of substantial numbers of putatively preexisting, partially assembled capsids that are more efficient substrates for pseudogenome packaging than are fully disassembled capsomers. To distinguish between these two possibilities, we examined the sensitivity of the viral particles to trypsin after exposure to nuclear extract (Fig. 4A). It has been shown previously that mature capsids are mostly trypsin resistant, while destabilized and immature capsids are trypsin sensitive (4, 43). We incubated disassembled or intact VLPs in a reassembly mix that was supplemented with nuclear extract or left unsupplemented. This mixture then was treated for 10 min with 0.125% trypsin at 37°C. Trypsin resistance was assessed by SDS-PAGE and detection of the HPV16 L1 protein by Western blotting. As a positive control, we treated the Cap/L2 preparation with trypsin, which showed the expected sensitivity to trypsin.

In the absence of trypsin, the disassembled and Cap/L2 preparations displayed, in addition to full-length L1, a prominent lower-molecular-weight band, presumably due to proteolytic cleavage of the capsomers during the purification process (Fig. 4A, lane 5). This second species is absent from the intact VLP preparation not incubated with nuclear extract (Fig. 4A, lane 3), indicating that the intact VLPs had essentially no disassembled character. After trypsin treatment of the disassembled or Cap/L2 preparations, several prominent cleavage products appeared in the presence or absence of nuclear extract (Fig. 4A, lanes 6, 7, and 10). In the absence of nuclear extract, trypsin generated relatively few cleavage products in the intact VLP preparation, whereas when nuclear extract was added to the intact VLPs, there was an obvious increase in the number of L1 cleavage products after trypsin treatment (Fig. 4A, compare lanes 8 and 9), and the digestion pattern was indistinguishable from that of the dissembled particles under the same conditions (Fig. 4A, intact, compare lanes 7 and 9). The results indicate that the L1 capsomers in the intact VLP preparation were in the form of assembled capsids initially and that the nuclear extract destabilized the VLP structure.

To further confirm that the nuclear extract triggers HPV16 capsid instability, we analyzed the buoyancy of particles subjected to reassembly reactions on an OptiPrep gradient (Fig. 4B). Smaller or distorted particles should float more than mature particles (44–46). For example, when Cap/L2 particles were run on the OptiPrep gradient, the L1-containing fractions were dispersed throughout the gradient but were mostly in the upper half of the gradient. The minor presence of some signal in the lower fractions probably is due to aggregation, as was evident in the electron microscopy analysis of this construct (Fig. 4D, image 1). Disassembled VLPs in the presence or absence of nuclear extract exhibited a pattern similar to that of Cap/L2, with L1 mostly migrating to fractions in the upper part of the gradient (Fig. 4B, disassembled). With or without the addition of nuclear extract, most of the disassembled/reassembled particles were found in fractions 5 and 6. For intact VLPs in the absence of nuclear extract, most of L1 was found in the middle-lower fractions, with the greatest concentration in fractions 4 and 5, where we typically find fully assembled particles (Fig. 4B, intact). When nuclear extract was added to the mixture, a shift in the L1 migration through the gradient was observed. At this point, a large portion of L1 was found associated with the upper-middle fractions, similar to that seen with the disassembled particles and Cap/L2 (Fig. 4B, intact). This shift confirms a change in the HPV16 capsid assembly state upon the addition of nuclear extract. Since some of the L1 portion floated to the top of the gradient, where capsomers typically are found, a release of some individual viral capsomers may have occurred in the presence of nuclear extract. To better quantify these shifts, we defined intact particles as those that migrate to the core fractions 4 and 5, as evaluated with intact particles without the addition of nuclear extract, and capsomers/disassembled particles as migrating to the top fractions 9 and 10 (Fig. 4C). We then performed densitometric analysis of the Western blots and quantified the amount of L1 present in fractions 4 plus 5 and fractions 9 plus 10. While for disassembled particles there was no obvious shift in the populations, for intact particles a decrease in the number of intact particles (fractions 4 plus 5) and a clear increase in the number of disassembled particles (fractions 9 plus 10) was evident when nuclear extract was added to the reassembly reaction (Fig. 4C).

Taken together, these results indicate that nuclear extract destabilized the viral capsid.

EM analysis of particles.

To visualize the changes that occurred in the HPV16 capsid under the various treatment conditions, we examined negatively stained particles by electron microscopy (Fig. 4D). The L1/L2 VLPs that had undergone disassembly had a high number of capsomers present, although some loosely assembled and incompletely assembled particles also were evident (Fig. 4D, images 3 and 4). In the presence of nuclear extract, some previously disassembled particles had a tighter, more uniform appearance (Fig. 4D, image 4). Although the untreated VLPs had a uniform morphology in the absence of nuclear extract (Fig. 4D, image 5), the addition of the nuclear extract resulted in the presence of many capsomers (Fig. 4D, image 6, red arrows). This observation confirms that the nuclear extract induced the destabilization of the viral capsid and release of capsomers.

Overall, these results indicate that nuclear extract can destabilize the HPV16 capsid, which allows the stable, nuclease-resistant packaging of an exogenous plasmid into the capsid. The fact that particles can reassemble in the absence of nuclear extract (compare Fig. 1B, image 2, to Fig. 4D, image 3) but are not infectious indicates the nuclear extract has a critical role in DNA packaging during the assembly process.

Disassembled and intact particles have similar productive encapsidation kinetics.

The kinetics of reassembly of L1-only HPV particles previously has been analyzed using dynamic light scattering (34, 47) in the absence of nuclear components. We decided to compare the assembly kinetics of disassembled and intact particles into pseudogenome-containing capsids, using infectivity as the readout of correctly assembled PsV. At different time points after the addition of viral particles to the reassembly reactions, samples were taken and nuclease treated for 10 min and then frozen to prevent further assembly. HeLa cells then were infected with the reassembly reaction products from different time points, and GFP expression was analyzed at 72 h p.i. by flow cytometry.

When nuclear extract was present in the reassembly mixture, the curve of the reassembly kinetics was sigmoidal both for disassembled and intact particles (Fig. 5A). There was a 2-h lag phase, followed by an increase in the production of infectious particles until 20 h, with no further increase in the resultant infection levels after this time (Fig. 5A). The observation of sigmoidal kinetics was consistent with that described for the assembly of other viral capsids (48–54), but the kinetics of HPV16 seem to be remarkably slower, with a half-life of about 6 to 7 h. There was no significant difference seen between disassembled and intact particles, although there was a trend for intact particles to have slightly slower kinetics, suggesting that assembly, rather than disassembly, was the primary rate-limiting step in the reaction. We also attempted to quantify capsid assembly by evaluating their light-scattering characteristics, but this type of analysis was not possible due to signal interference caused by the high concentration of complex material (proteins, nuclear membrane, etc.) in the nuclear extract.

FIG 5.

Disassembled and intact HPV16 L1/L2 particles have similar productive encapsidation kinetics. Disassembled or intact L1/L2 HPV16 VLPs were incubated with reassembly mix with (A) or without (B) nuclear extracts prepared from 293H cells. After the addition of the viral particles to the reassembly mix, samples at different time points were nuclease treated and frozen to stop the reaction. HeLa cells were infected with the product of the reassembly reactions. The number of GFP-positive cells (infected) was analyzed 72 h p.i. by flow cytometry. The sigmoidal fittings of the mean values for at least three independent experiments ± SD are shown. For each individual experiment, infection at a given time point was normalized for the plateau phase for mixes with nuclear extract.

When nuclear extract was omitted from the reassembly mixture, there was less infection, as expected, and the infection curve plateaued at 2 h (Fig. 5B). The low infection levels observed probably was attributable to the loosely assembled particles observed by EM (Fig. 4D). It should be noted that the level of infection with samples without nuclear extract was higher in these experiments than in previous ones. This difference probably is due to the reduction in the duration of nuclease treatment from 6 h to 10 min for the kinetics experiment. Although the extent of nuclease digestion appeared to be comparable, as evaluated by agarose gel electrophoresis (data not shown), we cannot exclude the possibility that when treatment time is reduced, more residual DNA remains.

DNA packaging in cell-free assembly is ATP dependent.

Assembly inherently decreases entropy and therefore seems likely to be an energy-dependent process. To address the requirement for ATP in reassembly, 5 U of apyrase, an ATP diphosphatase, was added during the reassembly process. The apyrase concentration was diluted 20-fold prior to addition to cells, and we verified that this residual amount of apyrase did not directly affect the infectivity of intracellularly assembled PsV (Fig. 6A). However, the inclusion of apyrase during assembly reduced infection by 84% and 43% for disassembled and intact particles, respectively, suggesting that ATP is required for productive encapsidation. The increased sensitivity of disassembled particles to ATP depletion may be due to the disassembled particles being in a mostly dispersed capsomeric form, while intact particles exposed to nuclear extract may more often be in a partially disaggregated form. In the latter case, capsomer reassembly may be less energy dependent.

FIG 6.

Reassembly is ATP dependent. (A) Disassembled or intact L1/L2 HPV16 VLPs were mixed with a reporter plasmid and nuclear extracts prepared from 293H cells in the presence (+) or absence (−) of 5 U of apyrase. HeLa cells were infected with the product of reassembly after nuclease treatment or with HPV16-GFP PsV in the presence of nuclear extract and the same concentration of apyrase in the reassembly mix samples. The number of GFP-positive cells (infected) was analyzed 72 h p.i. by flow cytometry. The mean values for at least three independent experiments ± SD are shown. (B) Disassembled or intact L1/L2 HPV16 VLPs were mixed with a GFP reporter plasmid, and nuclear extracts were prepared from 293H cells (+ nuclear extract) or without nuclear extract (− nuclear extract) in reassembly mix, which was supplemented with 1 mM (+ATP) or left untreated (−ATP). The number of GFP-positive cells (infected) was analyzed 72 h p.i. by flow cytometry. The mean values for at least three independent experiments ± SD are shown and are normalized for the samples with nuclear extract and without the addition of ATP.

Given these results, we decided to determine if the addition of ATP would promote assembly. Therefore, the reassembly mixture was supplemented with 1 mM ATP and analyzed by the subsequent infection of HeLa cells. The addition of ATP during the assembly process did not significantly affect the infectious units generated (Fig. 6B). Moreover, ATP was not sufficient to trigger capsid assembly in the absence of nuclear extract, indicating that the extract contains additional critical components.

Reporter plasmid packaging stabilizes the viral particle.

The results presented above raise the possibility that, in the host nucleus, HPV capsids undergo a series of assembly and partial disassembly reactions while screening to find a packageable, i.e., small episomal, DNA species. Such a process probably would involve transient association with the host cellular DNA. Consistent with this conjecture, we previously reported that cellular DNA of <8 kb in length is found within most capsids that are harvested with the standard protocol for generating PsV (4). This method includes a DNase step that would clip protruding DNA to a “headful” unit length. We hypothesize that packaging of an appropriately small DNA fragment would allow the capsid to close completely, as has been suggested before for HPV33 (55), which would result in a more stable state and be less subject to subsequent disassembly by nuclear factors. The site at which DNA fragments longer than ∼8 kb protrude from the capsid might be a weak spot that could make the particle subject to iterative rounds of disassembly and reassembly. To begin to address this concept experimentally, we designed a system to measure plasmid exchange during the assembly/disassembly processes. We used an HPV16 PsV that contained an RFP reporter plasmid (RFP-PsV) produced by the standard method (3, 4). Therefore, in addition to particles that contain the reporter plasmid, there are many “additional” particles that contain a linear ∼8-kb cellular DNA fragment or lack completely packaged DNA (4). During the reassembly procedure, we added an excess of exogenous GFP reporter plasmid to some samples. Using this strategy, we were able to assess the ability of one plasmid to displace a previously packaged plasmid from the capsid and thereby evaluate the stability of capsids containing small plasmids. We used GFP- and RFP-encoding plasmids with the same plasmid backbone and of a similar size to avoid spurious effects linked to these characteristics.

As shown in Fig. 7, when RFP-PsV were disassembled prior to reassembly in the absence of exogenous GFP plasmid, no RFP or GFP signal could be detected. This was the expected result, as the limited amount of RFP plasmid in the system would be released and diluted during disassembly, making detectable repackaging of the RFP plasmid unlikely. A GFP signal was readily observed when the GFP plasmid was added to the mixture, indicating the reassembly of the disassembled particles around the higher-concentration exogenous plasmid, as observed previously for VLPs (Fig. 2). However, if RFP-PsV were not disassembled prior to reassembly, the RFP infection signal was not affected by the addition of nuclear extract. The RFP infection also was not affected by the presence of exogenous GFP plasmid in the reassembly mixture, although the generation of a substantial number of infectious GFP PsV was observed in the presence of nuclear extract. These data indicate that the empty and/or unstable cellular DNA-containing capsids can be disassembled and package the exogenous GFP-encoding plasmid, but that the capsids that contain the RFP-encoding plasmid cannot. These results are consistent with previous findings of Fligge and colleagues for HPV33 (55), who suggested that DNA packaging stabilizes cell-derived viral capsids. The results are consistent with the idea that capsids achieve a higher level of stability following the encapsidation of a DNA molecule small enough to allow complete closure of the capsid. We speculate that nuclease treatments used in standard intracellular PsV production methods leave protruding nubs of DNA that prevent the capsids from forming an entirely closed structure that is resistant to disassembly.

FIG 7.

Pseudoviruses are stabilized by packaging a reporter plasmid. HPV16 RFP-PsV, either disassembled or intact, were incubated in reassembly mix with or without nuclear extract and with (+GFP) or without (−GFP) a GFP reporter plasmid. After nuclease treatment, HeLa cells were infected and RFP (A)- and GFP (B)-positive cells were analyzed by flow cytometry 72 h p.i. The mean values from at least three independent experiments ± SD are shown. RFP was normalized for the samples without nuclear extract and GFP reporter plasmid present in the reassembly mix. GFP was normalized for samples with nuclear extract and GFP reporter plasmid present in the reassembly mixture. Cells that were simultaneously positive for both RFP and GFP were added to both the RFP and GFP columns.

DISCUSSION

Viruses employ a range of assembly processes whose elucidation can inform their biology, and these processes also have translational implications. This knowledge can lead to the identification of new targets for antivirals (for example, hepatitis C virus [56]) and also can lead to improvements in viral vector production technologies for gene therapy or genetic vaccination.

To efficiently assemble, a virus must have a mechanism by which it preferentially packages its genome. Many viruses employ a viral packaging sequence encoded within their genome for this purpose (57–60). In contrast, HPVs produced as PsV within the mammalian 293TT system have been shown to package a variety of gene expression plasmids entirely lacking in viral sequences whose efficiency is not substantially augmented by the presence of PV sequences in the plasmids (3). This observation has raised the question of how HPVs are able to specifically package their own genomes in the nucleus by an apparently sequence-independent mechanism in the presence of a vast excess of cellular DNA.

The results reported in this study support a model in which HPV undergoes a series of partial disassembly and reassembly reactions around DNA until achieving a stable configuration (Fig. 8). The association with cellular DNA would result in an unstable capsid because that DNA would protrude from the capsid shell. However, a more stable structure would be formed when the virus happens to assemble around a DNA molecule that falls within the size limits for complete encapsidation (i.e., no larger than the natural ∼8-kb papillomavirus genome). During the productive phase of the HPV life cycle, the circular viral episome would be the only abundant DNA that would be of the appropriate size. Therefore, a simple size discrimination mechanism may be the key to the preferential incorporation of viral DNA into PV capsids. The possibility that several rounds of disassembly/reassembly cycles are required until the right DNA is found would explain the slow kinetics we observed for generating infectious capsids (Fig. 5). Based on our in vitro results, ATP and cellular factors likely are required for this disassembly/reassembly cycling within the nucleus. This model for viral genome packaging is supported by the plasmid exchange experiments in which we found that an encapsidated plasmid could not be displaced by the introduction of a second suitable plasmid (Fig. 7). The size discrimination model that we propose for HPV16 genome packaging would represent, to the best of our knowledge, a new mechanism of achieving specificity in viral DNA packaging.

FIG 8.

Model of HPV16 assembly. In the nucleus, HPV16 undergoes a series of disassembly/reassembly cycles as it samples various DNAs for packaging, including cellular DNA. Once an episomal DNA of the appropriate size is identified and packaged, its capsids are more stable and do not undergo additional cycles of disassembly/reassembly. Nuclear factors as well as ATP contribute to this process.

Our cell-free model has identified at least two distinct activities that depend on the nuclear extract: the reassembly of disassembled particles and the partial disassembly/reassembly of intact particles. The model should provide a tractable system to better understand the mechanisms involved in HPV16 assembly, for instance, through a subtractive analysis of the nuclear factors that are required for the process. Further studies also could address whether other PVs have the same mechanism of assembly. For example, the role of L2 in genome packaging and assembly seems to be heterogeneous for different PVs (3, 18, 19, 22, 61). It would also be interesting to test the idea that other morphologically similar double-stranded DNA viruses (e.g., polyomavirus) have a similar mechanism for size-based selective packaging of the viral genome. Since HPV16 DNA is chromatinized in authentic virions, future studies also should address the chromatinization status of the packaged plasmid and its importance for packaging. This could not be addressed in this study due to the small amounts of DNA and virus used.

The ability of HPV PsV to encapsidate expression plasmids lacking viral sequences makes them attractive candidates for gene delivery vectors. An important concern is that current production systems rely on the presence of the oncogenic SV40 LT in the producer cells, and pseudovirions containing cellular DNA fragments cannot be totally excluded from the preparation (33). There have been some attempts to generate HPV that would be more acceptable for therapeutic gene delivery to humans by disassembling and reassembling L1-only VLPs in the presence of the target DNA (62). However, the absence of L2 in these particles severely limits their efficiency of DNA delivery and expression. The utilization of L1/L2 (63, 64) VLPs has been attempted, but that production system required large amounts of viral proteins and DNA and appears to have yielded extremely poor particle-to-infectivity ratios. Therefore, there is a need for the development of a more efficient PV-based gene packaging and delivery protocol that might be designed through a better understanding of the mechanisms of PV assembly. Our assembly method may represent a considerable advance over previous attempts. The in vitro assembly of L2-containing capsids should promote a higher level of DNA delivery, as L2 allows the exodus of the packaged DNA from the endocytic system and accompanies it to the nucleus, where efficient transcription can occur. Additionally, the inclusion of nuclear extract in the assembly cocktail should greatly increase the efficiency of producing infectious PsV in vitro.

Funding Statement

This work was supported by intramural funding from the National Cancer Institute, National Institutes of Health.