Papillomavirus-Like Particles Induce Acute Activation of Dendritic Cells¹

To control invading microorganisms, the immune system has evolved to recognize viral pathogens and to mount protective defenses against them. Although so-called pattern recognition receptors (PRRs)⁴ for the recognition of invariant molecular structures in bacterial products have been well characterized, the role of pattern recognition of virion structural elements, especially of nonenveloped viruses, in the initiation of adaptive immune responses has not been extensively studied yet (1, 2). In this study, we have assessed the role of naked icosahedron virus structural characteristics on dendritic cell (DC) functions, given that these cells play a central role in the initiation and modulation of the antiviral host defense and that they form a bridge between innate and adaptive immunity (3, 4).

DCs are a family of bone marrow-derived, professional APCs with a unique capacity to initiate and modulate cell-mediated immune responses (4, 5, 6). Immature DCs located in peripheral tissues (e.g., epidermal Langerhans cells) function as sentinels of the immune system in that they capture and process Ags (7, 8, 9). A wide range of stimuli, including infectious virus, bacterial Ags, or inflammatory cytokines, can induce maturation of DCs, which is associated with up-regulation of costimulatory molecules as well as those encoded by the MHC (10, 11, 12, 13). Following the acquisition of Ag, DCs migrate to regional lymph nodes, where the presentation of MHC-Ag complexes together with costimulatory molecules leads to T cell activation (14, 15). Thus, DCs function as outposts of immune surveillance in that they trigger primary immune reactions against infectious pathogens including viruses (13).

Most studies on the role of DCs in viral immunity have focused on infectious virus models, which cannot distinguish between the effects of direct interaction with virion structure and those induced by de novo production of viral gene products. It has been shown previously that DC maturation can be induced by viral products such as dsRNA (16), or following interaction between glycosylated viral envelope proteins and lectin receptors expressed on DCs (17). However, it remained unclear whether structural features of nonenveloped virion surfaces can deliver DC-activating signals that are sufficient for the induction of an immune response.

To examine the effects of viral structural components on DC activation, we used papillomavirus-like particles (VLPs) as a model system. Papillomavirus VLPs are nonenveloped icosahedral particles, composed of the L1 major capsid protein, which form via self-assembly in the cell nucleus after high level expression of L1. Papillomavirus VLPs were chosen, because they resemble capsids from infectious virions morphologically and antigenically, while lacking encapsidated viral genes (18, 19, 20, 21).

In several animal models, systemic vaccination with papillomavirus VLPs, engineered and expressed in insect cells, can induce type-specific, high-titer neutralizing Abs against conformationally dependent L1 epitopes, and confer Ab-dependent protection against experimental infection with wild-type virus (20, 22, 23, 24). Based on these characteristics, papillomavirus VLPs are attractive candidates for prophylactic subunit vaccines, with the long-term goal of inducing protection against human papillomavirus (HPV)-induced diseases, including cervical cancer (25, 26). HPV VLPs are currently being studied in early phase human clinical trials, which have confirmed that systemic vaccination even without adjuvant can induce high titers of neutralizing Abs (27, 70). In addition, chimeric VLPs, in which peptides derived from an HPV oncoprotein are fused to a viral capsid protein, can induce specific CTL and protect mice from tumor formation by a syngeneic cell line (28, 29, 30). VLPs can also serve as immunogenic vehicles for fused nonpapillomavirus Ags. For example, CTL can be induced by murine P815 tumor-associated Ag P1A (31), while a strong Ab response to a central self Ag, CCR5, occurred when a cell surface-exposed CCR5 peptide was presented as part of the regular array of assembled VLPs (32).

Parenteral immunization with papillomavirus VLP-based vaccines can induce potent humoral and cell-mediated immune responses even without adjuvant. However, the mechanism(s) that accounts for the high intrinsic immunogenicity of papillomavirus VLPs has not been determined. Therefore, we have begun to address this issue in this study by using immature bone marrow-derived DCs (BMDCs), to assess early events occurring after encounter with noninfectious papillomavirus VLPs and to investigate their ability to initiate primary T cell responses. As controls, we investigated the DC-activating effects of disordered papillomavirus capsomers and of VLPs from human polyomaviruses (BK and JC), which, like papillomavirus VLPs, resemble naked icosahedral virions. We found that only the fully assembled papillomavirus VLPs were capable of inducing acute phenotypic and functional maturation of DCs, resulting in the induction of Th1-dominated, primary immune responses.

The following VLP preparations were generated using a baculovirus expression system established in our laboratory (20): 1) papillomavirus VLPs composed of L1 major capsid proteins, HPV16(K)-L1 VLPs, the assembly-deficient mutant HPV16(P)-L1, and bovine papillomavirus (BPV)-L1 VLPs; 2) papillomavirus VLPs containing L1 and L2 minor capsid protein, HPV18-L1/L2 and BPV-L1/L2 VLPs; and 3) human polyomavirus VLPs composed of VP1 major capsid proteins, JCV-VP1 VLPs and BKV-VP1 VLPs (Ref. 33 and S. A. Frye, P. N. Jensen, M. Gravell, and E. O. Major. Production of the major structural protein of the human polyomavirus JCV and BKV: self-assembly into virus-like particles and use in serological studies; manuscript in preparation). Briefly, recombinant virus was propagated in Sf9 insect cells (ATCC, CRL 1711; American Type Culture Collection, Manassas, VA) cultured in Graces’ insect medium supplemented with 10% inactivated FBS, 0.5% calcium chloride, 2.8% potassium chloride, 3.33% lactalbumin hydrolysate, 3.33% yeastolate, and 0.1% pluronic F-68 (all Life Technologies, Grand Island, NY). Sf9 cells were adjusted to 3 × 10⁶ cells/ml and infected at a multiplicity of infection of ∼10 with recombinant baculovirus. After 72 h at 26.6°C, cells were harvested and washed once in ice-cold PBS. The final pellet was either snap frozen in liquid nitrogen and stored at −70°C or processed immediately. For VLP preparation, the pelleted cells were lysed by short-pulse sonification on ice (Fisher sonic dismembrator). The lysates were loaded on top 40% (w/v) sucrose (ICN Biomedicals, Aurora, OH)/PBS cushions and centrifuged at 25,000 rpm for 2.5 h in a SW-28 rotor (Beckman, Fullerton, CA). The resulting pellet was resuspended in 2 ml of 27% (w/w) cesium chloride (CsCl; Life Technologies)/PBS and sonicated a second time. Cell lysates were subjected to two subsequent centrifugations to equilibrium in 27% (w/w) CsCl (20 h/28,000 rpm/10°C). Fractions were harvested and analyzed by SDS/PAGE and Coomassie blue staining or Western blotting using the cross-reactive anti-L1 mAb Camvir-1 (PharMingen, San Diego, CA) for detection of denatured L1. Immunoreactive fractions were dialyzed against 0.5 M NaCl in PBS and quantified by Coomassie blue staining of 10% Tris-HCl gels (Bio-Rad, Hercules, CA) by comparison with a BSA (Life Technologies) standard. To investigate possible effects of contaminating Sf9 proteins in the VLP preparations, a negative control preparation was generated by pelleting uninfected Sf9 cells and processing as for VLP purification.

To determine the efficiency of VLP assembly, transmission electron microscopy was performed by absorbing particles onto carbon-coated grids, negative staining with 1% uranyl acetate, and examination with a Philips electron microscope (model EM 400T) at a ×36,000 magnification.

C57BL/6 (H2^b) inbred mice, ages 6–8 wk, were obtained from the National Cancer Institute Frederick Cancer Research Facility Animal Production Area (Frederick, MD) and housed under pathogen-free conditions. Mice were euthanized by CO₂ asphyxiation for organ harvest.

BMDCs were obtained according to standard protocols with slight modifications (34). Briefly, hind extremities of C57BL/6 mice were collected, soft tissues removed, and bones rinsed in 70% ethanol. After cutting the ends of femurs and tibias, bone marrow was flushed out with RPMI 1640 (Life Technologies) and collected through a nylon mesh. Red cells were lysed with ACK lysing buffer (BioWhittaker, Walkersville, MD). DCs and precursors were enriched by negative selection of lineage marker-positive cells using magnetic beads (anti-CD4 (L3T4), anti-CD8a (Ly-2), anti-CD19, and anti-NK cell (DX5) MicroBeads; Miltenyi Biotec, Auburn, CA), according to the manufacturer’s protocol. The remaining cells were plated in 24-well plates at 1 × 10⁶ cells/ml in RPMI 1640 supplemented with 10% heat-inactivated FBS, 2 mM l-glutamine, 1 mM sodium pyruvate, 0.1 mM nonessential amino acids, 5 × 10⁻⁵ M 2-ME, 25 mM HEPES, 50 μg/ml gentamicin, and 1× antibiotic/antimycotic solution (all from Life Technologies) (normal culture medium, NCM). Added to the cultures were murine rGM-CSF (1000 U/ml) and murine rIL-4 (1000 U/ml; both PharMingen). After 3 days at 37°C, cultures were fed by aspirating 75% of medium and adding back fresh NCM supplemented with GM-CSF and IL-4. At day 6, cells were harvested and preparations contained 67–88% DCs, as assessed by CD11c expression.

To investigate DC maturation, day 6 BMDCs were replated at 1 × 10⁶ cells/ml in 24-well plates and exposed to various concentrations of VLPs (0.1–10 μg/ml). Control experiments were performed adding LPS or VLPs directly to the developing DCs without replating the cells. The results of these experiments were very similar to those obtained by replating the cells at day 6. By FACS analysis, the background staining of all the markers tested was lower on the cells that had not been replated as compared with background expressions on replated DCs. The phenotypic maturation of the DCs by LPS and VLPs was more pronounced if added directly to the cultures. Up-regulation of costimulatory molecules was observed at 24 h and did not increase after 48 h of exposure to LPS or VLPs. As a positive control, DCs were stimulated with LPS (Escherichia coli 026:b6; Sigma, St. Louis, MO) at 1 μg/ml. As a negative control, an uninfected Sf9 insect cell preparation, generated as described above, was used. At various time points, cells and cell supernatants were obtained and analyzed. For T cell assays, day 6 BMDCs were pulsed by adding HPV16(K)-L1 VLPs at 10 μg/ml. After overnight culture, excess VLPs were removed by thoroughly washing the cells.

To detect cell-associated VLPs, day 6 BMDCs were adhered to acid-washed number 01 cover slips in 24-well plates and cultured overnight. VLPs were added to the cells at a concentration of ∼0.7 μg VLPs/10⁵ cells. Dilutions were prepared in cold PBS, and the VLPs were incubated with the cells for 15 min on ice. Unbound VLPs were removed by washing four times with cold PBS. Subsequently, the cells were either fixed directly to evaluate cell surface binding of VLPs, or returned to 37°C in NCM supplemented with GM-CSF and IL-4 to allow for VLP internalization. Cells were then fixed at selected time points by a 15-min incubation in 1% paraformaldehyde diluted in PBS. The method of immunofluorescent staining has been previously described (35). Briefly, after fixation, the cells were washed three times with 200 mM glycine/PBS. They were then incubated with primary Ab diluted in 0.1% Brij 58/PBS and incubated at 4°C. Abs used were 5B6 for detection of BPV-L1 (36, 37), and 5.12.2 and 3.1.1 to detect JC and JC/BK VLPs, respectively (Novocastra Laboratories, Newcastle, U.K.). The secondary Ab, FITC-conjugated goat anti-mouse IgG (Jackson ImmunoResearch, West Grove, PA), was diluted to 5 μg/ml in 0.1% Brij 58/PBS. To distinguish VLP uptake by the BMDCs from that by other phagocytic cells, double stainings using a PE-labeled anti-CD11c Ab (PharMingen) were performed. Fluorescence was examined with a Bio-Rad MRC 1024 laser-scanning confocal system attached to a Zeiss Axioplan microscope. Images were acquired using the photon counting mode. Images were collaged and subjected to identical scale adjustment with Adobe Photoshop (Adobe Systems, Mountain View, CA) software.

Cells were collected into cold Hank’s BSS without Phenol Red (Life Technologies) plus 0.1% BSA (Life Technologies) and 0.1% sodium azide (Aldrich, Milwaukee, WI). To avoid nonspecific binding of labeled Ab, FcBlock (purified rat anti-mouse CD16/CD32; PharMingen) was added at 1 μg for 10⁶ cells. Single or double stainings were performed using mAbs against the following mouse surface Ags (all PharMingen): CD3 (145-2C11, hamster IgG), CD4 (GK1.5, rat IgG2b), CD8 (53-6.7, rat IgG2a), CD11c (HL3, hamster IgG), CD40 (3/23, rat IgG2a), CD54 (3E2, hamster IgG), CD80 (16-10A1, hamster IgG), CD86 (GL1, rat IgG2a), H-2Kb (AF6-88.5, mouse IgG2a), I-Ab (Aβb) (AF6-120.1, mouse IgG2a), and appropriate control Abs. A total of 10⁵ cells per assay was incubated with the respective Abs at a concentration of 2.5 μg/sample for 30–45 min at 4°C. After two rounds of washings, cellular fluorescence was monitored in a FACSCalibur and analyzed using CellQuest software (both Becton Dickinson, Mansfield, MA). Propidium iodide (Aldrich) was added to gate on viable cells.

Cytokine levels in conditioned cell supernatants were assayed by ELISAs for murine IL-4, IL-6, IL-10, IL-12p70, TNF-α, and IFN-γ (all R&D Systems, Minneapolis, MN), according to the manufacturer’s instructions. Adsorbance was read at 450 nm with an automated microplate ELISA reader. Cytokine levels were quantified from two to three titrations using standard curves, and results were expressed in pg/ml. The lower detection limits for the assays were as follows: IL-4, 7.8 pg/ml; IL-6, 15.6 pg/ml; IL-10, 15.6 pg/ml; IL-12p70, 7.8 pg/ml; TNF-α, 23.4 pg/ml; and IFN-γ, 9.4 pg/ml.

T cells were obtained from spleen cells of C57BL/6 mice by positive selection using magnetic beads (anti-CD4 (L3T4) and anti-CD8a (Ly-2) MicroBeads; Miltenyi Biotec) with a purity of >95%. T cells were placed into 96-well round-bottom plates at 2 × 10⁵ cells/well in NCM and stimulated with either VLP-pulsed or untreated syngeneic DCs at a ratio of 10:1. After 7 days in culture, T cells were harvested, replated at 2 × 10⁵ cells/well in NCM supplemented with murine IL-2 (20 U/ml; PharMingen), and restimulated with DCs as before. Three days after the second stimulation, T cell proliferation was assessed using the CellTiter96 Aqueous assay (Promega, Madison, WI). Cell proliferation was detected by 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (Owen’s reagent) bioreduction into formazan that can be determined by absorbance at 490 nm using a microplate reader. To correct for the presence of BMDCs in the cocultures, the absorbance obtained from DC cultures was subtracted (corrected absorbance 490 nm). In addition, culture supernatants were collected for cytokine detection (IFN-γ, IL-4, and IL-10).

Papillomavirus VLPs have been shown to bind to a wide range of cell types, including APCs from different sources (Refs. 38 and 39 ; W. M. Kast, unpublished data). However, the ability of BMDCs to bind and internalize VLPs has not been evaluated. To address this issue, we exposed BPV-L1/L2 VLPs to immature mouse BMDCs and used confocal microscopy to monitor the association of VLPs with the cell surface and their internalization (Fig. 1). We detected high levels of VLP binding to the BMDCs, with a pattern similar to that of control mouse fibroblasts, which are known to be infectable by BPV (40) (Fig. 1 and data not shown). However, internalization of bound VLPs by BMDCs was significantly more rapid than by the fibroblasts (data not shown). Particles were internalized by the immature BMDCs as early as 5 min after incubation at 37°C (Fig. 1). Within 15 min, high levels of VLPs were localized predominately in the dendrites and in cytoplasmic vesicles located adjacent to the cell membrane. After 30 min, uptake had reached saturation, and no further reduction of surface-bound VLPs was observed at later time points, suggesting that the uptake capacity of the cells was exhausted. Over the next 90 min, particles coalesced into dense vesicles within the cells and relocalized to the perinuclear region. Internalization of the VLPs by DCs was confirmed by costaining for CD11c (data not shown). Thus, immature DCs effectively bind and rapidly internalize papillomavirus VLPs.

When DCs use internalized proteins to stimulate T cell responses, the DCs must undergo maturation and migrate from the periphery to regional lymph nodes. To determine whether VLPs induce phenotypic DC maturation, BMDCs were exposed to serial dilutions of HPV16(K)-L1 VLPs, and expression of cell surface markers was determined by flow cytometry. We found a dose-dependent up-regulation of MHC class I and II molecules as well as costimulatory molecules (CD40, CD80, CD86) and the CD54 adhesion molecule (Fig. 2). At VLP concentrations of 10 μg/ml, we observed maturation of DCs equal to that after exposure to bacterial LPS (Table I). Phenotypic maturation occurred also after exposure to VLPs at a concentration of 1 μg/ml (∼3 × 10⁴ VLPs/DC) (Fig. 2), while no significant response was detectable at 0.1 μg/ml (data not shown). The kinetics of the induction of cell surface markers was similar for LPS and VLPs, in that it required 24 h for the activated phenotype to be present, and the activated state remained for 48 h with no further up-regulation of any of the markers tested (Fig. 3).

Phenotypic maturation of BMDCs was also observed after exposure to other papillomavirus VLPs (HPV18-L1/L2 and BPV-L1, Fig. 4), indicating that DC activation is a general characteristic of papillomavirus VLPs. To rule out the possibility that components other than the VLPs in the preparations might have induced the observed effects, a negative control sample, prepared from uninfected Sf9 cells, was added to BMDCs, but it failed to induce markers of activation (data not shown). Thus, papillomavirus VLPs induce acute phenotypic maturation of BMDCs.

To determine whether the assembly state of L1 influenced the induction of BMDC maturation, we compared the stimulatory capacity of fully assembled HPV16(K)-L1 VLPs with that of an assembly-deficient mutant HPV16(P)-L1. This L1 mutant, from an HPV16 genome isolated from cervical cancer, is defective in proper VLP assembly due to a single amino acid substitution at residue 202 (21). The wild-type and mutant L1s were purified concurrently using the same procedure. We confirmed by electron microscopy that HPV16(K)-L1 had efficiently assembled into complete spherical VLP structures of ∼55 nm in diameter with a regular array of capsomers (Fig. 5). In contrast, preparations of the mutant HPV16(P)-L1 contained predominately (>90%) unassembled pentamers or irregular aggregates, with properly assembled VLPs detected only rarely (Fig. 5). When added to immature BMDCs, the HPV16(P)-L1 preparation was at least 10-fold less effective in inducing phenotypic DC maturation than fully assembled HPV16(K)-L1 VLPs. Although 10 μg/ml of wild-type virus induced the up-regulation of a wide panel of activation markers, the mutant at this concentration induced only one of them (CD80) (Fig. 6). No up-regulation was seen with the mutant at 1 μg/ml (data not shown), although at this concentration the wild type still induced phenotypic maturation (Fig. 2). Therefore, in the interaction of HPV16 capsid proteins with DCs, well-ordered particles are superior in their capacity to activate DCs.

To investigate the possibility that the DC-stimulatory capacity of papillomavirus VLPs is a common characteristic of naked icosahedral VLPs, we tested VLPs of two human polyomaviruses, BK and JC virus, for their effect on BMDC maturation. These VLPs were chosen because polyomaviruses resemble papillomaviruses morphologically in that they all share an icosahedral structure composed of 72 pentamers of the major capsid protein arranged in T = 7 symmetry, although polyomavirus capsids are somewhat smaller than those of papillomaviruses (Fig. 5). By confocal microscopy, using a panel of monoclonal and polyclonal Abs, we found that immature BMDCs effectively bound and internalized BKV-VP1 and JCV-VP1 VLPs (data not shown). However, both polyomavirus VLPs failed to induce up-regulation of MHC and costimulatory molecules, even at concentrations of VLPs that were 100 times higher than that sufficient for phenotypic activation by papillomavirus VLPs (Fig. 6 and data not shown). Therefore, VLPs with very similar structures can differ markedly in their capacity to activate BMDCs.

Given that HPV16 is the papillomavirus type most commonly associated with cervical cancer (26) and that HPV16-L1 VLPs are currently being evaluated in clinical trials, we focused on VLPs from this HPV type for subsequent experiments.

It is well established that DCs are a significant source of a wide range of cytokines that are secreted in response to various stimuli, e.g., viruses or bacteria and their products, such as LPS (41, 42, 43). To investigate whether VLPs are capable of inducing cytokine production in DCs, levels of IL-6, IL-12p70, and TNF-α were determined in supernatants of BMDCs cultured in the presence or absence of different concentrations of HPV16(K)-L1 VLPs. Bacterial LPS (1 μg/ml) was used as the positive control. Although BMDCs secreted low to undetectable baseline levels of the above cytokines, HPV16(K)-L1 VLPs induced the secretion of IL-6 and TNF-α in a dose-dependent manner (Fig. 7). However, this treatment did not result in appreciable IL-12p70 production. When the kinetics of cytokine production was analyzed, secretion of IL-6 and TNF-α in response to HPV16 VLPs occurred more slowly than that induced by LPS, which rapidly induced inflammatory cytokines as early as 6 h after stimulation, with maximum levels reached at 18 h (Fig. 7, A and B). However, at 48 h, the VLPs (10 μg/ml) induced similar levels of TNF-α and higher levels of IL-6 compared with the levels induced by LPS at this time point. In parallel experiments, BKV-VP1 and JCV-VP1 VLPs did not induce DCs to secrete proinflammatory cytokines (IL-6, 31.6 ± 10 pg/ml; TNF-α, 23.4 pg/ml (not above background)).

The inability of papillomavirus VLPs to induce IL-12 secretion directly was not unexpected. Several previous studies have reported evidence that IL-12 induction by infectious agents requires additional stimulation that can be provided by CD40 ligation or by IFN-γ (44, 45, 46). Therefore, we cocultured BMDCs with syngeneic T cells in the presence or absence of VLPs. In this setting, HPV16(K)-L1 VLPs effectively induced IL-12p70 secretion by BMDCs. Moreover, rIFN-γ was able to substitute for T cells in inducing significant levels of IL-12 (Fig. 7 C). Control experiments revealed that HPV16(K)-L1 VLPs had no direct effect on T cell phenotype (the expression of CD25 and CD154, as measures of T cell activation, was unchanged over background) or IFN-γ production (data not shown). These results suggest that exposing BMDCs to VLPs can activate T cells, which in turn provide the signals, including IFN-γ, sufficient for IL-12 secretion by the BMDCs.

Because DCs are the only cell type capable of inducing primary T cell responses, we investigated the effect of VLP-pulsed BMDCs on activation of syngeneic, naive T cells. To exclude direct effects of VLPs on T cells, BMDCs were pulsed overnight with VLPs to allow processing and presentation of VLP proteins (39, 47) and were then added to the T cells. BMDCs pulsed overnight with HPV16(K)-L1 VLPs induced significant T cell proliferation measured at day 3 after the second stimulation (Fig. 8). In addition, culture supernatants obtained from T cells conditioned with VLP-pulsed DCs contained high levels of IFN-γ (Fig. 8), but lacked detectable levels of the Th2 cytokines IL-4 or IL-10 (data not shown). Thus, the VLP-pulsed BMDCs induce Th1-dominated immune responses in naive, syngeneic T cells.

In this study, purified papillomavirus VLPs (HPV16, HPV18, BPV) induced prompt and efficient phenotypic and functional maturation of BMDCs. The VLPs provided three different types of immunoregulator signals for the DCs: 1) those that control the induction of Ag-specific T cell activation (up-regulation of MHC class I and class II molecules, CD80, CD86, CD40, CD54); 2) those that mediate an inflammatory response (secretion of IL-6 and TNF-α); and 3) those that induce effector functions (production of IL-12p70). Papillomavirus VLPs composed of L1 alone or L1 and minor capsid protein L2 induced maturation of BMDCs (Fig. 4), correlating with in vivo data demonstrating that vaccination with L1 or L1/L2 VLPs resulted in protective immune responses (18, 22, 48). A threshold concentration of 10 μg/ml of VLPs was required for optimal phenotypic and functional activation of 10⁶ DC, although some response was seen at 1 μg/ml. Given that VLP concentrations applied during vaccination were usually 1–100 μg VLPs per injection (18) and that as few as 100 DCs are sufficient to induce an in vivo immune response in mice (49), our results are likely to reflect the physiologic situation after VLP vaccination. Therefore, these data provide an experimental explanation for the pronounced immunogenicity of papillomavirus VLP-based vaccines even in the absence of adjuvant.

Several observations support the conclusion that the DC activation seen in our assays resulted from specific recognition of VLP structural elements and not from copurified contaminants in the VLP preparations. First, neither similarly processed lysates of uninfected Sf9 insect cells nor similar preparations of polyomavirus VLPs, also derived from baculovirus-infected Sf9 cells, induced DC activation. Second, the assembly-deficient mutant L1 of the reference strain of HPV16 (21) inefficiently activated DCs. The low level residual activity of mutant L1 observed in our experiments may result from the low number of properly assembled VLPs in the preparation or from suboptimal activation of the DCs by disorganized aggregates of L1 capsomers.

Recently, so-called PRRs of the innate immune system have been identified that mediate the immune recognition of microbial Ags by means of certain conserved structural features (2). DCs express several types of PRRs, including the mannose receptor DEC 205 and Toll-like receptors (3, 50), suggesting that they may be important components of the innate immune system before initiation of Ag-specific immune responses. However, pattern recognition of virion structural elements by DCs is poorly understood. To our knowledge, there is only one previous report that examined the effect of noninfectious viral subunit vaccines on DC functions (51). In that study, two different influenza virus subunit vaccine preparations (one conventional viral subunit vaccine composed of viral hemagglutinin and neuraminidase proteins alone, and one virosomal subunit vaccine, in which influenza hemagglutinins were incorporated into the surface of liposomes) were found to have no effect on the phenotype or secretory functions of human DCs. Because an inactivated whole virus influenza vaccine preparation used in that same study induced DCs to up-regulate MHC and costimulatory molecules and to secrete IL-12p70 and TNF-α, the authors suggested that these positive effects on DCs might have resulted from the presence of viral RNA sequences analogous to bacterial DNA motifs containing unmethylated CpG dinucleotides (52). In contrast to inactivated enveloped viruses used in earlier studies (51, 53), our VLP preparations do not contain the viral genome, other viral proteins, or cellular membrane proteins of the heterologous cells used to produce the virions. Therefore, the ability of papillomavirus VLPs to activate immune functions in DCs is most likely related to the ordered repetitive array of virion surface proteins and the interaction of these proteins with yet to be identified DC cell surface receptors, presumably leading to receptor oligomerization and subsequent downstream signaling.

To characterize receptor-ligand interactions early after encounter of immature DCs with papillomavirus VLPs, we performed binding and internalization studies using confocal microscopy. The abundant association of VLPs to the cell surface of our highly enriched murine BMDC preparations and the rapid internalization observed support the notion that BMDCs possess specific VLP receptors. These findings are consistent with those of other investigators, who have shown, by FACS analysis of bulk cell populations, that APCs from mouse spleen or human PBMCs, as well as human Langerhans cells, bound HPV16 VLPs (W. M. Kast, unpublished data). FcγRIII (CD16) was proposed as a candidate receptor for papillomaviruses in that study. Therefore, the observed association of papillomavirus VLPs with the DC surface might result, at least partially, from the binding to Fc receptors, which are known to be highly expressed on DCs (54, 55, 56, 57). Other possible receptors include cell surface proteoglycans (58) and α₆ integrin (59). The latter is less likely to play a role in VLP-DC interaction, as monocyte-derived DCs (60) and BMDCs do not express this integrin (our own unpublished observation).

Secretion of cytokines such as IL-6, IL-12, or TNF-α is a critical component of immune regulation of B and T lymphocytes during encounter with an infectious agent (61, 62, 63). As the interaction of DCs with certain viruses (e.g., HIV) can induce mRNA for a wide range of cytokines, but without cytokine production, cytokine transcription and translation in DCs can in some instances be differentially regulated (42). The effect of noninfectious subunit vaccines on the cytokine secretory activity of DCs is less well understood and seems to be critically dependent on the vaccine preparation studied (51). We found that HPV16 VLPs were potent stimulators of DC-derived IL-6 and TNF-α, which were detectable at the protein level as secreted proteins. Compared with LPS, the overall cytokine response to VLPs was as pronounced, yet relatively delayed. By contrast, the kinetics of phenotypic activation of DCs after exposure to either stimulus was similar. If the time course of maturation relative to cytokine release in VLP-activated DCs is similar in vivo, these results suggest that most of the cytokines would be released after the DCs reached the regional lymph nodes, which appears to occur within ∼6–48 h after Ag exposure (64, 65). This may account, at least partially, for the observation that immunization with HPV L1 VLPs did not induce substantial inflammation at the side of injection in early phase clinical trials (27 , 70). This observation might also, in part, explain the characteristic persistence of localized productive papillomavirus infection in the absence of inflammatory responses. The rapid cytokine secretion induced by LPS is consistent with the strong local inflammatory reaction after injection of LPS in vivo. It will be interesting to determine whether viral and bacterial inducers of DC activation generally exhibit differences in the kinetics of cytokine release. An attractive hypothesis is that the acute response to bacterial products has evolved to produce immediate inflammation to focus the immune response to sites of infections, which tend to remain localized. In contrast, infections by most viruses (with papillomaviruses a notable exception) rapidly become disseminated, and an unrestricted inflammatory response would likely be counterproductive.

In previous studies, IL-12 production by DCs was tightly regulated and required additional signals from T cells (44, 45, 46). Similarly, we found that HPV16 VLPs failed to induce IL-12p70 directly, but required help from syngeneic T cells. Given that papillomavirus VLPs had no direct effect on IFN-γ production or on phenotypic activation of T cells (data not shown), these findings indicate that VLP-exposed DCs activate T cells, which in turn provide the signal(s), including IFN γ, necessary for the induction of IL-12p70 by the DCs. Consistent with this hypothesis, we found that rIFN-γ can substitute for T cells in stimulating VLP-exposed DCs to secrete IL-12p70 and that VLP-pulsed DCs stimulated T cells to secrete IFN-γ (Fig. 8).

Our data strongly suggest that DCs activated by papillomavirus VLP may be highly effective inducers of Ag-specific cellular immunity. We found that immature BMDCs pulsed with HPV16 VLPs induced proliferation of syngeneic T cells, which also produced IFN-γ (but not IL-4 or IL-10), thus initiating a strongly polarized Th1 immune response. Using an intracellular cytokine assay, we have preliminary data indicating that DCs armed with papillomavirus VLPs generate both CD4⁺ and CD8⁺ T cell responses (data not shown). These findings point to a central role of DCs in the initiation and modulation of T cell responses to papillomavirus VLPs and offer mechanistic explanations for the strong adjuvant-independent antitumor immune response to chimeric papillomavirus VLPs in mouse tumor models (28, 29, 30, 31). Because DCs are critically involved in B cell differentiation and isotype switching (66), VLP-activated DCs may also facilitate the pronounced Ab response against conformational L1 epitopes observed after VLP vaccination in the absence of adjuvant.

There are several possible explanations for the striking difference in the response of DCs to the structurally similar papillomavirus and polyomavirus VLPs. Immune responses have evolved under considerable selective pressure imposed by pathogens, which may have resulted in the recognition of naked icosahedral virion structures through yet to be identified receptors. During natural infection with papillomaviruses, virions are shed to the exterior by the terminally differentiated epithelial cells and therefore are not readily exposed to the systemic immune system in large numbers. Given that DC activation appears to require interaction with multiple virus capsids (>1000 VLPs per cell in our in vitro assays), there may not be a strong selective pressure against the recognition of papillomaviruses by DCs. In contrast, some lytic viruses that induce persistent systemic infections as part of their normal life cycles (e.g., BK and JC viruses) may have counterevolved to escape acute pattern recognition by DCs. Other viruses, e.g., herpes simplex, measles, or vaccinia virus, are known to inhibit DC maturation (67, 68, 69), but this inhibition is thought to require DC infection. It is interesting that DCs retain the ability to bind and internalize BK and JC VLPs, suggesting that the absence of activation in the case of polyomaviruses may be due to differences in the specific receptors bound by the two groups of viruses. Alternatively, intracellular events may account for the differences in DC activation upon internalization of viral Ags.

Collectively, our results point to a central role of DCs in acute immune recognition of papillomavirus virions. Although papillomavirus structural elements were necessary and sufficient to activated DCs, it is clear that gross structural features of a virus capsid do not allow predictions about its effect on DC functions or provide insights into which VLPs will most likely serve as potent vaccine vehicles. These results contribute significantly to a better understanding of the mechanisms involved in the immune response to virion capsids and may provide knowledge for the rational design and development of future vaccines.