HPV-E7 protein containing a short C-terminal GGC linker was recombinantly expressed in Escherichia coli, solubilized from the inclusion body fraction by 8M urea and purified using affinity and size-exclusion chromatography. The denatured protein migrated as a 14-kDa single band in reducing SDS-PAGE (data not shown). After refolding by dilution and dialysis against NaCl-MES-containing buffer, E7 protein spontaneously formed oligomers. Qβ VLP production and purification have been described in detail elsewhere (21).

CpGs 1668 with phosphorothioate backbone were purchased from Invivogen (sequence: 5′ tccatgacgttcctgatgct 3′). The protein and peptides E7_49–57 were produced in a modified version with additional 3 aa (GGC) added to the C terminus E7 (Proimmune, UK) to allow coupling to VLPs. E7 protein sequence UniProt database: P03129. E7_49–57 peptide sequence: RAHYNIVTFGGC.

Anti-VLP and anti-E7 antibody titers were measured in the serum of mice vaccinated 21 days earlier with unmodified VLP. A total of 96-well plates were coated overnight with 5 μg/mL of unmodified VLP or 5 μg/mL of E7 oligomers. After blocking for 2 h with 2% bovine serum albumin phosphate-buffered saline (PBS), serum obtained from vaccinated or control mice (diluted 1:500 to 1:12,500) was added and plates were incubated for 2 h at room temperature. After washing the plates three times with PBS-0.05% Tween, horseradish peroxidase-labeled goat anti-mouse immunoglobulin G (IgG-HRP) (Jackson ImmunoResearch, UK) was added for 1 h, followed by the addition of 3,3′,5,5′-tetramethylbenzidine (TMB) Sigma, as a substrate before reading the optical density (OD) at 450 nm (OD450). Titers are expressed as serum dilutions at the half-maximal OD (OD50).

Plates (Nunc-Immuno MaxiSorp) were coated with anti-E7 antibody, each of the vaccine preparations were added at a concentration corresponding to 60 ng/mL of E7 protein. Following washing and incubations, anti-Qβ monoclonal antibody was added followed by secondary detection antibody goat anti-mouse IgG-HRP and TMB. Data expressed in OD450.

Vaccines were prepared by chemical coupling as described elsewhere (2). Purified Qβ VLPs (2 mg/mL in PBS) were derivatized by a 1-h incubation at room temperature with a 10-fold molar excess of succinimidyl-6-(β-maleimidopropionamido)hexanoate (Pierce, Rockford, IL, USA). Free cross-linker was removed by diafiltration with Amicon Ultra Centrifugal Filters, 100 kDa MWCO. Derivatized Qβ VLPs and E7_49–57-GGC (peptide at fivefold molar excess) or E7 protein were then incubated for 3 h at room temperature to allow cross-linking. Unbound material was removed by diafiltration with Amicon Ultra centrifugal filters, 100 kDa MWCO for the peptide and size exclusion for the E7 protein. Efficiency of cross-linking was analyzed by SDS-PAGE.

Performed as described elsewhere (22). Briefly, bacterial RNA trapped into VLPs during recombinant expression was digested with RNAse A (Merck) for 5 h at 37°C. RNAse-treated VLPs were combined with 120 nM/mL of CpG and incubated for further 3 h at 37°C. Excess of CpG was removed by dialysis against PBS.

C57BL/6 mice (9–12 weeks old; Harlan) were injected s.c. with 80 μg of either vaccine formulation. Blood was collected from the tail vein and serum and PBMCs separated for further analysis.

E7 protein or E7_49–57 was labeled with AlexaFluor 488 C5-maleamide as described by the manufacturer and with Qβ labeled with AlexaFluor 647 or PE as instructed by the manufacturer (all from Thermo Fischer). C57BL/6 mice (9–12 weeks old; Harlan) were injected s.c. with 30 μg of VLP and peptide into the hind leg.

All mouse experiments have been performed in accordance with the local welfare legislation and under valid animal experimentation licenses.

Popliteal and inguinal LNs were isolated and a single-cell suspension was prepared by incubating LN with 1 mg/mL collagenase D (Roche) and 0.04 mg/mL DNase I (Boehringer) in 5% FSC containing DMEM, for 30 min at 37°C. Organ pieces were passed through a 70-μm cell strainer and stained for cell-specific markers. The following fluorochrome-labeled antibodies were used: CD11c, CD8a (all from eBioscience), F4/80, Live/Dead Aqua cell stain (all from Life Technologies).

Spleens were isolated smashed through a 70-μm cell strainer using the plunger of a syringe (Falcon), red blood cells were lysed with ACK solution (Lonza). The following fluorochrome-labeled antibodies were used: CD3, CD8a, Live-Dead dye, IFNγ, TNFα, and IL-17 (eBioscience).

Primary culture of DCs from LNs of a naïve mice were harvested and incubated for 24 h with 10 nM of E7_49-51 conjugated with Alexa488 and 1 μg/mL of Qβ conjugated with Alexa647. CD11c⁺ F4/80⁻ DCs were subsequently analyzed by flow cytometry for uptake of peptide and Qβ.

Female C57BL/6 mice at age of 10–11 weeks were injected with 1.5 × 10⁵ TC-1 cells expressing HPV16 E7 oncoprotein. Eight days post injection of tumor cells, mice developed palpable tumors and were subjected to a weekly immunization schedule with E7 coupled or mixed to Qβ(1668) for 49 days. Tumor size was recorded daily and was calculated as follows: W*W*L/2, where W is tumor width (centimeters) and L is tumor length (centimeters). Mice showing signs of suffering or reaching tumor size larger than 1,000 cm³ were euthanized.

Size-exclusion chromatography analysis of refolded E7 showed a wide distribution elution profile corresponding to molecular weights between 17 kDa and 10 MDa (void volume of the column, Figure 1A, upper panel). The refolded E7 was treated under reducing conditions, which narrowed the molecular weight distribution to a major peak with average size of 670 kDa (Figure 1A, bottom panel) and demonstrated the role of disulfide bonds in E7 multimerization, as described elsewhere (20). The 14 kDa capsid protein of the Qβ bacteriophage is expressed in E. coli and self-assembles to form a VLP with molecular weight of approximately 3 MDa. Electron microscopy and dynamic light scattering (not shown) demonstrates that Qβ VLPs are single particles with average diameter of 30 nm. The size-exclusion profile demonstrated a narrow size distribution of the VLP on a S500 column (GE) (Figure 1B). Thus, Qβ VLPs and non-reduced E7 oligomers exhibit roughly comparable sizes of 30–50 nm.

In order to compare the immunogenicity of mixed and covalently linked antigen and VLP, the E7 protein containing a short cysteine-containing linker (-GGC) was chemically cross-linked via the free cysteine to surface lysine on Qβ using the hetero-bi-directional cross-linker, SMPH. The efficiency of the cross-linking was assessed by separation on SDS-PAGE under reducing conditions and analyzed by Western blot (Figure 2). Alternatively, purified Qβ and E7 protein were simply mixed. The model of mixed and coupled formulations is represented in Figure 1C. To assess whether mixing would result in spontaneous association of Qβ to E7 oligomers, a sandwich ELISA assay was performed using anti-Qβ VLP capture antibodies and anti-E7 detection antibodies. The assay demonstrated that E7 was only associated with Qβ VLPs upon chemical coupling, while simple mixing did not result in any physical association (Figure 1D). To ensure that loading of VLPs with CpGs did not alter the binding properties of VLPs, the ELISA was repeated with Qβ VLPs loaded with B-type CpG 1668 (Qβ (1668)). Similarly, no association between E7 and Qβ VLPs could be observed (Figure 1D). VLPs loaded with CpG are represented as Qβ(1668), while “Qβ” will be used to indicate unmodified VLPs (naturally loaded with E. coli-derived ssRNA).

Drainage of Qβ and E7 was assessed by labeling particles with Phycoerythrine (PE) and Alexa488, respectively. Flow cytometry analysis demonstrated that a subpopulation of resident DCs of draining LNs simultaneously took up both E7 and Qβ VLPs (Figure 3A). A total of 23% of cDCs (CD11c^highF4/80⁻) from the popliteal LN were double positives for Qβ and the antigen E7-Alexa488. Thus, antigen and adjuvants, i.e., E7 and Qβ VLPs, were taken up simultaneously in vivo by the same DCs.

To confirm the hypothesis that antigen and VLP size was the limiting factor on antigen and VLP distribution, the E7 derived peptide H2-D^b E7_49–57 was used as antigen instead of the E7 protein. Injection of free peptide E7_49–57 labeled likewise with Alexa488 and Qβ VLPs labeled with Alexa647 did not result in simultaneous uptake by DCs in draining LNs (Figure 3B). As a matter of fact, the peptide could not be found in LNs 24 h after injection, indicating that small peptides do not drain efficiently to LNs and are not efficiently transported by DCs.

To exclude potential methodological limitations in visualizing peptide interaction with DCs by flow cytometry, an in vitro pulsing experiment was performed. Primary cultures of DCs from LNs of a naïve mice were incubated for 24 h with the peptide E7_49-51 conjugated to Alexa488 and Qβ conjugated to Alexa647. CD11c⁺ F4/80⁻ DCs were subsequently analyzed by flow cytometry for uptake of peptide and Qβ. As shown in Figure 3C, almost 50% of the DCs simultaneously interacted with the peptide and Qβ, which confirmed the capacity of small peptides to interact with DCs in vitro and suggested that subcutaneously injected small peptides do not efficiently traffic to the LN.

The contribution of the vaccine components for generation of CD4⁺ and CD8⁺ T cell responses was investigated by injecting mice with E7 protein either alone or mixed with CpG and Qβ(1668) either coupled or mixed with E7. The production of IFNγ by CD4⁺ and CD8⁺ T cells was measured 8 days after vaccination. E7 alone or mixed with free CpG induced low percentage of E7-specific T cells. E7 protein mixed or coupled with Qβ(1668) efficiently induced T cells producing IFNγ demonstrating the superior T cell immunogenicity of a vaccine containing VLPs-packaged CpGs (Figures 4A,B). More importantly, levels of CD4⁺ and CD8⁺ T cell responses induced by E7 protein either mixed or coupled to Qβ(1668) were similar, suggesting that chemical association of the antigen and adjuvant is dispensable for priming (Figures 4A,B) T cell responses.

After establishing that Qβ(1668) was necessary to induce IFNγ production by CD8⁺ T cells in response to E7 and that DCs simultaneously take up antigen and VLP if simply mixed, it was investigated whether such formulation was sufficient for induction of protective T cell responses or whether covalent linkage of E7 oligomers to Qβ(1668) VLPs was necessary. First, the kinetics and duration of the response were measured. E7 was either chemically coupled to or mixed with Qβ(1668) and C57BL/6 mice were immunized s.c. in a prime-boost scheme 70 days apart. T cell responses were followed in the blood by intracellular cytokine staining for IFNγ. Both experimental groups mounted strong primary CD4⁺ and CD8⁺ T cell responses that had declined by day 70 and were strongly increased after the boost injection (Figures 4C,D, respectively). CD4⁺ and CD8⁺ T cells from the blood were analyzed by intracellular cytokine staining upon in vitro stimulation on day 78.

To further investigate the cytokine production of splenic T cells, spleens were collected on day 78. Both vaccination regimens induced strong CD8⁺ and CD4⁺ T cell responses expressing multiple cytokines (Figure 5) and induction of multifunctional CD4⁺ T cells producing IFNγ, TNFα, and IL-17 (Figure 5A,B) were observed. For CD8⁺ T cells, a strong induction of TNFα and IFNγ-producing T cells was detected (Figure 5D,E). The number of tetramer-positive cells was roughly similar to the frequency of IFNγ-producing CD8⁺ T cells (Figure 5E). Thus, as observed in the blood, linked E7 protein induced similar frequencies of cytokine-producing T cells compared to the mixed formulation in splenic T cells (Figure 5).

In marked contrast to the results obtained with particulate E7 proteins, for small peptides, linkage to the VLP was essential. Specifically, mice immunized with the peptide E7_49–57 coupled to Qβ(1668) generated strong CD8⁺ T cell responses. In contrast, E7_49–57 mixed to Qβ(1668) failed to induce good production of IFNγ and TNFα by CD8⁺ T cells significantly above background (Figure 6).

In order to assess the therapeutic capacity of the vaccine-induced T cells, mice were injected with TC-1 cells expressing HPV16 E7 oncoprotein. Eight days post injection of tumor cells, mice developed palpable tumors and were subjected to a weekly immunization schedule with E7 coupled or mixed to Qβ(1668) for 49 days (Figure 7A). In untreated mice, as well as mice injected with Qβ(1668) only, tumors grew unrestrained and 50% of mice reached severity scores requiring euthanasia by day 32 (Figure 7B). In contrast, both coupled and mixed formulation of Qβ(1668) and E7 protein were able to induce strong protection against tumor growth resulting in survival rate of more than 80% until the end of the study. In an additional study, vaccination with E7 mixed with Qβ(1668) extended the survival of tumor bearing mice to more than 3 months (data not shown). Therefore, covalent linkage of the E7 to Qβ(1668) VLPs is not required for induction of protective T cell responses against solid tumor.

In order to understand whether the same rules would apply to B cells, the importance of covalent linkage of E7 to Qβ VLPs for induction of optimal antibody responses was assessed. Serum collected from mice immunized with E7 coupled or mixed with Qβ(1668) were analyzed for E7-specific antibody response by ELISA. Importantly, only covalent coupling but not simple mixing of E7 and Qβ (1668) was able to enhance the antibody response against E7 (Figure 8A). Conversely, the antibody levels raised against Qβ were higher in the mixed formulation when compared to the coupled counterpart (Figure 8B). Antibody induction against the E7_49–57 peptide also required covalent linkage, as in the absence of linkage, antibody levels induced were comparable to the background (Figure 8C). Hence, in contrast to T cell responses, B cell responses appear to require physical linkage of antigen and adjuvants in order to induce appropriate levels of antibody responses irrespective of the size of the antigen.