EPIclip: A Novel Approach for the Production of Decorated Virus-Like Particles Mediated by High-Affinity Protein Binding Partners

TL;DR

A new platform called EPIclip is developed to decorate virus-like particles with proteins using high-affinity binding partners, showing potential for treating pruritic diseases.

Contribution

The EPIclip platform enables decoration of virus-like particles with heterologous proteins using ColE7 and Im7 in a single prokaryotic host.

Findings

Decorated VLPs induce a durable IgG response against IL-31 lasting at least six months.

IL-31-displaying VLPs suppress IL-31-induced pruritus in mice, confirming target neutralization.

VLPs induce a T-cell response against the capsid but not the cytokine, indicating a B-cell-biased immune response.

Abstract

Background: Virus-like particles (VLPs) represent key tools for the development of vaccines due to their ability to induce a potent immune response to epitopes presented on their surface. However, the decoration of VLPs with a complete heterologous protein on the surface remains a bottleneck for clinical translation due to the complexity of manufacture. We present a novel platform, EPIclip™, for the decoration of VLPs mediated by high-affinity protein binding partners, colicin E7 (ColE7) and immunity protein 7 (Im7), within a single prokaryotic host. We evaluate this approach using a modified hepatitis B core capsid protein and IL-31 as a model epitope. IL-31 is a prominent therapeutic target for the development of pruritic diseases. Methods: We explore the design and development of the platform, including the use of T-cell-stimulating peptides. We demonstrate several small-scale purification methods for the candidate VLP, as well as morphological analysis by transmission electron microscopy (TEM). Further, we vaccinate mice with IL-31-displaying VLPs to evaluate immunogenicity and the ability to prevent IL-31-induced pruritus in vivo. Results: Our results demonstrate that decorated VLPs dosed in mice elicit an IgG response against IL-31 with at least six months of durability. In addition, IL-31-displaying VLPs suppress the development of IL-31-induced pruritus, confirming in vivo target neutralisation. Notably, IL-31-displaying VLPs induce a strong T-cell response against the VLP capsid but not against the cytokine, confirming a B-cell-biased immune response and the absence of detrimental autoreactive T cells. We further demonstrate the translation of this system with an additional virus capsid: tomato aspermy virus (TAV). Conclusions: Taken together, the novel EPIclip™ platform may represent a promising therapeutic approach for pruritic diseases. Additionally, this modular system could be adapted for a wide range of research as well as human and veterinary therapeutic applications.

Figures

Full text

1. Introduction

Virus-like particles (VLPs) are large biomolecules resulting from the self-assembly of the proteins encoding the capsids, cores, or envelopes of viruses. Importantly, while VLPs resemble the structure of the viruses they are derived from, they are devoid of the genetic material required for replication or infection [1]. VLPs are highly immunogenic and can induce high titres of neutralising antibodies due to the quasi-crystalline repetitive surface structure, which triggers crosslinking of the B-cell receptor, thereby activating B cells [2,3]. Their small size (20–200 nm) allows VLPs to efficiently enter the lymphatic system and gain access to lymphoid follicles [4]. As a result, they are promising platforms for the presentation of pharmaceutically relevant antigens to the immune system to induce an immune response against specific physiological or infectious diseases. In recent years, VLPs have been used for the development of both prophylactic and therapeutic vaccines for a wide range of diseases and disease states (e.g., COVID-19, malaria, inflammation, allergy, pain, neurodegenerative diseases, and cancer) [1,5,6,7,8,9,10,11].

Historically, many VLP-based vaccines approved for clinical use have been manufactured by self-assembly of individual protein subunits (e.g., hepatitis B surface (HB-S) antigen and human papillomavirus (HPV)) [12,13,14]. By contrast, VLPs decorated with epitopes for the induction of an immune response present a challenge for manufacture due to the requirement of chemical or genetic fusion of epitope and scaffold proteins. Frequently, decorated VLPs require complicated production processes where two different biologic products—the scaffold and the epitope—are produced separately and assembled into the final decorated VLP. To address this limitation, a variety of epitope-decorating platforms have been developed. The most widely known of these is the SpyCatcher/SpyTag system [15]. This system splits the CnaB2 domain from the Streptococcus pyogenes fibronectin binding protein (FbaB) to produce protein coupling partners: SpyCatcher and SpyTag. CnaB2 spontaneously forms a covalent isopeptide bond between Lys31 and Asp117 from SpyCatcher and SpyTag, respectively. These peptides can then be fused to the N- or C-terminus of scaffold and epitope proteins to allow VLP assembly. This system can be used to express decorated VLPs by expressing the scaffold in E. coli, while the epitope can be made in E. coli or a mammalian system. While there is no licensed drug made with this system, late-phase clinical trials have been reported for LYB001 and ABNCoV2 (Bavarian Nordic/AdaptVac, NCT05329220) [16]. Additional relevant epitope presentation systems include the linking of epitope proteins, such as RSV glycoproteins, to I53-50 nanoparticles, as well as glycosylphosphatidylinositol (GPI) anchoring to lipid bilayer structures [17,18,19].

We describe herein our novel platform, EPIclip™, for the decoration of VLPs by co-expression of scaffold and epitope proteins attached to high-affinity binding partners colicin E7 (ColE7) and immunity protein 7 (Im7). We fused the mutationally inactivated E. coli DNase ColE7 as a fusion tag to epitope proteins and simultaneously fused Im7 into the VLP scaffold protein. Simple co-expression of these proteins in a single prokaryotic host (E. coli) results in the generation of decorated VLPs. The colicin family and their corresponding immunity protein partners are ideal candidates for this system due to their ultra-high binding affinity (Kd~10^−14^ M), allowing for virtually irreversible binding under physiological conditions without the need for covalent linkage [20]. Additionally, these binding partners are hydrophilic to prevent aggregation, feature a small size (<12 kd), and have no equivalent homologous mammalian proteins, therefore preventing autoimmunity. We explore this approach in detail using the well-characterised HBc scaffold but also demonstrate the potential application of a plant-derived virus scaffold from the tomato aspermy virus (TAV), indicating a general applicability to numerous VLP scaffolds. We demonstrate the application of this platform with an interleukin-31 (IL-31) epitope. IL-31 is a T-cell-derived cytokine involved in the progression of pruritic skin diseases, such as atopic dermatitis, and has become an important target for drug development in this area [21].

2. Materials and Methods

2.1. Specific Structural Design Features of the ColE7–Im7 Fusion Platform

For ColE7, we did not use the wild-type catalytic DNase domain but instead the previously described triple mutant Arg538Ala, Glu542Ala, and His569Ala, which completely abrogates any toxicity [20,22]. For Im7, we employed the F41L mutation, which increases protein folding by 20-fold while retaining native structure [23]. For the HBc capsid protein, we incorporated the F97L mutation, which significantly boosts VLP capsid formation [24].

2.2. Construction of His-Tagged ColE7-mIL-31 and Scaffold-Only Expression Vectors

Hepatitis B capsid VLP (Supplementary Sequence 1, Internal ID: 67863) was prepared by replacing nucleotides (nt) 5071–5197 of a pET28(+) (Novagen, Darmstadt, Germany) vector with the amino acid sequence M1-V149 of the HBc core assembly domain (HBc149) (GenBank ADA56824). To prepare a VLP comprising HBc fused to Im7 (Supplementary Sequence 2, Internal ID: 67867), nt 71–149 in pETDuet-1 (Novagen) were replaced with the same HBc149 sequence as above. Next, the Im7 sequence was inserted between L76 and P79 of HBc149, followed by the addition of six histidine residues C-terminal of V149, as described in Schumacher et al. [25].

2.3. Cloning of Tet-Inducible HDE-Tagged ColE7-mIL-31

To prepare an inducible mouse IL-31 (mIL-31) epitope fused to colicin E7 (Supplementary Sequence 3, Internal ID: 70979), a tetracycline repressor protein (tetR) gene was placed downstream of the AmpR gene and ribosomal binding site (gaatta) in a pETDuet1 vector to allow constitutive expression as specified in [13]. Next, the T7 promoter in pETDuet-1 was replaced by the tetR/tetA promoter immediately upstream of the epitope start codon. Immediately after the ATG, the HDE tag with linker (HAHEHRHDHEHGGGS) was inserted upstream of the catalytic domain of ColE7. The design of the HDE tag can be accessed in the Supplementary Methods.

2.4. mIL-31-HBc and hIL-31-HBc VLP Design and Expression

To produce VLPs decorated with murine IL-31, the ColE7-mIL-31 plasmid was modified to remove the C-terminal cysteine of the mIL-31 native sequence (Supplementary Sequence 4, Internal ID: 73849). Here, proteins were expressed separately and combined with VLP scaffolds produced from the HbcIm7 construct (Supplementary Sequence 2, Internal ID: 67867) prior to use in the in vivo pruritus study detailed below.

To produce VLPs decorated with human IL-31, the wild-type sequence was codon-optimised and inserted downstream of the ColE7 sequence with a linker (GGGSSGSG) inserted between the peptide sequences. The HDE tag with linker (GGSGSGGG) was inserted downstream of ColE7. The epitope sequence was inserted downstream of the HbcIm7 protein coding sequence linked by an internal RBS (Supplementary Sequence 5, Internal ID: 79105).

2.5. Tomato Aspermy Virus VLP Design and Expression

To produce an equivalent VLP in a plant-derived virus, the N-terminal domain (M1-K30) of the Tomato aspermy virus (TAV) protein (GenBank, ABM46610) was deleted, and Im7 was inserted between R46 and E48. The epitope sequence was inserted downstream of the scaffold protein coding sequence linked by an internal RBS (Supplementary Sequence 6, Internal ID: 78555). For expression, we integrated both the VLP scaffold and the ColE7-fused epitope into a single message driven by a T7 promoter within a pET duet plasmid (Novagen).

2.6. Bacterial Expression

Plasmid constructs were transformed via heat shock into chemically competent cells of a modified BL21/DE3 E. coli strain. After recovery, cells were spread onto LB agar plates containing 100 µg/mL ampicillin (Sigma Aldrich, St. Louis, MO, USA, A9518) or carbenicillin (Fisher Scientific, Waltham, MA, USA, 1016319) and incubated overnight at 37 °C. An initial expression screen was performed using 3 single colonies, and 50% glycerol stocks were made and stored at −80 °C. For larger-scale expression, starter cultures were inoculated from glycerol stocks in LB broth containing 100 µg/mL carbenicillin and incubated overnight at 30–37 °C. The following day, cultures were diluted to an optical density (595 nm) of 0.05–0.1 in 400–500 mL of LB with appropriate antibiotic at 37 °C. OD was monitored, and cultures were cooled to 16 °C when they reached an OD of 0.6–0.8. Once cooled, cultures were induced with either 0.3 mM IPTG (Millipore, Burlington, MA, USA 420322) for T7 promoters or 40 ng/mL of anhydrotetracycline (ATC) (Fisher Scientific, 15478139) for tetR promoters for 4–18 h, depending on the construct. Cultures were isolated by centrifugation, pellets were weighed, and wet cell weight (WCW) was recorded before storage at −80 °C or immediate processing.

2.7. Cytosol Preparation

For mIL-31 and HBc constructs, cell pellets were resuspended in 3–5 mL lysis buffer (50 mM Tris/HCl, 200 mM NaCl, pH 7.0) and sonicated in 50 mL conical tubes in an ice bath using 30″ on/off pulsating cycles at 12 kHz for 1 min per mL of lysate. For some experiments, benzonase was added, as indicated, and incubated for 1 h at 4 °C. Lysed cells were spun at 5000× g for 10 min at 4 °C, then filtered sequentially through 0.45 and 0.22 µm PVDF syringe filters. For studies comparing soluble and insoluble recombinant protein expression, the pellets were resuspended in the same volume of the removed cytosolic supernatant.

For all hIL-31 and TAV constructs, pellets were thawed and resuspended in 10 mL of lysis buffer (300 mM NaCl, 25 mM Tris, 10% sucrose, pH 7.4) per 1 g WCW. The cell suspension was passed three times through an Emulsiflex-C3 homogeniser at 25 KPSI at 4 °C. The homogenate was then centrifuged at 4000 RPM for 10 min at 4 °C. The supernatant was collected and filtered sequentially through 0.45 and 0.22 µm PVDF syringe filters.

2.8. Immobilised Metal Affinity Chromatography (IMAC)

IMAC purification was performed using a Sartobind IDA membrane (93IDA-42DB-12-V). The membrane was charged with 15 mL of 100 mM ZnSO_4_, followed by washing with 10 mL of deionised water to remove excess Zn^2+^ ions. Next, the column was equilibrated with 10 mL of equilibration buffer (25 mM Tris, 300 mM NaCl, 30 mM imidazole, and 20% (w/v) sucrose, pH 7.4). The starting material was adjusted to match the equilibrium buffer and loaded onto the membrane. Subsequently, the membrane was washed with 10 mL of washing buffer (25 mM Tris, 300 mM NaCl, 50 mM imidazole, and 20% (w/v) sucrose, pH 7.4). Bound VLPs were eluted using 5 mL of elution buffer (25 mM Tris, 300 mM NaCl, 250 mM imidazole, and 20% (w/v) sucrose, pH 7.4). Eluted fractions were collected in aliquots of 1 mL and analysed by SDS-PAGE.

2.9. Capto Core 400 Chromatography

For polishing, 600 µL of IMAC elution was applied onto 200 µL of packed Capto Core 400 resin (Cytiva, Marlborough, MA, USA, 17372402) in a gravity-drip column. The mixture was roller-incubated for 5 min at room temperature to allow for binding and internal capture of impurities. The 500 uL of flow-through was collected and analysed by SDS-PAGE.

2.10. Sucrose Density Centrifugation

To confirm particle self-assembly, sucrose gradient centrifugation was performed to reduce background impurities and isolate fully assembled VLPs for further analysis. Freshly produced cytosol was loaded onto successively decreasing concentrations of sucrose gradients in the equivalent lysis buffer. Briefly, 5.5 mL each of 60, 50, 40, 30, 20, and 10% (w/v) sucrose was layered into open-top, thin-wall ultracentrifuge tubes (Beckman Coulter, Brea, CA, USA, 344058), modified as described in Benen et al. 2014 [26]. Next, 5.5 mL of freshly prepared cytosol was loaded on top of the final sucrose layer. In clones where the cytosol was prepared with 10% sucrose, 10 mL of freshly prepared cytosol was loaded on top of the 20% sucrose layer. After centrifugation in a Beckman SW32Ti rotor at 30,000 RPM (100,000× g) for 6 h at 4 °C, layers were collected from the top to the bottom and analysed by SDS-PAGE.

2.11. Native Agarose Gel Electrophoresis (NAGE)

First, 1% agarose gels were cast in 0.5X TBE buffer with a 1:10,000 dilution of SYBR™ safe stain (Thermo Fisher, Waltham, MA, USA). Samples were prepared in 5X TBE buffer with 10% glycerol and loaded with Quick-Load^®^ Purple 1 kb DNA ladder (New England Biolabs, Ipswich, MA, USA, B7025. The gel was run for 1 h at 60 V, followed by SYBR safe stain visualisation and imaging on the ChemiDock Gel Imaging System (BioRad, Hercules, CA, USA) and Coomassie blue staining overnight at 4 °C.

2.12. Tangential Flow Filtration (TFF)

Tangential flow filtration was carried out using a MidGee hoop ultrafiltration hollow fibre cartridge (UFP-750-E-H22LA, MWCO = 750 kDa, 60 cm flow path, 1.0 mm filter internal diameter, surface area = 0.0051 m^2^) on a KrosFlo Research II Tangential Flow Filtration system with three polysulfone pressure transducers attached to the feed, retentate, and permeate outlets. First, the hollow fibre cartridge was flushed with 300 mL of type I ultra-pure water, then equilibrated with 300 mL of buffer (25 mM Tris, 150 mM NaCl, pH 7.4). The feed flow rate was kept constant at 20 mL/min with the maximum transmembrane pressure (TMP) reaching 1.5 bar (permeate to waste, retentate recirculated). The system was drained, and the feed reservoir was filled with 15 mL of 50% sucrose cushion containing VLPs from the sucrose gradient (load material). The load material was recirculated for 5 min (retentate recirculated, permeate closed) to achieve stable system parameters. The load material was buffer exchanged into 25 mM Tris, 150 mM NaCl, pH 7.4, equivalent to 7 diafiltration volumes (105 mL) of the starting material. The reservoir volume was kept constant using the syphoning ability of the KrosFlo Research II system (dead flow rate 20 mL/min, TMP 1.5 bar). The buffer inlet line, which was connected to the reservoir, was closed, and ultrafiltration was started. The filtrate was further concentrated 2-fold using a 20 mL/min feed flow rate and 1.5 bar TMP throughout the concentration process. Ultrafiltration was stopped when 7.5 mL of permeate was collected through the permeate outlet. The permeate outlet was then closed, and the filtrate was recirculated for 5 min before the system was drained.

2.13. Transmission Electron Microscopy (TEM)

For HBc VLPs, the sample was added onto a carbon-filmed mesh copper grid, which was glow-discharged to enhance hydrophilicity and left to adsorb onto the grid for 30 min. The sample was fixed onto the grid with 2.5% glutaraldehyde and incubated for 1 min. Next, the grid was washed three times with water and negatively stained with 3% uranyl acetate for 1 min. The stain was removed and left to air dry. The grid was imaged using a Jeol JEM 1400 microscope at an acceleration voltage of 80 kV and magnifications between 10,000 and 50,000×.

For TAV VLPs, 5 µL of 0.1 mg/mL VLP protein sample (in buffer 25 mM Tris pH 7.4, 300 mM NaCl, 10% Sucrose) was applied onto a freshly glow-discharged (2 times for 10 sec at 5 mA) C300-Cu grid (Electron Microscopy Sciences, Hatfield, PA, USA), incubated for 3 min, and manually blotted. Afterwards, 5 μL of buffer (25 mM Tris, pH 7.4, 300 mM NaCl) was applied onto the same grid and incubated for 1 min before the solution was blotted off, followed by applying 5 μL of UranyLess EM Stain and incubating for 1 min before blotting off. Images were acquired at a nominal magnification of 40,000–60,000× on a JEM 1400 80 kV microscope.

2.14. Expression and Purification of ColE7-mIL-31

Cytosol was prepared by sonication as described above with lysis buffer containing 150 mM NaCl, 25 mM Tris, pH 7.0, and subsequently incubated with Nickel-NTA resin (Cube-Biotech, Monheim, Germany) in the presence of 50 mM imidazole. The supernatant was removed, and the resin was washed twice with the same buffer, followed by elution with PBS and 250 imidazole. Imidazole was removed by slide-a-lyzer buffer exchange (2 mL sample volume in 50 mL tubes) overnight at 4 °C, followed by incubation for 1 h with TEV protease (produced and donated by Thomas Eadsforth, University of Dundee, Dundee, UK) at room temperature. The TEV-treated sample was passed twice over 1 mL Ni-NTA resin to remove HDE-tagged Colicin E7.

2.15. Receptor Binding Assay of Purified mIL-31

mIL-31RA (RnDsystems, # 3028-ML) was coated at 5 µg/mL to ELISA plates, followed by blocking with 2% BSA/PBS. mIL-31 was isolated from colE7-mIL-31 processing and added to the plate at 0.01–5 µg/mL and incubated for 1 h. The plate was subsequently washed three times with PBS and incubated with rabbit anti-mIL-31 (1:2000, eBioscience, San Diego, CA, USA) and incubated for 1 h. Next, the plate was washed and incubated with anti-rabbit-AP (1:10,000, Abcam, Cambridge, UK) for TIME. Wells were then washed four times with PBS-T, and 100 µL of TMB substrate (Thermo Scientific; 34029) was added per well, and the reaction was stopped after 15–20 min with 50 µL/well of 0.2 M sulfuric acid. Absorbance was read at 450 nm using a plate reader.

2.16. Circular Dichroism

Purified mIL-31, after TEV-mediated cleavage of ColE7, was concentrated by vivaspin, rebuffered into 50 mM Tris/HCl pH 7.4, and analysed by Dr Barbara Ciani, Centre for Membrane Interactions and Dynamics & Krebs Institute, Dept. of Chemistry, University of Sheffield, Sheffield, UK. The mean molar ellipticity at 222 nm was found to be approximately ~−22,600 deg cm^2^ dmol^−1^. Based on −39,500 deg cm^2^ dmol^−1^ for a 100% helical protein, yielding an estimate of 57% helical content, compared to the predicted 49%, which is within experimental error, given the limited precision of the protein concentration estimate.

2.17. In Vivo Pruritus Study

Animal experiments were carried out under UK Home Office licences PE5B75232 and PP9514985. Studies were subject to internal veterinary review as per the University of Dundee (Dundee, UK) standard protocol. Female C57Bl/6j mice, aged 12 weeks at baseline, were injected subcutaneously with either 40 µg of mIL-31-decorated VLP, formulated in PBS containing 0.05% Polysorbate 80 and 5% Alum hydrogel, or PBS-only (vehicle control) in a volume of 100 µL. Each mouse received three total doses, each spaced two weeks apart. Tail vein blood was sampled at baseline and one week after each booster. Terminal cardiac puncture was performed at the end of the study. Two weeks after the second booster, mice were placed individually, but with a clear sight of cage mates, to perform baseline video documentation of itching behaviour. For 30 min, video recording was performed simultaneously using cameras mounted on top of two cages, each capturing two cages. Thereafter, a single priming dose of 0.2 µg of murine IL-31 in PBS (produced as recombinant protein in E. coli and verified for bioactivity by receptor binding assay) was administered subcutaneously, dorsally between the ears, in a 100 µL volume. The following day, 2 µg of mIL-31 or vehicle control was administered in the same way. Mice were then simultaneously video-observed as per above for 30 min. Pruritus behaviour was quantified by scoring (i) hindfoot-to-ear scratching, (ii) mouth-to-dorsal flank scratching, and (iii) prolonged forelimb-to-nose grooming exceeding 3 s. IL-31-specific IgG titres were determined by ELISA.

2.18. hIL-31 In Vivo Study

C57Bl/6j and Balb/c mice aged 12 weeks at baseline were injected subcutaneously with hIL-31-decorated VLP, formulated in buffer (300 mM NaCl, 25 mM Tris, pH 7.4). Dosing was repeated twice for three total doses, each spaced two weeks apart. Tail vein blood was sampled at baseline and one week after the second booster. Terminal cardiac puncture and spleen isolation were performed at the end of the study, one week after the last booster.

2.19. Measurement of Anti-hIL-31 Antibody Responses

The antibody titres were measured using a commercially available glycosylated IL-31 (Acro Biosystems, Newark, DE, USA, HEK293 cell-produced). Recombinant hIL-31 was diluted to 1 µg/mL in 50 mM bicarbonate buffer, pH 9.6, and 100 µL per well was added to a high-binding 96-well plate and incubated overnight at 4 °C. Wells were washed three times with PBS/Tween-20 (PBS-T) and then blocked with 100 µL/well blocking buffer (5% BSA in PBS-T). Plasma samples were serially diluted (three-fold) starting at 1:200, and 100 µL/well was added to the plates. Baseline plasma was used as a negative control. Serially diluted anti-human IL-31 antibody was used as a positive control. The samples were incubated for two hours at room temperature. Plates were washed four times with PBS-T, and 100 µL of HRP-conjugated rabbit anti-mouse IgG (1:10,000) was added per well and incubated for 1 h at room temperature. Wells were then washed four times with PBS-T, and 100 µL of TMB substrate (Thermo Scientific; 34029) was added per well, and the reaction was stopped after 15–20 min with 50 µL/well of 0.2 M sulfuric acid. Absorbance was read at 450 nm using a plate reader. Dilutions are log-transformed and plotted versus observed optical density for each sample. A linear curve fit is then used to determine slope and y-intercept. The limiting titre for each sample is defined as the theoretical dilution that would yield an absorbance 10% above the absorbance seen in the pre-immune samples.

2.20. Splenocyte Isolation and ELISPOT Analysis

Splenocytes were isolated from spleens by gently pressing the tissue through a 70 µm cell strainer into a 50 mL conical tube and washed with 2% FBS/PBS solution. The cell suspension was centrifuged at 300× g for 5 min. The supernatant was discarded and gently resuspended in 2 mL of RBC lysis buffer (Thermo Fisher, 004300-54) and incubated for 3 min, with gentle mixing. The RBS lysis buffer was neutralised with 30 mL of FBS/PBS solution and centrifuged at 300× g for 5 min. The supernatant was discarded and resuspended for cell counting and viability estimation prior to freezing in 90% FBS/10% DMSO solution.

Isolated splenocytes were thawed and seeded (300,000 cells per well) on a pre-coated, murine IFN-γ ELISPOT 96-well plate (ImmunoSpot, Cleveland, OH, USA, mIFNgP-2M/5). Cells were treated with either VLP capsid (0.8 µg/mL), hIL-31 (10 ng/mL), or concanavalin A (3 µg/mL) as a positive control. The cells were incubated overnight in a CO_2_ incubator at 37 °C. The ELISPOT assay was performed according to the manufacturer’s protocol, and plates were allowed to dry before imaging. Wells were imaged using a dissecting microscope with an attached camera. Within each image, a region of interest (ROI) of equal area was selected using the rectangle selection tool in ImageJ v. 1.54g (Fiji). ROIs were pre-processed using the bandpass filter operation (filter—large = 40, filter—small = 3, suppress = none, tolerance = 5, auto-scale, saturate), followed by an unsharp mask (radius = 4, mask = 0.4) and converted to 8-bit, grey-scale images. Pre-processed images were then manually analysed by thresholding using the Triangle method and adjusted to select spots. Next, the analysis particle tool (size = 0.008–infinity, circularity = 0.05–1.00) was employed to generate a sum of spots in each well.

2.21. Statistical Analysis

Data are presented as means ± standard deviation. Unless otherwise stated, statistical significance was assessed using a two-sided independent T-test with p-values ≤ 0.05 considered significant and <0.001 considered highly significant. All statistical analyses were performed using GraphPad Prism v. 10.6.0 (San Diego, CA, USA).

3. Results

VLPs decorated with full-length proteins can be produced in a single host system if high-affinity binding of the epitope and VLP scaffold can be mediated at the time of protein assembly. To achieve this, a binding pair of proteins can be employed to tag the epitope to the scaffold. To explore this concept, we selected the ColE7/Im7 pair of binding proteins based on their extreme binding affinity (Kd » 10^−14^) [27], small size, and surface charge distribution. Im7 has a high net negative charge that, when integrated into a VLP capsid, reduces aggregation through electrostatic repulsion (Figure 1A,B). The selection of the hepatitis B core antigen as a scaffold was based on several considerations: (i) it has been validated in clinical trials, confirming principal safety for use in humans [28,29,30]; (ii) it has a well-characterised insertion point for heterologous fusion tags; and (iii) it elicits a B-cell-biased IgG-focused response that is active even in a T-cell-independent context [31,32,33], suggesting the potential for robust therapeutic responses, even in potentially weak vaccine responders. As an insertion site for the Im7 protein, we used the major immunogenic region (MIR) localised on the tip of the HBc building block dimer (Figure 1C). We deleted the topmost residues—E77 and D78 (Figure 1D)—in order to abrogate any immunological cross-reactivity to HBcAg. Even though this deletion has been found to render capsids unstable by removing key negative charges [34], this design was possible due to the cupping of capsid spikes with highly negatively charged Im7. Additionally, histidine residues were added at the C-terminus of the capsid formation domain of HBc, which is located inside the VLP; we called this HBcIm7. This has been shown to significantly enhance capsid stability [25]. We also evaluated the standard HBc capsid (HBc149) alongside HBcIm7 as a control.

Wild-type HBc149 and HBcIm7 capsid proteins were placed under T7 promoter control in pET plasmids. HBcIm7 demonstrated comparable expression to HBc149 (Figure 2A) as well as relatively high cytosolic solubility (Figure 2A, right panel). Sucrose density gradient analysis confirmed the formation of particles, mostly partitioning to the 40%, less to the 50% sucrose cushions, similar to HBc149 (Figure 2B).

Native agarose gel analysis suggested that these two fractions corresponded to the T4 and T3 capsid structures, respectively (Figure 3A), as described in Schumacher et al. [25]. For wild-type HBc149, the 40% sucrose fraction contained mainly T3 capsids, while the 50% sucrose fraction contained a mix of T3 and T4 capsids. For HBcIm7 capsids, the addition of Im7 to the capsid was associated with an expected shift in the T3/T4 migration distance, respectively. As wild-type HBc149, the 40% fraction contained mostly T3 capsid, while the 50% cushion contained an equal abundance of T3/T4 capsids. However, it appears the altered double-layered rim structure inhibited particles from entering the 1% agarose gels. Next, electron microscopy was performed to confirm capsid formation (Figure 3B). Morphologically, HBcIm7 appeared fuzzy with a broader rim (Figure 3C). The broader rim suggests that there might be additional mass, which is consistent with the addition of Im7 to the HBc (a double-protein surface layer). The overall capsid size was not significantly different between HBcIm7 and HBc140 capsids when quantified via TEM; capsid size in HBcIm7 VLPs compared to HBc149 VLPs was 26.3 ± 3 nm and 25.9 ± 4 nm, respectively (Figure 3D). However, dynamic light scatter analysis exhibited a slightly larger apparent hydrodynamic diameter for HBcIm7 (Figure 3E). The larger diameter observed via DLS compared to TEM is in keeping with previously reported differences between the two methods [25].

We then performed purification experiments to demonstrate the utility of the construct for larger-scale production. Unexpectedly, the HBcIm7 capsid exhibited a highly selective affinity to the weak anion exchanger DEAE but not the strong anion exchanger, quaternary ammonium (Q), which we exploited for single-step purification of particles (Supplementary Figure S1). Taken together, the data show that Im7 can be incorporated into the HBc capsid, yielding VLPs of the expected size.

We next assessed the fusion of ColE7 to a model epitope, choosing the cytokine IL-31 as a clinically relevant target. The clone design, shown in Figure 4A, included a short N-terminal tag, followed by the ColE7 domain C-terminally fused to murine IL-31 (mIL-31). The tag, termed the HDE tag, was designed to simultaneously impart a negative charge onto the surface of the decorated VLP, as well as to facilitate purification. We screened various linker designs (Supplementary Table S1). In order to allow verification of the structural integrity of ColE7-fused IL-3, we inserted a TEV protease restriction site to facilitate ColE7 removal. ColE7-fused mIL-31 was expressed within the E. coli host and could be purified to near homogeneity via IMAC using the novel HDE tag (Figure 4B). After removal of ColE7, the purified mIL-31 was used to evaluate native folding by circular dichroism (Figure 4C, see methods), and our data demonstrate a similar pattern of alpha helices as the prediction. Next, we demonstrated intact binding to the IL-31 receptor by ELISA (Figure 4D). Taken together, HDE-tag-ColE7-fused epitope protein can be expressed with an E. coli host, and the purified mIL-31 can functionally bind to the mIL-31 receptor.

The inclusion of T-cell stimulatory peptides has been reported to enhance VLP immunogenicity [37,38]. Consistent with this, several papers suggest that the incorporation of host cell RNA into the VLP boosts target-specific IgG formation via TLR7 stimulation [39,40]. We therefore investigated the effect of VLP scaffold modifications accordingly. As a T-cell epitope, we selected the TpD peptide [41] shown to boost T-cell memory (Figure 5A). In addition, we reintroduced different lengths of the HBc capsid C-terminus (Figure 5A), which regulates RNA-binding capacity ([42]). We then expressed and purified the ColE7-mIL-31 epitope and VLP scaffold separately (Figure 5B), followed by mixing with the various individual scaffold constructs. Co-elution of both the scaffold and epitope on SEC chromatography from the mixed components (Figure 5B, bottom) suggested binding of the epitope to the VLP scaffold. SDS-PAGE showed an approximately even intensity of scaffold and epitope bands, further suggesting epitope decoration. Each of the constructs was then injected into C57Bl/6j mice, followed by determination of the epitope-specific IgG titre. To compare the immunogenicity of the resulting polyclonal IgG response relative to a monoclonal IL-31-specific antibody, we included a dose response of a monoclonal antibody on the same ELISA plate (Figure 5C), allowing quantification of the observed IL-31-specific signal in vaccinated mice as equivalent to a concentration of the monoclonal antibody. All four VLP scaffolds yielded mIL-31-specific IgG concentrations exceeding 10 µg/mL after two injections (Figure 5D) without exhibiting statistically significant differences in immunogenicity. We thus selected the parent scaffold HBc-Im7-149 (HBc-Im7), the scaffold featuring the T-cell epitope, and a partial RNA-binding domain (HBc-167-TpD) to study the long-term durability of the response. Both designs yielded sustained IL-31-specific IgG levels across the observational period of six months (Figure 5E). We conclude that the addition of RNA-binding regions or T-cell stimulatory peptides does not improve functional performance in this system. Subsequent studies were therefore conducted using the HBcIm7 VLP scaffold.

We next assessed the resilience of the design to genetic variability in vaccine recipients and the ability to functionally inhibit IL-31-induced pruritus in vivo. Dosing of outbred mice showed uniformly high IL-31-specific IgG titres, similar to dosing in Th1-biased C57Bl/6 mice (Figure 6A); however, we did not run monoclonal antibody equivalence curves for this assessment. Functionally, while challenge with subcutaneous recombinant bioactive mIL-31 led to the expected increase in pruritus in control mice, no increase in pruritus was noted in mice vaccinated with IL-31-decorated VLP or in mice that had only received challenge with vehicle injection instead of bioactive mIL-31 (control) (Figure 6B). The IL-31-challenge pruritus assay is known for notable variability between individual mice. We therefore quantified the change in pruritus as fold change before and after challenge for each mouse (Figure 6C). There was a marked increase in pruritus events in 80% of mice from the control group after the IL-31 challenge but not in mice that had received prior vaccination with mIL-31-decorated VLP. These data suggest that the VLP-decorating system described here is able to generate an IgG response capable of blocking the biological activity of the target in vivo.

For clinical translation of self-protein-targeting VLPs, target-specific T-cell activation remains a concern. To assess the propensity of IL-31-decorated HBcIm7 VLPs to generate autoreactive T cells, we vaccinated C57BL/6j mice, which have a strong Th1/gamma-IFN bias, with target-adapted human IL-31-displaying VLPs. As a control, we vaccinated a second group with non-decorated VLPs. One week after the final dose, splenocytes were collected and analysed with ELISPOT for the accumulation of IFN-gamma after stimulation. Evaluation of the entire splenocyte population was chosen to maximise the sensitivity of the assay by allowing signal production from all potentially relevant populations (e.g., macrophages and DC subsets) and to include the potential for MHC-II and MHC-I cross-presentation. Future experiments may seek to include CD4/CD8 sorting prior to analysis to confirm our present results. When splenocytes were stimulated with concanavalin A (positive control), a brisk IFN-gamma response was observed. Similarly, VLP capsid triggered a sizeable IFN-gamma response, confirming the capsid’s ability to generate strong T-cell activation. This was drug-specific, since splenocytes from vehicle-only vaccinated animals did not respond. By contrast, no IFN-gamma was produced after stimulation with native glycosylated human IL-31, ruling out a clinically relevant autoreactive T-cell response in this murine model. The entire study was carried out twice: once with a freshly prepared sample (Figure 7B, left) and once with a frozen–rethawed sample (right); both studies yielded similar results. Taken together, the data do not suggest potential for the generation of T-cell autoreactivity against the self-protein epitope.

Successful incorporation of Im7 may be facilitated by the prominent spike present in the HBc capsid. However, other virus capsids may be preferable for certain applications. We therefore explored the feasibility of extending the system to other capsids with less obvious insertion sites using two additional virus capsids: porcine circovirus 2 (PCV2) (Supplementary Figure S2) and tomato aspermy virus (TASV). We selected the TAV capsid as it is closely related to the cucumber mosaic virus (CuMV) capsid, which has validated applications as a VLP [43,44], while being more resilient to NaCl, RNase, and temperature challenges than CuMV [45,46,47]. We identified a potential Im7-insertion site in the S1/S2 loop, which indicated stable folding and retained native structure of the pentamer building block by structural modelling (Figure 8A). Based on the experiments detailed above (Figure 5), we deleted the capsid-internal N-terminal unstructured RNA-binding domain. For expression, we integrated both the VLP scaffold and the ColE7-fused epitope into a single expression cassette driven by a T7 promoter and separated by an internal ribosome binding site (RBS) (Figure 8B). Upon induction with IPTG, two bands of equal intensity were visible in cytosolic fractions corresponding to the expected size of the VLP scaffold and epitope bands, respectively, and both bands completely relocated to the high sucrose cushions in density gradient analysis, indicating complete capsid formation (Figure 8C). For purification purposes, we added the HDE tag to the C-terminus of the epitope. This afforded efficient single-step purification via IMAC (Figure 8D). Interestingly, when we polished the purified VLP fraction using CaptoCore 400 multimodal resin, we observed a high degree of retention, indicating a smaller-than-expected particle size. The reported secure cut-off size for this resin is approx. 20 nm [48]. Indeed, TEM analysis exhibited two distinct subpopulations of VLPs, with a bimodal diameter distribution of 28 nm and 20 nm, respectively (Figure 8E,F). The former size corresponds to the T = 3 capsid assembly reported for TAV [49]. The smaller size is consistent with a T = 1 architecture. While this has not been reported for TAV capsids, a number of closely related icosahedral capsids, including BMV, adopt a T = 1 architecture after removal of the N-terminal RNA-binding domain [50], corresponding exactly to the design chosen here for the TAV scaffold. Although we do not have the crystal structure of the construct, the observed data are consistent with the interpretation that N-terminal-deleted, Im7-integrated TAV capsids form a predominant T1 and minor population of T3 capsids, respectively.

Functionally, IL-31-decorated TAV VLPs generated target-specific IgG responses comparable to HBc-derived VLPs (Figure 9A–C). Therefore, we conclude that the ColE7/Im7 system is not limited to the HBc scaffold.

4. Discussion

In recent years, the use of VLPs as a platform for vaccine development has led to the clinical translation of many candidates. However, these have mainly focused on directing a response to the capsid itself, such as human papillomavirus (HPV) and influenza [12,51]. In comparison, few early-phase and late-phase VLP candidates have been developed that function to present an epitope to the immune system, such as the presentation of ACE2 for COVID-19 and the display of associated prefusion trimer proteins for the prevention of respiratory syncytial virus (RSV) and human metapneumonia virus (hMPV) [18,52,53].

In the present study, we describe a novel approach to the decoration of VLPs using high-protein binding pairs: colicin E7 and Im7. As described, this pair demonstrates exceptionally high binding affinity (Kd~10^−14^ M), small size, and hydrophilicity to prevent protein aggregation [20]. For all practical intents, the ColE7/Im7 interaction functions as a virtually covalent interaction under physiological conditions. A similarly favourable assessment has been made for the enzyme Barnase, together with its cognate Barstar [54].

In our platform, epitope and scaffold can be co-expressed within a single host or separated to allow expression in other hosts, depending on the requirement for post-translational modification (e.g., glycosylation). Importantly, we have demonstrated that epitope and scaffold proteins can be purified separately and subsequently mixed to produce a functional final product. We also demonstrated the concurrent expression of both system components from a single promoter, which may be advantageous for an adaptable bioprocess. However, we have also placed the epitope and scaffold under the control of different promoters while still located on a single plasmid, allowing for differential induction conditions (e.g., temperature) depending on the requirements of potential epitope proteins.

In addition to VLPs, several alternative antigen presentation platforms have been developed, including the use of nanoparticles and liposomes [17,18,19]. VLPs offer an advantage, as their highly organised and repetitive capsid structures are ideal for B-cell activation, they are of optimal size for uptake by antigen-presenting cells, and they offer high stability compared to other platforms [2,3]. The decoration of VLPs can be performed by either chemical conjugation, genetic fusion, or tag coupling [55]. Conjugation requires chemical treatment to crosslink antigen proteins to the scaffold, often leading to non-uniform decoration [55]. In comparison, genetic fusion and tag-coupling methods allow for stoichiometric and more precise antigen decoration. Several tag-coupling systems exist, including biotin–avidin, HaloTag, and SpyTag/SpyCatcher [15]. Another alternative is the use of GPI-anchoring to lipid membranes, which can be employed with liposomes or enveloped VLPs [19]. Due to post-translational incorporation of epitopes, this method reduces the risk of disrupting VLP capsid folding or assembly [19]. However, protein transfer to the membrane can result in variable surface density compared to fusion or tag-coupling methods. GPI anchors are also susceptible to cleavage by phospholipases and represent less stable alternatives compared to other VLPs, which may result in reduced immunogenicity over time [55].

The density of epitope proteins on the VLP surface is a critical parameter modulating the immunogenicity of the resultant drug [39,56,57]. Thus, a high density of the epitope achieves superior B-cell receptor crosslinking on cognate B cells, affording a more potent epitope-specific IgG response [39]. For SpyCatcher-system-produced drugs, the reported degree of epitope density on the VLP surface varies between 22% and 88% [58,59,60]. Therefore, a platform that achieves near-complete surface coverage of the VLP with the epitope would be desirable. While the results of this study (i.e., TEM morphological analysis and immunogenicity evaluation) indicate VLP decoration, we have not yet provided a quantitative analysis of VLP decoration. Future studies should apply advanced cryo-electron microscopy/tomography to quantify epitope structure and density of candidate VLPs.

We chose IL-31 as our epitope due to its clinical relevance in the development of various dermatological conditions, such as atopic dermatitis and pruritus [21]. IL-31 also has an important veterinary relevance, and IL-31 has been chemically coupled to scaffolds, yielding successful vaccines in both dogs and horses [61,62]. Monoclonal antibodies (mAbs) against IL-31 are currently used for the treatment of moderate to severe pruritic diseases (e.g., nemolizumab, lokivetmab) [21]. However, mAbs are expensive to develop and administer and require ongoing dosing. In addition, long-term use of mAbs can result in the development of anti-drug antibodies, which neutralise their activity and significantly reduce the efficacy of the treatment [63]. As a result, a VLP-based vaccine against IL-31 represents a promising and cost-effective therapeutic approach for pruritic disease.

Importantly, our ELISPOT data suggest the absence of autoreactive T cells. Since our model epitope is a self-protein (IL-31), it is important that the VLP design is optimised to prevent the development of autoreactive memory T cells, which can lead to chronic autoimmunity and tissue inflammation. VLPs represent a good strategy for the presentation of self-protein epitopes, as the repetitive structure strongly activates B cells and provides a foreign epitope to be presented and promote helper T cells without auto-reaction [64,65,66]. Previous studies with VLP-based vaccines against self-proteins, such as equine IL-5 and proprotein convertase 9 (PCSK9), resulted in the development of specific IgG antibodies without detrimental autoreactive T cells [66,67]. Interestingly, the addition of a highly potent T-cell epitope, combining both pertussis and tetanus sequences, with or without additional reintroduction of the sequences mediating incorporation of host cell RNA, did not further boost the immunogenicity of the HBcIm7 capsids. It is possible that such effects are no longer imperative when epitope density is sufficient, but this remains speculative. Consistent with this, previous studies have demonstrated that the addition of another potent T-cell epitope (PADRE) also failed to yield a useful boost to immunogenicity in another nanoparticle vaccine [68].

5. Conclusions

In conclusion, we provide a novel platform, EPIclip™, for the decoration of VLPs using high-affinity protein binding partners colicin E7 and Im7 produced in a single E. coli host. We employ a hepatitis B scaffold protein and an IL-31 epitope and successfully demonstrate immunogenicity without the generation of autoreactive T-cells. Importantly, dosing with IL-31-VLPs prevents the development of IL-31-induced pruritus in mice, suggesting in vivo target neutralisation. Taken together, these results suggest our platform may provide an effective alternative to monoclonal antibody treatment for pruritic diseases. We have also demonstrated that these VLPs can be purified at a small scale, and their manufacturing potential should be explored further. Larger mammal in vivo studies will provide further insight into immunogenicity before translation to clinical studies. Additionally, due to the modular nature of this platform, the epitope protein can easily be replaced with an alternative target, such as protein G or fluorescent markers (e.g., mini-GFP) for precipitation or detection assay development. Further, we show that the system can be applied to different virus capsids (e.g., TAV and PCV2). As a result, we provide a modular VLP-decorating platform, which will be a useful tool for applications in both research and pharmaceutical development.

6. Patents

GCS holds patents (UK 2519675.9, pending; UK 25038006.8, pending; US 18/996,838, pending; US 18/038,585 2024-0093159, pending) related to this present study. GCS also holds a UK trademark application for EPIclip (No. 00004267942). All other authors declare no competing interests