1.1. Antibodies for human therapeutic use

All three viruses contain surface glycoproteins that can elicit neutralizing antibodies: the RVFV Gn protein, the SBV Gc protein and the MERS-CoV spike protein. The structures of these proteins were elucidated during the ZAPI project, both from ZAPI participants and from groups outside the ZAPI project [[2], [3], [4]].

The conventional approach was to produce monoclonal antibodies (mAbs) against the viruses using wild-type mice. Alternatively, transgenic (“H2L2”) mice designed to produce human-rat chimeric antibodies that can be reformatted to fully human antibodies, and camelids (llamas and dromedary camels) engineered to produce heavy chain-only antibodies (HCAbs) and VHHs. HCAbs are devoid of light chains and have long complementary-determining regions with unique epitope binding properties, allowing them to recognize and bind with high affinity to epitopes not recognized by conventional antibodies. VHHs (sometimes referred to as nanobodies, due to their small size), are the antigen-binding domains of HCAbs that can be produced in bacteria and yeast in large amounts, at low cost. Furthermore, VHHs can be used as building blocks for the generation of multifunctional complexes. To test the antiviral activity of VHHs, in vitro neutralization assays were developed, in which viruses were incubated with the VHHs and with susceptible cells. Neutralization was screened for large numbers of VHHs using innovative virus neutralization tests (VNT), suitable for high-throughput analysis with automated read-out. Importantly, the same assays can be used to screen natural antibodies from human patients’ sera.

Studies with bunyaviruses demonstrated that neutralization of RVFV and SBV is most efficient when combining VHHs targeting different viral glycoprotein subdomains [5]. These findings stimulated the development of multivalent complexes, in which VHHs are covalently linked to scaffolds through an innovative technology known as “bacterial superglue” [6].

The most promising VHH combinations were subsequently genetically fused, yielding monospecific or bispecific constructs. The bispecific constructs were efficacious in preventing death of RVFV infected mice, especially when given prophylactically (6 h before virus challenge, 100% survival), and to a lower extent when given therapeutically (18 h after virus challenge, 60% survival). The effect was highest when VHHs targeting different regions on the Gn protein were combined, leading to a strong synergistic effect.

For SBV, as wildtype mice are not susceptible, interferon receptor knock-out (IFNAR^/-) mice were treated with monospecific or multispecific nanobody complexes one day before viral challenge. Prophylaxis with monospecific complexes led to 80% survival, whereas all mice survived when inoculated with multispecific complexes, demonstrating that combinations of VHHs targeting different antigenic sites result in very potent neutralization activity.

Antibodies against MERS-CoV were derived from dromedary camels that were first vaccinated and subsequently challenged with MERS-CoV. These dromedary camels developed exceptionally potent virus-neutralizing antibody responses [7]. Several MERS-CoV-specific VHHs were identified. In vitro, these VHHs efficiently blocked virus entry at picomolar concentrations, binding the receptor binding domain (RBD) of the viral spike protein with exceptionally high affinity. Furthermore, camel/human chimeric HCAbs with the camel VHH linked to a human Fc domain (HCAb-83) had an extended half-life in serum and protected mice against a lethal MERS-CoV challenge [7].

Initially a broad panel of functionally diverse, fully human antibodies was generated from immunized H2L2 mice. Eight human mAbs were developed that target functionally distinct domains of the MERS-CoV spike protein, interfering with the three critical entry functions of the MERS-CoV spike protein: sialic acid binding, receptor binding and membrane fusion. Neutralizing antibodies were shown to target multiple, non-overlapping epitopes within the receptor binding domain as well as in the fusion subunit of the spike protein. Passive immunization with these neutralizing antibodies protected mice from lethal MERS-CoV challenge [8].

Apart from MERS-CoV specific antibodies, antibodies with remarkable cross-reactivity were identified, that can be useful to pre-empt future epidemics [9]. The broadly reactive antibodies were shown to target the stem-helix region of the spike protein, which is conserved across betacoronaviruses. The antibodies provided protection in mice against MERS-CoV challenge, demonstrated by reduced weight loss, reduced pneumonia and pulmonary virus replication [9].

Generally, industry uses transient expression platforms to rapidly generate material during the early stage of drug development. To date gram per litre expression levels have been reported [[10], [11], [12]], however for resupply, a repeat transfection is required. To ensure process consistency, phenotypic characterization of the host cells and optimal process conditions were carefully assessed. The transient transfection process was assessed for process robustness at 5 L scale and then finally at 200 L scale to demonstrate scalability.

For purification, the traditional column process was replaced by membrane-based filtration, allowing the process to remain robust in a shorter time. The time required for a typical viral clearance study will delay the response time during a pandemic. However viral clearance can be claimed based on prior data on viral inactivation, anion exchange and viral filtration. Process performance of six different 5 L batches showed a recovery higher than 70% and purity close to 100%. At the same time, the host-cell protein levels were well below the 100 ng/mg threshold, typical of mAb products which further demonstrates the suitability of the transient expression process.

Normal release testing would require approximately 21 days, which could be reduced to 7 days by removing the physicochemical assays and relying on data obtained during the development phase. Furthermore, testing of bioburden could potentially be accelerated with ScanRDI, flow cytometry-based testing (Biomerieux, www.biomerieux-industry.com/products/scanrdi-rapid-and-sensitive-detection) that provides results in just 90 min, and with sterility testing using the Celsis Advantage system (Charles River, www.criver.com/products-services/qc-microbial-solutions/microbial-detection) that reduces time of analysis from 14 to just 5 days. However, these methods still require further validation.

A spike protein binding assay was selected to compare the potency and stability to a reference standard, through a MesoScale Discovery technology. Several factors were optimized such as spike protein concentration, incubation time, detection antibody concentration and optimization through a mAb dilution series.

During scale up, the process was optimized to maintain consistent yields and productivity by addressing vessel geometry, heat transfer, mass transfer and vessel fluid dynamics. Addition of the transfection complex was a challenge, as time and mixing are critical factors to achieve expression.

Product quality attributes were consistent over different batches and at different scales and were within typical criteria for material release. The relative binding activity was similar at the different scales, indicating functionality was comparable.

Long-term stability of mAb purified from the 200 L scale production runs was tested. Purified mAbs were stored at a concentration of 50 mg/mL in a standard buffer for 6 months at −80 °C, 5 °C (stress storage condition) and 25 °C (accelerated stress condition), and stability was tested at 1.5, 2, 3, 4.5 and 6 months. Differential calorimetry was used to assess the thermal stability of the molecules, while dynamic light scattering was used to investigate propensity for self-association. Additionally, samples were tested for visible and subvisible particles as well as size exclusion chromatography for assessment of purity. A recommended storage condition of −80 °C was proposed based on data generated, although a more suitable storage condition will be required under pandemic conditions.

1.2. Vaccine by design methodology for veterinary vaccines

Viral attachment and entry of SBV are mediated by the envelope glycoproteins Gn and Gc, of which Gc is targeted by neutralizing antibodies. The locations of the epitopes on the glycoproteins were identified by analyzing the reactivity of a set of mAbs and antisera against recombinant SBV glycoproteins [13]. One of two N-glycosylation sites was shown to be essential for interaction of Gc with a subset of Gc-specific mAbs. The epitopes could be grouped into two clusters, as revealed by fine mapping using chimeric proteins. Combination of disulfide bonding and epitope mapping generated a structural model of the SBV Gc N-terminus [14]. X-ray crystallography studies showed the projecting spike of SBV as the major target of the neutralizing antibody response, and generated X-ray structures of the spike protein in complex with two neutralizing antibodies [2]. Immunization of mice with the spike domains elicited virtually sterilizing immunity [2], whereas immunization with "Gc Amino"-encoding DNA plasmids, or "Gc-Amino" expressed in a mammalian system, conferred protection in up to 66% of the animals [15]. A multivalent antigen containing the covalently linked Gc domains of SBV and the related Akabane virus protected mice and cattle against SBV challenge infection [15]. The same domains of other orthobunyaviruses can be used as immunogens, showing the transferability of the methodology within the same viral genus.

Fold recognition identified four possible arrangements of both RVFV glycoproteins in the virion envelope. These models proposed that the ectodomain of Gn forms the majority of the protruding capsomer and that Gc is involved in the formation of the capsomer base [16]. Antigenic regions within Gn were mapped with Pepscan™, identifying six continuous and three discontinuous epitopes. Eight subdomains of the Gn glycoprotein were expressed and one domain showed high expression levels and was well recognized by neutralizing mAbs. This domain, Domain VII, corresponded with a subdomain within the Gn crystal structure referred to as “head domain” (Gn_head) [17]. The antigen was genetically fused to a SpyCatcher (SpyC) at the N-terminus, and produced by a recombinant baculovirus, in insect cells. Three self-assembling multimeric protein scaffold particles (MPSPs) with C-terminal SpyTags (SpyT) were produced in E. coli: lumazine synthase (LS), aldolase (ALD) and E2. Combining the SpyC-linked Gn head domain with SpyT-linked MPSPs resulted in particles displaying the antigen at the surface. Efficacy of the candidate vaccine was initially evaluated in the RVFV mouse model. Partial protection was achieved in the absence of adjuvant, whereas full protection was achieved in the presence of adjuvant. The neutralization titers in the mice correlated with protection [18].

Vaccination of young dromedary camels with a modified vaccinia virus Ankara (MVA-S) expressing the full-length Spike protein of MERS-CoV significantly reduced infectious MERS-CoV excretion from the nose [19]. The identification of dipeptidyl peptidase 4 (DPP4) as the MERS-CoV receptor facilitated the subsequent characterization of the RBD in the S1 region of the MERS-CoV spike protein. Vaccination with the S1 region reduced MERS-CoV excretion after challenge of animals. Using the SpyT/SpyC system, the RBD was coupled to LS MPSPs and used to immunize rabbits. This resulted in neutralizing antibodies, which could prevent the production of infectious virus in the rabbits. The technique can also be used to couple on the same particle multiple RBDs from different but related viruses [20].

1.3. Scale up for manufacturing MPSPs and antigen subunits

The choice of MPSPs for antigen display was based on their ability to self-assemble, and their high stability. Initial studies were done at small scale, from which the aldolase-SpyC and E2-SpyT were chosen to proceed with upscaling. The process was taken from small scale in shake flasks to medium scale - 5L fermenters. For the ALD-SC MPSP, a simple downstream processing program is needed: sonication for cell lysis, centrifugation, heat treatment, a second centrifugation step, size exclusion chromatography, and endotoxin polishing. The E2-SpyT could not be purified after heat treatment, so the downstream process relied on ion-exchange chromatography.

1.4. Antigen subunit production in baculovirus and C1 fungus expression systems

Using insect cells for expression, some issues were encountered: with the Gn head domain of RVFV, decreased expression was noted with increasing seed passages, whereas the harvest time of the Gc-Amino domain of SBV needed optimization, as the antigen was unstable in the culture supernatant. Although the baculovirus expression levels were 8–10 times lower than expected, the medium-scale production was found to be robust and reproducible. Yields were not improved using protease inhibitors. Until stability studies have been performed, the product should be stored at −80 °C.

For the C1 fungus expression platform, the initial aim was to achieve expression of at least one antigen for each of the three viruses, to produce stable antigen subunits, and to reach expression levels of at least 100 mg/L. Three antigens were produced and secreted into the medium, MERS-CoV RBD, SBV Gc-Amino and RVFV Gn_head. However, initially, the expression levels were relatively low, and the stability was limited. Further optimization was done by knocking out nine proteases in the C1 fungus genome, which led to a maximum yield of 1800 mg/L, 300 times higher than the baculovirus system.

1.5. MPSP/antigen subunit coupling step and quality control system

A broad range of conditions were tested to find the optimal conditions for coupling, showing that this can be done overnight at room temperature. The coupling kinetics for ALD-SC + ST-SBV and E2-ST + SC-RVFV were similar.

For the analytical QC system, the objective is to define the identity, quantity and quality of the two components and of the final vaccine complex. For the two components, this is relatively straightforward; the MPSP (first component) quality can be assessed as a homogeneous population of self-assembled particles, and the antigen subunit (second component) can be assessed by ELISA with virus neutralizing mAbs. For the MPSP-subunit complex, the amount of subunit needs to be quantified and quality assessment is based on the absence of free subunit in the final product; the coupling ratio also needs to be checked. Two techniques are used for QC of the MPSP-subunit complexes, asymmetrical flow field-flow fractionation (A4F) and quantitative mass spectrometry (qMS). A4F can provide information on the particle population homogeneity and the particle complex average molecular weight, both of which can be used to indirectly determine the coupling ratio. qMS can determine the exact identity and quantity of both the MPSP and subunit and can thus provide insight in the direct coupling ratio. Combining the two techniques provides a full picture for the quality of the final MPSP complex product.

1.6. In vivo efficacy

The candidate vaccines were validated in target species. The efficacy of the RVFV candidate vaccines was tested in the lamb model. RVFV Gn_head, coupled to LS was used and formulated with adjuvant. Whereas lambs vaccinated with coupled antigen were completely protected from the challenge virus infection, lambs vaccinated with uncoupled antigen were not. In a follow-up trial, three different MPSPs were compared side-by-side. All three MPSPs induced neutralizing antibodies and prevented fever and viremia, with no differences in efficacy. Therefore, the MPSP selection can be based only on the performance of the manufacturing process.

For SBV, LS was used as the MPSP and two candidate subunits were compared, the Gc Head (GcH) produced in cell line S2, and the Gc Head-Stalk (GcHS), produced in C1. Calves were vaccinated at day 0 and day 14 and challenged at day 35. The level of neutralizing antibodies was higher in the GcHS construct, but both constructs were equally efficient in inducing sterile protective immunity against the challenge. No viremia occurred and no viral RNA could be detected in tissue samples of calves immunized with GcH or GcHS.

The authors would like to thank all partners, scientists, post-docs, and graduate students involved in the ZAPI Project for their work and contributions, in particular: Dorota Kozub Gary Pettman, Maiken Kristiansen, Martyn Hulley, Aled Charles, Kieran Mistry, Ainul Zulkepli, Stephanie Davies, Katie Day, Sofia Ekizoglou, Peng Ke, Bradley Rawlins, Evangelia Ttofi, Laura Kerry, Katharina Mahal, Tom Witkos, Chloe Parry (AstraZeneca); Franziska Berlin, Sarah Henningsen, Benjamin Pelka, Dr. Wiebke Banz, Sandra Bauer, Dr. Michael Gaβel, Stanislav Kharchenko, Sina Pannemann, Katharina Paki, Dr. Erik Schacht, Stephanie Vielguth (BIAH BIVRC); Philippe Baudu, Dr. Natalia Bomchil, Dr. Frederic Reynard, Guillaume Rigaut, Dr. Catherine Charreyre, Britany Guillaumin, Dr. Mathieu Chevalier, Dr. Cecile Sigoillot-Claude, Marie-Line Sajous, Yves Giraud (BIAH France); Marika Vitikainen, Mari Mäkinen, Marilyn Wiebe, Markku Saloheimo (Dyadic/VTT); Nisreen Okba (EMC); Estelle Vincent (Finovatis); Abbie Charlet, Jean-Marie Préaud, Pieter Neels, Madinina Cox (IABS-EU); Mirriam Tacken, Sandra van de Water, Jet Kant, Lucien van Keulen, Michiel Harmsen, Erick Bermúdez-Méndez (WBVR); Ivy Widjaja, Frank van Kuppeveld (UU). The authors are grateful to Marc Baay and Ana Goios (both from P95 Epidemiology & Pharmacovigilance, Leuven, Belgium) for medical writing support.