Formulation development of a stable influenza recombinant neuraminidase vaccine candidate

ABSTRACT

Current influenza vaccines could be augmented by including recombinant neuraminidase (rNA) protein antigen to broaden protective immunity and improve efficacy. Toward this goal, we investigated formulation conditions to optimize rNA physicochemical stability. When rNA in sodium phosphate saline buffer (NaPBS) was frozen and thawed (F/T), the tetrameric structure transitioned from a “closed” to an “open” conformation, negatively impacting functional activity. Hydrogen deuterium exchange experiments identified differences in anchorage binding sites at the base of the open tetramer, offering a structural mechanistic explanation for the change in conformation and decreased functional activity. Change to the open configuration was triggered by the combined stresses of acidic pH and F/T. The desired closed conformation was preserved in a potassium phosphate buffer (KP), minimizing pH drop upon freezing and including 10% sucrose to control F/T stress. Stability was further evaluated in thermal stress studies where changes in conformation were readily detected by ELISA and size exclusion chromatography (SEC). Both tests were suitable indicators of stability and antigenicity and considered potential critical quality attributes (pCQAs). To understand longer-term stability, the pCQA profiles from thermally stressed rNA at 6 months were modeled to predict stability of at least 24-months at 5°C storage. In summary, a desired rNA closed tetramer was maintained by formulation selection and monitoring of pCQAs to produce a stable rNA vaccine candidate. The study highlights the importance of understanding and controlling vaccine protein structural and functional integrity.

KEYWORDS:

• Vaccine
• neuraminidase
• freezing/thawing
• recombinant
• CQA
• antigenicity
• conformational stability

Introduction

Current seasonal quadrivalent influenza vaccines are considered the best way to prevent human influenza disease and are crucial to decrease the burden of influenza illness and deaths, particularly in vulnerable populations at increased risk of serious complications.^1,2 The effectiveness of currently licensed inactivated split virus and recombinant influenza vaccines varies seasonally, and efforts to improve efficacy are ongoing. One approach is to broaden immunity and potential efficacy by introducing additional neuraminidase protein antigens. Manufacturing could also be improved by moving away from the traditional egg-based approach to a more defined cell-based system.³ To this end, we looked to develop a monovalent recombinant neuraminidase (rNA) candidate vaccine suitable for early clinical testing.

Influenza virus displays two major viral surface glycoproteins, hemagglutinin (HA) and neuraminidase (NA), considered virulence factors crucial for viral entry into and release from host cells, respectively.⁴ NA is an exo alpha sialidase hydrolyzing glycosidic linkages of terminal sialic acid residues from glycans on the host cell surface enabling release of budding viral particles from infected cells. NA is also theorized to cleave decoy receptors in mucus, facilitating infection of underlying epithelial cells.⁵ The full quaternary structure of NA is reported.⁶ The protein is composed of four identical monomers forming a complex tetrameric structure described as a “mushroom-shaped tetramer with circular symmetry.”^7–9 The tetramer is elevated by a stalk region embedded in the viral envelope. A more recent report describes tilting of the head region in relation to the stalk in the context of “whole-virion” and the possibility of additional antibody recognition sites.¹⁰

The ability to manufacture rNA protein with structural and functional integrity is crucial to produce a potent NA vaccine.¹¹ The NA protein has several functional activities that can be monitored to confirm structural and functional integrity of the recombinant antigen. One is its ability to bind Tamiflu (Oseltamivir). Tamiflu is a therapeutic anti-viral drug that binds directly within the central catalytic site of NA, blocking enzymatic activity, and Tamiflu binding can be impacted by changes in the central tetramer conformation. Another functional readout for NA protein is the ability to induce anti-NA antibodies which neutralize or block viral release from infected cells. Functional NA inhibiting (NAI) antibodies can be quantified using an enzyme-linked lectin assay (ELLA), and the ability to elicit NAI antibodies could be a potential potency indicator for candidate vaccines.^12,13 Multiple studies in humans have established an association between NAI antibodies and protection against influenza illness, as evidenced by a shorter duration of symptoms and reduced viral shedding. NAI has been confirmed as an independent correlate of protection by several authors.^14–17 The induction of NAI antibodies following vaccination can be optimally measured using an enzyme-linked lectin assay (ELLA)¹⁸ in human sera as well as in animal sera such as mice and ferrets.^19,20 In ferrets, post-vaccination NAI titers have also been linked to amelioration of influenza disease and found to be a correlate of protection.²¹ These findinds highlight the value of immunogenicity studies in animals for assessing the protective potential of neuraminidase in influenza vaccines.

There is a wealth of evidence pointing to inclusion of NA as a viable vaccine candidate. Several epidemiological and preclinical studies have been conducted and recently reviewed.⁵ However, rNA candidate vaccines have yet to be appropriately tested in the clinic.

In this paper, we describe the development of a formulation for a monovalent rNA candidate vaccine that is stable for at least 24-months at refrigerated (5°C) storage, and suitable for testing in early-phase clinical trials. We demonstrate that rNA tetramer conformation is a critical attribute impacting functional activity. Furthermore, we determine the potential mechanisms driving conformational change and define a strategy for control.

Materials and methods

Materials

Recombinant truncated NA protein from the A/Perth/16/2009 H3N2 strain of influenza virus was produced in Chinese hamster ovary (CHO) cell line. The recombinant protein was secreted and purified from clarified supernatants, yielding a homogeneous SEC profile with a major peak corresponding to the molecular size of the tetramer.

Sodium phosphate monobasic and dibasic, potassium phosphate dibasic (all ACS reagents) and monobasic (molecular grade), sodium chloride and sucrose (both BioXtra, ≥ 99.5%), were purchased from Sigma.

Freeze-thaw (F/T) studies

Studies were conducted to evaluate the effect of up to five repeated F/T cycles. Samples of rNA at 1.0 mg/mL, representative of drug substance, were aliquoted into 3-mL glass vials at 0.5 mL per vial. Vials with 0.5 mL aliquots of rNA were placed securely in a box upright in ≤ −60°C freezer overnight. The vials were thawed at room temperature for 10 to 15 min without disturbance, and this was considered one F/T cycle. The process was repeated four more times. The last thaw was completed in the analytical laboratory prior to testing.

Extrinsic fluorescence studies

Thermal ramping studies to obtain the melting temperature (Tm) of rNA were conducted by extrinsic fluorescence as described elsewhere.²² Briefly, the Mx3005p instrument (Stratagene, La Jolla, California) was employed to detect the unfolding process of rNA. The protein antigen was diluted in the different buffering conditions and folding detected by introducing the extrinsic dye SYPRO Orange (Invitrogen, Carlsbad, California). The instrument’s excitation and emission filters were set at 492 and 610 nm, respectively.

Thermal stability studies

Formulations of rNA at 360 μg/mL, representative of drug product concentration, were prepared in 10 mM potassium phosphate (KP) buffer with 10% sucrose (KP-10% sucrose), filled into 3-mL glass vials at 0.72 mL/vial and incubated for various times at temperatures of 5°C, 25°C, 37°C, and 45°C. The stability of rNA was evaluated as a function of time by measuring % closed tetramer by SEC, antigenicity by ELISA and Tamiflu-binding, as described below.

Stability prediction

The AKTS – Thermokinetics software (version 5.51, AKTS AG, Advanced Kinetics and Technology Solutions, Siders, Switzerland) was used as previously described^23,24 to identify kinetic models which best fit the rates of change in antigenicity and in the % closed tetramer measured for 5°C, 25°C, 37°C and 45°C over a six-month period. The design of studies ensured the “good modeling practices” recommended for accelerated stability predictions of bioproducts was followed.²⁵ After the model that best fits the collected data was identified by, bootstrapping analysis (resampling of data 100 times) was used to calculate the 95^th percentile prediction intervals (PI) for the selected model. Long-term predictions of loss of antigenicity by ELISA and % closed tetramer by SEC as a function of time and temperature were determined using the selected model and corresponding PIs. To assess prediction accuracy, the experimentally derived values of antigenicity and % of closed tetramer were compared to the predicted mean and its 95% PI.

Physical appearance and pH

Physical appearance and pH were recorded for every sample recovered from F/T and thermal stability studies as a routine monitor for gross abnormalities. The pH was measured using Mettler Toledo Seven Excellence Multiparameter S500 pH Meter.

Size exclusion chromatography (SEC)

Size exclusion chromatography (SEC) was performed on an ultra-high pressure low dispersion liquid chromatographic system (Agilent 1290 UHPLC or Waters H-Class UPLC) coupled with Agilent AdvanceBio SEC 200 Å, 1.9 µm, 4.6 × 300 mm column. Briefly, samples were injected onto a column held at 30°C and separated using an isocratic gradient of 10 mM sodium phosphate, 300 mM NaCl, pH 7 at 0.35 mL/min. Protein species were separated based on differences in hydrodynamic radius. The sub-2 µm particle size and narrow pore size facilitated high-resolution separation of different oligomers as well as open and closed conformational forms of rNA antigen detected by ultraviolet absorbance at 215 nm. The profile of a molecular weight protein standard mixture (e.g., Sigma 69385) was also acquired for reference. All data was reviewed in Empower software (Waters) and reported as the % closed tetramer.

Antigenicity by ELISA

Human monoclonal antibody (mAb) 1074 was isolated from a Fluzone® immunized individual and cloned for production in HEK293 cells (GeneArt GtP, ThermoFisher). Bivalent Fab (bFab) 42840 is an HRP conjugated recombinant human antibody fragment (HuCAL®, Bio-Rad), generated against rNA (N2, Singapore/INFIMH160019/2016), and produced in E. coli. Antibodies were characterized for specificity, functionality, and ability to detect heat treated NA. The antibodies were confirmed for N2 subtype specificity and ability to bind multiple strains of recombinant N2 (A/Perth/16/2009, A/Singapore/INFIMH160019/2016, A/South Australia/34/2019) and whole virus (A/Perth/16/2009)). The mAb 1074 was shown to be functional in a neuraminidase inhibition assay using a standard ELLA.^18,19 Both antibodies are conformational since the binding of the antibodies notably decreases upon heat denaturation; this was confirmed for each antibody individually by direct ELISA. As the capture mAb blocks viral rNA activity in ELLA, and both antibodies are sensitive to heat-induced conformational change, the sandwich ELISA can be considered to monitor both the conformational stability and functional integrity of rNA.

96-well microtiter plates were coated for 16 to 24-hours with anti-NA mAb 1074 (100 μL/well, 0.2 μg/mL in DPBS, Gibco) at 4°C. Subsequent steps include washing 3 times after each step (PBS; 0.1%Tween 20), 100 μL/well of each reagent, and incubations performed for 1 h at room temperature. Coated plates were blocked with assay buffer (PBS, 0.1% w/v skim milk (Difco), 0.1%Tween 20). Two independent replicates each, of reference standard (representative lot of in-house rNA) and test samples, were prepared in assay buffer. Preparations were serially diluted 3-fold for a total of eight dilution points and transferred to the blocked plate. Next bFab 42840 (0.15 μg/mL in antigen buffer) was added. Finally, 3,3′,5,5′-tetramethylbenzidine (TMB)(Sigma) was added, and the chromogenic reaction stopped after 10 minutes by the addition of 50 μL/well 2N H₂SO₄ (BDH) to wells containing TMB. Color development was quantified by measuring the absorbance of each well at a wavelength of 450 nm with a reference wavelength of 540 nm using a Spectramax plate reader (Molecular Devices). Antigenicity of test samples was calculated relative to the reference standard using SoftMax Pro (6.5.1, Molecular Devices) software for 4-PL curve with Relative Potency. The relative potency of each sample (compared to reference standard) was determined and reported in arbitrary AU/mL. The ELISA has been qualified to demonstrate accuracy, precision, linearity, and specificity for rNA.

Tamiflu-binding

The oseltamivir phosphate (Tamiflu)-NA binding assay was performed on Octet-Red96 instrument (ForteBio, Sartorius). The Octet BLI Detection System uses bio-layer interferometry (BLI) technology for real-time label-free analysis for assessing the binding interaction between rNA and Tamiflu. Tamiflu-Biotin (phospha-oseltamivir-biotin conjugate, synthesized by Life Tein) (10 μg/mL) in 1× KB buffer (containing PBS pH 7.4, 0.02% Tween-20, 0.1% albumin, 0.05% sodium azide) was first captured on a high precision streptavidin (SAX) biosensor (ForteBio, Sartorius). The interaction between rNA and Tamiflu was then monitored by dipping the biosensors into sample wells containing different dilutions of recombinant rNA (0 to100 μg/mL in 1X KB buffer) to generate a standard curve. The binding of rNA to the Tamiflu on the sensor generates a binding signal which is proportional to the concentration of rNA. Unknown samples were tested using the prepared Tamiflu-Biotin-SAX Biosensors and concentrations were calculated based on the standard curve. The Tamiflu-Biotin-SAX Biosensors were stored in 1× KB and regenerated with 0.5 M phosphoric acid after each experiment. Octet data acquisition software (Sartorius) was used for instrument control and data collection, and Octet data analysis software (Sartorius) was used for data processing.

Hydrogen-deuterium exchange mass spectrometry (HDX-MS)

The HDX-MS experiments were conducted as previously described²⁶ with minor modifications in the buffer compositions as follows. For comparison between never frozen (T0) and after 5 F/T cycles, 1 mg/mL rNA was used. For non-deuterated control, formulation buffer NaPBS (10 mM sodium phosphate buffer, 150 mM NaCl, pH 7.0) was used whereas for deuteration, a deuterated NaPBS formulation was used. For T0 and 5× F/T state comparison, five HDX-MS mixing timepoints were acquired in triplicate: 20 sec, 2 min, 10 min, 30 min, and 60 min. The quenched samples were then injected into nanoACQUITY UPLC HDX-MS module containing (1:1) pepsin/protease XIII (NovaBioAssays, MA, USA) for digestion. The digested peptides were subsequently desalted using Waters BEH C18 guard column followed by separation through ACQUITY CSH C18 analytical column. The eluted peptides were detected using Synapt GS-Si mass spectrometer (Waters) using GluFib (785.8426 m/z, doubly charged species) as a lock mass solution maintaining mass calibration of < 10 ppm.

For peptide identification, ProteinLynx Global Server (Waters) was used whereas for level of deuterium exchange analysis on identified peptides, DynamX software (Waters) was used. Only peptides with high spectral quality were considered. The summed difference over all timepoints were used for comparison evaluations.

Since there was no available structure for this specific NA from A/Perth/16/2009 H3N2 strain, a homology model was generated using PHYRE2 server.²⁷ The resulting homology structure was overlaid with three-dimensional structure of N1 Neuraminidase (PDB: 2HTY, H5N1)²⁸ and the structures were superimposable. The HDX-MS data was then mapped onto the homology structure as shown in Figure 2.

Mouse immunogenicity

Ethics

Animal experiments were performed in compliance with the Public Health Service Policy on Humane Care and Use of Laboratory Animals²⁹ and the Guide for the Care and Use of Laboratory Animals³⁰ and were conducted with approved animal protocols from the Sanofi Institutional Animal Care and Use Committee.

Mouse studies

The method used a well-established immunogenicity model previously reported.¹² Briefly, BALB/c mice (Mus musculus, females, 6 to 8 weeks) purchased from Charles River Laboratories (Wilmington, MA, USA) were housed in microisolator units. Mice (six per group, as determined by power analysis for 85% power) were vaccinated with rNA in the presence or absence of squalene-based oil-in-water emulsion adjuvant, AF03 (Sanofi, 5% squalene). The rNA from samples recovered from thermal stress studies was diluted in NaPBS or KP-10% to 40 μg/mL and then mixed with AF03 adjuvant at 1:1 ratio (25 µL vaccine:25 µL AF03) resulting in a final concentration of 2.5% squalene. Mixture was administered via intramuscular injection in 50 µL volume (1 μg rNA/dose/animal) on day 0 (D0) and then boosted with the same dose on day 21 (D21). Blood was collected on day 35 by cardiac puncture into serum tube containing clot activator and separator gel (Sarstedt, Nümbrecht, Germany) and tested for presence of NA-specific antibodies by ELLA.

Enzyme-linked lectin assay (ELLA)

Mouse sera from the rNA-immunized groups were tested for their ability to inhibit the activity of the viral NA using a standard ELLA as described previously.^18,19 Briefly, 2-fold serial dilutions of heat-inactivated sera (starting dilution of 1:20) were incubated with H6N2 A/Perth/16/2009 virus at a pre-determined concentration of virus to give 70% maximal NA activity for 30 min at 37°C in 2-(N-morpholino) ethanesulfonic acid (MES) (Alfa Aesar, Tewksbury, MA, USA) buffer supplemented with 1% bovine serum albumin (BSA), 20 mM CaCl₂ and 0.5% Tween20. Dilutions were added to fetuin (Sigma-Aldrich, St. Louis, MS, USA, 25 µg/mL)-coated plates (96-well MaxiSorp plates [Nunc]) and incubated for 16 to 18 hours at 37°C. Next plates were incubated with horseradish peroxidase (HRP)-coupled peanut agglutinin (PNA, Sigma-Aldrich, 2.5 µg/mL) at room temperature for 2 h in the dark and developed using o-phenylenediamine dihydrochloride (OPD, Sigma-Aldrich). The reaction was stopped by adding 1 N sulfuric acid and immediately read on BioTek plate reader at an optical density of 485 nm. The IC50 was calculated by non-linear regression analysis using asymmetrical sigmoidal 5PL curve (GraphPad Software, San Diego, CA, USA). IC50 is defined as the reciprocal of the serum dilution, resulting in 50% NA inhibition (NAI) titer.

Results

Freeze/thaw (F/T) triggers rNA conformational change in NaPBS buffer

Frozen vaccine storage accelerates timelines during early phases of clinical development and circumvents potential stability issues that may occur at higher temperatures. As knowledge accumulates, it may be possible to change to non-frozen storage at a later stage. The starting formulation of purified rNA was prepared in sodium phosphate buffered saline (NaPBS) to match the buffering conditions of other influenza recombinant proteins³¹ and was stored frozen at ≤ −60°C.

To investigate rNA robustness against F/T stress, five rounds of F/T were conducted with rNA in NaPBS buffer. After each F/T cycle physicochemical parameters were monitored to assess the impact of F/T on the structure, stability, and function of rNA.

There were no changes in physical appearance or pH, after several rounds of F/T, however, a change in rNA tetramer conformation was observed by high-resolution SEC. The SEC chromatogram of rNA showed an earlier eluting peak compared to the unfrozen, liquid formulation control (Figure 1a). An earlier eluting peak by SEC is consistent with a larger hydrodynamic radius and the previously reported “open tetramer” conformation.^32,33 A minor later-eluting population at ~6.6 min was also observed after 3 F/T and 5 F/T cycles, which exhibited a smaller hydrodynamic radius and a retention time consistent with a dimer (Supplemental Figure S4).

Figure 1. Freeze thaw (F/T) induced rNA conformational changes that impact physiochemical properties and functional activities. The rNA in NaPBS was subjected to 1, 3 and 5 F/T cycles and the impact was assessed by SEC, extrinsic fluorescence, and Tamiflu binding. (a) SEC profiles show a change from ‘closed’ to ‘open’ tetramer conformation after F/T stress. Changes in physiochemical properties were also observed in terms of (b) % Closed tetramer by SEC (n = 2), (c) Reduction in melting temperature Tm (n = 2) and (d) Reduced functional activity in terms of Tamiflu-binding (n = 3).

The peak area of the non-stressed rNA (Figure 1a) was integrated and expressed as % “closed tetramer” as a function of F/T cycles. A 90% total decline in % closed tetramer was detected and reached plateau after 3 F/T cycles (Figure 1b). The change from closed to open tetramer trended with a decrease in the conformational stability in terms of Tm (Figure 1c) and functionality, with a more gradual change over the 5 F/T cycles, by Tamiflu-binding (Figure 1d). The total decrease in Tamiflu-binding indicated the open rNA conformation has a reduced ability to bind Tamiflu.

Together, the results demonstrate the rNA tetramer can exist either as a closed or open configuration. The open configuration is less stable based on Tm and has reduced functionality based on Tamiflu binding. Therefore, maintaining and controlling the closed tetrameric conformation were considered desirable.

Hydrogen-Deuterium Exchange Mass Spectrometry (HDX-MS) was used to understand how F/T-induced conformational change impacts protein dynamics. Based on the HDX-MS heatmap analysis, the sequence coverage was approximately 70%, with missing regions primarily at the N-terminus and N-glycosylation sites (Supplemental Figure S1).

There were increases in deuterium uptake in residues 60–69, 88–97, and 393–398 for the sample that underwent 5 F/T cycles in NaPBS buffer (Supplemental Figure S1), indicative of increased solvent exposure in these regions. These residues contribute to the tetramer interface region as shown in Figure 2. The solvent exposure of these residues indicates a disruption of the noncovalent interactions between the protomers and at potential anchorage binding sites at the base of the rNA tetramer. Such disruption in anchor binding aligns with the change in SEC profile (Figure 1a, b) and offers a possible trigger for the tetramer to transition into the open conformation. The change is in the known “150 loop” region which also undergoes conformational change during substrate binding and known to be associated with open/closed conformation^33,34 suggesting a potential mechanism for impact on functional activities.

Figure 2. HDX-MS data of 5X FT rNA in NaPBS buffer mapped onto three-dimensional structure. Highlighted region (red) represents peptides that showed increase in deuterium uptake and is also known as the 150 loop which undergoes conformational change during substrate binding.

KP buffer preserves the desired rNA closed conformation upon F/T

Although protein drug substances are frequently stored under frozen conditions to improve stability, they can also undergo freezing-induced conformational changes leading to losses in biological activity. During freezing and thawing, changes in the physical environment such as freeze concentration, formation of ice crystals and buffer crystallization can lead to dramatic changes in the stability of the protein.³⁵

We hypothesized that the observed rNA instability after F/T cycling in NaPBS resulted from crystallization of the dibasic component of the NaPBS, leading to a pH-induced conformational change.³⁶

It was reasoned that switching to KP buffer where pH is minimally impacted by freezing³⁶ would protect against pH decline and conserve the desired closed tetramer. We investigated the impact of F/T on rNA prepared in KP buffer, pH 7.4. In this buffer, the desired closed tetramer conformation, conformational stability in terms of melting temperature, and Tamiflu-binding were all preserved, even after 5 F/T cycles (Figure 3a–c). Thus, a simple change in the buffer compositions from sodium salt to potassium salt was sufficient to prevent rNA instability during freezing and thawing. Having demonstrated KP buffer performed as reported elsewhere,³⁶ screening additional buffers was deemed unnecessary.

Figure 3. The freeze thaw-induced rNA conformational change is abolished when switching from NaPBS to KP buffer. The rNA in either KP buffer pH 7.4 or NaPBS pH 7.0 underwent 5X F/T cycles and (a) % Closed tetramer (n = 2), (b) Melting temperature Tm (n = 2), and (c) Tamiflu Binding (n = 4) were measured. Error bars represent the standard deviation from the mean.

Acidic pH is necessary but not sufficient to trigger conformational change

If F/T-induced pH reduction in NaPBS was the only cause of rNA conformational change, decreasing the pH at ambient temperature should have the same effect as freezing the protein in NaPBS. To emulate crystallization of the dibasic salt, a formulation of rNA in monobasic sodium phosphate in saline (10 mM NaH₂PO₄, 150 mM NaCl) (monoNaPS) was prepared. This formulation yielded a pH of 4.6, agreeing with the reported pH of frozen NaPBS.³⁶ The lower pH 4.6 formulation simulates the conditions where dibasic salt crystalizes at low temperature to leave the monobasic salt at a reduced pH. The formulations at ambient temperature were monitored for % closed tetramer by SEC. Unexpectedly, the monoNaPS formulation was able to maintain the rNA closed tetramer conformation despite having a reduced pH of 4.6 (Figure 4 – red bar). This finding prompted further investigation into whether a combination of low pH followed by F/T (indicated as blue bars) was responsible for inducing the shift from closed to open conformation. To do this, the same monoNaPS formulation at pH 4.6 was prepared but now an additional F/T step was included. Under these conditions, a change from closed to open conformation, as previously detected with frozen NaPBS buffer pH 7.0, was observed (Figure 4 – also see chromatograms in Supplemental Figure S2). Thus, in addition to acidic pH-induced stress there is a second F/T stress, and it is the combination of these stresses that ultimately drives rNA conformational change.

Figure 4. Acidic pH is necessary but not sufficient to trigger conformational change and closed tetramer is stabilized in KP-10% sucrose. Four rNA formulations in different buffers were analyzed (vertical axis). The % closed tetramer by SEC was monitored at ambient temperature, before (red bars) and after (blue bars) a single F/T cycle. Closed tetramer loss occurred following F/T in NaPBS pH 7.0 and monoNaPS pH 4.6. Tetramer was conserved in formulations with 10% sucrose.

Inclusion of 10% sucrose to optimize osmolality and stability

Since F/T was identified as a significant stress, it was hypothesized that inclusion of sucrose as a cryoprotectant³⁷ might contribute to preserving the rNA closed conformation. As hypothesized, inclusion of 10% sucrose to NaPBS was able to protect the desired closed tetramer against F/T stress (Figure 4 – blue bar). When switching from NaPBS to KP buffer, NaCl was specifically avoided to prevent the formation of the dibasic and monobasic phosphate salts. However, due to the lack of saline, the KP osmolality measured was only 23 mOSm/kg which is not compatible with parenteral administration. The inclusion of 10% sucrose restores a near physiological osmolality of 378 mOSm/kg. Therefore, KP with 10% sucrose (KP-10% sucrose) was considered the optimal rNA buffer (Figure 4). Finally, using extrinsic fluorescence methodology²² we demonstrated that inclusion of 10% sucrose in KP increased the Tm of rNA from 52.25°C to 53.25°C (Supplemental Figure S3). These results indicate sucrose not only prevents F/T-induced conformational changes but also improves the thermostability of rNA, which is paramount for longer-term stability. Sucrose was chosen as the preferred cryoprotectant based on its well-recognized safety profile as an excipient in humans and its extensive use across the vaccine industry.

In summary, we show the rNA tetramer conformation is sensitive to both pH drop as NaPBS freezes, as well as to F/T stress. HDX-MS results indicate the regions impacted by F/T are at the base of the tetramer (Figure 2), offering a potential structural basis for the way declining pH and F/T stress causes tetramer opening. The rNA tetramer in the desired closed position can be maintained by switching to a KP buffer where pH remains stable upon freezing. In this situation, functional attributes are preserved after five rounds of F/T (Figure 3). To optimize physiological osmolality, improve thermal stability and protect against cold denaturation, KP buffer was supplemented with 10% sucrose (Figure 4).

Assessment of rNA stability indicating assays

To evaluate stability indicating capability of assays and understand behavior of rNA when exposed to elevated temperatures, thermal forced degradation studies were initiated. Samples of rNA in KP-10% sucrose were exposed to increasing temperatures (5°C, 37°C, and 45°C) over incubation periods of up to 3-months. Testing included physical appearance and pH, % closed tetramer by SEC, and Tamiflu-binding. In addition, we introduced an antigenicity ELISA. In developing the antigenicity assay, specific mAbs were selected according to their ability to monitor conformational stability and functional integrity of rNA and as such the assay is considered a potentially critical attribute. We also used a well-developed mouse immunogenicity model¹² to link in vitro assays to an in vivo biological function.

Samples collected after 1 week at 45°C and 3-months at 5°C and 37°C were tested for % closed tetramer by SEC, Tamiflu-binding, and antigenicity by ELISA (Figure 5a–c). Samples at the same timepoints were also tested in a mouse immunogenicity study (Figure 5d), to confirm the induction of functional NAI antibodies after prime and boost immunization with rNA stored under different temperature conditions. Mouse sera were analyzed in in vitro ELLA assay to quantitate antibodies able to inhibit Neu activity of the recombinant influenza virus. Higher Geometric Mean Titers (GMT) of NAI antibodies are induced by properly folded and functional protein, while low or undetectable GMTs are induced by poorly folded or degraded protein. A significant 10-fold reduction in NAI titers was observed in a group of mice immunized with rNA incubated at 37°C (GMT of 268) in comparison to a standard 5°C storage (GMT of 2308). The highest 45°C exposure resulted in complete abrogation of NAI antibody titers (GMT of 7), which was significantly lower than GMTs of 37°C-immunized group, confirming stability studies (Figure 5d).

Figure 5. Antigenicity and % closed tetramer are potential critical quality attributes (pCQAs). Accelerated stability studies were initiated using rNA in KP-10% sucrose buffer. Samples from 3 months timepoint at 5°C and 37°C incubation and 1 week at 45°C were tested for (a) % closed tetramer by SEC (n = 2), (b) Tamiflu-binding (n = 1), (c) Antigenicity by ELISA (n = 2), and (d) immunogenicity, with symbols indicating NAI titer of individual animals, bars represent geometric mean titer (GMT) with 95% CI of 5–6 animals/group (* p < 0.05, Tukey’s multiple comparison test). All measurements decreased with increasing temperature and % closed tetramer by SEC and antigenicity by ELISA were considered pCQAs.

The SEC profile after 3-months at 37°C showed a reduction in closed tetramer from 100% to below 30% with a proportional increase in open tetramer population, and no change in main component rNA content based on UV absorbance in the SEC chromatogram (Supplemental Figure S4). The closed tetramer population was completely absent after 1 week at 45°C, having transitioned to open tetramer and smaller dimer species (Supplemental Figure S4). The SEC profiles were directly comparable to those obtained after F/T (Figure 1a). Together the data demonstrate the open tetramer has reduced functionality compared to the closed tetramer.

In conclusion, all tests were stability indicating and aligned directly with a significant change in immunogenicity. Changes in tetramer with loss of % closed tetramer trend directly with reduced function. Owing to the importance of detecting and monitoring structural changes, the SEC was considered a potential critical quality attribute (pCQA), according to quality by design (QBD) methodology.³⁸ Likewise, the antigenicity ELISA is linked indirectly to vaccine potency based on use of functional mAb as well as being aligned with changing functional immunogenicity and can therefore be considered an orthogonal pCQA. It was decided that the antigenicity ELISA could replace Tamiflu-binding to reduce unnecessary duplication and to be more aligned with immunogenicity.

Stability modelling using pCQAs accurately predicts 24-months stability at 5°C storage

Once the stability indicating assays and pCQAs were established, attention shifted to understanding longer-term stability of the vaccine candidate. Thermal stress studies were initiated to verify pCQAs, gain insight into rNA protein thermal stability, and to assess the possibility of moving from frozen to refrigerated storage conditions. A minimum target shelf-life of 24-months at 5°C was considered sufficient to support early phase clinical trials.

A thermal stress study was initiated to monitor identified pCQAs over a 6-month period. The temperatures and timepoints were specifically selected based on previous studies²⁴ to allow mathematical prediction modeling software to extrapolate stability and predict a shelf-life for the rNA vaccine. Making an early-stage shelf-life prediction would de-risk planning for subsequent, larger, more pivotal clinical trials. The study was designed to generate at least 44 data points across multiple informative temperatures over 6-months of varied temperature incubations.

The actual data for the two pCQAs are shown in Figure 6a, b. The SEC and antigenicity ELISA data trends were closely aligned, showing strong correlation at the informative and relevant 37°C forced degradation temperature (n = 11, r² = 0.932) (Supplemental Figure S5). AKTS analysis based on 6-months data predicts a target stability of at least 24-months at refrigerated storage as shown in Figure 6a, b (and Supplemental Table S6).

Figure 6. The rNA vaccine in KP-10% sucrose is predicted and verified to be stable for at least 24 months at 2–8°C. AKTS prediction software used 6 months accelerated stability data for (a) % Closed tetramer by SEC and (b) Antigenicity by ELISA to predict rNA stability for at least 24 months at 2–8°C. Stability data points up to 6 months were used for prediction, marked as filled circles from 5°C (black), 25°C (red), 37°C (blue), and 45°C (green). Solid line is the prediction mean; broken lines are the upper and lower 95% prediction intervals (PI). Real time data at 5°C on day 270 (9 M), 365 (12 M), 540 (18 M), and 730 (24 M) are bold open circles, showing no significant upward or downward trend, aligning with the prediction mean, and well within the PI for both assays.

To validate the prediction, real-time stability timepoints from an earlier stability study where samples were treated for up to 24-months at 5°C were assessed. The results, shown as open circles in Figure 6a, b, confirm that the rNA vaccine candidate remains stable for at least 24-months incubation at 5°C, as predicted by AKTS analysis.

Discussion

Introducing rNA protein as a candidate vaccine has challenges related to its complex structure, maintenance of longer-term stability and for combination as a potential multivalent vaccine. In this paper, we describe the development of a rNA monovalent protein vaccine candidate that is stable at refrigerated temperature for at least 24-months. We highlight the critical need to maintain a desired closed tetramer conformation and describe potential mechanisms driving change to the less-functional open form. We describe two distinct pCQAs that may be applied to monitor and control vaccine integrity throughout the vaccine lifecycle.

The quaternary structure of rNA is complex and not readily maintained given the degree of conformational plasticity described.¹⁰ Open/closed tetramer configurations have been described³² and recently, impressive cryo-electron microscopy (cryo-EM) images have given visual perspective to the open/closed structures.³² Stabilization of the closed tetramer conformation has shown improved thermal stability and enhanced binding affinity of protective antibodies.³² In the current study, we further evaluate the stresses driving closed to open tetramer changes and define ways to control and maintain the desired closed configuration, with a view to vaccine development.

Buffer selection is a key component of vaccine formulation. The use of NaPBS is often the first buffer of choice to maintain optimal physiological balance. However, when a protein prepared in NaPBS is frozen there are additional physiochemical stresses to consider. The dibasic sodium phosphate salt crystalizes before the monobasic sodium phosphate leading to a temporarily reduced buffering capacity and causing a pH to “dramatically decrease” from pH 7.0 to as low as pH 3.8 at lower freezing temperatures.³⁶

Our original hypothesis, that open tetramer change is triggered by decline in pH as NaPBS freezes, was only partially correct since additional F/T stress was also necessary to induce this change. After F/T, tetramer in KP buffer pH 7.4 remained closed (Figure 3), while tetramer in monoNaPBS with simulated lower pH 4.6 only caused tetramer opening after F/T (Figure 4). Thus, the combined stresses of pH reduction and F/T were required to trigger tetramer opening. A possible explanation is that acidic pH favors cold-induced conformational change of the rNA. The ability of acidic pH to compound F/T stress has been reported³⁵ giving rationale for the finding.

We show that the closed tetramer can be stabilized by controlling either pH or freeze/thaw independently. Thus, KP buffer, known for pH stability upon freezing,³⁶ preserves the closed conformation despite multiple rounds of F/T. Conversely, controlling F/T stress by including 10% sucrose as a cryoprotectant in NaPBS was also able to partially preserve the closed tetramer despite a well-known pH decline in this buffer system.³⁶ Controlling both stresses is expected to offer optimal closed tetramer stability, i.e., solution of rNA in KP buffer to stabilize pH and inclusion of sucrose as a cryoprotectant to control freeze/thaw stress. Lastly, NaCl was not included in KP buffer to avoid the formation of sodium salt. The addition of 10% sucrose serves a triple function: improving the conformational stability, maintaining isotonicity and lastly, acting as a cryoprotectant to prevent cold denaturation of rNA. Our study shows that individual stresses were unexpectedly compounded and highlights the need to be vigilant when choosing a buffer for frozen storage.

To better understand the closed/open tetramer change, HDX-MS was used to identify the region impacted by F/T in NaPBS. Differences in dynamics were observed in sequences at the tetramer interface, including the ‘150 loop,’ previously known to undergo conformational changes during substrate binding.^33,34 Indeed, it has been previously reported that improved atomic packing in the “interface-proximal” region is crucial for NA stability and maintenance of the closed tetrameric conformation.³² This conformational change was previously reported by cryo-EM and appears to be consistent with loss of rNA functionality.³² It is likely that destabilization at the tetramer interface causes the tetramers to shift outward, forming an open conformation. One possible explanation could be destabilization of the noncovalent (polar and hydrophobic) interactions at the interface due to increasing acidity, resulting in repulsion among monomers. It is conceivable that the central binding region within the tetramer is broadened enough to result in the observed reduction in Tamiflu binding. Additionally, there is a loss of antigenicity shown by ELISA, and a decrease in the functional antibody titers in animals. This loss in antigenicity and immunogenicity correlates with the observed loss of the closed tetramer structure, suggesting that the closed tetramer configuration is key for inducing functional antibodies.

Having developed a sound rationale for stabilizing rNA following F/T, studies progressed to an assessment of thermal stress and potential for longer-term storage stability. Forced degradation studies using aggressive incubation temperatures were conducted to verify stability indicating assays. For these studies, in vivo immunogenicity and an antigenicity ELISA were introduced. The ELISA uses two NA-specific mAbs, one of which can inhibit the enzymatic activity of NA, indicating that it binds an epitope important to NA function. Since the ability of NA to induce NAI antibodies is considered indicative of potency,¹³ and ELISA is monitoring a functional epitope, this assay has the potential to be a surrogate for vaccine potency. Under forced degradation conditions the decrease in % closed tetramer, as monitored by SEC, correlated closely with loss of antigenicity by ELISA. It is therefore likely that the functional site bound by the mAb is disrupted as the tetramer conformation changes from the closed to open configuration.

The SEC detected changes in tetramer conformation which trended with reduced thermal stability, antigenicity, and the ability to induce neutralizing antibodies in an in vivo animal immunogenicity model.¹² The in vitro antigenicity ELISA and SEC tests can therefore be considered critical to confirm quality of the vaccine. These can be referred to as pCQAs, according to the principals of quality by design (QbD), a methodology where quality is built into the vaccine manufacturing process at the earliest stages of development.³⁸ The pCQAs can be considered complementary, orthogonal methods, that monitor different aspects of the same characteristic, giving added assurance of vaccine integrity and performance.

Understanding longer-term storage capability of a vaccine as early as possible can greatly support clinical trial site logistics and planning. Verification of stability was also important since formulation development moved forward in the absence of calcium ion (Ca²⁺), which is reported to enhance NA thermal stability.^6,32 However, Ca²⁺ was specifically avoided to give more flexibility and prevent complications with calcium phosphate precipitation in NaPBS. Longer-term stability was an accepted initial risk. Rather than wait for lengthy real-time stability studies, mathematical predictive modeling software^23,24 was employed to gain early understanding and increased confidence in the potential to use liquid refrigerated storage.

Forced degradation thermal stability studies were designed to maximize predictive capability based on pCQA monitoring over 6-months. The model calculated that rNA in KP-10% sucrose remains stable for at least 24-months at refrigerated temperature. The prediction was the same regardless of which pCQA was being evaluated, giving added assurance both for the pCQAs and prediction. The prediction at 6-months was later confirmed with non-predicted real-time data at 24- months, validating the prediction and demonstrating Ca²⁺ is not a critical requirement in the current vaccine formulation. The study adds to other reports^24,39–41 giving confidence in predictive modeling to de-risk stability as early as possible in development.

In conclusion, our study shows the importance of monitoring rNA conformation in relation to different storage conditions, longer-term stability, and vaccine development. Furthermore, the study demonstrates a link between structural integrity and function and identified pCQAs for monitoring these parameters. The results may have implications for development of other vaccine candidates, where heterogeneity in a complex protein structure can significantly impact vaccine potency. The rNA formulation described herein is considered a suitable vaccine candidate for future clinical trial development.

Authors’ contribution

B.L., N.R., R.H.B., and S.F.A. were responsible for overall design of this study. S.Z. and D.A.J. carried out HDX-MS data analysis, and interpretation. L.S. was responsible for SEC data analysis and interpretation. M.M. was responsible for ELISA data analysis and interpretation. I.V.U. and G. C-G were responsible for the immunogenicity studies, data analysis and interpretation. M.H. and K.Y.Y. F were responsible for the AKTS design, data analysis and interpretation. R.H.B., B.L. and SFA wrote the paper and contributed to paper preparation. All authors contributed to and reviewed and gave approval to the final version of the paper.

Acknowledgments

We acknowledge many scientific and technical staff who supported testing at different locations including, Shilin Cheung and Cindy Xin Li for early recognition of open/closed tetramer in Toronto, Avanti Karkhanis and Jin Su for antigenicity ELISA development in Toronto, Marie-Julie Diana and Romain Pizzato who supported antibody characterization in Marcy l’Etoile, Timothy Farrell and Clint McDaniel who supported mouse immunogenicity studies and performed ELLA in Cambridge, Massachusetts. Finally, Swetha Vengadesh for editorial support.

Disclosure statement

This work was funded by Sanofi. At the time of preparation all authors were employees of Sanofi and may hold shares and/or stock options in the company.

Supplementary data

Supplemental data for this article can be accessed on the publisher’s website at https://doi.org/10.1080/21645515.2024.2304393.

Additional information

Funding

The work was funded by Sanofi.