Accelerating Vaccine Formulation Development Using Design of Experiment Stability Studies

Journal of Pharmaceutical Sciences xxx (2016) 1-11

Contents lists available at ScienceDirect

Journal of Pharmaceutical Sciences journal homepage: www.jpharmsci.org

Pharmaceutical Biotechnology

Accelerating Vaccine Formulation Development Using Design of Experiment Stability Studies Patrick L. Ahl 1, 2, *, Christopher Mensch 1, 2, Binghua Hu 1, 2, Heidi Pixley 1, 2, Lan Zhang 2, 3, Lance Dieter 1, 2, Ryann Russell 1, 2, William J. Smith 1, 2, Craig Przysiecki 2, 3, Mike Kosinski 1, 2, Jeffrey T. Blue 1, 2 1 2 3

Vaccine Bioprocess Research and Development, Merck Research Laboratories, Merck & Co., Inc., West Point, Pennsylvania 19486 Merck Research Laboratories, Merck & Co., Inc., Kenilworth, New Jersey 07033 Infectious Diseases and Vaccines Discovery, Merck Research Laboratories, Merck & Co., Inc., West Point, Pennsylvania 19486

a r t i c l e i n f o

a b s t r a c t

Article history: Received 12 April 2016 Revised 20 May 2016 Accepted 9 June 2016

Vaccine drug product thermal stability often depends on formulation input factors and how they interact. Scientific understanding and professional experience typically allows vaccine formulators to accurately predict the thermal stability output based on formulation input factors such as pH, ionic strength, and excipients. Thermal stability predictions, however, are not enough for regulators. Stability claims must be supported by experimental data. The Quality by Design approach of Design of Experiment (DoE) is well suited to describe formulation outputs such as thermal stability in terms of formulation input factors. A DoE approach particularly at elevated temperatures that induce accelerated degradation can provide empirical understanding of how vaccine formulation input factors and interactions affect vaccine stability output performance. This is possible even when clear scientific understanding of
particular formulation stability mechanisms are lacking. A DoE approach was used in an accelerated 37 C stability study of an aluminum adjuvant Neisseria meningitidis serogroup B vaccine. Formulation stability differences were identified after only 15 days into the study. We believe this study demonstrates the power of combining DoE methodology with accelerated stress stability studies to accelerate and improve vaccine formulation development programs particularly during the preformulation stage. © 2016 Published by Elsevier Inc. on behalf of the American Pharmacists Association.

Keywords: analytical biochemistry biotechnology stability vaccines vaccine adjuvants excipients

Introduction Successful vaccine development requires understanding of public health needs, disease pathology, and immunology. The overall benefits of vaccination are undeniable, but vaccine development is difficult, time consuming, and costly.1 Vaccine developers are continuously searching for ways to accelerate vaccine development. Identification of appropriate antigens and adjuvants Abbreviations used: AAHS, amorphous aluminum hydroxyphosphate sulfate adjuvant; AlPO4, aluminum hydroxyphosphate adjuvant; DoE, Design of Experiment; DSC, differential scanning calorimetry; ELISA, enzyme-linked immunosorbent assay; fHbpv1, factor H binding protein version 1; LOS, lipooligosaccharide; MenB, N. meningitidis serogroup B; 4-MUP, 4-methylumbelliferyl phosphate; NadA, Neisseria adhesion A; OMV, outer membrane vesicles; PBS, phosphate-buffered saline; Tm, thermal transition temperature midpoint; TRIS, Trizma base; WFI, sterile water for injection; WRAIR, Walter Reed Army Institute of Research. This article contains supplementary material available from the authors by request or via the Internet at http://dx.doi.org/10.1016/j.xphs.2016.06.014. * Correspondence to: Patrick L. Ahl (Telephone: 215-652-5534; Fax: 215-6525299). E-mail address: [email protected] (P.L. Ahl).

using animal studies is very time consuming and is often misleading.2 Automated high-throughput procedures are now being used to rapidly screen the immune responses of antigen and adjuvant.3 Acceleration of vaccine drug product formulation development is another urgent need within vaccine development. Numerous vaccine formulations must be screened to identify the formulation which provides the optimal efficacy, safety, and stability. Often, it is the formulation choices that are most important in determining vaccine drug product stability. Vaccine developers have automated several bioanalytical methods to monitor the stability of vaccine formulations.4,5 Successful optimization of vaccine formulation stability must typically overcome multiple challenges particularly in the early phases of development. These challenges often include (1) short formulation development timelines, (2) limited supplies of critical antigens and adjuvant, and (3) a wide range of available formulation conditions that should be examined. We have used a 2 formulation development concepts to overcome these 3 challenges. The short timeline challenge was overcome by conducting relatively short “stress testing” stability studies at a temperature

http://dx.doi.org/10.1016/j.xphs.2016.06.014 0022-3549/© 2016 Published by Elsevier Inc. on behalf of the American Pharmacists Association.

2

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11

slightly above the targeted storage conditions.6 We found that a short-term stress stability study, that is, an accelerated stability study, can more quickly provide valuable formulation information. Challenges 2 and 3 can be overcome by the use of half-factorial or quarter-factorial Design of Experiment (DoE) designs with small sample quantities in early formulation development studies.7,8 In this particular study, however, we used a full-factorial design. We found that a carefully planned DoE stability study can be used to build useful statistically significant mathematical models of formulation responses to key formulation input factors. A formulation screen of N. meningitidis serogroup B (MenB) outer membrane vesicles (OMVs) vaccines is used as an example in this report to demonstrate the advantage of the accelerated thermal stress DoE stability study approach for formulation development. Bacterial meningitidis caused by N. meningitidis is a devastating disease that can kill patients within days of infection despite the use of modern antibiotics. Currently, 2 MenB vaccines are on the market. The GSK MenB vaccine Bexsero® is approved in Europe, Canada, Australia, and United States, whereas the Pfizer MenB vaccine Trumenba® is approved in the United States.9,10 N. meningitidis is coated by a protein and lipooligosaccharide (LOS)-rich outer membrane which can bud off from the bacteria.11 Several methods are available to extract and purify OMVs from the bacteria and OMV isolated from the bacteria are 100-200 nm in diameter vesicles.12,13 Endemic protection to N. meningitidis serogroup B invasive disease is possible by using 2 relatively conserved OMV membrane associated proteins: (1) Neisseria adhesion A (NadA) and (2) factor H binding protein version 1 (fHbpv1).14 NadA is a major OMV surface antigen that appears to be important for N. meningitidis binding to host epithelial cells. The OMV antigen fHbpv1 is not embedded in the membrane bilayer. The molecule is anchored in the membrane by single fatty acid attached at the lipidated N-terminus.15 Recently, researchers at the Walter Reed Army Institute of Research developed a vaccine that contained 3 MenB OMVs from antigenically diverse strains of the bacteria.16,17 The LOS toxicity of the 3 OMVs was significantly reduced by disabling genes in all 3 OMV strains that were involved in the synthesis of LOS.16 Vaccines prepared with these 3 particular OMV sources have been clinically evaluated.18 Additional genetic modifications were performed at Merck with these 3 MenB strains to further increase the expression of NadA, fHbpv1, and other antigens to further optimize MenB strain protection.19 The MenB vaccines containing these 3 distinct OMVs with enhanced expression of NadA and fHbpv1 have potential to be both safe and effective to provide broad protection against most MenB bacterial strains. DoE is a key element of the Quality by Design approach recommended by the FDA to insure process understanding during pharmaceutical development.20 DoE methods are routinely used in pharmaceutical manufacturing and during the pharmaceutical development of small molecules and devices. Currently, however, the DoE approach is not extensively used in vaccine formulation development with some exceptions.21 The development of vaccine formulation is typically considered too complex for a simple DoE
approach. In addition, vaccine stability studies done at 4 C often require months or years for formulation related stability differences to be statistically reliable. Fortunately, formulation parameters that control thermal stability can often be identified in days or weeks by
higher temperature thermal stress studies at 37 C or higher.22 DoE methodology combined with thermal stress stability studies below the antigen unfolding Tm can often provide very predictive information about lower temperature formulation stability. Particularly, if the antigen degradation process basically follows Arrhenius kinetics, for example, simple protein unfolding. Thermal stress

stability studies may not be as predictive for antigen degradation processes that are non-Arrhenius such as protein aggregation.23 In any case, short-term high thermal stress DoE studies cannot unambiguously identify the optimally stable formulation for a vaccine. Longer stability studies at the targeted vaccine drug
product storage temperatures, for example, 4 C, will be required for regulatory approval. However, thermal stress DoE stability studies can rapidly provide critical formulation information during early vaccine formulation development which can significantly reduce formulation development timelines. This report will illustrate this approach to vaccine formulation
development by showing how a 37 C DoE stress stability study of 17 aluminum adjuvant containing vaccine formulations for N. meningitidis serogroup B OMV was used to quickly determine which formulation input factors are most important for the MenB vaccine formulation thermal stability. Of course, other vaccine formulation stresses, for example, light, freezing, oxidation, and agitation, can also be examined by a DoE stability study approach.
The temperature of 37 C was chosen because it is well below the differential scanning calorimetry (DSC) unfolding temperature of the antigens examined in the study. Although statistically significant stability effects were identified after only 15 days, the stability study was extended out to 77 days to confirm the 15-day stability trends. The following 4 formulation input factors were considered in this vaccine formulation DoE study: (1) pH, (2) aluminum adjuvant choice, (3) OMV to aluminum adjuvant (w/w) input ratio, and (4) added phosphate concentration. Two DoE formulation
output responses were measured during the course of the 37 C study to access stability: (1) the normalized change in the NadA enzyme-linked immunosorbent assay (ELISA) response ( D%NadA ) and (2) the normalized change in the fHbpv1 ELISA response (D%fHbpv1 ). The ELISA response changes of these 2 major OMV protein antigens, NadA and fHbpv1, were particularly important because these assays correlated well with mouse in vivo NadA and fHbpv1 immunogenicity results from thermally stressed MenB OMV vaccines. Thus, we consider the NadA and fHbpv1 ELISA response assays as “stability indicating” for this particular vaccine.
The MenB vaccine formulation DoE conclusions from the 37 C stress stability study were also found to be consistent with biophysical characterization of the antigens and results from a sepa
rate 11-month 4 C stability screen. This consistency demonstrates the value in the thermal stress DoE stability study approach for vaccine formulation development. Materials and Methods Materials The N. meningitidis OMVs used in this study were supplied by the Merck Vaccine Bioprocess Department as described in the following. The amorphous aluminum hydroxyphosphate sulfate adjuvant (AAHS) was obtained from Merck Manufacturing, whereas the aluminum hydroxyphosphate adjuvant (AlPO4) was supplied by the Merck Vaccine Formulation Department.24 The monoclonal antibodies, streptavidin-conjugated alkaline phosphatase and 4-methylumbelliferyl phosphate (4-MUP) used in the NadA and fHbpv1 ELISA assays, were from the Merck Vaccine Analytical Department. Mouse serum measurements to determine anti-NadA, anti-fHbpv1, and anti-OMV serum antibodies were done coating ELISA plates with recombinant NadA, recombinant fHbpv1, and OMV, respectively. These reagents were obtained from Merck Vaccine Basic Research. The goat antimouse Ig APeconjugated secondary antibody used in the mouse serum ELISA measurements were purchased from ThermoFisher Scientific. Physiological saline, phosphate-buffered saline (PBS), and sterile water for injection

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11

(WFI) was obtained from Merck Manufacturing. The following reagents were American Chemical Society reagent grade or better and were purchased from Sigma-Aldrich: Trizma base (TRIS), sodium phosphate monobasic, imidazole, sucrose, L-histidine, sodium tetraborate decahydrate, ethylenediaminetetraacetic acid, and Tween® 20. Succinic acid was purchased from Acros Organics. Isolation and Purification of OMV Antigens The 3 OMV preparations used in this study were genetically modified versions of the 3 N. meningitidis parent strains 44/76 HOPS-LD, 8570HOPS-LG1, and B16B6HPS-LG2. The genetic modifications of these parent strains are described in Zollinger et al.16 The growth, extraction, and purification of OMV from these bacterial strains were modifications of previously described methods.16 Following purification the OMV samples, diafiltration was used to concentrate and buffer exchange the extraction buffer with WFI to eliminate saline. The purified OMV preparations were typically concentrated to 1.0-0.4 mg/mL OMV protein as measured by the bicinchonmic acid protein assay.25 The WFI vaccine drug substance OMV solutions were then sterilized by passage through 0.22-mm filter. These OMV MenB vaccine drug substances were
then stored at 4 C until vaccine formulation. Biophysical Stability by UV Spectroscopy and Antigen Calorimetry Middle-UV second-derivative spectroscopy was used to examine the structural stability of all the proteins associated with OMV. Middle-UV second-derivative spectroscopy in the 250-350 nm range using diode-array detectors and curve-fitting algorithms is an excellent technique to detect conformational changes in proteins.26 A purified 1.0-mL OMV stock from parent strain 44/76 HOPS-LD was suspended in 1.6-mM succinate/1.6-mM imidazole/1.6-mM TRIS pH 7.0 buffer at a concentration of 0.66 mg/mL OMV protein. A 0.5-mL aliquot of this OMV suspension
was incubated in Thermomixer (ThermoFisher Scientific) at 70 C
for 16 h. The remaining portion of this sample was stored at 4 C until the adsorption measurements were made. The middle-UV second-derivative spectra of both samples were taken at room temperature in a small-volume quartz cuvette using an Agilent HP 8453 spectrometer and HP 89090A temperature controller. The

3

second-derivative of the middle-UV spectra was calculated after an 11-point filter length and polynomial degree 3 fit using ChemStation Software from Agilent Technologies. Capillary DSC measures the heat absorbed when a protein unfolds. This technique is an excellent method to compare the relative thermal stability of protein antigens. Meaningful information about the thermal protein unfolding midpoint temperatures (Tm's) of NadA or fHbpv1 is not possible in OMV DSC measurements because of extremely low levels of these antigens in OMV particles. The w/w% of NadA and fHbpv1 in OMV particles used in this study was approximately 1.1% and 0.75% respectively (unpublished results). Therefore, recombinant NadA and fHbpv1 protein was grown in Escherichia coli and was supplied as pure protein in unbuffered saline in a concentration range between 2.3 and 2.5 mg/mL. Both these recombinant OMV antigens were soluble in saline because (1) the transmembrane portion of the recombinant NadA antigen was removed and (2) the recombinant fHbpv1 antigen did not have the N terminal lipid found in the OMV-associated form.15 The concentrated recombinant protein was diluted to 0.3 mg/mL with the addition of formulation excipients made to make final formulation in Figure 1b; 20-mM NaCl, 9 % (w/v) sucrose, 10-mM histidine, 10-mM succinate at pH 6.0. The DSC measurements were made using a VP-DSC Capillary Cell Microcalorimeter equipped with temperature-controlled autosam
pler (Malvern Instruments Ltd.). The heating rate was 1 C/min from

20 C to 95 C. The thermal transition midpoint, Tm, was determined by the instrument software. Over 30 DSC thermograms at various formulation conditions were measured during this study. The DSC thermograms shown in Figure 1b were typical of the recombinant NadA or fHbpv1 thermograms measured at the formulation conditions described previously. Formulation of OMV MenB Vaccines The OMV MenB vaccine formulations used in the mouse preclinical studies to demonstrate the NadA and fHbpv1 ELISA responses were stability indicating were formulated at room temperature inside a biosafety cabinet under aseptic conditions. In these formulations, relatively small aliquots of the OMV stock suspensions were directly added to a larger volume of aluminum adjuvant in buffered saline. The formulations were prepared from 3 OMV antigen vaccine drug substances WFI stock solutions. The 3 purified




Figure 1. Biophysical characterization of NadA and fHbpv1 stability. (a) Room temperature UV second-derivative spectra of 4 C control ( ) and thermally stressed ( ) single-strain
MenB OMV sample suspended in WFI. The thermal stress was 16 h at 70 C. This region of the OMV UV spectra is primarily due to PorA protein tryptophans and tyrosines. Before the second-derivative processing, the UV spectra were subject to an 11-point filter length and polynomial 3 fit by the Agilent spectrometer software. (b) DSC of the purified recombinant

water-soluble OMV protein antigens NadA ( ) and fHbpv1 ( ). The Tm transition for NadA is 46 C. The fHbpv1 antigen has 2 Tm's at 70
C and 84 C. Unfortunately, detecting the NadA and fHbpv1 Tm transitions in intact OMV particles was not possible because the NadA and fHbpv levels in OMV particles were only approximately 1.1 and 0.75 w/w %, respectively.

4

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11

OMV antigen stocks were suspended in WFI with no pH buffer. Two 0.60-mL OMV MenB vaccine formulations were prepared by first mixing the appropriate volumes of WFI, NaCl, and AlPO4 adjuvant in a sterile 1.5-mL polypropylene centrifuge tube. After these additions, the appropriate aliquots of the 3 OMV preparations in WFI were separately added to buffered aluminum adjuvant suspension and gently mixed by multiple pipetting. An appropriate volume of 200-mM histidine buffer at pH 7.0 was then added after the OMV antigens were combined with the aluminum adjuvant. The final formulation composition following formulation of the 2 OMV MenB vaccine supplies for this study was 150-mg/mL OMV protein, 450-mg/ mL Al AlPO4, 150-mM NaCl, 20-mM histidine at pH 7.0. As part of the stability indicating mouse preclinical study, one of the samples was
incubated at 70 C in an Eppendorf Thermomixer (ThermoFisher Scientific) for approximately 96 h, whereas the other control sample
was stored at 4 C. Both samples were protected from light during the
incubation. The samples were then stored protected from light at 4 C before dosing for the mouse preclinical study. A portion of each vaccine formulation was diluted with saline to obtain the proper concentration for dosing the mice at 0.3-mg OMV protein with 2 50-mL injections. The dosing concentration was 3-mg/mL OMV protein, 9-mg/mL Al AlPO4 with 150-mM NaCl and 20-mM histidine at pH 7.0. The OMV MenB vaccine for the mouse dosing was supplied in stoppered 3.0-mL sterile glass vials.
The OMV MenB vaccines used in the DoE 37 C stability study were prepared aseptically in sterile 2-mL polyproprolene Eppendorf centrifuge tubes inside a biosafety cabinet at room temperature. All solutions were sterile filtered through 0.22-mm pore size filters (Millex GP filter unit PES membrane filter, 0.22-mm pore size, Merck Millipore, Ltd.), except for the aluminum adjuvant containing solutions. In general, the DoE study OMV MenB vaccines were prepared by adding a 2 OMV antigen solution in WFI to a 2 aluminum adjuvant solution in 300-mM NaCl. The 4 formulation input factors were (a) the pH (IpH), (b) the adjuvant choice (IAdj), (c) the OMV protein/aluminum (w/w) ratio (IRa), and (d) the added phosphate concentration (IPc). These DoE formulation input factors are summarized in Table 1. The adjuvant choice (IAdj) input factor was made quantitative rather than categorical, so that a DoE “center point” could be evaluated for this formulation input factor. The added phosphate concentration typically has major effect on the microenvironment pH and antigen adsorption. The final formulation pH was set at either 5.5, 7.0, or 8.5 by triple buffer of 10-mM succinate/10-mM imidazole/10-mM TRIS. Using a triple buffer allowed buffering through a wide pH range (IpH), without changing the buffer components in the different formulations. The 3 aluminum adjuvant input choices (IAdj) for the formulation were AAHS, APO4, or an equal 1 to 1 mixture of AAHS and APO4 for the midrange DoE level. The OMV protein/aluminum (w/w) formulation loading ratio (IRa) was set by the aluminum adjuvant concentration selection. The final aluminum adjuvant concentration in the 20 DoE formulations were 750-, 300-, and 187.5-mg/mL Al for the 0.20, 0.50, and 0.80 IRa formulation values, respectively. The final formulation input factor IPc was set by adding a pH adjusted Na2HPO4 stock solution to set the added phosphate concentration to either 0-, 10-, or 20-mM phosphate in the 2 aluminum adjuvant stock. The Na2HPO4 stock solution pH was preadjusted to 5.5, 7.0, or 8.5 for each of the respective formulation pH levels so that this phosphate addition did not alter the formulation pH

set by the triple buffer component. Thus, the final IPc formulation concentrations in the DoE samples were 0-, 5-, and 10-mM phosphate at the targeted pH. The 3 OMV antigens used in this study were from the same 3 N. meningitidis strains described previously. The purified OMV antigen stocks were suspended in WFI with no pH buffer. The final OMV protein concentration target for all the DoE MenB vaccine formulations was 3 50-mg/mL or 150-mg/mL total OMV protein. Thus, the concentration of each individual OMV stock in all the DoE vaccines was 50 mg/mL. A 0.80-mL volume of the 2 OMV antigen volume was slowly added to a 0.8-mL volume of the 2 aluminum adjuvant suspension for each of the 20 DoE formulations. The completion order of the 20 DoE MenB vaccine formulations were prepared according to the standard formulation order specified by the Design Expert version 7.0 software (Stat-Ease, Inc., Minneapolis, MN). The final 1.60-mL volumes of all of the complete OMV MenB vaccine formulations were gently rotated for 15 min at room temperature to allow for maximum OMV adsorption to the aluminum adjuvant to reach equilibrium at room temperature. The amount of adsorbed OMV was determined for all samples by centrifuging a 200-mL volume of the vaccine to pellet the aluminum adjuvant. The amount of nonadsorbed OMV in the supernatant was then measured. The concentration of nonadsorbed OMV was measured by a second-derivative UV spectroscopy method. The percent of nonadsorbed OMV ranged from 0% to 20% of the added OMV and depended on both the sample pH and the added free phosphate. This particular formulation output response is not part of any DoE modeling done for this report. The 20 formulation were then stored

protected from light at 4 C until the start of the 37 C DoE stability study.
The 3 OMV MenB vaccine formulations examined in the 4 C 11month stability screening study were formulated at room temperature inside a biosafety cabinet under aseptic conditions. These MenB formulations were prepared in a manner similar to the DoE formulations, that is, OMV stock suspensions in WFI were directly added to 2 aluminum adjuvant in 2 buffered saline. The 3 buffers used in this study were sodium succinate at pH 6.5, Na2HPO4 at pH 7.0, and TRIS at pH 7.5. These 2.5-mL formulations were prepared by mixing the appropriate volumes in a 3.0-mL sterile glass vial. A 1.20-mL aliquot of each of the 3 separate MenB vaccine formulations were transferred to 3.0-mL sterile vials which were sealed by Flurotec-coated stoppers. The vials of formulated OMV MenB vaccines were gently mixed by slow rotation at room temperature for 3 h. Each MenB formulation contained OMV from the 3 N. meningitidis serogroup B strains used in the DoE study at a total of 150-mg/mL total OMV protein and AlPO4 at 450mg/mL Al in 150-mM NaCl. However, the 3 vaccine formulations had a different 20-mM buffers and pH which were as follows: succinate at pH 6.5, NaPO4 at pH 7.0, and TRIS at pH 7.5. The 3 formulations
were stored protected from light at 4 C until the start of the stability study. OMV MenB Stability Studies


Immediately before initiation of the 37 C DoE stress stability study, 0.4 mL was removed from each of the 20 1.6-mL DoE study samples for the day 0 time point. The aluminum adjuvant particles present in MenB OMV vaccine samples are not compatible with either ELISA assay. However, we found that both aluminum

Table 1 DoE Formulation Input Factors and the Corresponding Input Levels Formulation Input Factor

Input Factor Variable

High Level

Middle Level

Low Level

pH Aluminum adjuvant choice OMV protein/Al (w/w) ratio Added phosphate concentration

IpH IAdj IRa IPc

pH 8.5 100% AlPO4 0.80 w/w 10 mM

pH 7.0 50% AlPO4/50% AAHS 0.50 w/w 5.0 mM

pH 5.5 100% AAHS 0.20 w/w 0.0 mM

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11

adjuvants were completely dissolved following an overnight room temperature incubation following a one-to-one dilution with an adjuvant dissolution buffer. The adjuvant dissolution buffer used in this study was 100-mM ethylenediaminetetraacetic acid, 100-mM borate, 5% (w/v) sucrose, 0.05% (w/v) Tween® 20 at pH 8.5. We established that this overnight adjuvant dissolution procedure did not significantly affect either the NadA or fHbpv1 ELISA response. A 150-mL adjuvant dissolution buffer was incubated overnight with 150 mL of each vaccine DoE sample to dissolve the adjuvant particles. From this, a 200-mL aliquot of 2 diluted adjuvant-dissolved formulation for each DoE sample was then transferred the Merck Vaccine Analytical Group for TECAN-automated NadA and fHbpv1 ELISA assays. These adjuvant-dissolved samples were stored in the
dark at 4 C until immediately before the ELISA assay. The remaining 1.2 mL of each DoE MenB vaccine sample was then transferred to a
37 C temperature-controlled stability incubator for the remainder of the study. The samples were protected from light during this incubation. Stress stability time point samples of 0.4 mL were then
removed at 15, 35, and 77 days into the 37 C incubation. These later time point samples were prepared as described previously for transfer to the Merck Vaccine Analytical Group for the NadA and fHbpv1 ELISA assays.

The 11-month 4 C stability study was done in a 4 C temperature-controlled incubator. A 100-mL aliquot was removed from each of the 4 samples at months 0, 1, 5, 9, and 11 of the study. The adjuvant in these 100-mL vaccine aliquot was dissolved with the adjuvant dissolution buffer and transferred to the NadA and fHbpv1 ELISA assays as described previously. NadA and fHbpv1 Enzyme-Linked Immunosorbent Assays The amount of properly exposed and properly folded NadA or fHbpv1 epitopes in purified OMV samples were measured by a TECAN-automated ELISA assays run by the Merck vaccine analytical department. The nadA and fHbpv1genes for the 3 MenB OMV strain preparations used in this study are identical. The ELISA assays where equally effective at quantitatively measuring NadA and fHbpv1 epitopes for each of the 3 OMV strain preparations. The ELISA responses in this report will refer to the sum of all 3 OMV strain preparations unless otherwise indicated. Frozen OMV standards with established levels of NadA and fHbpv1 were used to quantitate the amount of active NadA or fHbpv1 epitopes in stressed OMV vaccine samples. Two separate sandwich ELISA assays were developed to measure properly folded MenB OMV NadA and fHbpv1 epitopes. The OMV capture antibodies used to coat the plates were 14A10 and 1-7F2-11 for the NadA and fHbpv1 ELISA assays, respectively. The adjuvant-dissolved MenB OMV samples from all the studies in this report were serially diluted in antibodycoated ELISA well plates. This was followed by addition of the biotinylated detection monoclonal antibodies 14E-7 for the NadA ELISA or17G7-3 for the fHbpv1 ELISA. Streptavidin-conjugated alkaline phosphatase and 4-MUP was then used to quantitatively measure antibody binding from the fluorescence signal. OMV MenB Vaccine Mouse Preclinical Studies Mouse experiments were approved by the Institutional Animal Care and Use Committee at Merck and Co., Inc. 6- to 8-week old female CD1 mice were purchased from Charles River Laboratories and housed in microisolator cages in the animal facility at Merck Research Laboratories (West Point, PA). Food was provided ad libitum. Groups of 10 mice were intramuscularly immunized with 0.3 mg of OMV vaccines on weeks 0, 2, and 4. Mouse serum antiNadA and anti-fHbpv1 IgG antibodies were measured in 96-well plates were coated with 0.1 mg per well of purified protein

5

antigens in PBS overnight at 4
C. The anti-OMV total antibodies were measured by coating the ELISA plate with OMV. Unbound sites were blocked by addition of 1% bovine serum albumin in PBS containing 0.05% Tween 20 (PBS-T) per well following an incubation for 1 h at room temperature. Mouse sera were 5-fold serially diluted on the plates and incubated at room temperature for 1 h. Plates were washed with PBS-T and incubated with 1:2000 diluted goat antimouse IgG APeconjugated secondary antibody for 1 h at room temperature. Following an additional wash with PBS-T, 4MUP substrate was added. After incubation for 1 h at room temperature, the fluorescence was measured with excitation at 360 nm and emission at 465 nm using a Spectramax M series plate reader (Molecular Devices, LLC). Mice sera from animals immunized with purified protein antigen were used as positive control. Preimmunization mice sera were used as negative control in the assay. Fluorescence intensity was plotted against dilutions of serum samples, and data were fitted with a 4-parameter logistic curve. ED50 titer for each individual animal was reported as the effective dilution of the serum that elicits 50% of the fluorescence signal. Geomean ED50 titer for each of the immunization group was calculated. Results Thermal Stability of OMV Particles, NadA, and fHbpv1 MenB OMVs are composed primarily of highly stable LOS and very stable b-barrel transmembrane PorA proteins.14 LOS which is a lipidated carbohydrate should not be susceptible to normal protein thermal denaturation conditions. As a result, the overall structure
of OMV particles is relatively stable. Temperatures as high as 56 C 22 for 3 months are needed to thermally degrade OMV. Middle-UV second-derivative spectroscopy in the 260-330 nm wavelength range is highly sensitive to the protein structural environment around tryptophans and tyrosines.26 We observed, as shown in the
Figure 1a, that incubation of MenB OMV at 70 C in WFI did not significantly alter the middle-UV second-derivative spectra even after 16 h. Because most of the tryptophans and tyrosines are associated with the OMV PorA proteins, this observation indicates there is little reorganization of the OMV PorA protein structure
even after exposure to 70 C for several hours. The OMV proteins NadA and fHbpv1 comprise less than a few percent of the OMV mass and would not be expected to contribute significantly the middle-UV second-derivative OMV spectra. In contrast to the PorA proteins, NadA and fHbpv1 are key MenB antigens because they should provide broad vaccine protection in humans.14,15,27 The OMVs isolated from the 3 distinct strains of N. meningitidis serogroup B bacteria that were used this report contain the NadA and fHbpv1 antigens to maximize the MenB strain coverage for this vaccine.16-18,28 What about the thermal stability of OMV antigens like NadA and fHbpv1? This question was addressed in a biophysical study using capillary DSC of purified water-soluble recombinant versions of NadA and fHbpv1. NadA without the transmembrane domain and fHbpv1 without the fatty acid membrane anchor that were recombinantly expressed and purified were obtained from the Merck Vaccine Basic Research group. The DSC of recombinant versions of these OMV antigens is shown in Figure 1b. The protein unfolding Tm for the recombinant
NadA molecule was approximately 46 C. The truncated recombinant NadA used this study was most likely in the trimeric form that contained a considerable portion of the coiled-coil segment. Other recombinant versions of NadA with shorter coiled-coil a-helix segments are reported to have higher Tm's.29 The recombinant fHbpv1 protein antigen was significantly more stable than the recombinant NadA when examined by DSC showing 2 high

6

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11





temperature unfolding Tm's at 70 C for the N terminal and 82 C for the C terminal regions, respectively.30,31 The DSC results in Figure 1b also suggest the highest possible temperature for the accelerated DoE stability study. According to Figure 1b, NadA pro
tein begins to immediately unfold around 40 C, so the accelerated DoE stability study temperature should be below that temperature.
We selected 37 C because it was just below the beginning of the NadA unfolding temperature. The NadA and fHbpv1 proteins must have properly folded antibody epitopes to provide effective in vivo immunogenicity. To address this key question, the temperature dependence of the antiNadA and anti-fHbpv1 ELISA responses were used to quantify the amount of properly folded and exposed, that is, immunologically
active, NadA or fHbpv1 epitopes in OMV. Figure 2 shows that a 70 C incubation of isolated MenB OMV resulted in a large ELISA response decline for both NadA and fHbpv1 after 24 h. The isolated MenB
OMV used in this 70 C stress study were suspended in 9% (w/v) sucrose, 20-mM histidine at pH 6.5. The ELISA responses have been normalized to the time point 0. This normalization process was done for all the NadA and fHbpv1 ELISA response experiments in this report. The results in Figure 2 clearly indicate OMV-associated fHbpv1 is significantly more thermally stable than OMV-associated NadA. This result is consistent with the recombinant NadA and fHbpv1 DSC results shown in Figure 1b.

The NadA and fHbpv1 ELISA Response Assays Are Stability Indicating To adequately evaluate the stability of any vaccine formulation, an assay is needed which correlates well with the in vivo immunogenicity of the vaccine. Such assays are often referred to as “stability indicating” assays. Assay results which do not correlate with in vivo immunogenicity are generally not very useful for optimizing vaccine formulations. Incubation of isolated MenB OMV
at 70 C for 24 h significantly reduced both the NadA and fHbpv1 ELISA response (see Fig. 2). Does this treatment significantly reduce
the in vivo immune response? The effect of this 70 C stress on the in vivo immune response for a formulated MenB OMV vaccine was examined next. CD1 mice were dosed intramuscularly at 0, 2, and 4 weeks with 0.3-mg total OMV protein per MenB AAHS adjuvant vaccine injection. The mice were bled at 6 weeks. The mouse serum

Figure 2. Temperature dependence of OMV-associated NadA ( ) and fHbpv1 ( )
ELISA responses during a 70 C incubation normalized to time 0. Both ELISA responses
decrease significantly over 24 h at 70 C, but the NadA response appears to be less thermally stable. The OMVs in this study were suspended in 9% (w/v) sucrose, 20-mM histidine at pH 6.5. The error bars represent the CV% of 3 separate ELISA measurements on each sample.

anti-OMV, anti-fHbpv1, and anti-NadA total IgG titers were measured for each mouse. Figure 3 compares the normalized mouse serum anti-OMV, anti-fHbpv1, and anti-NadA ED50 titers of
control vaccines (left side) to vaccines that were incubated at 70 C (right side). Incubating the AAHS-formulated MenB OMV vaccine
for 96 h at 70 C significantly reduced the anti-NadA and antifHbpv1 immune responses by approximately 4 and 3 log units, respectively, compared to the nonincubated control vaccines. In
contrast, the 70 C thermal stress had a significantly smaller impact on the overall anti-OMV immune response possibly because antibody responses against other antigens such as PorA and LOS were
not affected as much by thermal stress. Incubation at 70 C clearly reduces both the NadA and fHbpv1 in vivo immune response and the in vitro ELISA assay response, that is, the in vitro assays correlate with mouse in vivo immunogenicity. Thus, the in vitro NadA and fHbpv1 ELISA response assays appear to be stability indicating for these antigens and can be used to examine the functional stability of the MenB vaccine formulations in the DoE study.


DoE Modeling of the 37 C Accelerated Stability Study DoE is a structured, organized mathematical method for determining the relationship between input factors affecting a process and the output responses of that process.7 The process in this specific DoE study was the stability of OMV vaccine formulations at
37 C after 15, 35, and 77 days. A 4-factor 2-level full-factorial model with 4 midpoints was chosen to characterize this process. This DoE model was chosen to investigate how 4 input factors and potential input inactions could control the stability outputs in a linear manner. The 4 input factors, the variables name for the input factors, and levels for the DoE model are summarized in Table 1. The analysis requires 24 þ 4 ¼ 20 OMV vaccine formulation test samples for this particular DoE study. Design Expert DoE software was used to calculate the required 16 OMV vaccine formulation combinations of 4 input factors along with the 4 identical middle level formulations for this 4-factor 2level full-factorial study. Twenty individual 1.6-mL OMV vaccine formulations were prepared for the study as described in the

Figure 3. The in vivo anti-LOS ( ), anti-fHbpv1 ( ), and anti-NadA ( ) immune responses of CD1 mice dosed with a control MenB OMV AAHS vaccine (left side) or a
thermally stressed MenB AAHS OMV vaccine (right side). The thermal stress was 70 C for 96 h. The thermal stress of the vaccine significantly decreased mouse immunogenicity of the OMV protein antigens fHbpv1 and NadA. In contrast, the thermal stress had much smaller impact on the OMV immunogenicity. This demonstrates that the NadA and fHbpv1 ELISA assays (Fig. 2) are “stability indicating,” so testing the in vivo immune response of the AlPO4 adjuvant was not necessary. The black bars indicate the geomean of the respective titers.

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11

Materials and Methods section. An aliquot of each OMV vaccine formulation was removed, and the aluminum adjuvant was dissolved at day 0 of the DoE stability study for both the NadA and fHbpv1 ELISA responses. The 20 OMV vaccine formulations were
then transferred to a temperature-controlled 37 C stability cabinet for an accelerated stress stability study. The samples were protected from light during the study. An aliquot from each OMV vaccine formulation was removed, and the aluminum adjuvant was dissolved at 15, 35, and 77 days for NadA and fHbpv1 ELISA response assay. The % ELISA response change ( D%Antigen ) relative to the day 0 ELISA response was calculated for each of the assay time points for each of the 20 OMV vaccine formulations in the DoE experiment. The formulation input factors and the corresponding D%NadA and D%fHbpv1 ELISA responses are summarized in Table 2. Design Expert DoE software was then used to mathematically model the Table 2 D%NadA and D%fHbpv1 ELISA responses for days 15, 35, and 77 of the stability study. Our DoE modeling approach did not fit the NadA and fHbpv1 ELISA responses changes to a time dependent ELISA response function. Rather, the 3 specific D%NadA or D%fHbpv1 linear response output models were calculated for the 3
time points in the 37 C stability study (see Tables 3 and 4). In addition, the DoE software also calculated a p value to indicate the statistical significant of the D%NadA or D%fHbpv1 output models to the actual D%NadA or D%fHbpv1 results. A p value <0.05 for the complete output response model indicated that the specific DoE model was statistically significant. The DoE models for the D%NadA and D%fHbpv1 ELISA responses used 16 separate OMV vaccine formulations and 4 identical midlevel formulations. Of course, these formulations were not identical. Each formulation was a different combination of the 4 input factors which are summarized in Table 1. A reasonable good view of the overall normalized NadA and fHbpv1 ELISA response for all the MenB OMV vaccines in stability study can be obtained without analysis by the DoE software. This is done by simply averaging the normalized ELISA responses changes for the 20 formulations for each time point during the study. This simple averaging analysis of all the normalized NadA and all the normalized fHbpv1 ELISA results as a function of time is shown in Figure 4. Each time point in Figure 4 is the average and standard deviation of 20 distinct formulations for each antigen.

7

Figure 4 illustrates how the OMV vaccine NadA ELISA response
declines during the course of the 37 C stability study even without differentiating the 4 formulation input factors used in the DoE study. However, this simple averaging of the results does not reveal any insight into how any of the 4 input factors might affect OMV NadA ELISA response stability. Surprisingly, OMV fHbpv1 ELISA
response actually increased during the course of the 37 C stability study as indicated in Figure 4. The DSC results presented in Figure 1b indicated that even the least stable portion of the fHbpv1
protein starts to unfold at 65 C. Thus, it is not surprising that the fHbpv1 ELISA response does not decline significantly over 77 days
at 37 C as indicated in Figure 4. However, how could the fHbpv1 ELISA response be greater at day 15 than day 0 and remain high
during the time course of the 37 C study? These increases in the fHbpv1 ELISA responses are probably due additional fHpv1 epitopes becoming exposed to the ELISA detection antibodies during the course of the stability study. We will speculate on how this might occur later in the discussion section. Determining the subtle effects of the 4 formulation input factors on the ELISA response of both these OMV antigens were not identified in Figure 4. This will require the more sophisticated DoE analysis which is presented in the following.


DoE Models for D% NadA and D% fHbpv1 ELISA Responses at 37 C The DoE software was used to calculated D%NadA ELISA response change models for days 15, 35, and 77 in the stability study. The 3 time point mathematical models are summarized in Table 3. The names of the input variables are presented in Table 1. The DoE modeling of the MenB OMV vaccine formulations for D%NadA are presented in Table 3. These models revealed the impact or lack impact of the 4 formulation inputs on NadA stability which were not immediately apparent in Figure 4. These response models are quite valuable because they (1) predict the D%NadA ELISA response output and (2) indicate which input formulation factors do and do not control D%NadA . All 3 DoE response models in Table 3 indicate that D%NadA ELISA
response declines from the day 0 at all the time points in the 37 C stability study. The negative coefficient for the formulation pH input (IpH ) term in the day 15 D%NadA equation dominates the

Table 2 The NadA and fHbpv1 ELISA Responses Used in DoE the Mathematical Models Label

Input Formulation Factors

Formulation ID#

pH (IpH)

Adjuvant Choice, % A (IAdj)

OMV to Adjuvant Formulation Ratio, w/w (IRa)

1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00 17.00 18.00 19.00 20.00

5.50 5.50 5.50 5.50 8.50 8.50 8.50 8.50 5.50 5.50 5.50 5.50 8.50 8.50 8.50 8.50 7.00 7.00 7.00 7.00

AAHS AlP04 AAHS AlP04 AAHS AlP04 AAHS AlP04 AAHS AlP04 AAHS AlP04 AAHS AlP04 AAHS AlP04 1:1 AAHS-AlPO4 1:1 AAHS-AlPO4 1:1 AAHS-AlPO4 1:1 AAHS-AlPO4

0.20 0.20 0.80 0.80 0.20 0.20 0.80 0.80 0.20 0.20 0.80 0.80 0.20 0.20 0.80 0.80 0.50 0.50 0.50 0.50

Day 15 ELISA

Day 35 ELISA

Day 77 ELISA

Added Phosphate, mM (IPc)

D%NadA

D%fHbpv1

D%NadA

D%fHbpv1

D%NadA

D%fHbpv1

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 10.00 5.00 5.00 5.00 5.00

20.23 5.06 6.21 14.27 15.12 17.93 16.24 5.66 1.55 17.10 0.84 11.25 33.03 23.78 30.28 30.73 19.01 15.04 10.89 17.33

29.35 44.80 34.19 30.95 22.34 23.60 24.39 9.24 26.87 40.58 42.16 24.15 12.54 12.48 13.43 6.09 28.82 34.99 34.39 23.93

32.98 4.57 24.22 28.15 32.92 34.37 21.46 20.11 15.52 26.40 13.89 29.07 45.24 37.06 36.00 41.97 27.14 25.86 22.69 17.15

14.85 31.63 18.72 18.40 5.71 4.76 4.15 9.55 19.75 32.00 21.67 15.05 13.72 5.49 14.75 2.80 16.83 21.21 13.54 4.50

46.09 10.11 35.63 27.15 34.97 42.92 42.89 35.89 18.56 19.64 17.42 12.78 61.23 41.48 61.31 61.81 40.56 19.58 17.75 28.19

44.31 61.08 31.88 4.52 11.33 12.20 13.81 12.58 28.78 52.46 42.44 38.46 25.56 1.83 17.39 9.23 16.68 41.04 50.22 28.62

8

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11

Figure 4. Single time point average NadA ( ) and fHbpv1 ( ) ELISA responses
for all samples in the 37 C DoE stress stability study. The single time point is average and standard deviation of all 20 MenB vaccine formulations in the study, although the compositions of the 20 formulations are significantly different. This figure clearly indicates that the average relative NadA ELISA response decreases during the time course of this stability study, whereas the average relative fHbpv1 ELISA response was greater than the day 0 response particularly at day 15.

output response and predicts that higher pH values will result in larger decreases in the NadA ELISA responses. Thus, lower pH values will increase NadA stability by reducing the net NadA ELISA response decrease. The negative effect of high pH is also observed at days 15, 25, and 77 of the stability study. In addition to pH affect, the DoE D%NadA models at days 35 and 77 into the stability studies indicate that the OMV vaccine formulation added phosphate concentration also has an impact on NadA stability. In addition, the formulation phosphate input (IPc) also has an interaction with the formulation pH as summarized in Table 3. The D%NadA models predict that other 2 OMV vaccine formulation input variables, that is, adjuvant choice (IAdj ) and the OMV to aluminum adjuvant formulation (w/w) ratio ðIRa Þ appear to have no significant effect on NadA stability. Half-normal and Pareto analysis of the D%NadA results by the DoE software also identified the input factors IpH and IPc for D%NadA as statistically significant (see Supplementary Material). The effect of pH and the interaction between the added phosphate concentration and pH are clearly illustrated in the DoE softwareegenerated phosphate-pH interaction graphs shown in Figure 5. Figure 5a shows how D%NadA is predicted to change after
15 days at 37 C over a wide range of formulation pHs. Figure 5b (day 35) and Figure 5c (day 77) predict the effect of formulation pH on the D%NadA with no added phosphate (black) or a 10.0-mM phosphate (red) phosphate addition. The error bars indicate the least significant difference in all the graphs which is calculated by the DoE software. These 3 formulation input factor graphs illustrate that the D%NadA is smallest at the low pH formulations. More OMV NadA epitope is present at days 15, 35, and 77 at the lower pH level, that is, pH 5.5. The effect of pH on NadA stability below pH 5.5 is not considered in the D%NadA model. Notice in Figure 5c that the addition of 10-mM phosphate to the formulation significantly improves the stability of NadA at pH 5.5, yet significantly decreases

the stability of NadA at pH 8.5. This small yet statistically significant interaction between the pH input and the added phosphate input to NadA formulation stability only becomes apparent following the DoE analysis of the D%NadA formulation output. The DoE software was also used to calculated D%fHbpv1 ELISA
response change models for days 15, 35, and 77 in the 37 C stability study. These mathematical models are summarized in Table 4. Unlike the D%NadA ELISA response models, the D%fHbpv1 output responses these models did not predict the conventional thermalinduced ELISA response decrease. In fact, the fHbpv1 ELISA response on days 15, 35, and 77 was greater than the day 0 fHbpv1 response. These DoE predicted results are completely consistent with the average normalized fHbpv1 results shown in Figure 4. The D%fHbpv1 DoE response models in Table 4 indicate that the
D%fHbpv1 is positive at all time points during this 37 C stability study over the range of pH and added phosphate covered in the study conditions. In fact, the fHbpv1 models in Table 4 generally indicate that that fHbpv1 ELISA responses at days 15, 35, and 77 are greater the response at day 0. So why is the total number of fHbpv1 ELISA epitopes greater than day 0 under these stability conditions? The
high fHbpv1 unfolding temperatures, Tm's >65 C, would account
for the lack of fHbpv1 epitope loss at 37 C. However, another process must be involved to account for the increased number of
fHbpv1 exposed epitopes during the 37 C stability study. We will speculate what this process might be in the discussion section. The negative value for the IpH term in all the D%fHbpv1 model equations predicts that higher pH values would decrease the positive D%fHbpv1 responses. Thus, pH 5.5 has more exposed fHbpv1 epitopes than pH 8.5 at days 15, 35, and 77. This suggests that fHpbv1 is more stable in the lower pH range of the DoE study. The 15-day D%fHbpv1 DoE model equation also indicates that there is a relatively small interaction between Iadj and IRa which also influences D%fHbpv1 . There is more fHbpv1 exposed epitope on AAHS than AlPO4, when IRa is 0.20, but if IRa is 0.80 more fHbpv1 epitopes are exposed on the AlPO4 adjuvant. Interestingly, this Iadj and IRa formulation factor interaction only applies to day 15. The pH input (IpH ) was the only vaccine formulation input factor that influenced the D%fHbpv1 output on days 35 and 77 of the stability study. Half-normal and Pareto analysis of the D%fHbpv1 results by the DoE software also identified the input factors IpH, Iadj, and IRa for D%fHbpv1 as statistically significant (see Supplementary Material). The effect of IpH on D%fHbpv1 is illustrated in the DoE softwareegenerated D%fHbpv1 output response graphs for days 15, 35, and 77 are shown in Figures 5d-5f, respectively. The 15-day Iadj and IRa interaction graph is not shown because this interaction is relatively small and not as significant as the effect of pH on D%fHbpv1 . The D%fHbpv1 formulation response output is more positive, that is, more exposed fHpbv1 epitopes, at pH 5.5 than pH 8.5 at days 15, 35, and 77. This suggests that the fHpv1 OMV antigen is more stable in the lower pH range of this study.


A 11-month OMV MenB Vaccine Stability Screen at 4 C for the NadA and fHbpv1 ELISA Response


Can these DoE-derived results from the accelerated 37 C stability described previously predict MenB vaccine stability at significantly lower temperatures? To address this question, the

Table 3 The NadA DoE Output Response Models for Days 15, 35, and 77 Time Point (Days)

Significant Input Factors

DoE Output Response Models for D%NadA

Model p Value

15 35 77

IpH IpH ; IPc IpH ; IPc

D%NadA  15:96  4:34  IpH D%NadA ¼ 12:89  1:58  IpH þ 2:71  IPc  0:47  IpH  IPc D%NadA ¼ 10:65  3:14  IpH þ 6:75  IPc  0:10  IpH  IPc

0.0046 0.0144 0.0005

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11

9

Table 4 The fHbpv1 DoE Output Response Models for Days 15, 35, and 77 Time Point (Days)

Significant Input Factors

DoE Output Response Models for D%fHbpv1

Model p Value

15 35 77

IpH ; Iadj ; IRa IpH IpH

D%fHbpv1 ¼ 65:43  6:21  IpH  0:14  IAdj þ 9:61  IRa  0:31  IAdj  IRa D%fHbpv1 ¼ 64:64  7:69  IpH D%fHbpv1 ¼ 96:21  10:33  IpH

0.0001 <0.0001 0.0020

results from this accelerated DoE study were compared to results
from a conventional MenB vaccine stability screen done at 4 C
which required 11 months. This parallel and separate 4 C OMV MenB vaccine stability formulation screen was started before the
accelerated 37 C DoE study described in this report. Each OMV MenB vaccine in this study contained OMV from the same 3 N. meningitidis serogroup B strains as used in the previously discussed DoE study although the exact OMV purification procedures were slightly different. The composition of all the MenB vaccine formulations in this study contained 75-mg/mL OMV protein, 250-mg/mL Al AlPO4 aluminum adjuvant, and 150-mM NaCl. The
three 4 C stability screen formulations also contained 1 of the following 3 buffer and pH conditions: (a) sodium succinate at pH 6.5, (b), Na2HPO4 at pH 7.0, and (c) TRIS at pH 7.5. Thus, these 3 formulations represent formulation conditions considered in the DoE study. There was a low pH formulation (sodium succinate at pH 6.5), high pH formulation (TRIS at pH 7.5), and a

phosphate-containing formulation (Na2HPO4 at pH 7.0). Both the NadA and the fHbpv1 ELISA responses of these formulations were measured at months 1, 5, 9, and 11 into the stability study and normalized to the initial ELISA responses at month 0. The ELISA assays were identical to assays used in the accelerated DoE stability study. Figure 6 shows the normalized NadA and fHbpv1 ELISA responses for the 3 OMV vaccine formulations described previously
over 11 months at 4 C. The individual points represent the average of 2 ELISA response measurements from same sample. The NadA ELISA responses from the 3 formulations did not
appear to change significantly over the first 5 months of this 4 C stability study as shown in Figure 6a. However, possibly significant differences between the 3 formulations appeared at months 9 and 11 at this temperature. In particular, the pH 6.5 formulation NadA ELISA response a these time points was the largest at 9 and 11 months, whereas the 20-mM phosphate-pH 7.0 formulation had the lowest ELISA response. In addition, the NadA ELISA responses

Figure 5. The NadA and fHbpv1 ELISA response changes ( D%NadA and D%fHbpv1 ) as a function of formulation pH calculated by the DoE software. (a) D%NadA ELISA response versus pH on day 15, (b) D%NadA ELISA response on day 35, (c) D%NadA ELISA response on day 77, (d) D%fHbpv1 ELISA response versus pH at day 15, (e) D%fHbpv1 ELISA response day 35, (f) D%fHbpv1 ELISA response day 77. The error bars indicate the least significant difference (LSD) which is calculated by the Design Expert DoE software. The black lines in Figure 5b and 5c predict the D%NadA output responses at 0.0-mM added phosphate, whereas the red lines are the responses at 10-mM added phosphate. The black lines in Figure 5d, 5e, and 5f predict D%fHbpv1 output as function of pH. There is no dependence of D%fHbpv1 on added phosphate for any of the DoE-generated models. All these DoE softwareegenerated graphs predict that more NadA and fHbpv1 epitopes are exposed at pH 5.5 relative to pH 8.5. This suggests that both NadA and fHbpv1 are more stable at the lower pH. Addition of 10-mM phosphate enhances this pH-driven stability difference with NadA on days 35 and 77 but has no effect on fHbpv1 stability at any time point.

10

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11




Figure 6. The average normalized ELISA responses of 3 AlPO4 adjuvant MenB vaccine formulations during an 11-month stability screen at 4 C. The coefficient of variation % for each time point (N ¼ 2) is shown by the dashed error bars. In some time points, the error bars are smaller than the size of the data point. (a) Normalized NadA ELISA response. (b) Normalized fHbpv1 ELISA response. The 3 formulations were identical expect for the formulation buffer and pH. The following 3 buffer and pH combinations were used; 20-mM succinate pH 6.5 ( ), 20-mM NaPO4 pH 7.0 ( ), and 20-mM TRIS pH 7.5 ( ).

were larger at pH 6.5 then pH 7.5. These relative stability trends at
4 C are consistent with the accelerated DoE NadA stability study.
However, these NadA ELISA response changes at 4 C must be considered as trends which are not sufficient to establish the statistical significance of any formulation input factor such as pH as
was done in the analysis of the 37 C DoE study results.
Similar to 37 C DoE study results, the normalized 3 fHpv1 ELISA
responses appeared to be greater at months 5 and 9 during the 4 C stability screen as shown in Figure 6b. However, there might be an
initial decline in the fHbpv1 response during the first month at 4 C particularly for the phosphate-containing formulation. Again, there is insufficient statistical information in this 11-month study to either support or oppose the fHbpv1 ELISA response dependence on pH that was identified in the accelerated DoE study. However, the fHbpv1 epitope exposure at month 9 was greater that month
0 for all the formulations in the conventional 4 C stability screen. We will speculate on how this might occur later in the Discussion section. Discussion The central theme of this report is that relatively simple DoE
methodology used for a short 37 C thermal stress stability study revealed several formulation input factors and interactions that control stability of aluminum adjuvant OMV MenB vaccines. The OMV MenB vaccine stability was measured by the normalized NadA and fHbpv1 ELISA responses during the course of the thermal stress. The normalized NadA and fHbpv1 ELISA responses were shown to be stability indicating assays that correlate with mouse immunogenicity studies. Understanding these formulation factors and interaction are clearly useful for further formulation development of OMV-based MenB vaccines. In contrast, clear identification of these factors was not possible in a separate conventional
11-month MenB vaccine stability screen at 4 C which did not use accelerated thermal stress and DoE methodology. The following 3 OMV MenB vaccine formulation conclusions were supported by the DoE study: (1) The fHpv1 OMV antigen has significant more thermal stability than the NadA OMV antigen, (2) The formulation pH (IpH) and phosphate concentration (IPc) has a significant effect on the thermal stability of NadA, and (3) The formulation pH (IpH) has an effect on epitope exposure and as a result the stability of the
fHbpv1 antigen during a 37 C incubation. NadA is a 362 amino acid protein with a predicted molecular weight of 35 kD.27 The secondary structure analysis based on the amino acid sequence predicts 3 distinct structural segments in the single polypeptide chain. The COOH-terminal segment is predicted

to be a b-sheet membrane anchoring domain. The middle segment is a a-helix region which forms a coiled-coil oligomeric structure with 2 other NadA monomers. The amino terminal head region is predicted to have a less-defined secondary structure. This segment of NadA is presumable responsible for N. meningitidis adhesion to host epithelial cells. An X-ray diffraction-based structure of the large ectodomain of NadA has been recently published.29 The truncated recombinant NadA segment evaluated by DCS in Figure 1b probably consists of a trimer of the amino terminal head regions and the large segment of the coiled-coiled regions. The DSC results in Figure 1b indicate the adhesion segment of NadA unfolds
at the relatively low Tm of 46 C. However, the overall thermal stability of the entire OMV-associated NadA trimer may be significantly higher.29 The DoE D%NadA ELISA response models in Table 3 indicate that the NadA ELISA response decrease is smaller at pH 5.5 than pH 8.5 at all the time points in the study, suggesting that NadA is more stable at pH 5.5. The least significant difference bars in the DoE NadA ELISA response graphs in Figure 5 indicate that this pH affect together with a pH and phosphate concentration interaction effect on the NadA ELISA response stability are statistically significant. The structure of fHbpv1 has also been determined.15,30-33 This 27kD protein has 2 predominately antiparallel b-sheet domains that bind to human factor H. The 2 fHbpv1 domains are relatively stable

with protein unfolding Tm's around 70 C and 85 C as indicated in Figure 1b. This high thermal stability is certainly the principle reason that the fHbpv1 ELISA response does not decline during incubation

at 4 C or even 37 C as indicated in Figures 4 and 6. However, why would the fHbpv1 ELISA response increase during these stability studies? Changes in fHbpv1 ELISA responses could result from the following 3 possible processes during the stability incubation: (1) degradation of the properly folded epitopes, (2) exposure loss of the properly folded epitopes, or (3) exposure gain of the properly folded epitopes. Temperature-induced epitope unfolding or degradation would surely cause the fHbpv1 ELISA response to steadily decline as
was observed in Figure 2 for fHbpv1 epitopes incubated at 70 C. Changes in fHbpv1 exposure to the ELISA assay capture or detection antibodies could cause the ELISA response to decrease or even increase. Because the fHbpv1 ELISA response at times appeared to increase during incubations of the OMV vaccines, this indicates that more properly folded fHbpv1 epitopes must have been exposed during this period. The fHbpv1 ELISA response increases do appear to be genuine and not the result of ELISA assay noise because it appeared to be present in all the OMV stability studies (Figs. 2, 4, and 6). Thus, in our studies, there is a high likelihood that a thermally induced change to the OMV structure or aggregation exposes more

P.L. Ahl et al. / Journal of Pharmaceutical Sciences xxx (2016) 1-11

fHbpv1 epitopes for the particular monoclonal antibody used in our ELISA assay. Unlike NadA, fHbpv1 is weakly tethered to the OMV membrane by a single lipid chain.14,15 This unresolved epitope exposure increase mechanism may involve physical rearrangement or movement of fHbpv1 molecules on the OMV surface for better antibody accessibility. Alternatively, increase fHbpv1 active epitope exposure could result from thermally induced OMV rearrangement or disaggregation. However, the NadA ELISA response did not show
corresponding epitope increase during the 37 C DoE incubation that might result from thermally induced OMV disaggregation (Fig. 4). However, there may have been NadA epitope increase during the
early phases of the 4 C incubation (Fig. 6a). More experimentation will be required to understand the complicated and unresolved molecular mechanism of incubation-induced increases in fHbpv1 epitope exposure. However, the DoE analysis of the fHbpv1 ELISA
response during the 37 C stability study did identify a significant effect of pH on fHbpv1 epitope exposure and/or stability. Conclusions These valuable DoE-derived formulation insights can be used to design better OMV MenB vaccine formulations. We found the re
sults from much longer formulation screening studies at 4 C and biophysical characterization of the antigens were consistent with
the 37 C DoE stability study. This supports the validity of the accelerated DoE stability study approach to identify formulation stability trends in early vaccine development. Compared to a long formulation stability screening study, we found that useful information can be obtained with a well-designed accelerated DoEbased stability study in less time and with less vaccine material. Of course, accelerated DoE-based thermal stress stability studies cannot replace the long-term multiyear stability studies required by regulators to unambiguously demonstrate vaccine drug product stability for the market approval. However, accelerated DoE-based studies can help formulators quickly pick the very best vaccine drug product formulations for those critical long-term studies needed at the end of the vaccine development process. Acknowledgments Edith Tan Senderak of MRL Non-clinical Statistics and WRAIR for supplying the original MenB strains.

7. 8. 9. 10. 11. 12.

13.

14. 15.

16.

17.

18.

19.

20.

21. 22. 23. 24.

25. 26. 27. 28.

References 29. 1. van Panhuis WG, Grefenstette J, Jung SY, et al. Contagious diseases in the United States from 1888 to the present. N Engl J Med. 2013;369(22):2152-2158. 2. Dey AK, Malyala P, Singh M. Physicochemical and functional characterization of vaccine antigens and adjuvants. Expert Rev Vaccines. 2014;13(5):671-685. 3. Higbee RG, Byers AM, Dhir V, et al. An immunologic model for rapid vaccine assessmentda clinical trial in a test tube. Altern Lab Anim. 2009;37(Suppl 1):19-27. 4. Maddux NR, Joshi SB, Volkin DB, Ralston JP, Middaugh CR. Multidimensional methods for the formulation of biopharmaceuticals and vaccines. J Pharm Sci. 2011;100(10):4171-4197. 5. Hu L, Joshi SB, Liyanage MR, et al. Physical characterization and formulation development of a recombinant pneumolysoid protein-based pneumococcal vaccine. J Pharm Sci. 2013;102(2):387-400. 6. Meers PR, Ahl PL. Stress testing to determine liposome degradation mechanisms. In: Baertschi SW, Alsante KM, Reed RA, eds. Pharmaceutical Stress

30.

31.

32.

33.

11

Testing: Predicting Drug Degradation. 2nd ed. Boca Raton, FL: CRC Press; 2011: 426-446. Mark J, Anderson PJW. DOE Simplified: Practical Tools for Effective Experimentation. 2nd ed. New York, NY 10016: Productivity Press; 2007:241. Haas J, Franklin A, Houser M, et al. Implementation of QbD for the development of a vaccine candidate. Vaccine. 2014;32(24):2927-2930. Esposito S, Principi N. Vaccine profile of 4CMenB: a four-component Neisseria meningitidis serogroup B vaccine. Expert Rev Vaccines. 2014;13(2):193-202. Brendish NJ, Read RC. Neisseria meningitidis serogroup B bivalent factor H binding protein vaccine. Expert Rev Vaccines. 2015;14(4):493-503. Rouphael NG, Stephens DS. Neisseria meningitidis: biology, microbiology, and epidemiology. Methods Mol Biol. 2012;799:1-20. Saunders NB, Shoemaker DR, Brandt BL, Moran EE, Larsen T, Zollinger WD. Immunogenicity of intranasally administered meningococcal native outer membrane vesicles in mice. Infect Immun. 1999;67(1):113-119. Zollinger WD, Shoemaker D, Saunders AG, Brandt BL. Vaccine against Gram Negative Bacteria. 2003. United States US patent application US20030059444A1; Publication date March 27, 2003. Feavers IM, Pizza M. Meningococcal protein antigens and vaccines. Vaccine. 2009;27(Suppl 2):B42-B50. Cantini F, Veggi D, Dragonetti S, et al. Solution structure of the factor H-binding protein, a survival factor and protective antigen of Neisseria meningitidis. J Biol Chem. 2009;284(14):9022-9026. Zollinger WD, Donets MA, Schmiel DH, et al. Design and evaluation in mice of a broadly protective meningococcal group B native outer membrane vesicle vaccine. Vaccine. 2010;28(31):5057-5067. Pinto VB, Moran EE, Cruz F, et al. An experimental outer membrane vesicle vaccine from N. meningitidis serogroup B strains that induces serum bactericidal activity to multiple serogroups. Vaccine. 2011;29(44):7752-7758. Keiser PB, Gibbs BT, Coster TS, et al. A phase 1 study of a group B meningococcal native outer membrane vesicle vaccine made from a strain with deleted lpxL2 and synX and stable expression of opcA. Vaccine. 2010;28(43):69706976. Zhang L, Wen Z, Lin J, et al. Improving the immunogenicity of a novel trivalent Neisseria meningitidis outer membrane vesicle vaccine by genetic modification. Vaccine. 2016;34:4250-4256. US Food and Drug Administration. Guidance for Industry. Q8 (R2) Pharmaceutical Development. 2009. Available at: http://www.fda.gov/downloads/Drugs/.../ Guidances/ucm073507.pdf. Accessed June 10, 2016. Kraan H, van Herpen P, Kersten G, Amorij JP. Development of thermostable lyophilized inactivated polio vaccine. Pharm Res. 2014;31(10):2618-2629. Arigita C, Jiskoot W, Westdijk J, et al. Stability of mono- and trivalent meningococcal outer membrane vesicle vaccines. Vaccine. 2004;22(5-6):629-642. Wang W, Roberts CJ. Non-Arrhenius protein aggregation. AAPS J. 2013;15(3): 840-851. Caulfield MJ, Shi L, Wang S, et al. Effect of alternative aluminum adjuvants on the absorption and immunogenicity of HPV16 L1 VLPs in mice. Hum Vaccines. 2007;3(4):139-145. Smith PK, Krohn RI, Hermanson GT, et al. Measurement of protein using bicinchoninic acid. Anal Biochem. 1985;150(1):76-85. Mach H, Middaugh CR. Ultraviolet spectroscopy as a tool in therapeutic protein development. J Pharm Sci. 2011;100(4):1214-1227. Comanducci M, Bambini S, Brunelli B, et al. NadA, a novel vaccine candidate of Neisseria meningitidis. J Exp Med. 2002;195(11):1445-1454. Keiser PB, Biggs-Cicatelli S, Moran EE, et al. A phase 1 study of a meningococcal native outer membrane vesicle vaccine made from a group B strain with deleted lpxL1 and synX, over-expressed factor H binding protein, two PorAs and stabilized OpcA expression. Vaccine. 2011;29(7):1413-1420. Malito E, Biancucci M, Faleri A, et al. Structure of the meningococcal vaccine antigen NadA and epitope mapping of a bactericidal antibody. Proc Natl Acad Sci U S A. 2014;111(48):17128-17133. Johnson S, Tan L, van der Veen S, et al. Design and evaluation of meningococcal vaccines through structure-based modification of host and pathogen molecules. PLoS Pathog. 2012;8(10):e1002981. van der Veen S, Johnson S, Jongerius I, et al. Nonfunctional variant 3 factor H binding proteins as meningococcal vaccine candidates. Infect Immun. 2014;82(3):1157-1163. Cendron L, Veggi D, Girardi E, Zanotti G. Structure of the uncomplexed Neisseria meningitidis factor H-binding protein fHbp (rLP2086). Acta Crystallogr Sect F Struct Biol Cryst Commun. 2011;67(Pt 5):531-535. Malito E, Faleri A, Lo Surdo P, et al. Defining a protective epitope on factor H binding protein, a key meningococcal virulence factor and vaccine antigen. Proc Natl Acad Sci U S A. 2013;110(9):3304-3309.