Influenza vaccines have become established for pro-phylaxis against influenza virus infections caused byseasonal epidemics.1 A typical seasonal vaccine com-prises three viral strains, including two antigens fromtype A influenza–subtypes H1N1 and H3N2 and onefrom influenza type B, and are administered intra-muscularly (IM) using a standard needle and sy-ringe. To respond to the annual changing strains, thevaccine is produced, distributed, and administeredwithin a relatively short period of time of about 6months.2 In the case of a pandemic, the response timeneeds to be as short as possible, making mass produc-
tion of the vaccine followed by distribution and admin-istration a significant time-bound challenge. More-over, because delivery is currently accomplished witha needle and syringe in most cases, administrationof the vaccine is a complex endeavor requiring theparticipation of a large number of skilled profession-als. Hence, there is an important need to expeditethe production of the vaccine, and also to facilitateeasier administration, particularly in the event of apandemic.
A range of novel technologies have the potentialto allow the development of improved influenza vac-cines, including improved immunogenicity and sta-bility, the latter of which could eliminate the require-ment for cold chain management. In the case of apandemic, any approach that could facilitate easierdistribution and administration of the vaccine wouldbe highly advantageous. Many potential technolo-gies are in various stages of R&D, including powderor liquid inhalers, solid, or dissolvable microneedles,
JOURNAL OF PHARMACEUTICAL SCIENCES 1
2 KOMMAREDDY ET AL.
prefilled syringes with needles for intradermal ad-ministration and stabilized influenza vaccine for oraladministration.3–10 Unfortunately, the current stan-dard IM administered vaccines will require signif-icant process changes to enable the vaccines to bedelivered effectively by some of these delivery sys-tems, but some innovations are proceeding. For pul-monary delivery, influenza subunit vaccine was suc-cessfully stabilized into a dry powder by freeze-dryingand spray drying.11,12 While for skin delivery, pre-formed solid coated or dissolvable microneedle arrayswill need antigens at high concentrations to coat orform the tips of the microneedles, in order to facilitateneedle-free administration of vaccine.
Subunit influenza vaccines, including those mar-keted by Novartis V&D, are produced by splitting theharvested virus and purifying the influenza hemag-glutinin (HA). The HA concentrations in the purifiedmonobulks, depending on the virus strain, are usu-ally in the range of 0.2–0.5 mg/mL. We believe thatthis is typical for a range of flu vaccines and alsofor processes described in the literature for the pro-duction of research material. However, it is our ex-perience that most of the novel delivery technologiesavailable require much higher concentrations of HAin the range of 20 mg/mL to enable them to delivera human dose of HA (15:g per strain). In the litera-ture, there have been attempts described to concen-trate influenza antigens by using microfiltration fil-ters from Sartorius Stedim Biotech (Goettingen, Ger-many) and ultrafiltration using Amicon filters fromMillipore (Billerica, MA).13,14 Unfortunately, the mi-crofiltration technique using Sartorius filters has holdup volumes as large as 10 mL making it difficult towork with in a research scale. In addition, concen-tration using Amicon filters also has significant lim-itations in scale up. Tangential flow filtration (TFF)of trivalent influenza vaccine has been previously at-tempted to improve concentration for spray freeze-drying, and this resulted in a total HA concentra-tion of 3.6 mg/mL (individual antigens at 1.2 mg/mL).However, this is not sufficiently concentrated to al-low efficient formulation into most novel delivery sys-tems. Although these approaches have been used toachieve concentrations able to coat low doses of anti-gen for testing in small-animal models, a significantimprovement is still needed to allow the preparationof a delivery system, which targets a full human dose,while retaining the quality of the antigen. In many ap-proaches described in the literature, the concentratedantigen is poorly characterized and there is little or noinformation on the structural and functional stabilityof HA after the concentration process.
To solve the problem described, we have evaluateda range of different concentration technologies thathad the potential to achieve the desired HA levels. Inthis work, we present TFF, as a robust and scalable
process for concentrating influenza vaccine. Impor-tantly, TFF is a well established and readily availableprocess, that can be easily scaled up for manufactur-ing and can be scaled down as needed.15,16 The suc-cess of the concentration process was validated usingan array of techniques, with the stability of HA con-firmed by the functional assay, single radial immun-odiffussion (SRID). In addition, the potency of the con-centrated antigen was confirmed in an in vivo mousemodel by comparison of immune responses againstunprocessed antigen.
MATERIALS AND METHODS
Materials
Seasonal influenza monobulks (purified antigen ofmonovalent strains) A/Brisbane/59/2007 (H1N1), A/Brisbane/10/2007 (H3N2), and B/Florida/4/2006 wereobtained from the manufacturing facility of No-vartis Vaccines and Diagnostics, Rosia, Italy. Thebuffer composition of the monobulk is: sodium chlo-ride 137 mM, potassium chloride 2.7 mM, potassiumdihydrogen phosphate 1.5 mM, and sodium phos-phate dibasic 10 mM. XCell surelock mini-cell gelboxes (EI0001), NuPAGE Novex 4%–12% Bis-TrisGel 1.0 mm 12 well (NP0322BOX), NuPAGE MESSDS 20× running buffer (NP0002), Novex sharppre-stained protein standard (LC5800), colloidal bluestaining kit (LC6025), and StainEase staining trays(NI2400) were obtained from Invitrogen (Grand Is-land, New York). Dithiothreitol (DTT) (20290) wasobtained from Pierce (Rockford, Illinois). Acetoni-trile (1.00030), trifluroacetic acid (1.08262), and hy-drochloric acid (UN1789) were obtained from VWR(Radnor, Pennsylvania). Phosphate buffered saline(PBS) (1×; pH 7.4; KCl 0.20 g/L, KH2PO4 0.20 g/L,NaCl 8.00 g/L, and Na2HPO4 1.15 g/L) (21-031-CM)was obtained from Mediatech (Manassas, Virginia).Zwittergent 3-14 (693017) was obtained from Cal-biochem (Billerica, Massachusetts). SeaKem agarose(50010) and DPBS (17-512F) were obtained fromLonza (Allendale, New Jersey). Glacial acetic acid(100001-74) and Coomassie brilliant blue R250 (3340)were obtained from EMD (Billerica, Massachusetts).Sheep reference antisera anti A/Brisbane/59/2007(AS396) and anti A/Brisbane/10/2007 (AS393-2) wereobtained from TGA (Woden ACT, Australia). Sheepreference antiserum anti B/Florida/4/2006 (07/356)was obtained from NIBSC (Ridge, Hertfordshire, UK).Reference standard antigens A/Brisbane/59/2007 (08/100), A/Uruguay/716/2006 (08/278), and B/Florida/4/2006 (08/140) were obtained from NIBSC. Fil-ter paper (1002 917) was obtained from Whatman(Piscataway, New Jersey). Distilled water (10977-023) was obtained from Gibco (Grand Island, NewYork). Methanol (2018-1GLP) was obtained from BDH
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
CONCENTRATED INFLUENZA VACCINE FOR NOVEL DELIVERY 3
(Radnor, Pennsylvania). Maxisorp plates were ob-tained from Nunc (Rochester, New York), and alkalinephosphatase-conjugated goat anti-mouse IgG was ob-tained from Sigma (St. Louis, Missouri).
Concentration of HA by Tangential Flow Filtration
The influenza monobulks were concentrated by TFFusing the KrosFlo system from Spectrum Labs (Ran-cho Dominguez, California), equipped with a 30-kDcutoff hollow fibers of polyethersulfone at 115 cm2
(Spectrum Labs P-D1-030E-100-01N). The filterswere connected using Masterflex tubing size 14, tothe KrosFlo pump and the digital pressure monitor.At the inlet, outlet, and permeate polysulfone pres-sure transducers were connected either in flow or in aT-junction. The pressure regulator was connected onthe outlet/retentate tubing. The system was checkedfor any leaks, and the hold up volume was mea-sured using 1× PBS. The system was equilibrated for20–30 min in 1× PBS, leaving the permeate line openat flow rate of 60 mL/min. At the end of equilibration,the feed and retentate lines were transferred into thereservoir with the monobulk, and the permeate linewas transferred to a permeate reservoir. Then the per-meate line was opened, and the system was run at anoperating flow rate of 120 mL/min. The permeate lineis left open, and permeate was measured by volume orby weight. The concentration process was continuedat a predetermined transmembrane pressure (TMP)of 8.0 psi. At the completion of the concentration pro-cess, the TMP was slowly reduced to zero, permeateline was closed, and the flow was stopped. The con-centrate material present in the hold up volume wasemptied by reversing the direction of flow. The finalpermeate volume was measured and the retentateconcentration factor was calculated on the basis ofthe volume of starting material. The collected con-centrate material was then characterized for proteinand the HA content.
Determination of Protein Content by BCA Assay
Protein content was determined to assess any lossesduring the concentration process. Even though it doesnot reflect the exact HA, the high sensitivity of theprotein assays was used to understand the concen-tration process. The protein content in the processsamples was determined by colorimetric estimationusing bicinchoninic acid (BCA) reagent with bovineserum albumin as a standard. Briefly, stock solu-tion of 2 mg/mL of albumin was sufficiently dilutedto prepare a set of standards. Working solution of thereagent is prepared by mixing 50 mL of BCA and 1 mLof cupric sulfate provided. For each sample, multipledilutions were prepared in 1× PBS. About 25:L ofthe standards and sample dilutions were transferredto micro plate. All sample dilutions were run in dupli-cate. Working reagent solution was added to the micro
plate at 200:L/well, and allowed to incubate at 37◦Cfor 30 min. At the end of the incubation, absorbanceof the color formed was measured at 562 nm using aVersaMax tunable microplate reader (Molecular De-vices, Sunnyvale, California). Protein concentrationwas calculated by plotting a standard curve with con-centration against absorbance, followed by a linearregression analysis and extrapolating the absorbanceof the samples.
Size-Exclusion Chromatography
To study the aggregation behavior of the protein, sam-ples were analyzed by size-exclusion chromatography(SEC) using Waters Alliance e2695 with UV detec-tion at 214 nm (Waters PDA detector 2996) and aTSKgel G6000PWxl (7.8 mm × 30 cm) 13:m columnwith a pore size of 1000 A (Tosoh Bioscience, GroveCity, Ohio). 1× PBS was used as the running bufferat a flow rate of 0.6 mL/min. The column was equili-brated for 60 min and Bio-rad’s gel filtration standardwas used for the calibration of SEC. The standard vialcontaining a lyophilized mixture of molecular weightmarkers was reconstituted with 0.5 mL of deionizedwater. About 10:L of the standard containing pro-teins of known molecular weight was injected to cali-brate the SEC. The monobulk and TFF samples wererun without dilution, 10:L of retentates and 50:L ofpermeates were injected and run for 30 min at a flowrate of 0.6 mL/min using 1× PBS. The column and thesamples were maintained at room temperature (RT)and 5◦C, respectively.
Gel electrophoresis was used to characterize the anti-gen after concentration process. Though qualitative,the technique helps to confirm degradation of antigenby the presence of stained protein bands of molecularweights lower than that of HA (70 kDa). The reten-tate and permeate from the TFF process were char-acterized by sodium dodecyl sulfate-polyacrylamidegel electrophoresis (SDS-PAGE) and compared withthe unprocessed monobulks that were used as stan-dards at concentrations 50, 100, and 200:g/mL. Thesamples were diluted in 1× PBS to 100:g/mL totalprotein, and the permeates were run undiluted. Thesamples prepared using 4× sample buffer, vortexed,and incubated for 5 min at 90◦C. Samples containing1.0:g (13.3:L) of protein were loaded on to each wellin a 4%–12% Bis-Tris gel and 6:L of a prestainedprotein molecular weight ladder was loaded in lane 1.The gel was run at 200 V for 45 min and then stainedovernight using the colloidal blue staining kit. The gelwas subsequently destained in deionized water to re-move any excess stain and imaged using a calibrateddensitometer (BioRad GS800).
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES
4 KOMMAREDDY ET AL.
Measurement of HA by SRID
Single radial immunodiffussion assay was used toanalyze the stability and integrity of influenza anti-gen following concentration process. Briefly, 1% (w/w)agarose solution was melted at 60◦C, antiserum tospecific HA strains was added, and the solution waspoured into the assay template. The gel was allowedto set and sample wells were created by punchingholes into the gel using vacuum. Dilute samples ofthe retentate and neat samples of the permeate wereused in analysis. Samples were prepared by additionof 50:L of Zwittergent (10%, w/v in water) to 450:Lantigen and incubated for 30 min at RT. Dilutions ofzwittergent-treated antigens were made using PBSpH 7.7 so that the standards and samples are at anapproximate concentration of 50:g/mL of HA to makeserial dilutions. About 20:L of serially diluted sam-ple mixture was loaded into appropriate wells at RT.Once the wells are empty, the assay plates were trans-ferred to humidified 25◦C incubator for 12–24 h. Thegel was allowed to cool and a sandwich was madeof gel, filter paper, and absorbent paper, followed byglass plate for weight. After 30 min, the filter paperwas allowed to dry in cabinet at 37◦C. The filter pa-per was then stained in Coomassie brilliant blue so-lution for at least 5 min. The excess stain was rinsedoff and the immunodiffusion rings were measured todetermine the HA content.17 The concentrates wereanalyzed for the HA content at 1, 2, and 4 week timepoints.
Measurement of HA by Reversed- Phase HPLC
As an alternate to SRID, the HA content in sampleswas also measured using chromatographic method(Waters Alliance e2695) with UV detection (WatersPDA detector 2996) and a Poros R1/10 (2.1 × 100 mm)column. The column was maintained at a tempera-ture of 60◦C, and samples were refrigerated at 5◦C.The method was run at a flow rate of 0.8 mL/min witha total run time of 15 min. The column was equili-brated with 80% mobile phase A, 0.1% TFA in 5%acetonitrile in water and 20% Mobile Phase B, 0.1%TFA in 100% acetonitrile. Mobile Phase B was in-creased to 30% in 2min and then to 35% B at 5.5 min.The column was washed with 100% B for 3.5 min andthen re-equilibrated at 20% B for 4min. The monob-ulks with known HA predetermined by SRID wereused as standards by diluting them in 1× PBS to therange of 3–100:g/mL of HA. The unprocessed or con-centrated monobulks were reduced and analyzed forthe HA content. To 500:L of diluted standard or sam-ple, 25:L of 500 mM DTT was added and the sampleswere incubated at 90◦C for 10 min to reduce the disul-fide bond between influenza HA1 and HA2. 100:L ofeach standard or sample was injected, and the HA1peak was determined by UV detection at 214 nm.18
Immunization Studies in Mice
To assess the effect of TFF concentration on immuno-genicity, groups of 8 BALB/c female mice, 6-week oldwere obtained from Charles River, were immunizedIM on days 0 and 21 with 100:L of antigen solu-tion split evenly between the tibialis anterior of twolegs of each mouse. The concentrated and unprocessedmonobulks were characterized by SRID and diluted tothe required concentrations to administer 0.1:g doseHA antigens. Serum samples of individual mice werecollected for 2 weeks following the first and second im-munizations and were analyzed for hemagglutinationinhibition (HI) and ELISA titers.
Evaluation of Serum Antibody Titers by ELISA and HI
Titration of HA-specific total immunoglobulin G (IgG)was performed on individual/pooled sera collected for2 weeks after each immunization. Maxisorp plateswere coated overnight at 2◦C–8◦C with 0.2:g/wellwith H1N1, H3N2, or B antigens in PBS and blockedfor 1 h at RT with 300:L of 3% poly vinyl pyrolidine.Serum samples and serum standard were initiallydiluted 1:5000–1:20,000 in PBS containing 1% BSAand 0.05% Tween 20, transferred into coated-blockedplates and serially diluted. Antigen-specific IgG wasrevealed with alkaline phosphatase-conjugated goatanti-mouse IgG. Antibody titres are respective dilu-tions that gave an optical density (OD) higher thanthe mean plus five times the standard deviation ofthe average OD obtained in the preimmune sera. Thetitres were normalized with respect to the referenceserum assayed in parallel. Geometric mean titres 2weeks after each immunization were calculated.
The HI assay was carried out on individual/pooledsera collected 2 weeks after each immunization. Toinactivate nonspecific inhibitors in serum samples,aliquots of sera were treated with receptor-destroyingenzyme before being tested with a final serum dilu-tion of 1:10 (starting dilution for the assays). Sampleswere serially diluted twofold into V-bottom 96-wellmicrotiter plates. Briefly, 25:L of twofold serially di-luted samples was incubated with 25:L of strain-specific influenza antigen (whole virus, containingfour hemagglutinin units) for 60 min at RT. A 0.5%(v/v) suspension of red blood cells obtained from adultturkeys were added and the mixture was incubatedfor another 60 min. Reactions were followed throughvisual inspection: a red dot formation indicates a pos-itive reaction (inhibition) and a diffuse patch of cellsindicates a negative reaction (hemagglutination). Asa negative control, serum samples of mice immunizedwith buffer were tested in parallel. Serum responseto vaccine antigens was considered positive if a rise inantibody titers > fourfold compared with backgroundwas detectable. All sera were run in duplicate. The HItiter is defined as the serum dilution in which the last
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
CONCENTRATED INFLUENZA VACCINE FOR NOVEL DELIVERY 5
complete agglutination inhibition occurs. Geometricmean titers of each group are shown.
RESULTS
Concentration of Monobulks
Influenza monobulks were concentrated by a scal-able TFF process, a hollow fiber filter was connectedvia the inlet and retentate lines to the reservoir asshown in Figure 1. The monobulks from 2008 to 2009seasonal vaccine were successfully concentrated al-most 50–80 fold by pumping through the hollow fil-ters at a set rate 120 mL/min within a shear range of3000–5000 s−1 at a predetermined TMP. A separateexperiment was performed to determine the optimalTMP for the TFF process by increasing the pressurestepwise from 2 to 12 psi at a constant inlet flow rate of20 mL/min. The resulting filtrate flux rate was plottedagainst TMP and the optimal TMP was determinedto be 8 psi. Throughout the concentration process, theTMP was maintained at this optimal rate. All threemonobulks from 2008 to 2009 seasonal vaccine wereconcentrated and further characterized.
Characterization for Protein and the HA Content
After the TFF process, the retentate and permeatewere collected and characterized. Each strain wasconcentrated individually to different extents. Thetotal protein was estimated by BCA assay–the B/
Figure 1. Line diagram of the TFF process.
Figure 2. Single radial immunodiffusion assay images forconcentrated B/Florida/4/2006 antigen and permeate alongwith the NIBSC standard. HA content was estimated bythe size of the immunodiffusion ring formed by comparisonagainst that of NIBSC standards.
Florida/4/2006 monobulk at 0.57 mg/mL was concen-trated to 28.7 mg/mL, A/Brisbane/59/2007 (H1N1) at0.58 mg/mL was concentrated to 46.0 mg/mL, and A/Brisbane/10/2007 (H3N2) at 0.39 mg/mL was concen-trated to 19.5 mg/mL (Table 1). The total proteincontent in permeate for all strains was found to be≤20:g/mL. The concentrate and permeate for eachstrain were analyzed for the HA content by SRID. Forthe three strains, HA was concentrated 50–80 fold: Bfrom 0.49 to 24 mg/mL, H1N1 from 0.34 to 28 mg/mL,and H3N2 from 0.24 to 14 mg/mL (Table 1). As can beseen from the immunodiffussion ring images of SRIDanalysis (Fig. 2), the HA proteins were concentratedin the retentate while it is only the non-HA proteins/peptides that are separated in permeate. The over-all recovery of the TFF process was found to be 80%based on an SRID analysis.
Reversed-phase HPLC (RP-HPLC) analysis wasused as an alternative method for the quantificationof HA. In this assay, we used SRID-certified monob-ulks as standards for an HA analysis of each strain.The samples were treated with DTT, and the HApeaks were chromatographically separated by RP-HPLC. Within the concentration ranges tested, allthree strains were found to have a linear responsein peak area. For the three strains analyzed, the HAwas found to elute at 2.9 min for B, 3.5 min for H1N1,and 3.2 min for H3N2 (Fig. 3). The individual concen-trates from the TFF process were analyzed by RP-HPLC and the estimated HA values were found to beat 38.6 mg/mL for antigen B, 34.7 mg/mL for H1N1,and 20.08 mg/mL for H3N2.
Table 1. Characterization of Concentrated Antigens from the 2008/2009 Seasonal Influenza Vaccine
Protein Content (mg/mL) Average ± SD HA Content by SRID (mg/mL) Average ± SD
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES
6 KOMMAREDDY ET AL.
Figure 3. Overlay of chromatographs for individual un-processed influenza antigens analyzed by RP-HPLC withUV detection. The HA peaks were separated on Poros R1/10 (2.1×100 mm) column using a gradient with 0.1% (v/v)TFA in 5% (v/v) acetonitrile as mobile phase A and 0.1%(v/v) TFA in 100% acetonitrile as B run at 0.8 mL/min.
Size-Exclusion Chromatography
The concentrates and permeates from the TFF pro-cess were analyzed by SEC and compared againstunprocessed monobulks as shown in Figure 4. Forall three strains, the monobulk elutes over multi-ple peaks. The earlier peaks 12.9 and 15.2 min for B/
Florida, 12.7 min for H1N1, and 12.8 and 17.3 min forH3N2 were retained in the concentrates and testedpositive for the HA proteins by SRID. These proteinswere found to elute after the void volume (9.5 min)but before the highest molecular weight standard ofbovine thyroglobulin, 670 kDa (15.5 min). Hence, theapparent molecular weights of the proteins in eachof the monobulks could not be estimated by SEC. Incase of the concentrated antigens in the retentate,there was no observed aggregation and the retentiontimes were comparable to that of the HA peaks inthe unprocessed monobulk. However, there are ob-served relative differences in peak areas/intensitiesbetween unprocessed monobulk and concentrates. Inthe TFF process, the material corresponding to theearlier peaks is concentrated, whereas the 19.7 minpeak present in the monobulks was eluted in thepermeate and was confirmed to be non-HA proteins/peptides by SRID.
The TFF samples were analyzed by gel electrophore-sis and compared with the unprocessed monobulks(Fig. 5). In each of the standards, for all three strainsof antigens run at 2.0, 1.0, and 0.5:g, a band was
Figure 4. Size-exclusion chromatographs of TFF samples. The unprocessed monobulk, con-centrate, and permeate of each strain were separated by SEC on a TSKgel G6000PWxl columnusing PBS as the mobile phase.
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
CONCENTRATED INFLUENZA VACCINE FOR NOVEL DELIVERY 7
Figure 5. Gel electrophoresis of TFF samples. The unpro-cessed monobulk controls, the concentrated aliquots, andthe permeates from the TFF were analyzed by gel elec-trophoresis on 4%–12% Bis-Tris gels.
observed at 70 kDa that corresponds to the molecularweight of HA. The concentrated antigens ran simi-lar to that of the monobulks, and there were no ad-ditional degradation fragments observed. In addition,multiple more slowly migrating bands were observed,which presumably correspond to oligomers that re-mained associated through the electrophoresis. Therewere no detectable HA proteins in the lanes loadedwith permeate.
Stability of Concentrated Monobulks
Aliquots of the retentate for each of the concentratedantigens were stored in glass vials at 4◦C were an-alyzed for the HA content by SRID at 1, 2, and 4week time points. At the end of 4 weeks, B/Floridaantigen was found to have HA activity in the rangeof 24 mg/mL retaining almost 100% of the activityand H1N1 showed HA activity of 29.1 mg/mL retain-ing 100% activity, while H3N2 showed reduced HAactivity at 12.1 mg/mL retaining about 85% activity(Fig. 6). Beyond 4 weeks, the concentrated antigenswere visibly aggregated with reduced antigenic con-tent by SRID.
Immunogenicity of Concentrated Antigens In Vivo
To demonstrate that the TFF concentration processdoes not impact the antigen immunogenicity, micewere immunized IM with either unprocessed controlantigens or with concentrated antigens. Formulationscontaining individual antigens were prepared and ad-ministered at 0.1:g dose IM. The HI titers were ana-lyzed for sera samples collected. The geometric meantiters of the concentrated antigens were comparableto that of the unprocessed control antigens when ad-ministered IM (Fig. 7). In case of H1N1, the control
Figure 6. Stability of concentrated antigens by SRID.Samples of concentrated antigens were stored in glass vialsat 4◦C and analyzed for HA content by SRID at time points1, 2, and 4 weeks. The immunodiffusion ring formed inSRID analysis is measured to calculate percent activity re-tained. Results are shown as an average of two measure-ments plus% relative standard deviation for each strain.
Figure 7. Serum HI antibody titers against A/Brisbane/59/2007 (H1N1), A/Brisbane/10/2007 (H3N2), and B/Florida/4/2006 strains 2 weeks after the second immuniza-tion. Mice, eight for each group, were immunized twice in-tramuscularly with 0.1:g of individual monobulks unpro-cessed control or concentrated by TFF. Results are shownas geometric mean titers plus standard deviation.
antigen resulted in titers of 99 and the concentratedat 95. In case of H3N2, the control antigen resultedin titers of 125 and concentrated antigens at 116. Incase of antigen B, the control antigen resulted in titersof 141 and concentrated antigens at 73. All the anti-gens tested resulted in seroconversion with resultingHI titers >40, and there were no significant differ-ences in immune response between the unprocessedand concentrated antigens.
Following immunization, the sera samples wereanalyzed for HA-specific immunoglobulin titers byELISA. All three concentrated influenza antigens re-tain their immunogenicity (Fig. 8). The IgG titersin case of antigen H1N1, the unprocessed controlantigen resulted in 3028 and concentrated antigenat 2471. In case of H3N2, the unprocessed controlantigen resulted in 4870 and concentrated antigen at3368. In case of B, the unprocessed control antigen re-sulted in 3403 and concentrated antigen at 2817. Allthe concentrated antigens resulted in antibody titerscomparable to that of unprocessed control antigensand there were no significant differences observed.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES
8 KOMMAREDDY ET AL.
Figure 8. Total IgG ELISA titers against A/Brisbane/59/2007 (H1N1), A/Brisbane/10/2007 (H3N2), and B/Florida/4/2006 strains two weeks after the second immunization.Mice, eight for each group, were immunized twice intramus-cularly with 0.1:g of individual monobulks unprocessedcontrol or concentrated by TFF. Results are shown as geo-metric mean titers plus standard deviation.
DISCUSSION
Recent advances in novel delivery system technolo-gies have led to a renewed interest in reformulat-ing influenza vaccines to facilitate easier delivery oraccess.3,4,19 Both seasonal and pandemic influenzavaccinations could potentially benefit significantlyfrom these innovations in delivery technologies. Al-though the cell-culture vaccine process is now estab-lished and already speeds up the production of in-fluenza vaccine, delivery systems such as micronee-dles could help in increasing the speed of distributionand enable administration with minimal requirementof trained personnel. Moreover, the possibility of self-administration of vaccines could eventually revolu-tionize vaccine distribution and enable the concept ofavailability of vaccines through mail order or over thecounter. However, from a manufacturing perspective,the reformulation of vaccines into novel delivery sys-tems such as microneedles requires significant pro-cess changes. Although there would be many chal-lenges and barriers to fully enable this concept, theability to prepare stable solutions of highly concen-trated HA monobulks for further processing would bea significant step forward.
In this work, after evaluation of a range of possibleconcentration techniques, we propose TFF as a simpleand scalable process for concentration of HA. The opti-mized TFF process using 115 cm2 filters enabled con-centration of monobulk volumes ranging from 200 mLto 4 L and was used for a mid-scale process in re-search. Using TFF, we have demonstrated that theHA concentrations as high as 14–28 mg/mL, almost50–80 fold higher than the monobulks were achieved.In the TFF process, the feed flow moves tangential tothe filter membrane in the hollow fibers. TFF in com-parison to “dead-end filtration” is more of a continu-ous operation with no filter cake formation. Above all,the technology is easily scalable to concentrate large
batches in manufacturing; at the same time, the avail-ability of smaller size filters (surface area of 8 cm2)with hold up volumes as low as 1 mL facilitates effec-tive use of the process in research scale. With little orno further optimization, the TFF process was broadlyapplicable for concentrating all the strains of egg-derived influenza antigens from 2008/2009 and 2011/2012 seasonal vaccine, and cell-culture-derived anti-gens from 2008/2009 and 2011/2012 seasonal vaccine(data not shown). Although the TFF process workedfor three different strains of seasonal monobulks, themaximum HA concentration achieved through TFFconcentration was strain specific. For 2008/2009 egg-derived antigens, the H3N2 monobulk could not beconcentrated as much as the B and H1N1 strains.This could be due differences in purity, hydrophobic-ity, nature of HA, and rosettes or soluble aggregates ineach of the strains. The TFF process could be furtheroptimized to narrow down the differences. The HAprotein with molecular weight 70 kDa was retainedin the concentrate using filters of cutoff size 30 kDa,and smaller non-HA proteins were present in the per-meate. We also evaluated filters with cutoff sizes up to100 kDa that could retain the HA proteins (data notshown), which can be explained by the nature of theHA proteins that form trimers and rosettes of highermolecular weights.
The in depth antigen characterization that was un-dertaken here confirmed the stability of HA duringand after the TFF process. Analysis by SRID indicatedthat the structural and functional integrity of anti-gens was maintained after the TFF process, whichwas run at a shear range of 3000–5000 s−1. SRIDresults for samples from the TFF process also indi-cated that HA was present in the retentate and thatthe HA proteins maintained their functional activ-ity after processing. The binding of the antigen withthe strain-specific antibodies was not affected follow-ing concentration by TFF and confirmed the stabil-ity of the antigens. Upon concentration, there wasno observed aggregation behavior by SEC. The ob-served relative differences in peak areas or intensi-ties in the concentrated monobulks by SEC could bedue to increase in soluble aggregates. The HA of fluantigen generally exists in trimeric form, and has ex-posed hydrophobic regions from the extraction frominfluenza viruses during manufacturing. Because ofthe exposed hydrophobic regions the HA exists assoluble aggregates of HA often called rosettes. Thenumber of trimers in each rosette is dependent onthe concentration of protein and detergent. The ab-sence of low- molecular-weight protein bands fromdegradation of HA in SDS-PAGE also provided ad-ditional confirmation that the structure and stabil-ity of the concentrated antigens were maintainedthroughout the TFF process. The concentrated anti-gens in solution were stable for up to 4 weeks without
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
CONCENTRATED INFLUENZA VACCINE FOR NOVEL DELIVERY 9
further modification. This allows sufficient time forfurther formulation with different vaccine technolo-gies and minimizes the need for additional unit op-erations such as lyophilization in the manufacturingprocess.
In addition to SRID analysis, the HA contentwas also measured by RP-HPLC assay, which useschromatographically separated peak of cleaved HA1for quantification.18 The RP-HPLC results also con-firmed that HA was concentrated in the retentate.For all strains, the HA quantified by RP-HPLC washigher than the HA measured by SRID, though thedifference varied depending on the strain. This dif-ference is likely due to the fact that SRID measurestotal HA based on the diameter of the immunodiffu-sion ring formed from interactions of HA with the an-tibodies, whereas in RP-HPLC only the cleaved HA1is quantified.
In the in vivo studies, we showed that all three con-centrated antigens retained their immunogenicity. IMadministration of the concentrated antigens resultedin comparable HI titers and ELISA IgG titers to theunprocessed control antigens. The HI titers showedstrain-specific induction of immune response afterimmunization with concentrated antigens and therewere no significant differences in immune responsesgenerated by unprocessed antigen and concentratedantigen. The concentrated individual antigens werealso evaluated in vivo in the presence of the MF59adjuvant as a trivalent vaccine, to amplify any dif-ferences in immunogenicity, but none were seen. Theconcentrated antigen was equally immunogenic to thecontrol antigens in the presence of the adjuvant forboth HI and ELISA titers (data not shown).
Overall, we have shown that multiple strains ofinfluenza monobulk from different sources (egg de-rived and cell culture derived) can be concentrated byTFF to achieve very high HA concentrations, whichfacilitate the subsequent formulation of HA into var-ious novel delivery approaches. Reproducibility ofconcentration with minimal optimization is criticalas the antigens change each year depending on theprevalent strains. The process was found to be ro-bust and reproducible for strains from different sea-sons and different sources. The concentrated anti-gens from TFF process, retained their potency andimmunogenicity. Although we cannot guarantee thatthis approach would work for all influenza antigens,we have shown that it is sufficiently robust to work formaterials derived from two very different processes.Therefore, we recommend that the approach that wedescribe here is a valid starting point on which todevelop a similar process for all influenza antigens.Moreover, we believe that this approach could be uti-lized for other vaccine antigens to achieve the desiredconcentrations for novel delivery applications. Our ex-perience in this area and our communications with
fellow researchers has highlighted that the availabil-ity of antigens of sufficiently high concentration foreasy use in novel delivery applications is a significantproblem in the field. We believe that we have high-lighted a potentially broadly applicable solution.
CONCLUSIONS
We have described the use of a scalable and robustprocess to prepare highly concentrated influenza anti-gens for subsequent easy use in novel delivery ap-plications. Importantly, we have demonstrated thatthe stability and immunogenicity of the antigensare maintained throughout the process. For antigenswhose nature and purity levels change every year, aprocess that could be adapted with very little opti-mization is needed. Most importantly, a process thatdoes not affect the stability of influenza antigens isrequired. This paper for the first time takes into con-sideration all of these challenges in establishing aprocess for concentration of influenza antigens. Thecharacterization and stability data generated, alongwith the systemic immune response in mice clearlyestablish TFF process for concentration of influenzaantigens. This scalable technique could help processlarge batches of antigens without affecting the time-lines for the development of seasonal or pandemic in-fluenza vaccine. Moreover, we believe that this pro-cess could be adapted to a wide range of alternativeantigens to facilitate their use in novel delivery appli-cations.
REFERENCES
1. Francis T, Magill TP. 1937. The antibody response of humansubjects vaccinated with the virus of human influenza. J ExpMed 65(2):251–259.
2. World Health Organization. WHO Media Centere influenzafact sheet at: http://www.who.int/mediacentre/factsheets/fs211/en/.
4. Dormitzer P, Tsai T, Del Giudice G. 2012. New technologiesfor influenza vaccines. Hum Vaccin Immunother 8(1):45–58.
5. Leroux-Roels I, Vets E, Freese R, Seiberling M, Weber F,Salamand C, Leroux-Roels G. 2008. Seasonal influenza vac-cine delivered by intradermal microinjection: A randomisedcontrolled safety and immunogenicity trial in adults. Vaccine26(51):6614–6619.
6. Van Damme P, Oosterhuis-Kafeja F, Van der Wielen M,Almagor Y, Sharon O, Levin Y. 2009. Safety and efficacy ofa novel microneedle device for dose sparing intradermal in-fluenza vaccination in healthy adults. Vaccine 27(3):454–459.
7. Ambrose CS, Luke C, Coelingh K. 2008. Current status oflive attenuated influenza vaccine in the United States for sea-sonal and pandemic influenza. Influenza Other Respi Viruses2(6):193–202.
8. Monto AS, Ohmit SE, Petrie JG, Johnson E, Truscon R, TeichE, Rotthoff J, Boulton M, Victor JC. 2009. Comparative efficacyof inactivated and live attenuated influenza vaccines. N EnglJ Med 361(13):1260–1267.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES
10. Sullivan SP, Koutsonanos DG, Del Pilar Martin M, Lee JW,Zarnitsyn V, Choi SO, Murthy N, Compans RW, Skountzou I,Prausnitz MR. 2010. Dissolving polymer microneedle patchesfor influenza vaccination. Nat Med 16(8):915–920.
11. Amorij JP, Huckriede A, Wilschut J, Frijlink HW, HinrichsWL. 2008. Development of stable influenza vaccine pow-der formulations: Challenges and possibilities. Pharm Res25(6):1256–1273.
12. Saluja V, Amorij JP, Kapteyn JC, de Boer AH, Frijlink HW,Hinrichs WL. 2010. A comparison between spray drying andspray freeze drying to produce an influenza subunit vac-cine powder for inhalation. J Control Release 144(2):127–133.
13. Raphael AP, Prow TW, Crichton ML, Chen X, Fernando GJ,Kendall MA. 2010. Targeted, needle-free vaccinations in skinusing multilayered, densely-packed dissolving microprojectionarrays. Small 6(16):1785–1793.
14. Zhu Q, Zarnitsyn VG, Ye L, Wen Z, Gao Y, Pan L,Skountzou I, Gill HS, Prausnitz MR, Yang C, Compans RW.
2009. Immunization by vaccine-coated microneedle arrays pro-tects against lethal influenza virus challenge. Proc Natl AcadSci USA 106(19):7968–7973.
15. Van Reis R, Goodrich EM, Yson CL, Frautschy LN,Dzengeleski S, Lutz H. 1997. Linear scale ultrafiltration.Biotechnol Bioeng 55(5):737–746.
16. Van Reis R, Zydney A. 2001. Membrane separations inbiotechnology. Curr Opin Biotechnol 12(2):208–211.
17. Schild GC, Wood JM, Newman RW. 1975. A single-radial-immunodiffusion technique for the assay of influenza haemag-glutinin antigen. Proposals for an assay method for thehaemagglutinin content of influenza vaccines. Bull WorldHealth Organ 52(2):223–231.
18. Kapteyn JC, Saidi MD, Dijkstra R, Kars C, Tjon JC,Weverling GJ, de Vocht ML, Kompier R, van Montfort BA,Guichoux JY, Goudsmit J, Lagerwerf FM. 2006. Haemag-glutinin quantification and identification of influenza A&Bstrains propagated in PER. C6 cells: A novel RP-HPLCmethod. Vaccine 24(16):3137–3144.
19. Hegde NR, Kaveri SV, Bayry J. 2011. Recent advances in theadministration of vaccines for infectious diseases: Micronee-dles as painless delivery devices for mass vaccination. DrugDiscov Today 16(23–24):1061–1068.
JOURNAL OF PHARMACEUTICAL SCIENCES DOI 10.1002/jps
