New technologies for influenza vaccines

Abstract

Influenza vaccine preparations have been administered to humans since the late 1930s,¹ and the diversity of approaches in licensed trivalent seasonal or monovalent pandemic products is unparalleled by vaccines against any other target. These approaches include inactivated whole virus vaccines, detergent or solvent “split” vaccines, subunit vaccines, live attenuated vaccines, adjuvanted vaccines, intramuscular vaccines, intradermal vaccines, intranasal vaccines, egg-produced vaccines and mammalian cell culture-produced vaccines. The challenges of influenza immunization, including multiple co-circulating strains, antigenic change over time, a broad age spectrum of disease, and the threat of pandemics, continue to drive the development of new approaches. This review describes some of the new approaches to influenza immunization that are the subjects of active research and development.

Keywords: :

• manufacture
• pandemic
• subunit
• technology
• vaccine
• virus-like particle

New Vaccine Platforms

Influenza vaccine manufacturing is unique in the production of biologicals because processes must be adjusted on an almost annual basis to allow the vaccine strain changes that track ongoing evolution of the virus. These process adjustments must take place in a compressed time frame because (for the northern hemisphere) each year’s strain selection is announced in February, and vaccine is typically shipped by manufacturers starting in July or August. There is insufficient time for a new technical and clinical development program each year. Therefore, influenza vaccine manufacturing and testing programs must be based on platforms that are flexible enough to accommodate adjustment for strain changes but also consistent and predictable enough to allow rapid production and regulatory approval of each year’s vaccines.

The requirement for speed is even greater during pandemics and is coupled with a need for surge capacity to protect all those at risk. The response to the 2009 H1N1 influenza pandemic was at once impressive and disappointing. A pandemic was declared on June 11, 2009.² Vaccines were first shipped approximately three months later, in early September. Yet, both the first pandemic wave (in the Northern Hemisphere spring) and the second pandemic wave (in the Northern Hemisphere fall) had peaked and begun to decline before substantial quantities of vaccine were available in developed countries, and most of the world’s population never had access to a pandemic vaccine. The necessity to do better next time, when the pandemic strain could prove more virulent, has been recognized by both manufacturers and governments and has encouraged a wave of innovation in influenza vaccine manufacturing. The 2009 pandemic also highlighted the constraints on implementation of innovation. Several new manufacturing technologies, such as virus-like particles (VLPs) and recombinant HA subunits, were in development at the time of the outbreak. Manufacturers of innovative influenza vaccine candidates scrambled to develop novel pandemic vaccines before those produced by the methods used for licensed seasonal vaccines were distributed. Yet, the pandemic vaccines to be licensed and distributed were those based on well-established platforms already in use to produce seasonal influenza vaccine, with the exception of a whole virus inactivated vaccine produced in cell culture (Celvapan^®, Baxter). That vaccine, based on a combination of technologies that had previously been used for seasonal vaccine production (cell culture production and whole virus inactivation), was introduced during the pandemic, without an already licensed seasonal equivalent, having achieved EU mock up licensure for an H5N1 pre-pandemic vaccine in advance of the H1N1 pandemic.

Why did the more innovative vaccine candidates not contribute to the 2009 pandemic response? It is not because the innovative approaches lack merit. Rather, regulatory processes are not equipped to evaluate and license new technologies as rapidly as would have been needed for innovative new vaccines to be deployed, and new production capacity takes years to establish. Practically, new manufacturing processes for pandemic vaccines must be developed, licensed, and used repeatedly for seasonal vaccines to establish tested and trusted production processes, manufacturing facilities, and regulatory pathways that can also be used for pandemic equivalents. A seasonal influenza vaccine that cannot be produced reliably with up to three new strains per year is not viable. Because conventional vaccine manufacture has proven itself able to deliver influenza vaccines each year despite strain changes (although in some years supplies have been inadequate),³ the market is unlikely to tolerate an influenza vaccine that is only available on time during years when the selected strains happen to be compatible with the manufacturing process. We will briefly consider several new platforms in this context—mammalian cell culture-based vaccines, novel influenza virus seed production techniques, VLP-based vaccines, recombinant subunit vaccines, and live attenuated vaccines.

Fully Mammalian Cell Culture-Based Influenza Vaccines

A recent review has detailed many aspects of cell culture-based influenza vaccine manufacturing.⁴ Mammalian cell culture-based pandemic and seasonal vaccines are available in many European and other countries (Celtura^® and Optaflu^® from Novartis and Celvapan^® and Preflucel^® from Baxter) but not, as of the writing of this review, in the US. To date, clinical trials of cell culture-based vaccines have shown equivalent performance over-all to egg-based vaccines.^5-15 There are pre-clinical reasons to believe that improvements to the vaccine seed virus generation process could lead to clinical improvements in cell culture-based vaccines.

At present, mammalian cell culture-based seasonal vaccines use influenza viruses that have been adapted to growth in eggs as their “seeds” (the influenza viruses used to inoculate eggs or cultures to generate vaccine antigens). When an influenza virus strain obtained from a patient’s respiratory secretions is isolated by a WHO-linked laboratory for use as a vaccine seed precursor, it is first adapted to growth in eggs. As it adapts to replication in eggs, the virus usually acquires a mutation in one of its most antigenically important regions—the region around the sialoside receptor binding site.^16-19 This mutation occurs because the virus adapts from binding the dominant sialosides in the human respiratory epithelium (α-2,6 sialosides) to binding the sialosides that dominate the surface of the avian allantoic cavity (α-2,3 sialosides).²⁰ Because some mammalian cells, such as MDCK cells, bear both α-2,3 sialosides and α-2,6 sialosides, egg-adaptive mutations do not generally revert during subsequent mammalian cell culture passage.^19,21 The egg-adaptive mutations can decrease the antigenic match between circulating viruses and the virus used as a vaccine seed.^16,17 We do not know how great an impact this mismatch has on vaccine efficacy in humans. In addition, a higher proportion of human influenza isolates can be isolated in mammalian cells, such as MDCKs, than in eggs, especially for H3N2 strains. Therefore, most virus isolation for disease surveillance is performed in mammalian cells.^22,23 In the 2003–4 influenza season, a season marked by unusually high pediatric death rates, no matching, well-growing H3N2 strain that had been isolated in cultured cells could be re-isolated in eggs in time, leading to a vaccine mismatch and low vaccine effectiveness.^24,25

Vaccine manufacturers and WHO-associated laboratories are now collaborating in an effort to eliminate the egg adaptation step from the process of preparing the viruses that will be provided to manufacturers of influenza vaccines produced in mammalian cells.²² Many questions are being asked. Can the reagents used to assess hemagglutinin (HA) content in mammalian cell-based influenza vaccines be the same as those used to assess HA content in egg-based influenza vaccines? Must additional adventitious agent testing be done for influenza vaccine seeds isolated in mammalian cell substrates? Which cell line should be adopted by WHO-affiliated laboratories for virus isolation and propagation? What quality standards should be applied to virus isolation on mammalian cells? As these questions are answered, it is likely that fully mammalian cell-based influenza vaccines will become available in the coming years. We do not yet know whether such vaccines will be more efficacious in ordinary years, when an all-mammalian cell-based vaccine may be better matched to circulating strains by a single amino acid identity. We also do not know whether there will be a year in which no well-matched vaccine can be produced in eggs but a well-matched vaccine can be produced in mammalian cells, producing a potentially larger gap in efficacy. Mammalian cell isolation and amplification of candidate strains could actually improve the strain match of egg-based vaccines, because amplifying clinical isolates by growth in mammalian cells before inoculation into eggs could raise the likelihood of successful adaptation of the virus to growth in eggs by a simple stochastic increase in the probability of a favorable egg-adaptive mutation occurring in a larger virus population.

Recombinant Subunit-Based Influenza Vaccines

Although all the vaccines licensed and deployed during the 2009 pandemic were derived from processed influenza virions, alternative candidates based on recombinant technologies were also produced, at least at pilot scale. For example, the chief influenza neutralization target, HA, was expressed in insect cells by Protein Sciences. Baculovirus-expressed, recombinant, uncleaved HA of avian and human seasonal influenza strains has been studied extensively in humans, and the trivalent seasonal formulation is under consideration for licensure in the US. The vaccine contains three times (45 µg) the usual quantity of HA for each of the three subtypes and has been highly immunogenic in young and older adults but not in children.^26-29 A field efficacy trial in healthy young adults during the 2007–08 influenza season demonstrated an efficacy of 44.6% against laboratory-confirmed CDC-defined influenza-like illness, which, in the context of a significant mismatch between strains contained in the vaccine and circulating viruses that season, was within expectations and consistent with the reported efficacy of mismatched conventional trivalent inactivated influenza vaccine (TIV).³⁰ Although the efficacy trial is the first to demonstrate that an HA-only vaccine is sufficient to protect against infection and disease, formulations containing recombinant NA are in development. An H5N1 vaccine also was shown to be immunogenic and primed for responses to an antigenically divergent H5N1 subunit vaccine 8 y later.³¹

The manufacturing strategy for the Protein Sciences recombinant subunit candidates is based on extraction of full length influenza HA from the membranes of recombinant baculovirus-infected insect cells followed by protein purification.³² Although producing just the HA ectodomain would simplify purification by eliminating the hydrophobic transmembrane region (which can mediate aggregation) HA does not reliably form trimers unless anchored into a membrane-bound, two-dimensional array during protein maturation.³³ Trimeric HA is considered desirable because some neutralizing epitopes are formed at the HA trimer interface, and HA trimers are likely to be more stable and immunogenic than HA monomers.^34-36 For research purposes, HA ectodomain trimers are often produced by tethering them together with C-terminal trimerization tags (such as foldon tags). For many but not all strains (A/California/04/2009 being a notable exception), HA trimers that form with the assistance of these tags remain stably associated after the tags are cleaved from the oligomerized proteins.^37,38 A constrained trimeric ectodomain approach has not been pursued in vaccine candidates that have entered clinical trials to date. In part, this is due to concerns about the effects of an immune response against the tag, the burden of demonstrating complete tag removal during purification, and the need for a consistent method of determining precisely how each HA gene will be modified by trimerization tag addition. Although retaining the transmembrane region of a full length recombinant HA complicates purification, it has the potential advantage that the resulting hydrophobic patch mediates the association of HA into rosettes, similar to the HA rosettes in conventional influenza vaccines. The multivalency of rosettes enhances immunogenicity.³⁵

A key challenge to the recombinant subunit approach is that changes in protein sequence can alter chromatographic behavior, and there is insufficient time for significant adjustment of purification protocols during seasonal strain changes or pandemic responses. Significant changes in vaccine production processes to accommodate changes in chromatographic behavior could trigger a requirement for new clinical trials to ensure that product performance is unaltered. A modern solution to the purification problem would be to use recombinantly-appended affinity tags to render purification essentially strain independent. Yet, affinity tags have not been used in the production of licensed human vaccine antigens, in part because of concerns similar to those that have prevented the use of trimerization tags, concerns about affinity ligands leaching from columns, and the expense, variability and limits to regeneration of affinity columns.³²

Alternatively, recombinant HA “heads” can be expressed in E. coli with yields of grams per liter. The heads are solubilized from bacterial inclusion bodies with denaturants and then refolded. With proper conditions, a soluble, properly folded domain can result.^39-41 With some constructs and conditions, such refolded heads are monomeric, but there is also evidence that other constructs and conditions may lead to oligomerization or anomalous inter-subunit disulfide bond formation.^40,41 The receptor-binding HA head is a compact domain that displays the bulk of known neutralizing epitopes, but not the broadly heterotypic neutralizing epitopes found on the HA stem region or, unless “properly” trimeric, the neutralizing epitopes that span the inter-subunit interfaces of intact HA trimers.^34,42,43

VaxInnate has advanced this technology in candidates that fuse the recombinant HA heads to Salmonella typhimurium type 2 flagellin, a potent TLR-5 agonist, which compensates for the otherwise lower immunogenicity of these stripped-down antigens.⁴⁰ In young adults, doses of 0.5, 1 and 2 μg of a monovalent A/Solomon Islands/3/2006 (H1N1) HA-fusion protein candidate induced geometric mean hemagglutination inhibition (HI) titers against egg-grown virus of 587, 698 and 640 (1 mo after vaccination), and these remained elevated above 300 for 6 mo after vaccination.⁴⁵ A dose ranging study in older adults with a mean age of 72 y found that a single 5 or 8 µg dose of monovalent vaccine led to a HI geometric mean titer of 211–222 with 98% of those subjects achieving seroprotective titers and 75% seroconverting, indicating that a higher antigen content is needed for that age group.⁴⁴ The geometric mean antibody titers in this older age group were similar to or higher than those obtained with the licensed unadjuvanted high dose TIV. The vaccine was well tolerated in both young and older adults. Although C-reactive protein was elevated 2–10-fold in a dose-dependent manner in the young adults, in general, elevated levels were not correlated with reported reactogenicity or with elaboration of inflammatory cytokines.

During the 2009 pandemic, several constructs with different fusions between the HA head and flagellin were made and shown to be immunogenic in mice.³⁹ If such a bacterially-expressed influenza vaccine were licensed, the bulk antigen could potentially be produced in sufficient quantity to actually cover the global population in a pandemic (provided that sufficient filling and distribution capacity existed). On the other hand, there is yet to be a demonstration the system can be adapted rapidly enough to strain changes to produce trivalent influenza vaccines reliably. The very high starting purities that can be obtained from an E. coli inclusion body preparation simplify but are unlikely to eliminate the need for optimization of purification procedures after refolding.

Recombinant VLP-Based Influenza Vaccines

VLP-based vaccines are recombinant equivalents to authentic virus-based vaccines (less the genome and often some internal virion proteins). VLPs can generally be purified by equivalent methods to those used to purify viruses, but the advanced VLP-based influenza vaccine candidates are whole particle vaccines rather than split vaccines or particle-derived subunits. VLPs form spontaneously when HA and M1 (the matrix protein) are co-expressed in a cell, with or without NA.^46,47 Formulation in a lipid bilayer may help maintain HA and NA stability and conformation in more sensitive strains, and the mounting of HA on a particle probably enhances its immunogenicity.⁴⁸ An advantage of VLPs is that the virus used to express the proteins that assemble into particles is not an influenza virus but rather a recombinant vector, such as the baculovirus used to infect insect cells in the vaccine candidates produced by Novavax or the Agrobacterium tumefaciens bacterial vector used to infect tobacco leaves in the candidates produced by Medicago.^46,47 These recombinant vectors do not depend on the function of the expressed influenza HA and NA for their replication. Thus, the growth of the vectors in plants or cultured cells should not vary much with the strain of the influenza antigens expressed. Because VLP purification with density gradients could be relatively strain independent (analogous to whole virus purification), a well-designed VLP production strategy could have advantages over a recombinant HA subunit approach (which requires chromatographic separation) in meeting the challenge of robustness to strain change. On the other hand, the stability and relative simplicity of HA trimer subunits could potentially provide advantages over the VLP approach in consistency from batch to batch and with storage.

The Novavax VLP-based vaccine has progressed to Phase II human trials.^49,50 A large trial of 2537 young adults receiving pandemic H1N1 vaccine showed that a 15 µg (HA) formulation stimulated seroprotective responses in 64–85% of seronegative subjects after one dose.⁵⁰ Dose-related trends for increased reactogenicity for formulations from 5 to 45 µg were reported. A detailed Phase I study of a H5N1 VLP showed that 90 µg of HA from A/Indonesia/05/2005 (H5N1 clade 2.1) elicited 4-fold rises in neutralizing antibodies in 76% of subjects after two doses, including antibodies cross-reactive to viruses in other clades. Neutralizing antibody titers were correlated with preferential binding to the oligomeric (vs. monomeric) form of recombinant HA1, which may present epitopes on virion spikes absent from monomeric forms.⁴⁹ That specificity is hypothesized to be important in providing cross-clade antibody responses. Anti-NA antibodies that were mapped to epitopes near the sialic acid binding site potentially could provide a functional benefit by inhibiting viral release. Trivalent seasonal formulations have been administered to young and older adults as well.

Production of VLPs in tobacco plants can rapidly provide large quantities of antigen from transfected plants grown in relatively small spaces.⁵¹ Human trials of a H5N1 HA-containing VLP detected HI and microneutralization antibody responses in 87.5% and 95.8% of young adult subjects receiving two doses of 10 or 20 µg of alum-adjuvanted vaccine. A non-adjuvanted formulation is under investigation as well as a pandemic H1N1 VLP. Tobacco plant-derived recombinant H5N1 and H1N1 HA subunit vaccines also are under clinical investigation.

Live Attenuated Influenza Vaccines

Currently licensed live-attenuated influenza vaccines (LAIV), such as those produced by MedImmune, are reassortants of cold-adapted (ca) and temperature sensitive (ts) master strain viruses with recommended seasonal strains (which donate the relevant HA and NA protein genes) or are reverse genetic engineered viruses recapitulating those ca and ts mutations as well as other attenuating mutations.⁵² The resulting viruses are limited in their ability to replicate at the warmer temperatures of the lower respiratory tract after intranasal administration. Live attenuated vaccines have superior efficacy to non-adjuvanted TIV in young children.⁵³ However, they are not licensed for use in children less than 2 y and carry a precaution for use in children under 5 y old who have a prior history of wheezing. Because wheezing episodes occur in as many as one third of young children by the age of three years, a substantial number of children under 5 y are excluded from receiving LAIV.⁵⁴ The efficacy of LAIV falls by young adulthood, probably because acquired, cross-reactive immunity limits replication of the attenuated virus after IN administration.^55,56 The risk of wheezing in children and reduced efficacy in older individuals are also likely to pose challenges for the new live attenuated vaccines in development.

A novel approach to generating a live attenuate influenza vaccine, deleting or altering the NS1 gene, is being explored by Vivaldi Biosciences and AVIR Green Hills Biotechnology. The NS1 protein inhibits host resistance to infection by blocking α interferon production and by other mechanisms. If NS1 function is abrogated or reduced, a greater host native immune response to vaccination leads to a more robust cellular and humoral immune response and inhibits replication of the NS1-mutated virus.^57-59 Limited replication attenuates the infection and, importantly from a clinical safety perspective, limits viral shedding and averts transmission of the vaccine virus with the potential for reassortment. Human volunteer recipients of a monovalent NS1-deleted vaccine candidate produced both nasal wash and serum antibodies, with almost no viral shedding, and experienced local symptoms that did not differ from those experienced by placebo recipients.⁵⁷ Serum antibodies that were induced despite the intranasal route of administration also cross-neutralized antigenically distinct, heterovariant viruses. A trivalent formulation is under clinical evaluation. Viruses attenuated with more limited deletions or mutations, coupled with mutations in other gene segments or in NS2, are in pre-clinical development.^58-60

Live Vectored Influenza Vaccines

Efforts are also underway to develop vectored influenza vaccines, which use replicating recombinant constructs based on viruses other than influenza to infect an immunized subject, thereby inducing immunity against influenza. A number of vector systems have been employed to express influenza HA and other influenza virus proteins in recombinant vaccine candidates for human seasonal as well as potentially pandemic avian influenza viruses.⁶¹ Some recombinant systems could allow faster and safer development of master seed viruses and vaccine production in large volumes at lower cost and with greater speed than conventional TIV. On the other hand, the vectored approach faces the challenge of anti-vector immunity, which can dampen the response to repeat immunization.

Adenoviruses naturally target mucosal receptors, and mucosal administration of the vector can overcome the anti-vector immunity seen with parenteral administration.⁶² Defective adenovirus 5 particles expressing HA delivered intranasally can induce innate and adaptive heterosubtypic responses and, in mice, may also induce a transient influenza-resistant state, simulating a prophylactic drug effect, without interfering with the development of persistent immunity.⁶³ More conserved viral genes and HAs of various subtypes also have been incorporated to produce more broadly protective candidate vaccines. Orally delivered adenovirus 4 and 5 vectored vaccines are in development.⁶⁴ There is also considerable interest in modified vaccinia virus Ankara as a vector system, with a clinical trial of a NP- and M-containing vaccine having been reported.⁶⁵

Relative Speed of Novel Influenza Vaccine Approaches

In a pandemic response, all existing influenza vaccine production systems face significant challenges. In conventional flu vaccines, the growth of the seed virus is dependent on the HA and NA chosen. A seed virus must be found with an HA and NA that not only match circulating strains antigenically but also support adequate growth in eggs or mammalian cell culture and performance in downstream purification. Virus growth and antigen purification often must be optimized in short order. Synthetic biology may provide means to speed seed generation for conventional influenza vaccines. Advances in the ability to synthesize nucleic acid molecules the size of an influenza genome could allow the generation of an influenza seed virus from an electronically distributed sequence in less than a week.⁶⁶ Indeed, the cloning steps for constructing baculovirus recombinants to generate subunits or VLPs in insect cells can be more rate limiting than the reverse genetic rescue steps needed for conventional seed generation because of the relatively long replication time of baculoviruses and the need to obtain clonal recombinant viruses. Advances in baculovirus engineering techniques could address these limitations.⁶⁷ The generation of a HA head-expressing E. coli strain can be exceedingly fast, but re-folding optimization and any required construct adjustment for new strains could be time consuming.

Quadrivalent Influenza Vaccines (QIV)

The current trivalent formulation of seasonal influenza vaccines has been challenged for decades, since appreciable co-circulation of influenza B viruses in two genetically and antigenically distinct genetic lineages was noted in the early 1980s.⁶⁸ The co-circulation of viruses from both lineages and, in some years, the absolute dominance of the lineage not represented in the vaccine have made clear the medical and public health gap left by a trivalent vaccine.⁶⁹ For example, in 2005–2006, influenza B strains of the Victoria lineage, which was not represented in the vaccine, caused 58% of all cases of influenza reported in Europe.⁷⁰ In the 10 years from 2001 to 2010, the B strain in the WHO vaccine composition recommendation has switched between the two lineages five times in reaction to the regular alternating dominance of the two lineages, resulting in a nearly predictable cycle of a season in which the B strain in the vaccine and in circulation match, followed by a year in which there is a lineage mismatch, leading to a change.⁶⁹ The cyclical transmission of the two lineages most likely reflects the acquisition of immunity in the human population, particularly children, as, unlike influenza A viruses, the B subtype is transmitted only between humans (with incidental animal infections). However, a role for interfering influenza A epidemics on the incidence of influenza B infections also has been proposed.⁶⁸

The movement toward making QIV (containing viruses from both B lineages) the standard seasonal vaccine has gained momentum, as manufacturers have overcome capacity constraints to produce a vaccine containing 15 ug more hemagglutinin and to improve processes to produce four monobulks in the same interval as they have for three. The slower drift of influenza B virus, 0.14–3.32 × 10⁻³ substitutions/site/year compared with A(H3N2), 2.68–12.50 × 10⁻³ substitutions/site/year means that fewer novel strains must be identified and selected for good growth characteristics. Therefore, in most years, seed viruses for a B strain from a previous season can be re-used.⁶⁸ At least two manufacturers, one producing an inactivated vaccine and one a LAIV, have announced their intentions to launch QIV products within two years of the writing of this manuscript.

For the licensure of inactivated QIV, manufacturers must show non-inferiority of the antibody response to the B strain contained in the candidate QIV that matches the B strain in TIV as well as a satisfactory response to the second B strain, not contained in TIV. Publically disclosed results for the candidate inactivated QIV have shown success in meeting those endpoints.⁷¹ For the quadrivalent LAIV, the potential for viral interference among four intranasally administered viruses evidently has been overcome.⁷² As influenza B virus attack rates are highest in older children, a live vaccine formulated with both B strains may have particular value in preventing influenza in the pediatric population.

Adjuvanted Inactivated Influenza Vaccines

Adjuvants represent the best known means to enhance immune responses to inactivated vaccines. Thus, adjuvanted TIV for adults, like LAIV for children over 2 y of age, provides benchmarks for new vaccine candidates. Influenza vaccines are among the very few non-replicating vaccines that do not, in general, include adjuvants. Efficacy of non-adjuvanted inactivated influenza vaccines likely occurs because individuals are already immunologically experienced with influenza antigens, based on previous annual vaccinations and/or previous influenza infections (clinically overt or asymptomatic). In this context, vaccination expands a pre-existing pool of memory cells without need for further “help” from an adjuvant.

The first generation of adjuvanted influenza vaccines, in the 1950s, contained mineral oil, which allowed dose sparing, increased antibody responses, and increased persistence of the elicited antibodies.^73-77 However, mineral oil-adjuvanted vaccines also caused serious adverse events, such as sterile abscesses, in almost 3% of vaccinees.⁷⁸ In the 1960s and 1970s, the next generation adjuvanted influenza vaccines, TIV with alum, increased reactogenicity but not immunogenicity.^79-82 The current generation of adjuvanted influenza vaccines was launched in 1997, when TIV containing the first oil-in-water adjuvant, MF59, was licensed in Italy for use in subjects more than 65 y old.⁸³ Strong adjuvant activity with a good safety profile was achieved by reducing the amount of the oil in the emulsion from 50% to 4–5% and replacing non-metabolizable mineral oil with a fully metabolizable oil, squalene. Squalene is a physiological component of the human body—the precursor of cholesterol and of corticosteroid hormones.⁸⁴

Based on the results of extensive clinical development, which included more than 20,000 subjects (mostly elderly), MF59-containing TIV is now licensed for the elderly in more than 20 countries. The studies showed that MF59-adjuvanted vaccine is more immunogenic than conventional non-adjuvanted influenza vaccine and is well tolerated, eliciting a low incidence of local, mild, and transient reactions, which do not increase upon subsequent immunization. Several clinical studies have shown that seasonal MF59-adjuvanted influenza vaccine induces high titer antibody responses against heterovariant strains.^85-90 The enhancement of the antibody response persists upon subsequent immunizations, suggesting that enhanced antibody responses in previous years do not affect the immunogenicity of later doses.⁹¹ Most encouragingly, clinical effectiveness data have now started to indicate that the increased immunogenicity of MF59 translates into improved protection.^92,93 The enhanced immunogenicity elicited by MF59 (also as compared with alum) has been shown in recent years with vaccines against avian H5N1 influenza viruses in all age groups, including those over 65 y old, and more recently against the 2009 pandemic H1N1 virus.^94-96 In studies of pre-pandemic immunization against H5 strains, an MF59-adjuvanted H5N3 vaccine primed for rapid boosting with an H5N1 vaccine 6–8 y later, resulting in a rapid rise in titers against H5N1 strains from both clade 1 and clade 2.^97,98

Pediatric MF59-adjuvanted TIV is now in development. Young children have a reduced immune response to TIV, requiring two doses of vaccine for primary immunization. Two phase I and II studies in 6–59-mo-old children that compared Fluad^® to a licensed inactivated split vaccine found that Fluad^® elicited significantly higher HI antibody titers and was as well tolerated as TIV, although it was associated with more induration at the injection site.⁹⁹ Based on these encouraging data, a Phase III efficacy field trial comparing Fluad^® to unadjuvanted TIV and control non-influenza vaccines was conducted in vaccine-naive 6–72 mo-old children, who were given age-appropriate doses of the respective vaccines.¹⁰⁰ More than 4500 children were studied over two influenza seasons for PCR-confirmed influenza-like illness. Serious adverse events were reported equally rarely in all three study groups, and Fluad^® had a similar tolerability profile to TIV and control vaccines in 6–36 mo-old children. The absolute efficacy of Fluad^® in the two seasons was 89% against all influenza viruses, irrespective of vaccine-match, while TIV had an efficacy of 43%, for a relative efficacy of Fluad^® over TIV of 75%. Absolute efficacy was slightly higher in 36–72 mo-olds compared with 6–36 mo-olds, but the relative efficacy advantage of Fluad^® was maintained in both age groups as well as in 6–24 mo-olds, yielding the highest efficacy for an inactivated influenza vaccine ever reported in these age groups.¹⁰¹

The success of MF59 has prompted other groups to develop oil-in-water adjuvants.¹⁰² One of these, referred to as AS03 and containing α-tocopherol as an immunopotentiator, has been developed by Glaxo Smith-Kline and tested in older subjects (greater than 65 y of age), together with a seasonal trivalent split influenza antigen. The ASO3-adjuvanted vaccine significantly enhanced HA-specific cellular immune responses.¹⁰³ Other data, for example on the antibody response induced by this adjuvant in association with seasonal inactivated influenza vaccines, have not been published as of the writing of this manuscript. In ASO3, the α-tocopherol component has an immunopotentiating effect, strongly activating innate immunity not only at the site of injection by also at the draining lymph nodes.¹⁰⁴ AS03 has been used to develop an H5N1 vaccine, which has shown a consistently strong ability to elicit cross-clade neutralizing titers and cellular immune responses as well as high potential for significant dose-sparing.¹⁰⁵

The pandemic H1N1 vaccine adjuvanted with AS03 has been used widely in many European countries, in Canada, and elsewhere. This vaccine has shown higher immunogenicity and efficacy than a non-adjuvanted whole-virion vaccine.¹⁰⁶ In a comparative study in immunosuppressed children in Switzerland and in another study in subjects with chronic inflammatory bowel disease, the ASO3-adjuvanted vaccine proved more immunogenic but also more reactogenic than a subunit vaccine adjuvanted with MF59.^107,108 Recently, the use of the AS03-adjuvanted pandemic H1N1 vaccine has been associated with cases of narcolepsy in subjects aged 4 to 19 y in Sweden and Finland, but not in non-Scandinavian countries (such as the UK and Canada), where this vaccine was more widely used.^109-111 It is not clear if a cause-effect relationship may link immunization with the ASO3-adjuvanted vaccine and narcolepsy. The use of MF59-adjuvanted pandemic or seasonal influenza vaccines has not been associated with narcolepsy.¹¹²

Another oil-in-water emulsion, referred to as AF03, is still at early stages of development.¹⁰⁵ It has been tested with H5N1 influenza vaccines, and it allowed significant dose sparing.¹¹³ Finally, an AF03-adjuvanted pandemic H1N1 vaccine was also licensed in Europe, but its use was extremely limited.

Novel Delivery Approaches

Macroneedles

Intradermal delivery of vaccines offers the advantage of favoring the migration of antigens to lymph nodes with consequent stimulation of resident dendritic cells (DCs), at the same time inducing activation of resident Langerhans cells (LCs) and activation and migration of dermal DCs. These antigen presenting cells (APCs) act together synergistically to activate antigen-specific T cells in the lymph nodes. By delivering the antigen at an anatomical site rich in professional APCs, intradermal vaccination also has the theoretical potential to allow dose sparing. Intradermal immunization of the elderly must overcome changes resulting from aging of the skin anatomy and physiology, including changes in the quantity and quality of LCs and DCs present in the skin.^114,115

Considerable experience with intradermal vaccination has been acquired in past years with smallpox vaccines, rabies vaccines, and other vaccines. Interest in the intradermal route for vaccination against influenza dates back to the 1940s and 1950s.¹¹⁶ This interest was boosted by the threat of an influenza pandemic due to the swine-origin A/New Jersey/76 H1N1 virus isolated in the US in 1976. Indeed, because intradermal immunization requires no more than 0.1 ml of vaccine (one-fifth of the standard volume of adult intramuscular influenza vaccines), intradermal vaccination offers dose-sparing, which could be essential to avert shortages during influenza pandemics. Some studies clearly showed that intradermal vaccination with low-dose monovalent or bivalent influenza vaccines induced antibody titers comparable to or higher than those induced by the full-dose intramuscular vaccination.^117,118 Interestingly, however, the beneficial effect of the intradermal route was less evident in subjects aged over 50 y, suggesting a potential role of aging in diminishing the immune response triggered by intradermal vaccination.¹¹⁸

More recently, a substantial number of clinical studies have evaluated intradermal administration of influenza vaccine to individuals of all ages, including the elderly. Two key studies conducted in adults and in the elderly showed that 20% or 40% of a regular human dose of a subunit or a split influenza vaccine induced immune responses comparable to or higher than those induced by the conventional full-dose vaccine.^119,120 However, in subjects older than 60 y of age, there was a trend toward a better response after intramuscular vaccination, which reached statistical significance for the H3N2 antigen, showing a clear effect of aging on the immune responsiveness to intradermal influenza vaccination.¹¹⁹ Other studies in healthy and immunocompromized adults have reported contrasting results.^121-125 It must be emphasized that in all these studies the most rigorous controls were missing – there was no comparison with equally low doses of vaccine given intramuscularly. When a study with such controls was conducted, the argument for a better immune response to intradermally administered influenza vaccine disappeared. All low doses (3, 6, or 9 μg), whether given intradermally or intramuscularly, were almost as immunogenic as the standard dose of 15 μg given intramuscularly.¹²⁶ These data confirmed that one half dose of trivalent seasonal vaccine given intramuscularly was as immunogenic as a full dose in boosting an influenza-specific antibody response,¹²⁷ although this boosting effect was less pronounced in people older than 50 y of age.¹²⁸

More recently, intradermal vaccination against influenza was extensively re-evaluated in a randomized, controlled study in 1107 subjects aged greater than 60 y, who received either 15 μg of HA intramuscularly or intradermally or 21 μg of HA, always in 0.1 ml volume. In this study with standard or increased HA dosage, the geometric mean HI titers, seroprotection rates, seroconversion rates, and mean antibody titer increases elicited by intradermal vaccination were superior to those observed in elderly individuals vaccinated intramuscularly with 15 μg of non-adjuvanted HA.¹²⁹ Recently, it has been asserted that this intradermal vaccine is non-inferior to adjuvanted seasonal influenza vaccine in the elderly despite inconsistent serologic results, the problem of vaccine leakage inherent to the route of administration, and the significantly higher reactogenicity of the intradermal vaccine.¹³⁰ An intradermal inactivated seasonal influenza vaccine produced by sanofi pasteur has now been licensed in Europe and in the US for adults and the elderly.¹³¹

A logical conclusion of these studies is that the intradermal route of vaccination is feasible, but dose sparing is not achieved for seasonal vaccines or for pandemic H5N1 vaccines, even after three doses of 3, 9 or 15 μg.^132,133 On the other hand, more basic and clinical research is needed to determine the role of the dermal environment in general and of professional antigen-presenting cells, such as DCs, in particular. This research may allow optimization of vaccine formulations for intradermal delivery and better understanding of the immune response required for effective protection. In addition, this research may provide precious information on the mechanisms leading to the side effects observed following intradermal immunization. Indeed, local inflammatory reactions (such as erythema) are significantly more frequent after intradermal than after intramuscular vaccination.^119-130 It is likely that these reactions are immunologically mediated, since they have been less frequently observed in immunocompromised subjects receiving influenza vaccine intradermally.¹²⁵

Microneedles and microstructures

Different approaches are currently being followed to deliver vaccine cutaneously, using various types of microneedle arrays to penetrate the stratum corneum and deposit vaccine in the epidermis and dermis. These approaches are particularly interesting because they may allow an easier way to administer the vaccine, possibly allowing dose sparing and even over the counter administration. However, most of these systems are still at very early stages of development, and very few clinical studies have been performed so far. This approach to delivering a monovalent recombinant H3N2 vaccine induced strong immune responses and protection in mice.¹³⁴ A device referred to as MicronJet was used to deliver 0.1–0.2 ml of influenza split vaccine intradermally to human volunteers.¹²⁴ Interestingly, the low-dose intradermal vaccine administered via microneedles gave antibody responses similar to those provided by the full-dose intramuscular vaccine, and met all the Committee for Medicinal Products for Human Use (CHMP) criteria. In this study, no low-dose intramuscular control was given. When such controls were included in studies using macroneedles, the low-dose intramuscular vaccine elicited antibody response similar to those elicited by the intradermal vaccine.¹²⁶ A fascinating new approach consists of preparing microneedle arrays using biocompatible polymers that encapsulate the inactivated influenza vaccines and rapidly dissolve after application to the skin. This method of administration was tested in mice using a monovalent H1N1 vaccine. Mice exhibited a strong antibody and cellular response and were completely protected against a lethal challenge with the mouse-adapted A/PR8/8/34 virus strain.¹³⁵

Vaccines can also be delivered through the skin using a transcutaneous approach, formulating the vaccine on a patch together with the Escherichia coli heat-labile enterotoxin (LT), which serves as an adjuvant. Alternatively, the patch (referred to as an immunostimulant patch) contains only LT and is applied on the skin at the site of injection of the vaccine. These approaches have been extensively tested in various experimental animal models as well as in clinical trials with enterotoxigenic E. coli (ETEC), influenza, and anthrax vaccines and have shown efficacy against diarrhea due to ETEC in mice but not in humans.^136,137 This immunostimulant patch, applied at the site of injection with influenza vaccine, was tested in aged mice¹³⁸ and then in elderly human subjects. Compared with subjects who had received the vaccine alone, those who had received the patch, applied on the skin, had greater antibody responses. In a few cases this difference reached statistical significance.¹³⁹

Intranasally administered inactivated influenza vaccines

Considerable work has been devoted in the past 10–20 y to the development of inactivated influenza vaccines for mucosal immunization, mainly via the IN route. This effort is motivated by the ease of IN immunization, evidence that good responses can be induced by immunizing via this route, and the fact that various strong mucosal adjuvants that work after IN administration in animals and humans have been developed.¹⁴⁰ However, two clouds hang over the development of novel influenza vaccines that are administered intranasally: the occurrence of facial palsy that followed administration of LT-adjuvanted virosomal TIV,¹⁴¹ and the higher risk for wheezing events and hospitalizations in young children who received live-attenuated influenza vaccine (see the section on LAIV). In both cases, the adverse events (AE) occurred at a relatively low incidence so they were not detected until after licensure or in late stages of pre-licensure development, respectively.

In addition, for both non-replicating and replicating antigens, IN administration carries a hypothetical risk that vaccine components may gain access to the central nervous system through the olfactory bulb. Experimental studies in ferrets have documented preferential central nervous system infection following IN rather than intratracheal inoculation of H5N1 influenza, suggesting direct spread from the nasal cavity.¹⁴² Inferences from animal studies must be tempered, however, by acknowledging the more prominent olfactory apparatus of rodents and mustelids compared with humans. Indeed, scintigraphic studies of human volunteers given ⁹⁹Tc-labeled diethylenetriaminepentaacetic acid in a simulated IN vaccination did not detect a signal in the region of the olfactory bulb (unpublished observations, TF Tsai). The facial palsy following IN administration of LT-adjuvanted TIV may have represented a spillover of the local immune response, leading to inflammation or entrapment of branches of the nearby facial nerves.¹⁴² The discovery that, like intranasal influenza immunization with an enzymatically active LT adjuvant, intranasal inoculation of volunteers with an enzymatically inactive mutant LTK63 adjuvant (together with HIV or tuberculosis antigens) was associated with Bell’s palsy brought a critical reconsideration of the use of these adjuvants via the intranasal route.¹⁴³ Based on these findings, it is likely that licensure of non-replicating antigens that are administered IN with adjuvants will require demonstration that Bell’s palsy does not occur at an increased rate in immunized subjects. Given the background rate for facial palsy, the number of subjects required in such a study could be very large.

Other strategies are now being followed for the delivery of influenza vaccines at the mucosal level, for example, using dry powder influenza vaccines or other systems for oral immunization (see Amorij et al.¹⁴⁴ for a review). These systems are still at very early stages of development and will require appropriate clinical studies to understand their applicability to human use.

New Assay Strategies

The assay used to estimate the HA content of most influenza vaccines, single radial immunodiffusion (SRID), is based on a rate-limiting technology that is more than 35 y old. It is essentially a modified Ouchterlony test.¹⁴⁵ To perform SRID, a well is cut in an antibody-suffused agarose slab; a vaccine preparation is mixed with Zwittergent and placed in the well; and after approximately 24 h of diffusion, the diameter of the precipitin zone is taken to be proportional to the HA content of the preparation. SRID has the benefit of conformational selectivity (preferentially detecting properly folded HA), which is achieved by coupling a simple physical separation (diffusion in agarose) to an antigenic assay (formation of precipitin), providing two checks on the state of HA—the antigen must be dispersed sufficiently by Zwittergent to diffuse, and it must also react with a polyclonal antiserum.¹⁴⁶ Denatured HA does not pass these tests.

Unfortunately, SRID also has many disadvantages. These include inconsistent reproducibility with different batches of strain-specific reagents, low through-put, poor dynamic range, quantification of calibration standards by methods that are insensitive to antigenic or conformational integrity, different dose-response curves for calibration standards and many vaccine samples, and a prolonged process for generating strain-specific reagents. SRID antibodies are generated by immunizing sheep with HA that has been isolated from the surface of egg-produced influenza virions and partially purified. The antisera typically take six weeks to two months to produce from the time of the first inoculation of the sheep. During the 2009 pandemic, the first clinical trials of potential vaccines were started at risk, without an agreed value for the HA content of the candidates, because SRID reagents had not yet been calibrated.¹⁴⁷

Because of the need for more a practical, timely, and accurate potency assay, there are many efforts underway to replace SRID. Techniques now under investigation include physical methods, such as reverse phase HPLC,^148,149 isotope dilution tandem mass spectrometry,¹⁵⁰ and SDS-PAGE with staining of protein bands. These techniques can identify and quantify HA, but they require denaturation of HA before or during the analysis. Therefore, they suffer from a common drawback—they cannot distinguish antigenically intact from denatured HA, and a conformationally inauthentic antigen that has limited or no ability to elicit neutralizing antibodies could produce a positive signal. Alternative antibody-based immunoassay formats, such as ELISA and surface plasmon resonance are also being explored.^152,153 These more modern formats could greatly increase the ease, accuracy, and throughput of testing for HA content, but, if based on the standard antisera, these techniques would not relieve the necessity to wait for experimental animals to seroconvert and produce antisera before vaccine can be released. Monoclonal antibodies and polyclonal antisera against conserved epitopes on HA are being evaluated as alternative immunoassay reagents.^151,154 Non-strain-specific reagents have the potential to speed vaccine release, but they also carry risks of inaccurate results. A single monoclonal antibody epitope could be preserved on HA despite lack of intact overall antigenic structure.¹⁵⁴ Indeed, some conserved epitopes actually require protein denaturation for exposure, so that recognition is evidence of lack of folded structure (although the HA could have been well-folded before denaturation in preparation for assay).¹⁵¹ Assays based on non strain-specific antibodies are also vulnerable to mutations of epitopes in future strains. Because antigen combining sites are large, a mutation in the less conserved periphery of a centrally conserved epitope could alter reactivity.

To address the need for conformational selectivity, SRID’s simple physical screen for “good” HA—the ability of Zwittergent-dispersed antigen to diffuse into an agarose gel—could be mimicked by more rapid separations of “good” and “bad” HA using modern techniques. After such a separation, the accurate and high throughput physical assays of HA content now being developed could then provide a determination of HA in the properly folded fractions. Thus, there is reason for optimism that a more practical assay that correlates better with actual vaccine potency in humans could be available for any strain immediately after strain identification. For such innovations to be accepted, it will be important that a new assay be judged on whether it predicts vaccine potency, not on whether it reproduces results obtained by SRID.

Universal Influenza Vaccine Approaches

Internal antigens and M2e

An influenza vaccine that need not be altered as strains change is the “holy grail” of influenza vaccine development. Such a finding would render influenza vaccines equivalent to other biologics—developed once and then mass produced, potentially stockpiled, and used as needed—rather than the result of an elaborate multi-continent performance acted out twice each year (once for the northern hemisphere and once for the southern hemisphere), always with the possibility of a misstep.

Many attempts to make vaccines using the conserved internal antigens of influenza virions have been attempted, although few have reached clinical trials.^155-160 Vaccines based solely on conserved internal antigens must rely on cell-mediated immunity, because the target antigens are not exposed on the surface of an influenza virion for antibody binding.¹⁶¹ In rodent models, such T cell-based vaccines can protect from challenge.^{155-157,159,162} In the more rigorous ferret model, they perform less well.¹⁵⁸ The recurrent nature of influenza infection, despite the presence of such conserved antigens in every circulating influenza virus strain, is an indication that immunity directed against these antigens is incomplete, although cell mediated responses appear to contribute to the clearance of infection in humans.¹⁶³ For a vaccine based solely on T-cell immunity against conserved internal antigens to work efficiently in adults, it would need to provide better protection than the exposure to these antigens that occurs during a natural influenza infection. On the other hand¸ during his or her first influenza infection, a child does not have pre-existing immunity to conserved viral antigens, and there could be a significant disease-mitigating effect of immunization with a vaccine that presents conserved antigens. In addition, a vaccine that combines conserved internal antigens with the variable surface antigens that are targets of protective antibodies could elicit T-helper responses that improve antibody responses to the surface antigens.¹⁶⁴

The surface of a group A influenza virion does present a conserved protein epitope—the external portion of the M2 ion channel. That external portion (M2e) is short (approximately 23 residues). Because the transmembrane portion of M2 is tetrameric, it is likely that the extracellular portion also associates into a larger quadrivalent structure.^165-167 There is little M2 on an influenza virion, but much more is presented on the surface of infected cells, and infected humans do produce antibodies recognizing M2e.¹⁶⁸ Antibodies against M2e do not neutralize virus but do provide weak protection in rodent models of infection, probably on the basis of antibody-dependent cellular cytotoxicity (ADCC).^169,170 In contrast to neutralizing antibodies, which block virus infection of cells, cell mediated immunity (including ADCC) contains viral infections by killing infected host cells. Thus, there is at least a theoretical possibility that vaccines that promote these mechanisms could exacerbate rather than reduce illness. In fact, disease enhancement has been reported when pigs immunized with a DNA vaccine that expressed a fusion between influenza M2e and NP were challenged with influenza virus.¹⁷¹ A fusion protein between Salmonella type 2 flagellin and M2e has been studied as a vaccine antigen in young adults. Although higher doses caused significant reactogenicity, 0.3 and 1 μg doses were well-tolerated and resulted in significant increases in antibody titers against M2e.¹⁷²

Conserved epitopes on HA and improved immunization regimens

The standard assay used for more than 60 y to assess immunity, including vaccine-induced immunity, against influenza is the hemagglutination inhibition assay (HI).¹⁷³ HI measures the ability of antibodies to prevent influenza from causing red blood cells to clump, a reaction predominantly mediated by HA binding to its receptor, sialic acid. Thus, the assay is a surrogate for testing antibody inhibition of influenza binding to host cells. The epitopes that mediate HI are located predominantly in the highly variable region near the sialic acid binding site on the HA head.⁴² Although some monoclonal antibodies that recognize the head neutralize relatively broadly, most do not.^174,175 Influenza viruses can also be neutralized by antibodies that prevent virus-cell fusion (and therefore cell entry) but do not block attachment to cells.⁴³ Such antibodies produce no signal in the HI assay. Intriguingly, the epitopes that mediate this mechanism of virus killing are located on the stem of HA and are much more conserved than the head epitopes.^176-178 Indeed, a monoclonal antibody that neutralizes all group A influenza viruses tested has recently been described.¹⁷⁹ Although the protective efficacy of such antibodies could be limited by accessibility to the stem on a crowded influenza virion surface or by competition from non-neutralizing anti-stem antibodies, some anti-stem antibodies do protect experimental animals from infection and therefore have potential as human prophylactic or therapeutic agents.^179-181

Efforts are underway to design HA-based antigens that preferentially elicit broadly neutralizing anti-stem antibodies rather than the more narrowly specific antibodies elicited by most exposures to influenza infections or vaccines.¹⁸² Redirecting immunity toward a desired epitope on a complex multi-epitope antigen is a long standing problem in vaccinology, which has been most rigorously explored in HIV vaccine development. The presence of broadly neutralizing epitopes on HIV has been known for many years, but a vaccine that can efficiently elicit potent, broadly neutralizing antibodies has yet to be developed.¹⁸³ The tools of antibody repertoire analysis and structural biology are now being deployed in both HIV and influenza vaccine development in search of an antigen that can elicit broadly protective antibodies more efficiently than does infection with the pathogens.

Recent laboratory and translational medicine findings have indicated that it may be possible for a sequence of antigen exposures to elicit broadly protective neutralizing responses. In experimental animals, a prime-boost strategy in which a DNA vaccine prime was followed by a conventional inactivated vaccine boost proved more efficient at eliciting cross-reactive antibodies than immunization with a conventional vaccine alone, and phase I trials of a DNA prime, split vaccine boost regimen provides preliminary evidence that humans may respond similarly.^184,185 Immunization with seasonal vaccines does elicit broadly neutralizing anti-stem antibodies in some individuals,¹⁸⁶ and the response to the 2009 pandemic influenza strain appears to have elicited a particularly broad response.¹⁸⁷ Although the reasons for this phenomenon are still being explored, it may have to do with the degree of antigenic relatedness of the 2009 strain to previous strains and the sequence of exposures to antigenic variants. Cross-reactivity of the stem epitopes, but not head epitopes, between the 2009 and previous strains may have provided potent booster responses to the conserved epitopes.

Thus, there are prospects for a vaccine regimen that, even if not truly universal, could significantly broaden and extend protection against influenza. The use of adjuvants is a relatively simple step that increases the breadth, potency, and efficacy of inactivated influenza vaccines. Engineered antigens, prime-boost vaccine regimens that set favorable immune profiles in previously unexposed infants, and sequential immunization with antigenic variants could produce vaccines that are more broadly protective than those in use today.

Declaration of Potential Conflicts of Interest

The authors are Novartis stockholders and employees of Novartis Vaccines.

Acknowledgments

We thank Jane Davis for help with formatting and compiling the references.