Challenges of Making Effective Influenza Vaccines

Abstract

Seasonal influenza vaccines prevent influenza-related illnesses, hospitalizations, and deaths. However, these vaccines are not as effective as other viral vaccines, and there is clearly room for improvement. Here, we review the history of seasonal influenza vaccines, describe challenges associated with producing influenza vaccine antigens, and discuss the inherent difficulties of updating influenza vaccine strains each influenza season. We argue that seasonal influenza vaccines can be dramatically improved by modernizing antigen production processes and developing models that are better at predicting viral evolution. Resources should be specifically dedicated to improving seasonal influenza vaccines while developing entirely new vaccine platforms.

INTRODUCTION

Influenza is a major respiratory disease caused by influenza viruses. According to the World Health Organization (WHO), seasonal influenza epidemics cause approximately 3 to 5 million cases of severe illness and 290,000 to 650,000 deaths worldwide each year (1). The most effective preventive measure against influenza virus infection is vaccination. Current seasonal influenza vaccines are administered in a trivalent or quadrivalent format with antigens from two influenza A strains (H1N1 and H3N2) and either one or two influenza B strains (Victoria and/or Yamagata) (2). The majority of influenza vaccines are inactivated influenza vaccines (IIVs) produced in embryonated chicken eggs, but egg-based live attenuated influenza vaccines (LAIVs) and alternative vaccines that do not rely on eggs are also commercially available (Table 1). Most influenza vaccines elicit antibodies against the major viral surface proteins, hemagglutinin (HA) and neuraminidase (NA). Unlike with other vaccines, influenza vaccine antigens need to be updated regularly as a result of antigenic drift in the HA. The WHO meets twice a year to select the vaccine components for the upcoming influenza seasons in the Northern and Southern Hemispheres. Despite these efforts, vaccine effectiveness (VE) is usually under 60% and can be as low as 10% (3). VE varies between seasons, and the lowest VE occurs in seasons in which there is an antigenic mismatch between the selected vaccine strains and circulating influenza strains. Here, we give a brief history of influenza vaccines and then present an overview of the current status of seasonal influenza vaccines. We then discuss how VE and immunogenicity differ between different types of seasonal influenza vaccines. While it is clearly important to develop new universal vaccines that elicit broader immunity, we argue that it is equally important to invest in efforts to improve current seasonal influenza vaccines.

Table 1.

FDA-approved influenza vaccines for the 2019–2020 season

Platform	Brand name	Manufacturer	Type	Inactivating agent	Component(s)	Formulation	Antigen amount	Licensed for (age)	Route of admini stration
Egg based	Fluarix	GlaxoSmithKline	Inactivated	Formaldehyde	Split with deoxycholate	Quadrivalent	15 ug HA	≥6 months	IM
	FluLaval	GlaxoSmithKline	Inactivated	UV light and formaldehyde	Split with deoxycholate	Quadrivalent	15 ug HA	≥6 months	IM or ID
	Fluzone	Sanofi Pasteur	Inactivated	Formaldehyde	Split with Triton X-100	Quadrivalent	15 ug HA	≥6 months	IM
	Fluzone High Dose	Sanofi Pasteur	Inactivated	Formaldehyde	Split with Triton X-100	Trivalent	60 ug HA	≥65 years	IM
	Afluria	Seqirus	Inactivated	BPL	Split with sodium taurodeoxy-cholate	Quadrivalent	15 ug HA	>5 years	IM or ID
	Fluad	Seqirus	Inactivated	Formaldehyde	Subunit	Trivalent + Adjuvant MF59	15 ug HA	≥65 years	IM
	FluMist	AstraZeneca	Live attenuated	Not applicable	Whole purified virus	Quadrivalent	106.5−7.5 FFU/0.2 mL	2–49 years	IN
Cell culture based	Flucelvax	Seqirus	Inactivated	BPL	Subunit	Quadrivalent	15 ug HA	>4 years	IM
Recombinant protein	Flublok	Sanofi Pasteur	Protein only	Not applicable	Subunit	Quadrivalent	45 ug HA	≥18 years	IM

Abbreviations: BPL, beta-propiolactone; FDA, Food and Drug Administration; FFU, focus-forming units; HA, hemagglutinin; ID, intradermal; IM, intramuscular; IN, intranasal.

HISTORY OF INFLUENZA VACCINES

The first human influenza virus was isolated by Smith et al. (4) in 1933 from human throat-washings (Figure 1). Soon thereafter, in 1937, Smorodintseff and colleagues (5) developed the first intranasal LAIV using influenza viruses adapted to ferrets and mice. In that same year, Burnet (6) demonstrated that influenza viruses passaged in eggs were immunogenic but not pathogenic in mice and ferrets and caused only mild symptoms when administered intranasally. Concurrently, Francis & Magill (7) determined that subcutaneous and intramuscular injections of tissue culture–propagated LAIV induced high antibody titers in humans. Two sequential trials carried out during the winters of 1935–1936 and 1936–1937 assessed protection from influenza infection after vaccination with LAIV among inmate populations (8, 9). While controversial by today’s ethical standards, these trials indicated that LAIVs were safe and provided protection with few adverse events. However, a subsequent long-term LAIV trial performed between 1937 and 1941 in a residential institution was inconclusive (10). Furthermore, stability, scalability, and safety issues precluded early use of LAIVs and led to the development of IIVs.

Figure 1.

Timeline of influenza virus pandemics, subtype circulation, major discoveries, and vaccine development. Abbreviations: HA, hemagglutinin; IIV, inactivated influenza vaccine; LAIV, live attenuated influenza vaccine; NA, neuraminidase.

Early IIVs included formaldehyde-inactivated whole influenza virions obtained from tissues of infected animals, but these vaccines were poorly immunogenic (11). Inconsistent antigen amounts and potency were common among early IIVs (12). In 1941, Hirst (13) discovered that chicken red blood cells can be agglutinated by the allantoic fluid of chick embryos infected with influenza virus. This pivotal observation resulted in an agglutination-based method to calculate influenza virus and antibody titers (14). Hirst and colleagues (12) also determined that higher antigen doses of both LAIV and IIV induced more robust immune responses. It became clear that influenza vaccines could be improved by developing methods to create large amounts of influenza virus antigens. Relatively simple methods of culturing influenza virus in the allantoic fluid of embryonated hen eggs made it feasible to obtain large quantities of virus for vaccine efforts (6). To increase antigen amounts, methods were developed to concentrate viruses using high-speed ultracentrifugation, adsorption-elution on chicken red blood cells, and freeze-thawing to form viral precipitates. In 1944, Stanley (15) compared different concentration and purification methods for influenza virus passaged in allantoic fluid. Stanley’s protocol for the preparation of influenza vaccine produced in embryonated chicken eggs became the basis for commercial influenza vaccine production (16, 17).

Refinement of inactivation methods was the other major obstacle to overcome for producing IIVs. In the 1940s, whole-virus preparations were concentrated and then treated with chemical inactivating agents (such as formaldehyde or beta-propiolactone) or irradiated with UV light (15). Chemical inactivating agents proved more feasible for large manufacturing operations and are still used today. Additional vaccine types were developed to avoid adverse effects associated with inactivated whole-virus vaccines, particularly among young children (18). Split and subunit vaccines were developed throughout the 1960s (19, 20). Split vaccines include antigens from concentrated inactivated whole viruses that are disrupted by detergent treatment, and subunit vaccines undergo an additional purification step that enriches the immunogenic surface antigens HA and NA (21). Split and subunit vaccines remain the most common formulations of influenza vaccines today. A single-radial-immunodiffusion technique was developed to standardize the amount of HA antigen (22, 23) because early antigen standardization methods that used chick cell agglutination assays were unreliable (24, 25).

IIVs prepared from allantoic fluid have been administered to humans since the 1940s (26–29). In 1969, Kilbourne (30) developed a method to produce virus that grew to high titers in eggs by coinfecting the virus of interest with an egg-adapted strain, thereby generating reassorted viruses, an approach that is still used today for the production of IIVs. Cold-adapted (ca) LAIV was first licensed in Russia in the 1970s (31). Meanwhile, a separate ca LAIV was developed in the United States by Maassab (32). The Russian ca donor strain (Leningrad) was obtained by multiple passages at low temperature (25°C), while the US ca donor strain (Ann Arbor) was produced by gradually decreasing the temperature between passages (33). The ca LAIV donor strains replicate at a lower rate than wild-type strains and can be updated through reassortment with HA and NA from circulating strains (34). The Ann Arbor–backbone LAIV (FluMist, AstraZeneca) was licensed in the United States and Europe in 2003 and 2008, respectively (35). The use of LAIV was temporarily suspended in the United States because it was not effective during the H1N1-dominated 2015–2016 season, but LAIV is currently recommended in the United States again as one of several options for influenza vaccination (35).

ANTIGENIC DRIFT AND VACCINE STRAIN SELECTION

Viral antigens in seasonal influenza vaccines need to be updated regularly because influenza viruses continuously acquire antigenically relevant HA and NA substitutions through a process called antigenic drift. HA and NA are both important influenza virus proteins: HA mediates virus attachment and entry (36), whereas NA is involved in virus release and is required to prevent virion aggregation (37–39). Current influenza vaccines elicit primarily HA antibodies, although some vaccines possess other viral proteins including NA (40). HA-specific antibodies are thought to be of greater importance than NA-specific antibodies, but antibodies to both surface proteins can independently confer immunity (41, 42).

Antigenic drift occurs in both influenza A and B viruses, but the evolutionary rates differ. H3N2 viruses evolve in a linear fashion and more rapidly than H1N1 viruses and influenza B viruses (43). The estimated rate of evolution for influenza B viruses is two to three times lower than for influenza A viruses (43). The polymerase of influenza B viruses has a lower mutation rate, and there seems to be a stringent genetic bottleneck between hosts for influenza B viruses, but it is unclear if these phenomena fully explain the differences in evolution rate (44, 45). Significant antigenic drift is often associated with the early onset of more severe influenza seasons, particularly for H3N2 viruses (46). It is thought that new H3N2 variants originate in East and Southeast Asia and subsequently seed epidemics in the Northern and Southern Hemispheres (47). Interestingly, East and Southeast Asia appear to play a more limited role in disseminating new H1N1 and influenza B virus variants (48). Unlike H3N2 viruses, H1N1 and influenza B virus variants can persist locally between epidemics (48).

Vaccine strain antigenic mismatch first became apparent during the winter of 1946–1947 when an antigenically distinct H1N1 virus circulated (10). In 1952, the first system for the global surveillance of influenza was created by the WHO (49). WHO vaccine strain recommendations are based on multiple factors: global surveillance, antigenic data, and the availability of suitable vaccine strains (50). Historically, antigenic characterizations of influenza viral isolates have relied on hemagglutinin inhibition (HAI) assays; however, virus neutralization assays are common today (50) because contemporary H3N2 viral strains agglutinate red blood cells inefficiently (51). Postinfection ferret antisera are typically used to identify antigenic variants; however, humans with extensive immune histories often have different antibody types compared with ferrets (52–54). For this reason, serum from vaccinated humans is also now routinely included for antigenically characterizing circulating viral isolates.

It is important to identify antibody specificities that are common among different aged humans that have distinct immune histories. Humans tend to have antibody responses that are biased toward HA epitopes present in viral strains that they encountered in childhood (55). For example, upon infection with the 2009 pandemic H1N1 strain, most individuals born between 1983 and 1996 mounted antibody responses focused on an epitope near the HA receptor-binding domain that was conserved between the 2009 H1N1 strain and H1N1 strains that circulated from 1983 to 1996 (52). An age-specific difference in antibody specificity was also observed during the 2013–2014 influenza season (53), which led to vaccine failures in individuals born between 1977 and 1985 (56). Age-specific differences in antibody specificities can be easily elucidated by analyzing serum from individual humans, and therefore, serum from different aged individuals should continue to be used for antigenic characterizations of new influenza virus strains.

Despite frequent vaccine updates, it remains challenging to adequately match vaccine candidate strains to circulating influenza viruses. Predictions about which viral strains are likely to circulate must be made well in advance of each influenza virus season because the current production process of making influenza vaccines takes up to 6 months (57). There have been great advances in generating predictive models of influenza virus evolution in recent years (58). Most of these models are based on antigenic cartography, a computational technique to quantitatively interpret antigenic variation based on serological data (59). Notably, most antigenic cartography studies continue to use data based on serological assays with ferret antisera. New models can estimate evolution of individual influenza clades based on epitope and nonepitope characteristics of HA genes (60). Other models have identified future predominant viral clades by using allele dynamics plots and antigenic trees (61) and branching patterns of reconstructed genealogical trees (58, 62). A recently published model combined metrics for antigenic novelty, intrinsic fitness, and recent clade growth to predict seasonal influenza circulation patterns up to a year in advance (63). Because most predictive models are based on large data sets, data sharing is critical to reduce inherent biases in surveillance data and to increase the robustness and accuracy of the models (58).

HUMAN INFLUENZA VACCINES ARE HETEROGENEOUS

Human influenza vaccines are remarkably heterogeneous (Table 1). Standard IIVs possess 15 ug of each HA; however, IIVs with higher antigen doses are licensed for use in older adults (≥65 years of age). LAIVs possess much lower amounts of antigen, but these vaccines are delivered intranasally and they replicate and produce additional antigen in vaccine recipients. There are also a limited number of commercially available influenza vaccines that are not produced in eggs, including cell-based and recombinant protein–based vaccines. Complicating matters further, WHO recommendations on the composition of influenza virus vaccines can sometimes differ between different vaccine types. For example, the 2019 Southern Hemisphere H3N2 vaccine component was updated for egg-based vaccines but not for cell-based vaccines (64). It is therefore difficult to directly compare immunogenicity and VE between vaccine types.

EGG-BASED INFLUENZA VACCINES HAVE INHERENT PROBLEMS

To make egg-based vaccine strains, reassortant viruses are isolated with the HA and NA of contemporary viral strains with the remaining six gene segments from an egg-adapted virus (65). These reassortant vaccine strains are then injected into the allantoic cavity of fertilized hen eggs, where they replicate in the cells that line the chorioallantoic membrane. After 2 days of incubation, the allantoic fluid is removed from the eggs and the virus is isolated. IIVs contain both HA and NA, but only the HA content is standardized (66). Because the NA content is not standardized, its contribution to VE is unclear and it is likely that there are inconsistencies in HA/NA ratios across vaccine types and different batches of the same vaccine.

Cultivation of human influenza viruses in eggs can result in egg-adapted amino acid substitutions, frequently around the HA receptor binding site (Figure 2). Human influenza viruses preferentially bind to α2,6-linked sialic acids, whereas allantoic cells contain predominantly α2,3-linked sialic acids (67, 68). Human influenza viruses grown in eggs rapidly acquire HA substitutions that enable efficient binding to α2,3-linked sialic acids (68), and these substitutions can sometimes alter antigenicity (69–72). Egg adaptations have been especially problematic for recent H3N2 strains. Since the 2014–2015 season, a new antigenically distinct clade (3C.2A) of H3N2 viruses began circulating (73). These viruses are unique compared with previously circulating strains because they possess an additional asparagine-linked glycosylation site (NXS/T) in an antigenically important region of HA. However, 3C.2A viruses cannot replicate efficiently in eggs without the acquisition of amino acid substitutions that abrogate the new glycosylation site. Egg-based 3C.2A H3N2 human vaccine strains possess an amino acid substitution at HA residue 160 (T → K) that abrogates the new glycosylation site, which subsequently dramatically alters antigenicity (70).

Figure 2.

Cultivation of human influenza viruses in eggs can result in egg-adapted amino acid substitutions in the hemagglutinin (HA) protein. Egg-adapted amino acid substitutions in important HA epitopes can cause an antigenic mismatch between the vaccine strain and circulating viruses. When this happens, antibodies elicited by the vaccine cannot bind and protect against circulating influenza virus strains.

INFLUENZA VACCINES THAT DO NOT RELY ON EGGS

There are alternatives to egg-based influenza vaccines that are also licensed in the United States and some other countries worldwide. In the United States, alternatives to egg-based influenza vaccines include vaccines with antigens prepared in mammalian cell cultures (Flucelvax, Seqirus) and vaccines with antigens prepared via recombinant protein–based technologies (Flublok, Sanofi Pasteur). Flucelvax is an IIV with antigens isolated from viruses propagated in Madin-Darby canine kidney (MDCK) cells (Table 1). Flucelvax has been licensed in the United States since 2013. While Flucelvax has always included antigens propagated in MDCK cells, egg-adapted candidate vaccine viruses (CVVs) were used as initial seed strains for the vaccine until 2017. In 2017, the WHO began recommending alternative H3N2 and influenza B CVVs that are isolated from cells, and now all four vaccine components of the 2019–2020 Flucelvax vaccine are isolated from cells (74). It is clear that human influenza viruses can acquire HA substitutions when passaged in MDCK cells (75, 76), but it is thought that these adaptive substitutions are less prevalent compared with egg-adapted substitutions. Unlike with Flucelvax, antigens for Flublok are produced in insect cell cultures using a baculovirus expression system. Glycans produced in insect cell cultures are relatively small compared with those produced in eggs or mammalian cells, but the effect of glycosylation on immunogenicity is not well understood (77). There are higher amounts of HA antigen in Flublok (45 ug of each HA compared with 15 ug of each HA in standard doses), and there is no NA antigen (78). Importantly, Flublok production does not require selection for egg- or cell-adapted HA substitutions, which potentially minimizes the chance for antigenic mismatches between vaccine antigens and circulating viral strains (78).

IMMUNOGENICITY AND VACCINE EFFECTIVENESS ARE VARIABLE BETWEEN DIFFERENT INFLUENZA VACCINE TYPES

Annual influenza VE data mostly represent VE of egg-based IIVs because the vast majority of influenza vaccine antigens are prepared in eggs. However, recent data suggest that different vaccine formulations have differences in immunogenicity and VE each season (70, 79–81).

Antigen dose clearly affects immunogenicity and VE of influenza vaccines. A high-dose egg-based vaccine elicited significantly higher HAI titers compared with standard-dose vaccines in adults ≥65 years of age during the 2011–2012 and 2012–2013 seasons (82). During the 2017–2018 season, high-dose vaccine recipients also had higher postvaccination neutralizing antibody titers to both egg-adapted and wild-type 3C.2A H3N2 viruses compared with standard-dose vaccine recipients (79). High-dose egg-based vaccines also appear to elicit more NA-specific antibodies than standard-dose vaccines (83), but the lack of information regarding NA content in egg-based influenza vaccines makes it difficult to interpret these data. High-dose egg-based vaccines were more effective than the standard-dose vaccine for prevention of influenza infections, influenza hospital admissions, and postinfluenza deaths in adults ≥65 years during the 2012–2013 season, when H3N2 viruses predominated in the United States (84, 85). Interestingly, no significant increase relative to standard-dose vaccines in prevention of postinfluenza deaths was observed during the 2013–2014 season, which was predominated by H1N1 viruses (85).

Antigenic differences between different vaccine formulations can also affect immunogenicity and VE. For example, the H3 component of the recombinant protein–based Flublok vaccine elicited higher antibody titers against circulating 3C.2A H3N2 virus strains compared with egg-based and cell-based vaccines during the 2017–2018 season (79, 80). Interestingly, the 2017–2018 cell-based Flucelvax vaccine elicited weak antibody responses against circulating 3C.2A H3N2 viruses even though a cell-based CVV was used for that vaccine (79). It is unclear if this is because the Flucelvax vaccine possesses unidentified HA cell culture adaptations or if H3 HAs produced in mammalian cells are inherently poorly immunogenic. An observational study in the United States in adults ≥65 years showed only an 11% increase in relative H3N2 VE for the cell-based Flucelvax vaccine and a 9% increase in relative H3N2 VE for the egg-based Fluzone High-Dose vaccine compared with the standard-dose egg-based Fluzone vaccine (86). Flublok VE estimates were not evaluated during the 2017–2018 season due to the small number of people who received the recombinant protein–based vaccine. When the protective efficacy of Flublok was evaluated in adults ≥50 years of age during the 2014–2015 season, which was predominated by H3N2 viruses that were antigenically mismatched to the H3N2 vaccine component, the probability of influenza-like illness was 30% lower with Flublok compared with the standard-dose egg-based vaccine (87).

Immunogenicity data are often used as a predictor for VE; however, immunogenicity data can be misleading unless proper antigenic assays are employed. For example, circulating 3C.2A H3N2 viruses do not agglutinate most types of red blood cells efficiently and so cannot be used in HAI assays. Egg-adapted 3C.2A H3N2 viruses can agglutinate red blood cells efficiently, but these viruses are not antigenically similar to circulating 3C.2A H3N2 viral strains (70). Therefore, HAI titers to egg-adapted 3C.2A H3N2 are irrelevant for predicting VE against 3C.2A H3N2 infection. During the 2016–2017 season, in vitro neutralization assays with a wild-type 3C.2A H3N2 viral strain showed that Flublok elicited better 3C.2A H3N2 antibody responses compared with Fluzone and Flucelvax (70). However, HAI assays with an egg-adapted 3C.2A H3N2 viral strain and samples from the same cohort were unable to detect these differences (88). A separate study demonstrated that neutralizing antibodies against a wild-type 3C.2A H3N2 viral strain, but not an egg-adapted 3C.2A H3N2 viral strain, were associated with protection during the 2017–2018 influenza season (89).

IMMUNE HISTORY AFFECTS INFLUENZA VACCINE EFFECTIVENESS

Most humans are infected with influenza viruses by 3–4 years of age (90) and subsequently infected with antigenically distinct strains later in life. Infections later in life often boost antibody titers against viral strains encountered in childhood through a process termed original antigenic sin (OAS) (91). Initial childhood encounters are important for establishing lifelong immune memory that affects susceptibility to seasonal and pandemic influenza virus strains later in life (55). For example, most H5N1 infections occur in younger individuals, whereas most H7N9 infections occur in older individuals (92). This unusual age pattern of infections is likely due to differences in initial childhood infections between younger and older individuals. Individuals born before 1968 were likely first exposed to an H1N1 or H2N2 virus in childhood because H1N1 viruses circulated between 1918 and 1957 and H2N2 viruses circulated between 1957 and 1968 (Figure 1). In contrast, most individuals born after 1968 were likely first exposed to an H3N2 virus in childhood because H3N2 viruses have dominated most influenza virus seasons since 1968 (Figure 1). Older individuals born prior to 1968 likely have some level of protection against H5N1 because the HA stalk domains of H5, H1, and H2 viruses are similar (92). Younger individuals born after 1968 likely have some level of protection against H7N9 because the HA stalk domains of H3 and H7 viruses are similar (92).

Immune history, and childhood exposures in particular, can also affect VE against seasonal influenza viruses. During the 2015–2016 season, low H1N1 VE was found among middle-aged individuals (93, 94) who were likely exposed to a unique H1N1 viral strain in the late 1970s and early 1980s (53). Similarly, a VE study of the 2007–2008 to 2017–2018 seasons found a reduction in the risk of H1N1 or H3N2 infections in individuals who were likely initially infected with H1N1 or H3N2 in childhood, respectively (95). During the 2018–2019 influenza season, individuals born around 1968 had increased protection against 3C.3A H3N2 viruses relative to other-aged individuals in the absence of vaccination; however, this protection apparently decreased upon vaccination through mechanisms that are not well understood (96, 97).

Multiple studies suggest that annual repeat vaccination has a negative effect on influenza VE during some influenza seasons, particularly for H3N2 (93, 98–101). The immunological basis for these observations remains unclear. Repeat vaccination can lead to diminished B cell responses, lower postvaccination antibody titers, and reduced antibody-affinity maturation (79, 88, 102–104). This could be related to higher prevaccination antibody titers in repeatedly vaccinated individuals (105, 106), although this is not the case in all studies (70, 79). Repeat vaccination may lead to faster postvaccination antibody waning, which could potentially increase the risk of influenza virus infections (107). It appears that negative effects of repeat vaccination are likely dependent on the antigenic relatedness of vaccine antigens and circulating strains (99). Repeat vaccination with the same vaccine strain over two consecutive influenza seasons reduces overall vaccine-induced B cell responses, whereas changing one of the vaccine strains results in more robust immune responses to the updated strain (108). Moreover, changing the vaccine formulation induces antibody repertoires that can more effectively react with antigenically drifted viral strains (109).

NEXT-GENERATION INFLUENZA VACCINES

New universal influenza vaccines are being sought to overcome problems associated with current seasonal influenza vaccines. The goal is to develop new vaccines that elicit broad immune responses against antigenically diverse viral strains. One of the most popular universal influenza vaccine targets is the HA stalk domain (110). The HA stalk domain, unlike the HA head domain, is partially conserved across multiple influenza A virus strains. HA can be divided into two phylo-genetically distinct groups: group 1 (e.g., H1, H2, H5) and group 2 (e.g., H3, H7) (111). Higher baseline group 1 HA stalk-specific antibodies were associated with a reduction in the duration of viral shedding and disease severity in a human H1N1 challenge study (112). H1 stalk titers were associated with protection against naturally acquired H1N1 infection during the 2015–2016 season (113), a finding that was subsequently confirmed with a larger cohort (42). Phase I clinical trials are underway to determine the immunogenicity and safety of two vaccine strategies aimed at eliciting group 1 HA stalk-specific antibodies (https://www.clinicaltrials.gov/ identifiers NCT03300050, NCT03814720) (114). Interim reports demonstrate that chimeric HAs elicited group 1 HA stalk antibodies in most vaccines at early time points after vaccination (114). Because HA stalk antibodies are not usually as potent as HA head antibodies, HA stalk-based vaccines will need to be designed to elicit high levels of antibodies that do not wane over time.

Another potential target for new influenza vaccines is NA. Preexisting anti-NA antibodies are associated with reduced viral shedding and protection against infection in humans (115–117). Anti-NA titers were an independent predictor of protection and reduced disease severity in a human challenge study (118). Unlike most anti-HA head antibodies, anti-NA antibodies are largely non-neutralizing and likely confer immunity through preventing the release of virus particles from the host cells (119). NA evolves slower than HA; however, it is clear that NA routinely acquires substitutions in key epitopes recognized by antibodies (120, 121). For example, broadly reactive NA monoclonal antibodies were identified in a 2018 study (122); however, a subsequent study demonstrated that many of these monoclonal antibodies could not bind to a drifted viral strain that has been circulating since 2016 (123). Similar to HA stalk antibodies, NA-based universal influenza vaccines will likely need to be designed to elicit persistently high levels of antibodies because most NA antibodies do not prevent viral infections.

Other universal influenza vaccine targets include conserved viral proteins such as matrix protein 2 (M2) and nucleoprotein (NP). M2 forms a tetrameric ion channel necessary for viral entry and egress, while NP interacts with the viral RNA genome within the virion and is essential for genome packaging (reviewed in 124). M2 and NP are highly conserved across human seasonal viruses and avian influenza viruses, making them attractive targets. M2 ectodomain (M2e)– specific antibodies are protective in animals (125, 126), and NP antibodies reduce disease severity (127). In addition to the production of anti-NP antibodies, NP stimulates T cell immune responses and is the major target for cross-reactive cytotoxic T lymphocytes (128). Fused-M2e recombinant vaccines are under investigation in phase I/II clinical trials (identifiers NCT0378939, NCT00921947). Trial outcomes are pending, but preliminary data indicate that an M2e-flagellin vaccine was well-tolerated and elicited increased M2e antibody titers (129). Multiple phase I/II clinical trials are investigating novel vaccine platforms aimed at boosting NP immunity either alone or in combination with current seasonal influenza vaccines (identifiers NCT00942071, NCT03883113).

THE IMMEDIATE FUTURE: IMPROVING SEASONAL INFLUENZA VACCINES

It is exciting to develop entirely new influenza vaccine platforms; however, it is important to also employ simple solutions that could immediately improve current-season influenza vaccines (130). Seasonal influenza vaccines can be improved by applying our knowledge about influenza virus evolution, egg adaptation, OAS, and repeat vaccination. For example, the vaccine strain selection process has been improved in recent years, and it is exciting to see that sophisticated serological analyses and viral forecasting models are now being used to select more appropriate seasonal influenza vaccine strains (54, 55). Significant progress has been made with respect to viral forecasting models during the past few years (58, 63), and this area needs further investment and expansion.

We should also consider moving away from producing influenza vaccine antigens in embryonated chicken eggs because egg-adapted substitutions can alter influenza virus antigenicity. While studies clearly show that egg-adapted HA substitutions can alter antigenicity (69–72), additional studies need to measure antibody responses in larger cohorts and directly compare the VE of different vaccine formulations across multiple influenza virus seasons. Several ongoing studies are measuring immunogenicity and effectiveness of the Flublok vaccine relative to other vaccines. For example, the Marshfield Clinic has initiated the Randomized Influenza Vaccine Evaluation of Immune Response (RIVER) study that will compare immune responses elicited in 350 adults aged 18–64 years old immunized with Flublok, Flucelvax, and FluLaval (egg based) (identifier NCT03598439). The US Centers for Disease Control and Prevention has initiated a study that will compare the immunogenicity of Flublok, Flucelvax, Fluarix (egg based), and Fluzone in 864 healthcare personnel aged 18–64 years (identifier NCT03722589). Hong Kong University is continuing the Study on Influenza Vaccination Strategy (Project PIVOT) that compares immunogenicity of four vaccines, including Flublok, in 2,200 adults aged 65–82 years of age (identifier NCT03330132). Finally, the US Department of Defense is completing the Pragmatic Assessment of Influenza Vaccine Effectiveness in the DoD (PAIVED) trial, which will measure vaccine efficacy in 10,650 adults receiving Flublok, Flucelvax, and egg-based vaccines (identifier NCT03734237). If these studies conclusively determine that recombinant protein–based vaccines provide better protection relative to egg-based vaccines, then governments and philanthropic organizations should give incentives and support to vaccine manufactures to move away from egg-based vaccines.

Overcoming immune biases caused by prior influenza virus infections might be the biggest challenge to improve seasonal influenza vaccines. Additional studies are needed to better understand why human antibody responses to influenza vaccines are often biased toward epitopes conserved in viral strains that were encountered early in life. While this can be advantageous when these responses are directed against neutralizing epitopes, antibodies elicited by seasonal influenza vaccines often bind with a much higher affinity to past circulating strains relative to the vaccine strain that elicited the response. Further studies also need to be completed to determine how previous vaccinations affect the induction of protective responses and why repeat vaccinations are associated with lower VE in some influenza virus seasons (56, 99, 100).

Influenza vaccines will undoubtably be improved in the coming years as basic science guides vaccine design and development. There have been so many important advances in improving influenza vaccines over the past decade, yet there are many scientific and practical hurdles that must be overcome. It is an exciting time to study influenza vaccines, and the future looks bright.