In sharp contrast to viruses such as measles, in which a single exposure generates long-lasting immunity, individuals are repeatedly exposed to influenza viruses throughout their lifetime without developing broad and durable protection against infection. As a consequence, substantial worldwide morbidity and mortality associated with influenza infection remain, as herd immunity by natural infection is never established to protect the population at large. In this regard, influenza epidemics (caused by influenza A and B viruses) account for an estimated 291,243–645,832 respiratory deaths worldwide each year (1), including ∼12,000–56,000 deaths in the United States (2). Additionally, influenza A viruses (IAVs) have caused severe worldwide pandemics of varying degrees of mortality, the most devastating of which occurred in 1918 and resulted in an estimated 50–100 million deaths worldwide (3).
The mainstay of seasonal influenza prevention and control is vaccination. Influenza vaccines are important tools in our armamentarium against influenza infection and disease. They reduce infection rate and influenza-related complications and deaths in some populations. However, they are less efficacious than vaccines for many other infectious diseases. For example, in the United States, influenza vaccine effectiveness (VE) against medically attended illness ranges between 10% and 60% depending on the season (4). Additionally, as a result of time constraints imposed by current manufacturing practices, influenza vaccines have historically had limited public health impact during pandemics (5).
Antigenic drift and shift
One fundamental stumbling block in the current approach to influenza vaccine design is that the virus itself is a constantly moving target, often driven by immune pressure from previous exposures (6–9). IAVs rapidly escape preexisting immunity because of the evolution of an error-prone RNA polymerase that lacks proofreading capacity and garners mutations in two key viral surface proteins, hemagglutinin (HA) and neuraminidase (NA) (10). Consistent mutability combined with strong Ab-mediated selective immune pressure at the population level leads to important antigenic changes from year to year, termed “antigenic drift.” Despite this evident problem of antigenic drift, our current vaccination strategy primarily targets the highly variable head of the HA molecule. This strain-specific approach relies heavily on an exact match between the HA of the vaccine virus and that seen in circulating viral strains. If antigenic change occurs in circulating viral HAs, new vaccines must be developed to keep pace with this viral evolution. Furthermore, IAVs from other animal species can sporadically infect humans de novo or recombine segments of their genomes with circulating human IAVs, leading to the evolution and transmission of an IAV with an entirely antigenically novel HA (antigenic shift). Pandemics can result from outbreaks of influenza viruses that manifest antigenic shift. Thus, to protect susceptible populations from the effects of “antigenic shift,” entirely new influenza vaccines must be developed specifically to target the novel strain. Traditionally, theories regarding suboptimal influenza VE have centered on antigenic mismatch between circulating viral strains and the vaccine strain (11). Therefore, it seems logical that VE data for any given population should be fairly predictable from year to year based solely on characteristics of the circulating viral strain and its relationship to the vaccine virus. However, VE estimates show remarkably complex temporal and demographic trends that differ from season to season, vary among age cohorts within a season, and seem to be influenced by prior vaccination history (11). Hence, to better understand the limitations of our current influenza vaccines, we must first acknowledge that influenza immunity is unlikely to be static and that immunological memory generated through both natural infection and vaccination may impact all subsequent influenza exposures.
When we are exposed to or infected with a virus such as measles, which remains fairly constant over time, each subsequent exposure generally elicits consistent and reproducible B cell responses (12). However, with a constantly moving target such as influenza, each re-exposure to an antigenically distinct variant elicits mostly recall responses and usually to a lesser degree, some novel Ab and B cell responses (13–15). Practically, this means that the immune response to each re-exposure will only be slightly different from the previous immune response. The dynamics of these differential responses, in particular, pertaining to competitive dominance, are postulated to have critical implications for understanding protective immune responses to influenza. In this regard, it has been hypothesized that a preferential boosting of Ab and B cell responses against influenza viruses encountered in early life occurs with each subsequent influenza exposure. This concept was first postulated 50 y ago by T. Francis (16), who analyzed serologic data from several human cohorts in the 1940s–1950s (15). At the time, Francis and others (16) noted that Ab patterns varied among birth cohorts, that the breadth of Ab responses increased with age, and that influenza vaccination preferentially boosted Ab levels to historical influenza strains. These findings led to the theory that Ab first established by childhood influenza infection will continue to characterize and dominate subsequent immune responses to influenza both for the individual and the entire birth cohort within the general population. Francis (16) proposed the phrase “original antigenic sin” (OAS) to describe this phenomenon. Recently, this theory has been revisited as a potential explanation for the complexities seen in VE data as well as age-specific mortality patterns seen with influenza infection (15). Immunological and epidemiologic evidence is accumulating to lend credibility to both the positives and negatives of the OAS hypothesis, although the mechanisms involved in the process and how much impact every re-exposure plays in shaping subsequent immune responses are largely unknown.
Birth cohort data showing differential protection by age against influenza have long suggested that OAS may impact immunity throughout one’s lifetime. Mortality data from the 1918 influenza pandemic revealed that the majority of deaths occurred in individuals 20–40 y of age. This is unlike mortality associated with seasonal influenza, which is typically concentrated among infants and the elderly (17). To further underscore this point, the elderly actually had less influenza-related mortality during the 1918 pandemic than in any of the preceding influenza seasons (1911–1917). A potential explanation for these observations centers on the concept of OAS. Elderly individuals and the very young may have been relatively more protected from the 1918 H1N1 pandemic than other demographic groups because of an early-life exposure to an antigenically related H1 virus (15). Young adults, in contrast, were more likely exposed to an H3 virus in childhood, potentially hindering robust immunity to H1N1 influenza during 1918 and explaining their more severe clinical courses. Similarly, during the 2009 H1N1 pandemic, elderly individuals again had a decreased incidence of disease, suggesting that they were protected by exposure to H1N1 viruses circulating prior to 1957 (18). Individuals born after 1977, when H1N1 viruses returned to human circulation, did not (as a birth cohort) have the same level of protection against pandemic H1N1. This may have resulted from the fact that only a proportion of this age group would have had an initial encounter with H1N1 viruses when H3N2 viruses were cocirculating at the time. This observation may also be related to the degree of antigenic similarity between the post-1977 priming virus and the 2009 H1N1 pandemic strain or alternatively be related to some yet-undefined factor (18, 19). Interestingly, age-specific mortality patterns have also been seen in individuals infected with H5N1 (phylogenetic group 1) and H7N9 (phylogenetic group 2) avian influenza viruses (20). These data suggest that early-life exposure to seasonal human influenza strains can lend protection against divergent avian influenza viral subtypes if they are in the same phylogenetic group.
From an immunological perspective, it would be important to determine the basis of OAS and to modify this phenomenon to improve influenza vaccine approaches (21). During an initial influenza encounter, an individual develops Abs and memory B cell responses that target multiple different epitopes. Upon subsequent exposures, there is a recall of humoral and memory B cells to conserved epitopes that were likely initially elicited by the priming virus and that predominate and perhaps skew the overall response to the newly encountered strain (15, 21, 22). It is uncertain whether this phenomenon is unique and defined entirely by the first influenza encounter, be it infection or vaccination; defined by the first influenza encounter within a subtype or phylogenetic group; or a result of cumulative experiences over time. This question is currently a subject of intense scrutiny.
In any case, preexisting immunity to influenza viruses can have unforeseen consequences when the conserved epitopes generating the focus of the Ab-mediated immune response mutate. For example, during the 2013–2014 H1N1 influenza season, middle-aged adults were surprisingly disproportionately affected. A particular mutation, K166Q, occurred in the H1N1 HA head (23). This region of the HA was conserved between H1N1 strains circulating in the late 1970s and the pandemic H1N1 strain that emerged in 2009. Consequently, a substantial proportion of middle-aged adults generated a focused Ab response to this region in response to the 2009 influenza pandemic, which protected them during the 2009 H1N1 pandemic and any subsequent H1N1 exposures (23). However, the mutation at position 166 may explain why they were disproportionately affected during the 2013–2014 H1N1 season.
Another example of how the immune response can be affected by a seemingly innocuous single amino acid change arises from the vaccine manufacturing process itself. The majority of the world’s influenza vaccine supply is still generated in eggs, and during this process, mutations can occur in areas of the HA head that help adapt the virus to grow better in eggs but have the collateral effect of impacting an individual’s immune response to the vaccine (24). For example, a recent study by Garretson et al. (25) showed that 5% of a 159-person cohort had a ≥4-fold higher Ab response to a single egg-adapted mutation in the HA head of the egg-adapted H1N1 vaccine strain than to the circulating H1N1 influenza strain. Historically, it is not known whether such egg-adapted Ab responses impact overall immunity. However, based on recent data, we can speculate that Ab responses focused on an egg-adapted mutation in the 2016–2017 H3N2 influenza vaccine strain may have contributed to an overall decrease in VE against that strain (24).
HA stem immunity
Thus far, our discussion of OAS has focused primarily on neutralizing Ab responses to highly mutable regions of the HA head, which is arguably the dominant protective response against influenza and the best understood mechanism of immunity. However, protection seen against extremely divergent viral strains, such as the 2009 H1N1 pandemic strain, and the cross-group protection described by Gostic et al. (20), suggest that, at least in some cases, conserved regions of the influenza virus play a critical role in defining the immune response. In this regard, functional Ab responses do occur to less-mutable regions of the HA, such as the HA stem. This region of the HA remains relatively unchanged from year to year despite antigenic drift and is conserved between influenza subtypes within the two respective phylogenetic HA groups. Despite repeated exposures to antigenically drifted influenza viruses, most individuals have low levels of these HA stem Abs in serum (21), suggesting an immunologic hierarchy in which HA head Abs are generally immunodominant. Conversely, when exposed to more-divergent viruses created by antigenic shift, HA stem responses may be more pronounced. This was recognized following the 2009 H1N1 pandemic when infected individuals had increased HA stem–specific serum Ab and higher memory B cell frequencies because the HA head epitopes antigenically shifted, whereas the HA stem region remained antigenically similar (13, 26). However, evidence suggests that HA stem responses are transient in circulation, dissipating quickly following influenza exposure, yet they generally accumulate with increasing age (15). Therefore, it is unclear from currently available data whether functional Ab responses directed toward the HA stem are protective and, more so, if they are durable without a specific boosting strategy (27).
Immunity to NA
Abs directed to other influenza proteins may also play a role in protection against infection and transmission and theoretically contribute to shaping an individual’s immunity over time. In this regard, Ab to NA has emerged as a potentially important target of an individual’s immune response to influenza (28). Although Abs to NA are not neutralizing in vitro, they are effective at limiting viral spread. In animal models, NA is immunogenic (29) and is able to prevent overt disease despite not leading to sterilizing immunity (28). Multiple clinical trials (28) have shown that the NA inhibition titer is an independent predictor of protection against influenza. Additionally, recent data from human challenge models suggest that NA inhibition may actually be more predictive of clinical outcome indicators in this particular setting than measurements that assess Ab responses to the HA head (30) (hemagglutination inhibition titer) or HA stem (anti-HA stem Ab titers) (31).
It has been postulated for decades that NA immunity was critical in reducing the severity of infection during the 1968 H3N2 influenza pandemic (28, 29). It is also well established that NA drifts over time, again suggesting that it is subject to immune pressure and is an important target for population immunity (28). The role of NA immunity in an individual’s immune history to influenza exposure, including the concept of OAS, remains uncertain. However, Rajendran et al. (32) reported that Abs to NA differ in both magnitude and breadth among different age groups. Furthermore, similar to Abs to the HA head, NA Ab titers in middle-aged adults and the elderly are highest against HA subtypes encountered in childhood. Although additional research is needed, this lends credibility to the hypothesis that lifelong NA responses are also greatly impacted by early influenza infection.
Universal influenza vaccine
As discussed above, our current approach to influenza vaccination leaves us constantly trying to chase a moving target. The ideal influenza vaccination strategy would be a “universal” influenza vaccine approach that targets the immutable components of influenza and puts us one step ahead of this elusive virus. Such a vaccine would ideally give ≥75% protection against symptomatic influenza disease caused by both group I and group II IAVs, have a durability of at least 1 y and preferably much longer, and be effective in all demographic groups (33). To achieve this goal, we must improve our research tools and fill the knowledge gaps in our scientific understanding (33). First, our current animal models for studying influenza are for the most part immunologically naive with regard to influenza. One problem with this approach has been illustrated in the past by surveillance data collected by the World Health Organization and its collaborating centers. Historically, antigenic distance between circulating and vaccine viruses was judged using ferret sera (animals without preexisting influenza immunity) and, in some cases, produced very distinct results from those generated using human sera (15). This also has important implications for developing new vaccination strategies as a vaccine may work well in the context of a naive host; however, it may fail to protect in the setting of historical immunity. Second, vaccine efficacy is often linked to the HAI titer, which is the best-studied correlate of protection against influenza (33). This assay relies entirely on hemagglutination, which is a property of the HA head, specifically the receptor binding site. The HAI titer has a number of limitations, including no defined threshold for 100% protection, variability between age groups and patient populations, and an inability to measure other non-HA head components of the immune response. To develop and test a universal influenza vaccine, we most certainly need novel ways of measuring immunity and animal models that can recapitulate the human experience.
Even if we are able to overcome the limitations in our current immunological tools for assessing influenza vaccines, we must still address the gaps in scientific knowledge needed to achieve the ultimate goal of universality. First, we must determine the choice(s) of antigenic target or targets for vaccine design. It seems clear that NA is an underused target that could improve the effectiveness and breadth of currently available vaccines. Most seasonal influenza vaccines do contain NA; however, the NA content for each vaccine is not standardized and may be too low to generate an optimal immune response. This was recently illustrated by Chen et al. (34) who demonstrated that influenza vaccines rarely induce NA-reactive B cells, whereas natural infection induces an NA-specific B cell response with a magnitude similar to that of HA-reactive B cells. A proportion of these B cells produced broadly cross-reactive Abs capable of binding a range of IAVs. Because of slower rates of drift, NA epitopes may be more resistant to antigenic change and potentially impact VE in the short term. Ultimately, however, it seems likely that to overcome selective immune pressure, more conserved epitopes may have to be identified as either a component of influenza vaccines or as a standalone universal vaccine design. In this regard, there is currently significant focus in the influenza research community on targeting conserved epitopes of the HA stem. To successfully generate immune responses to these epitopes, we will need to overcome the immunodominant response to the HA head. This will likely involve presenting the stem (or other subdominant epitopes) to the immune system in new ways, either through novel conformational constructs, different routes of immunization, exposure of naive individuals to different influenza subtypes sequentially in the context of platform technologies that boost the immune response to the target Ag, or administration with adjuvants that have previously been postulated to help overcome the phenomenon of OAS (21, 35, 36). Additionally, a better understanding of how T cell responses (particularly CD4+ T follicular helper cells) facilitate protection and shape Ab responses will also be important in optimizing vaccine design (37, 38). In this regard, several universal influenza vaccine candidates in clinical trials target T cell epitopes in hopes of increasing the breadth of protection against IAVs (39, 40). Finally, it is conceivable that we may find that immune history cannot be overcome. In that case, universal influenza vaccines may need to be targeted to various birth cohorts based on their immune history or alternatively given to infants in hopes of shaping immunity in a desired way from the start.
We will never achieve a universal influenza vaccine unless we understand the complexity of the immune response to influenza much more fully than we do now. This special topical review of The Journal of Immunology will highlight some of the unanswered questions regarding immunity to influenza, the tools needed to address these questions, and the research that is underway. In addition, understanding influenza immunity in the context of developing a truly universal influenza vaccine is the focus of a new strategic plan for the National Institute of Allergy and Infectious Diseases (41). The process almost certainly will be iterative and progressive, with vaccines first improving immune response against drifted versions of a single subtype such as H3N2, followed by vaccines that protect against all subtypes within an entire phylogenetic group (e.g., either group 1 or group 2 influenza viruses), followed by a truly universal vaccine that would protect against all IAVs. Until this is achieved, we will continue to chase and fail to consistently catch the constantly moving target of influenza.
The authors have no financial conflicts of interest.