Control and clearance of influenza virus infection involves most components of the innate and adaptive immune system. Innate immune components keep the viral infection in check while the adaptive immune response, composed of T and B cells, develops with the aid of antigen-presenting cells (APCs) [14]. Influenza-specific antibodies, produced by activated B cells, play a major role in resistance to influenza infection, while antibodies to other viral proteins, in particular the neuraminidase (NA), also contributes, most likely to the amelioration of clinical disease. In addition to directly limiting virus entry or release from infected cells, virus-specific antibodies may also aid in virus elimination through initiation of the complement cascade and antibody-dependent cellular cytotoxicity; mechanisms that have not been well examined with influenza.

Because of their ability to block virus attachment and entry into host cells, neutralizing antibodies directed against the globular head of the HA molecule are recognized to be the most powerful mediators of resistance to influenza infection and are considered the primary immune correlate of protection. The HI assay, because of its relative simplicity, has long been used as a surrogate assay for detection of virus-neutralizing antibodies in serum. Numerous studies have demonstrated resistance to influenza infection in persons with pre-exposure HI titers ≥32 or 40 [15–17]. This has led to the general convention of using a threshold HI titer of ≥40 as a measure of 50% reduction in the risk of influenza, sometimes referred to as seroprotective titer. Meta-analyses support the findings of individual studies but also demonstrate that higher serum HI titers are associated with higher rates of protection [18,19]. Although this HI titer threshold is the only universally accepted laboratory correlate of protection against influenza, several limitations of these earlier studies need to be considered in its generalized use for vaccine evaluation in all age groups. These include the use of attenuated challenge viruses to assess protection [17] and the fact that studies were generally conducted in younger adults that had acquired antibodies through natural infection rather than vaccination. In one study that did assess protection in adults following receipt of IIV, only serologic methods were used to detect influenza infection [16]. Recently, it has been recognized that serological end points, evaluated as a fourfold or greater rise in postinfluenza season HI antibody titer compared with preseason titer, may overestimate the protective effect of inactivated vaccines in adults because detection of such rises is compromised by already high preseason titers due to vaccination [20]. Indeed, a recent study in children that received an inactivated vaccine demonstrated that HI titers greater than 100 were associated with a 50% reduction in clinical influenza illness in laboratory-confirmed cases and HI titers over 200 predicted higher rates of resistance to clinical illness [21]. On the other hand, Ohmit et al. demonstrated that although absolute postvaccination HI antibody levels correlated with protection in adults vaccinated with either IIV or LAIV, IIV failures in particular had high HI titers that were not protective, indicating that perhaps quality as well as quantity of HI antibody, or indeed other immune parameters, need to be considered [11].

Recently, multiple groups have described the isolation of human monoclonal antibody (mAb) directed against the stem region of HA from memory B cells of adults exposed to influenza virus through infection or vaccination. These mAbs bind to non-contiguous epitopes within the highly conserved stem region of HA, blocking conformational changes in HA needed for membrane fusion, thus inhibiting virus replication [22–25]. These broadly cross-reactive antibodies exhibit neutralization activity in vitro and prophylactic and therapeutic protection in vivo against challenge with multiple virus subtypes, making them a target for broadly cross-reactive vaccines [26,27]. Such antibodies were also shown to contribute to clearance of virus-infected cells through antibody-dependent complement cytotoxicity, further demonstrating that anti-HA antibodies may inhibit and/or reduce virus replication and disease impact by multiple discrete mechanisms [28].

NA, the second glycoprotein on the influenza virus surface, has sialidase activity that promotes virus release from infected cells [29]. Antibodies to NA inhibit release of the virus from infected cells, reduce virus replication and prevent disease in animal models and humans [30–34]. In humans immunized with inactivated vaccines bearing relevant human NA but irrelevant (equine) HA antigens, the level of serum anti-NA antibody was inversely related to the clinical illness observed. The highest anti-NA titers in these studies were associated with asymptomatic infection following homotypic experimental virus challenge or natural infection [32–35]. Clements et al. also found that anti-NA serum antibody acquired through natural infection or immunization with live-attenuated or inactivated vaccines was significantly associated with reduced virus replication in volunteers experimentally challenged with wild-type influenza A viruses [36]. Epidemiological studies suggest that anti-N2 NA anti-body induced by prior exposure to H2N2 viruses may have reduced the disease burden of the 1968 H3N2 pandemic [37].

In addition to the antibody responses generated against HA and NA, repeated exposure to influenza viruses elicits antibody responses to more highly conserved viral proteins; however, the range of conserved protein epitopes recognized and the role of such antibodies in promoting viral clearance or disease prevention in humans remain poorly understood. Studies in the mouse model provide some evidence as to the ability of antibodies to influenza A M2 protein and nucleoprotein (NP) to reduce virus replication and ameliorate disease. Although antibodies directed against the ectodomain of the M2 protein (M2e) do not neutralize the virus, numerous mouse studies have demonstrated the benefits of passive transfer of anti-M2e antibodies in reducing lung viral titers and ameliorating disease [38–40]. Antibodies that bind to M2, which is abundant on the surface of influenza A virus-infected cells, can mediate antibody-dependent cell-mediated cytolysis, which contributes to protection in mice [41]. In humans, influenza A virus infection appears to elicit only weak, transient M2 antibody responses [42,43], although mAbs derived from human memory B cells that recognize native M2 protein show similar properties to murine counterparts [44,45]. The relative conservation of the M2e sequence, particularly among human influenza viruses, makes it an attractive target for the development of more broadly cross-reactive influenza vaccines [7,46]. Early studies demonstrated that mAbs directed against the NP and matrix 1 (M1) proteins exhibited no protective effect in passive transfer experiments in mice [47]. However, more recently, polyclonal anti-NP antiserum derived from mice hyperimmunized with recombinant NP protein was shown to modestly reduce lung virus titers and lessen morbidity using a low-dose challenge model [48]. The use of whole-genome fragment phage display libraries to characterize the serum antibody response in influenza A H5N1-infected persons has identified antibodies with strong reactivity to the viral PB1F2 protein, a proapoptotic protein associated with virulence of influenza A viruses in mice [49,50]. Whether anti-PB1-F2 antibodies are elicited by other influenza A subtypes and their role, if any, in protective immunity remain to be determined.

While it has long been recognized that immunoglobulin in the respiratory mucosa contributes to control of influenza virus, the relative roles and importance of locally produced IgA versus plasma-derived IgG remains controversial, but likely depend on the site of action within the respiratory tract. Polymeric IgA (pIgA) is the primary mucosal antibody that protects mucosal surfaces together with pentameric IgM [51]. Mucosal IgA can neutralize influenza viruses inside secretory epithelial cells [52]. In young adults, protection from virus replication or illness was significantly correlated with HA-specific nasal wash IgA acquired through natural infection or vaccination with LAIV [36]. By contrast, nasal wash IgG derived from plasma by passive transudation was associated with resistance to influenza in individuals that had acquired antibody through parenteral receipt of inactivated vaccine [36,53]. These results illustrate a fundamental difference in immune correlates of protection between LAIV and IIV [54]. The mouse model has been used to better understand the ability of pIgA and IgG to control influenza virus infection. Studies in IgA-deficient mice have suggested that IgA is not essential for reduction of virus replication in the nasal cavity, although others have argued that enhanced secretory IgM may compensate for a lack of IgA both in mice and humans [51,55]. Mice impaired in the transepithelial transport of pIgA are impaired in their ability to reduce influenza virus titers in the nasal cavity [56]. In the absence of other influenza-virus specific responses, pIgA alone was shown to neutralize and eliminate the virus from the murine upper respiratory tract and prevent virus-induced damage to respiratory epithelium. However, in mice with only high levels of influenza-specific serum IgG, transudated plasma IgG can neutralize newly replicating virus in the murine lung [57,58].

T-cell immunity requires antigens to be processed within cells and presented on their surface bound to MHC molecules, known as HLA in humans. CD4 T cells recognize exogenous antigens that have been internalized, processed and presented in the context of HLA class II, while CD8 T cells recognize endogenous antigens produced inside the cell, such as in the case of an infecting virus, that have been processed and presented in the context of HLA class I. Although less efficient, some APCs, primarily dendritic cell subsets, have the capacity to cross-present internalized antigens in the context of HLA class I to CD8 T cells [59]. This exemplifies one of the fundamental differences between the mechanisms of IIV and LAIV. IIVs stimulate antibody production and can be internalized by APCs to stimulate CD4 T cells, but are incapable of active replication in cells and therefore less effective at stimulating CD8 T cells. LAIVs are capable of limited replication in cells, and therefore more effective at stimulating CD8 T cells in addition to CD4 T cells and antibody production.

Antibodies are capable of binding and neutralizing live virions, making the development of vaccines that stimulate a strong antibody response attractive for influenza protection. A strong, strain-specific antibody response has the potential to afford neutralization without infection of host cells. T cells can be activated without infection of host cells, albeit less effectively in the case of CD8 T cells; however, they can only eliminate viruses after cellular infection via lysis of infected cells or by inducing an antiviral state in infected cells. Virus-specific CD4 T cells are critical to the development of protective immunity, primarily by helping B cells and CD8 T cells by secreting cytokines to support the immune response. However, a subset of CD4 T cells has been shown to have a direct cytotoxic function [60–63]. Virus-specific CD8 T cells eliminate infected cells by releasing cytotoxic granules such as perforin and granzymes or by inducing apoptosis through Fas/Fas ligand interactions. CD8 T cells also support the immune response through cytokine production. While CD8 T cells have been shown to play a role in protection from influenza, these responses are likely subordinate to antibodies in vaccinated individuals when vaccines are well matched to the circulating strains. Unfortunately, vaccines are not always well matched to the circulating strain due to the amount of time required to produce the seasonal influenza vaccine and the speed at which influenza mutates. In contrast to antibody epitopes, T-cell epitopes are mainly derived from internal proteins that are more conserved between subtypes and are able to confer immunity to heterologous as well as homologous influenza viruses [64–68]. For this reason, there is now heightened interest in the ability of influenza vaccines to generate antigen-specific T-cell responses, especially in the context of vaccines designed to protect against pandemic influenza, a situation in which antibodies may not be well matched to the emerging strain [69–71].

To determine the role of T cells in protecting against influenza and identify potential cellular correlates of protection, a few experimental human infections have been conducted using live, unattenuated influenza viruses. In earlier studies, 63 volunteers were inoculated intranasally with A/Munich/1/79 virus [67]. In these experiments, both antibody and cytotoxic T cells correlated with protection. In subjects lacking neutralizing antibodies, the level of influenza-specific cytotoxic T lymphocytes (CTLs) correlated with viral clearance, but not susceptibility to infection. These initial studies did not distinguish CD4 and CD8 T cells.

More recently, human subjects experimentally infected with either H1N1 A/Brisbane/59/2007 or H3N2 A/Wisconsin/67/05 demonstrated that pre-existing influenza-specific T cells were associated with protection in subjects with no detectable neutralizing antibodies [72]. The magnitude of the peak CD4 responses in these subjects correlated with reduced viral shedding, illness duration and total symptom scores. Responding T cells in these studies primarily recognized antigens from the internal M1 protein and NP. The T cells in these trials were good producers of IFN-γ as determined by enzyme-linked immunosorbent spot (ELISpot) assay, and likely played an important part in the antiviral response. In addition to pre-existing influenza-specific CD8 T cells capable of cytotoxic killing, a subset of subjects in this study were demonstrated to exhibit influenza-specific CD4 T cells capable of direct cytotoxic activity pre-vaccination. Despite the demonstration of substantial CD8 T-cell responses, CD8 T cells were not well correlated with reduced shedding or severity of disease in these experiments, likely due to the high variability in CD8 responses between subjects and the small number of subjects in the study. Sampling time in this study was also likely to contribute to difficulties in the evaluation of CD8 T-cell responses as a significant proportion of influenza-specific CD8 T cells were likely localized to the site of infection at 7 days postinfection.

Older adults are considered at high risk for influenza-associated complications due to dysregulation of the immune system, termed immunosenescence. While this topic is beyond the scope of this article, immunosenescence and challenges associated with influenza vaccination in older adults are more comprehensively reviewed by McElhaney [73] and Reber et al. [74]. Young children are also considered at high risk for influenza-associated complications, due in part to a naive and developing immune system. Bodewes et al. [75], PrabhuDas et al. [76] and Hodgins and Shewen [77] provide comprehensive reviews of childhood immune development and challenges associated with influenza vaccination in children. Experiments evaluating protection in these populations emphasize the importance of T-cell responses. Experiments performed in adults aged 65 years and older demonstrated that T-cell responses correlated with protection from influenza infection [78,79]. Experiments in young children (age 6–36 months), also demonstrated a protective correlation with the level of IFN-γ-producing cells induced by a LAIV vaccine [80]. In these experiments, conventional HI titers were shown to be poorer correlates of protection in these at-risk populations. These experiments highlight the need not only for the establishment of alternative correlates of protection, but also for confirmation in populations of more diverse age and health status.

As described earlier, T-cell responses require presentation of antigens in the context of HLA. HLA complexes are some of the most polymorphic genes in the human genome. Over 1000 allelic variants have been identified, each with its own unique peptide-binding properties, and thus its ability to stimulate a respective T-cell response [81–83]. This has presented a major hurdle for the study of T-cell immunity and indirectly, their use as a correlate of protection.

The understanding of CD8 T-cell responses has increased significantly with the development of HLA class I tetramers. HLA class I tetramers are four linked HLA class I molecules presenting a defined peptide antigen. This technology allows identification of antigen-specific CD8 T cells. However, the polymorphic nature of the HLA genes has resulted in most studies being limited to a handful of common alleles, HLA-A2 being the most prominent due to its high frequency in western populations [84–86]. Despite this drawback, significant progress has been made in understanding CD8 T-cell immunity. By contrast, understanding of CD4 T-cell responses has lagged behind due to difficulties associated with the development of HLA class II tetramers. This is unfortunate considering the importance of CD4 T-cell responses in the development of protective immunity.

Mouse studies have shown that mice lacking functional CD4 T cells exhibit severely impaired immune responses to influenza and significantly shorter lived immune memory [87]. In humans, antigen binding to HLA class II, and thus presentation to CD4 T cells, is important for immune development. The HLA class II alleles HLA-DRB1*03 and DQA1*0201 have been associated with seronegative or low antibody responses to hepatitis B and/or measles vaccines, while DQA1*0104 and DPA1*0202 alleles were associated with high antibody levels [88–93]. A small study examining individuals who failed to mount a neutralizing antibody response to influenza vaccination found an unusually high frequency of individuals expressing the HLA-DRB1*0701 allele and an unusually low frequency expressing the HLA-DQB1*0603 allele, suggesting potential genetic correlates with vaccine failure [94]. Further study in this area may provide useful information to improve vaccines and provide protection in a broader proportion of the population.

The ability to elicit HA strain-specific serum antibodies is often used as a primary end point of influenza vaccine immunogenicity. Multiple host characteristics influence the ability to mount antibody responses to influenza vaccines, the primary ones being age and history of prior exposure to the virus, as well as the health status of the individual. A number of different methods are available to measure antibodies to the HA and have been described in detail elsewhere [95]. The HI assay is the most widely used to measure strain-specific anti-HA antibodies because of its relative simplicity and correlation with protection. This assay and several others are described in brief below.

Although it is well accepted that cellular immunity plays a significant part in immunological protection from influenza, there are currently no well-defined, broadly accepted cellular correlates of protection. However, there are a wide variety of techniques that have been developed and are frequently utilized for the quantitation of cell-mediated immune responses to influenza infection and vaccination. Any future cellular immune correlate of influenza protection is likely to have its basis in one of the following techniques.

Interlaboratory variability in assay techniques and determination of assay end points poses a considerable challenge for comparing the immunogenicity of influenza vaccines in clinical trials. While there are currently accepted criteria for HI and SRH assays, many of the serological and cellular assays lack such established criteria. Lack of standardized protocols and knowledge as to what constitutes a positive response to vaccination and how it relates to protection leads to variability in results between laboratories. Some laboratories utilize fold increases in pre- to post-vaccination titers for some serological assays, using criteria similar to the HI assay, while others rely on statistical significance to determine immunogenicity. T-cell assays in particular are prone to these concerns. A wide variety of assay protocols are utilized by a number of laboratories. Statistical significance is usually used as the determining criteria for immunogenicity without an understanding of how the responses relate to protection from disease. A concerted effort by stakeholders is needed to establish standardized techniques and develop the knowledge base needed for establishment of assay criteria.

Recent global interest in development of H5N1 prepandemic vaccines and the 2009 H1N1 pandemic has renewed interest to evaluate the extent of interlaboratory variability and develop tools for improved standardization to aid the regulatory process. International studies that have compared both inter- and intra-laboratory variation among HI and VN assays have demonstrated poor reproducibility of these assays, with the more technically demanding VN assay showing even greater interlaboratory variability than the HI assay [117,184,185]. For the HI assay, the species and method of standardization of RBC as well as efficiency with which sera are treated to remove nonspecific inhibitors may contribute to variability in results between laboratories. For VN assays, differences in protocols, quality of cells used and, in particular, the amount and standardization of virus used, likely contribute to poor reproducibility and interlaboratory variability. The development and use of a standard antibody reagent to normalize results within a laboratory reduces interlaboratory variability in HI and VN assays by at least 50% [184–186]. Recently, the EMA assessed serologic assay variability between vaccine manufacturers and European regulatory agency laboratories using a defined subset of sera from A(H1N1)pdm09 clinical trials [187]. The substantial variability in antibody titers observed among manufacturer and regulatory laboratories was greatly reduced when absolute titers were calibrated relative to the A(H1N1)pdm09 international standard. Therefore, the use of antibody international standards is a new and powerful approach to improve inter laboratory agreement for serologic assessment of influenza vaccines. The use of standard operating procedures to qualify serologic assays should also improve intra- and interlaboratory variability [188]. The newly formed Consortium for the Standardization of Influenza Seroepidemiology is currently investigating options for the improved harmonization of influenza serological assays [189].

The conduct of well-designed clinical efficacy or effectiveness studies in different age groups with defined immunological endpoints is key to the development of new and reliable immune correlates of protection against influenza illness. Although the HI titer of ≥40 remains a useful and important benchmark for the qualitative assessment of protection in a population, there are limitations for the generalized use of this long-held correlate for assessing IIV-induced protective immunity in populations of different ages and risk groups [11,21,78,79]. New influenza vaccine strategies may target specific segments of the population, and it will be important to develop immune correlates that are similarly age group specific. The lack of knowledge and qualified methods to measure alternate immune markers of protection against influenza remain a substantial barrier to the development of more immunogenic, broadly cross-reactive and effective influenza vaccines. The lack of any laboratory correlate of protection for LAIV is problematic for future licensure of this technology, which is an important component of the WHO Global Influenza Action Plan to build global pandemic vaccine capacity in developing countries [190].

There is a growing consensus among influenza experts and regulatory authorities that qualified, well-standardized assays that quantify non-HI anti-HA antibodies, antibody directed against other viral components (e.g., NA or M2) or T-cell responses are needed to establish additional immune correlates of protection to support licensure of next-generation influenza vaccines [191–193,306]. TABLE 1 outlines prospective correlates of protection for influenza and assays currently used for assessment. European regulatory agencies are currently reassessing guidelines for influenza vaccine licensure to include vaccine type and subpopulation-specific considerations as well as a broader range of criteria for immunogenicity [307].

Since the relevance of VN assays is now well recognized, there is an urgent need to incorporate these assays into vaccine efficacy studies to identify a correlate of protection against laboratory-confirmed seasonal influenza. This should also aid in regulatory evaluation for vaccines against pandemic threats. Furthermore, additional attention to the development of standardized methods to quantify stem region-specific subtype cross-reactive antibodies, and/or those that mediate cell- or complement-mediated cytotoxicity is warranted. For stem-region antibodies, assays that detect inhibition of cell–cell fusion or proteolytic activation (trypsin cleavage) of HA have been used to understand the neutralizing mechanism of these mAbs [22,24]. The development of higher throughput assays that quantify fusion inhibition in HA-transfected cells may be one means to measure functional stem-region antibodies in human sera. Although assays to measure cross-reactive antibody-dependent cell or complement-mediated cytotoxicity activity on virus-infected cell lines have been described [28,194], further development of standardized methods is needed. For cell cytotoxicity assays, there is a need to standardize both the target and effector cells. This could be achieved by the use of a continuous cell line expressing the desired viral target (e.g., HA, NA or M2) and a human cloned effector cell (e.g., NK cell); appropriate human positive and negative control antibodies are also needed. Recently, Jegaskanda et al. described a novel antibody-dependent cell-mediated cytotoxicity assay that measured intracellular IFN-γ and degranulation (CD107a) expression by NK cells after incubation with immobilized influenza antigen and non-neutralizing antibody complexes [195]. Although this approach was evaluated using whole virus or purified HA, the use of alternate purified viral proteins may broaden the applicability of such an assay.

Substantial progress has been made in developing antibody standards for serologic assessment of pandemic and prepandemic vaccines, thereby establishing a process to reduce interlaboratory variability and promote comparability of different vaccines. This approach has not yet been adopted for seasonal influenza vaccines, but should be explored. Efforts are underway to unify and standardize methods for detection of NI antibodies, which may be a first critical step towards developing a laboratory-defined correlate of clinical protection. This may be particularly challenging for immune parameters such as anti-NA that may reduce disease severity rather than prevent infection and may require more intensive and costly studies to detect milder illness or asymptomatic infection. The human experimental challenge model may be a more controlled environment for this purpose, and has recently been used to evaluate a vector-based vaccine strategy [196].

Two areas of assay development warrant particular attention. Mucosal antibody responses likely contribute an important component of protection that is poorly quantified by existing assays, but could be particularly beneficial in describing correlates of protection for LAIV, and potentially some next-generation vaccines. Future LAIV trials should consider standardized sample collection and ELISA procedures for evaluation of nasal immuno globulin and cross validation of methods by a central expert laboratory. Although quantification of cell-mediated immune responses to influenza vaccines has become more common, multiple methods are in use to assess vaccine immunogenicity. Of these, it is likely that for T-cell responses the IFN-γ T-cell ELISpot coupled with detection of CD4⁺ and CD8⁺ T-cell phenotypes by flow cytometry following influenza virus-specific stimulation of peripheral blood mononuclear cells may be the most amenable to standardization in method, reagents and response criteria. While some progress has been made toward validation of T-cell assays, more work is needed, ideally by an international consortium of laboratories that can identify optimal methods, reagents and criteria for assessment of T-cell responses such as has been established for HIV vaccine development [197,198].

In the short term, those supporting and conducting clinical studies, particularly those assessing efficacy or effectiveness, should consider incorporating a broader range of laboratory assessments to identify additional immune markers that may predict protection from laboratory-confirmed influenza illness or disease. Recognized challenges in this regard include the altered timing and larger volume of blood required for optimal detection of cell-mediated immunity compared with serum antibodies. In the longer term, new high-throughput global gene expression technology will allow for expanded analyses of the immune system networks that may identify immunological signatures associated with protective responses. This approach has already revealed fundamental differences in signatures elicited by IIV versus LAIV, and/or identified early molecular signatures correlated with subsequent anti-HA antibody responses [199,200]. In the future, this systems biology approach should have the power to define a complex of immune correlates of protection against influenza that are age group and vaccine specific.