Universal Influenza Vaccines: Progress in Achieving Broad Cross-Protection In Vivo

Abstract

Despite all we have learned since 1918 about influenza virus and immunity, available influenza vaccines remain inadequate to control outbreaks of unexpected strains. Universal vaccines not requiring strain matching would be a major improvement. Their composition would be independent of predicting circulating viruses and thus potentially effective against unexpected drift or pandemic strains. This commentary explores progress with candidate universal vaccines based on various target antigens. Candidates include vaccines based on conserved viral proteins such as nucleoprotein and matrix, on the conserved hemagglutinin (HA) stem, and various combinations. Discussion covers the differing evidence for each candidate vaccine demonstrating protection in animals against influenza viruses of widely divergent HA subtypes and groups; durability of protection; routes of administration, including mucosal, providing local immunity; and reduction of transmission. Human trials of some candidate universal vaccines have been completed or are underway. Interestingly, the HA stem, like nucleoprotein and matrix, induces immunity that permits some virus replication and emergence of escape mutants fit enough to cause disease. Vaccination with multiple target antigens will thus have advantages over use of single antigens. Ultimately, a universal vaccine providing long-term protection against all influenza virus strains might contribute to pandemic control and routine vaccination.

The deadly spread of influenza is a risk we all hope to reduce. Current vaccination strategies are inadequate because they induce mainly strain-specific immunity that is ineffective against unexpected strains. Universal influenza vaccines protective against all virus strains would be a great improvement. Work on universal influenza vaccines has been going on for decades; however, there has recently been increased interest in this topic. This commentary includes discussion of cross-protection against influenza virus infection and progress with universal vaccine candidates, specifically breadth and durability of protection in animals, escape mutation in target antigens, findings in human surveillance and challenge studies, and reported clinical trials. The discussion cannot cover all studies, but examples are presented.

HISTORICAL BACKGROUND

Influenza virus surface glycoproteins hemagglutinin (HA) and neuraminidase (Figure 1), targets of classical strain-matched vaccines, evolve rapidly. Nucleoprotein (NP), matrix 1 (M1) and matrix 2 (M2), polymerases (i.e., PB1, PB2, and acidic polymerase), and the HA stem change more slowly, so vaccines based on these antigens may protect more broadly. Cross-protection against virus of a particular influenza A subtype by immunity to another (i.e., mismatched HA and neuraminidase), for example, protection against H3N2 induced by H1N1, is known as heterosubtypic immunity, and has been studied in animal models since the 1960s (1–3). Cross-protection by wild-type virus was found to be induced by internal viral proteins (4–6). In mechanism studies, cross-protection was shown to be mediated by antibodies and/or T cells, depending upon conditions (7, 8). For the HA stem as target antigen, cross-protection by monoclonal antibody was demonstrated (9).

Figure 1.

Diagram of the influenza virus, indicating major protein components. Hemagglutinin (HA), neuraminidase (NA), nucleoprotein (NP), matrix 1 (M1), and matrix 2 (M2) are the main target antigens proposed for universal vaccines. NEP, nuclear export protein; NS2, nonstructural protein 2; PA, acidic polymerase; RNP, ribonucleoprotein.

Influenza A and B viruses do not generally protect against each other (10, 11) and are more divergent from each other than are influenza A subtypes (12), so vaccines based on one are not expected to protect against both. However, analogous approaches can be applied to influenza B vaccination. Influenza B virus HA is far less diverse than that of influenza A (13). In addition, conserved protein NP, a universal vaccine target, has been estimated to evolve more slowly in influenza B than influenza A (14). Thus, a universal influenza B vaccine, if it induced durable protective immunity, might be useful over a longer period. Some vaccine candidates cover both influenza A and B by including multiple components (15).

STRENGTH AND BREADTH OF PROTECTION BY ACTIVE IMMUNIZATION TO VIRUS PROTEINS

The outcomes described in the following paragraphs were chosen to show the extent of protection that has been achieved by an antigen, though the same antigen may fail to protect under other circumstances. Results are influenced by vectors or adjuvants, administration route (intranasal/mucosal vs. parenteral/systemic), vaccine or challenge virus doses and strains, and responder genetics. Protection experiments should include animals given control constructs with irrelevant antigens, not just buffer, because many vector backbones, production cell components, and matrices have nonspecific or innate immune effects.

Nucleoprotein

NP differs modestly between influenza A strains: There is greater than 90% amino acid identity between H1 and H3 viruses (group 1 vs. group 2) (16), but only 36% identity between influenza A and B (12). Immunization with NP and the resulting T-cell protection have been explored since the 1980s, using protein (17–20), DNA vaccines (21–23), and recombinant vectors (24). In mice and ferrets, NP vaccination prevented death and greatly reduced morbidity after challenge with mismatched strains, including highly pathogenic avian influenza virus (24). Enhanced protection was achieved by various prime-boost combinations (24, 25). NP epitopes shared among divergent viruses are recognized by human T cells, even when using viruses the donors have never encountered, including avian influenza viruses (26–28).

Matrix

M2 has a sequence shared by most human influenza viruses (i.e., H1, H2, H3), whereas M2s from avian influenza viruses, such as H5, are more divergent and less serologically cross-reactive (29). M2 has been extensively studied as a vaccine candidate. In a classic study, a fusion protein of the M2 ectodomain with hepatitis B core protein protected against lethal challenge, with protection transferable by passive antibody (30). An M2 ectodomain construct (31) and vectored M2 expression (32) were later shown to protect mice against challenge. Antibodies and T cells contribute to protection (32). A construct combining swine, human, and avian M2 sequences protected broadly (33), but even a single sequence can cross-protect against viruses with different M2 sequences. Vaccines based on M2 sequences of human influenza viruses protected against avian influenza viruses with differing M2 sequences (32, 34). An M2 conformational epitope may be particularly important; passive antibodies to it protect mice against both H1 and H5, according to research on human monoclonal antibodies (mAbs) (35).

M1 is not a dominant antigen in mice and is studied only occasionally (36, 37). It is a dominant human T-cell target and vaccine component in human trials, as is discussed later in this article.

Hemagglutinin stem

HA has a globular head (HA1 domain) that varies considerably and a more conserved stem (HA2 domain) (Figure 1). Figure 2 shows the sequence relationships among HAs, with divergence between group 1 and 2 subtypes, and influenza B lineages. In 1983, polyclonal antibodies to the conserved HA2 region were described (38). In 1993, Okuno et al. (39) discovered mAb C179, recognizing a conserved epitope on multiple HA subtypes. It did not inhibit hemagglutination but had neutralizing activity. The investigators prepared “headless” HA as an immunogen. The headless construct, but not whole HA, protected mice against cross-subtype challenge (H1 vs. H2) (40). Years later, with renewed interest in the HA stem, studies focused on characterizing broadly reactive mAbs (41), and many have now been identified (42). These recognize HAs of group 1 (43), group 2 (44), occasionally both groups 1 and 2 (45, 46), influenza B of both lineages (47), or even both influenza A and B (47).

Figure 2.

Phylogenetic tree of hemagglutinin sequences, showing different influenza A virus subtypes in groups 1 and 2, and influenza B lineages, based on amino acid sequences. Reproduced from Figure 1 of Jang et al. (156), available by open access and subject to Creative Commons license https://creativecommons.org/licenses/by-nc-sa/3.0/.

Although rare antibody clones may be useful for passive antibody therapy, vaccines must generate active immunity with protective titers—a more demanding problem. For active immunization to stem epitopes, various constructs have been developed, including engineered stem fragments (48–50), peptides (51), recombinant expression vectors (52), HA-ferritin nanoparticles (53), DNA followed by virus (54), and whole HA, including chimeric constructs (55, 56). There has been some progress with in vivo protection. Some H1-based constructs protected against H5N1, which is highly pathogenic but in the same group 1 (49, 53, 56). Candidates based on H3 or H7 (group 2) protected mice well against matched virus, but only weakly against a different group 2 subtype; they did not protect ferrets (50). Sequential immunization with HA chimeric constructs partially controlled group 1 virus in ferrets (57). An HA-stem trimer construct protected mice partially against viruses in groups 1 and 2 (58). Unless an immunogen efficiently induces antibody to sites shared between groups 1 and 2, a mixture of constructs may be better for optimum protection.

Neuraminidase

Antibodies to neuraminidase can protect against influenza virus infection (59), especially those that inhibit enzymatic activity, so neuraminidase contributes to effectiveness of seasonal vaccines (60). Cross-reactivity (61) and protection (60) extend to mismatched strains within a subtype (62). Neuraminidase virus-like particles (63), but not neuraminidase protein (64), gave some heterosubtypic protection. Discouragingly, with an mAb recognizing all neuraminidase subtypes, an extremely high concentration gave partial protection (65). There are T-cell responses to neuraminidase in animals (66, 67) and humans (27, 68, 69), but their role in protection is unknown. With growing interest in neuraminidase protection, a consortium has proposed extensive research (70).

Polymerases

PB1, PB2, and acidic polymerase are T-cell target antigens in mice (71) and humans (27, 69). PB1 has been investigated as a vaccine antigen (37, 72), but the polymerases are not a major focus.

Combinations of antigens

Since the 1990s, various combinations of antigens have been studied for cross-protection, including vectored NP + M1 (73, 74), HA + NP peptides (75), headless HA + M2 (76), HA + NP + M1 DNA vaccines (77), HA + NP + M2 DNA vaccines (78), and combinations of 5 or 6 antigens (79, 80). The combination NP + M2 provided more powerful protection than either NP or M2 alone and was enhanced by intranasal administration (81, 82). Protection was seen against viruses mismatched for M2 sequence (83), and against H1N1 and H3N2 challenge, preventing morbidity and death 10 months after a single intranasal dose (82). NP + M2 constructs protected against highly pathogenic H5N1 in mice and ferrets (81, 82). Constructs presenting consensus sequences of known epitopes protected broadly against human, avian, and swine challenge viruses, including H9N2 (84). Other proposals have included combining seasonal vaccines with M2 (85) or NP + M1 (86). Combinations often provide protection superior to individual antigens (76, 82, 87), but interference should be considered (88).

Live attenuated influenza vaccines have been considered as universal vaccines. They induce cell-mediated immunity to conserved sites and, in animals, provide some protection against mismatched challenge viruses (83, 89–91). In human studies, T-cell cross-reactivity increases in recipients of live attenuated vaccines (92, 93), with NP a prominent target (93). Clinically, in a drift year, live attenuated vaccine protected against a viral variant not present in the vaccine (94). New types of live attenuated vaccines have been proposed (95, 96). However, for live attenuated vaccines, compared with other conserved vaccine candidates, the dominant antigen of the HA head group may reduce focus of the response on conserved components.

ACTIVE IMMUNIZATION: DURABILITY OF PROTECTION AND RESPONDER AGE

Duration of protection is important for a universal vaccine to cover an outbreak or pandemic wave and ideally would permit less frequent vaccination. For HA stem vaccine candidates, protection of mice or ferrets against challenge was demonstrated 3–4 weeks after a boost (52, 58, 76). For vectored NP + M2 vaccination, a single intranasal dose protected mice for at least 10 months (82), and memory from priming can persist for 16 months (97).

An additional concern is protection of the young and the elderly. HA stem vaccination protected mice 6–8 weeks old (52, 58, 76). Vectored NP + M2 vaccination protected mice 2–20 months old (97). Recent interest in imprinting highlights the importance of birth year in susceptibility to influenza (98). Because imprinting appears related to HA group/subtype (not conserved sites in NP/M or HA stem), it likely would not restrict effectiveness of universal vaccines in different age groups.

STERILIZING IMMUNITY VERSUS INFECTION-PERMISSIVE IMMUNITY

If vaccination induces high-titered neutralizing antibodies to the HA globular head, viral entry is blocked. If immunity is instead mediated by other mechanisms, such as T cells or nonneutralizing antibodies, then transient infection is permitted, but rapid clearance avoids severe disease. Some investigators have considered infection-permissive immunity inadequate, stating, for example, that pandemic spread in the presence of cross-reactive T-cell immunity demonstrates that protection is by antibody (99). The reality is more complex. Individuals differ greatly in their extent of preexisting T-cell immunity, and during a pandemic not all individuals become ill. Assessing whether preexisting T-cell immunity controls infection requires surveillance studies linking individual preexposure T-cell responses with individual monitoring for infection and severity. A pandemic might have been worse without this prior immunity.

Other investigators have suggested possible advantages of mild infection and debated this as an argument against annual vaccination (100, 101). Cellular immunity to conserved antigens is induced in a host who experiences infection, even asymptomatic or mild infection, and this immunity can control divergent virus encountered later (102). If a vaccine prevents symptomatic disease but not all viral entry or replication, that may be sufficient to protect the recipient while permitting development of additional immune responses (Figure 3).

Figure 3.

Comparison of sterilizing immunity with infection-permissive immunity. The diagram depicts the differing consequences of types of vaccination. Upper box: Classical vaccine induces antibodies that neutralize virus, preventing infection by a matched virus strain. Lower box: Nucleoprotein plus matrix (NP + M) universal vaccine induces nonneutralizing immune responses that partially control infection and reduce the severity of disease, but allow some degree of viral replication. The consequences can differ in a subsequent year when a new viral strain is encountered.

Many universal vaccine candidates do not induce sterilizing immunity. Antibodies to NP are nonneutralizing, but they passively transfer protection in animals (103), which involves Fc receptors (104). Antibodies to M2 also passively transfer protection (32), which can depend on Fc-receptor mechanisms (105) and natural killer cells (106).

Similarly, antibodies to HA can protect by multiple mechanisms and do not always confer sterilizing immunity. Human sera have antibody-dependent cellular cytotoxicity activity reactive with influenza viruses (107–109). This activity was associated with protection in some reports (110) but not in others (111). Some antibodies to the HA stem neutralize in vitro, but Fc receptor–dependent mechanisms are required in vivo (112, 113). Antibody to the HA stem that does not neutralize in vitro can protect animals (114). Enzyme-linked immunosorbent assay binding has been suggested as a correlate for protection, based on passive transfer of human antibodies to mice (115). Many questions remain regarding mechanisms of protection by stem-reactive antibodies. In any case, although HA stem immunity includes some neutralizing antibodies, viral titer results show that available candidate vaccines are operationally infection permissive, including one protective against both group 1 and 2 challenge (58).

The fact that NP and M vaccines do not induce sterilizing immunity was often viewed as a drawback compared with HA-based vaccines. Ironically, not only may that property sometimes be an advantage rather than a drawback, but it also turns out to be a property of many HA stem vaccines.

SELECTIVE PRESSURE AND EPIDEMIOLOGY OF ESCAPE MUTANTS

For NP, naturally occurring infection induces strong antibody and T-cell responses. The relative conservation of NP sequence may be due to structural constraints and to the fact that anti-NP immunity is infection permissive. However, there is selective pressure and escape mutants are detected (8, 116–118). Sequences of viruses from a human outbreak revealed emergence of new NP mutations in cytotoxic T-lymphocyte epitopes, consistent with evasion of T-cell immunity that otherwise helped contain the virus (116). These mutations spread through the population (116, 119). Using a cytotoxic T-lymphocyte epitope as a probe, variants were shown to emerge in an order suggesting immune selection (117). Mutations in dominant T-cell epitopes also emerged in infected mice but reverted in the absence of selective pressure, suggesting a fitness cost (118).

Mutation in an epitope does not necessarily mean lack of immune recognition for the whole protein and overall escape. The 2009 pandemic virus had a different NP_418–426 sequence than previously circulating viruses, and that peptide was not recognized by prepandemic human T cells (120). However, a pool of pandemic sequence NP peptides restimulated T-cell responses in prepandemic samples from 97% of normal human donors (69), presumably due to recognition of other epitopes.

For M2, the limited naturally occurring sequence variation could be due to structural constraints, but in addition, infection induces only weak immune responses to M2 and thus little selective pressure. Strong selective pressure was tested to see if it would reveal greater variation. Escape mutants emerged in infected mice with severe combined immunodeficiency that were given anti-M2 mAbs. However, only 2 different mutations emerged, suggesting most positions were constrained (121).

Infection generates some HA stem–specific antibodies (41, 122), but selective pressure on the stem likely has been modest in the past. Some investigators hoped the HA stem could not mutate because of structural constraints (8, 123), so an immune response to it would not select escape mutants and an HA stem vaccine would remain effective indefinitely. However, escape mutants were isolated in the original study of broadly reactive mAb C179 (39). Although some later attempts to isolate stem escape variants were unsuccessful (55), others have demonstrated in vitro selection and propagation of viral variants resistant to neutralization by antistem antibodies (123), including broadly neutralizing antibody (114). Multiple mechanisms of escape from neutralization by stem antibodies have been demonstrated (124).

Another hope was that HA stem mutations would carry a high viral fitness cost, reducing replication and ability to cause disease even if detectable in vitro. However, mutations at a site recognized by a broadly neutralizing antibody have been found in H7 viruses from recent human outbreaks (125). In another study (126), analysis of HA sequences of circulating H1N1 viruses since 1918 suggested that immune pressure contributed to the emergence of HA stem variants. The investigators then selected HA stem escape mutants by virus passage in vitro in the presence of polyclonal or monoclonal antibodies. Some stem mutants selected that were less sensitive to neutralization by the antibodies still replicated well in vivo and still caused morbidity and death in mice.

EPITOPE-BASED VACCINES

Escape mutation is a drawback of vaccines based on minimal epitopes and peptides rather than whole proteins. A peptide vaccine providing an immunodominant epitope may not protect against a virus mutated at that site. Another drawback of the approach is that multiple peptides would likely be required to provide epitopes for coverage of diverse human responders. In contrast, use of entire proteins as antigens can enhance population coverage by providing different T-cell epitopes restricted by different human leukocyte antigen alleles, even rare ones. Immunity to entire proteins can also provide some protection against a virus mutated in an immunodominant epitope, by providing recall responses to other epitopes, though compensation can be complex (127, 128).

ROUTE OF ADMINISTRATION AND MUCOSAL IMMUNITY

Mucosal immunization in the respiratory tract can provide better protection against influenza virus infection than systemic immunization—up to several million-fold better, as shown in 1950 (10). Universal vaccine candidates have often been given intramuscularly; mucosal administration has been explored subsequently. Intranasal administration of NP and M vaccine candidates proved superior to intramuscular administration (81, 82, 129), with primed immune effectors located at the site of infection (81, 82). NP and M candidates given intranasally have sometimes been given with adjuvant (130). HA stem vaccine candidates given intramuscularly can protect against infection (53, 58), but the intranasal route was superior for a chimeric HA (52).

TRANSMISSION

Strain-matched vaccines inducing neutralizing antibody prevent infection and thus transmission. Universal vaccines based on other mechanisms of protection permit some infection and thus might allow transmission. Some investigators have worried that use of an infection-permissive vaccine would allow continued spread of virus in the population. However, mathematical modeling provided a more optimistic estimate. Based on data from work in ferrets indicating an infection-permissive vaccine reduced virus shedding (81), a mathematical model projected that reduced transmission could limit the size of outbreaks or pandemics and slow virus drift (131). At the time of that work, linkage of reduced virus shedding to reduced transmission was just an assumption. It has now been demonstrated. In a proof-of-concept mouse transmission model, for mice vaccinated with NP + M2 and then infected, transmission to naïve contacts was reduced (132).

EPIDEMIOLOGY OF HUMAN PROTECTIVE IMMUNITY AND CONTEMPORARY HUMAN STUDIES

Evidence has accumulated, some suggestive and some more definite, that immunity due to previous influenza exposures can protect humans against a novel strain despite absence of neutralizing antibodies to it (133–135). Because of prior infections, humans have readily detectable immune memory responses to conserved influenza virus antigens, including T-cell responses to NP, M1, PB1, and other antigens (26, 27, 136, 137); and antibodies to M2 (69, 138), NP (69), and the HA stem (61, 139). The classic McMichael et al. (140) study pointed to a role of cytotoxic T-lymphocyte immunity in protection, with a reduction in influenza challenge virus shedding but not symptoms. Additional evidence for protection by T cells has been collected recently from surveillance during the 2009 pandemic (68, 141) and from challenge studies (142).

Human challenge studies with live influenza virus were standard practice years ago, continued in some parts of the world, and recently have been undertaken more broadly. Carefully characterized challenge stocks are used in volunteers to study parameters, including transmission, antibody and T-cell responses, cytokines, and correlations between immune parameters and clinical outcomes. In a recent study of antibodies, high preexisting anti-HA stem titers were associated with reduced virus shedding but not reduced symptoms. Neuraminidase-inhibiting antibody titer correlated with reduced symptom severity (139).

Various candidate universal vaccines, including epitope-based vaccines (143–145) and viral vectors expressing NP + M1 (146–150), have been tested in clinical trials. Table 1 summarizes some results. In 1 study, MVA−NP + M1 vaccination induced T-cell responses, and vaccinees had reduced symptoms and duration of viral shedding compared with controls (146). Safety and efficacy of MVA−NP + M1 given with inactivated vaccine are being tested in a phase IIb study (86). A phase III study is in progress of a protein containing 9 conserved epitopes (HA, NP, M1; influenza A and B) (151). Clinical experience with HA stem vaccines is beginning. A DNA vaccine expressing whole HA induced antibody to the HA stem in humans (152), and a trial of live attenuated and inactivated viruses with chimeric HAs is beginning (153). Regarding immune correlates, it was reported in 1 trial that cellular immunity (detected by interferon-γ enzyme-linked immunospot assay) correlated with reduced viral shedding and symptoms (144).

Table 1.

Examples^a of Universal Vaccine Candidates in Clinical Trials and Studies

Vaccine/Route	Population and Age, years	Immune Responses, and Protection Outcomes, if Tested	References
Biondvaxb M-001, recombinant protein containing 9 B- and T-cell conserved epitopes of HA, NP, and M1, of influenza A and B, or placebo, intramuscular, followed by TIV	Adults 18–45	M-001 alone induced CD4 and CD8 responses to influenza antigens. M-001 followed by TIV primed for enhanced seroconversion (HI), even to strains not in the TIV.	15
Biondvax intramuscular followed by suboptimal dose of H5N1 investigational vaccine	Adults 18–60 (phase IIb)	Antibody and cellular response endpoints. Completed according to ClinicalTrials.gov, but results not yet public.	143
Biondvax intramuscular (A Pivotal Trial to Assess the Safety and Clinical Efficacy of the M-001 as a Standalone Universal Flu Vaccine)	Adults >50 (phase III)	PCR-positive influenza (% confirmed cases, vaccine vs. placebo), frequency of IFN-γ restimulated CD4 T cells. Estimated data collection completion according to ClinicalTrials.gov: May 2020.	151
Mixture of M1, NPA, NPB, and M2, peptides (20–32-mers) vs. placebo, with ISA-51 adjuvant, subcutaneous, followed by challenge	Men 18–45	Cellular responses; postchallenge viral titers and symptoms. IFN-γ response to stimulation with vaccine correlated with reduced viral shedding and symptoms after challenge.	144
6 modified 35-mer peptides of NP, M, PB1, and PB2, intramuscular, vs. placebo	Adults 22–55	Increases in CD4 and CD8 IFN-γ responses, with recognition of multiple viral strains	145
Mixture of H5, NP, and M2 DNA given twice intramuscular, compared with H5 DNA alone	Adults 18–45	Antibody responses (HI, MN, and ELISA antibody to M2e), NP, M2, and H5 T-cell IFN-γ responses. Subset of subjects responded, more often T-cell response to NP or H5 T cells than the other responses.	157
H5 DNA vaccine intramuscular followed by monovalent inactivated H5N1 (not a universal vaccine, study included here only because of testing of responses to HA stem)	Adults 18–60	Antibody to H5 by multiple methods, CD4 intracellular cytokine responses to H5. Antibody to HA stem tested for 1 subject per group; some activity detected.	152
LAIV intranasal, and trivalent inactivated vaccine intramuscular, in prime-boost sequences	Healthy children	HI antibodies induced in all groups. CD4+, CD8+, and γδ T-cell IFN-γ responses to live virus or M1/M2 and NP peptide pools increased in groups receiving LAIV, not groups with TIV only. Cellular responses were not correlated with HI.	158
Chimeric cH8/1N1 LAIV or chimeric cH8/1N1 IIV or placebo, followed 3 months later by chimeric cH5/1N1 IIV. AS03 adjuvant in some groups.	Adults 18–39	Will measure immune responses. Active according to ClinicalTrials.gov, but not yet recruiting.	153
MVA−NP + M1, intramuscular, followed by influenza challenge vs. challenge in unvaccinated controls	Adults 18–45	Vaccination induced T-cell IFN-γ responses to NP + M1. Compared with control subjects, vaccinees had reduced symptoms and duration of viral shedding.	146
MVA-NP + M1, intramuscular	Adults >50	CD4 and CD8 T-cell IFN-γ responses to NP + M1 increase with vaccination. Magnitude and duration depend on responder age.	147
ChAdOx1 NP + M1 intramuscular ± MVA−NP + M1 intramuscular boost	Adults 18–50	Cellular IFN-γ responses to NP + M1 peptide pools increase after single ChAdOx1 vaccination.	148
MVA−NP + M1, intradermal or intramuscular	Adults 18–50	Cellular IFN-γ responses to NP + M1 peptide pools increase after vaccination. Majority are CD8 T cells, fewer are CD4. IFN-γ, IL-2, TNF-α, and CD107a multifunctional cells also increase.	149
MVA−NP + M1 plus TIV vs. TIV plus placebo, all intramuscular	Adults >50	T-cell IFN-γ responses to NP + M1 peptide pools were greater for MVA-NP + M1 plus TIV than for TIV alone. ELISA antibody to TIV antigens was slightly, though not significantly, higher for the group given MVA-NP + M1 plus TIV (adjuvant-like effect) than TIV alone.	150
MVA−NP + M1 plus TIV vs. TIV plus placebo, all intramuscular	Adults ≥65 (phase IIb)	Plan: Surveillance of 2030 subjects for influenza symptoms, severity, and duration of infection. Immunology substudy (n = 50 per group): antibody and T-cell responses. Currently in progress, results not available.	86

Abbreviations: CD4, cluster of differentiation 4; CD8, cluster of differentiation 8; ELISA, enzyme-linked immunosorbent assay; HA, hemagglutinin; HI, hemagglutination inhibition; IFN, interferon; i.m., intramuscular; LAIV, live attenuated influenza vaccine; M1, matrix 1; M2, matrix 2; MN, microneutralization; MVA, modified vaccinia Ankara; NP, nucleoprotein; PCR, polymerase chain reaction; TIV, trivalent inactivated vaccine.

^a For a more complete list, see World Health Organization tables on clinical evaluation of influenza vaccines (159). The webpage includes vaccines other than universal candidates, but a linked table (160) includes broadly protective strategies.

^b BiondVax Pharmaceuticals Ltd., Ness Ziona, Israel.

THE PATH FORWARD

Vaccine protection need not be perfect, and achieving a proposed goal of 75% or greater protection against symptomatic disease (154) would be a desirable outcome. Even short of that, a major reduction in severity would be valuable, as would a reduction in transmission. Enthusiasm for HA stem vaccines is prompted by discovery of potent and broadly neutralizing HA stem antibodies. NP + M vaccines are further along in demonstrating various parameters of effectiveness in animals (i.e., breadth, duration, range of recipients tested, reduction in transmission) and immunogenicity in humans. Vaccines based on all these antigens should be pursued, with further characterization for potency, breadth and duration of protection in animals, immune responses, and correlates of protection. We need to understand how each vaccine works, because the immune correlates of protection will not be the same for all components in a combination.

Different universal vaccine candidates can then be combined. NP and M1 are usually expressed from vectors to achieve endogenous expression optimal for inducing T cell immunity, while M2, the HA stem, and neuraminidase can be given as protein or vectored. Some formats (expression vectors, virus-like particles) may be amenable to all these antigens. Ultimately, a combination of NP, M1 and/or M2, HA stem and neuraminidase may be advantageous, but only after optimizing each component and checking for interference in the mixture.

There are many vaccine candidates and limited resources for clinical trials; thus, prescreening may be useful. The scientific community could share materials and conduct simultaneous testing (“tournaments”) comparing routes of administration and vaccine components for potency, breadth and duration of protection in animals, as well as immune response markers. A strategic plan of the US National Institute of Allergy and Infectious Diseases for universal influenza vaccine work calls for a consortium (155), which could perhaps coordinate such comparisons.

We have learned a great deal since 1918, but the 2018 influenza season provides a harsh reminder of the serious public health problem we still face with influenza. Research on universal vaccines has made encouraging progress, and these vaccines could contribute to reducing the future toll.

Abbreviations

HA

hemagglutinin

M1

matrix 1

M2

matrix 2

mAb

monoclonal antibody

NP

nucleoprotein