Avian influenza viruses of H7 hemagglutinin (HA) subtype in combination with several neuraminidase (NA) subtypes pose significant pandemic risk.1 In poultry, these viruses are typically low pathogenicity avian influenza (LPAI) viruses causing little disease. However, spontaneous mutations can generate highly pathogenic avian influenza (HPAI) viruses.2 On multiple occasions, LPAI or HPAI H7 viruses have sporadically infected humans, usually causing mild disease.1 In two instances, H7 viruses have infected larger contingents and caused severe disease. In the Netherlands in 2003, at least 89 people were infected with an HPAI H7N7 influenza virus, including one fatality due to severe pneumonia.3,4 In China during 2013–2014, a LPAI H7N9 virus has infected several hundred humans, with many experiencing severe respiratory disease and a case fatality rate of more than>30%.5,6

Serological testing of people for evidence of infection with H7 viruses has proven difficult. Titers measured using hemagglutination inhibition (HI) and microneutralization (MN) assays have often been low or undetectable7–9 in many human cases, even with confirmed H7 virus infections.9–12 HI and MN antibodies have been readily detected in a majority of H7N9 patients, but titers are often lower than those induced by human seasonal viruses.13–15 In pre-clinical and clinical studies, vaccines against H7 viruses have typically elicited low titers or infrequent serological responses.16–20

However, the intrinsic immunogenicity of H7 HA has not been directly compared to that of human influenza viruses, which are known to induce HI antibody titers associated with protection.21–23 We therefore used recombinant HA to compare the immunogenicity of two H7 HAs with those of a seasonal H3N2 HA and a pandemic H1N1pdm09 HA in naïve mice. A/Perth/16/2009(H3N2) and A/California/7/2009(H1N1pdm09), which were components of seasonal influenza vaccines from 2010 to 2012, were chosen as representative contemporary H3 and H1 viruses. The H7 HAs were from A/Netherlands/219/2003 (NL/219) (H7N7), an HPAIV representing the Eurasian lineage of H7,3 and A/New York/107/2003 (NY/107) (H7N2), a LPAIV representing the North American lineage of H7.24

Human serological responses to H7 viruses appear to be weaker than those to seasonal influenza strains.16–20 In particular, even people known to have been infected with H7 subtype influenza viruses rarely develop high titers of antibodies, as measured by HI or MN assays.9,10,34,35 While this may reflect the often-superficial nature of infection with H7 viruses, which have until recently often caused localized conjunctivitis rather than respiratory or systemic disease, this was also seen with people immunized with H7 vaccines,16–18 suggesting that humans may generally respond poorly to H7 subtypes of HA. However, in the absence of studies directly comparing the H7 antibody response to H1N1pdm09 and H3N2 strains, it has not been clear that H7 is exceptional. Furthermore, it is not clear whether any difference in response might be related to intrinsic aspects of the HA, to pre-existing immunity in humans, or to some aspect of the virus or vaccine. To distinguish between these possibilities, we immunized naïve mice with purified recombinant HA from two human viruses, H1N1pdm09 and H3N2, and two H7 viruses and compared the antibody responses.

All four HA were roughly similar in their ability to induce IgG that bound to HA (as measured by ELISA), although the H7s induced slightly lower responses than seasonal HA. When measured by HI, however, both H7 HAs were considerably less immunogenic than either H3 or H1pdm HAs. HI assays measure the ability of antibodies to prevent RBC hemagglutination by influenza viruses, mainly due to high-affinity binding near the receptor-binding site (RBS),36 while ELISA can also detect lower-affinity binding to any part of the HA. Our results here are consistent with clinical findings that HI titers following infection with H7 influenza viruses or immunization with inactivated H7 vaccines tend to be low.9–12,16–20

The microneutralization (MN) assay is another widely used assay for measuring antibody responses to influenza viruses. MN assays detect antibodies that block virus infection and/or replication and thus are relevant for protection against infection. By this assay, the difference between responses to HA from H7 and from human viruses was most obvious. Even when limiting analysis to sera from H7-immunized mice showing clearly detectable HI responses, MN responses were absent or very low (Figure 4A). In contrast, all mice immunized with H3 HA, and almost all those immunized with H1pdm HA, developed strong MN titers. Again, these results are consistent with observations in humans, in which H7 vaccinees had low neutralizing antibody titers despite easily detectable ELISA antibody responses.37

Thus, the antibody response to H7 HAs showed discordant results depending on the assay used. If antibodies are being generated to H7 viruses (as shown by ELISA), why are they poorly detected in HI and MN assays? One possible explanation is that the assays used are not capable of properly detecting H7-specific antibodies. The fact that both HI and MN assays, which measure antibodies in very different ways, show very low titers to H7, while positive-control sera gave clear reactions in each, argues against this possibility. The low MN titers for the H7 viruses raise the possibility that these viruses may have some intrinsic resistance to antibody-mediated neutralization. The fact the HI titers were also lower for H7 viruses makes this somewhat less likely, although it remains possible that some aspect of the receptor-binding site in these H7 viruses makes it resistant to antibody blockade. Other possibilities are that H7 antibodies target non-neutralizing regions of HA or that they bind with such low affinity that they are unable to block infection.

To test whether the antibody response to H7s disproportionately targets the stalk region of HA,37 we used HA1 as the antigen in ELISAs. While titers against HA1 were low compared to those against the full-length HA, NL/219 and the seasonal HAs showed a similar reduction in titer of about 40–50% (Figure 1D). However, as the HA1-specific responses for NL/219 and NY/107 were similar to that for CA/07 (P > 0·05), our data do not support the explanation that the H7 and H1pdm HAs induced different levels of HA1-specific antibodies.

Antibodies produced by B cells normally increase in affinity through the course of an infection, as activated B cells undergo somatic hypermutation of their immunoglobulin variable regions.38 Avidity ELISA showed that NY/107 did indeed induce low-affinity antibodies, as a brief urea wash reduced the bound antibody titer significantly (Figure 4B,C). Antibodies induced by NL/219 were also of relatively low affinity, although only modestly (but significantly) lower than those induced by CA/07. Not only did Perth/16 HA induce the highest titers of antibodies as measured in all assays, the affinity of these antibodies was also high, with titers being unaffected by the urea wash. Thus, the most likely explanation for the difference in HI and MN titers induced by H7 compared with seasonal and pandemic HA is that the former induce predominately low-affinity antibodies.

The molecular mechanism(s) for the difference in immunogenicity is unknown. As in these experiments (unlike in humans), mice had no prior exposure to seasonal HA, previous priming is not a critical factor. In addition, as purified recombinant HA was used as an immunogen, viral, or vaccine factors other than HA would not affect the response. This suggests that intrinsic features of the HA itself may affect immunogenicity, perhaps through differential induction of inflammatory responses.

It is important to note that our results are based on immunization with purified recombinant protein, demonstrating that immunogenicity is intrinsic to the protein itself. Purified proteins are intrinsically less immunogenic than the same protein in standard inactivated split vaccines, which provide adjuvant factors such as viral nucleic acids.39 Similarly, viral infection provides strong inflammatory signals that might be able to overcome low intrinsic immunogenicity. For example, even though immunization of mice with recombinant H7 from NL/219 and NY/107 induced low or no MN titers, positive-control serum from ferrets infected with these viruses had readily detectable MN antibodies (Figure 3D). Similarly, H7N9 human infections, most of which caused severe disease with inflammation,40 often lead to detectable if relatively low serological responses.13–15

In humans infected with avian influenza, weak immunogenicity may also reflect the lack of prior exposure to avian HA, compared with the repeated exposure to seasonal virus HA that may prime for subsequent responses. Similarly, vaccine or virus features other than HA may influence immunogenicity. Nevertheless, our findings suggest that, all other factors being equal, NL219 and NY107 HAs may be less immunogenic than seasonal H3 or H1pdm HA. In particular, the low level of H7-neutralizing antibody responses observed in this study suggests that H7 vaccines may require strong adjuvants or other strategies to induce antibody titers associated with protection in a majority of vaccinees.