The present studies were prompted by the very poor immunogenicity in humans of an inactivated monovalent subunit influenza A (H7N7) vaccine despite SRID values indicating the HA dosages given were those that should induce good serum antibody responses 
. Similar antigen dosages have regularly induced good antibody responses with seasonal vaccines 
. The A/H7N7 vaccine was developed as a potential vaccine for one of the avian influenza A viruses thought to be a possible cause of an influenza pandemic. Monovalent inactivated avian influenza A virus vaccines have been prepared by various manufacturers for influenza A subtypes considered likely causes of a pandemic and tested for safety and immunogenicity in humans. These studies have focused on influenza A (H5), (H7) and (H9) 
. The NA in these vaccines has usually been N1 or N2 but the NA in the H7 vaccine we tested was N7. In general, those avian virus vaccines without adjuvant have exhibited poor immunogenicity in humans (see ) despite the fact that all were prepared by manufacturers using their standard methods for preparing seasonal influenza vaccines. All were prepared from virus grown in embryonated chicken eggs but a study not in the table with an influenza A (H7N1) vaccine made in a tissue culture system also exhibited low immunogenicity 
. As indicated earlier, this low immunogenicity was not the experience in 1957, 1968, 1976, 1977, and 2009 with monovalent inactivated influenza vaccines for new subtypes or major new variants of influenza A viruses that emerged and spread worldwide 
One solution for correcting the low immunogenicity of these vaccines has been to give them with an oil-in-water adjuvant. Alum as an adjuvant for the avian virus vaccines has not reliably improved antibody responses but oil-in-water adjuvants have, so far, shown consistent improvement in responses 
. While the adjuvant remedy for the low immunogenicity of avian influenza A vaccines prepared using seasonal vaccine methods is available, understanding the reason for the low immunogenicity of these nonadjuvanted vaccines is desirable. For this reason, we undertook a series of studies of the vaccines to seek some understanding.
The first steps leading to an immune response are uptake, processing and presentation of antigenic determinants to T cells. We evaluated this sequence using established in vitro methods. We used available monovalent vaccines of H5, H7, and H9 that had been used in clinical trials and we also included a recent 2009 pandemic H1N1 vaccine which was highly antigenic in clinical trials in primed adults as a single dose and in unprimed children in standard two dose schedules 
. Since the content of vaccines can include proteins other than the HA, we included purified recombinant HA proteins expressed in a baculovirus system in the studies. In these studies, avian influenza vaccine and HA antigen uptake, processing and presentation to human T cells for initiating an immune response appeared normal.
In view of the immunogenicity reported and the enhancement with some adjuvants, it seemed unlikely that there was a general defect in the conformation of the avian HA in the vaccines although variation and a defect in the HA in the H7 vaccine seemed possible. To evaluate the conformation of the HA antigens, we tested the vaccine HA protein interactions with antisera in ELISA assays and in gel electrophoresis and western blots. Interactions with HA proteins in ELISA assays were evaluated for polyclonal and monoclonal antisera against native and denatured (unfoldon) proteins for each vaccine and for a matching virus. No differences in the ratio of native/unfoldons between the virus and the vaccine for each vaccine and virus HA, including the H7 vaccine, were detected; however, reagents for determining the actual ratios were not available.
HA trimers were detected in the H5, H9 and H1 but not in the H7 western blots; however, trimers in the H7 vaccine could not be excluded as the antiserum showed considerable background and the Coomassie stained gel pattern was not interpretable. Although not a definitive evaluation, no abnormalities in the state of the HA in the various inactivated avian influenza virus vaccines, including the H7 vaccine, were identified.
In order for vaccine virus to hemagglutinate red blood cells (RBC), the HA must exhibit the biological activity of binding to the receptor on RBCs and be in a morphologic configuration that can bridge to other RBCs to induce hemagglutination. The requirements for this are intact conformational HA that can bind to receptor and a morphologic structure with a number of HA units such as a virus particle that can bridge between RBCs and lead to hemagglutination. Before SRID was adopted as the standard for HA quantitation in vaccines, vaccine antigen quantitation was done using hemagglutination and was expressed as HA units or chick cell agglutinating units (CCA) 
. In hemagglutination titer comparisons (), turkey RBCs were most sensitive, chicken RBCs exhibited the same pattern as turkey cells but with lower titers and horse red cells were lowest. Horse RBCs are reported to exhibit higher titers of anti-HA antibody for the avian viruses than either avian RBC despite requiring more virus per HA unit 
. Notable in these comparisons was the poor correlation between the HA quantity in SRID assays and the HA titer with RBCs and the complete absence of hemagglutination for the H7 vaccine despite a SRID concentration of 60 µg/ml. The H7 vaccine with no HA titer was the poorest immunogen in humans and the HK/G1/99 (H9N2) vaccine with very low titers was next poorest. The highest HA titer was exhibited by the CK/G9/97 (H9N2) vaccine and it appeared to be the best immunogen among the avian vaccines.
The evaluations of morphology of the various subunit vaccines suggested an association with immunogenicity of the vaccines in humans. The best immunogenicity in humans was exhibited by the vaccines that contained residual virion particles, particle-like structures, or pieces of viral particles of varying sizes, some of which are large pieces clearly containing surface structures corresponding to the HA and NA; these vaccines were the two p2009 (H1N1) vaccines and the CK/G9/97 (H9N2) vaccine. This finding is similar to that of the many inactivated trivalent seasonal vaccines we have viewed and reported by others 
. The H7 vaccine morphology was primarily small structural units in the size range of the HA and NA glycoproteins. Perhaps these include individual HA units as they would not be able to induce hemagglutination and yet could be detected in antibody binding assays like SRID and ELISA. However, in immunogenicity assays they might be more like peptides as immunogens and require an effective adjuvant for inducing an antibody response. In an immunization study of the H7 vaccine in ferrets with and without an oil-in-water adjuvant vaccine performed after the poor immunogenicity in humans was known, immunogenicity was negligible with vaccine alone but was significantly improved when AS03 adjuvant was used (R. Webby, personal communication).
The classical stellate structures of HA and NA as described by Laver in the 1970s were prominent structures in the A/VN/04 (H5N1) vaccine and the A/CK/G9/99 (H9N2) vaccine. The A/VN/04 (H5N1) vaccine appeared intermediate in immunogenicity (); however, the A/Indo/05 vaccine was of similar immunogenicity without a clear morphologic association. The degree of prevalence of morphologic units with no distinct structure () did not appear to relate to immunogenicity.
The studies reported here have suggested that, despite the variable and sometimes poor immunogenicity in humans, the inactivated avian influenza A virus vaccines all contain conformationally intact HA proteins capable of inducing some HAI antibody. Uptake, processing and presentation to human T cells and the state of the HA proteins, as determined in antibody binding assays and gel analyses, appear normal. These findings are reassuring for the potential of making useful avian influenza A vaccines. All the vaccines evaluated by us in this study are subunit vaccines that resulted from detergent treatment and other proprietary manipulations. It is known that these processes do not always split virus completely. An example of this was seen with one of the monovalent p2009 (H1N1) vaccines evaluated in this study which contained residual intact virus particles. The poorest immunogenicity was exhibited by the H7 vaccine which was predominantly small units that may be HA and NA units. One of the intermediate immunogenicity vaccines (H5N1) contained typical stellates. These findings suggest that the morphology of the vaccines may have influenced immunogenicity of these subunit vaccines in humans.
Vaccine morphology has been known to relate to immune responses to influenza vaccines in animal models and humans for decades. In general, whole virus vaccines have been more immunogenic than subunit vaccines and the smaller the subunit, the less immunogenic is the vaccine in naïve, unprimed hosts, best demonstrated in humans with a single dose 
. At the peptide/epitope level, an adjuvant is generally required to elicit good responses. Use of subunit vaccines became common after the extensive immunogenicity studies of A/New Jersey/76 (H1N1) vaccines in humans in 1976 
. This trend was primarily for the reduced reactogenicity of subunit vaccines as whole virus vaccines were frequently shown to be more immunogenic than subunit vaccines but also more reactogenic. Proponents for superiority of whole virus vaccines continue to report findings in animal models and whole virus influenza vaccines are the products distributed by many companies throughout the world.
An effect of morphology on immunogenicity of influenza vaccines in humans was clearly demonstrated in the publications of Laver and Webster in the 1970s. Laver successfully removed the HA and NA from virus particles and purified the subunits 
. The resulting subunits formed stellates when the hydrophobic ends attached to each other in aqueous solution. Those subunit vaccines, called hanaflu, were reduced in immunogenicity compared to whole virus vaccines in hamsters unless some whole virus was included. Webster and Laver showed that the whole virus could be influenza B even though the hanaflu vaccine evaluated was influenza A 
. This finding could relate to immunogenicity of seasonal subunit trivalent vaccines which commonly contain virus particles as the type and subtype of the particles is unknown for these immunogenic vaccines. When the hanaflu stellate vaccines were tested with and without influenza B whole virus in humans, no enhancement in antibody responses was seen although the subjects were primed adults 
. However, when the hanaflu/whole virus concept was tested in subjects unprimed for A/NJ/76 (H1N1) and A/USSR/77 (H1N1), enhancement was demonstrated when the hanaflu stellate vaccine included some whole virus 
. The intermediate immunogenicity of an H5N1 vaccine that contained stellate structures is compatible with the reduced immunogenicity seen in unprimed subjects with the pure hanaflu “stellate” vaccines by Webster and Laver.
Exceptions to a uniform proposal for whole virus vaccine superiority for immunogenicity in humans are in the varied immunogenicity reports in the clinical trials in 1976 and 1977 and for recent trials with avian influenza A inactivated whole virus vaccines 
. The reported experience with avian monovalent whole virus vaccines without adjuvant is insufficient for conclusions on immunogenicity of whole virus versus split product vaccines without adjuvant. Two different A (H5N1) studies apparently using the same whole virus vaccine with and without alum adjuvant varied in reported immunogenicity 
. The only other identified trial of A (H5N1) whole virus vaccine without adjuvant was with A (H5N1) in healthy adults and 69% achieved HAI titers of ≥1
40 from a dosage of 7.5 µg of HA 
. A trial with an A (H9N2) whole virus vaccine is notable for finding low immunogenicity 
. Most reported avian influenza whole virus vaccine trials were with alum adjuvant; the reported range of percent achieving HAI titers of ≥1
40 is 20–64% for 5–7.5 µg and 45–86% for 15 µg of HA 
. Reported percentages achieving ≥1
40 HAI for 7.5 and 15 µg HA of subunit vaccine with alum adjuvant are 1–43% for 5–7.5 µg HA and 2–44% for 15 µg HA 
. These findings suggest but do not prove a generally greater immunogenicity for whole virus vaccines; however, they are inferior to those reported for subunit vaccine with oil-in-water adjuvant 
Limitations of the current study include the limited number of vaccines and different manufacturers evaluated, the incomplete evaluations of the conformational state of the HAs because of lacking some required reagents, the fact that the SRID concentrations were provided by different manufacturers who did not always prepare and use reagents as originally described by Schild, et al., and lack of a single starting antigen vaccine constructed to exhibit different morphologies and HA titers for correlating with immunogenicity 
. Additionally, although the immunogenicity reports in the literature generally support the importance of vaccine morphology in inactivated avian influenza A vaccines for immunogenicity, there are inconsistencies. The clinical trial data compared were developed in a number of different laboratories and HAI titers are known to exhibit considerable variation between laboratories