For this study, we developed an immunization schedule that elicited broadly-neutralizing antibodies against H3 influenza viruses in vivo
. The finding that such antibody specificities can be elicited by vaccination of mice suggests that with the proper immunogen(s) and vaccination protocol, such a response might also be elicited in humans. Several recent studies describe antibodies isolated from human phage display libraries that have cross-neutralizing activity against Group 1 influenza viruses 
. Mabs isolated from human display libraries have proved extremely useful in the characterization of structural epitopes that mediate heterosubtypic neutralization. Caveats to this methodology exist, however, since the diversity of combinatorial display libraries is typically orders of magnitude greater than the diversity of the true human variable region repertoire 
. Additionally, phage display libraries are generated by random combination of immunoglobulin VH and VL genes and are therefore not restricted, as the in vivo
repertoire is, by mechanisms regulating the production of auto-reactive specificities.
Until now, broadly neutralizing antibodies reactive with H3 viruses have not been described. Interestingly, mAbs 7A7 and 39A4 react by immunofluorescence with the hemagglutinin of multiple subtypes, though neutralizing activity appears to be limited to H3 viruses. Binding by these mAbs to other subtypes may be of relatively low avidity such that they no longer mediate neutralization, or, they may simply bind an epitope of non-H3 hemagglutinins that does not mediate neutralization. The identification of anti-H3 mAbs 12D1 and 39A4 complements recent works describing antibodies F10 and CR6261 that neutralize an array of Group 1 viruses 
. One might envision a passive transfer therapy consisting of multiple broadly neutralizing mAbs for general use against pandemic and seasonal influenza virus strains. With the increasing resistance of influenza virus isolates to available anti-viral drugs, such an antibody cocktail could be of great value in severe disease.
Mouse monoclonal antibodies such as the anti-H3 mabs described herein are commonly used in the development of therapeutic antibodies for use in humans. Once characterized, rodent antibodies are readily humanized by methods typically involving grafting of non-human complimentary determining regions into appropriate human frameworks followed by cloning of variable region segments into complete human immunoglobulin constructs 
The fact that escape mutants were not selected after multiple passages of virus in the presence of anti-H3 mabs 12D1 and 39A4 is intriguing. Sui et al. reported that they were similarly unable to isolate escape mutants using their fusion-inhibiting mAb F10 
. MAb F10 makes multiple interactions with the hydrophobic pocket of the hemagglutinin including with the fusion peptide itself and prevents the low-pH triggered conformation change required for fusion. Considering the rigid structural and electrostatic requirements involved in membrane fusion, the hemagglutinin might not readily accommodate mutations at the F10 binding epitope. Anti-H3 mAb 39A4 binds a conformational epitope of the hemagglutinin trimer; the region of binding may bridge two monomers, therefore interacting with two different portions/faces of each monomer. A mutation at one region of contact (that does not affect trimer formation) may not be sufficient to ablate 39A4 binding. Anti-H3 mAb 12D1 likely binds within the long alpha-helix of HA2. This region may not accommodate changes that would affect 12D1 binding due to required secondary helix structure and specific van der Waals interactions that stabilize the hemagglutinin trimer 
. Generally, mutations in the stalk of the hemagglutinin are more likely to affect the architecture of the entire molecule than are mutations in the classical antigenic sites 
The development of HA2-based vaccine constructs is of significant interest given recent reports of anti-HA2 mAbs with broad neutralizing activity against influenza viruses. Original studies of immunogens consisting of virus particles lacking the HA1 subunit demonstrated that design of an effective construct, however, will likely not be straightforward 
. This is in large part due to the difficulty involved in maintaining the native configuration of the hemagglutinin stalk, which has complex tertiary structure and incorporates a portion of HA1 in addition to the HA2 subunit. Recent reports of mAbs with broad neutralizing activity against influenza viruses that are not active by western blot and that make contacts with amino acids in both HA1 and HA2 underscore the importance of maintaining non-contiguous epitopes in HA2 vaccine contructs 
In contrast to these mAbs, anti-H3 mAb 12D1 does not rely on a structural/non-contiguous epitope of the hemagglutinin stalk for binding. The observation that 12D1 makes dominant contacts within a continuous segment of the HA2 subunit suggests the design of an immunogen, perhaps consisting of that HA2 segment coupled to a carrier protein, that would direct an immune response to the region. The identified region, HA2 76–106, is 100% conserved between the H3 viruses used in this study and all other H3 viruses that we have examined. In contrast, the H1 viruses A/New Caledonia/20/99 and A/PR/8/34 share only 56.7% identity with the equivalent region in the H3 hemagglutinin. A vaccine construct incorporating this region, therefore, would likely not provide protection against H1 influenza viruses. This study and other structural studies 
of the influenza hemagglutinin provide groundwork for the design of novel vaccine constructs aimed at providing broad-spectrum immunity to influenza viruses.