In this study, we compared the plasma cell repertoires present at 7 days following TIV and EI. Previous studies have shown that influenza-specific plasma cells appear transiently in the blood with a peak at 7 days after immunization following which they rapidly decline 
. Studies of infection, however, have been limited to natural infection where timing must be estimated based on exposure and symptoms 
, or where the infection occurred months to years before antibodies were isolated 
. This study of vaccination and experimental infection allowed us to compare the two stimuli at exactly the same point after challenge. We found a similar degree of plasmacytosis in the two conditions but found that the frequency of HA-specific plasma cells and the degree of clonality of those cells was lower in the EI group. These findings differ from studies of natural infection where both a higher frequency and clonality of influenza-specific antibodies were found 
. Trafficking of B cells after infection may have a different kinetic pattern than that following vaccination. This EI study stopped collecting cell samples at 7 days post-infection and so future studies of later time points will be required to determine if the polyclonality of the plasma cell response we observed is specific for the period early after infection, specific for H3N2 A/Wisconsin/67/2005 infection, or if infection with novel pandemic strains consistently stimulates B cell clonal expansion.
For each subject there was an increase in rHA binding and neutralization titers 3–4 weeks after antigen stimulation. The magnitude of rise was greater in the TIV cohort although all EI subjects showed a rise in titer against the infecting strain that was similar to that observed in outbreak situations 
. The presence of asymptomatic but productive infection is common and has been observed for seasonal influenza 
, pandemic H1N1 
, and human infection with avian H5N1 
. Thus, we found it interesting that similar frequencies of rHA-specific antibodies were recovered from subjects with the highest and lowest symptom scores (Fig. S3
online), and that the most broadly cross-reactive rmAbs were recovered from a subject with mild symptoms and a modest rise in binding and neutralization titer. Whether this response contributed to the mild symptoms experienced by this subject is unknown and future studies will have to address the potential therapeutic role for this group of rmAbs.
The plasma-cell-derived mAbs from TIV subjects were more frequently rHA-specific and showed evidence of clonal expansion while mAbs derived from EI subjects were less frequently rHA-specific and had less evidence of clonal expansion. In all subjects, anti-HA responses were primarily strain-specific, but mAbs derived from EI subjects were more frequently cross-reactive for multiple influenza strains or consistent with OAS compared with mAbs derived from TIV subjects. In the EI cohort, responses to influenza proteins induced by infection that were not detected in our assays may have been present. Although split-virus vaccines like that we used for screening contains antigens other than HA 
, the presence of antibodies reactive with antigens not tested by our assays cannot be excluded. Isolation of rmAbs from both EI and TIV subjects reactive only with split-virus vaccine preparations (Fig. S8
online) suggests that some recovered rmAbs were reactive with antigens other than HA.
Vaccination is the primary means to prevent seasonal 
and pandemic 
influenza infection, however, antibody responses to TIV are generally not broadly-neutralizing but rather strain-specific and directed at highly variable domains on HA 
. Such responses are not absolute—one rmAb isolated from TIV01 was recently shown to have cross-reactivity related to its ability to recognize the sialic acid receptor-binding pocket of HA 
. Vaccines that bypass regions of diversity by targeting other influenza proteins [e.g.
, the virus-associated proton-channel M2 
] have not been successful in clinical trials 
. Targeting conserved regions on HA could provide a route to a “universal” influenza vaccine 
, but seasonal vaccines have not consistently induced these specificities of antibodies as shown in our study. Influenza vaccination strategies using novel adjuvants may induce a greater antibody breadth against HA 
and it will be of interest to compare mAbs derived from newer vaccines with those currently in use.
By producing inferred intermediate antibodies of one clonal lineage, we found evidence of induction and modulation of antibody breadth stimulated by TIV (clonal lineage 641; ). This evidence of affinity maturation in humans is similar to findings in mice 
that were used as evidence of V(D)J mutation as the source of affinity maturation 
. Clonal lineage 641 was remarkable in that some inferred intermediate rmAbs showed a greater breadth of rHA binding than the recovered rmAbs, suggesting that continued affinity maturation might have eliminated rHA cross-reactivity. Our novel observation of a branch displaying cross-reactivity in an otherwise strain-specific clonal lineage implies that making such cross-reactive branches immunodominant through novel vaccine strategies could lead to improved cross-strain protection.
In contrast to the findings in TIV subjects, we found that EI subjects showed less evidence of clonal expansion seven days after infection despite having plasma cell antibodies with a higher frequency of VDJ mutations. Our observation of polyclonal activation after EI is similar to that seen in animals after γ-herpesvirus infection 
. This suggests that in our EI subjects either B cell clonal expansion was not required for circulation of B cells at seven days after infection expressing affinity matured cross-reactive antibodies, and/or that the circulation dynamics of HA-specific plasma cells seven days after infection was different than that observed after vaccination.
The human mAb response to influenza infection has been studied in survivors of H5N1 avian influenza 
and in isolation of mAbs from survivors of the 1918 H1N1 pandemic 
; in these studies some mAbs were cross-reactive with related strains 
and some mAbs also displayed cross-protection for novel influenza strains 
. While broadly cross-reactive neutralizing antibodies against influenza can be induced with TIV, these responses are not sufficiently immunodominant to result in high-titer neutralizing antibodies that provide protection against infection with highly divergent strains. Our data demonstrate this phenomenon directly, in that cross-reactive rmAbs and inferred intermediates were found in the plasma cell repertoires of both TIV and EI subjects while such cross-reactive antibodies were less prominent or not detected in day 21 or day 28 plasma samples.
Work in mice suggested that some anti-influenza responses can be associated with restriction of Ig gene usage 
, and recent work has suggested that human antibodies with broad cross-reactivity may preferentially derive from restricted Ig gene pools 
. Phage-displayed antibody libraries screened for binding to H5 HA were enriched for VH
1-69 usage 
and broadly reactive anti-HA stem mAbs using VH
1-69 and VH
3-21 have been produced from IgM+
memory B cells 
. Corti et al. recently reported a series of TIV-induced cross-reactive antibodies using primarily VH
1-69 demonstrating that cells expressing those antibodies can be isolated following seasonal vaccination 
; a strict requirement for VH
1-69 was not shown as there were additional antibodies using VH
3-53 and VH
4-39. It was interesting that none of our clonal lineages used VH
1-69 and only three cross-reactive antibodies in this study used that VH
gene. This could be due to a lack of VH
1-69 gene usage in our subject population as the frequency of VH
1-69 expressing B cells is related to copy number variation 
, and whether this is the case for our current subject groups is not known. In our study, other Ig genes were found to make cross-reactive antibodies including VH
3-74, and VH
1-46. One group of these antibodies, clonal lineage 2569, was remarkable both in the degree of mutation and that its members were IgM and IgA1. These findings are consistent with the recall of an IgM+
memory B cell lineage expressing a broadly cross-reactive antibody response that appeared transiently following infection but that did not predominate in plasma four weeks after infection. A similar phenomenon has been observed following TIV with isolation of broadly reactive subdominant antibodies that did not predominate in convalescent plasma 
. In our EI subject, stimulation of a subdominant response may have contributed to the mild symptoms experienced by this subject and could reflect one mechanism by which antibody responses contribute to control of a virus infection without resulting in long-lasting antibody titers. These findings are also consistent with Pape et al. who reported that class-switched memory B cells have different circulation kinetics compared to IgM memory B cells, with the latter persisting for longer periods but also being subject to suppression in the presence of cross-reactive plasma antibodies 
Finally, OAS antibodies were more frequent among mAbs derived from EI subjects compared with TIV subjects. Prior studies have shown that OAS occurs following both influenza vaccination 
and infection 
, although work by Wrammert et al. suggested that OAS in humans following influenza vaccination was uncommon; our study of TIV subjects is consistent with Wrammert et al. 
. That the majority of OAS mAbs were reactive with HA from the H3 A/Johannesburg/33/1994 strain is consistent with both prior exposure to that strain and recall by immunization with H3N2-containing vaccine or by infection with H3N2 influenza virus.
Regardless, the presence of both OAS and cross-reactivity in mAbs from EI subjects suggests that both kinds of antibodies were stimulated in parallel during infection. Whether these two processes can be decoupled to only stimulate broadly cross-reactive antibodies by a vaccine remains unknown. Corti et al. showed that the H5 anti-HA antibody response following TIV was detectable but generally weak, both for serum antibodies and memory B cell frequency 
. Recent reports have described vaccine designs that may stimulate more broadly reactive anti-HA antibodies. Khurana et al. showed increased anti-HA antibody breadth using an adjuvanted vaccine 
. Wei et al. 
and Wang et al. 
have demonstrated induction of broad of plasma antibody responses by sequential immunization with heterologous HAs, perhaps replicating the cross-protective response that can occur after multiple seasonal influenza infections in ferrets 
. The fact that both seasonal influenza vaccination and infection involve exposure to heterologous HAs on an annual basis, and that these exposures do not lead to high levels of influenza resistance in the general population, suggests that sequential exposure to heterologous HAs by itself is insufficient. Targeting of the conserved HA stalk to induce antibodies in mice reactive across influenza subtypes has recently been demonstrated through the use of synthetic peptides 
and rHA subunits 
. It remains to be seen whether such vaccine strategies will work in humans.
In summary, we have shown that while the anti-HA response in vaccination and H3N2 A/Wisconsin/67/2005 infection differs in the degree of clonal expansion present at seven days, broadly reactive antibodies are induced in both settings. One strategy for vaccine design to expand B cell clonal lineages of broadly cross-reactive antibodies may be to create novel HA molecules with enhanced binding to germline B cell receptors of HA-responsive naïve B cells 
. Thus, the study of clonal lineages of antibodies with HA binding and breadth of influenza neutralization in both vaccination and infection could provide guidance for design of influenza vaccines capable of inducing immunodominant broadly cross-reactive antibody responses.