Induction of broadly cross-reactive antibody responsesto the influenza HA stem region following H5N1vaccination in humansAli H. Ellebedya,b, Florian Krammerc, Gui-Mei Lia,b, Matthew S. Millerc, Christopher Chiua,b,1, Jens Wrammerta,d,Cathy Y. Changa,b, Carl W. Davisa,b, Megan McCauslanda,b,2, Rivka Elbeine, Srilatha Edupugantif, Paul Spearmand,Sarah F. Andrewsg, Patrick C. Wilsong, Adolfo García-Sastrec,h,i, Mark J. Mulliganf,j, Aneesh K. Mehtae,f,j, Peter Palesec,h,and Rafi Ahmeda,b,3

bDepartment of Microbiology and Immunology, dDepartment of Pediatrics, fDivision of Infectious Diseases, Department of Medicine, jHope Clinic of theEmory Vaccine Center, aEmory Vaccine Center, and eEmory Transplant Center, Emory University School of Medicine, Atlanta, GA 30322; cDepartment ofMicrobiology, hDepartment of Medicine, and iGlobal Health and Emerging Pathogens Institute, Icahn School of Medicine at Mount Sinai, New York, NY10029; and gDepartment of Medicine, Section of Rheumatology, Committee on Immunology, The Knapp Center for Lupus and Immunology Research,University of Chicago, Chicago, IL 60637

Contributed by Rafi Ahmed, July 24, 2014 (sent for review June 24, 2014; reviewed by Yoshihiro Kawaoka and Rino Rappuoli)

The emergence of pandemic influenza viruses poses a major publichealth threat. Therefore, there is a need for a vaccine that caninduce broadly cross-reactive antibodies that protect againstseasonal as well as pandemic influenza strains. Human broadlyneutralizing antibodies directed against highly conserved epitopesin the stem region of influenza virus HA have been recentlycharacterized. However, it remains unknown what the baselinelevels are of antibodies and memory B cells that are directedagainst these conserved epitopes. More importantly, it is also notknown to what extent anti-HA stem B-cell responses get boostedin humans after seasonal influenza vaccination. In this study, wehave addressed these two outstanding questions. Our data showthat: (i) antibodies and memory B cells directed against the con-served HA stem region are prevalent in humans, but their levelsare much lower than B-cell responses directed to variable epitopesin the HA head; (ii) current seasonal influenza vaccines are effi-cient in inducing B-cell responses to the variable HA head regionbut they fail to boost responses to the conserved HA stem region;and (iii) in striking contrast, immunization of humans with theavian influenza virus H5N1 induced broadly cross-reactive HAstem-specific antibodies. Taken together, our findings provide apotential vaccination strategy where heterologous influenza im-munization could be used for increasing the levels of broadly neu-tralizing antibodies and for priming the human population torespond quickly to emerging pandemic influenza threats.

stalk | breadth | immunoglobulin | neutralization

The emergence of novel influenza virus strains poses a con-tinuous public health threat (1, 2). The World Health Or-

ganization estimates that influenza viruses infect one-billionpeople annually, with three- to five-million cases of severe illness,and up to 500,000 deaths worldwide (3). Following influenza virusinfection, humoral immune responses against the viral hemagglu-tinin (HA) protein may persist for decades in humans (4). Theseanti-HA responses correlate strongly with protection againstinfluenza infection (5). Serological memory is maintained byantibody-secreting long-lived plasma cells and reinforced bymemory B cells, which can rapidly differentiate into antibody-secreting cells upon antigen reexposure (6).Influenza vaccine efficacy is constantly undermined by antigenic

variation in the circulating viral strains, particularly in the HA andneuraminidase (NA) proteins. Current influenza vaccination strat-egies rely on changing the HA and NA components of the annualhuman vaccine to ensure that they antigenically match circulatinginfluenza strains (7, 8). Developing an influenza vaccine that iscapable of providing broad and long-lasting protective antibodyresponses remains the central challenge for influenza virus research.

HA is a trimer, with each monomer comprised of two subunits:HA1, which includes the HA globular head, and HA2, whoseectodomain together with the N- and C-terminal parts of HA1constitute the HA stem region (9). Phylogenetically, the 18 HAsubtypes characterized so far are divided into two groups. Amongstrains that have recently caused disease in humans, H1 and H5HAs belong to group 1, whereas H3 and H7 HAs belong to group2 (10). Conventional anti-HA neutralizing antibodies primarilytarget a few immunodominant epitopes located in proximity to thereceptor-binding domain within the globular head region of themolecule (11, 12). Although these antibodies are potentially pro-tective, they are strain-specific because of the high variability ofsuch epitopes, and thus lack, in general, the much-desired broadneutralizing activity. Recently, broadly neutralizing human (13–18) and murine (19) monoclonal antibodies (mAbs) directedagainst distinct epitopes within the HA stem region have beenextensively characterized. These mAbs were shown to interferewith the influenza viruses’ life cycle in different ways (20). By

Significance

Vaccination is the most effective means of attaining protectionagainst influenza viruses. However, the constantly evolvingnature of influenza viruses enables them to escape preexistingimmune surveillance, and thus thwarts public health efforts tocontrol influenza annual epidemics and occasional pandemics.One solution is to elicit antibodies directed against highlyconserved epitopes, such as those within the stem region ofinfluenza HA, the principal target of virus-neutralizing anti-body responses. This study shows that annual influenza vac-cines induce antibody responses that are largely directedagainst the highly variable HA head region. In contrast, het-erologous immunization with HA derived from influenzastrains that are currently not circulating in humans (e.g. H5N1)can substantially increase HA stem-specific responses.

Author contributions: A.H.E., M.J.M., A.K.M., P.P., and R.A. designed research; A.H.E., F.K.,G.-M.L., M.S.M., C.C., J.W., C.Y.C., M.M., and R.E. performed research; F.K., S.E., P.S., S.F.A.,P.C.W., and A.G.-S. contributed new reagents/analytic tools; A.H.E. and C.W.D. analyzeddata; A.H.E. and R.A. wrote the paper; R.E. conducted patient recruitment; and M.J.M.and A.K.M. provided the clinical study design.

Reviewers: Y.K., University of Wisconsin–Madison; and R.R., Novartis Vaccines.

The authors declare no conflict of interest.1Present address: Centre for Respiratory Infection, National Heart and Lung Institute,Imperial College London, London W2 1PG, United Kingdom.

2Present address: Assay Development Department, Quintiles, Marietta, GA 30067.3To whom correspondence should be addressed. Email: rahmed@emory.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1414070111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1414070111 PNAS | September 9, 2014 | vol. 111 | no. 36 | 13133–13138

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

1

generating monoclonal antibodies from plasmablasts isolated exvivo, we demonstrated that these broadly neutralizing antibodiescould be retrieved from patients infected with or vaccinatedagainst the pandemic H1N1 2009 influenza virus (18, 21). Recentobservations that HA stem epitopes are accessible on the ma-jority of HA trimers on intact virions (22), and that a stable HAstem protein that is immunologically intact could be produced(23), provided further hope for the feasibility of a stem-baseduniversal influenza vaccine (24).Notably, HA stem-specific mAbs isolated from humans showed

a high degree of affinity maturation, suggesting a memory B-cellorigin. These results raised two important questions that we ad-dress in the current study. First, what are the baseline levels ofbroadly cross-reactive stem-binding antibodies and memory Bcells? Second, using current influenza vaccines, to what extent canHA stem-specific responses be boosted in comparison with thosedirected against the HA globular head?Structural studies have clearly demonstrated that the main

neutralizing antibody epitopes within the HA stem region areconformation-dependent, and that the integrity of these epitopesrequires the presence of the HA1 subunit in addition to the HA2subunit, which constitute the bulk of the HA stem (16, 17). To beable to directly measure HA stem-reactive antibodies and memoryB cells, we used a chimeric HA molecule that expresses theglobular head of H9 HA on H1 backbone (25). Our data dem-onstrate that post-2009 trivalent inactivated vaccines (TIV) in-duced minimal stem-specific responses in comparison with head-specific responses. On the other hand, immunization with H5N1generated relatively strong anti-HA stem responses, demonstratingthat it is feasible to elicit broadly neutralizing responses in humansgiven the right immunogen design.

ResultsTIV Immunization Induces Minimal Stem-Specific Plasmablast andAntibody Responses. We used a recombinant, trimeric globularHA head protein from the 2009 pandemic H1N1 virus (pH1N1)and a chimeric HA that expresses the globular head of an H9HA and H1 stem to measure vaccine-induced anti-pH1 head andstem antibody and B-cell responses, respectively (Fig. 1A) (25).The chimeric protein will hereafter be referred to as the H1stem. We also used the trimeric H9 head protein to determinethe chimeric H9/H1-specific signals/spots that are caused by H9head binding. First, we wanted to determine the portion of theplasmablast response induced by TIV that is specific to the stemregion. We examined B-cell responses in 17 healthy adult volun-teers immunized with the 2012/13 TIV (Table 1). It is important tonote that HA and NA from the 2009 pH1N1 virus constituted theH1N1 component of seasonal TIVs for the 2010–2014 influenzaseasons. We determined the kinetics of anti-pH1 plasmablastresponses in blood after vaccination by ELISPOT (Fig. 1B).Antigen-specific plasmablast response peaked at day 7 before re-turning to background levels by day 14 postvaccination (Fig. 1B).At the peak of the response, plasmablasts directed against the stemwere barely detectable in comparison with those directed againstthe pH1 HA (Fig. 1C). The latter were detectable in all but threeindividuals at day 7 postvaccination, whereas H1 stem-specificplasmablasts were detectable in only four (of 17) individuals (Fig.1D). In those four individuals the average frequency of H1 stem-specific antibody secreting cells (ASCs) was 10 per millionperipheral blood mononuclear cells (PBMCs), which was 10-fold lower than the average frequency of ASCs directed againstthe pH1 HA (Fig. 1D).We next examined serum antibody responses to the pH1 globular

head and H1 stem in those subjects. We observed an average of 1.6-fold increase in H1 stem-specific IgG antibody titers from pre-vaccination [geometric mean titer (GMT) = 410] to 30 d post-TIVvaccination (GMT = 626), and a sevenfold increase in anti-pH1

head-specific IgG antibody titers (GMT = 390 and 2,777 at pre-vaccination and 30 d postvaccination, respectively) (Fig. 2 A and B).To confirm these results, we used a competition-based ap-

proach that was recently described to measure the increase inserum antibody responses to epitopes within the HA head andstem regions (Fig. S1) (13). In this assay we measured the extentto which pre- and postvaccination sera blocked the binding ofhead- and stem-specific mAbs to the pH1 HA. For stemresponses, we used a broadly neutralizing, stem-binding mAb,SF70, which was isolated from a 2009 pH1N1-infected patient(18). The epitope that SF70 binds to is highly similar to thosedescribed for the group 1 prototypic broadly neutralizing mAbsCR6261 and F10 (16, 17). For the HA head, we used 15-2G04(21), a mAb that, unlike SF70, exhibits hemagglutination in-hibition (HAI) activity against the pH1N1 virus, confirming itsbinding to an epitope within the pH1 globular head. After vac-cination, there was a small (threefold) increase in the 50%blocking dilution or BD50 values against SF70 (Fig. 2 C and D).In contrast and consistent with the ELISA data, we observeda greater increase (eightfold) in BD50 titers against the HA head-binding mAb, 15-2G04 (Fig. 2 C and D). In summary, these datashow that anti-HA stem plasmablast and antibody responsesfollowing TIV vaccination are modest in comparison with thosedirected against the HA globular head.

Fig. 1. Plasmablast responses after vaccination with the 2012/13 TIV arelargely directed against the HA head region. Healthy adult volunteers werevaccinated with the 2012/13 TIV (n = 17). PBMCs were isolated at baseline andat days 7, 14, and 30 postvaccination. (A) Structural depiction of therecombinant proteins used to measure HA head vs. HA stem responses. Asdiscussed in theMethods, the globular head region of the H1 stem protein wasderived from H9 HA (blue) whereas the stem region is from H1 (red). (B) Ki-netics of the pH1 HA-specific IgG-secreting plasmablasts. (C) A representativeIgG-specific ELISPOT analysis of day 7 plasmablasts taken from one subject.ELISPOT wells were coated with the proteins displayed above each column.The numbers on the left represent the number of PBMCs plated on each row(threefold dilution). The number given below is the spot count for that well.(D) Day 7 plasmablast responses to the 2012/13 seasonal TIV detected by ELI-SPOT. Each symbol represents one individual (n = 17). Shown is the frequencyof TIV-specific plasmablasts (black), pH1 HA-specific plasmablasts (red), and H1stem-specific plasmablasts (blue). P values are from Student t tests. Dotted linesrepresent limit of detection.

13134 | www.pnas.org/cgi/doi/10.1073/pnas.1414070111 Ellebedy et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

1

Before the emergence of the 2009 pH1N1 virus, the vastmajority of young adults who constitute the various cohorts ex-amined in the present study did not have any detectable HAItiters against this virus (26). We wanted to examine the baselinelevels of the pH1 head- and H1 stem-specific serum antibodiesover the 4-y period following the emergence of the pandemic. Itis important to note here that the H1 component of the TIVsadministered over this period was the pH1 HA and most of ourstudy subjects had detectable baseline HAI titers against thepH1N1 virus. To this end, we determined anti-pH1 head andanti-H1 stem IgG titers in preimmunization sera collected fromsubjects enrolled during the 2010/11, 2011/12, 2012/13, and 2013/14 influenza seasons (Table 1). Serum anti-pH1 head IgG levelswere lowest in the 2010/11 cohort (Fig. 3A). These levels showedonly a slight increase in the 2011/12 cohort. However, in the2012/13 season, we detected a significant (P = 0.026) increase inanti-pH1 HA head-specific antibody titers compared with the2010/11 cohort. There was no significant gain in such titers be-tween the 2012/13 and 2013/14 cohorts (Fig. 3A). In contrast,anti-H1 stem titers did not significantly change over the sameperiod (Fig. 3B). These data confirm our earlier observation thatimmunization with TIV results in a greater antihead—in com-parison to antistem—antibody response.

Head-Specific Memory B-Cells Dominate After Immunization withTIV. We next determined the baseline and post-TIV immuni-zation frequency of blood memory B cells using the previouslydescribed memory B-cell assay (27). For detecting influenza HA-specific responses we used the antigens shown in Fig. 1A. Wedetermined the frequency of IgG+ memory B cells that are di-rected against the pH1 head (n =12) and H1 stem (n =16) afterTIV (2011/12 and 2012/13) immunization. Consistent withplasmablast and antibody responses, we observed a large in-crease in the frequency of anti-pH1 head IgG+ memory B cells(median = 0.033% and 0.45% at day 0 and day 30 post-vaccination, respectively, P = 0.04) and a modest increase in anti-H1 stem IgG+ memory B cells (from 0.02% to 0.09% at days0 and 30 postvaccination, respectively) (P = 0.012) (Fig. 4).These data show that although stem-specific IgG+ memory Bcells are detectable in most individuals, they are minimallyboosted by TIV immunization in comparison with the head-specific ones.

Enhanced Anti-HA Stem Antibody Responses After H5N1 Vaccination.We have shown that cross-reactive B cells dominated the plas-mablast response following the 2009 pH1N1 vaccination (21).We wanted to determine whether immunization with a similarly

heterologous (relative to the seasonal antigens) influenza vac-cine would affect the trend of serum antibody responses to theHA head vs. stem regions. Therefore, we determined anti-H5HA head and anti-H1 stem antibody levels in 17 paired serumsamples collected before and after immunization with an inac-tivated H5N1 vaccine derived from A/Vietnam/04/1203 orA/Indonesia/05/2005 (Table 1) (28). Those subjects received abooster H5N1 immunization with a vaccine that was derivedfrom A/Indonesia/05/2005 6 mo later (28). Blood samples wereanalyzed at four time points; baseline, 28 d following the primaryimmunization and before the booster immunization, and 28 dafter the booster immunizations. Both H5 and H1 belong togroup 1 HAs and have a significant degree of homology in theamino acid sequence of their stem regions; therefore, we usedthe chimeric H9/H1 HA molecule to measure anti-H5 HA stem-specific antibody responses by ELISA. We also measured anti-body titers against H7 HA, a representative of group 2 HAs.Interestingly, there was an enhanced (an average of fourfold)increase in stem-specific IgG antibody titers from prevaccination(GMT = 925) to day 28 postprimary H5N1 vaccination (GMT =3,330, P = 0.0013). On the other hand, the increase in anti-H5head antibody titers was modest (1.8-fold increase, GMT = 150and 270 at prevaccination and 28 d postprimary vaccination,

Table 1. Number of subjects, year of enrollment, and influenzavaccines used in the study

Vaccine Year Vaccine strains No. of subjects

TIV 2010/11 A/California/7/09 (H1N1) 18A/Perth/16/2009 (H3N2)B/Brisbane/60/2008

TIV 2011/12 A/California/7/09 (H1N1) 16A/Perth/16/2009 (H3N2)B/Brisbane/60/2008

TIV 2012/13 A/California/7/09 (H1N1) 17A/Victoria/361/2011 (H3N2)B/Wisconsin/1/2010

TIV 2013/14 A/California/7/09 (H1N1) 10A/Victoria/361/2011 (H3N2)B/Massachusetts/2/2012

H5N1 2008/09 A/Vietnam/1203/2004 17A/Indonesia/05/2005

Fig. 2. Serological analysis of HA head- and stem-specific antibodyresponses following immunization with the 2012/13 TIV. Pre- and 30 dpostvaccination sera from individuals vaccinated with the 2012/13 TIV wereexamined by ELISA (A and B) and blocking-of-binding assay (C and D). (A)Pre- and 30 d postvaccination IgG antibody titers against the pH1 HA head(Left, red) or H1 stem (Right, blue). Each symbol represents one individual(n = 17). P values are from paired Student t tests. Dotted lines represent limitof detection. (B) Fold-change in IgG antibody titers against the pH1 HA head(red) and H1 stem (blue). (C) ELISA plates were coated with the pH1 HA andincubated with serial dilutions of the sera followed by either the HA head-binding biotinylated mAb 15-G04 (Left, red) or the HA stem-binding SF70(Right, blue). The binding of the mAbs was detected by enzyme-conjugatedstreptavidin. Shown are the reciprocal serum dilutions that blocked 50%(BD50) of the mAb binding. Dotted lines represent limit of detection. (D)Fold-change in BD50 values against 15-2G04 (red) and SF70 (blue) after TIVvaccination.

Ellebedy et al. PNAS | September 9, 2014 | vol. 111 | no. 36 | 13135

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

1

respectively; P = 0.018) (Fig. 5 A and B). The increase in stem-specific antibody responses was largely group 1-specific, as therewas barely any increase (1.2-fold) in anti-H7 HA IgG antibodyresponses (Fig. 5B).Responses after the H5N1 booster immunization resembled

those observed after TIV immunization; there was a strong anti-H5 head antibody response (ninefold increase, GMT = 265 and2,360 at days 180 and 208 postprimary vaccination, respectively;P < 0.0001) and a feeble stem-specific response (1.8-fold in-crease, GMT = 2,000 and 3,780 at days 180 and 208 postprimaryvaccination, respectively; P = 0.004) (Fig. 5 C and D). As withresponse after the priming immunization, minimal increase (1.4-fold) in anti-H7 HA titers were observed after the booster im-munization (Fig. 5D).Next, we compared the increase in group 1 neutralizing anti-

bodies following a single TIV vs. H5N1 immunization. To avoidany interference from H1 head-, H5 head-, or N1 NA-specificantibodies, we used a chimeric H9N3 virus that was previouslygenerated and characterized (25). After TIV immunization, wedetected an increase in neutralizing antibody titers in only 2 ofthe 15 subjects tested (mean titers were 145.8 and 132.9 at days0 and 30 postvaccination, respectively, P = 0.6) (Fig. S2A). Incontrast, stem-directed neutralizing antibody titers modestly in-creased in 11 of the 16 subjects tested (mean titers were 198.6and 397.3 at days 0 and 28 postvaccination, respectively, P =0.042) (Fig. S2B).In summary, these data show that the balance between vaccine-

induced anti-HA head vs. stem antibody responses can be shiftedin favor of stem responses by vaccination using HA moleculesagainst which humans have minimal preexisting antibody titers/memory B cells, such as those derived from avian H5N1 viruses.However, upon boosting with the same HA, the dominance ofantihead responses is restored.

DiscussionIn this study we report three key findings in regards to humanB-cell and antibody responses to the conserved influenza HAstem region: (i) baseline antibodies and memory B cells directedagainst the HA stem region are widely prevalent, albeit at lowlevels; (ii) although immunization with TIV did induce someB-cell and antibody responses to epitopes within the HA stemregion, the responses were disproportionately higher againstthe HA globular head epitopes; and (iii) immunization with aninactivated H5N1 vaccine considerably improved the cross-reactive anti-HA stem responses. Clearly, these findings haveimportant implications for the pursuit of HA stem-based

“universal” influenza vaccines and for influenza immunizationstrategies in general.Our data show that there is a clear bias in favor of the HA

globular head over stem in terms of the magnitude of antibodiesdirected against either region in response to influenza vaccines ingeneral. This finding is indicated by our observation that theoverall anti-H1 stem-specific antibody levels, unlike their anti-pH1 HA head counterparts, did not significantly change overthe 4 y following the emergence of the 2009 pH1N1 (Fig. 3).We have previously proposed that broadly cross-reactive anti-

bodies, such as those directed against the HA stem, are producedby low-frequency memory B cells that are specific to conserved butsubdominant epitopes (21). However, after introduction of a rel-atively novel HA from a noncirculating flu strain, cross-reactiveB-cell responses make up a relatively greater proportion of thehumoral immune response. The relatively strong HA stem-specificresponses after primary H5 vaccination provide additional supportto this hypothesis. Additional lines of evidence for the enhancedantistem responses after H5 vaccination or infection in humansinclude: the isolation of stem-specific mAbs from combinatorialantibody libraries generated from the bone marrow of five survi-vors of an avian H5N1 outbreak (29); the increase in stem-bindingantibodies after H5 immunization as reported by Ledgerwoodet al. (30); and isolation, using a flow cytometry-based approach,of B cells expressing the broadly neutralizing HA stem-specificmAbs from an individual immunized with H5N1 vaccine (31).Finally, Khurana et al. observed that serum antibodies that boundto linear peptides from the HA2 subunit increased in individualsimmunized with either unadjuvanted or alum-adjuvanted H5N1vaccines (32).Notably, all vaccines used in our study were unadjuvanted, and

it has been proposed that adjuvants, such as the oil-in-wateremulsion MF59, could expand the breadth of the antibodyresponses to influenza vaccines (32). In the latter study, theauthors showed an increase in antibody titers directed againstepitopes within the HA1 subunit in sera from individuals im-munized with MF59-adjuvanted H5N1 vaccine. This result canbe explained by the fact that they analyzed sera obtained aftera second or a third booster immunization, whereas we observedpotent antistem responses after primary H5N1 immunization. Ingeneral it is important to make a distinction between the ability

Fig. 3. Analysis of the pH1 HA head- and H1 stem-specific IgG antibodytiters over the 2010–2014 period. Prevaccination sera collected from subjectsenrolled during the 2010/11 (n = 18), 2011/12 (n = 16), 2012/13 (n = 11), and2013/14 (n = 10) influenza seasons. Geometrical mean IgG titers directedagainst the pH1 head (A, red) and H1 stem (B, blue) are depicted. Eachsymbol represents one individual. P values are from Student t tests. Dottedlines represent limits of detection.

Fig. 4. Memory B-cell responses induced following immunization with TIV.PBMCs isolated either before- or 30 d after immunization with either the2011/12 or the 2012/13 TIV. The frequency of pre- and 30 d postvaccinationlevels of IgG+ memory B cells directed against the pH1 head (Left, red) or H1stem region (Right, blue). P values are from paired Student t tests. Dottedlines represent limit of detection.

13136 | www.pnas.org/cgi/doi/10.1073/pnas.1414070111 Ellebedy et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

1

of adjuvants to increase the overall magnitude of immuneresponses and their ability to qualitatively modulate an immuneresponse with an already established immunodominance hier-archy. To test the latter point in context of influenza, the impactof adjuvants on the ratio and longevity of antihead vs. stemresponses after vaccination should be examined.The most probable explanation for our results with primary H5

vaccination is that vaccinees had relatively lower frequency ofpreexisting memory B cells specific to most of the prominent H5HA head epitopes in comparison with the stem-specific ones.Thus, upon immunization more HA stem-specific memory B cellsgot recruited to the immune reaction and differentiated into ASCs.However, these conditions have probably changed before theH5N1 booster immunization, and the preexisting frequency of H5head-specific memory B cells has relatively increased. Therefore,the booster immunization resulted in significantly larger head-specific responses in comparison with stem-specific ones. Thisconcept has been clearly demonstrated in animal models (33, 34).The effect of homologous boosting is another point raised by

our study. Although this strategy does eventually lead to signifi-cant accumulation of HA-specific antibody titers, these antibodiesare mostly directed against the strain-specific head region. Under

this scenario, boosting the cross-reactive HA stem-specificresponses is largely marginalized.The high prevalence of HA stem-specific antibodies at base-

line in tested subjects is consistent with earlier studies reportingthe detection of baseline neutralizing antibody titers againstH5N1 in healthy volunteers (13, 35). Our data indicate that mosthumans are capable of establishing a humoral immune memorythat is specific to the conserved HA stem region. This finding isintriguing in light of what was recently suggested that antibodiesdirected against the HA2 subunit might enhance virus fusion andthus virus-induced pathology in the absence of virus-neutralizingresponses (36). The latter suggestion indicates that other factors,such as the quantity and the exact epitopes targeted by antistemantibodies, play a role in determining the overall impact of suchantibodies on immunity to influenza. It is important to note herethat the main neutralizing epitope within the HA stem is con-formational, and thus cannot be detected by measuring bindingto linear peptides from the HA2 subunit. Overall, our data raisethe important question of what would be the minimal “concen-tration” of antistem antibodies required to provide in vivo pro-tection (37). Therefore, it will be important in future studiesto determine the quantity of HA stem-specific antibodies ormemory B cells that would positively correlate with better clin-ical outcomes against influenza infections (38).In summary, the isolation and characterization of broadly

neutralizing human mAbs directed against the conserved stemregion of influenza HA represents a potentially important steptoward developing a universal influenza vaccine. Our data showthat low levels of antibodies and memory B cells directed againstthe HA stem region are widely prevalent in humans. However,TIVs induce B-cell responses that are largely directed against theHA head region. In contrast, heterologous immunization withHA derived from influenza strains that are currently not circu-lating in humans has greatly increased HA stem-specific re-sponses. Our findings delineate a potential vaccination strategywhere H5N1 or H7N9 immunization could be used not only forimmunologically priming the population to quickly respond toserious pandemic influenza threats, but also for generatingbroadly neutralizing antibodies against influenza in humans.

MethodsAll studies were approved by the Emory University institutional review board.Healthy adult volunteers were given either the TIV from the 2010/11, 2011/12,2012/13, and 2013/14-influenza seasons or a monovalent inactivated H5N1vaccine. Generation and characterization of the chimeric H9N3 virus waspreviously described (25). Direct ELISPOT to enumerate total and HA specificplasmablasts were performed as previously described (39). RecombinantHA-specific ELISA, as well as HAI (done with the chimeric H9N3 virus) wereperformed as previously described (40). More detailed materials andmethods are presented in the SI Methods.

ACKNOWLEDGMENTS. This work was funded in parts by National Instituteof Allergy and Infectious Diseases Contracts HHSN266200700006C (to R.A.and P.C.W.) and HHSN26620070010C (to A.G.-S. and P.P.); ContractHHSN272200800005C (to M.J.M.), which supported the H5N1 vaccinationtrial; Grant 1P01AI097092 (to P.P., P.C.W., and R.A.); Grant 1U19AI109946-01(to P.P.); the Georgia Research Alliance (M.J.M.); Children’s Healthcare ofAtlanta (M.J.M.); the National Center for Advancing Translational Sciencesof the National Institutes of Health under Award UL1TR000454 (to M.J.M.);Training Grant T32AI074492 from the National Institute of Allergy and In-fectious Diseases (to A.H.E.); Erwin Schrödinger Fellowship J3232 from theAustrian Science Fund (to F.K.); a Canadian Institutes of Health ResearchPostdoctoral fellowship (to M.S.M.). Some of clinical studies were supportedby the National Center for Advancing Translational Sciences of the NationalInstitutes of Health under Award UL1TR000454.

1. Rappuoli R, Dormitzer PR (2012) Influenza: Options to improve pandemic prepara-

tion. Science 336(6088):1531–1533.2. Webby RJ, Webster RG (2003) Are we ready for pandemic influenza? Science

302(5650):1519–1522.

3. WHO (2008) Immunization, vaccines and biologicals: Influenza. Available at

www.who.int/immunization/topics/influenza/en/. Accessed June 20, 2014.4. Yu X, et al. (2008) Neutralizing antibodies derived from the B cells of 1918 influenza

pandemic survivors. Nature 455(7212):532–536.

Fig. 5. Robust antistem antibody responses following primary H5N1 im-munization. Sera collected before- and 28 d after priming (A and B) andbooster (C and D) H5N1 immunization. The time interval between thepriming and the booster immunizations was 6 mo. (A) Pre- and 28 d post-primary immunization IgG endpoint titers directed against Indonesia H5 HAhead region (Left, magenta) and H1 stem region (Right, green). P values arefrom Student t tests. (B) Fold-change in IgG antibody titers against theIndonesia H5 HA head (magenta), H1 stem (green), and H7 HA (black) afterprimary H5N1 immunization. (C) Pre- and 28 d postboost IgG endpoint titersdirected against Indonesia H5 HA head region (Left, magenta) and H1 stemregion (Right, green). P values are from Student t tests. (D) Fold-change inIgG antibody titers against the Indonesia H5 HA head (magenta), H1 stem(green), and H7 HA (black) after secondary H5N1 immunization.

Ellebedy et al. PNAS | September 9, 2014 | vol. 111 | no. 36 | 13137

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

1

5. Francis T, Jr (1953) Vaccination against influenza. Bull World Health Organ 8(5-6):725–741.6. Slifka MK, Antia R, Whitmire JK, Ahmed R (1998) Humoral immunity due to long-lived

plasma cells. Immunity 8(3):363–372.7. Ellebedy AH, Webby RJ (2009) Influenza vaccines. Vaccine 27(Suppl 4):D65–D68.8. Monto AS (2010) Seasonal influenza and vaccination coverage. Vaccine 28(Suppl 4):

D33–D44.9. Skehel JJ, Wiley DC (2000) Receptor binding and membrane fusion in virus entry: The

influenza hemagglutinin. Annu Rev Biochem 69:531–569.10. Ellebedy AH, Ahmed R (2012) Re-engaging cross-reactive memory B cells: The in-

fluenza puzzle. Front Immunol 3:53.11. Caton AJ, Brownlee GG, Yewdell JW, Gerhard W (1982) The antigenic structure of the

influenza virus A/PR/8/34 hemagglutinin (H1 subtype). Cell 31(2 Pt 1):417–427.12. Gerhard W, Yewdell J, Frankel ME, Webster R (1981) Antigenic structure of influenza

virus haemagglutinin defined by hybridoma antibodies. Nature 290(5808):713–717.13. Corti D, et al. (2010) Heterosubtypic neutralizing antibodies are produced by individuals

immunized with a seasonal influenza vaccine. J Clin Invest 120(5):1663–1673.14. Corti D, et al. (2011) A neutralizing antibody selected from plasma cells that binds to

group 1 and group 2 influenza A hemagglutinins. Science 333(6044):850–856.15. Dreyfus C, et al. (2012) Highly conserved protective epitopes on influenza B viruses.

Science 337(6100):1343–1348.16. Ekiert DC, et al. (2009) Antibody recognition of a highly conserved influenza virus

epitope. Science 324(5924):246–251.17. Sui J, et al. (2009) Structural and functional bases for broad-spectrum neutralization

of avian and human influenza A viruses. Nat Struct Mol Biol 16(3):265–273.18. Wrammert J, et al. (2011) Broadly cross-reactive antibodies dominate the human

B cell response against 2009 pandemic H1N1 influenza virus infection. J Exp Med208(1):181–193.

19. Tan GS, et al. (2012) A pan-H1 anti-hemagglutinin monoclonal antibody with potentbroad-spectrum efficacy in vivo. J Virol 86(11):6179–6188.

20. Brandenburg B, et al. (2013) Mechanisms of hemagglutinin targeted influenza virusneutralization. PLoS ONE 8(12):e80034.

21. Li GM, et al. (2012) Pandemic H1N1 influenza vaccine induces a recall response inhumans that favors broadly cross-reactive memory B cells. Proc Natl Acad Sci USA109(23):9047–9052.

22. Harris AK, et al. (2013) Structure and accessibility of HA trimers on intact 2009 H1N1pandemic influenza virus to stem region-specific neutralizing antibodies. Proc NatlAcad Sci USA 110(12):4592–4597.

23. Lu Y, Welsh JP, Swartz JR (2014) Production and stabilization of the trimeric influenzahemagglutinin stem domain for potentially broadly protective influenza vaccines.Proc Natl Acad Sci USA 111(1):125–130.

24. Nabel GJ, Fauci AS (2010) Induction of unnatural immunity: Prospects for a broadlyprotective universal influenza vaccine. Nat Med 16(12):1389–1391.

25. Pica N, et al. (2012) Hemagglutinin stalk antibodies elicited by the 2009 pandemicinfluenza virus as a mechanism for the extinction of seasonal H1N1 viruses. Proc NatlAcad Sci USA 109(7):2573–2578.

26. Hancock K, et al. (2009) Cross-reactive antibody responses to the 2009 pandemic H1N1influenza virus. N Engl J Med 361(20):1945–1952.

27. Crotty S, Aubert RD, Glidewell J, Ahmed R (2004) Tracking human antigen-specificmemory B cells: A sensitive and generalized ELISPOT system. J Immunol Methods286(1-2):111–122.

28. Belshe RB, et al.; National Institute of Allergy and Infectious Diseases-Funded Vaccineand Treatment Evaluation Units (2011) Safety and immunogenicity of influenza A H5subunit vaccines: Effect of vaccine schedule and antigenic variant. J Infect Dis 203(5):666–673.

29. Kashyap AK, et al. (2008) Combinatorial antibody libraries from survivors of theTurkish H5N1 avian influenza outbreak reveal virus neutralization strategies. ProcNatl Acad Sci USA 105(16):5986–5991.

30. Ledgerwood JE, et al.; VRC 306 Study Team (2011) DNA priming and influenza vaccineimmunogenicity: Two phase 1 open label randomised clinical trials. Lancet Infect Dis11(12):916–924.

31. Whittle JR, et al. (2014) Flow cytometry reveals that H5N1 vaccination elicits cross-reactive stem-directed antibodies from multiple Ig heavy-chain lineages. J Virol 88(8):4047–4057.

32. Khurana S, et al. (2010) Vaccines with MF59 adjuvant expand the antibody repertoireto target protective sites of pandemic avian H5N1 influenza virus. Sci Transl Med2(15):ra5.

33. Krammer F, Palese P (2014) Universal influenza virus vaccines: Need for clinical trials.Nat Immunol 15(1):3–5.

34. Krammer F, Pica N, Hai R, Margine I, Palese P (2013) Chimeric hemagglutinin influenzavirus vaccine constructs elicit broadly protective stalk-specific antibodies. J Virol87(12):6542–6550.

35. Sui J, et al. (2011) Wide prevalence of heterosubtypic broadly neutralizing humananti-influenza A antibodies. Clin Infect Dis 52(8):1003–1009.

36. Khurana S, et al. (2013) Vaccine-induced anti-HA2 antibodies promote virus fusionand enhance influenza virus respiratory disease. Sci Trans Med 5(200):200ra114.

37. Yewdell JW (2011) Viva la revolucion: Rethinking influenza a virus antigenic drift.Curr Opinion Virol 1(3):177–183.

38. Corti D, Lanzavecchia A (2013) Broadly neutralizing antiviral antibodies. Annu RevImmunol 31:705–742.

39. Crotty S, et al. (2003) Cutting edge: Long-term B cell memory in humans aftersmallpox vaccination. J Immunol 171(10):4969–4973.

40. Wrammert J, et al. (2008) Rapid cloning of high-affinity human monoclonal anti-bodies against influenza virus. Nature 453(7195):667–671.

13138 | www.pnas.org/cgi/doi/10.1073/pnas.1414070111 Ellebedy et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

1