Influenza virus infection history shapes antibody responses to influenza vaccination

Studies of successive vaccination suggest that immunological memory against past influenza viruses may limit responses to vaccines containing current strains. The impact of memory induced by prior infection is rarely considered and is difficult to ascertain, because infections are often subclinical. This study investigated influenza vaccination among adults from the Ha Nam cohort (Vietnam), who were purposefully selected to include 72 with and 28 without documented influenza A(H3N2) infection during the preceding 9 years (Australian New Zealand Clinical Trials Registry 12621000110886). The primary outcome was the effect of prior influenza A(H3N2) infection on hemagglutinin-inhibiting antibody responses induced by a locally available influenza vaccine administered in November 2016. Baseline and postvaccination sera were titrated against 40 influenza A(H3N2) strains spanning 1968–2018. At each time point (baseline, day 14 and day 280), geometric mean antibody titers against 2008–2018 strains were higher among participants with recent infection (34 (29–40), 187 (154–227) and 86 (72–103)) than among participants without recent infection (19 (17–22), 91 (64–130) and 38 (30–49)). On days 14 and 280, mean titer rises against 2014–2018 strains were 6.1-fold (5.0- to 7.4-fold) and 2.6-fold (2.2- to 3.1-fold) for participants with recent infection versus 4.8-fold (3.5- to 6.7-fold) and 1.9-fold (1.5- to 2.3-fold) for those without. One of 72 vaccinees with recent infection versus 4 of 28 without developed symptomatic A(H3N2) infection in the season after vaccination (P = 0.021). The range of A(H3N2) viruses recognized by vaccine-induced antibodies was associated with the prior infection strain. These results suggest that recall of immunological memory induced by prior infection enhances antibody responses to inactivated influenza vaccine and is important to attain protective antibody titers. Recent prior influenza A infection is associated with elevated hemagglutinin-inhibiting antibody responses and greater breadth of reactivity to influenza strains following vaccination, suggesting that infection history boosts vaccine responses.

R NA viruses undergo relatively rapid mutation, which can critically impact vaccination strategies 1 . Influenza viruses are particularly prone to substitutions within the major surface protein, hemagglutinin (HA), as a consequence of viral RNA replication without proofreading 2 and selection of human antibody escape mutants. This process, termed antigenic drift, facilitates recurrent influenza infection throughout life. In turn, prevention by vaccination requires repeated administration of vaccine containing regularly updated virus strains. Vaccine effectiveness (VE) has been found to be poor against A(H3N2) viruses from 2010 onward, when VE estimation by subtype became more widely implemented 3 . This could, in part, be due to greater mismatch between vaccine and circulating strains. A(H3N2) viruses have undergone greater antigenic evolution compared to A(H1N1) and B influenza viruses 4 , and more often acquire substitutions within antigenic sites when propagated in eggs to produce vaccine 4,5 . It is further speculated that vaccine immunogenicity and effectiveness may be limited by recall of immunological memory against past strains, a hypothesis that was first proposed in the 1960s and termed original antigenic sin 6 . Interest in this phenomenon has been revived by a series of recent reports that antibody responses 7 and VE against A(H3N2) viruses [8][9][10][11] are attenuated among people who received vaccine in prior years. A meta-analysis indicates that although adverse effects of repeat vaccination are more pronounced for A(H3N2) than for other subtypes, there is substantial heterogeneity in effects 12 Fig. 1 | Participant selection and investigation of previously circulating A(H3N2) viruses. a, Study design and timeline. d, day; Dec, December; Nov, November; Pre, before. b, Phylogenetic tree of the HA genes of viruses recovered from Ha Nam cohort ILI cases (colored by season) and viruses used to construct antibody landscapes (colored black if cell grown or red if egg grown). Viruses from participants of the vaccine study are indicated by the suffix 'vax'. Clades (cl.) and subclades are delineated using parentheses. c, Model of the globular head of HK14e HA (SWISS-MODEL: A0A0K0yAS1), showing amino acid positions within antigenic sites A to E that differed from at least one of the prior infecting strains and receptor binding site (RBS) residues. d, Antigenic site positions that varied between HK14e and at least one prior infecting strain are tabulated and shaded according to amino acid properties. Substitutions that result in gain (+) or loss (−) of glycosylation are colored in pink. Egg-adapted substitutions are indicated by a superscript e. The asterisks in b and d indicate the viruses that were used for landscape serology, and that are listed in the associated table.

Nature MediciNe
The cellular and molecular mechanisms underlying the variable effects of prior vaccination and pre-existing immunity remain largely undefined. The antigenic distance hypothesis postulates that when successive vaccine strains are antigenically similar, existing antibodies or memory B cells attenuate vaccine immunogenicity by masking or clearing vaccine antigen, resulting in attenuated VE if the vaccine and epidemic strains differ, but not if they are similar 13 . Alternately, it is hypothesized that memory B cells induced by prior vaccination dominate and focus responses on epitopes that are conserved between prior and prevailing vaccine strains, compromising responses against epitopes that have changed 14 . This could enhance antibody responses and VE, if epidemic strains retain those conserved epitopes but reduce VE if these epitopes have changed 14 . The epitopes recognized by influenza virus neutralizing antibodies are largely located on the globular head of HA, surrounding the receptor binding site 15 . Up to 131 amino acid positions in the head of HA of A(H3N2) viruses have been associated with antigenic variation and assigned to one of five antigenic sites, designated A to E 16,17 . Antigenic sites A and B are immunodominant 16 , and single amino acid substitutions in these sites can result in escape from vaccine-induced immunity, particularly if glycosylation sites are introduced 4,18 .
Few studies consider how prior influenza infections affect the immunogenicity and protection afforded by influenza vaccines. Understanding infection history is contingent on detecting asymptomatic or subclinical infection, which may account for up to three-quarters of influenza virus infections 19,20 . To this end, we investigated vaccine immunogenicity among participants of a cohort in northern Vietnam (Ha Nam cohort) 19 who were influenza vaccine naive and had been monitored for clinical or subclinical influenza infection for 9 years.

Results
Study design. The primary study objective was to determine the effect of recent influenza A(H3N2) infection on vaccine-induced antibody responses against A(H3N2) viruses. Therefore, adult participants of the Ha Nam cohort were purposefully selected based on influenza infection history during the preceding 9 years of cohort participation (ANZCTR 12621000110886) ( Fig. 1a and Extended Data Fig. 1). The Ha Nam cohort commenced in December 2007, when 945 members of 270 randomly selected households were enrolled to participate in monitoring for influenza infection via active influenza-like illness (ILI) surveillance, RT-PCR on swabs and serology on blood samples collected annually or biannually to detect seroconversion 19 Table 2). The primary outcome was fold-rise in antibody titer, determined from geometric mean ratios (GMRs) on days 14 and 280 after vaccination, comparing vaccinees with and without A(H3N2) infection during the preceding 9 years.  Table 1). The year that participants were last infected with an A(H3N2) virus ranged from 2008 to 2015. A(H3N2) viruses circulating during these years belonged to a range of genetic clades ( Fig. 1b and Extended Data Table 1), and were distinct from the vaccine strain-A/Hong Kong/4801/2014 (HK14e), which belongs to clade-3c2a. Twenty-six antigenic site positions differed between at least one prior strain and HK14e (Fig. 1c,d). Viruses circulating in 2014 (HN14/Sw13-like, clade-3c3a) differed substantially from HK14e in sites A and B, where seven amino acids were substituted, including six to amino acids that had different properties or affected a glycosylation site. Viruses circulating in 2012 (HN12-like, clade-3c1) and earlier differed more from HK14e in site C, where five positions were substituted to an amino acid with different properties or affecting a glycosylation site (Fig. 1d).

Antibody responses to the vaccine A(H3N2) strain.
Analysis of the kinetics of antibody production against the HK14e vaccine strain (Fig. 2a) showed that vaccination induced robust antibody production within 7 days. Titers were highest on day 14 and remained at least fourfold higher than prevaccination titers on day 280 for 54% of participants. Geometric mean titers (GMTs) were higher among participants who had recent A(H3N2) virus infection at all time     Articles Nature MediciNe points. GMRs for participants with and without prior infection were 9.8-fold (7.3-to 13.1-fold) versus 9.3-fold (5.8-to 14.8-fold) on day 14 and 4.0-fold (3.0-to 5.3-fold) versus 2.5-fold (1.6-to 4.0-fold) on day 280. Proportions seropositive (titer ≥40) and seroconverted (titer rise ≥4-fold) were also higher among participants with recent infection (Table 1). Titers and titer rises were at least as high for older compared to younger adults, particularly for those with recent prior A(H3N2) infection (Extended Data Fig. 2). Recent A(H3N2) virus infection had little effect on the proportion of participants seropositive against A(H1N1)pdm09 in the vaccine (Table 1). These results indicate that recent A(H3N2) virus infection enhanced the capacity of the vaccine to induce and maintain A(H3N2)-reactive, but not A(H1N1)-reactive, antibodies. Therefore, effects of recent infection were likely to be mediated by subtype-specific memory B cells rather than by broadly cross-reactive B or T cells.

Cross-reactivity of vaccine-induced antibodies.
The strain coverage of antibodies induced by vaccination was examined using generalized additive models (GAMs) to estimate titer and titer rise landscapes for viruses arranged by circulation year. Prevaccination antibody titers were relatively high against strains encountered early in life, as well as against 1993-2002 strains ( Fig. 2b and Supplementary Fig. 2). Titer rise from day 7 to day 280 after vaccination diminished as virus genetic and temporal distance from HK14e increased and was negligible against the oldest strains ( Fig.  2c,d and Supplementary Fig. 3). The boosting of titers against past strains, referred to as back-boosting, could reflect low-avidity antibody binding to past strains when antibody concentrations are high, as titer rise extended across more strains on day 14 after vaccination than on day 280 (Fig. 2d). However, back-boosting was largely limited to strains circulating after participant's birth years (Extended Data Fig. 3). This finding is consistent with a previous study 21 and suggests that back-boosting reflects recall of memory B cells induced by prior infections.  Table 3). GMRs averaged against vaccine and subsequently circulating strains differed most on day 280, when titers were on average 2.6-fold (2.2-to 3.1-fold) higher than baseline titers among participants with recent infection compared to 1.9-fold (1.5-to 2.3-fold) higher among participants without recent infection (Fig. 3e). This boosting of titers against future strains was accompanied by higher rates of seroconversion against subsequent A/Kansas/14/2017 (Ka17) and A/Brisbane/60/2018 (Br18) strains ( Table 1). Effects of recent infection were observed across all ages (Extended Data Fig. 3) and whether participants had one, two or three recent infections (Extended Data Fig. 4). However, titers and titer rises tended to increase with proximity of prior infection (Extended Data Fig. 5). Importantly, participants with prior infection had higher GMTs on day 280 against viruses from clades that were causing infections in the cohort by that time (Fig. 3b and Extended Data Table 2). We have shown previously that HI titers of 40 can be associated with substantial protection in this cohort 22 .

Effect of recent infection on antibody cross-reactivity.
These results indicate that memory from recent infection enhances the magnitude and breadth of A(H3N2)-reactive antibodies induced by vaccination.
Strain-specific effects of prior infections. As described above, amino acids that were substituted between prior A(H3N2) strains and the HK14e vaccine strain were concentrated within antigenic sites A and B for HN14-like (clade-3c3a) viruses that were detected in the cohort from 2013 to 2015 but within site C for 2009 and 2012 strains (Fig. 1d). We therefore examined whether the cross-reactivity of antibodies induced by vaccination differed between participants who had been infected with an A(H3N2) virus that was HN14-like versus HN09-or HN12-like. Participants infected multiple times between 2009 and 2012 were excluded (Supplementary Table 3). Antibody landscapes, modeled against a two-dimensional map of virus antigenic distances, differed between participants infected with HN14-like viruses versus earlier viruses (Fig. 4). Most notably, day 14 postvaccination landscapes were relatively skewed toward HN14 and other clade-3c3a viruses among the group with prior HN14-like virus infection (Fig. 4b). Similar trends were observed for landscapes on day 280 after vaccination and titer-rise landscapes (Extended Data Fig. 6a-d). These results suggest that memory recall may drive antibody production toward epitopes that are shared between the vaccine strain and prior infecting strains.
To further investigate whether the prior infecting strain affects antibody production against site B of the HK14e vaccine, sera were titrated against reverse-engineered viruses bearing wild-type HK14e HA or HA containing a substitution in site B (Fig. 4c). Y159S was chosen because substitutions at this position are known to have large antigenic effects 18 and because Sw13e has an S at position 159 and is antigenically distinct from HK14e ( Fig. 1d and Fig. 4a). Characterization using ferret antisera indicated that the Y159S virus was antigenically distinct from HK14e, as well as from Sw13e (clade-3c3a) (Supplementary Table 4). Antibody titers against wild-type versus Y159S virus were compared using   Articles Nature MediciNe microneutralization (MN) and HI assays, which were strongly correlated (Fig. 4d,e). MN titers were higher against wild-type compared to Y159S virus regardless of the prior infecting strain (Fig. 4f,h), which could be in part because the infectious dose of virus in the assay was marginally higher for the Y159S virus ( Supplementary  Fig. 3). Nevertheless, differences between wild-type and Y159S virus titers were greater among participants with prior HN09 and/or HN12-like virus infection than among those with prior HN14-like virus infection (Fig. 4f-i and Extended Data Fig. 6e,f). HI titers at baseline were on average 1.6-fold higher against Y159S virus among participants with prior HN09/HN12-like virus infection (Fig. 4g).
This ratio increased after vaccination to around 3-fold on day 14 ( Fig. 4i) and 2.6-fold on day 280 (Extended Data Fig. 6h), indicating that vaccination induced antibodies against site B of HK14e among participants with prior HN09 and/or HN12 infection. In contrast, postvaccination HI titers were equivalent against wild-type and Y159S virus among participants with prior HN14-like virus infection (Fig. 4h,i), suggesting that less of the antibody induced was directed against immunodominant site B. In turn, we speculate that there may have been better induction of antibodies against subdominant sites, such as site C, which was relatively well conserved across past and future strains (Supplementary Table 5).    Postvaccination antibody titers were relatively low among vaccinees who developed symptomatic A(H3N2) virus infection (Fig. 5a versus Fig. 5b), and titer rises were transient and did not increase from day 7 to day 21 ( Fig. 5c versus Fig. 5d). Titer rise was greater after infection than after vaccination in these participants and increased further from day 7 to day 21 (Fig. 5b,d), indicating that infection was more immunogenic than vaccination. The vaccinee who developed symptomatic infection despite prior infection with HN09-like and HN12-like viruses had relatively high postvaccination titers against the vaccine strain (Fig. 5e). However, this individual was infected with an A/Switzerland/8060/2017 (Sw17)-like (clade-3c2a2) virus and had low titers and no titer rise against Sw17 (Fig. 5e). Symptomatic A(H3N2) cases came from four households, which together contained five vaccinees who were not symptomatically infected despite possible exposure (Extended Data Table 4). Three of the five unaffected vaccinees from these households had recent prior infection compared to only one of five cases (Extended Data Table 4). Taken together, these results suggest that adults who were infected with an A(H3N2) virus up to 9 years before vaccination were better protected against antigenically drifted A(H3N2) viruses and that protection was mediated by subtype-specific rather than cross-reactive immune responses.

Discussion
In the current study, adults who had undergone active investigation to detect influenza virus infections since December 2007 received inactivated influenza vaccine in November 2016, and antibody titers were assessed against A(H3N2) viruses spanning 1968-2018. Antibody titers against older strains were associated with year of birth, whereas titers against post-2007 strains were associated with recent A(H3N2) infection status. However, titers were also relatively high against 1993-2002 strains. This deviates from antigenic sin and seniority hypotheses, which suggest that strains encountered earlier in life are higher in the antibody hierarchy because later infections back-boost antibodies against earlier strains and/or because immune responses to earlier strains mitigate responses to later strains 6,23 . Antibody titers against the vaccine A(H3N2) strain, as well as recent past strains, rose substantially within 7 days of vaccination, indicating that memory B cells were recalled. In contrast, young children produce negligible antibody within 7 days of their earliest influenza infections 24 , and the antibody induced mostly targets the HA of the strain that caused infection 25 . Participants with an A(H3N2) virus infection during 9 years before vaccination had higher antibody titers and more persistent titer rise against the vaccine virus and future circulating viruses. Similarly, in this cohort, symptomatic A(H3N2) infections were predominantly detected among vaccinees who lacked prior A(H3N2) virus infection, indicating that both vaccine immunogenicity and effectiveness are enhanced by immunological memory associated with prior infection. The boosting effect of prior infection, observed here, contrasts with reports of negative effects of prior or repeated vaccination [7][8][9][10][11] , suggesting that the type of prior exposure is highly relevant. It was also notable that vaccine responses were at least as good among older compared to younger adults, contrasting with studies in more highly vaccinated populations 26,27 . The recommended annual interval between influenza vaccinations is typically shorter than that between influenza infections. However, titers and titer rises following vaccination tended to increase with proximity of prior infection, indicating that time between exposures does not directly account for the different effect of prior infection. Several groups have demonstrated that neutralizing antibodies can become focused on limited virus epitopes that have remained conserved across successively encountered strains 28,29 . It is hypothesized that recalled memory B cells dominate and focus responses on epitopes that are well conserved in successively encountered strains, which could either enhance or compromise protection depending upon whether these targeted epitopes undergo mutation in subsequent strains 13,14 . In the current study, the strain coverage of antibodies, and capacity to generate antibodies against a prominent site B epitope, were shaped by the prior infecting strain, consistent with memory B cell dominance. These findings present a paradox whereby memory B cell recall is pivotal for inactivated egg-based influenza vaccine to elicit sufficient antibody for protection but may also be problematic in terms of the capacity for vaccination to update immunity by generating memory B cells and antibodies against epitopes that have mutated in a new vaccine strain. To generate antibodies and memory B cells against variant epitopes, influenza vaccines must either induce memory B cells to undergo further affinity maturation 30 or induce naive B cell differentiation. Memory B cells may have a competitive advantage, because they have undergone affinity maturation and may compete more successfully for antigen to engage T cell help for further differentiation and they additionally less reliant than naive B cells on T cell help for activation 31,32 . Inactivated influenza vaccines deliver antigen transiently and induce minimal innate costimulation and hence may have little capacity to activate naive B cells and generate new B cell clones and antibodies in the presence of vaccine-reactive memory B cells.
Infection induced higher antibody titers against a broader antigenic range of A(H3N2) viruses than vaccination among individuals who developed A(H3N2) ILI in the season after vaccination. This suggests that infection may have greater potential to expand the antibody repertoire than vaccination. In turn, as the epitope range of the memory B cell pool increases, the potential to recognize epitopes in a new vaccine strain will also increase, providing a mechanism for the differential effects of prior infection and vaccination. Similarly, in ferrets and mice, priming with inactivated influenza vaccine induces little to no measurable antibody and no protection against variant virus strains, whereas priming by infection induces more antibody and substantial protection against variant strains 33,34 . These differences in antibody responses may reflect a greater capacity for influenza virus infection, as opposed to vaccination, to activate both the innate and adaptive immune systems 35 and in turn activate naive B cells. Additionally, antigen may be retained for longer periods after infection than vaccination and may be available to engage naive B cells after the memory B cell response starts to contract 36 .
The study has several limitations. The objective, to investigate the effect of prior A(H3N2) infection on vaccine immunogenicity against A(H3N2) viruses, required an observational design, and the sample size was constrained by the rarity of people who lacked A(H3N2) infection over a 9-year period. Therefore, inferences are suggestive rather than conclusive. It would be possible to perform larger studies looking at effects of infection in the prior season only; however, the results presented here indicate that it is important to consider infections over a number of years. Although we used seroconversion in addition to ILI surveillance to determine participant's prior infection status, some asymptomatic infections may be missed using a fourfold or greater titer rise as the criterion for seroconversion 37 . It was also clear that titers and titer rises were higher among participants with infection confirmed by RT-PCR than by Articles Nature MediciNe seroconversion, which could reflect the potential for serologically confirmed infections to be false positive 37 . Alternately, recent studies indicate that antibody titers can increase with severity of influenza infection 38 .
Taken together, the results of this study indicate that prior A(H3N2) virus infection may increase the titer and breadth of antibody responses induced by a new A(H3N2) vaccine strain, and thereby enhance protection despite antigenic drift. However, the range of strains against which antibodies are induced may be dictated by the prior infecting strain, consistent with a memory-dominated response. Such memory dominance may need to be overcome in future vaccine strategies to increase protection against drifted A(H3N2) viruses.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41591-022-01690-w.

Nature MediciNe
Methods Study design and participants. Participants were purposefully selected from a household-based influenza cohort established in Thanh Ha Commune, Thanh Liem District, Ha Nam Province, northern Vietnam (Extended Data Fig. 1). The cohort has been described previously 19 . In brief, the Ha Nam cohort enrolled 270 randomly selected households between 17 November and 7 December 2007. Household members were asked to participate in active ILI surveillance, provide swabs if they developed ILI and provide blood samples annually or biannually, at times spanning influenza transmission peaks 19 . Swabs were assessed by RT-PCR to detect influenza virus RNA, and sera were assessed to determine HI antibody titers against circulating strains. Infection was defined as detection of RT-PCR-confirmed ILI or a fourfold or greater rise in antibody titer (seroconversion).
The primary objective was to determine whether vaccine-induced antibody titer rises against A(H3N2) viruses differ between participants with and without documented prior A(H3N2) infection. Sample size was based on the assumption that GMRs of post-to prevaccination antibody titers would differ by 0.7 with a standard deviation of 1.0, giving an effect size of 0.7. It was estimated that 33 participants per group would provide 80% power to detect this effect with 95% confidence. This was inflated to 50 participants per group to account for loss to follow-up and facilitate comparison of people infected with A(H3N2) in different years since 2007.
Inclusion criteria were age ≥18 years and continued participation in cohort investigations to ascertain prior infection status. Participants with a history of allergic reactions were excluded. A total of 371 of 556 adults registered interest in participating in a vaccine study at the time of reconsent for the Ha Nam cohort in July 2016 (Extended Data Fig. 1), and 161 had participated continually in ILI surveillance and all blood sample collections, including 32 without and 129 with A(H3N2) infection since December 2007 (Extended Data Fig. 1). For each of the 32 participants without recent A(H3N2) infection, two or three participants with prior infection were selected based on proximity of their ages and sex to obtain a similar ratio of males to females (Supplementary Table 1). A total of 100 of 114 selected participants consented to the vaccine study between 1 October and 6 November 2016. This included 28 of 32 without recent infection and 72 of 82 with recent A(H3N2) infection. Ages and proportions female were similar among nonselected, selected and consenting participants (Extended Data Fig. 1). Selected participants were from 79 of 210 household remaining in the cohort, with three households contributing three participants each and 16 households contributing two participants each.
Participants received licensed, locally available Trivalent inactivated influenza vaccine (TIV; Vaxigrip, Sanofi Pasteur) in November 2016. Blood samples were collected before and 4, 7, 14, 21 and 280 days after vaccination. Blood samples were also collected 7 and 21 days after confirmed influenza illness occurring in the season after vaccination.
Study protocols were approved by ethics committees of the University of Melbourne (1646470), the National Institute of Hygiene and Epidemiology in Vietnam (IRB-VN01057 -08/2016) and the Oxford Tropical Medicine Research Unit . All participants provided written informed consent (conducted in Vietnamese). Participants were compensated financially for each investigation commensurate with the time required. The study was not prospectively registered as a clinical trial, because participants were not assigned to intervention versus control groups. However, study protocols were retrospectively included on the Australian New Zealand Clinical Trials Registry (12621000110886).
Virus propagation and characterization. Viruses were propagated in mammalian cell lines and/or in 10-to 12-day-old embryonated chicken eggs (Supplementary Table 2). Madin-Darby canine kidney (MDCK) cells and MDCK cells transfected with 2,6-sialtransferase (SIAT) were grown in DMEM (Gibco) containing penicillin/streptomycin and 10% fetal bovine serum (Bovagen). A number of viruses acquired neuraminidase (NA) substitutions, which that have been associated with erythrocyte agglutination via NA 39 , when propagated in MDCK cells (Supplementary Table 2). HA titers of most of these viruses decreased when oseltamivir was added to inhibit NA, but HI titers did not uniformly increase in the presence of oseltamivir. Therefore, viruses were plaque-selected on SIAT cells to produce stocks that lacked NA T148X or D151X substitutions and were more sensitive to detect HI antibodies (Supplementary Table 6 and Supplementary Fig. 4).
Reverse genetics viruses were produced using the eight plasmid system based on A/Puerto Rico/8/1934 (PR8) (ref. 40 ). The Y159S substitution was introduced into the HA of HK14e using the following primers: forward, 5′-CTTAAACAGCAAATACCCAGCATTGAACGTGACT-3′; reverse, 5′-TATTTGCTGTTTAAGTGGGTCAACCAATTT-3′. Wild-type and Y159S HA were cloned into the vector PHW2000 (ref. 40 ). The 7:1 reassortant viruses were generated using plasmids encoding PR8 internal and NA genes, and HA of HK14e or HK14e-Y159S. Plasmids were transfected into cocultured 293T/SIAT cells, and then recovered viruses were propagated in eggs. Reverse genetics viruses were assessed by HI assay using antisera raised against HK14e and Sw13e and a human mAb (Q129C) that recognizes site B of A/ Victoria/361/2011 (generously provided by A. Townsend, MRC Weatherall Institute of Molecular Medicine).
HA and NA genes of viruses used for serology and/or from swabs of Ha Nam cohort participants (isolates or clinical specimens) were sequenced via Sanger sequencing and aligned using the multiple alignment using fast Fourier transform algorithm in MegAlign Pro 13 (DNASTAR Lasergene 13). Phylogenetic trees were edited in FigTree version 1.4.4 (2006-2018, A. Rambaut, Institute of Evolutionary Biology, University of Edinburgh; http://tree.bio.ed.ac.uk/). HA antigenic site positions (Fig. 1c), defined by Lee et al. 17 , that varied between HK14e and at least one recent prior strain were tabulated to determine whether antigenic variation from HK14e was clustered within particular sites and if this varied between prior infecting strains (Fig. 1d).
Viruses circulating since 2007 were antigenically characterized by HI assay using ferret antisera generated for routine virus characterization by the WHO Collaborating Center for Reference and Research on Influenza, Melbourne (Supplementary Table 7). A two-dimensional map of virus antigenic distances was generated from the matrix of two-way titers of each sera against each virus using antigenic cartography software (Racmacs, https://acorg.github.io/Racmacs/). Serological assays. Sera were assessed by HI assay to determine antibody titers against influenza viruses. Assays were performed according to WHO Global Influenza Surveillance Network protocols 41 with the exception that volumes were reduced to 25 µl each of diluted sera, virus and 1% erythrocytes (0.33% final). Guinea pig erythrocytes were used for titration of antibodies against all A(H3N2) viruses, based on initial comparisons of titers obtained using guinea pig versus turkey erythrocytes ( Supplementary Fig. 5). Sera were treated with receptor destroying enzyme (Denka Sieken), adsorbed with 5% erythrocytes, and then tested over twofold serial dilutions from 1:10 to 1:10,240. Each individual's complete set of sera were tested against all viruses using the same batch of erythrocytes. Quality control viruses and sera were run with each new batch of samples/erythrocytes and were accepted if HA and HI titers were within twofold of initial values. HI titers were read using an automated reader (CypherOne, InDevR). Instrument settings for plate reading were determined by comparison with manual titer reads ( Supplementary Fig. 6) and then applied to all plates. Antibody titers against reverse engineered viruses were validated by MN assay, conducted according to World Health Organization protocols 41 using SIAT cells, and plasma treated as per the HI assay protocol above.
Outcomes. The primary outcome was vaccine immunogenicity, comparing GMRs of antibody titers among participants with and without recent A(H3N2) virus infection. GMTs and proportions seropositive (defined as a titer of 40 or more) or seroconverting (defined as a fourfold or greater titer rise) were also compared. The strain coverage of antibodies induced by vaccination was further compared by fitting antibody titer landscapes across all A(H3N2) viruses tested 21 . Titers were determined at a range of time points, but comparison focused on day 14 after vaccination, when titer peaks were detected, and on day 280, when titer decay plateaus 42 .
Post-hoc comparisons of participants who had been infected with viruses from distinct genetic clades, and participants who did or did not develop A(H3N2) ILI in the season after vaccination, were also performed.
Statistical analysis. GMTs and GMRs were estimated from log 2 HI titers and from differences of log 2 titers at post-minus prevaccination time points. GMTs and GMRs were calculated for individual viruses (n = 40) and groups of viruses representing prior exposure and postvaccination periods, which were averaged for each person. To estimate the size of the effect of recent infection on GMTs and GMRs, a mixed-effects linear regression model was used, which included a random effects term to account for within-person correlations of antibody titers over time and an interaction term for time of serum collection by recent infection status. Fisher's exact test was used to compare proportions with and without prior infection who seroconverted on day 14, maintained a fourfold titer rise on day 280 or became infected after vaccination.
To construct and compare antibody landscapes across strains, we used either GAMs or regression models with locally weighted scatterplot smoothing 21 to fit log 2 titers against A(H3N2) viruses organized temporally or antigenically, respectively. We used the GAM function from the R package mgcv and accounted for repeated measurements on each individual through specification of a random effect 43 . Plots were generated with ggplot2 (ref. 44 ). Previously published code, available in GitHub, was used to generate two-dimensional maps of virus antigenic distances (https://acorg.github.io/Racmacs, version 1.1.4) and fit antibody landscapes using the locally weighted scatterplot smoothing regression model (https://github.com/acorg/ablandscapes, version 1.0.2.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section. n/a Confirmed D � The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement D � A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly D � The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.

D �
A description of all covariates tested D � A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons D � A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection Data analysis Participant data was recorded into a secure, auditable online data bate called CliRes, developed by the Oxford University Clinical Research Unit, Viet Nam https://clires.oucru.org/. Serological data was linked to participant data using Microsoft Access Version 15.0.5349.1000.
Data was analysed in R using code that is available as part of existing packages (mgcv, ggplot2). R Code used to produce Figures has been uploaded to GitHub (https://github.com/afoxmarsh/FluVax_prior_infection_study). Previously published code, availablle in GitHub was used for antigenic cartography (https://acorg.github.io/Racmacs, version 1.1.4) and antibody landscapes (https://github.com/acorg/ablandscapes, version 1.0.2) For manuscripts utilizing custom algorithms or softwarehttps that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a communit y repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability The data set used for analysis (wide-format "HI_timecourse.csv", long format "HI_long_diff.csv") will be made available on request and will be publically available at https:

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
Sample size was based on a sample size calculation for the difference in geometric mean titre ratio (GMR, mean post -pre log 2 titer). We assumed that GMRs would differ by at least 0.7 between the groups with and without recent prior A(H3N2) infection, and that the standard deviation will be up to 1  It is not possible to replicate the longitudinal cohort. Antibody titres in pre-vaccine and d21 post-vaccine sera of all participants were measured against the A(H3N2) vaccine strain in two independent experiments by different researchers. HI titers against reverse engineered viruses were also measured in two independent experiments. Titers reproducibility was generally high (Supp Fig. 7) Participants were not randomly selected. Rather participants were purposefully selected based on their history of prior A(H3N2) virus infection. Participants who lacked any confirmed A(H3N2) infection during 9 prior years of cohort participation were all selected, then 82 participants who had an infection were selected based on proximity of age and sex distribution of the non-infected participants. The final sample included 28 participants without and 72 with prior A(H3N2) infection. Serology was performed using lab numbers to identify samples so that researchers were blinded to participant information.

I I
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response. This study was nested into the Ha Nam Cohort, an ongoing, prospective study of influenza infection within households. 100 Cohort Participants were included, and received influenza vaccine. 62 were female, 38 were male, and ages ranged from 20 to 81 years. A(H3N2) infection was detected during the preceding 9-years among 72 of 100. The median age of participants with and without prior A(H3N2) infection was the same (49Y), and 64% and 60% were female (Fig. 1a, Extended Fig.1) I Recruitment Ethics oversight commencement of the cohort, and who had no history of reactions to vaccination or condition that rendered them ineligible for influenza vaccine as listed in the summary of product characteristics were asked to participate. Participants who lacked A(H3N2) virus infection since 2007 were recruited first, then participants of similar age and gender distribution who had at least one A(H3N2) virus infection detected were recruited until the sample size was met. Participants were compensated financially for each investigation commensurate with the time required. The requirement to recruit participants purposefully from an existing cohort based on prior infection status, rather than randomly, and the rarity of people who lacked A(H3N2) infection over 9 years meant that the population was biased towards females and older adults. These factors limit interpretation, and inferences are suggestive rather than conclusive. The majority of selected participants consented, and age and sex were similar among selected versus consenting participants (Extended Data Fig 1) indicating that there was little self-selection bias. The study was approved by ethics committees of the University of Melbourne, the National Institute of Hygiene and Epidemiology in Viet Nam, and the Oxford Tropical Medicine Research Unit. All participants provided written informed consent. I Note that full information on the approval of the study protocol must also be provided in the manuscript.

Clinical data
Policy information about clinical studies All manuscripts should comply with the ICMJE guidelines for publication of clinical research and a completed CONSORT checklist must be included with all submissions.
Clinical trial registration ACTRN: ACTRN12621000110886 Study protocol Data collection

Outcomes
The study protocol is available on the Australian New Zealand Clinical Trials Registry ACTRN12621000110886 The vaccine study was conducted in Ha Nam, Viet Nam. Participant samples were processed at the National Institute of Hygiene and Epidemiology (NIHE), Viet Nam. Virus isolation and sequencing were performed at NIHE and at the WHO Collaborating Centre for Reference and Research in Influenza (WHOCCRRI) in Melbourne, Australia. Serology to detect infections during 9 years prior to the vaccine study was conducted at NIHE. Serology on vaccine sera was conducted at WHOCCRRI.
The primary outcome was vaccine immunogenicity, comparing geometric mean ratios (GM Rs) of antibody titres among participants with and without recent A(H3N2) virus infection. Geometric mean titres (GMTs), and proportions seropositive (defined as a titre of 40 or more) or seroconverting (defined as a four-fold or greater titre rise) were also compared. The strain-coverage of antibodies induced by vaccination was further compared by fitting antibody titre landscapes across all A(H3N2) viruses tested21. Titres were determined at a range of time points, but comparison focused on day 14 post-vaccination, when titre peaks were detected, and on day 280, when titre decay plateaus 42. Post hoc comparisons of participants who had been infected with viruses from distinct genetic clades, and of participants who did or did not develop A(H3N2) ILi in the season after vaccination were performed. Within participant comparisons of log2 transformed titres against revere genetics viruses bearing wild-type or Y159S HA were compared using paired t-Test. GM Rs of wild-type to Y159S virus titres were compared across groups infected with different prior strains were using non-paired t-Test following verification that data passed the Shapiro Wilks Normality test.