Opposing Effects of Prior Infection versus Prior Vaccination on Vaccine Immunogenicity against Influenza A(H3N2) Viruses

Prior vaccination can alternately enhance or attenuate influenza vaccine immunogenicity and effectiveness. Analogously, we found that vaccine immunogenicity was enhanced by prior A(H3N2) virus infection among participants of the Ha Nam Cohort, Viet Nam, but was attenuated by prior vaccination among Australian Health Care Workers (HCWs) vaccinated in the same year. Here, we combined these studies to directly compare antibody titers against 35 A(H3N2) viruses spanning 1968–2018. Participants received licensed inactivated vaccines containing A/HongKong/4801/2014 (H3N2). The analysis was limited to participants aged 18–65 Y, and compared those exposed to A(H3N2) viruses circulating since 2009 by infection (Ha Nam) or vaccination (HCWs) to a reference group who had no recent A(H3N2) infection or vaccination (Ha Nam). Antibody responses were compared by fitting titer/titer-rise landscapes across strains, and by estimating titer ratios to the reference group of 2009–2018 viruses. Pre-vaccination, titers were lowest against 2009–2014 viruses among the reference (no recent exposure) group. Post-vaccination, titers were, on average, two-fold higher among participants with prior infection and two-fold lower among participants with 3–5 prior vaccinations compared to the reference group. Titer rise was negligible among participants with 3–5 prior vaccinations, poor among participants with 1–2 prior vaccinations, and equivalent or better among those with prior infection compared to the reference group. The enhancing effect of prior infection versus the incrementally attenuating effect of prior vaccinations suggests that these exposures may alternately promote and constrain the generation of memory that can be recalled by a new vaccine strain.


Introduction
Influenza viruses can evolve relatively rapidly because viral RNA replicates without proofreading. Substitutions that increase virus fitness or facilitate escape from host immune responses are positively selected. Influenza virus hemagglutinin (HA) mediates infection vaccine formulation, we also conducted a study among 157 Australian healthcare workers (HCWs), with varying prior exposures to vaccination, but unknown prior exposures to infection [43]. Vaccine-naïve HCWs and those who had been vaccinated only once before demonstrated greater post-vaccination responses against the A(H3N2) component of the vaccine than frequently vaccinated (3+ prior vaccines) HCWs [43]. Here, we have combined these studies to compare titers across 35 A(H3N2) antigens spanning 1968 to 2018 among participants with prior exposure through infection versus vaccination.

Vaccination Study Designs
The Ha Nam Cohort and the vaccination sub-study have been described previously [14,44]. In brief, 945 members of 270 households commenced participation in active surveillance for influenza infection causing RT-PCR-confirmed illness or seroconversion without illness in December 2007. In November 2016, 100 participants aged at least 18 Y, who had participated in all investigations to detect laboratory confirmed influenza infection received Southern Hemisphere trivalent inactivated split egg-grown influenza vaccine (Vaxigrip, Sanofi). This included 28 who lacked A(H3N2) virus infection since 2007, and 72 of similar sex and age who had at least one A(H3N2) virus infection. Venous blood was collected before and at multiple post vaccination time points, including days 21 and 280-282. The study was approved by ethics committees of the University of Melbourne (1646470), the National Institute of Hygiene and Epidemiology in Viet Nam (IRB-VN01057 -08/2016), and the Oxford Tropical Medicine Research Ethics Committee .
The HCW vaccine study has been described previously [43]. In brief, HCWs aged 18-65 years were recruited in April 2016 when attending the staff influenza vaccination clinic of the Royal Melbourne Hospital, Victoria, Australia to receive the 2016 Southern Hemisphere quadrivalent inactivated split egg-grown influenza vaccine (Fluarix Tetra 2016, Glaxo Smith Kline). Participants were asked if they had been vaccinated against influenza in the preceding 5 years since 2011. Venous blood was collected prior to vaccination and post vaccination on days 21-28 and 210-227. The study was approved by the human research ethics committee at the Royal Melbourne Hospital (reference number: HREC/16/MH/11).

Definition of Prior Exposure Groups, and Criteria for Inclusion in Each Group
Four groups were defined based on prior exposure by infection or vaccination to A(H3N2) viruses going back to A/Perth/16/2009-like viruses included in the 2011 vaccine (Table 1). These groups were referred to as "no recent exposure", "recent infection", "1-2 prior vaccinations" and "3-5 prior vaccinations". Ha Nam cohort participants were all vaccine-naïve, so were assigned to the "no recent exposure" group if A(H3N2) virus infection was not detected since December 2007, or to the "prior infection" group if infection was detected since 2009. HCWs were assigned to prior vaccination groups based on the number of times they were vaccinated in the preceding 5 years (HCWs) since 2011. HCWs who had no vaccinations in the preceding 5 years were excluded because it was not known if they had recent infection.
Inclusion and exclusion criteria for this analysis are shown in Figure 1. Participants who developed ILI with A(H3N2) infection confirmed by RT-PCR in the season after vaccination were excluded from both comparison groups because infection was expected a priori to lead to higher post-season titers, hindering interpretation. Ha Nam vaccinees were excluded if aged > 65 Y. Further, we excluded two participants who were last infected in 2008 with an A/Brisbane/10/2007-like (H3N2) virus in order to match the range of strains that HCWs were exposed to by vaccination between 2011 and 2015. HCWs were excluded if they did not complete the study, if vaccination histories were incomplete, or if insufficient sera remained for titration against 35 viruses. HCWs who had less than five prior vaccinations were under-represented, and all were selected. A similar number of HCWs with five prior vaccinations were selected randomly after stratifying by sex in order to obtain similar sex ratios to the other exposure groups.

Viruses
Viruses that circulated between 1968 and 2018 (n = 35) were propagated in mammalian cell lines and/or in eggs for use in serology. Where necessary, viruses were plaqueselected on Madin Darby Canine Kidney cells that had been transfected with human 2,6-sialtransferase (MDCK-SIAT cells) to produce stocks that lacked NA T148X or D151X substitutions, which can reduce the sensitivity of the assay for detecting HI antibodies [45]. Virus HA genes were sequenced and aligned with vaccine strains and strains circulating in the Ha Nam Cohort during the study period ( Figure S1). A(H3N2) strains included in vaccines between 2011-2016 (Table 1) were further scrutinized for substitutions at 131 amino acid positions that have been associated with antigenic variation and assigned to antigenic sites [46] ( Figure S2).

Serology
Sera from both studies were tested in hemagglutination inhibition (HI) assay against 35 A(H3N2) viruses (Table 1). Viruses were selected to represent each main antigenic or genetic cluster detected between 1968, when A(H3N2) emerged in humans, and 2018, four years after the strain included in the 2016 vaccine (A/Hong Kong/4801/2014). HI assays were performed according to WHO Global Influenza Surveillance Network protocols [47], with the exception that the red blood cell percentage and volume per well were 1% and 25 µL, respectively. Quality controls were performed for each new batch of red blood cells used to enable comparison of titers across multiple viruses and time points. HI titer reading was automated using a CypherOne reader (InDevR, Inc., Boulder, CO 80301).

Analysis
To compare antibody responses across strains, and between prior exposure groups, titers were presented as fitted landscapes across strains, and as averages for sets of strains, focusing on strains circulating since 2009. Generalized additive models (GAMs) were used to fit log 2 titers and log 2 titer differences (geometric ratios) between timepoints against 35 A(H3N2) viruses arranged temporally. Fitted titer estimates were presented with 95% confidence intervals. An antibody landscape package (https://github.com/acorg/ ablandscapes, accessed on 22 August 2021), which uses Lowess models, was used to fit log 2 titers against viruses arranged on a two-dimensional map of antigenic distances, and present estimated titers as contours [13]. Strains that circulated prior to a participants' birth year were excluded to reduce the impact of age on landscapes. Birth years ranged from 1950 to 1996, meaning that up to 10 strains circulating between 1968 to 1995 could have been excluded. We used the GAM function from the R package mgcv and accounted for repeated measurements on each individual through specification of a random effect [48]. To further compare effects of prior vaccination versus prior infection, the no prior exposure group was assigned as a reference, and the titer ratios were estimated for each prior exposure group compared to the reference. Ratios were estimated using the linear model function (titre_ratio~lm(log(titre)~exposure_group) of the tidyverse package in R. All plots were generated with the ggplot2 package in R [49].

Characteristics of Participants in Each Prior A(H3N2) Exposure Group
Eighty-six of 100 Ha Nam vaccinees, and 41 of 112 eligible HCW vaccinees were included in this analysis ( Figure 1). HCWs included 13/13 with 1-2 prior vaccinations; 8/8 with 3-4 prior vaccinations; and 20/91 with 5 prior vaccinations. These participants were classified according to exposure to A(H3N2) viruses by infection or vaccination since 2009 ( Figure 1, Table 2). Sex and age distributions, as well as sampling times, were similar across exposure groups (Table 2). Compared to participants with no recent exposure, participants with prior vaccination had higher pre-vaccination antibody titers against 2009-2014 strains, approximating strains present in prior vaccines (Figures 2A,B, S1a). Pre-vaccination titers were higher against a greater range of strains, extending back to circa 2002, among participants with prior infection (Figure 2A). Titers against strains circulating before 2000 were not consistently different between prior exposure groups ( Figure 2A). Post vaccination, a clear hierarchy emerged, with titers against strains spanning 2009-2018 being higher among participants with prior infection, and lower among those with 3-5 prior vaccinations compared to participants with no recent exposure ( Figure 2B,C). Participants with 1-2 prior vaccinations had intermediate post-vaccination antibody titers against recent strains, which did not exceed the titers of participants with no recent exposure. These trends reflected titer rise, which was negligible among participants with 3-5 prior vaccinations and relatively poor among participants with 1-2 prior vaccinations compared to those with no recent exposure ( Figure 2D).
Post-season titers remained relatively high among participants with prior infection, but not among those with prior vaccination compared to the group with no recent exposure ( Figure 2E). This reflected better maintenance of titer rise from pre-vaccination levels among participants with prior infection, but poorer maintenance among participants with prior vaccination, compared to those with no recent exposure ( Figure 2F). HCWs who developed ILI in the season after vaccination were excluded from these comparisons. However, separate analysis of this group showed that infection was substantially more immunogenic than vaccination, inducing titer rises exceeding four-fold and extending across a broad range of strains ( Figure S3A,B). The average titer rise against cell-grown viruses was notably poor post vaccination, but exceeded four-fold post season (after infection) ( Figure S3C,D), consistent with the antigenic differences between egg-grown vaccines and circulating strains and their cell-grown equivalents.
The impact of prior infection versus prior vaccination on vaccine responses was further examined by assigning the no recent exposure group as a reference and estimating titer ratios to reference for each prior exposure group. Pre-vaccination geometric mean titers (GMTs), shown for all groups, were generally lower among the no recent exposure group ( Figure 3A). Accordingly, pre-vaccination titer ratios to reference exceeded one for all viruses for the prior infection group, and for a narrower range of viruses for the prior vaccination groups ( Figure 3B). Post vaccination, titer ratios for most viruses were similar to pre-vaccination levels for the prior infection group, but had decreased for the prior vaccination groups, falling substantially below one for participants with 3-5 prior vaccinations ( Figure 3C). Titer ratios for A/Victoria/361/2011 (Vi11) were higher among groups with prior exposure at all time points, indicating that antibodies may be particularly well maintained and back-boosted against this virus. Post season, the prior infection group maintained high ratios, the ratios for the 3-5 prior vaccination group were still largely below 1, and the 1-2 prior vaccination group appeared to maintain titers no worse or better than the no recent exposure reference group ( Figure 3D).
The combined data indicate that the immunogenicity of the 2016 Southern Hemisphere vaccine against A(H3N2) viruses was enhanced by recent infection but attenuated by annual administration (3+ years) of vaccines containing strains similar to those causing infection.  Table 3. Results are presented as geometric mean titers (A) or as estimated mean ratios to reference (B-D) with 95% confidence intervals. The red shaded panel indicates the current vaccine strain.
a: Amino acids that differ from the prior vaccine strain are shaded yellow, while those that differ from the prevailing 2016 vaccine strain are colored red; + indicates introduction of a potential glycosylation site.

Investigation of Antigenic Site Substitutions across Vaccine Strains and the Potential for Antibody Focusing
The antibody focusing theory suggests that memory B cells against epitopes that are retained from past strains are recalled, and competitively dominate vaccine responses so that memory is not induced against variant epitopes [41].  (Table 3). Hence, the pool of vaccine-reactive memory may decline with each new vaccine strain encountered if memory is not induced against antigenic sites bearing substitutions. We investigated whether successive exposure to different vaccine strains affected antibody titer and titer rise distribution across a two-dimensional map of virus antigenic distances ( Figure 4). Titer/titer rise distribution differed between HCWs who had one versus two prior vaccinations, indicating that the pool of A/Hong Kong/4801/2014 (HK14) vaccinereactive memory may have been altered by exposure to A/Texas/50/12 (Tx12) prior to A/Switzerland/9715293/2013 (Sw13) vaccine. Titer distribution differed again among HCWs who had five prior vaccinations.

Discussion
The analysis presented here shows that the immunogenicity of inactivated split egggrown influenza vaccine against A(H3N2) viruses was enhanced by recent A(H3N2) virus infection but attenuated by recent vaccination. These findings are reminiscent of Hoskins' studies, as described in the introduction [16], and with animal models, which indicate that infection provides more potent priming than vaccination [50][51][52]. Additionally, a recent study from Japan found that the negative effect of prior vaccination on VE can be mitigated by prior infection [53]. Most repeat vaccination studies, including the current study, involve inactivated egg-grown virus vaccines. Inactivated cell-grown, recombinant HA, and adjuvanted vaccines have been developed in recent years, and early evaluations indicate that they are more immunogenic [54][55][56][57][58] and effective [59][60][61] than standard egggrown virus vaccines. It will be important to determine whether these vaccine formulations can overcome or alleviate the attenuating effects of repeated vaccination. While vaccine effectiveness against A(H3N2) viruses is poor and attenuated by prior vaccination, current estimates suggest that vaccination in the current and prior season affords better protection against A(H3N2) than being vaccinated in the prior season only [62]. This suggests that it is better to vaccinate annually than in alternate years. However, there have

Discussion
The analysis presented here shows that the immunogenicity of inactivated split egggrown influenza vaccine against A(H3N2) viruses was enhanced by recent A(H3N2) virus infection but attenuated by recent vaccination. These findings are reminiscent of Hoskins' studies, as described in the introduction [16], and with animal models, which indicate that infection provides more potent priming than vaccination [50][51][52]. Additionally, a recent study from Japan found that the negative effect of prior vaccination on VE can be mitigated by prior infection [53]. Most repeat vaccination studies, including the current study, involve inactivated egg-grown virus vaccines. Inactivated cell-grown, recombinant HA, and adjuvanted vaccines have been developed in recent years, and early evaluations indicate that they are more immunogenic [54][55][56][57][58] and effective [59][60][61] than standard egggrown virus vaccines. It will be important to determine whether these vaccine formulations can overcome or alleviate the attenuating effects of repeated vaccination. While vaccine effectiveness against A(H3N2) viruses is poor and attenuated by prior vaccination, current estimates suggest that vaccination in the current and prior season affords better protection against A(H3N2) than being vaccinated in the prior season only [62]. This suggests that it is better to vaccinate annually than in alternate years. However, there have been no formal comparisons of VE among people vaccinated in alternate versus successive years to determine whether there is any benefit to protection in the vaccinated years, and whether this outweighs the increased risk of infection in unvaccinated years. Additionally, the results presented here indicate that vaccine immunogenicity was particularly poor in people who had been vaccinated in three to five, as opposed to one to two, prior years, indicating a need to determine the impact of multiple years of vaccination on VE.
The findings presented here suggest that existing memory (B cell) responses induced by infection enhance vaccine immunogenicity, and in turn, that repeatedly vaccinated individuals lack memory responses that cross-react with the vaccine antigen. Several studies have indicated that, similar to antibody responses, B cell responses decline with repeated annual vaccination [21,22,63]. Some suggest that existing antibodies may sequester (mask or clear) antigens and thereby dampen adaptive immune cell activation [40,63]. However, poor vaccination responses were not associated with high pre-vaccination antibody titers in the present study. Recent clinical trials demonstrate that inactivated influenza vaccines are immunogenic and effective in previously-naïve infants [64][65][66][67], indicating that these vaccines can prime naïve B cells. However, memory B cells are intrinsically programed to out-compete naïve B cells upon B cell receptor engagement [68][69][70], and it has been proposed that this could account for the observation that antibody responses to A(H1N1pdm09) vaccines were highly focused on epitopes shared with previously encountered A(H1N1) viruses in selected age groups [41,[71][72][73]. Similarly, it was speculated that vaccination in the 2018/19 season was less effective against clade 3c3a A(H3N2) viruses in persons born before 1983 because HA S159T/Y substitutions emerged in 1983 and have been retained in subsequent strains, except those belonging to clade 3c3a [32,35]. It is plausible that responses diminish with repeated vaccination due to antigenic change across vaccine strains, in combination with competitive memory dominance. While memory may become limited against variant-neutralizing antibody epitopes with repeated vaccination, memory against more conserved non-neutralizing epitopes could be a source of substantial competition [74,75].
Direct comparisons of immune responses induced by infection and vaccination indicate that infection is more immunogenic. Specifically, A(H3N2) infection has been shown to induce more cross-reactive and less clonally expanded anti-hemagglutinin antibodies than influenza vaccination [75,76]. Infection also induces higher frequencies of CD4+ T cells with increased functional capacity than are induced by inactivated vaccine [77]. While the broader back-boosting capacity of infection has been interpreted to indicate that infection is a more potent driver of original antigenic sin than vaccination [75,78], this does not translate to antibody responses. It has been reported that infection promotes the maturation of memory B cell affinity against the HA head of infecting A(H3N2) strains [78]. Adjuvants increase the capacity of inactivated vaccine to induce naïve B cells and reduce the effects of prior vaccinations [79]. It is highly plausible that infection will induce greater naïve B cell differentiation than vaccination with inactivated virus because of greater innate immune stimulation [80] and antigen retention so that naïve B cells can be engaged after the memory B cell response starts to contract [81].
This study had several limitations. Sample sizes were small, particularly for HCWs who had less than five prior vaccinations. The effects of prior infection versus vaccination reported here could be biased by ethnic and socioeconomic differences between community members from Viet Nam and HCWs from Australia. However, we could detect effects of vaccination history within the HCW cohort, consistent with our previous studies [25], and with studies showing effects on VE in various populations and countries [17,18,53]. HCW infection histories were unknown, so we were unable to account for effects of prior infections, which may have alleviated the attenuating effects of prior vaccination [53]. These limitations could be remedied by the longitudinal study of other populations who have more varied uptake of the influenza vaccine, and who are monitored for clinical and sub-clinical influenza infection.

Conclusions
In conclusion, the results presented here indicate that recall of existing memory can enhance antibody titers induced by inactivated egg-based influenza vaccines, but may concurrently limit the generation of memory against variant epitopes of vaccines, accounting for the detrimental effect of repeated vaccination. Direct examination of how exposure history affects the magnitude and repertoire of memory B cells induced by different vaccine formulations will be important to inform the development of vaccines that provide better protection against A(H3N2) viruses.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/v14030470/s1, Figure S1: Phylogenetic tree of HA genes of A(H3N2) strains that participants were exposed to by infection or vaccination, and that were used to detect serum antibodies by HI assay, Figure S2 Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Data Availability Statement: HI titre data of participants who have consented to use of samples for future research will be made available on request and will be publically available for research that adheres to the original ethics approval conditions at https://melbourne.figshare.com/ (accessed on 22 August 2021) within one year of this publication. HA (+/− NA) sequences of influenza viruses included in the analyses are available on GISAID.
Acknowledgments: We thank Elizabeth Matchett and Kathryn Williams from the Infection Prevention and Surveillance Service of Royal Melbourne Hospital for their assistance with study logistics. Staff of Melbourne Health and people of Thanh Ha Commune have generously given their time and donated study samples. We are grateful to the Ha Nam Preventive Medicine Centre and People's Committees of Thanh Liem District for their support. We would like to thank the Thanh Ha Commune health workers for their dedication to conducting active surveillance and cross-sectional surveys. We also wish to thank the Ministry of Health of Vietnam for their continuing support of the research collaboration between the Oxford University Clinical Research Unit and the National Institute for Hygiene and Epidemiology. We are grateful to members of the Oxford University Clinical Research Unit including Peter Horby for his role in establishing the Ha Nam Cohort; Heiman Wertheim for management support of the Ha Nam Cohort during his period as Director of OUCRU Ha Noi; Nghiem Nguyen Minh Trang for project coordination; and Bui Huyen Trang for administrative support. Ongoing support of the Ha Nam Cohort has also possible due to the efforts of Directors and Division Heads at the National Institute for Hygiene and Epidemiology in Viet Nam including Dang Duc Anh, Pham Quang Thai and Tran Nhu Duong. Heidi Peck, Yi-Mo Deng, and their staff at the WHO Collaborating Centre for Reference and Research in Melbourne performed initial isolation and characterization of many of the Influenza viruses used.