High Preexisting Serological Antibody Levels Correlate with Diversification of the Influenza Vaccine Response

ABSTRACT

Reactivation of memory B cells allows for a rapid and robust immune response upon challenge with the same antigen. Variant influenza virus strains generated through antigenic shift or drift are encountered multiple times over the lifetime of an individual. One might predict, then, that upon vaccination with the trivalent influenza vaccine across multiple years, the antibody response would become more and more dominant toward strains consistently present in the vaccine at the expense of more divergent strains. However, when we analyzed the vaccine-induced plasmablast, memory, and serological responses to the trivalent influenza vaccine between 2006 and 2013, we found that the B cell response was most robust against more divergent strains. Overall, the antibody response was highest when one or more strains contained in the vaccine varied from year to year. This suggests that in the broader immunological context of viral antigen exposure, the B cell response to variant influenza virus strains is not dictated by the composition of the memory B cell precursor pool. The outcome is instead a diversified B cell response.

IMPORTANCE Vaccine strategies are being designed to boost broadly reactive B cells present in the memory repertoire to provide universal protection to the influenza virus. It is important to understand how past exposure to influenza virus strains affects the response to subsequent immunizations. The viral epitopes targeted by B cells responding to the vaccine may be a direct reflection of the B cell memory specificities abundant in the preexisting immune repertoire, or other factors may influence the vaccine response. Here, we demonstrate that high preexisting serological antibody levels to a given influenza virus strain correlate with low production of antibody-secreting cells and memory B cells recognizing that strain upon revaccination. In contrast, introduction of antigenically novel strains generates a robust B cell response. Thus, both the preexisting memory B cell repertoire and serological antibody levels must be taken into consideration in predicting the quality of the B cell response to new prime-boost vaccine strategies.

INTRODUCTION

A primary immune response induced upon first exposure to a given antigen is characterized by a wave of low-affinity IgM B cells activated from the naive B cell pool with few to no mutations in their immunoglobulin (Ig) genes. This is followed by isotype-switched, higher-affinity B cells generated from germinal center reactions with greater numbers of genetic mutations (1, 2). Subsequent exposures, or secondary responses, are largely driven by activated and differentiated memory B cells and thus dominated by mutated, isotype-switched, medium- to high-affinity B cells (1, 3, 4). While this profound difference between the primary and secondary immune responses is relatively straightforward upon repeated exposure to the same antigen, it is less clear how the immune system responds to challenge with a varying antigen such as the influenza virus. Due to the evolving nature of the influenza virus, individuals are repeatedly exposed over their lifetimes to viral strains containing both novel and immune-experienced epitopes (5). In keeping with the very definition of immune memory, the memory response to conserved epitopes encountered before should dominate the response to novel epitopes introduced by mutations in divergent influenza virus strains. If the magnitude of the immune response is driven solely by the memory B cell repertoire, then one would expect these repeated exposures to influenza virus to progressively focus the B cell repertoire toward viral strain epitopes encountered multiple times.

Two virus types, influenza A and B viruses, circulate in the human population. Influenza A virus is further subdivided into different subtypes based on characterization of the two major envelope proteins, hemagglutinin (HA) and neuraminidase (NA). H1N1 and H3N2 are the two predominant influenza A virus subtypes currently circulating in the human population. Influenza B virus can be divided into two coevolved lineages, with one or the other being more dominant from year to year. The annual inactivated influenza trivalent vaccine (TIV) contains influenza A H1N1 and H3N2 and influenza B virus strains projected to be in circulation. Produced every year, the vaccine may contain the same strain(s) as in the previous year's vaccine or may include a more divergent strain in the case of antigenic drift away from previously circulating strains. This allows us to ask if from year to year the vaccine-induced B cell response is favored toward repeated or divergent strains. If the extent of the immune response is determined predominantly by the prevaccination levels of memory B cells specific to a given strain, then the immune response should be dominated by B cells recognizing a repeated strain at the expense of more variant strains. However, early influenza vaccine studies showed an inverse correlation between preexisting serum antibody (Ab) levels to a given influenza virus strain and subsequent serological Ab levels upon vaccination with that strain (6-8). One complication with these studies, however, was their reliance on serum Ab levels to determine the magnitude of the influenza virus strain-specific response. It is difficult to tease apart newly generated Abs from preexisting serum Abs. In addition, differentiating the quantity from the quality of the response and the level of strain-specific versus cross-reactive Abs is a challenge when the polyclonal mixture of Abs found in serum is considered. In a recent study, the magnitude of the overall vaccine-induced plasmablast response was shown to be lower in individuals vaccinated the previous year with TIV than in those who were not, and this was again linked to higher prevaccination serum titers (9). As novel prime-boost strategies and exposure to variant influenza virus antigens provide the most promising strategies being developed for generating a universally protective influenza vaccine (10, 11), it is important to understand how recent preexisting immunity biases a response to subtly or dramatically changing influenza virus strains. In this study, we set out to directly determine how the specificity of B cell responses in humans are biased by recent past exposures.

To clearly address how the immune system responds to antigenic boosts with homologous and heterologous strains, we looked directly at the Abs produced by B cells activated early in the immune response. To do so, we took advantage of the fact that 5 to 7 days after influenza vaccination, an expanded population of vaccine-induced plasmablasts appears in human peripheral blood (12). By analyzing the strain specificity and magnitude of this plasmablast population by enzyme-linked immunosorbent spot (ELISPOT) assay and monoclonal antibody production, we can directly trace at a single-cell level the effect of recent viral strain exposure on subsequent immune challenges with similar and more novel influenza virus strains.

We thus undertook a comprehensive longitudinal analysis of the plasmablast, memory, and serological responses upon vaccination in multiple individuals between 2006 and 2013. During this time, several changes were made in the viral strains included in the vaccine, both because of antigenic drift and shift of H1N1 and H3N2 and alternating predominance of the two influenza B virus lineages circulating from year to year. Taking our observations together, we found that repeated vaccination with the same three strains from year to year reduced the overall vaccine-induced B cell response. Upon introduction of a variant strain in the vaccine, the immune response was more robust and dominated by B cells recognizing the divergent vaccine strain. This allows for diversification of the humoral response, ensuring maintenance of an expansive B cell repertoire ready to respond to evolving pathogens.

MATERIALS AND METHODS

Influenza vaccine donors.

All individuals vaccinated between 2009 and 2013 were recruited at the University of Chicago as approved by the Institutional Review Board. These individuals were adults between 18 and 53 years of age and received the 2009 inactivated monovalent H1N1 vaccine and/or the yearly seasonal inactivated influenza vaccine (Fluzone; Sanofi Pasteur). Samples from 2006 to 2008 were historical from our previously published studies (12, 13).

Virus and recombinant HA.

Influenza virus was grown in eggs and purified using polyethylene glycol (PEG) virus precipitation (BioVision Research Products, CA) according to the manufacturer's instructions. Viral hemagglutination activity units (HAU) were measured by incubating serial dilutions of virus with 0.5% turkey red blood cells (Lampire Biological Laboratories); 1 HAU was determined as the minimum virus amount needed to induce red blood cell agglutination. A/Victoria/351/2011 and B/Massachusetts/2/2012-like recombinant HA (rHA) proteins were obtained from Protein Sciences Corporation (Meriden, Connecticut). All other rHA proteins were obtained from the NIAID BEI Resources Repository or Influenza Reagent Resource.

PBMC isolation and single-cell sorting.

Peripheral blood mononuclear cells (PBMCs) were obtained by centrifugation of whole blood through a Ficoll gradient and resuspension in phosphate-buffered saline (PBS)–0.2% bovine serum albumin (BSA) for staining and sorting, or cells were viably frozen at −80°C in fetal calf serum (FCS)–10% dimethyl sulfoxide (DMSO) for later use. At days 5 to 7 after vaccination, PBMCs were enriched for B cells by incubation of whole blood with a Rosette Sep B cell enrichment cocktail (Stem Cell Technologies), followed by centrifugation through a Ficoll gradient. PBMCs were then stained with fluorescently labeled Abs recognizing CD3 (7D6), CD19 (H1B19), CD27 (O323), CD38 (HIT2), and CD20 (2H7). CD19⁺ CD3⁻ CD27^hi CD38^hi plasmablasts (80 to 90% CD20^lo) were then bulk sorted using a FACSAria followed by single-cell sorting into 96-well plates containing a Tris buffer with RNase inhibitors (Promega). Anti-CD3 was purchased from Invitrogen Life Technologies; all other Abs were obtained from Biolegend. Immunoglobulin genes from each sorted B cell were then amplified and cloned into expression vectors for 293T transfection and protein expression to produce monoclonal Abs as described previously (12, 13). Serum samples were obtained by spinning down whole blood and collecting the supernatant. Serum was collected at day 14 and/or 21 (day 14/21) after vaccination. No difference in influenza virus-specific serological Ab levels was detected between these two time points.

Memory B cell activation.

Freshly isolated PBMCs or thawed viably frozen PBMCs containing 5 to 10% CD19⁺ B cells were resuspended at 3 × 10⁶ to 4 × 10⁶ cells/ml in complete medium (RPMI medium supplemented with 4 mM l-glutamine, 10% FCS, 1:1,000 penicillin-streptomycin [Pen/Strep], mM Na pyruvate 50 μM β-mercaptoethanol, 10 mM HEPES) containing 2 μg/ml CpG (InvivoGen), 1:10,000 Staphylococcus aureus Cowan I (Sigma-Aldrich) and 1:1,000 pokeweed mitogen (PWM; a kind gift from Shane Crotty, La Jolla Institute for Allergy and Immunology) for 5 days in 24-well plates with 1 × 10⁶ to 4 × 10⁶ cells/well. All activated cells were then pooled, washed three times with complete medium, and transferred to antigen-coated ELISPOT assay plates. All cells were 80 to 90% viable as measured by trypan blue staining before activation.

ELISPOT assays.

Filter plates (96-well; Millipore) were incubated overnight at 4°C or at 37°C for 2 h with anti-human IgG, IgA, and IgM (KPL) and a 1:100 dilution of inactivated influenza vaccine, 2 μg/ml rHA, or 16 to 32 HAU/well whole virus. After a blocking step with RPMI medium–10% FCS, freshly isolated PBMCs or activated memory cells washed three to five times were resuspended in medium (RPMI medium supplemented with 4 mM l-glutamine, 10 mM HEPES, 100 U/ml Pen/Step, 10% FCS), added to each plate, and serially diluted 2-fold down the plate. After an overnight incubation at 37°C, the plate was washed extensively with PBS and PBS–0.05% Tween, and antibody-secreting cells were detected with biotinylated anti-human IgG or IgA (Southern Biotech), followed by streptavidin-alkaline phosphatase (Southern Biotech) and developed with nitroblue tetrazolium/5-bromo-4-chloro-3-indolylphosphate (NBT/BCIP; Thermo Scientific). Spots were counted at a dilution point when 10 to 50 spots were visible in a single well to determine the number of total antibody-secreting cells or antigen-specific antibody-secreting cells.

ELISAs.

High-protein-binding microtiter plates (Costar) were coated overnight at 4°C with 8 HAU of whole virus/well or 2 μg/ml rHA. Each plate was washed with PBS–0.5% Tween 20 and blocked for 1 h at 37°C with PBS–20% FCS. Serum samples were serially diluted 3-fold seven times, starting at a 1:100 dilution, and incubated on an antigen-coated plate for 1 h at 37°C, followed by horseradish peroxidase (HRP)-conjugated goat anti-human IgG (Jackson ImmunoResearch) and development with Super AquaBlue enzyme-linked immunosorbent assay (ELISA) substrate (eBioscience). Absorbencies were measured at the optical density at 405 nm (OD₄₀₅). To standardize the results, all plates coated with the same virus strain were allowed to develop for the same amount of time. Serum antibody 50% effective concentration (EC₅₀) levels were determined by nonlinear regression analysis of the serially diluted serum OD values, as calculated by GraphPad Prism. The EC₅₀ takes into account both the concentration (maximum binding) and affinity of Abs (slope of the binding curve) and is therefore an accurate measure of total binding and generally correlates with the endpoint titer. To measure monoclonal antibody (MAb) binding, MAbs were diluted to a starting concentration of 10 μg/ml, followed by serial 3-fold dilutions down the plate. To standardize results, a control antibody was included on each plate, and the plate was allowed to develop until this control reached an OD of 3.0.

Hemagglutination inhibition (HAI).

Human serum was treated with receptor-destroying enzyme (RDE II) overnight at 37°C and heat inactivated at 56°C for 30 to 60 min before being diluted to a final dilution of 1:10 in PBS and then diluted 2-fold down a U-bottom 96-well plate. An equal volume of 8 HAU of live virus/well (final, 4 HAU) was added for a final volume of 50 μl/well. After a 15-min incubation at room temperature (RT), 50 μl of 0.5% turkey red blood cells was added, and 45 min later the minimum serum dilution which inhibited hemagglutination was recorded.

Statistical analysis.

A two-tailed unpaired nonparametric Mann-Whitney test and paired nonparametric Wilcoxon tests were performed using GraphPad Prism. Nonparametric correlation values (Spearman r) were also determined using GraphPad Prism.

RESULTS

First vaccination against influenza virus has the highest serological response.

If the magnitude of the immune response is driven solely by the memory B cell repertoire, then one would expect that adults vaccinated recently against the influenza virus would have a higher plasmablast response upon subsequent vaccination than those vaccinated for the first time. To test this, we vaccinated individuals who had received an influenza vaccine the year prior or who reported never having been vaccinated before. Interestingly, we observed by flow cytometry a higher overall CD19⁺ CD38^hi CD27^hi plasmablast response 7 days after vaccination in newly vaccinated donors (Fig. 1A). This was true when we combined the vaccine response from individuals vaccinated between 2010 and 2013 (Fig. 1A, left panel) and when we looked specifically at the vaccine response to the 2010-2011 TIV (Fig. 1A, right panel). In the 2009-2010 season, two vaccines were available: the seasonal TIV containing drifted variants of H1N1, H3N2, and influenza B virus and a monovalent vaccine that protected against the pandemic 2009 H1N1 strain (A/California/4/2009 [A/Cal]) (Table 1). When we measured the fold change in influenza virus-specific Abs in serum between day 0 and day 14/21 following vaccination with the 2010-2011 TIV, we found that individuals who had received the 2009-2010 TIV had a reduced serum Ab response to the H3N2 and B virus strain present in the vaccine compared to those who had not been vaccinated or had received only the monovalent pandemic H1N1 vaccine the prior year (Fig. 1B and Table 1). Similarly, fewer Abs recognizing H1N1 were produced in individuals who had received the monovalent pandemic H1N1 vaccine or both 2009-2010 vaccines (Fig. 1B). Receiving the monovalent pandemic H1N1 vaccine the year prior also correlated with a slight reduction in the 2010-2011 H3N2 response, likely due to generation of cross-reactive influenza A virus H1/H3 Abs by the H1N1 vaccine (Fig. 1B). Thus, individuals have a higher plasmablast and serological Ab response when they are vaccinated against influenza virus for the first time. It should be noted that even those who reported never having been vaccinated had been naturally exposed to influenza virus through the upper airway tract over the course of their lifetimes and had preexisting influenza virus-positive Abs in their sera (data not shown).

FIG 1.

Influenza vaccine-induced serological Ab response varies with vaccine history. (A) Percentage of peripheral blood CD19⁺ B cells at day 7 after vaccination that were CD27^hi CD38^hi plasmablasts as determined by flow cytometry. The left panel compares the percentage of plasmablasts of all donors vaccinated between the 2010-2011 and 2013-2014 influenza vaccine seasons who had or had not been vaccinated the year prior. The right panel compares the plasmablast response to the 2010-2011 vaccine only. (B) The serum antibody EC₅₀ to the three virus strains present in the 2010-2011 vaccine was determined on the day of vaccination (day 0) with the 2010-2011 TIV and at day 14/21 by ELISA. Each dot represents the fold change increase in the binding EC₅₀ to each strain between day 0 and day 14/21 in an individual donor. Donors are divided according which vaccine(s) they received the year prior (2009-2010 season). The line represents the median fold change within each vaccine group. (C) Fold change in serum Ab EC₅₀ between day 0 and day 14/21 in five individuals to the vaccinating H1N1, H3N2, or influenza B virus strain in the indicated year. Each line represents the yearly serological response to the vaccinating influenza strain of a given donor. None of the donors received the 2009-2010 TIV, while four of them (except 007) received the A/Cal monovalent vaccine. (D) Fold change in serum Ab EC₅₀ as described for panel C to the vaccinating H1N1, H3N2, or influenza B virus strain. As in the data shown in panel C, all 2010-2011 donors had received the A/Cal monovalent vaccine but not the 2009-2010 TIV. After the 2010-2011 season all donors had been vaccinated the year prior as well. Each dot represents the fold increase in serum Abs in each donor with the median indicated by the line. All serum Ab EC₅₀ data are the average from three independent experiments. Statistical analysis was determined using a Mann-Whitney test. Vacc, vaccine; Yam, Yamagata lineage; Vic, Victoria lineage; n.s., not significant.

TABLE 1.

Influenza virus strains in the yearly inactivated influenza vaccine^a

^aThe influenza A H1N1 virus strains in light red represent antigenically similar drifted variants; the A/California/4/2009 strain (dark red) was an antigenically novel shifted H1N1 strain. The influenza A H3N2 virus strains have remained relatively similar with only minor variations due to drift since the 2006-2007 season. For the influenza B virus, the Victoria (Vic) or Yamagata (Yam) lineage is indicated in parentheses.

We next wanted to know if the introduction of a more divergent influenza virus strain in the vaccine would also magnify the B cell response. Between the 2010-2011 and 2013-2014 vaccinating seasons, the H1N1 strain was always the same, and H3N2 strains varied little (Table 1). However, while the 2009-2010, 2010-2011, and 2011-2012 TIVs contained an influenza B virus strain from the Victoria lineage, the 2012-2013 and 2013-2014 TIVs included a strain from the Yamagata lineage (Table 1). As an initial test of whether shifting the vaccine influenza B virus strain between lineages changes the level of the B cell response, we followed the serological response to the TIV in five donors between 2010 and 2011 and between 2013 and 2014. All but subject 007 had received the 2009-2010 A/Cal monovalent vaccine. Subject 007 received no influenza vaccines in the 2009-2010 season. None of these five individuals had been vaccinated against the influenza B virus strain in the 2009-2010 season. Consistent with this, this cohort had a higher influenza B virus-specific response in the 2010-2011 than in 2011-2012 season upon revaccination with the same influenza B virus strain (Fig. 1C, left panel). However, upon vaccination in the 2012-2013 season, when a different influenza B virus lineage was introduced, the serum Ab response to this strain was higher in all five subjects and, again, decreased in four of five subjects in the 2013-2014 season upon revaccination with the same B virus lineage (Fig. 1C, left panel). Though this influenza B virus response pattern was fairly consistent across the five subjects, the differences did not reach statistical significance due to the small cohort size. To verify this observation in more people, we extended our serological analysis to 10 to 15 donors per year with the same vaccine history as the original 5 donors. These subjects received the Cal/09 pandemic H1N1 vaccine but not the 2009-2010 TIV and reported having been vaccinated the year prior in all other years. With this larger cohort we observed significant drops in the influenza B virus-specific serological responses upon revaccination with the same linage in sequential years (Fig. 1D, left panel). In contrast to the influenza B virus response, the magnitude of the serological response to H3N2 was a little higher in the 2010-2011 season in individuals not vaccinated the year prior against H3N2 but remained low in subsequent years (Fig. 1C and D, center panels). The response to the pandemic H1N1 was more variable, likely due to this shifted strain being antigenically novel, although it did not statistically vary from year-to-year upon repeated vaccination (Fig. 1C and D, right panels). Thus, vaccination with more divergent vaccine strains induces a stronger serological response than repeated vaccination with similar strains.

Vaccination against a divergent influenza virus strain magnifies the plasmablast response.

In order to understand if this difference in the serological response corresponds to differences in the plasmablast response, we looked, using ELISPOT assays, at the proportion of plasmablasts that were specific for H1N1, H3N2, and the influenza B virus strain in response to the 2012-2013 TIV. We first looked at the level of cross-reactivity of plasmablasts generated in the 2012-2013 season to the strains present in the 2011-2012 vaccine as a measure of antigenic variation between strains in the vaccine in these 2 years. We found that few of the plasmablasts generated in response to the 2012-2013 vaccine containing an influenza B virus Yamagata lineage strain cross-reacted with the Victoria lineage strain present in the 2011-2012 TIV (Fig. 2A, left panel), confirming that these influenza B virus strains are quite antigenically distinct. In contrast, the H3N2 plasmablast response to the 2012-2013 TIV was highly cross-reactive to the H3N2 strain present in the 2011-2012 vaccine, indicative of the antigenic similarity between the two strains (Fig. 2A, right panel).

FIG 2.

Alternating the influenza B virus lineage in the vaccine magnifies the plasmablast response. (A) An ELISPOT assay was performed on PBMCS at 7 days after influenza vaccination with the 2012-2013 TIV to determine the proportion of IgG⁺/IgA⁺ plasmablasts (PB) capable of binding each influenza virus strain. ELISPOT assay plates were coated with the given year's vaccine or influenza virus rHA and incubated with fresh PBMCs overnight, and spots were counted under each condition. Shown is the proportion of total vaccine-positive plasmablasts capable of binding rHA from the vaccinating 2012-2013 influenza B virus lineage (Yamagata) and H3N2 strain or the influenza B virus lineage (Victoria) and H3N2 strain present in the previous year's vaccine. Each line links the proportion of vaccine-positive plasmablasts that can bind rHA of each of the two strains in an individual. Statistical significance was determined using a paired Wilcoxon test. (B) Plasmablasts present in peripheral blood at day 7 after vaccination with the 2012-2013 or 2013-2014 TIV were tested for specificity to the vaccinating H1N1, H3N2, or influenza B virus rHA or whole virus by ELISPOT assay as described for panel A. Shown is the percentage of vaccine-positive antibody-secreting cells that were specific to the 2012-2013 vaccinating influenza B virus strain in each donor who had or had not been vaccinated in the 2011-2012 season or to the 2013-2014 vaccinating B strain after vaccination with the 2013-2014 TIV in donors who had also received the 2012-2013 TIV. (C) Proportion of vaccine-positive plasmablasts generated in response to the 2013-2014 TIV specific to the H1N1 (H1), H3N2 (H3), and influenza B virus strain in the vaccine as measured by ELISPOT assay. Each bar represents the response in one donor. (D) Percentage of CD19⁺ B cells that were CD27^hi CD38^hi plasmablasts as detected by flow cytometry in the peripheral blood 7 days after either the 2012-2013 or 2013-2014 TIV. Shown is the percent plasmablasts in each donor, with the median indicated by the horizontal line. Statistical significance was determined with a Mann-Whitney test. (E) The proportion of total antibody-secreting cells specific to the H1N1, H3N2, or B strain was determined by ELISPOT assay and multiplied by the percentage of total plasmablasts detected by flow cytometry to calculate the percentage of total B cells that were influenza virus strain-specific plasmablasts at day 7 after vaccination with the 2012-2013 or 2013-2014 TIV. Shown is the percentage of B cells in each donor specific for each strain, as indicated, with the median represented by the horizontal line. Statistical significance was determined with a Mann-Whitney test. (F and G) Immunoglobulin genes from single-cell sorted plasmablasts were cloned and expressed as MAbs and tested for binding to the vaccinating influenza virus strains by ELISA. Shown is the proportion of total influenza virus-specific MAbs able to bind the vaccinating influenza B virus strain as detected by ELISA in the years indicated. Each dot represents the proportion from a given donor. All donors in the 2007-2008 and 2008-2009 seasons were vaccinated the year prior as well. Shown is the percentage of influenza B virus-specific MAbs (Flu MAbs) comparing all donors as a group between years (F) or a paired comparison of the same individual across years in the subset of donors for whom we tracked the immune response over multiple years (G). Statistical significance was determined using a Mann-Whitney test (F) or paired Wilcoxon test (G).

If there is no strain bias to the response, 33% of the plasmablasts will recognize each of the three strains present in the vaccine. And, indeed, on average, in individuals who had not received the influenza vaccine the year prior, a third of the vaccine-specific plasmablasts bound the vaccinating influenza B virus strain in the 2012-2013 season (Fig. 2B). However, individuals who had been vaccinated in the 2011-2012 season against the Victoria lineage B virus strain had a biased vaccine-induced plasmablast response to the 2012-2013 Yamagata lineage B strain (Fig. 2B). Upon revaccination in the 2013-2014 season against the same influenza B virus lineage, the average proportion of plasmablasts recognizing the B strain decreased (Fig. 2B). As both influenza A and B virus strains varied little between 2012 and 2013 and between 2013 and 2014, the plasmablast response was fairly evenly distributed between the H1N1, H3N2, and B strain contained in the vaccine in the 2013-2014 season (Fig. 2C). In addition, the overall plasmablast burst at day 7 induced by vaccination in these individuals was higher in the 2012-2013 season upon introduction of a new influenza B virus lineage (Fig. 2D). All together, the percentage of B cells composed of influenza B virus-specific plasmablasts at day 7 after vaccination was significantly lower in the 2013-2014 than in the 2012-2013 season, while the plasmablast responses to the influenza A H1N1 and H3N2 virus strains were similar between the 2 years (Fig. 2E). Thus, both the proportion and number of influenza B virus-specific plasmablasts produced upon vaccination were higher when the influenza B virus lineage was alternated from year to year than when vaccination was with the same lineage in consecutive years.

To confirm that this phenomenon was not unique to the particular B strains present between 2010 and 2013, we also tested the strain specificity of monoclonal Abs (MAbs) generated from vaccine-induced plasmablasts in individuals vaccinated in the 2006-2007 through 2008-2009 seasons. During this time, the influenza B virus strain lineage included in the vaccine alternated from the Victoria lineage in the 2007-2008 vaccine to the Yamagata lineage in the 2008-2009 vaccine (Table 1). At day 7 postvaccination, plasmablasts were single-cell sorted, and immunoglobulin (Ig) gene transcripts were amplified by reverse transcription-PCR (RT-PCR) and sequenced. The Ig heavy and light chains were then cloned into mammalian expression vectors, and MAbs were purified from the supernatant of 293T cells transfected with these expression vectors as described previously (12, 13). Each monoclonal Ab therefore represents the Ab secreted by one plasmablast. Similar to what we observed by ELISPOT assay for the 2012-2013 and 2013-2014 vaccine responses, when we calculated the percentage of monoclonal Abs produced from each individual that bound the vaccinating influenza virus strains, we again saw a marked correlation between altering the influenza B virus strain lineage in the vaccine from year to year and a high proportion of vaccine-induced influenza B virus-specific plasmablasts (Fig. 2F). This was true both when we analyzed the plasmablast vaccine response as separate cohorts of donors each year and when we did a pairwise comparison of donors vaccinated in consecutive years (Fig. 2G). Having analyzed the vaccine-induced plasmablast response using two different methods and across several years with multiple influenza B virus strains of either the Victoria or Yamagata lineage, we conclude that alternating the vaccinating influenza B virus lineage in consecutive years leads to a biased response to the influenza B virus strain and an overall stronger plasmablast response.

Preexisting serological levels negatively correlate with the magnitude of the B cell response.

We then wanted to know if the difference in the plasmablast response a week after vaccination also reflected a difference in the vaccine-induced serum Ab induction and memory B cell response. In response to the 2012-2013 and 2013-2014 TIVs, there was a strong correlation between the proportion of plasmablasts at day 7 and IgG⁺/IgA⁺ memory cells at day 14/21 specific for each viral strain, as detected by ELISPOT assay (Fig. 3A). This included individuals who had or had not been vaccinated the year prior. In addition, the number of plasmablasts specific to a given influenza virus strain present at day 7 in the peripheral blood correlated well with the level of strain-specific serum Abs induced by vaccination, as measured by fold change in EC₅₀ values (Fig. 3B, top panels) and HAI titers (Fig. 3B, bottom panels). The plasmablast response was therefore a good reflection of both the serological and memory responses induced by viral strains present in the vaccine.

FIG 3.

High prevaccine serological Ab titers dampen the B cell response. (A and B) Correlation between the proportion of influenza virus-specific plasmablasts generated 7 days after vaccination with the 2012-2013 or 2013-2014 TIV and memory B cells specific for the indicated strain detected 12 to 14 days after vaccination (A) or the fold increase in serum Abs levels (upper panels) or fold change (FC) in serum HAI titer (lower panels) between day 0 and day 14/21 after vaccination (B). Plasmablasts specific to the indicated viral strain rHA or whole virus was determined by ELISPOT assay 7 days after vaccination by placing freshly isolated PBMCs on an antigen-coated ELISPOT assay plate overnight. To detect influenza virus specificity of memory B cells, PBMCs were incubated for 5 days with B cell mitogens to induce differentiation of memory B cells into antibody-secreting cells. Activated cells were then placed on an antigen-coated ELISPOT assay plate overnight. Shown is the percentage of total vaccine-positive IgG⁺ IgA⁺ antibody-secreting cells specific for the indicated influenza virus strain (A) or the total number of strain-specific antibody-secreting B cells per 1 × 10⁶ PBMCs (B). (C) Correlation between percentage of influenza virus strain-specific plasmablasts induced by vaccination and the percentage of memory cells (Mem) specific to that strain present just before vaccination determined as described for panel A. (D) Correlation between the total number of strain-specific antibody-secreting B cells per 1 × 10⁶ PBMCs and serum Ab EC₅₀ present on the day of vaccination (day 0). The Spearman r (rs) value for non-Gaussian distributions and corresponding P value were used to determine the degree of correlation for all analyses. d, day; PB, plasmablasts.

Taking these observations together, we show that upon yearly vaccination with the TIV, B cells specific for more variant strains are preferentially activated while revaccination with similar strains produces a relatively weak response. This would suggest that the level of memory B cell precursors specific for a given viral strain do not correlate with the plasmablast response upon revaccination. Indeed, we saw no correlation between viral strain-specific memory B cells detected on the day of vaccination and the subsequent vaccine-induced plasmablast response to the 2012-2013 and 2013-2014 vaccines for either H1N1 or influenza B virus though we did some see correlation with the H3N2 response (Fig. 3C). However, we found an inverse correlation in all three influenza virus types between the serological levels of a given influenza virus strain present on the day of vaccination and the magnitude of the subsequent vaccine-induced plasmablast response (Fig. 3D). Thus, preexisting serological Ab levels play a significant role in determining the magnitude of the secondary immune response to influenza vaccine.

DISCUSSION

The B cell response to influenza virus in adults is principally a secondary immune response mediated through activation of memory B cells (12, 14, 15). Classically, the secondary immune response is thought to be primarily focused on generating a targeted, high-affinity immune response. This is very effective upon reexposure with antigenically identical proteins. However, it is unclear how the preexisting memory repertoire influences the response to variant strains of influenza virus encountered throughout the lifetime of an individual. The observation that viral variants encountered early in life are more prevalent in serum led to the concept of original antigenic sin, i.e., that the immune response will be stronger to past antigen variants at the expense of the vaccinating variant (16). Studies in mouse models and humans provided some evidence for this theory (17-20). Somewhat contrary to the theory of original antigenic sin, however, were early influenza vaccine studies showing an inverse correlation between preexisting serum Ab levels to a given influenza virus strain and subsequent Ab production upon vaccination with that strain (6-8). In other words, recent exposure to a given influenza virus strain might inhibit subsequent Ab responses to that same strain. In addition, a recent longitudinal study showed progressive increases in Abs recognizing both initial pandemic strains and subsequent drifted variants, suggesting that higher levels of the first strains encountered in life may simply be a product of being boosted more often by heterologous viral strains (21). We therefore undertook to clarify the effect of B cell memory, both at the serological and cellular levels, on subsequent immune challenges with similar and divergent influenza virus strains by analyzing the specificity of vaccine-induced plasmablasts at the single-cell level.

Overall, we found that the preexisting serological level to a given influenza virus strain was the main predictor of the magnitude of the B cell response upon vaccination with that same strain. Lower preexisting serological Ab levels corresponded to a higher number of activated plasmablasts and subsequently larger numbers of vaccine-induced memory B cells and serum Ab levels. This negative correlation between preexisting serum Ab levels and the subsequent serological response upon revaccination was first observed many years ago (6–8), but at what level preexisting serum Abs affected the response was unclear. It was hypothesized that serum Abs might form immune complexes with vaccine-induced Abs and thus prevent detection of these new Abs (7, 22) or that the negative correlation simply represented a mathematical artifact (23). Sasaki et al. demonstrated that levels were lower not only of serological Ab but also of vaccine-induced plasmablasts in individuals with higher prevaccine serological Ab levels, suggesting that serum Abs might be blocking the general activation of B cells upon vaccination (9). We confirm that high preexisting serum Ab levels correlate with activation of fewer strain-specific plasmablasts and show further that this corresponds to fewer vaccine-induced memory B cells and lower serum titers. This would suggest that serum Abs bind and mask viral epitopes in the vaccine, thus preventing binding and activation of memory B cells recognizing that particular epitope. We also demonstrate that a direct outcome of this phenomenon is an uneven distribution of the B cell response to the three strains contained in the vaccine. If a divergent strain is introduced into the vaccine compared to the year prior, individuals vaccinated in both years have a higher response to the vaccine and to this variant strain in particular.

In contrast to serological levels, the proportion of preexisting memory B cells seemed to have little effect on the magnitude of the response to a particular influenza virus strain. We did observe that the vaccine-induced serum response to the pandemic 2009 H1N1 strain remained high in donors the first couple of years they were vaccinated against this strain, and only upon the third and fourth year did we observe a weaker response. We hypothesize that this may be a function of its antigenic novelty. The vaccine donors in this study were for the most part between 20 and 30 years of age, and it has been shown that individuals born after 1980 have little preexisting immunity to the antigenically shifted A/Cal H1N1 strain (24). Though this study looked at serum HAI and neutralization titers to A/Cal and not total serum binding as we do here, based on the antigenic disparity between recent seasonally drifted H1N1 strains and the pandemic H1N1 strain, it seems likely that memory precursors recognizing this shifted strain would indeed be lower. Thus, these younger individuals likely required multiple immunizations before generating an Ab serum response high enough to affect subsequent exposures to the same strain. Both influenza B virus lineages have cocirculated in the population for many years, and it is likely, in contrast to the pandemic H1N1 strain, that most individuals have significant prior immunity to both lineages, and so we did not observe this booster effect.

It is probable that the preexisting memory pool does play a role in modulating the particular viral epitopes targeted by vaccination with a heterologous variant, particularly one with few matching epitopes, as in the case of the 2009 pandemic H1N1 strain. A recent paper showed that individuals born between 1983 and 1996, when H1N1 strains contained a particular shared residue between the seasonal and 2009 pandemic H1N1 strain not present in the most recent drifted H1N1 strains, generated more Abs cross-reactive with that shared residue (25). However, in the case where there is a substantial presence of preexisting memory to a particular strain, the level of preexisting serum Abs, not the memory B cell pool, modulates the magnitude of the response upon revaccination. The overall effect is diversification of B cell memory toward new viral variants, ensuring a broad and diverse memory pool ready to respond to a wide range of influenza virus variants. In support of this, studies in ferrets showed that sequential priming with multiple variant H1N1 strains was far more effective at protecting against the antigenically novel A/Cal H1N1 strain than priming with one antigenically distant strain (26). This suggests that the memory repertoire is not simply composed of B cells reflecting the sum of influenza virus strains to which an individual has been exposed but, instead, the result of influenza virus responses tailored by the breadth of and sequence in which an individual has been exposed to strains over time. For this reason, though the overall size of the vaccine-induced B cell response is lower after multiple vaccinations, this in no way implies that they are not effective. A study in 1997 that enrolled close to 1,000 participants by the end of the study looked at the incidence of influenza virus infection in individuals between 1983 and 1988 who were unvaccinated, received their first vaccine, or were vaccinated in consecutive years (27). They concluded that repeated annual vaccinations were beneficial at reducing disease despite a lower serological response than that in those vaccinated for the first time. Our study demonstrates how serological Ab levels allow multiple exposures to various influenza virus strains to diversify the memory repertoire and overall provide improved protection.