Broad Hemagglutinin-Specific Memory B Cell Expansion by Seasonal Influenza Virus Infection Reflects Early-Life Imprinting and Adaptation to the Infecting Virus

Rapid and vigorous virus-specific antibody responses to influenza virus infection and vaccination result from activation of preexisting virus-specific memory B cells (MBCs). Understanding the effects of different forms of influenza virus exposure on MBC populations is therefore an important guide to the development of effective immunization strategies. We demonstrate that exposure to the influenza hemagglutinin via natural infection enhances broad protection through expansion of hemagglutinin-reactive MBC populations that recognize head and stalk regions of the molecule. Notably, we show that hemagglutinin-reactive MBC expansion reflects imprinting by early-life infection and that this might apply to stalk-reactive, as well as to head-reactive, MBCs. Our findings provide experimental support for the role of MBCs in maintaining imprinting effects and suggest a mechanism by which imprinting might confer heterosubtypic protection against avian influenza viruses. It will be important to compare our findings to the situation after influenza vaccination.

ABSTRACT

Memory B cells (MBCs) are key determinants of the B cell response to influenza virus infection and vaccination, but the effect of different forms of influenza antigen exposure on MBC populations has received little attention. We analyzed peripheral blood mononuclear cells and plasma collected following human H3N2 influenza infection to investigate the relationship between hemagglutinin-specific antibody production and changes in the size and character of hemagglutinin-reactive MBC populations. Infection produced increased concentrations of plasma IgG reactive to the H3 head of the infecting virus, to the conserved stalk, and to a broad chronological range of H3s consistent with original antigenic sin responses. H3-reactive IgG MBC expansion after infection included reactivity to head and stalk domains. Notably, expansion of H3 head-reactive MBC populations was particularly broad and reflected original antigenic sin patterns of IgG production. Findings also suggest that early-life H3N2 infection “imprints” for strong H3 stalk-specific MBC expansion. Despite the breadth of MBC expansion, the MBC response included an increase in affinity for the H3 head of the infecting virus. Overall, our findings indicate that H3-reactive MBC expansion following H3N2 infection is consistent with maintenance of response patterns established early in life, but nevertheless includes MBC adaptation to the infecting virus.

IMPORTANCE Rapid and vigorous virus-specific antibody responses to influenza virus infection and vaccination result from activation of preexisting virus-specific memory B cells (MBCs). Understanding the effects of different forms of influenza virus exposure on MBC populations is therefore an important guide to the development of effective immunization strategies. We demonstrate that exposure to the influenza hemagglutinin via natural infection enhances broad protection through expansion of hemagglutinin-reactive MBC populations that recognize head and stalk regions of the molecule. Notably, we show that hemagglutinin-reactive MBC expansion reflects imprinting by early-life infection and that this might apply to stalk-reactive, as well as to head-reactive, MBCs. Our findings provide experimental support for the role of MBCs in maintaining imprinting effects and suggest a mechanism by which imprinting might confer heterosubtypic protection against avian influenza viruses. It will be important to compare our findings to the situation after influenza vaccination.

INTRODUCTION

The B cell response to influenza A virus (IAV) infection or vaccination generates antibodies (Abs) and memory B cells (MBCs) reactive with the surface hemagglutinin (HA) of the virus, the key molecular target of efficiently neutralizing Abs (1). HA-specific MBCs dictate the strength and character of subsequent HA-specific Ab responses and are key determinants of immunity to influenza (2–4). Studies of responses to noninfluenza antigens indicate that MBCs are more broadly reactive than are long-lived plasma cells generated by exposure to the same antigen (5, 6). Thus, MBCs extend the breadth of protection provided by circulating Abs.

An increasingly comprehensive picture of HA-reactive MBC differentiation pathways and population dynamics after antigen exposure has emerged (7–11). However, much less attention has been paid to the fine specificity and breadth of reactivity of expanded HA-reactive MBC populations and how this relates to the form of antigen exposure. Recent studies of HA-specific Ab and MBC responses to inactivated influenza vaccines (IIVs) have distinguished between reactivity to the immunodominant and more variable head domain and the conserved stalk domain. Ellebedy et al. (12) demonstrated that IgG MBC expansion after seasonal IIV was much larger for head-specific than for stalk-specific MBCs, reflecting an Ab response that was mostly against the HA head. The Ab response to HA includes a much stronger antistalk component when humans are vaccinated with relatively novel HAs, such as HAs of the 2009 pandemic IAV or emerging avian IAVs (12–14). This reflects greater activation of MBCs specific for conserved stalk epitopes, since they are not outcompeted by more abundant head-reactive MBCs for limited antigen resources (9). Halliley et al. (15) demonstrated that strong HA stalk-specific Ab production in humans given an H7 IIV was associated with marked expansion of stalk-specific IgG MBCs. Overall, the impression from limited human vaccine studies is that MBC expansion reflects the specificity and magnitude of Ab production.

The HA-specific Ab response to clinical IAV infection in humans has been well characterized, but little is known about associated changes in MBC populations. Based on Ab titers measured by hemagglutination inhibition (HAI) assay, which detects Abs that bind in the vicinity of the receptor binding domain on the HA head, IAV infection typically generates an Ab response that is broadly HA cross-reactive (16). The pattern of cross-reactive HA binding frequently reflects preferential boosting of Abs against sets of HAs related to those of viruses that caused significant early-life infection, a phenomenon referred to as “original antigenic sin (OAS)” (17–19) or “antigenic seniority” (20). OAS is explained by the competitive dominance of MBCs reactive to conserved epitopes on the HA head domain (3, 21), but experimental support at the level of MBC population dynamics is lacking. In addition to OAS-type Ab production, IAV infection generally induces increased Ab titers against HAs of the infecting virus and related strains (16, 22–24), but it is unclear whether MBC adaptation to a novel HA occurs.

IAV infection in humans also generates Abs that bind to immunosubdominant epitopes in the HA stalk (11, 25). Because of high conservation of the HA stalk, especially within the group 1 and group 2 phylogenetic categories of HA, anti-stalk Abs have broadly protective potential both within and across HA subtypes (26, 27). Notably, anti-stalk Abs are generated by seasonal IAV infection, but few are produced by seasonal IIV, suggesting that infection is a key source of circulating stalk-specific Abs (25, 28). Expansion of stalk-specific MBCs by infection may be even more important for long-term broad protection if circulating stalk-specific Ab levels wane. However, the effect of infection on stalk-specific MBC populations has not been examined.

Our goal was to investigate the relationship between HA-specific Ab production and changes in the size and character of HA-specific MBC populations following human infection with seasonal H3N2 IAV. In particular, we were interested in the breadth of HA reactivity of the Ab response and whether the associated patterns of MBC expansion could provide a basis for similarly broad Ab production after future exposures to related HAs. We show that (i) H3 head-reactive MBC expansion is very broad and closely reflects patterns of H3-specific Ab production, including OAS responses, (ii) H3 stalk-specific MBC expansion accompanies stalk-specific Ab production, and (iii) MBC adaptation to the HA of the infecting virus occurs together with broad H3-reactive MBC expansion. Findings also raise the possibility that expansion of group 2 HA stalk-reactive MBCs after H3N2 infection reflects early-life H3N2 infection, suggesting a mechanism for the greater resistance of H3N2-imprinted individuals to avian H7N9 infection (29).

RESULTS

Subject characteristics.

The subject cohort consisted of 8 individuals who presented at the University of Rochester Vaccine Research Unit with medically unattended influenza-like illness in the 2012–13 season and had H3N2 IAV infection confirmed by reverse transcription-PCR (RT-PCR) (Table 1). Plasma and peripheral blood mononuclear cells (PBMCs) were collected on the day of presentation (designated day 0) and on days 3, 10, and 28 thereafter. All subjects had received the 2012–13 trivalent IIV composed of the A/California/7/2009 (H1N1), A/Victoria/361/2011 (H3N2), and B/Wisconsin/1/2010-like (Yamagata lineage) viruses. H3s carried by viruses causing human infection in 2012–13 frequently possessed genetic changes compared to the H3 of A/Victoria/361/2011 (Vic11; see Table 2 for virus abbreviations), but the overall antigenic relatedness with H3 Vic11 was strong (22). HAI activity against the vaccine H3N2 virus was present in all subjects on day 0 and was increased ≥ 4-fold (indicating clinically significant seroconversion) in five of eight subjects on day 28 (Table 1).

TABLE 1.

Subject details, clinical features, and HAI titers

Subject no.	Birth yr	Age (yr)	Sex	Symptom duration (days)a	Feverb	Prolonged PCR positivityc	HAI titerd
							Day 0	Day 28
55	1960	53	F	1	N	Y	80	320
56	1956	56	F	2	Y	Y	80	320
58	1990	23	F	1	N	N	160	320
65	1992	21	M	1	N	N	320	640
72	1968	45	M	2	Y	N	40	160
73	1972	41	F	1	N	N	80	80
90	1949	64	F	3	Y	N	80	1280
91	1973	40	M	2	N	Y	80	320

^aPrior to presentation.

^bOn the day of presentation.

^cSubjects determined to be positive for virus by RT-PCR on day 10 after presentation, but infectious virus could not be isolated.

^dMeasured against the 2012–13 vaccine strain A/Victoria/361/2011 (H3N2).

TABLE 2.

Recombinant HA proteins used in this study

HA reagent	Abbreviation	Straina	Subtype(s)
H1	PR34	A/Puerto Rico/8/1934*	H1N1
	USSR77	A/USSR/1977*	H1N1
	Tex91	A/Texas/36/1991*	H1N1
	NC99	A/New Caledonia/20/1999*	H1N1
	SC18	A/South Carolina/01/1918*	H1N1
	Cal09	A/California/04/2009*	H1N1
H3	HK68	A/Hong Kong/1/1968*	H3N2
	PC73	A/Port Chalmers/1/1973*	H3N2
	Ala81	A/Alabama/1/1981*	H3N2
	Phil82	A/Philippines/2/1982*	H3N2
	Pan99	A/Panama/2007/1999*	H3N2
	Wyo03	A/Wyoming/3/2003*	H3N2
	Hiro05	A/Hiroshima/52/2005*	H3N2
	Wis05	A/Wisconsin/67/2005*	H3N2
	Per09	A/Perth/16/2009*	H3N2
	Vic11	A/Victoria/361/2011*	H3N2
	Tex12	A/Texas/50/2012*	H3N2
	Switz13	A/Switzerland/9715293/2013*	H3N2
H5	Indo05	A/Indonesia/05/2005	H5N1
	AH05	A/Anhui/1/2005	H5N1
H7	maNeth00	A/mallard/Netherlands/12/2000	H7N3
	SH13	A/Shanghai/1/2013	H7N9
Influenza B HA	B Bris08	B/Brisbane/60/2008*
	B Mass12	B/Massachusetts/2/2012*
Chimeric HA	cH4/7-SH13	A/duck/Czech/1956; SH13	H4N6, H7N9
	cH5/3-Per09	A/Vietnam/1203/2004; Per09	H5N1, H3N2
	cH6/1-Cal09	A/mallard/Sweden/81/2002; Cal09	H6N1, H1N1
HA head trimer		H5 head (A/Indonesia/05/2005)	H5N1
		H7 head (A/Shanghai/1/2013)	H7N9
HA1 (Vic11)		A/Victoria/361/2011	H3N2
HA1 (Switz13)		A/Switzerland/9715293/2013	H3N2
HA1 (Cal09)		A/California/04/2009	H1N1

^a*, Multiplex assay panel.

HA-specific IgG production following H3N2 infection.

Since our goal was to relate HA-specific Ab and MBC responses, we comprehensively evaluated H3-specific IgG production using approaches that measured HA head versus stalk reactivity and the breadth of binding to HA variants and subtypes.

(i) H3N2 infection generates H3 head- and stalk-specific IgG.

The Ab response to infection generates circulating Ab-secreting plasmablasts (PBs) that initially reflects activation of preexisting MBCs (11, 30). IgG PBs specific for H3 Vic11 were present in all subjects on day 3 and/or day 10 postpresentation with influenza-like illness (Fig. 1A). H3 stalk-specific IgG PBs were detected in four of eight subjects in whom they formed 20% or more of H3-specific IgG PBs. IgG PBs specific for H1 Cal09 were detected at low frequencies in some subjects, suggesting cross-reactive Ab induction. HA-specific PBs were also enumerated by flow cytometry as CD19⁺ CD71⁺ IgD⁻ CD20⁻ CD38^hi cells that bound an H3 Vic11 probe (see Fig. 4C and D). The results were consistent with PB measurement by enzyme-linked immunospot (ELISpot) assay and identified a transient peak of H3-specific PBs within 2 weeks of presentation.

FIG 1.

HA-specific Ab induction by H3N2 infection. (A) Frequencies of circulating IgG-secreting plasmablasts (PB) enumerated by direct ELISpot analysis of PBMCs. PB specificity was assessed against H3 Vic11, the chimeric HA cH5/3 to identify H3 stalk reactivity, and H1 Cal09. The dotted line identifies the limit of PB detection. (B) Plasma IgG concentrations measured by ELISA against the above antigens, as well as against the HA1 domains of H3 Vic11 and H3 Switz13, the chimeric HA cH6/1 to identify H1 stalk reactivity, the HA of IBV Mass12, and tetanus toxoid (TTd). (C) H3 stalk-specific IgG in plasma as a percentage of IgG binding to the H3 of Vic11. (D) Neutralizing Ab levels in plasma measured by microneutralization assay against H3N2 Vic11 and H1N1 Cal09 viruses. (E) Neutralizing anti-stalk Ab levels. IgG was purified and concentrated 10-fold and analyzed by microneutralization assay against the cH14/3N3 virus. In all panels, values are shown for individual subjects (n = 8). Significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001; NS [not significant]) was determined by repeated ANOVA with adjusted P values.

FIG 4.

HA-specific MBC expansion by H3N2 infection. (A) MBC expansion represented as the frequency of MBC-derived Ab-secreting cells (MASCs). PBMCs were stimulated in vitro to induce MBC differentiation into Ab-secreting cells, followed by enumeration of antigen-specific IgG-secreting cells by ELISpot assay. Ab-secreting cell specificity was assessed against H3 Vic11, the chimeric HA cH5/3, and H1 Cal09. The dotted line identifies the limit of MASC detection. (B) H3 stalk-specific MBCs as a percentage of H3 Vic11-specific MBCs. Percentages were calculated from IgG MASC frequencies. (C) Gating strategy for identification of H3-specific plasmablast (PB), activated B cell (ABC), and MBC populations by flow cytometry. PBMCs were stained with an H3 Vic11 probe and for expression of CD3, CD19, CD71, IgD, CD20, and CD38. Gated CD19⁺ cells were analyzed to identify the recently proliferated (CD71⁺) PB (CD19⁺ CD71⁺ IgD⁻ CD20⁻ CD38^hi) and ABC (CD19⁺ CD71⁺ IgD⁻ CD38⁻ CD20⁺) subsets and the resting MBC (CD19⁺ CD71⁻ IgD⁻ CD20⁺) subset. Analysis for probe⁺ PBs, ABCs, and MBCs is shown for day 3 (peak PB frequencies) and day 28 (increased ABC and MBC frequencies). (D) Frequencies of H3⁺ PBs, ABCs, and MBCs determined by flow cytometry. Baseline frequencies (mean and range) in PBMCs from healthy adult donors (n = 5) were 0.6 (0 to 2) for H3⁺ PBs, 10.4 (6 to 14) for H3⁺ ABCs, and 15.6 (5 to 22) for H3⁺ MBCs. In panels A, B, and D, values are shown for individual subjects (n = 8). Significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001; NS [not significant]) was determined by repeated ANOVA with adjusted P values.

Enzyme-linked immunosorbent assay (ELISA) measurements of plasma IgG specific for H3 Vic11, the HA1 domain of H3 Vic11, and the H3 stalk increased significantly from days 0 to 10 and then remained stable or slightly decreased to day 28 (Fig. 1B). H3 stalk-specific IgG as a percentage of the H3 Vic11-specific IgG concentration remained relatively constant in most subjects over the course of the response (Fig. 1C). However, the percentage ranged from approximately 8% to >20% across subjects, indicating subject-to-subject differences in the induction of HA head- versus stalk-specific Abs. The increase in IgG specific for H3 Vic11 HA1 was similar to that against the H3 HA1 of Switz13, an emerging H3N2 virus with a variant HA, indicating strong Ab cross-reactivity between H3 Vic11, H3 Switz13, and H3s of infecting viruses (Fig. 1B). IgG levels specific for H1 Cal09, the stalk domain of H1, an influenza B virus (IBV) HA, and tetanus toxoid (TTd) did not change significantly (Fig. 1B).

To provide information on the biological activity of Ab responses, H3 Vic11-, H1 Cal09-, and H3 stalk-specific plasma Ab levels on days 0 and 28 were also measured by microneutralization (MN) assay. Levels of Abs that directly inhibited viral activity increased significantly against H3N2 Vic11 and did not change significantly against H1N1 Cal09 (Fig. 1D), reflecting Ab binding to the respective HAs. Neutralization by anti-stalk Abs was measured using 10-fold concentrated plasma IgG and the reassortant cH14/3N3 virus. Neutralizing anti-stalk Ab levels increased significantly from days 0 to 28 (Fig. 1E), consistent with the ELISA data and suggesting a boosting of broadly protective Abs by infection.

(ii) H3N2 infection generates broadly H3 cross-reactive IgG.

To assess the breadth of H3 cross-reactivity of IgG generated by infection, plasma IgG concentrations against H3s of viruses isolated from 1968 to 2013 were measured by multiplex assay. IgG generated by infection and produced as early as day 3 in some subjects bound to a broad range of H3s (Fig. 2A). Within each subject, the kinetics of the response to each H3 was strongly similar. H3-reactive IgG concentrations on day 0 differed between H3 variants in each subject and formed a profile or hierarchy of concentrations that was largely maintained on day 28 (Fig. 2B). This profile differed markedly between some subjects. However, a consistent feature was IgG concentrations on days 0 and 28 that were highest against H3s of older circulating viruses, indicating an effect of OAS on Ab production.

FIG 2.

Analysis of breadth of HA binding by plasma Abs after H3N2 infection. (A) Heatmap representation of HA-reactive plasma IgG concentrations on days 0, 3, 10, and 28 after subject presentation with H3N2 infection. Concentrations were measured by multiplex assay against the indicated IAV and IBV HAs. The results are shown subject by subject, with subjects in order of birth year. (B) Within-subject comparison of plasma IgG reactive with a range of H3s on days 0 and 28. Multiplex measurements of plasma IgG concentrations (y axis) are plotted against the H3s listed by year of virus isolation (x axis). To facilitate comparison of response patterns, local regression (Loess) curves to smooth data are fitted through IgG concentrations on day 0 (blue symbols and curve) and day 28 (red symbols and curve). Subjects are in order of birth year. (C) Heatmap representation of neutralizing Ab levels in plasma. Titers on days 0 and 28 were measured by microneutralization assay against the H3N2 viruses HK68, Wyo03, Wis05, and Switz13. Titers below the limit of assay sensitivity (<25) are identified by white heatmap panels. The results are shown subject by subject, with subjects in order of birth year. (D) Plasma IgG levels on days 0 and 28 against H1s of SC18, USSR77, NC99, and Cal09. Values for individual subjects (n = 8) are shown antigen by antigen in panels D and E. (E) Plasma IgG levels on days 0 and 28 against HAs of IBVs Bris08 and Mass12. Differences in panels D and E were determined to be “NS” (not significant) by repeated ANOVA.

The immune history of an individual that establishes OAS can be complex, but it is generally accepted that the first significant IAV infection plays a key role. In four subjects (subjects 65, 72, 73, and 91), the highest H3-specific IgG titer on day 0 was against a virus that circulated during the first 10 years of the subject’s life (Fig. 2B), suggesting that this or a closely related virus caused early infection and established an OAS response pattern (20, 31). Representation of response magnitude as the difference in plasma H3-specific IgG concentration from days 0 to 28 (delta) reflected the OAS responses in subjects 65, 72, and 73 (Fig. 3A). Notably, however, representation of responses in these and other subjects as the fold change in H3-specific Ab concentrations indicated relatively strong responses to the H3s of recently circulating viruses (Fig. 3B). Overall, our analysis of plasma IgG specific for a broad range of H3s identified OAS-type profiles on day 0 that were maintained after H3N2 infection and also indicated a degree of adaptation of the Ab response to the HA of the infecting virus.

FIG 3.

H3-specific plasma and MBC-derived polyclonal Ab (MPAb) IgG concentrations following H3N2 infection represented as delta or fold change. Concentrations of plasma and MPAb IgG reactive to a range of H3s were measured by multiplex assay. H3s are listed by year of virus isolation on the x axis. (A) Difference in concentration of plasma IgG reactive to the indicated H3s from days 0 to 28 after presentation of the infected subject (delta). (B) Fold change (days 0 to 28) in concentration of plasma IgG reactive to the indicated H3s. (C) Fold change (days 0 to 28) in concentration of MPAb IgG reactive to the indicated H3s. To emphasize patterns, local regression (loess) curves are fitted to the data.

Identification of OAS patterns of Ab reactivity have traditionally been based on functional measurements of anti-HA Ab levels by assays such as HAI and MN. To provide support for our conclusions based on binding data, we measured plasma MN titers on days 0 and 28 against H3N2 viruses isolated in 1968, 2003, 2005, and 2013. Generally, higher binding titers against a particular H3 on day 0 and/or on day 28 (Fig. 2A and B) were associated with high MN titers against the matching H3N2 virus (Fig. 2C). For example, this was evident for binding and MN titers against H3 Wyo03 in subjects 90, 55, 91, and 65 (birth year order) and against H3 HK68 in subjects 72 and 91. However, clear increases from days 0 to 28 in binding titers against H3 HK68 in subjects 72, 73, and 91 were not associated with increased MN titers against H3N2 HK68, perhaps reflecting the large amount of Ab production required for a detectable increase in the already high day 0 MN titers. Notably, MN titers against H3N2 Switz13 increased from days 0 to 28 in all subjects (P < 0.001 for the cohort as a whole), consistent with binding data (Fig. 1B) and an antigenic relationship to infecting viruses. Overall, our limited analysis of MN titers was consistent with OAS patterns of Ab reactivity demonstrated by binding titers, but also indicated that binding data provided a more sensitive measure of Ab reactivity.

Plasma IgG levels measured by multiplex assay against a range of H1s and IBV HAs did not change significantly (Fig. 2A, C, and D).

H3-specific MBC expansion after H3N2 infection includes stalk-reactive MBCs.

To assess the effect of H3N2 infection on HA-specific MBC frequencies, MBCs were stimulated in vitro to induce differentiation into Ab-secreting cells (ASCs). Enumeration of poststimulation ASCs by ELISpot assay provided a measure of precursor MBCs specific for H3 Vic11, the H3 stalk, and H1 Cal09 (Fig. 4A). All subjects had measurable H3 Vic11-specific IgG MBCs on day 0, perhaps reflecting recent expansion of this population by the 2012–13 IIV. IgG MBCs specific for H3 Vic11 and the H3 stalk increased approximately 4- to 20-fold from days 0 to 28, with the stalk-specific MBCs representing approximately 10 to 60% of H3 Vic11-specific MBCs on days 10 and 28 (Fig. 4B). Notably, H1 Cal09-specific IgG MBCs also increased significantly, suggesting a response to cross-reactive epitopes. Patterns of MBC expansion were similar when specific IgG MBCs were expressed as a percentage of the total IgG MBCs (data not shown).

Flow cytometry with an H3 Vic11 probe was used to distinguish H3-specific “activated B cells” (ABCs; recently proliferated precursors of resting MBCs) and resting MBCs (non-recently proliferated) populations (30) (Fig. 4C and D). H3-specific ABCs (CD19⁺ CD71⁺ IgD⁻ CD38⁻ CD20⁺ H3⁺⁾ increased significantly from days 0 to 10, indicating that the increase in H3-specific MBCs resulted from cell proliferation. H3-specific resting MBCs (CD19⁺ CD71⁻ IgD⁻ CD20⁺ H3⁺) were detected in most subjects on day 0 and increased significantly from days 0 to 28.

H3 head-specific MBC expansion after H3N2 infection is broad and reflects OAS-type Ab production.

To investigate the effect of H3N2 infection on MBC populations specific for a broad range of HAs, we analyzed Abs in the culture supernatants of stimulated MBCs (designated MBC-derived polyclonal Abs [MPAbs]). This approach provides an alternative to ASC measurement by ELISpot assay as a readout for antigen-specific MBC activation and permits a far more extensive analysis of MBC reactivity. Antigen-specific MPAb IgG concentrations correlated strongly with frequencies of antigen-specific IgG ASCs derived from stimulated MBCs (determined for H3 Vic11-specific and H1 Cal09-specific responses, r = 0.82 and 0.73, respectively, P < 0.0001) and were used as a measure of the size of specific MBC populations.

Concentrations of MPAb IgG were measured by multiplex assay using the same Ag panel used in the plasma analysis. H3-specific MPAb IgG concentrations, reflecting the numbers of IgG MBCs reactive to each of the H3s, increased markedly from days 0 to 10 in most subjects, with the increase generally continuing to day 28 (Fig. 5A and B). Notably, significant positive correlations between MPAb and plasma IgG concentrations specific for all H3s on day 28 (Fig. 5D) and for 11 of 12 individual H3s on days 0, 3, 10, and 28 combined (r = 0.55-0.78, P ≤ 0.001 for HK68, PC73, Ala81, Phil82, Wis05, Hiro05, Per09, Vic11, Tex12, and Switz13; r = 0.40, P = 0.027 for Wyo03; and r = 0.34, P = 0.058 for Pan99), indicated that the size hierarchy of H3-specific IgG MBC populations after expansion by H3N2 infection reflected that of plasma Ab concentrations (Fig. 5C and D). In other words, Ab response patterns after infection were accompanied by similar patterns of MBC expansion. As was the case for plasma Ab production, the fold changes in H3-specific MPAb concentrations were relatively high for recent H3s (Fig. 3C), suggesting adaptation to the infecting virus in expanded MBC populations.

FIG 5.

Analysis of MBC-derived polyclonal Abs (MPAbs) to investigate the breadth of HA reactivity of MBCs expanded by H3N2 infection. PBMCs were stimulated in vitro to induce MBC differentiation into Ab-secreting cells. MPAb IgG in stimulation culture supernatants was analyzed as a measure of the abundance of specific IgG MBCs. (A) Heatmap representation of HA-reactive MPAb IgG concentrations on days 0, 3, 10, and 28 after subject presentation with H3N2 infection. Concentrations were measured by multiplex assay against the indicated IAV and IBV HAs. The results are shown subject by subject, with subjects in order of birth year. Day 3 data for subject 55 were excluded because of poor in vitro MBC stimulation. (B) Within-subject comparison of MPAb IgG reactive with a range of H3s on days 0 and 28. Multiplex measurements of MPAb IgG concentrations (y axis) are plotted against the H3s listed by year of virus isolation (x axis). Local regression (Loess) curves to smooth data are fitted through IgG concentrations on day 0 (blue symbols and curve) and day 28 (red symbols and curve). Subjects are in order of birth year. (C) Within-subject comparison of H3-reactive MPAb IgG concentrations (reflecting relative sizes of H3-reactive IgG MBC populations) and plasma IgG concentrations on day 28. Multiplex measurements of MPAb (right y axis) and plasma (left y axis) IgG concentrations against each H3 are plotted against the H3s listed by year of virus isolation (x axis). To facilitate comparison of patterns, local regression (Loess) curves to smooth data are fitted through the MPAb (green symbols and curve) and plasma (red symbols and curve) values. (D) Correlation between H3-reactive MPAb IgG and plasma IgG concentrations on day 28. The data are combined from all subjects (8 subjects × MPAb/plasma IgG binding concentrations against 12 H3s = 96 data points). (E) Heatmap representation of neutralizing activity of MPAb IgG on days 0 and 28. IgG was purified and concentrated 10-fold and analyzed by microneutralization assay against the H3N2 viruses HK68, Wis05, and Vic11. Titers below the limit of assay sensitivity (<2) are identified by white heatmap panels. Sample availability limited the analysis to seven subjects shown in order of birth year.

To compare patterns of MPAb reactivity demonstrated by multiplex assay to a functional measure of Ab levels, MPAbs were also analyzed by MN assay. This analysis was limited by sample availability and the low concentrations of H3-reactive Abs in MPAb samples. IgG in MPAbs from seven subjects on days 0 and 28 was concentrated 10-fold and tested for neutralizing activity against the H3N2 viruses HK68, Wis05, and Vic11 (Fig. 5E). MN activity against the three viruses increased from days 0 to 28 in subject 72 in a pattern that reflected binding titers against the matching H3s (Fig. 5A and B). However, in other subjects, the pattern of MN activity less closely matched the pattern of binding titers. This might primarily reflect the sensitivity of the MN assay against H3N2 HK68, which was only positive in a subject (subject 72) who had relatively high binding titers against the matching H3. MN titers against H3N2 Wis05 increased in five of seven subjects, largely in agreement with binding data. The increase in MN titers against H3N2 Vic11 in only four of seven subjects might reflect subject differences in the level of affinity maturation of HA-reactive MBCs generated in response to an HA variant virus. Overall, our analysis of MPAb IgG by MN assay suggests that Ab binding assays identify a greater breadth of reactivity within the MBC pool than do assays that measure Ab function.

MPAbs were also analyzed to confirm H3 stalk-specific MBC expansion. ELISA measurements of MPAb IgG demonstrated that H3 Vic11-specific and H3 stalk-specific MBC populations expanded from days 0 to 28, with the stalk-specific MBCs representing approximately 10 to 20% of the HA-specific MBCs on day 28 (Fig. 6A). These findings were consistent with our analysis based on ASC enumeration after in vitro MBC stimulation (Fig. 4A). Marked expansion of H3 head-specific IgG MBCs was confirmed by ELISA measurements of MPAb IgG specific for the HA1 domains of H3 Vic11 and H3 Switz13 (Fig. 6B).

FIG 6.

Analysis of MBC-derived polyclonal Abs (MPAbs) to investigate HA stalk-specific MBC expansion by H3N2 infection. MPAb IgG generated by in vitro PBMC stimulation was analyzed as a measure of the relative sizes of IgG MBC populations. (A) Comparison of expansion of IgG MBCs specific for H3 and the H3 stalk. MPAb IgG was measured by ELISA against H3 Vic11 and the chimeric HA cH5/3. (B) Expansion of IgG MBCs reactive to H3 head domains. MPAb IgG was measured by ELISA against HA1 domains of H3 Vic11 and H3 Switz13. (C) Analysis of IgG MBC populations reactive with the H1s of SC18, USSR77, and NC99. MPAb IgG was measured by multiplex assay. (D) Comparison of IgG MBC populations reactive with H1 and the H1 stalk. MPAb IgG was measured by ELISA against H1 Cal09 and the chimeric HA cH6/1. (E) Analysis of IgG MBC populations reactive with the HAs of IBVs Bris08 and Mass12. MPAb IgG was measured by multiplex assay. All data from the multiplex analysis of MPAb IgG (including binding to additional H1s and IBV HAs) are presented in Fig. 5A. (F) Analysis of IgG MBC populations reactive with the non-influenza virus protein tetanus toxoid (TTd). MPAb IgG was measured by ELISA. In all panels, values are shown for individual subjects (n = 8). Significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001) was determined by repeated ANOVA with adjusted P values.

To assess the biological activity of MPAb IgG specific for the H3 stalk, IgG was concentrated from day 28 MPAb samples from five subjects and tested for MN activity against the reassortant cH14/3N3 virus. Neutralization was demonstrated for three of the five subjects at minimum H3 stalk-specific MPAb IgG concentrations of 39 to 108 ng/ml. The minimum neutralizing concentration of the control stalk-specific monoclonal antibody (MAb) CR9114 was 1,250 ng/ml. The results indicate that the H3 stalk-reactive MBCs express Abs with antiviral activity and the potential to provide broad protection.

Levels of MPAb IgG that bound to different H1s consistently increased from days 0 to 28, suggesting expansion of cross-H1-reactive MBCs (Fig. 5A and 6C). To investigate expansion of broadly reactive MBCs specific for the H1 stalk, MPAb IgG specific for H1 Cal09 and for the stalk of this H1 was measured by ELISA (Fig. 6D). H1 Cal09 stalk-specific MPAb IgG increased from days 0 to 28 and represented 36 to 86% of H1 Cal09-specific MPAb IgG on day 28, indicating that a large proportion of H1-reactive MBCs expanded by H3N2 infection recognized the stalk region. IBV HA-specific MPAb IgG also consistently increased, suggesting cross-reactivity at the level of expanded MBCs (Fig. 5A and 6E). There was a small but significant increase in MPAb IgG specific for the non-influenza virus protein TTd from days 0 to 28, raising the possibility that infection might also have a nonspecific effect on MBC numbers (Fig. 6F).

Overall, our findings indicate that H3-reactive MBC expansion after H3N2 infection closely reflects Ab production and includes broad MBC reactivity to head epitopes that matches OAS-type Ab responses and MBC reactivity to the H3 stalk. H1-reactive and IBV HA-reactive MBC populations were also expanded after infection, even though plasma Ab levels against these HAs did not change significantly. This might reflect cross-reactivity at the level of the stalk, less stringent selection of MBCs than ASCs in germinal centers (GCs), and the difficulty in demonstrating small increases in plasma Abs against H1 and IBV HA stalks in assays with high baseline levels.

Stalk-specific MBC expansion by H3N2 infection reflects HA imprinting.

Gostic et al. (29) recently provided evidence that an individual’s first IAV infection establishes a form of immune memory to the HA group of the infecting virus. This process, designated HA imprinting, confers a potentially lifelong level of heterosubtypic protection effective against viruses in the same HA group. Thus, individuals imprinted by a first infection with an H1N1 (group 1 HA) or an H3N2 (group 2 HA) virus would have less severe disease if infected with the novel avian IAVs H5N1 (group 1 HA) or H7N9 (group 2 HA), respectively. The mechanism of HA imprinting is unclear, but a role for Abs directed against the HA stalk was proposed. Since the birth years and OAS-type plasma Ab profiles of some subjects in our study indicated childhood H3N2 infection, we evaluated HA stalk-specific Ab and MBC responses subject-by-subject for evidence of group 2 HA imprinting.

Concentrations of plasma and MPAb IgG that bound group 2 HA stalk domains and H7 HAs were measured by ELISA. For comparison, responses to group 1 HA stalks and to H5 HAs were also measured. As described above, our findings indicated that subjects 65, 72, 73, and 91 had experienced early-life H3N2 infection and thus were potentially imprinted on group 2 HAs. However, these subjects did not generate stronger plasma IgG responses to group 2 HA stalks or to H7 HAs than did other subjects (Fig. 7A). Notably, the increase in H3 stalk-specific MPAb IgG was largest in subjects 65, 72, 73, and 91, indicating that infection generated greater H3 stalk-specific MBC expansion in these subjects (Fig. 7B). This trend was similar for H7-specific and H7 stalk-specific MPAbs (Fig. 7B). As would be expected, H7 stalk-specific MPAbs accounted for most of the H7-specific MPAbs. An exception was subject 72, in whom H3N2 infection increased MPAb IgG specific for the H7 head domain (Fig. 7B), suggesting expansion of MBCs specific for a broadly cross-reactive HA head epitope.

FIG 7.

Relationship between potential H3N2 imprinting and formation of group 2 and group 1 HA stalk-reactive plasma IgG and IgG MBCs following H3N2 infection. The levels of plasma and MPAb IgG on days 0 and 28 after subject presentation were measured by ELISA against the indicated HAs and HA constructs to assess responses to group 2 and group 1 HA stalk domains. Specific subjects (n = 8) are identified by symbol shape and color; red symbols identify subjects with potential H3 imprinting by early-life H3N2 infection. (A) Plasma IgG responses to stalk regions of group 2 HAs (H3 and H7). Binding to an H7 head-only reagent was used to identify the component of H7-binding that was stalk-reactive. (B) MPAb IgG responses to stalk regions of group 2 HAs (H3 and H7). (C) Plasma IgG responses to stalk regions of group 1 HAs (H1 and H5). Binding to an H5 head-only reagent was used to identify the component of H5-binding that was stalk-reactive. (D) MPAb IgG responses to stalk regions of group 1 HAs (H1 and H5). (E) Model depicting a proposed mechanism for the resistance of H3N2-imprinted individuals to avian H7N9 influenza virus infection. Data from the present study were used to estimate, for H3N2-imprinted and non-H3N2-imprinted individuals, the proportions of preexisting MBCs reactive to the H3 head domain (shown in green) and to the stalk domain of H3 and other group 2 HAs such as H7 (yellow). A small proportion of H3 head-reactive MBCs in H3N2-imprinted individuals might cross-react with the H7 head; these are shown in orange. The same colors identify HA epitopes of the H7N9 virus that are recognized by MBCs; novel epitopes on the H7 head are shown in purple. The H7 head domain is largely novel in most individuals, leaving stalk-specific MBC activation and PB formation as the major source of early HA-reactive Abs after H7N9 influenza virus infection. Individuals imprinted by early-life H3N2 influenza virus infection have more MBCs reactive to the H7 stalk than do non-H3N2-imprinted individuals. As a result, H3N2-imprinted individuals produce a stronger early stalk-specific Ab response to H7N9 influenza virus infection and have a greater level of protection.

The increase in MPAb IgG to group 1 HA stalks and to H5s was consistently stronger in subjects 72, 73, and 91, but this was not the case for plasma Abs (Fig. 7C and D). MPAb IgG specific for the H5 head was not detected in any subject, indicating that the increase in H5-specific MPAbs reflected expansion of stalk-specific MBCs (Fig. 7D). Subjects 72, 73, and 91 also had marked expansion of group 2 HA stalk-reactive MBCs, raising the possibility of preferential expansion of cross-group stalk-reactive MBCs by H3N2 infection.

Although our findings can only be suggestive because of the small number of subjects, they raise the possibility that expansion of HA stalk-specific MBCs following seasonal IAV infection reflects early-life imprinting and reinforces heterosubtypic protection. A larger pool of group 2 stalk-reactive MBCs could explain the greater level of protection against avian H7N9 infection in H3N2-imprinted subjects proposed by Gostic et al. (29). The H7 head domain would be largely novel in H7N9-infected subjects, leaving the response of H7 stalk-reactive MBCs as the key determinant of the strength of early HA-specific Ab production and the severity of disease (32). This hypothesis is shown diagrammatically in Fig. 7E.

The H3-specific B cell response to H3N2 infection includes MBC adaptation to the infecting virus.

Our findings indicated that the H3-specific B cell response to H3N2 infection was largely a cross-reactive response against conserved HA head epitopes. To investigate adaptation of the B cell response to novel features of the HA of the infecting virus, we used surface plasmon resonance (SPR) to assess IgG binding to the H3 HA1 domains of the H3N2 viruses Vic11 and Switz13 and to the H1 HA1 domain of H1N1 Cal09. H3 Switz13 represents an emerging HA that is genetically and antigenically divergent from H3 Vic11 and the H3s of infecting viruses (22, 33), but our results indicate considerable Ab cross-reactivity between these HAs. Furthermore, H3 Switz13 might share some antigenic features with the H3s of infecting viruses that are absent on H3 Vic11 (22). In addition to analyzing plasma and MPAbs, we analyzed Abs produced by preexisting PBs on day 3, the peak of circulating virus-specific PB numbers. These Abs, designated PPAbs for PB-derived polyclonal Abs, were collected after overnight culture of PBMCs prior to stimulation for MBC activation. Since virus-specific PBs circulating during the early phase of infection result directly from activation of preexisting MBCs (11, 30), PPAbs represent unmodified Abs expressed by those MBCs. For SPR analysis, IgG was purified from plasma, PPAbs, and MPAbs to remove potential influences of other Ab classes.

SPR determinations of Ab dissociation rates are independent of Ab concentration and provide a measure of the net affinity of polyclonal Ab binding (34). The off-rates for plasma Abs bound to the HA1 domains of H3 Vic11 and H3 Switz13 decreased significantly from days 0 to 28, indicating an increase in net affinity (Fig. 8A). Similarly, the off-rates for MPAb binding to the HA1 domains of both H3s decreased significantly from days 0 to 28, suggesting MBC formation after affinity maturation (Fig. 8B). Interestingly, the net affinity of plasma Abs but not MPAbs increased significantly from days 3 to 10; the significant increase in MPAb affinity occurred after a longer period of time. Changes in the net affinity of plasma Ab and MPAb binding to the H1 HA1 domain were relatively small, consistent with a limited B cell response to this region of H1 (Fig. 8A and B).

FIG 8.

Net binding affinities of HA-reactive plasma, PB-derived polyclonal Ab (PPAb), and MBC-derived polyclonal Ab (MPAb) IgG following H3N2 infection. Off-rate constants for purified IgG binding to HA head domains were measured by SPR. (A) Kinetic analysis of off-rates for plasma IgG binding to H3 Vic11 HA1, H3 Switz13 HA1, and H1 Cal09 HA1. (B) Kinetic analysis of off-rates for MPAb IgG binding to H3 Vic11 HA1, H3 Switz13 HA1, and H1 Cal09 HA1. (C) Comparison of PPAb IgG off-rates on day 3 and MPAb IgG off-rates on days 0 and 28 for binding to H3 Vic11 HA1, H3 Switz13 HA1, and H1 Cal09 HA1. In all panels, values are shown for individual subjects (n = 8). Significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001; NS [not significant]) was determined by repeated ANOVA with adjusted P values.

Since PPAbs represent Abs expressed by preexisting MBCs, we expected that off-rates for Ab binding to H3 Vic11 HA1 would be similar for day 3 PPAbs and day 0 MPAbs. However, day 3 PPAb off-rates were significantly lower than day 0 MPAb off-rates, suggesting preferential activation of preexisting MBCs expressing high-affinity Ab (Fig. 8C). This could account for the early increase in plasma Ab affinity that we observed. MPAb off-rates on day 28 were similar to the day 3 PPAb off-rates, consistent with ongoing affinity maturation of H3 Vic11-specific MBCs (Fig. 8C). The pattern was similar for PPAb and MPAb binding to the HA1 domain of H3 Switz13 (Fig. 8C). Relationships between day 3 PPAb off-rates and day 0/day 28 MPAb off-rates for binding to H1 Cal09 HA1 resembled those for binding to the H3 HA1 domains, but indicated less affinity maturation of MBCs (Fig. 8C). Overall, the SPR analysis demonstrates that the broadly cross-reactive, H3-specific B cell response to H3N2 infection includes adaptation of MBCs to novel features of the HA of the infecting virus.

DISCUSSION

Unlike the production of circulating virus-specific Abs following IAV infection, associated changes in the MBC pool have received little attention. The present study related HA-specific Ab production following human infection with H3N2 IAV to changes in HA-specific MBC populations. Our key findings are as follows. (i) H3-specific MBCs expanded by H3N2 infection are, collectively, broadly cross-reactive with the head domain of H3 variants. (ii) The size hierarchy of MBC populations reactive to individual H3 variants after expansion by infection closely reflects patterns of Ab production, providing direct experimental support for current thinking about mechanisms of OAS recurrence. (iii) H3N2 infection expands HA stalk-reactive MBCs, including MBCs cross-reactive with group 1 and group 2 HA stalks. (iv) Broad H3-reactive MBC expansion after infection is accompanied by MBC adaptation to novel features of the HA head domain of the infecting virus. In addition, we present evidence that the expansion of HA stalk-reactive MBCs by H3N2 infection reflects HA imprinting, suggesting a mechanism for imprinting effects on long-term heterosubtypic protection.

There is considerable current interest in the conserved HA stalk as a target of broadly protective Abs and a potential basis for a universal influenza vaccine (26, 27). It is well established that anti-stalk Ab production is favored when HA-immune individuals are exposed to relatively novel HAs, such as those of avian viruses or the 2009 pandemic virus (11, 12, 14, 32, 35). This reflects responses by stalk-specific MBCs with limited competition from MBCs reactive to the immunodominant HA head (9). Although H3 stalk-reactive Abs are poorly generated by seasonal IIV (25, 28), we found significant anti-stalk Ab production after H3N2 infection, consistent with a previous report (25). Infection-associated factors that might promote stalk-specific Ab production include greater epitope accessibility, longer antigen persistence and response duration, and a stimulatory microenvironment that facilitates MBC activation. Importantly, we demonstrate that stalk-specific Ab production after H3N2 infection is associated with expansion of stalk-reactive MBCs. Thus, seasonal H3N2 infection increases an individual’s capacity to generate strong stalk-specific Ab responses on subsequent exposure to the HA stalk.

Recently, Andrews et al. (36) characterized MBCs expanded after prime/boost vaccination of humans with group 1 (H5) or group 2 (H7) HA vaccines. As expected because of the novelty of the avian HA head domains, both vaccination regimens generated a high proportion of stalk-binding MBCs. Notably, the number of stalk-binding MBCs that bound both group 1 and group 2 HAs was significantly higher after H7 (group 2 HA) vaccination than after H5 (group 1 HA) vaccination, suggesting that cross-group stalk-binding MBCs are preferentially expanded in the response to a group 2 HA. Our findings indicate that this applies when the response is against the group 2 HA of an infecting seasonal H3N2 IAV and occurs together with a robust response to the HA head. We demonstrate significant expansion of MBCs reactive to a range of H1s and show that this is largely accounted for by expansion of stalk-reactive MBCs (Fig. 6C and D). In addition, H3N2 infection expanded MBCs reactive to IBV HAs (Fig. 6E), perhaps also reflecting cross-reactivity at the level of the HA stalk (37, 38). Unexpectedly, circulating TTd-specific MBC numbers significantly increased after H3N2 infection (Fig. 6F). Bernasconi et al. (39) proposed that nonspecific B cell activating signals commonly associated with viral infections activate human MBCs in vivo and induce differentiation into plasma cells; nonspecific MBC expansion could occur as part of this process. MPAb binding to non-influenza proteins could also reflect the low level polyreactivity of stalk-specific Abs in a situation where stalk-specific MBC numbers are increased (8).

It is widely accepted that OAS reflects a response dominated by activation of preexisting MBC populations initially established by early-life HA exposure and reactive to HA epitopes conserved on later circulating viruses (3, 21). However, experimental support for this idea is lacking. Our analysis enabled us to consider the question of whether the expansion of H3-reactive MBCs after H3N2 infection would be consistent with the recurrence of OAS responses. In our cohort of eight subjects, four (subjects 65, 72, 73, and 91) had levels of H3-reactive plasma Abs on days 0 and 28 that were generally highest against the H3s of viruses that circulated during the individual’s childhood and thus were strongly suggestive of OAS (Fig. 2B). Three older subjects (subjects 55, 56, and 90) had distinctive profiles of H3-reactive plasma Abs generated by unknown sequences and forms of exposure to influenza virus HAs. We demonstrated that MBC expansion collectively maintained broad H3 reactivity. Importantly, we found that the sizes of MBC populations reactive to individual H3s correlated positively with levels of plasma Abs reactive with each H3. Thus, the expansion of MBC populations maintained the hierarchy of H3 reactivity that would be required for OAS responses to recur, consistent with current thinking about the role of MBCs in this phenomenon.

Our analysis of plasma Ab levels after H3N2 infection identified OAS patterns in the early increase in H3-reactive Abs. This is consistent with analyses of MAbs derived from PBs circulating early in the response to H1N1 or H3N2 infection (11, 28). Typically, the MAbs carried a high number of somatic mutations, indicating PB formation from MBCs. Among the MAbs were those that had an OAS phenotype and reacted with a range of HA variants from previously circulating viruses; some bound more strongly to an earlier HA than to the HA of the infecting virus. Binding was to the HA head, since the MAbs mediated HAI. Overall, the current picture is that OAS Ab responses to infection result from early activation of preexisting MBCs that, individually, respond to HA head epitopes that are sufficiently conserved on a range of HA variants; MBC affinities differ and might be strongest against an earlier HA. The mechanism that balances MBC depletion through activation and differentiation into PBs with replenishment of MBCs required for the maintenance of OAS is not clear. Consistent with other studies (30), our flow cytometric analysis demonstrated that H3⁺ PBs and H3⁺ ABC precursors of resting MBCs were formed early in the response after proliferation of activated precursor cells (Fig. 4D). B cell lineage analysis indicates that both PBs and ABCs are formed from preexisting MBCs, but it is unclear whether both lineages originate from the same precursor MBC or are formed from functionally distinct MBC subsets (30, 40, 41). Early PB formation from activated, high-affinity MBCs is likely to occur in extrafollicular locations without GC involvement (42). MBC replenishment from activated precursors might involve extrafollicular proliferation alone and/or GC seeding with proliferation and affinity-based selection (43).

Recent studies have reported that H3N2 infection generated Ab production that was strongest against the H3s of contemporary viruses or the infecting strain, even though Abs against older H3s were also boosted (16, 23). Our findings are similar when the H3-specific Ab response is represented as the fold change (Fig. 3B). However, it is important to note that representing the Ab response as delta, a measure of the amount of Ab produced, results in a different picture with stronger production of Abs reactive with older H3s (Fig. 3A). In our analysis, we demonstrate that broad H3-reactive MBC expansion after H3N2 infection included MBC adaptation to novel features of the HA head domain of the infecting virus. This is indicated by increased net affinity of MBCs for the H3 HA1 domains of Vic11 and Switz13 (Fig. 8B). Interestingly, PPAbs collected from day 3 PBs, which are formed directly from preexisting MBCs, had net affinities for Vic11 HA1 and Switz13 HA1 that were significantly higher than those of MPAbs on day 0, but similar to those of MPAbs on day 28 (Fig. 8C). This could reflect early PB formation from preexisting MBCs preferentially activated because of high affinity for the H3 of the infecting virus (44), with later formation of MBCs in GCs after selection for high affinity. This scenario is consistent with our finding that net affinity for Vic11 HA1 and Switz13 HA1 increased earlier for plasma Abs than for MPAbs (Fig. 8A and B). Overall, our study indicates the importance of infection in modulating MBC specificities through sustained GC reactions (4). It is likely that GCs generated in the response to IAV in immune adults are primarily seeded by preexisting MBCs, rather than by cells stemming from de novo activation of naive B cells (45). All subjects in our study had received the 2012–13 seasonal influenza vaccine and this might have generated a preexisting MBC pool well suited for activation and specificity remodeling in GCs. In a separate study, DeDiego et al. (22) identified genetic differences between antigenic sites in the HA head domain of the 2012–13 influenza vaccine component H3 Vic11 and HAs of viruses infecting subjects in the present study, although HAI assays indicated strong cross-reactivity between all viruses.

In conclusion, our analysis of the HA-specific B cell response to seasonal H3N2 infection in immune adults focused on changes in HA-reactive IgG MBC populations. We show that infection expanded MBCs with, collectively, very broad H3 reactivity. Importantly, the size of MBC populations reactive to particular H3 variants after infection related closely to the strength of Ab production against those variants, including OAS patterns of Ab production. Thus, our analysis provides direct experimental support for current thinking about the role of MBCs in the recurrence of OAS responses. OAS is commonly considered to be mediated by responses to conserved epitopes in the HA head; we also demonstrated that infection expanded MBCs reactive to the highly conserved HA stalk and that these MBCs might, in particular situations, form the basis of OAS patterns of stalk-specific Ab production. Overall, our analysis emphasizes the potential of IAV infection to enhance broad protection through HA-reactive MBC expansion, but with differences between individuals that depend on their early-life influenza exposure history. Comprehensive studies of MBC expansion after various forms of influenza vaccination will be important for comparison with the situation following infection. Findings could guide development of a broadly protective universal influenza vaccine (46).

MATERIALS AND METHODS

Antigens.

All recombinant HA proteins used in this study are listed in Table 2. These included a range of baculovirus-expressed trimeric HA ectodomains and globular head domains (47) used for coating assay plates and generating the bead-coupled antigen panel for the multiplex assay. The chimeric HA ectodomains cH4/7 (35), cH5/3 (25), and cH6/1 (48) primarily identified HA stalk-binding Abs, since they carried exotic head domains that are novel to most humans. HA1 protein derivatives of H3 Vic11, H3 Switz13, and H1 Cal09 were expressed in Escherichia coli and purified and characterized as previously described (49, 50). The HA1 protein preparations consisted of properly folded trimers and oligomers with functional activity (49). Expression of all recombinant HA proteins was based on sequences derived from egg-grown viruses. TTd (MilliporeSigma, Burlington, MA) was used as a non-influenza control protein.

Viruses.

A range of viruses was used in MN assays. The H3N2 influenza viruses A/Hong/Kong/1/1968 (NR-28620), A/Wisconsin/67/2005 (NR-41800), and A/Victoria/361/2011 (NR-44022) were obtained from the Biodefense and Emerging Infections Research Resources Repository (BEI Resources), and influenza A/Switzerland/9715293/2013 (H3N2) (FR-1368) was obtained from the Influenza Reagent Resource (IRR). Influenza A/Wyoming/3/2003 (H3N2) and A/California/04/2009 (H1N1) were generated by reverse genetics as described previously (51). The reassortant virus cH14/3N3 (25, 52) expressed a chimeric HA consisting of the exotic H14 head and the Per09 (H3N2) stalk, together with the irrelevant NA N3. Thus, this virus could be used to identify neutralization by anti-stalk Abs in an MN assay with multiple growth cycles. Virus stocks for use in MN assays were grown and titrated by plaque assay in MDCK cells.

Ab-secreting cell analysis.

Frozen PBMCs were thawed and washed and a proportion was analyzed directly by ELISpot assay to enumerate preexisting Ab-secreting PBs (14). A minimum of 500,000 PBMCs were analyzed for each antigen and the limit of PB detection was set at two spots (PBs)/10⁶ PBMC. A separate proportion of PBMCs was “rested” overnight at 1.5 × 10⁶ to 2.0 × 10⁶ PBMCs/ml in six-well tissue culture plates. Rested cells were harvested and pelleted by centrifugation. Undiluted supernatant containing Abs secreted by preexisting PBs (PB-derived polyclonal Abs or PPAbs) was collected and stored (53).

MBC analysis.

Measurement of antigen-specific MBCs was based on previously described methods (54–56). Rested PBMCs at 0.5 × 10⁶ PBMCs/ml in 24-well tissue culture plates were stimulated for 6 days to induce MBC differentiation into ASCs (MBC-derived ASCs or MASCs). The stimulation cocktail consisted of complete medium with 1/100,000 dilution pokeweed mitogen (a gift from Shane Crotty), 1/10,000 dilution protein A from Staphylococcus aureus (MilliporeSigma), 6 μg/ml CpG ODN-2006 (InvivoGen, San Diego, CA), and 25 ng/ml interleukin-10 (Invitrogen, Carlsbad, CA). After stimulation, cells were harvested and pelleted by centrifugation. The undiluted supernatant containing Abs secreted by ASCs generated from stimulated MBC precursors (MBC-derived polyclonal Abs or MPAbs) was collected and stored for later analysis. Supernatants from unstimulated cultures of rested PBMCs were collected to control for Abs produced by preexisting PBs. Antigen-specific and total IgG ASCs in the cell pellet (MASCs) were enumerated by ELISpot assay. A minimum of 500,000 stimulated PBMCs were analyzed by ELISpot assay for each antigen and the limit of MASC detection was set at four spots (MASCs)/10⁶ PBMCs. Based on ELISpot assay results, antigen-specific MBCs in peripheral blood were quantified as (i) antigen-specific IgG MASCs as a proportion of stimulated PBMCs (55) and (ii) antigen-specific IgG MBCs as a percentage of total IgG MBCs (calculated as the percentage of antigen-specific MASCs among the total number of IgG MASCs) (54). Antigen-specific IgG concentrations in MPAb samples (after subtraction of Ab concentrations in supernatants from unstimulated PBMC control cultures) were also used as a measure of the relative size of reactive MBC populations (56).

Flow cytometry.

Thawed and washed PBMCs were stained as described previously (30) with an H3 Vic11 probe and the following panel of reagents: anti-CD3 PE-Cy5.5 (UCHT1; SouthernBiotech, Birmingham, AL), anti-CD19 PE-TR (J3-119; Beckman Coulter, Brea, CA), anti-CD71 FITC (CY1G4; BioLegend, San Diego, CA), anti-IgD PE (IA6-2; BD Biosciences, San Jose, CA), anti-CD20 BV450 (L27; BD Biosciences), CD27 PC5 (1A4CD27; Beckman Coulter), anti-CD38 PE-Cy7 (HIT2; eBioscience, San Diego, CA), and streptavidin-APC (BD Pharmingen, San Jose, CA). A Live/Dead Fixable Aqua Dead cell stain kit (Invitrogen) was used to discriminate dead cells. The H3 Vic11 probe was a biotinylated trimer expressed in 293 cells (30). Specificity of the H3 Vic11 probe for human B cells expressing H3-specific Ig was established using 293 cells transfected with probe-specific and non-probe-specific Ig and PBMCs from subjects acutely infected with H3N2 or H1N1 IAV. Data were acquired using an LSR II flow cytometer (BD Biosciences) and analyzed using FlowJo software.

Elisa.

Levels of Abs specific for selected antigens in plasma and MPAb samples were measured by ELISA. Briefly, Nunc MaxiSorp 96-well plates were coated with optimized concentrations of antigens. ELISAs were completed by addition of serial 2- or 3-fold dilutions of sample, followed by alkaline phosphatase-conjugated anti-human IgG (clone MT78; Mabtech, Cincinnati, OH) and then p-nitrophenyl phosphate substrate. Well absorbance was read at 450 nm after color development.

Specific Ab levels in test samples were quantified using a human serum standard that contained Abs of the required specificities. Weight-based concentrations of specific Abs in the standard were assigned as described previously (57). Initially, the total IgG concentration in the standard was determined by an IgG capture ELISA using plates coated with Abs to human kappa and lambda light chains (clones MHK-49 and MHL-38, respectively; BioLegend). The IgG concentration was calculated from a curve constructed using a human IgG subclass standard serum (Accurate Chemical and Scientific Corporation, Westbury, NY). Control experiments established that the anti-IgG secondary Ab bound comparably to all human IgG subclasses. To assign antigen-specific Ab concentrations to the standard, the standard was run concurrently in antigen-specific ELISAs and the IgG capture ELISA. The slopes of regression lines for log of the absorbance versus log of IgG concentration (IgG capture ELISA) or reciprocal dilution (antigen-specific ELISA) were similar for titration of the standard in the IgG capture ELISA (0.9 to 1.1) and in antigen-specific ELISAs (0.8 to 1.0), indicating similar equilibrium binding in the respective assays. The known IgG concentration at a selected absorbance in the IgG capture ELISA was thus, assigned as the antigen-specific concentration at the dilution of the standard that gave the same absorbance in antigen-specific ELISAs. Unknown Ab concentrations were calculated from standard curves by nonlinear regression analysis with four-parameter logistic curve fitting.

Multiplex assay.

Levels of binding Abs in plasma and MPAbs were measured against a broad range of HAs (listed in Table 2) by multiplex assay (mPlex-Flu assay) (58). Based on the method of Quataert et al. (57), a standard was generated from pooled human sera for conversion of median fluorescence intensity (MFI) measurements of HA-binding Ab levels to weight-based units. Briefly, the serum pool was titrated in the multiplex assay with the panel of HA-coupled beads and a bead population coupled to goat anti-human IgG(H+L) (KPL, Gaithersburg, MD) for total IgG capture. Bead-bound Abs were detected using PE-conjugated goat anti-human IgG. Regression line slopes were similar for total and HA-specific IgG titrations. A curve constructed from known total IgG concentrations in the serum standard and the corresponding MFI measurements was thus, used to assign HA-specific IgG concentrations to the standard, assuming that a selected MFI reflected similar amounts of bound Ab. Titration of the serum standard in multiplex assay runs was used to generate a standard curve for each required HA specificity. HA-specific Ab concentrations in test samples were assigned by nonlinear regression analysis using standard curves constructed by four-parameter logistic curve fitting.

HAI assay.

HAI titers were measured against egg-grown influenza virus A/Victoria/361/2011 (H3N2) (22).

IgG purification.

IgG was purified using protein G-Sepharose 4 Fast Flow (GE Healthcare Life Sciences, Pittsburgh, PA) according to the manufacturer’s instructions. After elution and neutralization, the IgG was concentrated and buffer exchange (using phosphate-buffered saline [PBS; pH 7.2]) was performed simultaneously using 50-kDa Amicon Ultra centrifugal filters (MilliporeSigma).

MN assay.

The MN assay to measure neutralization of infection by HA-specific Abs was performed as described previously (59, 60) against the H3N2 viruses HK68, Wyo03, Wis05, Vic11, and Switz13, and H1N1 Cal09. Briefly, triplicates of heat-inactivated plasma samples or duplicates of 10-fold-concentrated purified IgG from MPAbs were serially diluted 2-fold in virus diluent and incubated with 100 to 200 PFU of virus for 1 h at room temperature. The virus/sample mixtures were then added to confluent MDCK cell monolayers in 96-well flat-bottom plates. After virus adsorption for 1 h at room temperature, virus/sample inocula were removed and MDCK cells in TPCK-trypsin-supplemented infection medium were cultured for 60 to 72 h at 33°C (until cytopathic effect was observed in the virus-only control wells). The wells were washed with PBS and then stained with crystal violet for 1 h. The MN titer was identified as the highest dilution of the tested sample that completely neutralized the virus in 50% of wells.

Neutralization by anti-stalk Abs was measured by MN assay against the reassortant virus cH14/3N3. IgG was purified and concentrated 10-fold from plasma samples and 30- to 60-fold from MPAb samples to increase the likelihood of detecting neutralizing activity; samples were tested in duplicate. The assay was performed as described above, but with modifications. After incubation of serially diluted samples with virus (100 PFU) and the virus adsorption step, the same concentration of IgG that was used to neutralize virus in each well was included in the infection medium added to the well. Cells were then incubated for 72 h at 33°C, and the assay was completed as described above.

Net affinity measurement by SPR.

The net binding affinity of polyclonal IgG in plasma, PPAbs, and MPAbs was measured by SPR against H3 and H1 head domains. IgG was purified from samples for SPR analysis. Steady-state equilibrium binding of IgG was monitored at 25°C using a Bio-Rad ProteOn SPR biosensor (34, 61). Recombinant HA1 globular head domains of H3 Vic11, H3 Switz13, or H1 Cal09 were coupled to a GLC sensor chip. The spatial density of antigen on the chip was adjusted to measure only monovalent Ab binding. Ab off-rate constants, which describe the fraction of antigen-Ab complexes that decay per second, were determined directly from sample interactions with HA antigens using SPR in the dissociation phase (34, 61).

Statistics.

Pairwise comparisons of values at different sampling times were performed by paired t test with data on a log scale and by its nonparametric version the signed rank test. In addition, repeated analysis of variance (ANOVA) was applied to log-transformed data to analyze changes in values over time, with consideration of within-subject correlation and simultaneous use of all information. The statistical findings presented are from repeated ANOVA with the Tukey method for type I error adjustment, but the results of all analytical methods supported the same conclusions. Correlations were tested by Pearson correlation analysis and confirmed by Spearman correlation analysis.

Study approval.

This study was approved by the University of Rochester Human Research Subjects Review Board (protocol 09-0034). Written informed consent was obtained from subjects prior to inclusion in the study.