To identify HA-specific B cells, we used recombinant HA from A/Puerto Rico/8/34 (PR8) secreted by insect cells with its carboxy-terminal domain modified to promote trimerization and enable detection with an anti-His tag monoclonal antibody (MAb). This HA preparation (rHA^PR8) is remarkably native, as shown by biochemical analysis in conjunction with a panel of MAbs to assess conformation (20). To establish the properties of rHA^PR8 as a B cell probe, we screened hybridoma cell lines to identify a line that expresses cell surface Ig, which is unusual (21). H17-L7, an HA Cb site-specific hybridoma, exhibits surface IgG expression (Fig. 1A) and binds rHA^PR8 in a dose-dependent fashion but does not bind to the serologically distant H5 rHA^{Vietnam 04}, demonstrating the specificity of staining with rHA^PR8 (Fig. 1B and C).

We next stained mediastinal lymph node (MLN)-resident B cells from B6 mice infected intranasally (i.n.) with PR8. We focused on GC B cells (defined as B220⁺ CD38^lo GL7^hi), as they exhibit a distinct differentiation marker profile and retain strong surface Ig expression (22). Among MLN CD3-negative cells (>90% B220⁺), rHA^PR8 staining was essentially limited to GC B cells (Fig. 1D). To confirm the specificity of rHA^PR8 binding to GC B cells, we infected mice with a reassortant IAV (J1) with 7 PR8 genes but a serologically distinct HA from an H3 virus (A/Hong Kong/1/68). Little reactivity was observed from J1-infected GC B cells up to an rHA^PR8 concentration of 66 nM, providing an upper concentration limit for detecting Ab-specific binding of this probe to GC B cells (Fig. 1E).

Using a staining concentration of 66 nM rHA^PR8 we quantified the HA-specific GC B cell response following i.n. infection in the MLNs and spleen. HA-specific GC B cells were first detected at 7 days postinfection (dpi) in MLNs and 10 dpi in the spleen. Numbers of splenic HA-reactive GC B cells peaked at 14 dpi, while their MLN-resident counterparts peaked a week later (Fig. 2A). Expressed as a fraction of total GC B cells, splenic B cells peaked at 21 dpi while MLN B cells continued to rise in frequency at 28 dpi (Fig. 2B).

These findings confirm the utility of rHA^PR8 in tracking HA-specific responses and clearly demonstrate distinct kinetics of the HA-specific GC B cell response in proximal versus distal lymphoid organs.

As initially described for hapten-specific B cells (23), the binding of rHA^PR8 to GC B cells should obey the law of mass action and thereby provide a measure of Ab affinity for HA. In theory, for any B cell, Ab affinity should equal the rHA^PR8 concentration that gives half-maximal binding (rHA^PR8₅₀). There is no need to reach saturation to define AC₅₀, as B cell receptor (BCR)-specific and “nonspecific” antigen binding is a continuum, where “nonspecific” means low-affinity binding of germline BCRs. However, during an immune response, Abs with high affinity will outcompete those with lower affinity. Therefore, our maximum binding is the binding at a concentration where naive BCRs do not bind. While each B cell can be tested only for binding to a single rHA^PR8 concentration, at the population level, if cell surface Ig levels are equally distributed among B cell clones, the average Ab affinity will track the rHA^PR8 concentration that detects half the number of HA-specific B cells. We define this as the 50% antigen concentration (AC₅₀) (Fig. 3A). It is important to emphasize that since HA-specific GC B cells comprise a heterogeneous population of clones with various affinities and surface Ig expression, AC₅₀ provides (at best) the mean affinity of Abs expressed by the HA-specific B cell population studied.

We verified the validity of AC₅₀ measurements by comparing the calculated H17-L7 AC₅₀ versus purified H17-L7 MAb affinity determined via ELISA (Fig. 3B). We observed remarkable agreement between AC₅₀ (1.1 nM) and K_D (equilibrium dissocation constant) (0.6 nM), supporting the validity of our approach. We next compared the avidities of 10 MAbs generated 5 days after primary infection (24) to the AC₅₀ of GC B cells 7 days after i.n. infection. Again, there was strong concordance of AC₅₀ with the average avidity, with both values being ~10 nM (Fig. 3D).

We next investigated the extent to which the AC₅₀ is influenced by the fraction of HA-specific cells in a B cell population, an important factor given the low and variable fraction of HA-specific B cell in ex vivo populations used for studying HA-specific GC B cell responses. We mixed naive splenocytes with H17-L7 hybridoma in different proportions and determined the binding curves and respective AC₅₀s. The AC₅₀s were statistically indistinguishable between the different mixtures (Fig. 3E).

Based on these findings, we conclude that by simple antigen titration in flow-cytometric staining, we can derive a value, the AC₅₀, that provides a measure of the affinity of the B cell Ig receptor.

We next used AC₅₀ to characterize maturation of specific GC B cell responses in MLNs and spleens during IAV infection. GC B cells at 28 dpi had a significantly higher frequency of HA reactivity at lower rHA^PR8 concentrations than 14-dpi counterparts (Fig. 4A), with a clear increase in HA-specific GC B cell frequency and a decrease of AC₅₀ over 4 weeks following infection (Fig. 4B). AC₅₀ decreased approximately 30-fold, from 10 nM at 7 dpi to 0.36 nM at 28 dpi (Fig. 4C).

While we were unable to obtain reliable AC₅₀ data for spleen-resident B cells at early and late time points due to low frequencies of HA-specific GC B cells (Fig. 2B), we did calculate AC₅₀ at 14 and 21 dpi, which follow a surprisingly similar pattern of AC₅₀ decrease to MLN-resident GC B cells (Fig. 4D). Importantly, we cannot trivially attribute the time-dependent decrease of AC₅₀ in MLN and splenic GC B cells to temporal increases in GC B cell surface Ig levels, which are statistically indistinguishable between 14 and 28 dpi (Fig. 4E) (note that likely due to steric hindrance, we were unable to costain GC B cells with rHA^PR8 and anti-Ig Abs to directly correlate Ig surface levels with antigen binding).

A potential pitfall of HA staining is the requirement to treat cells with neuraminidase to prevent HA binding to the abundant cell surface terminal sialic acids. To circumvent this problem, we used an rHA with a substitution in the sialic acid binding site that prevents sialic acid binding (rHA^PR8-Y98F) (19). This enabled GC B cell staining without neuraminidase treatment of cells and gave essentially identical results (see Fig. S1 in the supplemental material).

Using the rHA^PR8-Y98F probe, it was further possible to costain with anti-Ig Abs, allowing us to correlate IgG surface levels and rHA binding. We show that the amount of IgG on the surface of B cells does not determine rHA binding (see Fig. S2 in the supplemental material), strengthening our conclusion that AC₅₀ is not trivially due to different levels of surface B cell receptor.

Lastly, we used the mean fluorescent intensity (MFI) of stained cells to calculate the AC₅₀. As predicted from the law of mass action, this yielded results similar to those obtained by using the percent positive cells (an example is shown in Fig. S3 in the supplemental material) but exhibited greater variability between samples. Therefore, in all additional experiments, we calculated the AC₅₀ based on the percent positive cells.

If the observed decrease in AC₅₀ is due to affinity maturation, it should not occur in mice with activation-induced deaminase knockout (AID^−/− mice), which are incapable of somatic hypermutation (25). In response to intrperitoneal (i.p.) injection of UV light-inactivated PR8, WT mice generated a robust HA-specific GC B cell response exhibiting 12-fold-decreased AC₅₀ (5.8 nM to 0.49 nM) between 14 and 28 dpi (Fig. 5A and B). As seen in other studies, the AID^−/− mice generate a large germinal-center reaction, with a robust HA-specific GC B cell response (26). However, the AID^−/− response’s AC₅₀ remained constant from 14 to 28 dpi (Fig. 5). Since Ig class switching is also abrogated in AID^−/− mice, these data strongly support the conclusion that kinetic changes in AC₅₀ reflect the selection of higher-affinity HA-specific clones by classical B cell somatic mutation and selection mechanisms.

As proof of principle of the utility of AC₅₀ for characterizing immune responses, we compared i.n. infection with infectious PR8 to a single intramuscular (i.m.) injection of UV-inactivated PR8 in Titermax adjuvant. The latter is known to be less protective against lethal PR8 challenge (27). Following i.m. immunization, serum anti-HA Abs were always detected, confirming successful vaccination (data not shown). HA-specific GC B cell responses were uniformly observed in the most proximal popliteal draining lymph nodes, 33% of the time in the more distant inguinal nodes, but never within the spleen (Fig. 6A), indicating only localized generation of HA-reactive GC B cells.

The HA-specific B cell response in the inguinal LN was similar in frequency to that in i.n. infection (Fig. 6B) but was numerically inferior; i.m. vaccination of mice induced only 5% and 7% of GC B cells compared to i.n. infection at 14 and 28 dpi, respectively (Fig. 6C). While the popliteal LN B cell numbers induced by i.m. immunization were too low to establish AC₅₀ due to lymph node size, the inguinal LN GC B cell response decreases in AC₅₀ from 14 to 28 after i.m. vaccination with similar kinetics to infected responses (~7 to 10-fold decrease, vaccination to infection, respectively). However, at both time points, the GC B cell AC₅₀ after vaccination was approximately 4-fold higher than those responding to i.n. infection (Fig. 6D). Thus, i.m. vaccination resulted in a 14- to 20-fold lower HA-specific GC B cell response that was also 4-fold higher in AC₅₀ than that achieved with i.n. infection.

These findings predict that i.m. immunization will generate less anti-HA Ab of lower affinity. We tested this prediction by surface plasmon resonance (SPR) analysis of serum Ab responses, which provides a measure of total Ab responses and accurately measures the average Ab off-rates, typically the most important factor in Ab affinity. In agreement with our measurement of GC B cell numbers, serum antibody HA-binding activity from i.m.-vaccinated mice was 38% and 17% lower than in sera from infected animals at 14 dpi and 28 dpi, respectively (Fig. 6E). Further confirming our AC₅₀ data, Ab off-rates decreased in vaccinated mice between days 14 and 28, suggesting increased affinity, but retained statistically significant faster off-rates compared to postinfection sera (Fig. 6F). The disparity in off-rates (1.3-fold and 1.4-fold higher at days 14 and 28 postvaccination, respectively) remained stable relative to values for postinfection sera, similar to observed GC B cell AC₅₀ (Fig. 6D).

Taken together, these data support the validity of the AC₅₀ method for determining Ab avidity and demonstrate its utility for analyzing immunity to viral infection and immunization.

Recombinant PR8 Mount Sinai hemagglutinin (rHA^PR8) was expressed and purified as previously described (19, 36). The PR8 HA protein possessed a C-terminal thrombin cleavage site, a “foldon” sequence, and a His tag at the extreme C terminus for protein purification and detection. In later experiments, we also used a PR8 HA with a Y98F mutation that abrogates binding activity with a biotin C-terminal tag (19). For Biacore serum affinity measurements, we used recombinant PR8 H1N1 HA obtained through BEI Resources (NIAID).

Allophycocyanin (APC)-conjugated and purified mouse anti-6×His (clone ad 1.1.10) was purchased from Abcam. Fluorescein isothiocyanate (FITC)-conjugated rat anti-mouse CD38 (90/CD38), APC- and phycoerythrin (PE)-conjugated rat anti-mouse GL7 (GL7), PE-Cy7- and Pacific blue-conjugated rat anti-mouse CD45R/B220 (RA3-6B2), and Pacific blue-conjugated hamster anti-mouse CD3ε (500A2) were purchased from BD Biosciences. APC-conjugated streptavidin was purchased from eBioscience. DyLight 6470-conjugated AffiniPure goat anti-mouse IgG (all subclasses), Fcγ fragment specific, was purchased from Jackson ImmunoResearch.

C57BL/6 mice were purchased from Taconic Farms or Charles River Laboratories. For all experiments, 6- to 12-week-old AID^−/− mice (25) were kindly provided by A. Nussenzweig with gracious permission of T. Honjo. All mice were housed under specific-pathogen (including murine norovirus [MNV], mouse parvovirus [MPV], and mouse hepatitis virus [MHV])-free conditions. All animal procedures were approved by and performed in accordance with the NIAID and CBER Animal Care and Use Committee guidelines.

For intranasal infections, mice were anesthetized in isoflurane and inoculated intranasally with 50 TCID₅₀ (50% tissue culture infective doses) of influenza virus A H1N1 PR8 (Mt. Sinai strain) or influenza virus A J1, a PR8 H3 reassortant (49). For i.p. injections, virus was UV irradiated for 4 min, and 4 × 10⁸ TCID₅₀ (pre-UV inactivation titer) was administered by intraperitoneal injection. For intramuscular challenge, UV-inactivated virus was mixed at a ratio of 50:50 with Titermax gold adjuvant (Sigma), and 5 × 10⁶ TCID₅₀ (pre UV-inactivation titer) were injected bilaterally into the caudal thigh.

At various times after injection/infection, the mediastinal lymph nodes, inguinal lymph nodes, and/or spleens were harvested from euthanized mice. Tissues were dispersed into single-cell suspensions and treated with red blood cell lysis buffer (Sigma) prior to staining with antibodies. In order to account for the number of specific B cells, a low input cell number (on average 10,000 to 20,000 antigen-specific B cells per tube) combined with excess rHA probe and large washing volumes was used. When stained with recombinant HA (rHA), cells were treated with either filtrate-derived (1:20 dilution) or purified (6 mU/ml in RPMI) neuraminidase from Vibrio cholerae for 60 min at 37°C in RPMI (Gibco) to remove surface sialic acid from cells. Cells were stained with primary surface antibodies (30 min, 4°C), with a complete nine-step titration of rHA^PR8 for each experiment (starting at 66 nM down to 0.0066 nM) (60 min 4°C), and finally with anti-6× antibodies (30 min 4°C), to detect rHA^PR8 binding. Each experiment included appropriate fluorescent minus one (FMO), unstained, and single-stained controls to ensure proper staining and gates that provided the internal controls for the titration. Stained cells were analyzed on an LSR-II flow cytometer (BD Biosciences). For staining with rHA^PR8-Y98F (19), the same procedure was followed but with neuraminidase treatment omitted and with final staining performed using streptavidin-APC conjugated (30 min, 4°C).

Using a titration of rHA^PR8 probe with J1 (H3N1)-reactive germinal center B cells, the maximal specific binding of the HA^PR8 probe was established. rHA^PR8 probe was titrated over a dose range of 66 nM to 0.66 nM, and data were plotted as the frequency of rHA^PR8 probe-positive GC B cells. Using a single one-site binding with Hill slope calculation, the 50% maximal binding rHA^PR8 probe concentration was established to evaluate the antigen concentration that stains 50% of the maximum specific population, or AC₅₀. Where noted, the geometric mean fluorescent intensity (MFI) was used instead of percentage of the population. In these experiments, AC₅₀ is defined as the antigen concentration staining at 50% of the maximal MFI.

Ninety-six-well plates (Immunlon 4HBX) were used to test MAb reactivity. Plates were coated overnight at 4°C with saturating amounts of purified PR8 HA in phosphate-buffered saline (PBS), then washed three times with PBS containing 0.05% (vol/vol) Tween 20, and blocked with PBS–0.5% fetal bovine serum (FBS) for 2 h at room temperature. MAb was titrated from nondetectable to saturating binding and incubated with virus for 2 h at room temperature. Rat anti-mouse kappa conjugated to horseradish peroxidase (Southern Biotech) was used to detect MAb binding. Plates were developed for 5 min using TMB (tetramethylbenzidine) substrate (KPL biomedical) and halted by the addition of 0.1 N HCl, after which plates were read at 450 nm. Ab avidities were determined using Prism software, and all avidities reported demonstrated excellent fit for one-site binding with Hill slope curve fitting (R² > 0.98).

Steady-state equilibrium binding of polyclonal sera was monitored at 25°C using a ProteOn surface plasmon resonance biosensor (Bio-Rad Labs). The recombinant functional PR8 H1N1-HA0 (BEI resources) was coupled to a GLC sensor chip using amine coupling with 100 resonance units (RU) in the test flow cells. Samples of 60 µl of freshly prepared polyclonal sera at various concentrations were injected at a flow rate of 30 µl/min (120-s contact time). Flow was directed over a mock surface to which no protein was bound, followed by the recombinant functional PR8 H1N1-HA0 coupled surface. Responses from the protein surface were corrected for the response from the mock surface and for responses from a separate, buffer-only injection. MAb 2D7 (anti-CCR5) was used as a negative-control antibody in various binding experiments. Binding kinetics for the sera and the data analysis was analyzed using Bio-Rad ProteON manager software (version 2.1.1). Affinity measurements were calculated using the heterogeneous model.

Ab avidities were determined using Prism software, and all avidities reported demonstrated excellent fit for one-site binding with Hill slope curve fitting (R² > 0.98). Significances were calculated by GraphPad Prism (GraphPad Software, Inc.) using unpaired Student’s t test.