Intranasal Live Influenza Vaccine Priming Elicits Localized B Cell Responses in Mediastinal Lymph Nodes

ABSTRACT

Pandemic live attenuated influenza vaccines (pLAIV) prime subjects for a robust neutralizing antibody response upon subsequent administration of a pandemic inactivated subunit vaccine (pISV). However, a difference was not detected in H5-specific memory B cells in the peripheral blood between pLAIV-primed and unprimed subjects prior to pISV boost. To investigate the mechanism underlying pLAIV priming, we vaccinated groups of 12 African green monkeys (AGMs) with H5N1 pISV or pLAIV alone or H5N1 pLAIV followed by pISV and examined immunity systemically and in local draining lymph nodes (LN). The AGM model recapitulated the serologic observations from clinical studies. Interestingly, H5N1 pLAIV induced robust germinal center B cell responses in the mediastinal LN (MLN). Subsequent boosting with H5N1 pISV drove increases in H5-specific B cells in the axillary LN, spleen, and circulation in H5N1 pLAIV-primed animals. Thus, H5N1 pLAIV primes localized B cell responses in the MLN that are recalled systemically following pISV boost. These data provide mechanistic insights for the generation of robust humoral responses via prime-boost vaccination.

IMPORTANCE We have previously shown that pandemic live attenuated influenza vaccines (pLAIV) prime for a rapid and robust antibody response on subsequent administration of inactivated subunit vaccine (pISV). This is observed even in individuals who had undetectable antibody (Ab) responses following the initial vaccination. To define the mechanistic basis of pLAIV priming, we turned to a nonhuman primate model and performed a detailed analysis of B cell responses in systemic and local lymphoid tissues following prime-boost vaccination with pLAIV and pISV. We show that the nonhuman primate model recapitulates the serologic observations from clinical studies. Further, we found that pLAIVs induced robust germinal center B cell responses in the mediastinal lymph node. Subsequent boosting with pISV in pLAIV-primed animals resulted in detection of B cells in the axillary lymph nodes, spleen, and peripheral blood. We demonstrate that intranasally administered pLAIV elicits a highly localized germinal center B cell response in the mediastinal lymph node that is rapidly recalled following pISV boost into germinal center reactions at numerous distant immune sites.

INTRODUCTION

Live attenuated influenza vaccines (LAIVs) are licensed for prevention of seasonal influenza infection in children and adults (1–3). LAIVs bearing the six internal protein genes of the A/Ann Arbor/6/60 cold-adapted (caAA) donor virus and the hemagglutinin (HA) and neuraminidase (NA) genes from seasonal human influenza viruses are currently licensed in several countries for healthy individuals 2 to 49 years of age. We along with others have explored the potential use of the LAIV platform for pandemic influenza vaccines in some detail (4–6), and several pandemic LAIVs (pLAIVs) have been evaluated in phase 1 clinical trials. While the caAA-based pLAIVs were safe, their replication was highly restricted to the upper respiratory tract, and their immunogenicity was variable (7–11). Although the pLAIVs did not elicit a reliable serum antibody (Ab) response, we have now demonstrated that H5N1, H7N7, and H7N9 pLAIVs generated long-term immune memory that was revealed on administration of a dose of an antigenically matched pandemic inactivated subunit vaccine (pISV) (12–14). While these findings support the use of pLAIVs in the event of a pandemic, the immunologic events associated with pLAIV priming are not well understood.

In the case of H5N1 vaccines, we found that subjects who received two doses of an H5N1 pLAIV were primed to generate a rapid and robust neutralizing Ab response that was detectable as early as 7 days following administration of pISV more than 4 years later (12). The kinetics of this neutralizing Ab response suggests that the pLAIV generated long-lived memory B cells that were recalled upon pISV boost. We used sensitive HA-specific flow cytometric probes (15, 16) to enumerate H5-specific B cells in the peripheral blood of subjects who had received two doses of the H5N1 pLAIV followed by a single dose of H5N1 pISV (12) but could not detect a significant difference between H5N1 pLAIV-primed and unprimed subjects prior to pISV vaccination. Consequently, we hypothesized that H5-specific memory B cells induced by the pLAIV were localized to draining lymph nodes (LN). Because we could not readily test this hypothesis in humans, we utilized an African green monkey (AGM) model in which we had previously demonstrated that replication of intranasally (i.n.) administered H5N1 pLAIV was mostly restricted to the upper respiratory tract and that one dose did not elicit a serum neutralizing Ab response (17). We vaccinated three groups of 12 AGMs with the H5N1 pLAIV alone, H5N1 pISV alone, or a prime-boost combination of H5N1 pLAIV followed by H5N1 pISV 28 days later and examined the serological and antigen-specific B cell responses in peripheral blood and lymphoid tissues to define the mechanistic basis of pLAIV priming.

RESULTS

H5-specific memory B cells and plasmablasts in peripheral blood do not increase after pLAIV priming but increase after pISV boost.

Recombinant influenza virus HA probes (HAΔSA, for HA that does not bind sialic acid) have been utilized to detect and isolate HA-specific memory B cells following DNA-pISV prime-boost vaccination (15, 16). Using well-characterized H5-specific probes (15), we sought to determine whether long-lived H5-specific memory B cells could be detected in stored peripheral blood mononuclear cells (PBMCs) from H5N1 pLAIV-primed human subjects before or after H5N1 pISV boost. We compared samples collected on days 0, 7, and 28 following H5N1 pISV (A/Vietnam/1203/04 [VN/04]) boost from two cohorts of subjects primed with H5N1 pLAIV and one cohort of subjects primed with a mismatched H7N3 pLAIV approximately 4.5 years earlier and from two cohorts of unprimed subjects (Fig. 1a).

FIG 1.

Frequency of H5-specific plasmablasts and memory B cells in PBMCs from vaccine recipients following H5N1 pLAIV prime and pISV boost vaccinations. PBMC samples at days 0, 7, and 28 post H5N1 pISV (45 μg) were obtained from subjects who were previously vaccinated with either (a, row 1) two doses of A/Vietnam/1203/04 (H5N1) pLAIV (4 to 6 weeks apart, n = 11; blue circles in b and c), (a, row 2) two doses of A/Hong Kong/213/2003 (H5N1) pLAIV (4 to 6 weeks apart, n = 10; red circles), (a, row 3) two doses of A/chicken/British Columbia/CN6/2004 (H7N3) pLAIV (4 to 6 weeks apart, n = 8; green circles), or one prior dose of H5N1 pISV (n = 20, brown circles) or who had no prior pLAIV (n = 20, purple circles). Short horizontal bars represent median values. PBMCs were measured for CD19⁺ CD27⁺ IgM⁻ IgG⁺ H5-specific memory B cells (b) or CD19⁺ CD27⁺ CD38⁺ CXCR5⁻ H5-specific plasmablasts (c). Samples were compared by either separate Mann-Whitney U tests (between groups) or Wilcoxon matched-pair rank tests (within each group). **, P < 0.05; ns, not significant.

Prior to the pISV administration (day 0), a modest frequency of H5-specific memory B cells was detected in both H5N1 pLAIV-primed and unprimed subjects (Fig. 1b). The moderate frequency of H5-specific memory B cells in subjects who were not exposed to H5N1 viruses or vaccines is likely a consequence of prior seasonal influenza vaccination and/or infection (18, 19). Notably, at this time point there was no significant difference in the frequencies of H5-specific memory B cells between the H5N1 pLAIV-primed and unprimed or between the H5N1 pLAIV-primed and H7N3 pLAIV-primed subjects (P > 0.05, separate Mann-Whitney U tests) (Fig. 1b). Therefore, H5-specific memory B cell frequencies in peripheral blood after pLAIV administration could not explain the observed differences in neutralizing Ab responses to subsequent pISV boost.

H5-specific B cell responses increased at day 7 post-pISV boost in cohorts that received a matched or mismatched pLAIV or that were unprimed (Fig. 1b and c) (P < 0.05, Wilcoxon matched-pairs signed-rank test), likely due to cross-reactive H5-specific memory B cells induced by prior seasonal influenza virus infection and vaccination (15, 20). Interestingly, H5-specific memory B cells but not H5-specific plasmablasts were moderately higher in H5N1 pLAIV-primed subjects than in unprimed subjects on day 7 after the receipt of pISV (P = 0.01 and P = 0.18 Mann-Whitney U test, respectively) (Fig. 1c). However, the frequency of H5-specific plasmablasts and memory B cells on day 7 post-pISV was not significantly higher in H5N1 pLAIV-primed (either A/Hong Kong/213/2003 [HK/03] or VN/04) subjects than in recipients of the mismatched H7N3 pLAIV (P > 0.05, Mann-Whitney U test), suggesting that despite modest increases in the level of H5-specific B cells following pISV boost, there was no clear signature that reflected H5N1 pLAIV priming in the peripheral blood either prior to or following pISV boost.

We considered three potential explanations for the observed differences in serum antibody responses in the H5N1 pLAIV recipients. The first was that the pLAIV induced CD4 T cell memory, which recalled the B cell response following pISV boost several years later. We have previously shown that although influenza virus-specific T cell responses increased following H5N1 pLAIV administration in 12 of 21 subjects (based on enzyme-linked immunosorbent spot [ELISPOT] assay), this increase did not correlate with antibody responses following pISV boost (21). The second possible explanation was that the pLAIV changed the clonal composition of the memory B cell pool. We have attempted without success to detect antigen-specific B cells from peripheral blood from human subjects following pLAIV administration (14, 22), so unfortunately we cannot comment on the effect of pLAIV on the clonal composition of the memory B cell pool. The third possibility was that H5-specific B cells generated following pLAIV administration were localized to local lymph nodes or tissues. We tested this hypothesis in nonhuman primates.

Reproducing the human pLAIV-pISV prime-boost serological responses in an AGM model.

We have previously demonstrated that AGMs provide a suitable model for evaluation of pLAIVs (17). However, it was not known whether a dose of H5N1 pLAIV administered via nasal atomizer would prime for an Ab response that could be boosted subsequently with pISV, thereby reproducing the results in humans. Therefore, we vaccinated 12 AGMs with one dose of H5N1 pLAIV intranasally (i.n.), followed by an intramuscular (i.m.) injection of H5N1 pISV (group 3). For controls, we vaccinated 12 AGMs with H5N1 pLAIV alone (i.n.) or pISV alone (i.m.) (Fig. 2a, groups 1 and 2, respectively). AGMs were serially bled, and four AGMs per group were euthanized at days 14, 35, and 56 after initial vaccination; lymphoid tissues were collected for B cell analysis.

FIG 2.

Serum neutralizing Ab response against H5N1 wt virus in AGMs. (a) Groups of AGMs received either H5N1 VN/04 pLAIV alone, H5N1 pISV alone, or H5N1 VN/04 pLAIV followed by pISV. The AGMs that received either vaccine alone were sham immunized with L15 medium at the indicated time points as follows: group 1, 12 AGMs received 10⁷ TCID₅₀ of H5N1 VN/04 pLAIV delivered i.n., and 8 AGMs at day 28 received L15 medium i.m. (mock) into the right deltoid muscle; group 2, 12 AGMs received L15 medium delivered i.n. (mock), and 8 AGMs at day 28 received H5N1 pISV (45 μg) i.m. into the right deltoid muscle; group 3, 12 AGMs received 10⁷ TCID₅₀ of H5N1 VN/04 pLAIV delivered i.n., and 8 AGMs at day 28 received L15 medium i.m. (mock) into the right deltoid muscle. Four AGMs were sacrificed in each group at days 14, 35, and 56 from the start of the study (as indicated by X), and blood and lymphoid tissues were collected. PBMCs and sera were collected at days 0, 7, 14, 28, 32, 35, 42, and 56. (b) Serum samples from AGMs vaccinated with pLAIV-mock, mock-pISV, and pLAIV-pISV were assayed for hemagglutination inhibition (HAI) antibodies using the A/Vietnam/1203/2004 (H5N1) wild-type virus and horse red blood cells. (c) Serum samples from AGMs vaccinated with pLAIV-mock, mock-pISV, and pLAIV-pISV were assayed for neutralizing Abs against the A/Vietnam/1203/2004 (H5N1) wild-type virus. (d and e) Serum samples from AGMs vaccinated with pLAIV-mock, mock-pISV, and pLAIV-pISV were assayed for IgG-specific and IgA-specific binding antibodies to recombinant A/Vietnam/1203/2004 HA protein. The arrows indicate when the pISV boost or L15 injection was administered. Horizontal bars represent median values.

Consistent with our previous findings (17), we observed very low or undetectable neutralizing antibodies to the H5N1 wild-type (wt) virus by hemagglutination inhibition (HAI) and microneutralization (MN) assays at day 28 following one dose of H5N1 pLAIV (geometric mean titer [GMT] of 18.3 by HAI and 10 by MN at day 28) (Fig. 2b and c). Low levels of H5-specific serum IgG but not IgA were detected by enzyme-linked immunosorbent assay (ELISA) at day 28 following H5N1 pLAIV administration (Fig. 2d and e). There were no detectable HAI, MN, or IgG/IgA binding Abs following a dose of pISV alone (Fig. 2b to e). In contrast, in pLAIV-primed AGMs there was an increase in Ab titers from day 28 to day 35 (7 days following H5N1 pISV boost), with HAI titers rising from a GMT of 20 to 134.5 and MN titers rising from a GMT of 10 to 86.6 (Fig. 2b and c). Both H5-specific IgG and IgA ELISA Ab titers were also increased following H5N1 pISV boost (Fig. 2d and e). These findings replicate our clinical observation that H5N1 pLAIV efficiently primes for Ab responses that can be detected 7 days following H5N1 pISV boost (12).

These findings prompted us to ask whether simultaneous vaccination with H5N1 pLAIV and pISV would be immunogenic. In four AGMs, coadministration of H5N1 pLAIV and pISV did not elicit a detectable MN Ab response at day 14 or 28 (data not shown). This highlights the importance of an interval between priming and boosting to generate a neutralizing Ab response.

Early expansion of H5-specific plasmablasts predict serum neutralizing Ab responses in prime-boost-vaccinated AGMs.

The frequency of plasmablasts produced following pLAIV administration may provide a measure of B cell priming. H5-specific plasmablasts were therefore quantified by ex vivo B cell ELISPOT assay or flow cytometric analysis of CD19⁺ CD20^low CXCR5^low CD27^+/− CD38⁺ B (where CD27^+/− indicates either CD27⁺ or CD27⁻) cells (23). A modest frequency of H5-specific plasmablasts was detected at day 14 post-pLAIV administration in influenza virus-naive AGMs by flow cytometry (Fig. 3a) and as H5-specific IgG and IgA antibody-secreting cells (ASCs) (P > 0.05, data not shown). In contrast, there were low or undetectable H5-specific plasmablasts by flow cytometry (Fig. 3b) (P > 0.05, Wilcoxon matched-pairs signed-rank test) as well as low or undetectable H5-specific IgA and IgG ASCs by ELISPOT assay in AGMs that received pISV alone (data not shown). In contrast, in H5N1 pLAIV-primed AGMs there was a significant increase in the level of H5-specific plasmablasts at days 32 and 35 (4 to 7 days following pISV boost; P < 0.05, Wilcoxon matched-pairs signed-rank test) (Fig. 3c) that were confirmed as H5-specific IgG and IgA ASCs (data not shown). Additionally, we detected an increase in H5-specific IgG and IgA specific ASCs in bone marrow at terminal time points following pLAIV priming and pISV boost (Fig. 4a and b, respectively), indicating that the prime-boost regimen may seed influenza virus-specific plasma cells in the bone marrow. Only a modest number of H5-specific IgG and IgA ASCs were observed at day 35 (day 7 post-pISV) in AGMs that received pISV alone. Thus, although only low levels of H5-specific plasmablasts are observed in the peripheral blood after pLAIV administration, the frequency of H5-specific plasmablasts and plasma cells increased significantly along with serum antibody titers following pISV boost.

FIG 3.

Frequency of H5-specific plasmablasts in peripheral blood. PBMC samples from AGMs vaccinated with either pLAIV-MOCK (a), mock-pISV (b), or pLAIV-pISV (c) were assayed for H5-specific CD20⁻ CD38⁺ CD27^+/− CXCR5⁻ plasmablasts. Blue arrows indicate pLAIV vaccination, and red arrows indicate pISV vaccination. **, P < 0.05; ns, not significant.

FIG 4.

Ex vivo bone marrow ELISPOT assay. Fresh bone marrow collected at necropsy from AGMs vaccinated with pLAIV-mock, mock-pISV, and pLAIV-pISV was assayed for either IgG-specific (a) or IgA-specific (b) spots against rHA H5 protein (A/Vietnam/1203/2004) at the indicated time points. Horizontal bars represent median values.

Dynamics of H5-specific memory B cell responses in the peripheral blood of AGMs following pLAIV and pLAIV-pISV vaccination.

As noted earlier, we were unable to detect a robust circulating H5-specific memory B cell response in H5N1 pLAIV-pISV vaccine recipients prior to pISV boost (Fig. 1b). Here, we investigated whether AGMs have a corresponding frequency of H5-specific CD19⁺ CD20⁺ CD27⁺ IgG⁺ memory B cells. We identified a clear but low-frequency population of H5-specific memory B cells as early as day 14 that continued to expand up to day 28 following pLAIV administration in some animals (Fig. 5a). The frequency of H5-specific memory B cells also expanded significantly after H5N1 pISV boost in pLAIV-primed animals (Fig. 5c) (P < 0.05, Wilcoxon matched-pairs signed-rank test). We could not detect any H5-specific memory B cells after pISV alone (Fig. 5b). This indicates that pLAIV induces a modest frequency of H5-specific memory B cells that can be expanded following pISV boost.

FIG 5.

Frequency of H5-specific memory B cells in peripheral blood. PBMC samples from AGM vaccinated with either pLAIV-mock (a), mock-pISV (b), or pLAIV-pISV (c) were assayed for H5-specific CD19⁻ CD20⁺ CD27⁺ IgD⁻ IgM⁻ IgG⁺ memory B cells. Blue arrows indicate pLAIV vaccination, and red arrows indicate pISV vaccination. (D) Proportions show Ig sequences found in blood at day (d) 28 and day 35, those found only at day 28, and those found only at day 35. (E) Phylogenetic tree of one antibody lineage isolated from PBMCs of animal 8033 at day 28 and day 35. **, P < 0.05; ns, not significant.

To understand the relationship between the pLAIV-primed B cell responses and the recall response following pISV boost, we amplified and sequenced the AGM immunoglobulin (Ig) heavy and light chains from single-sorted H5-specific memory B cells at day 28 (preboost) and day 35 (7 days postboost) from four AGMs that received pLAIV followed by pISV and annotated and characterized heavy-chain and light-chain sequences using a newly defined AGM Ig sequence database (24) (see Table S1 in the supplemental material). Among 306 Ig sequences isolated from PBMCs of four vaccinated AGMs, V_H3 and V_H4 were the predominant IgH families used (Tables S1 and S2), matching heavy-chain variable region (V_H) family usage based on alignment with the limited number of sequences in the reference database for this species (24). In AGMs 7837 and 8033, more than 50% of the H5-specific Ig sequences recovered at day 28 and day 35 overlapped. In contrast, AGMs 7692 and 7843 showed only modest overlap between H5-specific Ig sequences recovered at days 28 and 35 (Fig. 5d and S1 Table). The AGM Ig sequences demonstrated a mean somatic mutation frequency of 3.23 to 5.99% (range from 4 animals) from the inferred germ line sequence (Table S1), which is similar to that described in humans following influenza vaccination (25). However, there was no significant difference in the expansion of specific sequences from germ line Ig, CDR3 (complementarity determining region 3) length, or mutation frequency of the matched Ig between days 28 and 35 in any of the AGMs (Table S1). Interestingly, distinct clones were evident in all AGMs at day 35, suggesting that pISV may induce additional new clones. Moreover, the H5-specific V_H sequences isolated from day 28 and day 35 in the same clonal lineage were equally distributed in the phylogenetic trees without evidence of distinct evolution (Fig. 5e) and even appeared within groups of genetically identical variable genes, potentially indicating clonal expansion in the periphery. This suggests that pLAIV-primed H5-specific memory B cells are likely coming from active germinal center (GC) reactions and that peripheral expansion of these clonotypes likely contributes to the circulating H5-specific memory B cell pool following pISV boost.

pLAIV generates H5-specific plasmablasts in the MLN.

Primary immunization with H5N1 pLAIV induced only a modest frequency of H5-specific plasmablasts in the peripheral blood; thus, we sought to identify the lymph nodes in which the H5-specific plasmablasts are generated. AGMs were sacrificed at days 14, 35, and 56 after pLAIV administration, and lymphoid tissues associated with the upper respiratory tract (tonsil-like tissue and submandibular and cervical LN), lower respiratory tract (mediastinal LN [MLN]), periphery (ipsilateral and contralateral axillary LN), and spleen were collected. Tissues were processed as single-cell suspensions, stained, and analyzed for H5-specific B cells by flow cytometry. In AGMs vaccinated with H5N1 pLAIV, we found that H5-specific plasmablasts were localized mostly to the MLN at day 14, with frequencies greater than 2% in four of eight animals (Fig. 6a). This frequency declined over time, with a detectable plasmablast response in the MLN at days 35 and 56 in only one of four AGMs. AGMs that received pISV alone had very low to undetectable H5-specific plasmablasts in LN and tissues; only one of four AGMs had an expansion of H5-specific plasmablasts detected in the ipsilateral axillary LN at day 7 post-pISV alone (Fig. 6b). In contrast, following pISV boost, pLAIV-primed AGMs had high levels of H5-specific plasmablasts in the MLN (mean frequency of H5-specific plasmablasts of 5.07%), ipsilateral LN (mean frequency of H5-specific plasmablasts of 11.92%), contralateral axillary LN (mean frequency of H5-specific plasmablasts of 7.33%), and spleen (mean frequency of H5-specific plasmablasts of 13.88%) at day 35 (7 days post-pISV boost) (Fig. 6c). By day 56 (28 days post-pISV boost), the H5-specific plasmablast response was detectable only in the ipsilateral axillary LN (Fig. 6c). These findings suggest that in prime-boost-vaccinated AGMs, H5-specific plasmablasts are likely generated in the MLN following pLAIV administration and are detected in both axillary LN and spleen following pISV boost.

FIG 6.

Frequency of H5-specific plasmablasts in lymphoid tissues. AGMs vaccinated with pLAIV-mock (a), mock-pISV (b), or pLAIV-pISV (c) were assayed for H5-specific CD19⁺ CD20⁻ CD27^+/− CD38⁺ CXCR5⁻ plasmablasts in submandibular, cervical, mediastinal, ipsilateral, and contralateral axillary LN, spleen, tonsil-like tissue, bone marrow, and blood at days 14, 35 and 56. Note that the day 14 time point from the pLAIV-pISV group is the equivalent to day 14 from the pLAIV-mock group, and the results are therefore shown together (green and red, respectively, in panel a). Horizontal bars represent median values.

H5-specific memory B cells also originate in the MLN following pLAIV administration.

To identify the proportion and localization of H5-specific memory B cells in tissues following primary pLAIV vaccination, we measured H5-specific CD19⁺ CD20⁺ CD27⁺ IgG⁺ memory B cells in the MLN and compared them with those found in the peripheral blood. Consistent with the H5-specific plasmablast data, we identified large frequencies of H5-specific memory B cells in the MLN at day 14 (five of eight AGMs), day 35 (two of four AGMs), and day 56 (two of four AGMs) post-pLAIV administration (Fig. 7a) (based on a cutoff 3-fold above the frequency at day 0 of peripheral blood H5-specific memory B cell frequencies). AGM Ig sequence analysis demonstrated a relationship between H5-specific memory B cells found in PBMCs and MLN following pLAIV administration (Fig. 7d and Table S2). In contrast, we could not detect H5-specific memory B cells in lymphoid tissues of AGMs that received pISV alone (Fig. 7b). H5-specific memory B cells were weakly expanded in lymphoid tissues following pISV boost in H5N1 pLAIV-primed AGMs, with a low frequency detected in the spleen (three of four AGMs) and MLN (one of four AGMs) at day 35 (7 days post-pISV boost) and ipsilateral axillary LN (four of four AGMs) at day 56 (day 28 post pISV boost) (Fig. 7c). One H5-specific B cell clonal lineage contained members isolated from the periphery and MLN (lineage 8), including 5 immunoglobulin sequences from PBMC and 14 from MLN isolated at day 14. Interestingly, phylogenetic analysis revealed that the V_H sequences of all five immunoglobulin sequences isolated from PBMCs shared 100% amino acid sequence identity and were 4% mutated from the inferred unmutated common ancestor (UCA). In contrast, the 14 immunoglobulin sequences isolated from MLN shared 95 to 98% amino acid sequence identity (Fig. 7e) and were 4 to 7% mutated from the UCA. These immunoglobulin sequence data suggest that, following vaccination, somatic hypermutation occurred mainly in the lymph node, whereas clonal memory B cells primarily expanded in peripheral blood.

FIG 7.

Frequency of H5-specific memory B cells in lymphoid tissues. AGMs vaccinated with either pLAIV alone (a, red), pISV alone (b), or pLAIV-pISV (c) were assayed for H5-specific CD19⁻ CD20⁺ CD27⁺ IgD⁻ IgM⁻ IgG⁺ memory B cells in submandibular, cervical, mediastinal, ipsilateral, and contralateral axillary LN, spleen, tonsil-like tissue, bone marrow, and blood at days 14, 35 and 56. Note that the day 14 time point from the pLAIV-pISV group is equivalent to day 14 from the pLAIV group, and the results are therefore shown together (green and red, respectively, in panel a) (D) Comparison of H5-specific memory B cell clones isolated from the peripheral blood or mediastinal lymph nodes from animals 7538 or 7796 vaccinated with pLAIV only. Proportions show those found in both blood and MLN, those found only in MLN, and those found only in blood. (E) The phylogenetic tree of one antibody lineage isolated from PBMCs and MLN of animal 7538 at day 14. Horizontal bars represent median values.

Expansion and phenotype of GC B cells in mediastinal and axillary lymph nodes.

To gain a better understanding of the GC phenotype of H5-specific B cells generated in the MLN following pLAIV administration and ipsilateral axillary lymph nodes following pISV boost, we measured the expression of Ki67 and Bcl-6 in lymphoid tissues and blood as previously described (26). Characterization of H5-specific memory B cells in AGMs following administration of pLAIV alone demonstrated a high frequency of a phenotype related to GC B cells (Ki67⁺ Bcl6⁺), with very few non-GC B cells (Ki67⁻ Bcl6⁻) (Fig. 8a and b). In contrast, H5-specific B cells detected in pLAIV-primed animals following pISV boost demonstrated a phenotype related to both GC-B cells (Ki67⁺ Bcl6⁺) and non-GC B cells (Ki67⁻ Bcl6⁻), with the highest frequency of H5-specific GC B cells found in the ipsilateral axillary LN (Fig. 8a and c). Taken together, these data indicate that H5-specific GC B cells are generated in the MLN following H5N1 pLAIV administration, and subsequent parenteral administration of pISV expanded primed H5-specific GC and non-GC B cells in the local axillary LN and peripheral blood.

FIG 8.

Frequency of GC B cells and extrafollicular B cells in lymph nodes following pLAIV and pLAIV-pISV administration. (a) Representative plots of H5-specific CD19⁺ CD20⁺ B cells stained for intracellular Ki67 and Bcl6 in mediastinal lymph node and ipsilateral axillary lymph node. (b and c) Four AGM vaccinated with either pLAIV alone (day 35 or 56) or pLAIV-pISV (day 28) were assessed for H5-specific CD19⁺ CD20⁺ Ki67⁺ Bcl6⁺ memory B cells in the mediastinal, ipsilateral, and contralateral axillary LN, spleen, and blood.

Recently Lau et al. defined a population of influenza virus-specific CD27⁺ CD21⁻ B cells generated during the peak of GC reactions following seasonal inactivated vaccine (27). Similarly, others have shown that memory B cells can be defined into distinct subsets based on their differential CD21 and CD27 expression levels (28–30). These populations include CD27⁻ CD21⁺ naive B cells, CD27⁺ CD21⁺ resting memory B cells, CD27⁺ CD21⁻ activated memory B cells, and CD27⁻ CD21⁻ “tissue-like” memory B cells. The last, tissue-like memory B cells, have been shown to develop during chronic HIV and other persistent infections and represent a population of B cells that bear many of the features of exhausted B cells (29). To determine whether GC B cells generated after pLAIV-pISV administration represent a similar expansion, we analyzed CD27 and CD21 levels on H5-specific class-switched B cells in the peripheral blood of AGMs following administration of pLAIV alone or pLAIV-pISV vaccination (Fig. 9a). We found a modest increase in CD27⁺ CD21⁻ (activated memory) and CD27⁺ CD21⁺ (resting memory) H5-specific B cells at day 14 post pLAIV administration (Fig. 9b), which was dramatically expanded at day 35, 7 days following pISV boost (Fig. 9c). Interestingly, we also observed an increase in the CD27⁻ CD21⁻ (tissue-like) population of H5-specific B cells that was not observed following seasonal influenza vaccination (27). To determine the source of these H5-specific B cells, we analyzed CD21 and CD27 expression on H5-specific class-switched B cells in lymphoid tissues. H5-specific B cells following primary pLAIV vaccination were mostly CD27⁺ CD21⁺ (resting memory B cells) at days 14, 35, and 56 (Fig. 9d) though some CD27⁻ CD21⁺ cells were also expanded at days 35 and 56 (Fig. 9e and f). However, in pLAIV-primed animals that were boosted with pISV, we observed an increase in CD27⁻ CD21⁻ (tissue-like memory) and CD27⁺ CD21⁻ (activated memory) cells in the ipsilateral axillary lymph node at both day 7 and day 28 post-pISV boost and in CD27⁺ CD21⁺ (resting memory) cells at day 28 post-pISV boost (Fig. 9g and h). These markers allowed us to identify distinct patterns, with primary pLAIV administration eliciting a predominant CD27⁺ CD21⁺ resting memory cell population and pISV boost eliciting a CD27⁺ CD21⁻ activated memory cell population.

FIG 9.

CD21 expression on B cells following H5N1 pLAIV-pISV vaccination. (a) Representative plot of class-switched B cells (CD19⁺ CD20⁺ IgD⁻) gated for either CD27⁺ CD21⁺ (resting memory), CD27⁺ CD21⁻ (activated memory), or CD27⁻ CD21 (tissue-like memory) B cells. (b and c) Frequency of H5-specific B cells that express either CD27⁻ CD21⁻, CD27⁺ CD21⁻, or CD27⁺ CD21⁺ at the indicated days (0, 7, 14, 28, 32, 35, and 56) following initial vaccination with pLAIV or pLAIV-pISV in peripheral blood. (d to h) Frequency of H5-specific B cells that express either CD27⁻ CD21⁻, CD27⁺ CD21⁻, or CD27⁺ CD21⁺ days 14, 35, or 56 following initial vaccination with pLAIV (d to f) or days 35 (day 7 postboost) and (H) 56 (day 28 postboost) (g and h) in lymphoid tissues of pLAIV-primed animals post-pISV boost. Bars represent means ± standard errors of the means.

DISCUSSION

Pandemic LAIVs clearly induce long-term immune memory in human subjects, which can be recalled by administration of a matched inactivated subunit vaccine to generate a robust neutralizing Ab response (12–14). The rapid kinetics and magnitude of the antibody response following pISV boost suggest that pLAIV generates memory B cell responses. However, expansion of HA-specific memory B cells by pLAIV was not directly observed in peripheral blood in humans. Therefore, we turned to a nonhuman primate (AGM) model that permitted a detailed analysis of the H5-specific memory B cell responses in systemic and local lymphoid tissues following pLAIV prime-pISV boost. We found that pLAIV elicited robust GC B cell responses localized largely in the MLN. Subsequent intramuscular administration of pISV led to expansion and mobilization of H5-specific memory B cells in a wide range of lymphoid tissues and the peripheral blood. Additionally, we observed a robust H5-specific neutralizing Ab response in the peripheral blood of AGMs, recapitulating our clinical observations.

The localization of pLAIV-elicited GC reactions within the MLN rather than submandibular or cervical LN was surprising because the pLAIV is a temperature-sensitive virus and therefore is expected to be restricted to replication in the upper respiratory tract of AGMs. While it is possible that some drainage of vaccine antigen into the lungs could occur because the animals are anesthetized for handling and vaccine administration, vaccine virus replication was largely restricted to the upper airways (17). We could detect only low virus titers and inconsistent replication in the lungs of AGMs on the first day following intranasal administration of H5N1 pLAIV (17). Infection of anesthetized mice and NHPs with wild-type influenza viruses has been shown to generate influenza virus-specific memory B cells in the MLN (31, 32), but in these models, replication occurs in the lungs as well as the upper airways. Our findings are consistent with a recent report from Pichyangkul et al. where repeated infection of rhesus macaques with influenza A/California/07/2009 (H1N1pdm09) virus led to high concentrations of memory B cells in the MLN detected by ELISPOT assay (33). Active replication of the vaccine virus in the respiratory tract was critical for the induction of H5-specific B cell responses because AGMs given an equivalent dose of UV-inactivated H5N1 pLAIV showed no detectable H5-specific memory B cell responses in the peripheral blood or any lymphoid tissues, including the MLN (data unpublished).

The influenza virus-specific B cell pool in humans is shaped by a lifetime of influenza exposure from vaccination and infection. This results in a pool of highly cross-reactive memory B cells that, with CD4 T cell help, can recognize both circulating seasonal and avian-origin influenza viruses, with varied frequency (16, 21). Prior exposure to influenza is critical for the recall response upon subsequent vaccination or infection, but it complicates the study of human influenza virus-specific B cell responses in the peripheral blood. In contrast, because AGMs are influenza-naive they lack the pool of cross-reactive B cells, making analysis of the peripheral blood easier to interpret, and, more importantly, we are able to sample lymphoid tissues at different time points after primary or prime-boost vaccination. The use of the AGM model was key to our finding that pLAIV induces a pool of memory B cells in the mediastinal lymph node that can be recalled following pISV boost. The need for priming is well recognized in preclinical studies of inactivated influenza vaccines; ferrets require prior influenza virus infection to generate neutralizing antibodies following seasonal trivalent influenza vaccine (TIV) (34). Thus, although the immune history of influenza virus infections and vaccination in humans cannot be recapitulated in animal models, going back from the bedside to the bench has been very illuminating.

H5-specific memory B cells in the MLN were clonally related to those in the peripheral blood yet demonstrated markedly higher diversity than that of the clonally expanded peripheral blood memory B cells. An increase in HA-specific memory B cells was demonstrated in the tonsils of children following seasonal LAIV administration, suggesting that this may also be a site of memory B cell induction (35, 36). However, AGMs do not have discrete tonsils, so we sampled tonsil-like tissue but could detect only modest levels of CD19⁺ B cells. Our findings have important implications for both pandemic and seasonal LAIVs. Although ISVs induce higher rates of seroconversion than LAIVs (37–39) and although an HAI antibody response following seasonal LAIV is associated with vaccine efficacy, LAIV efficacy has been documented even when seroconversion rates were low (40). Our data support the notion that the route of vaccination influences the site at which immune responses can be detected and generated. Indeed, a higher proportion of extrafollicular H5-specific B cells are present in numerous tissues following pISV boost than after initial pLAIV vaccination. Furthermore, H5-specific GC B cell responses following pISV boost were significantly smaller in the ipsilateral axillary lymph nodes than in the mediastinal lymph nodes following primary pLAIV administration. This may reflect the poor immunogenicity of the unadjuvanted H5 subunit vaccine and its inability to sustain GC-formation. The strongest response following pISV in the ipsilateral axillary lymph node may indicate an ability to target responses to distinct sites based on targeted intramuscular vaccination. Our findings in the AGM models suggest that enumeration of B cells or serological responses in the peripheral blood yields an incomplete picture of LAIV immunogenicity.

Pandemic influenza vaccines based on other platforms, including recombinant HA (41, 42), DNA (43, 44), and recombinant adenovirus-expressed HA (rHA) (45), also prime for robust Ab and B cell (16) responses following subsequent pISV, even in individuals who had undetectable Ab responses following the initial vaccination. These observations are similar to our observations with pLAIV-pISV prime-boost (12–14). Understanding the mechanisms underlying these prime-boost strategies is critical for improving influenza vaccines. Clinical trial results suggest that the quality and magnitude of antibody responses are influenced by the interval between priming and boosting (43, 44) though the optimal interval may be specific to the vaccine platforms used. We have recently found that an interval of 4 weeks between H7N9 pLAIV and pISV administration was sufficient to elicit a robust neutralizing antibody response in humans (K. R. Talaat, C. J. Luke, K. Chang, B. Plunkett, P. Grier, W. Sun, R. Adkinson, M. Eichelberger, Z. Chen, H. Jin, M. Levine, J. Treanor, K. Coelingh, and K. Subbarao, unpublished data) and therefore used this interval in AGMs. Simultaneous administration of pLAIV-pISV to AGMs demonstrated that although an interval between prime and boost is important for generating neutralizing antibodies, it is not required to generate and clonally expand H5-specific memory B cells (data not shown). The reason why a particular interval is important in generating more durable neutralizing antibody responses following boost is still under debate and requires further evaluation.

Recent studies have demonstrated that the inclusion of an adjuvant greatly enhances the ability of inactivated influenza vaccines to induce neutralizing antibodies (46, 47). This has been credited to the ability of particular adjuvants to induce higher levels of T follicular helper (Tfh) cells to promote GC formation (48). The CD4⁺ T cell response likely also plays an important role in pLAIV priming. It is likely that pLAIV induces Tfh cells that coordinate GC reactions and somatic hypermutation in the MLN and ipsilateral axillary LN following pISV boost. In mice, LAIVs induce a robust T cell response in local draining LN (49) and in the lungs (50), but in humans we previously demonstrated only a modest population of T cells in the peripheral blood following pLAIV administration (21).

In this study, we demonstrate that intranasal administration of pLAIV generates a robust antigen-specific GC-B cell response that can be recalled following pISV boost into GCs at numerous distinct sites. There are some caveats to our study. (i) Due to the limited amount of available H5N1 pISV, we could not include a group that was vaccinated with two doses of H5N1 pISV. However, it is clear that 28 days following pISV alone we could not detect evidence of significant priming in local axillary LN or even serological signs of priming. This finding is consistent with an earlier study in which naive macaques given two doses of seasonal ISV (28 days apart) failed to develop a significant vaccine-specific neutralizing Ab response and also did not demonstrate a significant boost following homologous H1N1 infection, indicating a lack of immunological priming (51). In contrast, pLAIV induced a significant H5-specific memory B cell response in the mediastinal LN. (ii) We cannot be certain that the observed changes in the B cell repertoire following pLAIV administration were responsible for the rise in serum neutralizing Ab titers following pISV boost. Qualitative differences in the composition of influenza virus-specific B cell clones prior to seasonal influenza vaccination have been linked with seroconversion (52–54). Unfortunately, we were limited in our ability to assess B cell clonality in AGMs before and after pLAIV administration. In the limited set of samples from our study, we found some shared clones following administration of pLAIV alone and pLAIV-pISV vaccination and some unique clones following pISV boost suggesting expansion of H5-specific memory B cells as well as recruitment of new clones following pISV boost. Comparison of clones between pISV- and pLAIV-primed AGMs was difficult because H5-specific B cells were not detected following a dose of pISV alone. (iii) Memory CD4 T cells provide help in recruitment and generation of antibodies following pISV boost. Unfortunately, due to technical difficulties, we were unable to detect influenza virus-specific T cells in AGMs by flow cytometry. Recent data from Mohn et al. suggest that ELISPOT assay techniques may be more useful for detection of low frequencies of antigen-specific T cells; they detected CD8 T cells following seasonal LAIV administration in children (55) and transient induction of CD4 T cells following seasonal LAIV administration in adults (56). In contrast, we have been able to detect transient increases in influenza virus-specific T cells by ELISPOT assay following pLAIV administration only in adults (21) and speculate that the frequency of cross-reactive T cells that recognize pandemic influenza viruses may be below the sensitivity of the assay.

In summary, we demonstrate that intranasally administered pLAIV elicits a highly localized and somatically hypermutated GC B cell response in the MLN that is rapidly recalled following pISV boost into GC reactions at distant immune sites, most notably the ipsilateral axillary LN. These results provide important new evidence for the mechanisms that may determine the success of prime-boost strategies in humans and bring us closer to understanding why serum antibody is not a reliable correlate of immunity for seasonal and pandemic LAIVs.

MATERIALS AND METHODS

Experimental design.

The objective of this study was to understand the immunological mechanisms underlying the pLAIV prime-pISV boost strategy, and we used the AGM model because the findings in AGMs paralleled findings in humans more closely than those in models using smaller animals such as mice and ferrets (17). We hypothesized that vaccine-specific memory B cells induced by pLAIV were localized to draining lymph nodes (LN). Due to the ethical considerations for the use of nonhuman primates in research, we did not use a power analysis to calculate sample size. However, we selected group sizes (n = 4) of AGMs at each necropsy time point that would provide robust data in outbred animals. Thus, 36 AGMs were randomly assigned by animal number to receive the H5N1 pLAIV alone, H5N1 pISV alone, or H5N1 pLAIV-pISV prime-boost. An additional four AGMs were vaccinated with UV-inactivated H5N1 and four AGMs were vaccinated with both H5N1 pLAIV-pISV simultaneously. We did not alter this number during the course of the study. Any data considered outliers are identified in figure legends and excluded from analysis. Experiments were performed without blinding, and each experiment represents a single replicate, with four sampling replicates of tissue or blood collected at each time point.

The clinical samples that were tested as part of this research were from phase I, open-label clinical trials. The inclusion and exclusion criteria were described in the primary reports of the studies (11, 12).

Clinical study.

Stored peripheral blood mononuclear cells (PBMCs) were obtained from a previously reported study in which healthy adult subjects (n = 17) received two doses of H5N1 HK/03 pLAIV (A/Hong Kong/213/2003), 28 days apart (ClinicalTrials.gov registration number NCT00488046) (11). PBMCs were collected at days 0, 7, and 28 days after the first dose and 7 (day 35) days after the second dose.

Stored PBMCs were obtained from a previously reported clinical trial in which subjects who had previously received pLAIV were recalled several years later and were vaccinated with monovalent influenza A(H5N1) pISV (Sanofi Pasteur) (ClinicalTrials.gov registration number NCT01443663) (12). Briefly, subjects who had previously received two doses of H5N1 VN/04 pLAIV (A/Vietnam/1203/04, group 1, n = 11), H5N1 HK/03 pLAIV (group 2, n = 10), and H7N3 BC/04 pLAIV (A/chicken/British Columbia/CN6/2004, group 3, n = 8, representing an irrelevant subtype pLAIV) received a single 45-μg intramuscular dose of H5N1 pISV (12). Forty subjects received a single intramuscular dose (group 4, n = 20) or two intramuscular doses of H5N1 pISV (45 μg, n = 20) 28 days apart. Blood was collected, and PBMCs were stored from subjects at 0, 7, and 28 days post-H5N1 pISV vaccination (as shown in schematic Fig. 1a).

Viruses and vaccine.

As previously described, the hemagglutinin (HA) and neuraminidase (NA) genes of the cold-adapted pLAIV viruses were derived from avian influenza A/Vietnam/1203/2004 (H5N1), A/Hong Kong/213/2003 (H5N1), or A/chicken/British Columbia/CN6/2004 (H7N3) virus. The H5 HA had an engineered deletion of the multibasic cleavage site, and all of the vaccine viruses were generated by reverse genetics (57). Viruses were propagated in 10- to 11-day-old embryonated chicken eggs. Virus was titrated in Madin-Darby canine kidney (MDCK) cells maintained in modified Eagle's medium with 10% heat-inactivated fetal calf serum (HyClone) and l-glutamine (Gibco). The monovalent subvirion vaccine influenza rgA/Vietnam/1203/2004 H5N1 pISV (A/Vietnam/1203/2004) was obtained through the NIH Biodefense and Emerging Infections Resources Repository (BEI), NIAID, NIH (catalog number NR-4143).

Ethics statement.

Stored peripheral blood mononuclear cells (PBMCs) were obtained from a previously reported open-label phase 1 study. Clinical protocols were reviewed and approved by Western IRB. Written informed consent was obtained from all study subjects.

Embryonated specific-pathogen-free chicken eggs, 10 to 11 days old, were purchased from a vendor (Charles River Laboratories, North Franklin, CT) and used to grow influenza virus.

This study was carried out in strict accordance with the recommendations described in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, the Office of Animal Welfare, and the U.S. Department of Agriculture. All animal work was approved by the NIAID Division of Intramural Research Animal Care and Use Committees (IACUC), in Bethesda, MD (protocol LID3E). Animals were housed individually within isolators and provided with food twice a day and water ad libitum. They were provided enrichment daily in the form of numerous types of food items and had access to toys as enrichment within their cages at all times. All procedures were performed under sedation with ketamine (10 to 25 mg/kg) administered intramuscularly with minimal pain and distress to the animals. Animals were euthanized with an overdose of phenobarbital (20 to 25 mg/kg) administered intravenously. The animal facility is accredited by the American Association for Accreditation of Laboratory Animal Care.

Virus inoculation and tissue collection.

Twenty-four male and female adult AGMs (Chlorocebus aethiops) were randomly selected and inoculated intranasally (i.n.) with 1 × 10⁷ 50% tissue culture infective doses (TCID₅₀) in 0.4 ml (0.2 ml in each naris) of H5N1 pLAIV (A/Vietnam/1203/2004) using a mucosal atomization device (Teleflex), and 12 AGMs were mock vaccinated with 0.4 ml (0.2 in each naris) of Leibovitz-15 (L15) medium (Gibco). Twenty-eight days later, 12 of the H5N1 pLAIV-immunized AGMs and the 12 mock-vaccinated AGMs were given a single dose of H5N1 pISV (45 μg) intramuscularly in the right deltoid muscle, and the remaining 12 H5N1 pLAIV-immunized AGMs were given a single dose of L15 as a sham control for pISV. This study represents only one replicate, and statistical methods were not used to estimate sample size. The AGM experiments were performed without blinding. A summary of the vaccination regimen is given in Fig. 2a. In a separate experiment, four AGMs received 1 × 10⁷ TCID₅₀ in 0.4 ml (0.2 ml in each naris) of H5N1 pLAIV (A/Vietnam/1203/2004) using a mucosal atomization device (Teleflex) and a single dose of H5N1 pISV (45 μg) intramuscularly (i.m.) in the right deltoid muscle, simultaneously. The AGMs were euthanized at day 28 postvaccination, and tissues were collected as described below. In an additional study, four AGMs received 1 × 10⁷ TCID₅₀ in 0.4 ml (0.2 ml in each naris) of UV-inactivated H5N1 pLAIV (A/Vietnam/1203/2004) using a mucosal atomization device (Teleflex). AGMs were euthanized at day 14 postvaccination, and tissues were collected as described below.

Blood (sodium heparin tubes) and serum samples (serum separator tubes) were collected at days 0, 7, 14, 28, 32, 35, 42, and 56 from the start of the study. At days 14, 35, and 56, 4 AGMs from each group were euthanized, and tissue samples, including cervical LN, submandibular LN, mediastinal LN, ipsilateral and contralateral axillary LN, spleen, tonsil-like tissue, and bone marrow (EDTA tubes) were collected. Tissues were divided, and a small piece was transferred to cryomolds containing optimal cutting temperature (OCT) medium and snap-frozen on dry ice. Frozen samples were transferred to −80°C for storage until sectioning for confocal microscopy. Attempts to identify antigen-specific germinal centers using rHA probes were unsuccessful. The remaining tissue was further processed by passing tissues through a 70-μm-pore-cell cell strainer and washed into R10 medium, which contains RPMI medium (Life Technologies) and 10% heat-inactivated fetal calf serum (HyClone), penicillin-streptomycin, and l-glutamine (Gibco), and 10 U/ml of DNase (Roche). Blood, bone marrow, and splenocytes were processed by layering cells over a Ficoll gradient (GE Healthcare). Cells were washed two times with R10 medium and counted for both number and viability (Cellometer). Cells were resuspended in freeze mix containing 10% dimethyl sulfoxide (Sigma) and 90% heat-inactivated fetal calf serum (HyClone) and quickly transferred to cryostorage containers, where they were stored at −80°C overnight and subsequently transferred to liquid nitrogen for long-term storage.

MN assay.

Heat-inactivated serum (56°C for 1 h) collected at least weekly was tested for neutralizing activity against wild-type (wt) influenza A/Vietnam/1203/2004 (H5N1) virus in a microneutralization (MN) assay as previously described (58, 59). Briefly, 2-fold dilutions of heat-inactivated serum were tested for the presence of antibodies capable of neutralizing the infectivity of 100 TCID₅₀ of virus in MDCK cells (ATCC). The serum dilution that completely prevented cytopathic effect in 50% of the wells on day 4 was calculated by the Reed-Muench formula (60).

HAI assay.

Serum was treated with receptor-destroying enzyme (RDE; Seiken Denka) and tested for hemagglutination inhibition as previously described (61). Briefly, 8 HA units of wild-type H5N1 influenza virus in a V-bottom 96-well plate was incubated with serial dilutions of serum, and the dilution of serum that inhibited agglutination of a 1% suspension of horse red blood cells (Lampire Biological Laboratories, Pipersville, PA) was recorded.

ELISA IgG or IgA.

ELISAs were performed as previously described (51, 62). Ninety-six-well, flat-bottom MaxiSorp plates (Nunc) were coated overnight at 4°C with 1 μg/ml of recombinant H5 HA protein from A/Vietnam/1203/2004 (FR-39; obtained from the Influenza Reagent Resource). The plates were blocked with phosphate-buffered saline (PBS) plus 10% bovine serum albumin (BSA; Sigma) for 2 h at room temperature. The plates were washed at least four times with PBS with 0.01% Tween 20 (PBST). Heat-inactivated serum samples were diluted in PBS with 5% BSA (20-10240; 4-fold dilutions) and incubated on 96-well plates for 2 h at room temperature. Plates were washed six times with PBST and incubated with either a goat anti-monkey IgG gamma peroxidase-conjugated Ab (Rockland) or goat anti-monkey IgA(α) peroxidase-conjugated Ab (Rockland) diluted in PBS with 5% BSA for 1 h at room temperature. Subsequently, the plates were washed six times with PBST, incubated with TrueBlue solution (KPL), and allowed to develop for 10 min. Color development was stopped with stop solution (KPL) and read on a microplate reader immediately at 650 nm. Ab titers were expressed as the reciprocal dilution giving an optical density (OD) reading 3 times that of background (serum without protein).

Ex vivo B cell ELISPOT.

A direct ELISPOT assay to determine the frequency of Ab-secreting cells (ASCs) present in PBMC or bone marrow samples was performed as previously described (63). Briefly, 96-well Multiscreen-HA 96-well plates (Millipore) were incubated with either recombinant H5 HA protein from A/Vietnam/1203/2004 (H5N1) (FR-39; provided by the Influenza Reagent Resource) at a concentration of 2 μg/ml or Ig Ab at 5 μg/ml (Rockland) in triplicate overnight. Plates were washed and incubated with R10 medium at 37°C for 2 h. Plates were removed, 5 × 10⁶ cells/ml of PBMCs or bone marrow cells in R10 medium were added to the top well, and 3-fold dilutions were performed. Plates were incubated at 37°C in 5% CO₂ overnight. Plates were removed from the incubator and washed repeatedly with PBS/PBST. Diluted anti-monkey IgG, IgA, or IgM biotin-conjugated antibody (Rockland) was added to appropriate wells and incubated for 2 h at room temperature. Plates were washed repeatedly with PBS/PBST and incubated with diluted horseradish peroxidase (HRP)-avidin D (Vector Laboratories) for 1 h at room temperature. Plates were washed repeatedly with PBS/PBST and incubated for 10 min with ELISPOT substrate (BD Biosciences). Plates were washed repeatedly with distilled water and left overnight to dry before being read on an automated ELISPOT reader (Immunospot).

HA-specific B cell phenotyping and sorting.

HA probes were previously generated and validated for use in characterizing HA-specific B cells in PBMC samples (15, 16). HA-specific memory B cells and plasmablasts were identified using cryo-preserved PBMCs or processed tissue samples using recombinant H5 (A/Vietnam/1203/2004) conjugated to streptavidin-allophycocyanin (APC) (catalog number S-32362; Life Technologies) (26). Cell viability was determined using Aqua Live/Dead (L34957; Life Technologies). Human PBMCs were stained with the following antibodies: CD8 BV510 (catalog no. 301048, clone RPA-T8; Biolegend), CD3 BV510 (catalog no. 317332, clone OKT3; Biolegend), CD56 BV510 (catalog no. 318340, clone HCD56; Biolegend), CD14-BV510 (catalog no. 301842, clone M5E2 Biolegend), CD19–energy-coupled dye (ECD) (catalog no. IM2708U, clone J3-119; Beckman Coulter), CD27 BV605 (catalog no. 302830, clone O323; Biolegend), CD11c BV650 (catalog no. 563404, clone B-ly6; BD), CD38 AF700 (catalog no. 303524, clone HIT2; Biolegend), CD21-phycoerythrin (PE)-Cy5 (catalog no. 551064, clone B-ly4; BD), IgG-BV421 (catalog no. 562581, clone G18-145; BD), IgM-peridinin chlorophyll protein (PerCP)-Cy55 (catalog no. 561285, clone G20-127; BD), CD95 BV711 (catalog no. 563132, clone DX2; BD), CD22 PE-Cy7 (catalog no. 302514, clone HIB22; Biolegend), CXCR5 APC-Cy7 (catalog no. 356926, clone J252D4; Biolegend). AGM PBMCs were stained with the following antibodies: CD14-BV510 (catalog no. 301842, clone M5E2; Biolegend), CD19-ECD (catalog no. IM2708U, clone J3-119; Beckman Coulter), CD20-BV605 (catalog no. 302334, clone 2H7; Biolegend), CD27-BV650 (catalog no. 302828, clone 0323; Biolegend), CD38-AF680 (kindly provided by Margaret Beddall, Vaccine Research Center, NIAID, NIH), IgG-BV421 (catalog no. 562581, clone G18-145; BD), IgD-fluorescein isothiocyanate (FITC) (2030-02; Southern Biotech), IgM-PerCP-Cy55 (catalog no. 561285, clone G20-127; BD), CD21-PE-Cy5 (catalog no. 551064, clone B-ly4; BD), and CXCR5 PE-Cy7 (catalog no. 25-9185-42, clone MU5UBEE; eBioscience). To identify H5-specific GC B cells, additional staining was performed for intracellular Ki67-PerCP-Cy5.5 (catalog no. 561284, clone B56; BD) and Bcl-6 APC-Cy7 (catalog no. 563581, clone K1129-19; BD) using a FoxP3/transcription factor staining buffer set (eBioscience) as previously described (26). A minimum of 2 million total events was collected on a custom LSR II instrument (BD Immunocytometry Systems), and analysis was performed using FlowJo software, version 9.7.6 (TreeStar). Individual H5-specific memory B cells were sorted with a modified three-laser FACSAria cell sorter using the FACSDiva software (BD Biosciences). Probe-positive B cells were sorted as single cells into wells of a 96-well plate containing lysis solution as previously described (64). Flow cytometric data were subsequently analyzed using FlowJo, version 9.9.5. Gating strategy or individual fluorescence-activated cell sorter (FACS) plots are available on request.

RT-PCR, cloning, and expression of immunoglobulin genes.

Single B cell RNA was reverse transcribed as previously described (64). Briefly, samples were diluted 2-fold by addition of 26 microliters of nuclease-free water, and the cDNA plates were stored at −20°C. Individual AGM immunoglobulin (Ig) heavy-chain (H), light kappa (Lκ)-chain, and light lambda (Lλ)-chain genes were amplified by nested PCR using rhesus Ig-specific primers as previously described (65). Amplified PCR products were analyzed on 2% agarose gels (Embi-Tec), and positive reaction products were sequenced directly.

Immunoglobulin sequence analysis and SHM estimation.

The Ig heavy-chain and light-chain nucleotide sequences were analyzed as described previously (24). Briefly, the sequences were aligned with consensus human V_H and V_L genes with IMGT/HighV-Quest (66) to identify the variable (V) region, complementarity determining region 3 (CDR3), and CDR3-framework 4 (FR4) region (junction). Then the identified sequences of the V region, CDR3, and junction were aligned with AGM germ line V genes, D genes, and J genes to infer the V(D)J rearrangement and estimate the somatic hypermutation (SHM) according to the following: SHM (%) = (number of mutated nucleotides in the V region/number of V region nucleotides) × 100%.

The V region in the calculation inferred the region from the start of framework 1 (FR1) through the end of framework 3 (FR3).

Lineage analysis.

The unmutated common ancestor (UCA) was inferred as described previously (67). Briefly, the germ line gene segments of the V gene, D gene, and J gene were determined by comparing the mature Ig V_H genes with the germ line gene libraries and then computing the recombination sites of the gene segments and the recombination sequences including P nucleotides and N nucleotides. The inferred UCA was obtained by arranging the sequence segments as described above (see Fig. S9 in the supplemental material).

Statistics.

Statistical analyses used Prism GraphPad, version 6 (GraphPad Software, San Diego, CA). Mann-Whitney U tests were performed between groups; Wilcoxon matched pair-rank tests were used for analyses within each group.