Research ArticleVACCINES

Prolonged evolution of the memory B cell response induced by a replicating adenovirus-influenza H5 vaccine

See allHide authors and affiliations

Science Immunology  19 Apr 2019:
Vol. 4, Issue 34, eaau2710
DOI: 10.1126/sciimmunol.aau2710

Prolonged B cell responses

Replicating viral vaccines are known to induce durable and protective B cell responses, but previous studies have suggested that these responses may only evolve during active viral replication. Here, Matsuda et al. characterize memory B cell responses after vaccination with a replication-competent adenovirus-based influenza H5 recombinant vaccine over several months. Unexpectedly, they observed changes in B cell specificity and antibody affinity at 6 to 12 months after vaccination, long after viral replication ended. They detected antibodies with broad specificity, including a stem-specific antibody that shared similarities with another recently described antibody, thus defining a new multidonor class. These findings indicated that the B cell responses can evolve long after a single vaccination and provide insight into optimizing vaccine schedules.


Induction of an antibody response capable of recognizing highly diverse strains is a major obstacle to the development of vaccines for viruses such as HIV and influenza. Here, we report the dynamics of B cell expansion and evolution at the single-cell level after vaccination with a replication-competent adenovirus type 4 recombinant virus expressing influenza H5 hemagglutinin. Fluorescent H1 or H5 probes were used to quantitate and isolate peripheral blood B cells and their antigen receptors. We observed increases in H5-specific antibody somatic hypermutation and potency for several months beyond the period of active viral replication that was not detectable at the serum level. Individual broad and potent antibodies could be isolated, including one stem-specific antibody that is part of a new multidonor class. These results demonstrate prolonged evolution of the B cell response for months after vaccination and should be considered in efforts to evaluate or boost vaccine-induced immunity.


Antibodies induced after viral infection or vaccination can provide durable sterilizing immunity against many human viruses. For some viruses, such as influenza virus or HIV, the surface glycoproteins to which antibodies bind are highly diverse. For these two pathogens in particular, induction of antibodies capable of binding diverse isolates is currently the subject of intense interest and is an important obstacle to the development of effective vaccines (1, 2).

In the case of influenza virus, the surface hemagglutinin (HA) glycoprotein, which is the major target of neutralizing antibodies, exhibits an extraordinary level of genetic diversity. A total of 18 HA subtypes have been identified that are broadly classified into two groups based on the genetic and antigenic diversity of the virus. For influenza A, group 1 includes the H1 and H5 subtypes, and group 2 includes H3 and H7 subtypes. Most variation occurs in the head of HA, and the HA stem is more conserved and is the target of antibodies capable of neutralizing diverse strains (2). Passive transfer of HA stem–specific antibodies can provide protection against lethal challenge of animals with group 1 or 2 viruses (3, 4). However, the HA stem–specific response induced by either existing licensed influenza virus vaccines or natural infection is of insufficient magnitude or durability to protect against infection with highly varied isolates (2). Because of these features, the protection afforded by existing influenza virus vaccines is not long lived, and annual vaccinations use antigens that match circulating strains as closely as possible. In addition, existing seasonal vaccines provide very limited protection against new pandemic strains that may arise through zoonotic transmissions. Thus, understanding the dynamics of the B cell and antibody response after vaccination is of critical importance to the design of vaccines capable of providing broader and more durable protection against influenza.

Of the available vaccine platforms for presenting HA, replicating vectors have the potential to be highly immunogenic and offer several advantages over most nonreplicating vectors (5). They can express viral surface glycoproteins such that the total dose and duration of exposure to antigen likely exceed those of nonreplicating vectors. In addition, viral surface glycoproteins may be expressed by the host cell in the appropriate conformation and glycosylation state. They may therefore more closely match those expressed during natural infection compared with vaccines produced in heterologous cell lines or eggs. In addition, replicating vectors can cause inflammation through the induction of proinflammatory cytokines or through nucleic acid stimulation of Toll-like receptors on B cells or antigen-presenting cells.

In this study, we examine the dynamics of B cell expansion and evolution after vaccination with a replication-competent adenovirus type 4-H5 HA (A/Vietnam/1194/2004) expressing recombinant virus (Ad4-H5-Vtn). Fluorescent H1 or H5 probes were used to measure the frequency of peripheral blood B cells specific for H5 only or those that cross-reacted with H1. We also used these probes to isolate these cells and clone and re-express their antigen receptors to examine their evolution. This permits an analysis of the response of memory B cells that were primed by prior infection and cross-react with the H5 HA expressed by the vaccine and the response of naïve B cells over 12 months after a single vaccination. These methods provide a detailed examination of the shifts in antigen specificity and evolution of antibody lineages that occur after a single vaccination that was not apparent by standard analysis of serum specificity, antibody magnitude, or avidity. Unexpectedly, we observed evolution of the B cell response specificity and antibody affinity maturation 6 to 12 months beyond the period of active viral replication. Although serum neutralization after a single infection was modest, individual antibodies that were broad and potent could be isolated, including one stem-specific antibody that shares genetic elements and a mode of recognition with a previously described antibody, named 39.29, thus forming a new multidonor class. These results suggest that the evolution of B cell specificity and antibody affinity maturation can persist for months after a single vaccination. These findings also suggest that the optimal timing of boosting to increase breadth of the response to highly variable antigens may be considerably longer than currently thought.


Upper respiratory tract administration of Ad4-H5-Vtn vaccine induces durable antibody responses

In a prior trial, the Ad4-H5-Vtn vaccine was only modestly immunogenic when given orally without a subsequent boost, and serum antibodies rapidly declined after 4 weeks (6). However, the benefits of replication were limited by attenuation of the vaccine by two means. The first was the incorporation of the H5 HA transgene in the E3 region of Ad4, which can reduce its replication by as much as 10-fold (7, 8). The second was through administration via the gastrointestinal tract. The vaccine given to U.S. military personnel is a wild-type Ad4 that does not cause respiratory disease when given by an enteric-coated tablet (9). For this reason, we sought to remove gastrointestinal tract–mediated attenuation by administering the Ad4-H5-Vtn vaccine in the upper respiratory tract by either applying it to the tonsils or spraying it intranasally. For these vaccinations, escalating doses were given to groups of three participants that ranged from 103 to 108 virus particles (VPs). To provide a direct comparator, we orally administered Ad4-H5-Vtn to a separate group in the form of an enteric-coated capsule containing 1010 VPs, consistent with the prior trial (6). In total, 56 volunteers were enrolled to receive the Ad4-H5-Vtn vaccine by one of the three routes (fig. S1A).

Administration of 103 VPs by the tonsillar or intranasal route did not result in Ad4 seroconversion of vaccinees as defined by a fourfold or greater rise in serum neutralizing antibody titer. Immunization with 104 VPs or greater did result in seroconversion to Ad4, and of those that seroconverted to Ad4, there was no effect of dose on the development of neutralizing antibodies to the H5 transgene at week 8 or 26 (fig. S1B). Consistent with the prior trial (6), sera from oral vaccinees showed very low H5 neutralizing activity in a pseudotyped virus entry inhibition (PVEI) assay by week 8 [median 50% inhibitory dilution (ID50) = 210], and these titers dropped by 26 weeks (median ID50 = 44). In contrast, tonsillar and intranasal vaccination with 104 to 108 VP induced neutralizing antibody responses at 8 weeks [median ID50 = 836 (tonsillar) and 352 (intranasal)], and the antibody responses remained elevated at 26 weeks [median ID50 = 497 (tonsillar) and 429 (intranasal)] (Fig. 1A). In addition to the PVEI assay, H5-specific neutralization was also quantified by microneutralization (MN) and hemagglutination inhibition (HAI) assays. The MN and HAI results showed that the H5 antibody responses in serum were modest (fig. S2A).

Fig. 1 Intranasal Ad4-H5-Vtn vaccination induces durable serum neutralizing antibodies with breadth that extends to group 2 viruses.

(A) Serum neutralization of H5N1 elicited by Ad4-H5-Vtn immunization by the indicated route. Serum samples collected at weeks 8 (W8) and 26 were analyzed in the H5N1 PVEI assay. P values were calculated using t tests. ns, not significant. (B) Serum neutralization of group 1 or 2 influenza A or influenza B strains in the PVEI assay. (C) ELISA binding of serum from six vaccinees to group 1 or 2 influenza A or influenza B HA proteins. AUC, area under curve.

Ad4-H5-Vtn elicits HA stem–specific broadlyneutralizing antibodies

To characterize the response at the level of individual antibodies, we used H5 or H1 fluorescent probes to stain HA-specific peripheral blood B cells to sort and re-express the respective antibodies (10). We derived antibodies from three donors selected by the H5-specific potency of their sera, which was in the upper half of the cohort at week 26 in the PVEI assay (fig. S2B). The breadth of neutralization or binding of sera from these participants was modest against group 1 and 2 viruses (Fig. 1, B and C, and table S1). We sorted 1488 B cells and were able to amplify complementary DNA (cDNA) from 851 cells. Of these antibody sequences, 203 had both a productive heavy and light chain, and 95 antibodies were expressed at levels sufficient for further testing.

These antibodies varied widely in their potency against the target H5-Vtn HA in the PVEI assay (table S2). Among H5-specific antibodies with neutralizing activity against A/Vietnam/1203/2004, we observed an increase in potency of those isolated at weeks 34 to 103 [median 50% inhibitory concentration (IC50), 0.025 μg/ml] compared with antibodies isolated at week 4 (median IC50, 0.82 μg/ml). A similar increase in potency was not observed among H1H5-specific antibodies. To better understand the potency of these antibodies in other assays, we also tested them against live H5N1 virus in plaque reduction neutralization assays. The influenza virions used in plaque reduction neutralization assays have a higher density of surface glycoproteins compared with pseudotyped virus assays in which HA spikes are less dense. Although there is not uniform agreement (11, 12), this may limit access of stem-specific monoclonal antibodies (mAbs) to their respective binding sites on the HA stem in plaque reduction assays. The neutralization potencies of antibodies derived from H5-specific B cells showed no difference in the plaque reduction neutralization assay compared with the PVEI assay (fig. S2C, left). However, the neutralization potencies of antibodies derived from H1H5-specific B cells, which typically target the HA stem, were significantly lower in the plaque reduction assay compared with the PVEI assay (P < 0.001, fig. S2C, right).

To examine the breadth of these antibodies, we then tested them against a panel of five pseudotyped viruses in the PVEI assay, including HAs from the group 2 viruses H3N2 A/Hong Kong/1/1968 and H7N9 A/Anhui/1/2013 (table S2). In this assay, four antibodies neutralized group 2 viruses (429 F05, 429 F08, 429 B01, and 550 C06). Of the 95 antibodies tested for neutralization, we chose 37 based on potency for further testing of their breadth of binding in an enzyme-linked immunosorbent assay (ELISA). Of these, three antibodies were particularly broad and bound both group 1 and 2 viruses (429 F05, 429 B01, and 550 C06) (fig. S3), consistent with the neutralization results.

We then tested the four antibodies that neutralized group 2 pseudotyped viruses against a panel of 16 pseudotyped viruses (Fig. 2A) in comparison with other previously described broad and potent influenza virus neutralizing antibodies. Of the antibodies tested, mAb 429 B01 was particularly broad and potently neutralizing (Fig. 2, A and B). The 429 B01 neutralized 14 of 15 pseudotyped viruses, including H1, H5, and H9 viruses from group 1 and H3, H7, and H10 viruses from group 2. This is in the same range of neutralization breadth and potency of two previously described broadly neutralizing antibodies, CR9114, and FI6v3. The median IC50 of 429 B01 was 0.090 μg/ml. These data show that broad and potent antibodies are present within the repertoire of peripheral blood memory B cells, even among vaccinees with modest serum neutralizing breadth or potency.

Fig. 2 A subset of isolated antibodies are particularly broad and potent.

(A) Neutralization of multiple group 1 or 2 influenza strains assessed using the PVEI assay. Antibodies isolated from sorted B cells of Ad4-H5-Vtn vaccinees are displayed on the left and controls on the right. IC50 values of <0.1 μg/ml are highlighted in red, between 0.1 and 1 μg/ml in orange, between 1 and 10 μg/ml in yellow, and between 10 and 50 μg/ml in green. Any IC50 values of >50 μg/ml were excluded from median IC50 calculations for each antibody displayed. (B) Breadth and potency curve of four mAbs isolated from sorted B cells. The IC50 against influenza B was excluded from the calculation. (C) SPR binding of 429 B01 to HA1, HA2, or HA0 of the indicated influenza viruses. RU, resonance units. NB, no binding. (D) ELISA binding to H5 HA protein of 429 B01 alone (empty circles) or in competition with F10 (solid circles). CR6261 was used as a positive control (right). Ab, antibody. RLU, relative light units.

We then sought to determine the binding site of 429 B01 on the HA protein. HA is synthesized as a polypeptide HA0 that is proteolytically cleaved into two subunits. They are HA1, which forms the globular head, and HA2, which, together with N- and C-terminal HA1 residues, forms the stem domain. Binding to HA1 or HA2 from H1N1 A/California/07/2009, H5N1 A/Vietnam/1203/2004, or the HA0 protein from four different viruses (H1N1 A/California/07/2009, H3N2 A/Brisbane/10/2007, H5N1 A/Vietnam/1203/2004, and H7N7 A/Netherlands/219/2003) was measured by surface plasmon resonance (SPR) (Fig. 2C). 429 B01 showed no binding to either HA1 or HA2 alone but displayed strong binding to all the tested HA0 proteins. These data suggest that 429 B01 recognized an epitope that includes both HA1 and HA2 domains. The antibody was also tested for the ability to bind to H5 protein in a competition ELISA with a stem-binding antibody, F10. In the presence of F10 mAb, we observed decreased binding of 429 B01 to H5, suggesting that it bound the HA stem (Fig. 2D) (13).

Structural definition of antibody recognition identifiesa new multidonor class

To provide initial images of antibody recognition of HA, we performed negative-stain electron microscopy (EM) for the antigen binding fragments (Fabs) of 429 B01, 550 C06, and 429 F05 in complex with HAs from diverse influenza virus strains. Reference-free two-dimensional classification revealed that all antibodies interacted with the stem region of HA in similar modes (Fig. 3A and fig. S4A), except that 429 B01 in complex with A/Hong Kong/1-4-MA21-1/1968 (HK68) HA appeared to be monomeric, consistent with the dissociation of this HA observed in a previous report (13). To understand the detailed structural basis of HA stem recognition by these antibodies, we determined the crystal structure of 429 B01 Fab in complex with the HK68 HA to 3.2-Å resolution (Fig. 3B and table S3). The structure confirmed that 429 B01 recognized a conserved region on the HA stem with HA1 and HA2, providing 23 and 77% of the 940-Å2 binding surface, respectively (Fig. 3C and table S4A). 429 B01 interacted with HA primarily through its heavy chain, which contributed 68% of the 910-Å2 paratope surface (table S4B). The Phe99HC at the tip of the third heavy chain complementarity-determining region (CDR H3) was situated in the hydrophobic pocket formed by HA2 residues Trp21HA2, Ile45HA2, and Ile48HA2. In addition, the main chain atoms of Thr97HC, Gly100HC, and Leu100BHC formed hydrogen bonds with glycan Asn38HA1 and Ile18 HA2. Overall, the CDR H3 provided 56% of the total paratope surface (Fig. 3D and table S4C). Extensive hydrogen bonds were also formed between CDR L1, L2, and residues of HA2 helix A (Fig. 3D and table S5).

Fig. 3 Structural basis of HA recognition by the 429 B01 class of antibodies.

(A) Negative-stain EM of antibody in complex with HA as indicated. Scale bars, 10 nm. (B) Cocrystal structure of 429 B01 in complex with HK68 H3 HA. HA1, HA2, and antibody heavy and light chains are depicted in cartoons colored green, cyan, orange, and slate, respectively. Inset: A 90° view of the interface with interacting CDR loops shown in cartoon and HA as surface representation. (C) Epitope of 429 B01 colored in orange. A black dashed arrow indicates the orientation of the antibody defined by the line connecting Cα atoms of heavy chain Cys22 (orange circle with H inside) and light chain Cys23 (purple circle with L inside). (D) Detailed interactions of 429 B01 with HA. Residues from CDR H3, CDR L2, and L3 of 429 B01 form extensive hydrogen bonds with HA. (E) Epitope of antibody 39.29 (salmon) derived from the same germline gene as 429 B01 overlapped with that of 429 B01 (black lines). (F) Position of the CDR H3 and residues conserved between 429 B01 and 39.29 after superposition over the HA2 region. (G) Sequence alignment of 429 B01 with deduced germline sequences. Paratope residues are colored as in (B). Sequences of 39.29 are shown to highlight the conservation of a Val-Phe-Gly motif in the CDR H3.

The 429 B01 antibody shared several features with some previously described stem-binding antibodies. The 429 B01 antibody, along with FI6v3 and 39.29, was derived from the heavy chain variable gene HV3-30. Although the 429 B01 epitope overlapped that of FI6v3, the orientation of HA-bound 429 B01 was rotated about 90° relative to FI6v3 (fig. S4B). The 429 B01 did share a similar mode of HA recognition with 39.29 (14). The epitopes of these two antibodies shared a high degree of overlap, and the orientation of the heavy and light chains were very similar (Fig. 3E). They shared critical contacts with HA at Ser52HC, Tyr52AHC, Asp53HC, Asn56HC, and Tyr58HC in the CDR H2, and especially the Val98HC-Phe99HC-Gly100HC-Leu/Ile100AHC motif at the tip of the CDR H3 that provided ~40% of the overall paratope surface (Fig. 3, F and G). Although they were not derived from the same germline gene, the light chains of 429 B01 and 39.29 shared 4 of 10 429 B01 paratope residues. The shared genetic elements in the heavy chain, convergent affinity maturation, and similar mode of recognition indicate that these antibodies form a multidonor class (Fig. 3G).

The tip of CDR H3 of 429 B01 was encoded by the HD3-3 gene. Junctional analysis indicated that HD3-3–encoded Phe99-Gly100 provided 35% of the paratope surface on the heavy chain and remained unmutated in the mature antibody (Fig. 4A and table S4). VH3-30–derived 429 B01 and 39.29 shared the same heavy chain HD3-3 gene with two previously described HV6-1–derived antibodies MEDI8552 and 56.a.09 (13, 15), and the Phe99HC and Gly100HC at the tip of CDR H3 were conserved (Fig. 4B). Alignment of the structures of 429 B01 and 56.a.09 in complex with HA indicated that the overall binding modes of HA-bound antibodies were different (Fig. 4C, left). However, the HD3-3–encoded CDR H3 tip recognized the HA stem in the same conformation (Fig. 4C, right). The selection of a Val/Ile98-Phe99-Gly100-Leu/Ile/Val101 motif among these antibodies suggested that antibodies derived from different heavy chain variable (VH) genes can achieve recognition of conserved epitope using a motif in their CDR H3 encoded by a shared germline D region.

Fig. 4 Dominant interaction at the tip of CDR H3 encoded by the IGHD3-03 gene.

(A) Analysis of the junctional sequence of 429 B01. Germline gene-encoded nucleotide and amino acid residues are shown in black with the corresponding junctions colored in light blue. Somatically mutated amino acids and nucleotides are colored red. Nucleotides deleted by exonuclease trimming are crossed out. (B) Sequence alignment of the CDR H3 of 429 B01 in comparison with published stem-specific antibodies. The HD3-3–encoded motif was conserved between antibodies derived from varied immunoglobulin heavy chain variable region (IGHV) genes. (C) Comparison of binding modes of HA stem–specific antibodies sharing the same HD3-3 gene.

Ad4-H5-Vtn drives prolonged H5-specific B cell expansion and affinity maturation

To better understand the dynamics of H5-specific B cells in the peripheral blood after Ad4-H5-Vtn vaccination, we measured their frequencies over time by flow cytometry in the three vaccinees from whom we derived individual mAbs and an additional three vaccinees selected for their serum neutralization potency (Fig. 5A). We observed an early expansion of H1H5 dual-specific B cells at the 4-week time point in each vaccinee (Fig. 5B). It is likely that this is a population of memory B cells that was primed by prior infections with H1 viruses, cross-reacted with H5-Vtn, and was expanded by vaccination. This H1H5-specific cell population, measured as a percent of the memory B cell compartment, then contracted over the subsequent time points that were sampled. In contrast, those cells that were single H5-specific continued to expand over 26 to 52 weeks, well beyond the 7 to 28 days in which viral DNA was detected (Fig. 5C). At the prevaccination time point, antibodies isolated from H1H5 dual-specific B cells showed moderate to strong binding to both H5 and H1 HA proteins, whereas antibodies from H5-specific cells displayed only weak binding and likely represent nonspecific binding (fig. S5). These data suggest that, although vaccinees received only a single dose of Ad4-H5-Vtn that replicated for 1 to 4 weeks, B cells specific for the H5 globular head that are not primed by prior infection continued to expand over 6 to 12 months.

Fig. 5 Expansion of H5-specific B cells is prolonged after Ad4-H5-Vtn vaccination.

(A) Expansion of H5- or H1H5-specific B cell populations at sequential time points after Ad4-H5-Vtn vaccination. Vaccinee 849 is presented. B cells were stained with HA probes H5 (A/Vietnam /1203/2004) and H1 (A/New Caledonia/20/1999). (B) Summary of expansion of H5-, H1-, or H1H5-specific peripheral blood memory B cell populations in all six subjects. (C) Last day of Ad4 shedding detected in each of the six participants.

We then sought to further understand the development of H5-specific antibodies at the level of individual B cell clones. The H5-specific monoclonal antibodies isolated from each vaccinee at later time points exhibited significantly higher neutralization potency against H5-Vtn pseudovirus and affinity toward H5 HA than the antibodies isolated from week 4 (Fig. 6, A and B, left, and fig. S6A, left). In contrast, no difference in neutralization potency or affinity was observed for the H1H5-specific antibodies between week 4 and the later time points (Fig. 6, A and B, right, and fig. S6A, right), suggesting that these specificities, presumably induced by prior infections, did not undergo the same increases in potency and affinity as H5-specific cells stimulated from the naïve compartment.

Fig. 6 Increase in neutralization potency of H5-specific antibodies parallels increases in SHM and affinity.

(A) Neutralization by H5- or H1H5-specific mAbs against H5-Vtn in the PVEI assay over time. The median IC50 for each time point is displayed. P values were calculated using t tests. (B) Affinity of H5-specific or H1H5-specific mAbs against H5N1 Thailand HA over time. The median affinity for each time point is shown. (C) Antibody pairs isolated from week 4 and later time points with the same heavy and light chain alleles, CDR3 lengths, and junction sequences. H5-specific antibodies are indicated in red, and H1H5-specific antibodies are indicated in blue. (D) Mutation rate (left), affinity toward H5 HA protein calculated on the basis of biolayer interferometry analysis (middle), and neutralization of H5-Vtn assessed by the PVEI assay (right) of antibody pairs. KD, dissociation constant.

Among the mAbs isolated, five pairs of antibodies for which a clonal relative was found at an early and late time were identified. Each pair shared the same heavy and light chain variable region alleles, CDR3 length, and junction cDNA sequences indicating their clonal relationship (Fig. 6C). H5-specific clonal relatives had higher rates of somatic hypermutation (SHM) at late time points compared with week 4, whereas little or no change was observed among H1H5-specific pairs (Fig. 6D, left). This increase in SHM was paralleled by an increase in affinity to trimeric H5 (Fig. 6D, middle, and fig. S6B), with one exception (A09-D11), which had a low starting affinity but was within the range of data from nonrelated clones. The changes in neutralization potency against H5 A/Vietnam/1203/2004 in the PVEI assay for antibody pairs were more mixed (Fig. 6D, right), likely a result of sampling error in the low numbers of pairs isolated.

The serum avidities to the HA1 or HA2 subunits of H1N1 (A/California/07/2009) or H5N1 (A/Vietnam/1203/2004) were also measured. The serum avidity to the H5 HA1 domain rapidly increased and plateaued by 8 weeks and was maintained up to 52 weeks after vaccination (fig. S6C). However, we did not observe prolonged increases in avidity to the trimeric H5 HA1 that paralleled the changes observed in the monoclonal antibodies derived from memory B cells. Serum avidity to the heterologous H1 HA1 domain was lower than avidity for the H5 HA1 for most donors, peaked at 4 weeks after vaccination, and did not increase further (fig. S6D). Serum avidity to the HA2 domain showed a similar pattern of reactivity to the HA2 domains derived from either H5 or H1 (fig. S6, E and F), suggesting that most anti-HA2 antibodies induced after vaccination in adults are primarily cross-reactive antibodies generated against the conserved sequences within HA2 between the H5N1 and H1N1 strains. Although serum avidities to HA2 peaked by 8 to 12 weeks after vaccination, these came to baseline levels in most of individuals 52 weeks after Ad4-H5-Vtn vaccination. Together, these results suggest that after a single Ad4-H5-Vtn infection, there was a prolonged parallel increase in neutralization potency, affinity, and SHM of individual monoclonal antibodies, although this is not readily detectable at the seum level.

Temporal evolution of B cell lineages is detected by next-generation sequencing

To gain a broader understanding of the evolution of individual influenza virus HA specificities in the peripheral blood memory B cell pool associated with the prolonged expansion of H5-specific B cells and increases in antibody potency, we performed next-generation sequencing (NGS) of immunoglobulin G (IgG) transcripts from memory B cells of six donors at time points corresponding to the collections of influenza-specific B cells (weeks 0, 4, and 34 or 52). We searched the NGS data to identify transcripts clonally related to the HA-specific clones identified in earlier HA sorts. Because two previous studies (13, 15) revealed that H5 vaccination can elicit three multidonor classes of stem-specific broadly neutralizing antibodies—two originating from VH1-18 and one from VH61-1 genes—we investigated whether similar clones were elicited by the Ad4-H5-Vtn vaccination. A search of existing NGS datasets identified transcripts with the VH usage and signatures in CDRH3 matching the three previously described classes of stem-specific antibodies (fig. S7). We found only one relative of the 429 B01 antibody at the 4-week time point, and no relatives were detected at week 52.

To understand the rate of antibody maturation, we calculated SHM levels for transcripts from all sorted B cells (Fig. 7A). The median frequencies of SHM among sequences from H1H5- and H1-specific B cells at weeks 3 and 4 were 5.9 and 6.6%, respectively, and did not increase over time (Fig. 7A, right, and fig. S8, left), suggesting that many of these sequences were derived from recalled matured memory B cells. In contrast, there was a progressive increase at each time point in the median frequency of SHM among heavy chain sequences from H5-specific B cells from 1.0% (range, 0.0 to 20.4%) at week 4 to 6.1% (1.4 to 35.0) at week 52. A similar pattern was observed in the paired light chain transcripts (Fig. 7B and fig. S8, right). The low SHM level for many transcripts from H5-specific B cells at week 4 suggests that these cells were elicited from naïve B cell repertoires by Ad4-H5-Vtn. The mutation frequencies of some antibodies induced by Ad4-H5-Vtn even at 4 weeks were in the range of 10 to 20%.

Fig. 7 Evolution of SHM and clonal expansions is prolonged after Ad4-H5-Vtn vaccination.

(A) Mutation frequency of H5- or H1H5-specific antibody heavy chain sequences isolated from participants at various time points after vaccination by single-cell sorting. Red horizontal bars indicate the median. (B) Mutation rate of light chain sequences. (C) Dynamics of clonal size of H5- or H1H5-specific antibody sequences. Clonal family members were identified from the NGS dataset of each time point. The size of each clonal family at a time point was normalized by the number of NGS reads. Each color represents a clonal family with the width of the stacked ribbon showing the number of members.

The temporal nature of clonal expansions was also examined. The relative size of expansions in clonal relatives of sorted antibodies and NGS-derived sequences is shown in Fig. 7C. In the case of the H1H5-specific cells, there was a rapid polyclonal expansion at 4 weeks after vaccination that rapidly contracted (Fig. 7C, right). The timing of this clonal expansion paralleled the increase in frequency of the H1H5-specific cells observed after vaccination (Fig. 5A). In the case of the H5-specific memory B cells there was a progressive increase in the frequency of individual clones that began as a highly oligoclonal response that remained oligoclonal at 52 weeks (Fig. 7C, left). The analysis further showed that members of some H5- and H1H5-specific clones were found at week 0, supporting the idea that Ad4-H5-Vtn activated matured memory B cells. Together, these data indicate that, although cross-reactive stem-specific expansions occur early after vaccination, there is a remarkably prolonged maturation of the H5-specific response at the level of monoclonal antibodies in their frequency, SHM, affinity, potency, and clonality.


These results add to recent observations on the characterization of human monoclonal antibodies directed to influenza HA and provide new data regarding the dynamics of maturation of the antibody response after vaccination. After a single vaccination with a replicating vector, we observed several features of the response that were not detected at the level of sera. This included a very rapid expansion and contraction of B cells expressing H1H5-specific antibodies, which typically bind the HA stem. It also included a more prolonged increase in the frequency of B cells expressing H5-specific antibodies that paralleled an increase in antibody SHM, affinity, and neutralization potency. The duration of these changes was 6 to 12 months, considerably longer than the 2 to 4 weeks of detected adenoviral replication.

The duration of maturation of the H5-specific response is well beyond the period that is commonly used in immunization strategies to boost or measure the response to viral antigens (16, 17). It is relatively common for studies in animals or humans for boost immunizations to be given at 2 to 8 weeks, likely based on the relatively rapid increase and plateau in the level of binding or neutralizing antibodies at the level of sera usually observed after immunization. On the basis of the results of the present study, such a boost may occur well before full maturation of the response at the level of monoclonal antibodies. It is possible that boosting at 6 to 12 months may induce higher-magnitude or higher-quality responses given that B cells expressing higher affinity antigen receptors would be available for competition for antigen in germinal center light zones. Although infrequently directly tested, there are some examples in which delaying a booster immunization to 6 to 12 months induced higher levels of antibodies than boosting at 1 to 2 months (1823). These results suggest that there is a benefit with regard to greater neutralizing titers or serum avidity conferred by delaying a boost beyond 12 weeks. In addition to the prolonged duration of SHM and affinity maturation, the rate of increase in SHM after vaccination in some transcripts was quite high. This rate is potentially of significance for vaccines where the goal is to stimulate antibodies that are highly somatically mutated such as for HIV. Most broadly neutralizing antibodies to HIV are mutated 20 to 30% from germline VH segments. Because these antibodies may take more than 2 years to develop in some patients, it has been suggested that this level of SHM is one potential barrier to their stimulation. However, in this study, we observed some antibodies that had SHM levels of 20% only 28 days after a single infection.

There are several important limitations to this study that should be noted. First, the overall serum neutralization induced by a single dose of this vaccine was modest. Second, it remains uncertain whether some of the highly mutated antibodies observed in the first 4 weeks after immunization were derived from cells that were naïve on day 0 or represent cross-reactive memory cells. In addition, we were unable to define a mechanism that underlies the rapid contraction of the H1H5-specific B cells. Although speculative, it is possible that these cells, primed by prior infections, represent a pool of cells that is programmed to undergo more rapid expansion and contraction. Perhaps the greatest limitation of the study was our inability to document the impact of persistence of antigen to the prolonged maturation of the H5-specific B cell response.

The presence of antigen is thought to be central to the process of affinity maturation of the B cell response. Protein and viral antigens are thought to persist on follicular dendritic cells for weeks to months after immunization, although we do not have direct measurements using current techniques (24, 25). Expression of green fluorescent protein by a recombinant Ad5 host range mutant has been observed 25 weeks after the last vaccination in the rectal tissue macrophages of Rhesus macaques (26). Although some adenoviruses can chronically infect cells in lymphoid tissues, this has not been observed with adenovirus type 4. It should be noted that, although persistence of antigen may be required for the protracted maturation of the H5-specific antibody response that we observed, chronic viral replication is likely not required. In a genetically restricted response to the hapten 4-hydroxy-3-nitrophenyl acetyl in mice, although most of the affinity maturation in germinal centers occurs in the first 2 to 3 weeks, the affinity of bone marrow antibody-producing cells increased for at least 60 days and serum for 120 days, presumably through persistent clonal competition (27, 28). In response to a more complex protein antigen, in one recent study, the level of SHM of H5-specific B cells at 26 weeks was higher than those from cells at 4 weeks in vaccinees that received an H5 DNA vaccine followed by an H5 MIV boost (13). Thus, it is possible that prolonged maturation of a response to neo-antigens is a more generalizable phenomenon that follows immunization with either nonreplicating vaccines or replicating viral infections.

Together, the results of this study demonstrate the utility of probe-based analysis and sorting of peripheral blood B cells after vaccination. The probes used in the current study permitted an analysis of the evolution of the memory B cell response specificity, SHM, affinity, and potency at a level of detail that was not apparent by analyses of serum alone. Construction of similar probes for other viral glycoproteins and application of these to peripheral blood cells or tissues may potentially advance vaccine efforts directed toward highly variable viruses such as HIV or influenza virus.


Study design

Ad4-H5-Vtn vaccine recipients were enrolled at the National Institutes of Health under a clinical protocol approved by the Investigational Review Board of the National Institute of Allergy and Infectious Disease (NIAID-IRB) ( identifier NCT01443936). All vaccine recipients and their household contacts provided written informed consent in accordance with the Declaration of Helsinki and were selected on the basis of the following criteria: healthy men and nonpregnant women, between the ages of 18 and 49 years, and with an Ad4 antibody titer of ID80 < 100. Ad4-H5-Vtn recombinant vaccine produced by PaxVax Inc. is a replication-competent Ad4-based vaccine carrying a full-length HA gene from influenza A H5N1 virus (A/Vietnam/1194/2004). Vaccine regimen is indicated in fig. S1A.

PVEI assay

HA-pseudotyped lentiviruses were generated as previously described (29). Briefly, 293T cells were transfected with the plasmids expressing the following: 1 μg of a plasmid expressing HA, 0.1 μg of a plasmid expressing a matched neuraminidase, 17 μg of luciferase under the control of a CMV promoter, and 0.1 μg of type II transmembrane serine protease. Supernatants were harvested after 72 hours, filtered, and frozen. The HA and matched NA genes used in transfection are listed in table S6.

Influenza group 1 and 2 neutralizing activities of sera and monoclonal antibodies derived from Ad4-H5-Vtn vaccine recipients were tested using single-round pseudovirus infection of 293T or 293A cells as described previously (13, 30). Screening of volunteer sera was similarly performed using Ad4-luciferase virus infections of A549 cells. The protocol for human sera influenza virus neutralization assays is identical to that used for monoclonal antibodies, except that it was pretreated with receptor-destroying enzyme II (Denka Seiken, Japan), in accordance with the manufacturer’s instructions, to eliminate serum nonspecific inhibitors (31). The ID50 was calculated as the serum dilution that reduced infection by 50%.

Flow cytometry, cell sorting, and cloning

Memory B cell staining and sorting were performed as previously reported (fig. S9) (10). Antibody heavy and light chains were recovered by reverse transcription polymerase chain reaction (PCR) as previously described (32). PCR products were sequenced by ACGT Inc. Antibody sequences were analyzed by IMGT/HighV-QUEST ( Heavy and light chains (κ or λ) from the same well were cloned into pVRC8400 expression vectors containing the respective IgG heavy or light chain constant region via enzyme recognition sites AgeI/SalI, AgeI/BsiwI, or AgeI/XhoI. Antibodies were re-expressed by cotransfection of heavy and light chain plasmids into 293T cells purified with protein A–Sepharose resin (GE Healthcare).

ELISA and competition ELISA assay

ELISA assays were performed as previously described (33). Briefly, mAb dilutions in phosphate-buffered saline (PBS) starting at 30 μg/ml and followed by threefold serial dilutions were prepared and kept overnight at 4°C (34, 35). ELISA plates were coated overnight with recombinant protein (2 μg/ml) (table S6) in coating buffer and stored at 4°C (34, 36). The next day, 3% milk in 0.1% Tween 20–PBS (PBS-T) was used to block the plates for 1 to 2 hours at room temperature (RT). Antibody dilutions were applied, and plates were incubated for 80 min at RT, shaking. After washing plates three times with PBS-T, goat antihuman IgG conjugated to horseradish peroxidase (HRP; 1:3000) was added onto plates and incubated 60 min at RT, shaking before the final washing step. O-phenylenediamine (SIGMAFAST OPD, Sigma-Aldrich) was used as a substrate for HRP, and the reaction was stopped by 3M hydrochloric acid. Obtained signal was measured using a Synergy H1 microplate reader (BioTek) at 490 nm. Human sera were assayed as above except that ELISA plates were blocked with 3% goat serum and 0.5% milk in PBS-T for 1 hour, and sera were incubated on plates for 2 hours.

The competition ELISAs were performed as described previously (30). Briefly, streptavidin-coated 384-well plates [MesoScale Discovery (MSD), MD] were blocked with a 5% (v/v) solution of MSD Blocker A for 1 hour and then washed with PBST (PBS + 0.05% Tween 20). The plates were coated with biotinylated HA proteins (H5 Vtn04) at 1 mg/ml for 1 hour and then washed. Single-chain variable fragments derived from the HA stem-specific mAb F10, or the HIV-1 Env-specific control mAb VRC-01, were diluted to 4 mg/ml in MSD Blocker A buffer and added to bound HA antigens. After 1 hour, plates were washed, and serially diluted mAb preparations were added. After 1 hour, plates were washed, and bound antibody was detected using goat antihuman IgG SULFO-TAG–labeled secondary antibody (1 mg/ml). Plates were read using an MSD sector imager 2400.

Detection of viral shedding

Viral DNA was extracted from nasal, rectal, or throat swabs and vaccine standards using the KingFisher Flex Purification system (Thermo Fisher Scientific). Quantitative PCR was performed using the QuantStudio 3 system (Thermo Fisher Scientific) to measure the levels of viral DNA in nasal, rectal, or throat swabs. DNA extract (8.25 μl) was combined with TaqMan Fast Advanced Master Mix (Thermo Fisher). Reaction conditions were as follows: one 20-s period at 95°C, followed by 40 cycles of 1 s at 95°C and 20 s at 60°C. The H5-Vtn insert was amplified using the following primer/probe set: forward primer 5′-CAG CTT GTG CAC GGT GCT AA-3′, reverse primer 5′-TGC AAA AAG AAG CAC TAT TTT CTC C-3′, and probe 5′-/56-FAM/CGC GAT CGT GTG CCT GAG CAT TC/36-TAMSp/-3′. The standard curve was used to associate cycle threshold with VP per microliter, and averages were calculated for each swab sample.

Statistical analysis

When comparing two groups with each other, we used the two-sample t test assuming unequal variances. When comparing two groups with each other where each sample from one group can be paired with another sample from the second group, and as specified in figure legends, we used the paired t test. When comparing multiple groups together, we used one-way analysis of variance (ANOVA). All analyses were conducted in R or GraphPad Prism. Additional methods may be found in the Supplementary Materials.


Materials and Methods

Fig. S1. Influence of vaccination route and dose on H5 antibody responses.

Fig. S2. H1H5- or H5-specific antibody responses in PVEI, MN, and HAI assays.

Fig. S3. ELISA binding of H1H5- or H5-specific monoclonal antibodies to group 1 or 2 influenza A or influenza B (Massachusetts/2/12) HA proteins.

Fig. S4. Recognition of HA by IGHV3-30–derived antibodies.

Fig. S5. ELISA binding of H1H5- or H5-specific monoclonal antibodies isolated before vaccination to H5 Vietnam or H1 New Caledonia HA proteins.

Fig. S6. Increase in affinity of H5-specific antibodies and avidity maturation of sera against HA1 or HA2 subunits.

Fig. S7. Alignment of antibody clones that are members of three previously described multidonor classes of HA-specific broadly neutralizing antibodies.

Fig. S8. SHM of H1-specific antibodies over time.

Fig. S9. Gating strategy for analyzing and sorting.

Table S1. Serum neutralization of multiple influenza strains in the PVEI assay after Ad4-H5-Vtn vaccination.

Table S2. Neutralizing activity of monoclonal antibodies isolated from vaccinees against a five-virus panel.

Table S3. Crystallographic data collection and refinement statistics.

Table S4. Composition of 429 B01 epitope and paratope.

Table S5. Hydrogen bonds between HA and 429 B01 antibody.

Table S6. Virus strain abbreviations.

Data file S1.

References (3746)


Acknowledgments: We thank members of the Vaccine Research Center for discussions of the manuscript. We thank S. Strohmeier and F. Amanat for technical assistance with protein expression in the Krammer laboratory. J.A., J.S., and M.G. were employed by PaxVax Inc. (Redwood City, CA 94065, USA), which was acquired by Emergent BioSolutions. J.S. is currently employed by ClearPath Vaccines (Rockville, MD 20855, USA). M.G. is currently employed by Verndari Inc. (Sacramento, CA 95817, USA) and Jiangsu Atom Bioscience and Pharmaceutical Co., Ltd. (Zhenjiang, China). J.H. is currently employed at Fudan University, Shanghai City, China. Funding: This work was supported by the Intramural Research Programs of the Vaccine Research Center and of the National Institute of Allergy and Infectious Diseases (NIAID), NIH. Use of sector 22 (Southeast Region Collaborative Access team) at the Advanced Photon Source was supported by the U.S. Department of Energy, Basic Energy Sciences, Office of Science, under contract number W-31-109-Eng-38. Work by Leidos was supported, in part, with federal funds from the Frederick National Laboratory for Cancer Research, NIH, under contract HHSN261200800001E. Funding to the Krammer laboratory was provided by NIAID grant U19 AI109946 and the Centers of Influenza Virus Research and Surveillance (CEIRS) contract HHSN272201400008C. Author contributions: K.M., J.H., J.S., J.A., M.G., and M.C. conceived, designed, and coordinated the study. B.H.K., E.I., T.G., S.S., L.B., A.A.P., J.H., S.R.N., and K.M. performed B cell isolations, antibody cloning, and characterizations. The clinical study was executed by A.P. and S.A.M. A.C., K.L., S.A., A.W., and A.B.M. conceived and executed the sorting of probe-specific B cells. T.Z., J.P., B.Z., Y.Y., S.D., M.G.J., Y.T., and P.D.K. contributed to the structural analysis of antibody-HA complexes. R.T.B., T.J.W., V.C., and F.K. conceived or performed ELISA, MN, and HAI assays. H.G. and S.K. conceived and performed SPR and serum avidity assays. C.S.F.C. performed antibody affinity assays. C.J.L. performed the statistical analyses. Z.S., Y.G., C.S., and L.S. performed bioinformatics analyses of NGS data. K.M., J.H., T.Z., Z.S., T.G., S.S., L.B., S.K, A.B.M., R.A.K., L.S., F.K, P.D.K., and M.C. contributed to the writing and revision of the manuscript. All authors read and approved this manuscript. Competing interests: F.K. is an inventor on patent applications held/submitted by the Icahn School of Medicine at Mount Sinai that covers universal influenza virus vaccines. The authors and affiliated companies do not benefit financially from publication of this manuscript. Data and materials availability: Antibody sequences are deposited in GenBank. Accession numbers are MK510883-MK510886 and MK518386-MK518389. Protein structures are available through the Protein Data Bank (PDB) under PDB code 6NZ7. Recombinant HAs are available from F.K. under a material transfer agreement with the Icahn School of Medicine at Mount Sinai. Plasmids encoding the heavy and light chains of antibodies 550 C6, 429 F08, 429 F05, or 429 B01, or sequences for other antibodies identified in this work are available from M.C. under a material transfer agreement with NIAID/NIH. Plasmids used to generate pseudoviruses used in this work are available from R.T.B. under a material transfer agreement with NIAID/NIH.

Stay Connected to Science Immunology

Navigate This Article