Characterization of Novel Cross-Reactive Influenza B Virus Hemagglutinin Head Specific Antibodies That Lack Hemagglutination Inhibition Activity

Understanding the immune response that older individuals mount to influenza virus vaccination and infection is critical in order to design better vaccines for this age group. Here, we show that older individuals make broadly neutralizing antibodies that have no hemagglutination-inhibiting activity and are less potent than strain-specific antibodies. These antibodies could drive viral escape from neutralization but did not result in escape from binding. Given their different mechanisms of action, they might retain protective activity even against escape variants.

ABSTRACT

Humoral immune responses to influenza virus vaccines in elderly individuals are poorly adapted toward new antigenically drifted influenza virus strains. Instead, older individuals respond in an original antigenic sin fashion and produce much more cross-reactive but less potent antibodies. Here, we investigated four influenza B virus hemagglutinin (HA) head specific, hemagglutination inhibition-inactive monoclonal antibodies (MAbs) from elderly individuals. We found that they were broadly reactive within the B/Victoria/2/1987-like lineage, and two were highly cross-reactive with B/Yamagata/16/1988-like lineage viruses. The MAbs were found to be neutralizing, to utilize Fc effector functions, and to be protective against lethal viral challenge in a mouse model. In order to identify residues on the influenza B virus hemagglutinin interacting with the MAbs, we generated escape mutant viruses. Interestingly, escape from these MAbs led to numerous HA mutations within the head domain, including in the defined antigenic sites. We observed that each individual escape mutant virus was able to avoid neutralization by its respective MAb along with other MAbs in the panel, although in many cases binding activity was maintained. Point mutant viruses indicated that K90 is critical for the neutralization of two MAbs, while escape from the other two MAbs required a combination of mutations in the hemagglutinin. Three of four escape mutant viruses had increased lethality in the DBA2/J mouse model. Our work indicates that these cross-reactive antibodies have the potential to cause antigenic drift in the viral population by driving mutations that increase virus fitness. However, binding activity and cross-neutralization were maintained by a majority of antibodies in the panel, suggesting that this drift may not lead to escape from antibody-mediated protection.

IMPORTANCE Understanding the immune response that older individuals mount to influenza virus vaccination and infection is critical in order to design better vaccines for this age group. Here, we show that older individuals make broadly neutralizing antibodies that have no hemagglutination-inhibiting activity and are less potent than strain-specific antibodies. These antibodies could drive viral escape from neutralization but did not result in escape from binding. Given their different mechanisms of action, they might retain protective activity even against escape variants.

INTRODUCTION

Influenza virus infections occur seasonally, causing significant global morbidity and mortality, despite annual vaccinations. Like influenza A viruses, influenza B viruses (IBVs) cause significant morbidity and mortality worldwide, especially in children and the elderly. Additionally, IBVs have been predominant during some influenza virus seasons, with a strong presence during the 2019–2020 season in the United States (1). There are two lineages of IBV that cocirculate annually, the B/Victoria/2/1987-like (B/Victoria) and B/Yamagata/16/1988-like (B/Yamagata) lineages, defined based on their hemagglutinin (HA) sequences.

Individuals are exposed to IBVs from both lineages repeatedly throughout their lifetimes either by natural infection or by vaccination. These exposures cause antibody responses that are typically focused on the HA head domain (2, 3). It has been shown that influenza virus-specific immune memory can change future adaptive immune responses. Elderly individuals have had the most opportunities for multiple exposures to infection and vaccination over their lifetimes. Interestingly, elderly individuals tend to have limited adaptability to antigenically drifted IBV strains, forcing their antibodies to target more-conserved epitopes (4). Because of this, antibodies isolated from these individuals are much more cross-reactive than those from younger adults who received the same vaccination, but they are often inactive in hemagglutination inhibition (HI) assays (4). Thus, these antibodies may not be as potent as typical HI antibodies induced in younger individuals (4). Recently (in 2019), Henry et al. isolated cross-reactive monoclonal antibodies (MAbs) from elderly individuals that were HI inactive and did not compete for epitopes targeted by known stalk MAbs (4). Similar MAbs have been shown previously to be neutralizing, protective in vivo, and cross-reactive, although the neutralizing potency in vitro was lower than that of HI-active MAbs (5). This unique subset of MAbs (HI inactive, neutralizing, and HA head specific) has been largely understudied for IBVs. Therefore, we chose to characterize these antibodies specifically because of their unique properties.

Here, we describe the characterization of four MAbs isolated from elderly individuals (aged 71 to 74 years) after vaccination. These elderly individuals were recruited between 2006 and 2011 after being vaccinated with the seasonal trivalent vaccine containing either the B/Malaysia/2506/2004 or the B/Brisbane/60/2008 vaccine strain (both from the B/Victoria lineage). The antibodies, designated AG13-3F04, AG5-E04, AG10-A05, and AG13-3C02, were described as HI inactive, neutralizing, and not targeting known stalk MAb epitopes (4). To further understand the interactions of these MAbs with the IBV HA, we characterized their cross-reactivity, neutralization activity, and ability to provide protection from lethal infection in vivo. Using escape mutagenesis, we identified several different residues involved in escape phenotypes and observed that escape mutant viruses did not lose fitness in the mouse model.

RESULTS

HI-inactive MAbs broadly bind B/Victoria and B/Yamagata lineage viruses.

We decided to investigate the breadth of each MAb in order to determine if they are only specific to the previously tested viruses (B/Malaysia/2506/2004, B/Brisbane/60/2008, and occasionally B/Florida/04/2006) or if they have cross-reactivity to other B/Yamagata and B/Victoria lineage viruses (4). To do this, we performed an enzyme-linked immunosorbent assay (ELISA) using purified IBVs or recombinant IBV HA proteins. These viruses or recombinant HA proteins were from both the B/Victoria and B/Yamagata lineages and spanned several decades. Their antigenic differences are depicted in a phylogenetic tree based on the amino acid compositions of the different HA proteins (Fig. 1A). All four MAbs had high reactivity within the B/Victoria lineage (Fig. 1B). MAbs AG13-3F04 and AG5-E04 had the most cross-reactivity with B/Yamagata lineage viruses; however, binding was reduced against more-recent strains (B/New York/PV00094/2017, B/Wisconsin/1/2010, B/Massachusetts/02/2012, and B/Phuket/3073/2013). MAb AG10-A05 had cross-reactivity only against the ancestral B/Lee/1940 and B/Yamagata/16/1988 strains. AG13-3C02 was able to bind all strains tested aside from the most recent B/Yamagata lineage virus, B/New York/PV00094/2017. We also included CR8033 (an HI-active MAb) and CR8059 (which binds the conserved vestigial esterase region of the HA) in these ELISAs in order to determine how the binding of these MAbs to IBV HAs changed with time (6). We found that CR8033 did not cross-react with B/Yamagata/16/88 or B/New York/PV01181/2018 (a B/Victoria lineage virus) but otherwise shared the cross-reactivity profile of MAbs AG5-E04 and CR8059. Overall, these MAbs are very cross-reactive, with some binding viruses that span 60 years of evolution.

FIG 1.

HI-inactive MAbs broadly bind B/Victoria and B/Yamagata lineage viruses. (A) A maximum likelihood tree composed of approximately 200 influenza B virus HA protein sequences from both the B/Victoria/2/1987-like (green) and B/Yamagata/16/1988-like (purple) lineages was built. Viruses used in ELISAs are labeled. (B) Minimum binding concentrations of each MAb for B/Victoria/2/1987-like (V) and B/Yamagata/16/1988-like (Y) lineage viruses.

HI-inactive MAbs are neutralizing, utilize effector functions, and are protective in vivo.

After determining the binding breadth of these HI-inactive MAbs, we wanted to further characterize these MAbs using the B/Victoria vaccine strains B/Malaysia/2506/2004 and B/Brisbane/60/2008, since all MAbs displayed strong binding to these strains (Fig. 1). Using plaque reduction neutralization assays, we found that B/Malaysia/2506/2004 was neutralized efficiently by AG13-3F04 (50% inhibitory concentration [IC₅₀], 0.03328 μg/ml), AG5-E04 (IC₅₀, 0.0333 μg/ml), and AG13-3C02 (IC₅₀, 0.6429 μg/ml). However, B/Malaysia/2506/2004 was not neutralized efficiently by AG10-A05 (IC₅₀, 11.5 μg/ml). Each MAb neutralized B/Brisbane/60/2008, with IC₅₀s of 0.5957 μg/ml (AG13-3F04), 1.028 μg/ml (AG5-E04), 0.1926 μg/ml (AG10-A05), and 0.07664 μg/ml (AG13-3C02) (Fig. 2A and B).

FIG 2.

HI-inactive MAbs are neutralizing and active in an ADCC reporter assay. HI-inactive MAbs were characterized using vaccine strains B/Malaysia/2506/2004 and B/Brisbane/60/2008. (A) Neutralization curves for each MAb as determined by a plaque reduction assay (n = 2). (B) IC₅₀ values for plaque reduction assays of each MAb. (C) ADCC reporter activity curves for each MAb (n = 2). MAb CR9114 was used as a positive control. (D) AUC values for ADCC reporter activity.

Fc effector functions are utilized by a variety of broadly cross-reactive MAbs to provide protection (7). To investigate if our MAbs utilize effector functions similarly to other cross-reactive MAbs, such as the anti-stalk MAb CR9114 (6), we performed antibody-dependent cell-mediated cytotoxicity (ADCC) reporter assays against both wild-type B/Malaysia/2506/2004 and B/Brisbane/60/2008 (8). All MAbs had ADCC activity against both B/Victoria lineage strains, B/Malaysia/2506/2004 and B/Brisbane/60/2008 (Fig. 2C). Compared to the stalk MAb CR9114, these MAbs have approximately 2- to 5-fold-lower activity. Area-under-the-curve (AUC) analyses indicate that the ADCC activities of each MAb are similar for B/Malaysia/2506/2004 and B/Brisbane/60/2008 (Fig. 2C and D).

It has been demonstrated that MAbs with neutralization and/or ADCC activity can mediate protection against viral challenge in vivo (9–13). Here, we chose to use B/Malaysia/2506/2004 for protection experiments, because the murine 50% lethal dose (mLD₅₀) of B/Brisbane/60/2008 is considerably higher. DBA2/J mice were used because they are more susceptible to B/Malaysia/2506/2004 infection than wild-type mice (14, 15). Mice were given 3 mg of MAb/kg of body weight intraperitoneally (i.p.), and 3 h later, the mice were challenged intranasally with 7.5× mLD₅₀ of B/Malaysia/2506/2004 (2.37 × 10⁴ PFU/mouse). Body weight was monitored for 14 days postchallenge, as shown in Fig. 3A, and mice were euthanized after a loss of 25% of their initial body weight (Fig. 3B). The most neutralizing MAbs, AG13-3F04 and AG5-E04, were 100% protective, with mice losing approximately 10% of their body weight before fully recovering (Fig. 3A). Prophylactic administration of AG10-A05 was not protective in mice, and AG13-3C02 was 60% effective, with protected mice losing 10 to 15% of their body weight before recovering (Fig. 3A and B). Our data suggest that the most neutralizing antibodies, AG13-3F04 and AG5-E04, can confer 100% protection against lethal influenza virus, while survival decreases for less-potent neutralizing antibodies. Protection appears to be independent of ADCC activity, considering that all MAbs had similar activities, indicating that these MAbs may provide protection using neutralization alone.

FIG 3.

Antibodies show ranges in protectiveness against lethal challenge in the DBA2/J mouse model. DBA2/J mice (n = 5) were given 3 mg/kg of each MAb 3 h prior to a lethal challenge with B/Malaysia/2506/2004. Data are shown in red for mice vaccinated with AG13-3F04, in green for AG5-E04, in orange for AG10-A05, in blue for AG13-3C02, and in gray for the irrelevant IgG control MAb. (A) The percentage of initial body weight was measured daily for 14 days. (B) Survival curves.

Identifying residues critical for MAb neutralization via escape mutagenesis.

These MAbs were found to be inactive in hemagglutinin inhibition assays but did not compete with known stalk antibodies (4). These data suggest that there are other residues on the HA that are important for the binding of the MAbs to HA. To determine which HA residues were important in MAb-mediated neutralization, escape mutagenesis was performed. Briefly, 1 × 10⁴ PFU of B/Brisbane/60/2008 was mixed and incubated with each MAb at 0.5× IC₅₀. Then the mixture was injected into 8-day-old embryonated eggs and incubated at 33°C for 3 to 4 days. Subsequent passages used a 1:1 ratio of allantoic fluid and the MAb at double the IC₅₀ concentration used in the previous passage. After 8 to 9 passages (64× or 128× IC₅₀), viruses were plaque purified in the presence of 100 μg of MAb. The viruses isolated contained 3 to 4 mutations in the HA head domain (Table 1). Mutations K90E, G164E, D212N/S, and T214A are shared by two or more escape mutant viruses (EMVs). Of note, the D212S mutation was also identified in the copassaged irrelevant IgG control virus. Interestingly, many mutations were found within or adjacent to identified antigenic sites (Fig. 4A). Together, the D212N mutation and the wild-type T214 residue would introduce a potential N-linked glycosylation site. However, we observed that when D212N occurred, it was coupled with T214A, avoiding this addition. No other mutations were found to be involved in HA glycosylation. To examine the prevalence of mutations at the residues identified, we conducted in silico selection analysis of 104 B/Victoria lineage viruses. This analysis indicated that I91 and S135 are under high positive selection, D212 is under moderate positive selection, and the remaining mutated residues are not under positive selection (Fig. 5A). The findings for the HA head domain for these isolates were converted into a sequence logo to allow for the examination of residues present at the mutated sites. Residues K90, I91, G164, and S270 are variable within B/Victoria lineage viruses; however, residues S135, N212, and T214 are more conserved (Fig. 5B).

TABLE 1.

Mutations found in segments of IBV HA EMVs

Virus	Mutation(s) in:
	PB2	PB1	HA
AG13-3F04 EMV			K90E, D212N, T214A, S270F
AG5-E04 EMV		D109N	K90E, G164E, D212S
AG10-A05 EMV			I91T, G164E, D212S
AG13-3C02 EMV	A552T		S135L, D212N, T214A
Irrelevant IgG control virus			D212S

FIG 4.

Characterization of MAb EMVs. (A) Location of each HA mutation on the structure of B/Brisbane/60/2008 (PDB ID 4FQM) (6), visualized using PyMOL software. The major antigenic sites of the head domain—the 120 loop (blue), the 150 loop (green), the 160 loop (teal), and the 190 helix (red)—are indicated on the structure (30). Mutations K90E, G164E, and T214A are located within antigenic sites. (B) (Top) Immunofluorescence of MAb binding to EMVs. MDCK cells were infected at an MOI of 1 and were stained using 50 μg of the antibody indicated at the top. Images are representative of binding from two IF experiments. (Bottom) A corresponding heat map describes the percentage of luminescence for each image (with luminescence for B/Brisbane/60/2008 set to 100%). (C) The neutralization of each EMV by each MAb was determined using a plaque reduction assay. The fold change was calculated by comparing the IC₅₀ value for each EMV to that for the wild-type virus B/Brisbane/60/2008. Higher fold changes indicate stronger escape phenotypes. The maximum concentration of MAb used to determine IC₅₀ values was 100 μg/ml.

FIG 5.

Positive selection and conservation of B/Victoria/2/1987-like viruses. (A) Selection values (ω) for each residue of the influenza B virus HA head domain. Amino acid positions (residues 57 to 308) are listed along the x axis, while ω values are on the y axis. Residues mutated in EMVs are marked with a red letter “X.” (B) Sequence logo (created by WebLogo) illustrating the amino acid substitutions at each residue of the HA head domain. Residues where escape mutations were present are boxed in red.

HA escape mutations ablate MAb neutralization but do not always alter MAb binding.

We next investigated how the mutations acquired affected MAb-HA interactions relative to those for the wild-type virus B/Brisbane/60/2008. To begin characterizing our mutant viruses, we tested whether the mutations identified changed antibody binding or neutralization. Immunofluorescence (IF) of these mutant viruses showed that while some MAbs no longer bound the virus against which they were raised, a majority of MAbs retained binding to their respective mutant viruses (Fig. 4B). MAbs AG13-3F04 and AG5-E04 had reduced binding to both the AG13-3F04 and AG5-E04 EMVs, which share the K90E mutation. Similarly, MAb AG10-A05 showed reduced binding to its EMV. Finally, all MAbs showed somewhat reduced binding to the AG13-3C02 EMV, at least when the IF was quantified, as shown in the heat map in Fig. 4B. Overall, we observed little to no changes in MAb binding to most EMVs. To confirm escape from MAb neutralization, we performed plaque reduction neutralization experiments with each EMV and the panel of MAbs (Fig. 4C). We were able to confirm that each escape mutant was no longer neutralized by the specific MAb against which it was raised. Additionally, the AG13-3F04 and AG5-E04 EMVs had increased resistance to AG10-A05.

Escape mutant viruses exhibit fitness changes in mice.

Frequently, escape mutations were found to alter viral fitness in vivo (16–19). Our escape mutant viruses contained mutations that are found in naturally occurring strains, and thus, we wanted to investigate if there were fitness changes in the mouse model. We performed mLD₅₀ experiments for each escape mutant virus and the irrelevant IgG control virus in DBA2/J mice. The irrelevant IgG control virus had an mLD₅₀ of 1.78 × 10⁴ PFU/mouse, while the majority of EMVs had lower mLD₅₀ values. The AG13-3F04 and AG5-E04 EMVs had decreased mLD₅₀ values, at 178 and 25 PFU/mouse, respectively. The AG10-A05 EMV had a decreased mLD₅₀ value as well, at 316 PFU/mouse. Only the AG13-3C02 EMV had an mLD₅₀ higher than that of the irrelevant IgG control virus, at 3.16 × 10⁴ PFU/mouse. AG13-3F04 and AG5-E04 were the strongest neutralizing MAbs, and their EMVs had the largest changes in pathogenicity. EMV mLD₅₀ values are listed in Table 2. These two EMVs share the mutation K90E, which was the driving mutation for escape from both AG13-3F04 and AG5-E04 (Fig. 6). MAb AG10-A05 was the poorest neutralizer; however, its EMV was still more lethal than the control virus, and its mLD₅₀ was equivalent to that of wild-type B/Malaysia/2506/2004 (316 PFU). It may be that changes at residues important for strong neutralizing MAbs push the virus toward a more fit phenotype in mice, e.g., through stronger binding to sialic acid.

TABLE 2.

Summary of EMV mLD₅₀ values

Virus	mLD50 (PFU/mouse)
B/Malaysia/2506/2004	316
Irrelevant IgG control virus	1.78 × 104
AG13-3F04 EMV	178
AG5-E04 EMV	25
AG10-A05 EMV	316
AG13-3C02 EMV	3.16 × 104

FIG 6.

Impact of HA point mutations on MAb neutralization. The neutralization of rescued point mutant viruses by each HI-inactive MAb was determined by a plaque reduction assay (n = 2). The fold change was determined by comparing the IC₅₀ value for each rescued mutant virus to that of the rescued virus carrying wild-type B/Brisbane/60/2008 HA. Higher fold changes indicate stronger escape phenotypes.

K90E is a critical residue for MAb neutralization.

To determine which mutations were directly linked to the escape mutant phenotypes, point mutations were introduced into the HA of B/Brisbane/60/2008, and the viruses were rescued in a B/Malaysia/2506/2004 backbone. Immunofluorescence assays using point mutant rescue viruses showed that all were bound equivalently by each MAb (data not shown). Plaque reduction assays showed that a majority of point mutations caused modest changes in IC₅₀ values, ranging from 1.1-fold to 7-fold increases. However, these IC₅₀ values still remained less than 1 μg/ml, indicating that the MAbs could still neutralize each virus efficiently. The K90E mutation increased the IC₅₀ values of AG13-3F04 and AG5-E04 almost 500-fold over that of the wild-type B/Brisbane/60/2008 virus, from 0.01 μg/ml to 5 μg/ml, indicating complete escape in our experimental setup. These data are summarized in Fig. 6. It can be inferred that the majority of escape phenotypes observed are mediated by a combination of mutations rather than by one single escape mutation (aside from K90E).

DISCUSSION

Here, we described four HI-inactive MAbs that were isolated from elderly individuals after vaccination with B/Victoria/2/1987-like lineage viruses (either B/Malaysia/2506/2004 or B/Brisbane/60/2008). Their lack of HI activity suggests that while these MAbs bind to the globular head domain of HA, their orientation does not block access to sialic acid. Binding of more-recent strains was reduced. We observed one MAb (AG10-A05) that was not cross-reactive with B/Yamagata/16/1988-like lineage viruses other than B/Yamagata/16/1988 and B/Lee/1940.

Cross-reactive, stalk-specific MAbs show high ADCC activity and use this to confer protection (5, 6). While our MAbs are cross-reactive and have ADCC activity (2- to 5-fold lower than that of the stalk MAb CR9114), they do not depend on this activity for protection in this mouse model. All MAbs had similar ADCC activity against B/Malaysia/2506/2004, but the poorest neutralizer, AG10-A05, was not protective. AG13-3C02 was an intermediate neutralizer of B/Malaysia/2506/2004 and was found to be 60% protective. MAbs AG13-3F04 and AG5-E04 were both very strong neutralizers and 100% protective. These results may have been different if we had used B/Brisbane/60/2008 as the challenge virus, because the MAbs neutralize it more effectively than they do B/Malaysia/2506/2004. However, because of the high mLD₅₀ value for B/Brisbane/60/2008, we chose not to use this virus in our protection studies. The caveat in this case is that these are human antibodies that were used in a mouse model, and their interactions with murine Fc receptors might be suboptimal.

By comparing the HA sequences of B/Malaysia/2506/2004 and B/Florida/04/2006 to that of B/Brisbane/60/2008, we were able to identify natural variation at residues shared with EMVs (Table 3). B/Malaysia/2506/2004 had mutations K90N, D212N, and T214I, while B/Florida/04/2006 had mutations K90T, I91T, G164K, and S270P. None of these mutations appeared to cause changes in HA N-linked glycosylation. It is interesting that while B/Florida/04/2006 has several mutations at positions similar to those in EMVs, all MAbs aside from AG10-A05 exhibited binding activity. MAb AG10-A05 caused an I91T mutation. B/Florida/04/2006, like the AG10-A05 EMV, has the I91T mutation, which may explain the lack of binding.

TABLE 3.

Amino acid differences between B/Malaysia/2506/2004, B/Brisbane/60/2008, and the B/Yamagata lineage virus B/Florida/04/2006

Position	Residue in:			Escape mutation
	B/Malaysia/2506/2004	B/Brisbane/60/2008	B/Florida/04/2006
90	N	K	T	K90E
91	I	I	T	I91T
135	S	S	S	S135L
164	G	G	K	G164E
212	N	D	D	D212N/S
214	I	T	T	T214A
270	S	S	P	S270F

As influenza B viruses adapt to present antibody responses, they acquire mutations in the HA, with a majority in the head domain (20). In fact, point mutant viruses revealed that a majority of escape mutant phenotypes were not caused by a single amino acid mutation. This is in contrast to previous work, which identified a single mutation, T214P, that led to the loss of binding and neutralization activity of MAb CR8033 and other recently identified MAbs (5). It is interesting that we identified a mutation at T214; however, this change was to an alanine, not a proline, and did not significantly impact the binding or neutralization of the MAbs tested, including CR8033. This may be because both threonine and alanine are small, nucleophilic amino acids that differ by the addition of a hydroxyl group. Based on the similar properties of these amino acids, the change may not be significant enough to fully ablate the binding and neutralization of antibodies that recognize that epitope. However, a threonine-to-proline substitution adds an α-amino group and a larger, more structured side chain and also effects a change from a nucleophilic to a hydrophilic residue. These differences may lead to biochemical/structural changes that could ablate antibody binding and neutralization more effectively than the substitution with an alanine.

Only the K90E mutation caused complete escape from antibodies AG13-3F04 and AG5-E04; however, it did not cause a loss of antibody binding. Antibodies such as AG13-3F04 and AG5-E04 are strong neutralizers, a characteristic that could be attributed to their targeting of the residue K90, located in the 120-loop antigenic site. AG13-3F04 and AG5-E04 EMVs became more resistant to both MAbs, along with AG10-A05. These MAbs neutralized B/Malaysia/2506/2004 (N90) better than B/Brisbane/60/2008 (K90), suggesting that not all substitutions at residue 90 contribute to escape from neutralization. Several point mutations affected neutralization by MAb AG10-A05, but not to the extent of the K90E mutation. This MAb is the only antibody we tested that does not cross-react between B/Victoria and B/Yamagata lineages. Additionally, it is not protective against a lethal challenge with B/Malaysia/2506/2004, whose head domain contains mutations at K90 and D212. Antibody AG13-3C02 was also affected by the K90E mutation. This antibody does not cross-react well with recent B/Yamagata lineage isolates. This phenotype is most likely caused by a combination of the residues that are not conserved between B/Brisbane/60/2008 and B/Florida/04/2006 (K90T, I91T, G164K, and S135P).

The K90 residue is interesting because it affects the neutralization efficiency of every MAb tested. Escape from the most cross-reactive and strongest neutralizing MAbs, AG13-3F04 and AG5-E04, was most affected by the K90E mutation. Additionally, the AG13-3F04 and AG5-E04 EMVs had the greatest increase of fitness in the mouse model. It should be noted that the K90 residue is conserved between B/Victoria/2/1987 and B/Brisbane/60/2008. It may be that the K90 residue was a key target for MAb neutralization and that changes at that position continue the evolution of the lineage by increasing fitness. However, it should be noted that the AG5-E04 (the most fit) and AG13-3C02 (the least fit) EMVs contained mutations in PB1 and PB2, respectively. Previous reports identified mutations in PB1 (A193T) and PB2 (Q78K) of mouse-adapted B/Brisbane/60/2008 (21). These mutations were found in viruses with increased lethality in the mouse model. However, these mutations are not the same as those that we identified in our EMVs (D109N in PB1 and A552T in PB2). Additionally, the authors hypothesized that an identified PA mutation (K338R) might be the main influencer for increased lethality (21). The effects of the polymerase mutations found in our study on viral fitness have not been described.

The mechanism of neutralization for our MAbs is currently unknown. Since they are HI inactive, we can assume that they are not directly inhibiting access to sialic acid. One possible mechanism could involve steric hindrance of the NA. By binding the HA head domain at odd angles, the antibodies may block access to sialic acids by the NA, preventing virus spread to neighboring cells. A second possibility is that the MAbs could cross-link HA molecules, connecting the virions to each other or anchoring them to cells expressing HA on their surfaces (22).

The development of universal influenza B virus vaccines involves chimeric or mosaic IBV HA antigens that are intended to boost broadly protective antibodies to conserved epitopes. The chimeric and mosaic constructs contain conserved stalk domains but have either the whole HA head domain (chimeric) or the HA antigenic sites (mosaic) replaced with HA sequences from influenza A virus subtypes such as H5, H8, and H13 (12, 23). Chimeric HA-based vaccines boost stalk-reactive antibodies, while mosaic HA-based vaccines are designed to boost both anti-stalk and cross-reactive anti-head antibodies. It is therefore important to better understand the epitopes of cross-reactive antibodies targeting influenza B viruses as well as the consequences of immune pressure on these epitopes. The data generated in this study will help to optimize these universal influenza virus vaccine candidates.

MATERIALS AND METHODS

Viruses and recombinant protein.

The B/Brisbane/60/2008 and B/Malaysia/2506/2004 viruses were obtained from the International Reagent Resource and were propagated in embryonated 10-day-old specific-pathogen-free (SPF) chicken eggs (Charles River Laboratories) at 33°C for 3 to 4 days. Eggs were inactivated by overnight storage at 4°C. Next, allantoic fluid was collected and centrifuged at 2,000 × g for 10 min at 4°C. Viruses were aliquoted and stored at –80°C. For purified viruses, allantoic fluid was added to round ultracentrifuge tubes (Thermo Scientific). NTE buffer (100 mM NaCl, 10 mM Tris-HCl, 1 mM EDTA) containing 30% sucrose was slowly added to each tube to form a sucrose cushion. The tubes were ultracentrifuged at 25,000 rpm using an SW28 rotor for 2 h at 4°C. The supernatant was removed, and pellets were resuspended in a total of 1 ml of 1× PBS (phosphate-buffered saline; Gibco). Purified virus was then aliquoted and stored at –80°C. Recombinant protein was produced using a baculovirus expression system as described previously (24). Protein concentrations for purified viruses and recombinant proteins were determined using a Bradford developing reagent and were based on a standard curve.

MAbs.

Variable-heavy and variable-light gene sequences were recovered from a single plasmablast isolated from each elderly individual (71 to 74 years of age) after vaccination with the B/Malaysia/2506/2004 or B/Brisbane/60/2008 virus as described previously (25). Briefly, the B cell receptor heavy- and light-chain variable regions were amplified and cloned into IgG-AbVec expression vectors carrying human immunoglobulin constant regions followed by a simian virus 40 (SV40) polyadenylation sequence (25). HEK293T cells were cotransfected, and antibody was purified as described previously (4).

Enzyme-linked immunosorbent assay.

Flat-bottom Immulon 4HBX microtiter plates (Thermo Scientific) were coated with 50 μl of either 2-μg/ml recombinant HA proteins or 5-μg/ml purified virus in 1× KPL coating solution (SeraCare) and were incubated overnight at 4°C. The coating buffer was discarded, and wells were blocked with 200 μl of blocking solution (1× PBS containing 0.1% Tween 20, 3% goat serum [Gibco], and 0.5% milk powder [AmericanBio]) per well. Antibodies were diluted to 30 μg/ml in blocking solution, serially diluted 2-fold, and incubated at 20°C for 2 h. The plates were then washed three times with 220 μl of PBS containing 0.1% Tween 20 (PBS-T) per well using the AquaMax 3000 automated plate washer. Next, a peroxidase-conjugated anti-human IgG (Fab-specific) antibody (Sigma) at a 1:3,000 dilution in blocking solution was added at 50 μl per well, and the mixture was incubated for 1 h at 20°C. The plates were washed again three times and were developed with 100 μl per well of SigmaFast OPD (o-phenylenediamine dihydrochloride; Sigma) for 10 min at room temperature. Development was stopped by the addition of 3 M hydrochloric acid, and the optical density (OD) at 490 nm was read using a Synergy H1 hybrid multimode microplate reader (BioTek).

Plaque assay.

Virus titers were determined using a plaque assay. One day prior, Madin-Darby canine kidney (MDCK) cells (ATCC no. PTA-6500) were seeded onto sterile 12-well plates at a concentration of 8 × 10⁵/ml in complete Dulbecco’s modified Eagle medium (cDMEM) (1× DMEM [Gibco] containing 10% heat-inactivated fetal bovine serum [Sigma-Aldrich], 1% penicillin [100 U/ml]–streptomycin [100 μg/ml] solution, and 1% 1 M HEPES) and were incubated at 37°C under 5% CO₂ overnight. The following day, allantoic fluid was serially diluted 1:10 six times in 1× minimal essential medium (1× MEM) (10% 10× MEM [Gibco], 2 mM l-glutamine [Gibco], 0.1% sodium bicarbonate [Gibco], 1% 1 M HEPES [Gibco], 1% penicillin [100 U/ml]–streptomycin [100 μg/ml] solution [Gibco], and 0.2% bovine serum albumin [BSA]). Cells were washed once with 1× PBS and were replaced with 200 μl of diluted virus per well. Plates were incubated at 33°C under 5% CO₂ for 40 min, with shaking every 10 min. The inoculum was removed and replaced with an overlay containing 2× MEM supplemented with 0.1% diethylaminoethyl (DEAE) dextrane, 1 μg/ml tosylsulfonyl phenylalanyl chloromethyl ketone (TPCK)-treated trypsin, and 0.64% Oxoid agarose. B/Malaysia/2506/2004 virus plates were then incubated for 3 days, while B/Brisbane/60/2008 virus plates were incubated for 5 days, at 33°C under 5% CO₂. The cells were fixed with 3.7% paraformaldehyde (PFA) overnight at 4°C. The following day, the overlay was removed, and cells were stained with a solution of 20% methanol containing 0.5% crystal violet powder.

Plaque reduction neutralization assay.

MDCK cells were seeded onto 12-well plates at a concentration of 8 × 10⁵/ml in cDMEM and were incubated overnight at 37°C under 5% CO₂. The following day, antibodies were first diluted to 100 μg/ml in 300 μl 1× MEM and then serially diluted 1:5 six times to a final concentration of 0.032 μg/ml in sterile 24-well plates. Approximately 1 × 10³ PFU was added to each well, and the MAb-virus mixture was incubated at room temperature for 1 h with shaking. The cells were washed once with 1× PBS, infected with 200 μl of the MAb-virus mixture, and incubated at 33°C under 5% CO₂ for 40 min, with shaking every 10 min. In the meantime, the MAb was first diluted to 100 μg/ml in 500 μl 2× MEM and then serially diluted 1:5 six times in 24-well plates. Next, 180 μl of a mixture containing 0.1% DEAE dextrane and 1 μg/ml TPCK-treated trypsin in sterile water for injection (Gibco) was added to each well. The inoculum was removed and immediately replaced by an overlay containing the diluted MAbs and 360 μl of 2% Oxoid agarose. B/Malaysia/2506/2004 virus plates were then incubated for 3 days, while B/Brisbane/60/2008 virus plates were incubated for 5 days, at 33°C under 5% CO₂ and were fixed with 3.7% PFA. The overlay was removed, and cells were stained with a solution of 20% methanol containing 0.5% crystal violet powder.

ADCC assay.

Antibody-dependent cell-mediated cytotoxicity (ADCC) assays were performed using the Promega ADCC kit. Briefly, sterile opaque 96-well plates were seeded with 3 × 10⁴ cells/well in cDMEM and were incubated overnight at 37°C under 5% CO₂. Cells were washed once with 1× PBS, infected with 3 × 10⁵ PFU of virus, and incubated overnight at 33°C under 5% CO_2. The following day, antibodies were serially diluted 1:2 in Roswell Park Memorial Institute (RPMI) 1640 medium from a starting concentration of 100 μg/ml. Each plate also contained a no-MAb, virus-only control. The cDMEM was carefully aspirated and replaced with 25 μl RPMI 1640 medium and 50 μl of the antibody dilutions. ADCC bioassay effector cells (human FcγRIIIa) diluted to 75,000/well in 25 μl of RPMI 1640 medium were added to each well. The plates were then incubated at 37°C under 5% CO₂ for 6 h. The ADCC kit luciferase assay reagent was added to each well at a volume of 75 μl, and the result was immediately read using the luminescence setting on the Synergy H1 hybrid multimode microplate reader (BioTek). The data were analyzed by subtracting the no-antibody control luminescence from the total luminescence per well.

HA assay.

Virus containing allantoic fluid was diluted 1:2 in 96-well V-bottom plates (Thermo Fisher). Next, turkey red blood cells (catalog no. 720943; Lampire Biological) were diluted to 0.5% and added to virus dilutions at a 1:1 ratio. Plates were incubated for 40 min at 4°C. HA titers were based on the first well to contain a pellet.

Mouse experiments.

All animal experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee at the Icahn School of Medicine at Mount Sinai. Female mice (species, Mus musculus; strain, DBA2/J) at the age of 6 to 8 weeks (The Jackson Laboratory) were used for all experiments. Mice were randomly selected, 3 per group, for determination of the mLD₅₀ of each virus. For protection experiments, 5 mice were randomly selected per group and were given 3 mg/kg of antibody diluted in 1× PBS via intraperitoneal (i.p.) injection. Three hours later, mice were challenged intranasally with 7.5 times the mLD₅₀ of the B/Malaysia/2506/2004 virus. Weight loss was measured daily for 14 days, and mice were sacrificed when weight loss was >25% of the initial body weight.

Escape mutagenesis.

MAb EMVs were produced by serially passaging the wild-type B/Brisbane/60/2008 virus in 8-day-old embryonated chicken eggs. The virus (1 × 10⁴ PFU) was first mixed with the MAb at 0.5 times the IC₅₀, and the mixture was incubated for 1 h at room temperature with shaking. Then 200 μl of the mixture was injected into eggs, which were incubated for 4 days at 33°C. Allantoic fluid was harvested as described above. The presence of a virus was confirmed using the HA assay. If a virus was present, the virus and the MAb at 1× IC₅₀ were incubated at a 1:1 ratio for 1 h and were then injected into the eggs (total volume, 200 μl). This procedure was repeated until 64× or 128× IC₅₀ was reached, at which point the virus was plaque purified and deep sequenced. EMVs were propagated and stored as described above.

Generation of point mutation viruses.

B/Brisbane/60/2008 RNA was extracted, and the HA segment was amplified using a SuperScript III reverse transcription-PCR (RT-PCR) kit. The primers used for HA amplification were designed based on the N-terminal and C-terminal portions of the B/Malaysia/2506/2004 virus with its noncoding regions according to the In-Fusion cloning kit (Clontech) specifications. The pDZ expression vector was linearized using the SapI restriction enzyme (New England Biolabs). The wild-type HA segment cDNA was amplified and inserted into the linearized plasmid using the Infusion Cloning Kit according to the manufacturer’s protocol. Additionally, point mutations were introduced into the wild-type HA using a 2-step PCR with primers containing the single point mutation coupled with the In-Fusion cloning primers. The presence of point mutations was confirmed using Sanger sequencing (Macrogen). The mutated HA segments were then inserted into the linearized pDZ vector. Reassortant viruses were rescued in HEK293T cells as described previously (23). Briefly, HEK293T cells were seeded at 7 × 10⁵/ml in cDMEM and incubated overnight at 37°C under 5% CO₂. The next day, 0.7 μg of the HA segment and 2.8 μg of a pRS 7 segment plasmid that expresses the seven gene segments of the B/Malaysia/2506/2004 virus (26) were mixed in 250 μl Opti-MEM (Gibco) and then incubated at a 1:5 ratio with the TransIT-LT1 transfection reagent (Mirus) at room temperature for 20 min. The cells were then washed once with 1× PBS, and the medium was replaced with DMEM containing 1% penicillin (100 U/ml)–streptomycin (100 μg/ml) solution (Gibco) and 0.3% BSA. The transfection mixture was added in a dropwise fashion and then incubated at 33°C under 5% CO₂ for 48 h. The cell supernatant was collected and homogenized using an insulin syringe, and 200 μl was injected into 8-day-old SPF eggs and incubated for 3 days at 33°C. Allantoic fluid was collected, and a plaque assay was performed. A single isolated plaque was used to make the final viral stock. Point mutation viruses were grown in eggs as described above.

RNA extraction and deep sequencing.

RNA from allantoic fluid was extracted using the TRIzol LS reagent (Ambion) as per the manufacturer’s instructions. RNA was quantified using a NanoDrop One Microvolume UV-Vis spectrophotometer (Thermo Scientific) on the RNA setting and was then stored at –80°C for future use. The RNA was sequenced by next-generation sequencing using the MiSeq reagent kit, v2 (300 cycles; Illumina). Genome assembly was performed as described previously using a customized pipeline implemented at the Icahn School of Medicine at Mount Sinai (27). Assembled sequences for each segment were aligned using MUSCLE in MEGA 7.0 (28). Mutations were determined based on differences from the deep-sequenced wild-type B/Brisbane/60/2008 virus used at the beginning of escape passaging.

Phylogenetic tree.

The phylogenetic tree was built from 200 IBV HA protein sequences from both the B/Victoria/2/1987-like and B/Yamagata/16/1988-like lineages. Sequences were downloaded from the Global Initiative on Sharing All Influenza Data (GISAID) (http://platform.gisaid.org/) and were aligned using MUSCLE in MEGA 7.0 (28). The tree was constructed using RAxML and was visualized through FigTree (29).

Immunofluorescence.

Sterile 96-well flat-bottom plates (Corning) were seeded with 2 × 10⁴ MDCK cells per well and were incubated overnight at 37°C under 5% CO₂. Cells were then washed once with 200 μl of sterile PBS. The virus was diluted to a multiplicity of infection (MOI) of 1 (1 × 10⁵ PFU) in 1× MEM and was added at 100 μl per well. The plates were incubated at 33°C under 5% CO₂ for 18 h. Plates were then inactivated with 200 μl per well of 3.7% PFA for 1 h at room temperature or overnight at 4°C. The PFA was removed, and cells were blocked with 1× PBS containing 3% milk powder (AmericanBio) and were incubated for 1 h at room temperature. The blocking solution was removed, and cells were incubated with 50 μg of MAb per well diluted in 1× PBS containing 1% milk powder for 1 h at room temperature with shaking. The liquid was then removed, and plates were washed three times using 200 μl of 1× PBS per well. Alexa Fluor 488-conjugated goat anti-human IgG(H+L) (Invitrogen) was added to 1× PBS containing 1% milk powder at a 1:500 dilution. The plates were then incubated in the dark at room temperature with shaking. The liquid was removed, and cells were again washed three times with 1× PBS. A final volume of 50 μl 1× PBS was added to each well. Immunofluorescence was visualized using the Celigo S imaging cytometer for adherent cells (Nexcelom Bioscience) using the 2-channel Target 1+2 (merge) setting. Exposure time and gain were automated by the machine. The focus was set on a single channel using image-based autofocus with the 488-nm signal as the target. Images were captured from each well after the entire plate had been scanned. Additionally, total fluorescence values were collected via analysis using default settings and were downloaded in a .csv format. The total integrated intensity of fluorescence for the wild-type B/Brisbane/60/2008 virus was set at 100%, and all other viruses were compared to these values in order to calculate the percentage of fluorescence.

Data analysis.

Calculations were performed using PRISM 7.0. IC₅₀ and AUC values were calculated using nonlinear regression (4 parameters) based on log₁₀-transformed antibody concentrations. Graphs were generated using the PRISM 7.0 program as well.