A computationally optimized broadly reactive hemagglutinin vaccine elicits neutralizing antibodies against influenza B viruses from both lineages

Influenza B viruses (IBV) can cause severe disease and death much like influenza A viruses (IAV), with a disproportionate number of infections in children. Despite moving to a quadrivalent vaccine to include strains from both the B/Victoria and B/Yamagata lineages, vaccine effectiveness rates continue to be variable and low in many past seasons. To develop more effective influenza B virus vaccines, three novel IBV hemagglutinin (HA) vaccines were designed using a computationally optimized broadly reactive antigen (COBRA) methodology. These IBV HA proteins were expressed on the surface of a virus-like particle (VLP) and used to vaccinate ferrets that were pre-immune to historical B/Victoria or B/Yamagata lineage viruses. Ferrets vaccinated with B-COBRA HA vaccines had neutralizing antibodies with high titer HAI titer against all influenza B viruses regardless of pre-immunization history. Conversely, VLPs expressing wild-type IBV HA antigens preferentially boosted titers against viruses from the same lineage and there was little-to-no seroprotective antibodies detected in ferrets with mismatched IBV pre-immune infections. Overall, a single IBV HA developed using the COBRA methodology elicited protective broadly-reactive antibodies against current and future drifted IBVs from both lineages.


A computationally optimized broadly reactive hemagglutinin vaccine elicits neutralizing antibodies against influenza B viruses from both lineages
Michael A. Carlock 1,3 & Ted M. Ross 1,2,3,4,5* Influenza B viruses (IBV) can cause severe disease and death much like influenza A viruses (IAV), with a disproportionate number of infections in children.Despite moving to a quadrivalent vaccine to include strains from both the B/Victoria and B/Yamagata lineages, vaccine effectiveness rates continue to be variable and low in many past seasons.To develop more effective influenza B virus vaccines, three novel IBV hemagglutinin (HA) vaccines were designed using a computationally optimized broadly reactive antigen (COBRA) methodology.These IBV HA proteins were expressed on the surface of a virus-like particle (VLP) and used to vaccinate ferrets that were pre-immune to historical B/Victoria or B/Yamagata lineage viruses.Ferrets vaccinated with B-COBRA HA vaccines had neutralizing antibodies with high titer HAI titer against all influenza B viruses regardless of pre-immunization history.Conversely, VLPs expressing wild-type IBV HA antigens preferentially boosted titers against viruses from the same lineage and there was little-to-no seroprotective antibodies detected in ferrets with mismatched IBV pre-immune infections.Overall, a single IBV HA developed using the COBRA methodology elicited protective broadly-reactive antibodies against current and future drifted IBVs from both lineages.
Influenza viruses cause seasonal epidemics and occasional worldwide pandemics.In addition, severe infection with influenza viruses results in hundreds of thousands of deaths every year 1,2 .Prevention of influenza virus infection by vaccination is the best way to minimize severe cases.Two types of influenza, A and B, co-circulate in the human population.Influenza A viruses (IAV) are subdivided into subtypes based upon the surface hemagglutinin (HA or H) and neuraminidase (NA or N).Currently, H1N1 and H3N2 subtypes have co-circulated in people since 1968.Influenza B viruses (IBV) were first isolated in 1940 and, in some seasons, IBVs are the dominant type isolated from people during flu season, though this can vary regionally and throughout a season as IAV and IBV can peak at different times 1,2 .During the 1988-1989 influenza season, an IBV strain antigenically distinct to the vaccine strain, B/Victoria/2/1987 (B/VIC/87), was detected.And by the next season, half of the isolates matched this new strain, B/Yamagata/16/1988 (B/YAM/88) 3 .As a result, IBV strains were categorized based on whether they were Victoria-like (B/VIC) or Yamagata-like (B/YAM) [4][5][6] .Viruses from the two lineages co-circulated afterwards, but similar to H1N1 and H3N2 IAVs, one lineage would out compete the other each flu season.From 2001 to 2012, the lineage recommended as the vaccine strain component changed six times 7 and the predominant influenza B lineage selected for the vaccine matched the dominant circulating IBV lineage ~ 50% of the time 8 .In 2012, representatives from both IBV lineages were recommended for a quadrivalent influenza vaccine that resulted in increased vaccine effectiveness against influenza B virus infections 9,10 .Despite this, vaccine effectiveness is variable from season to season ranging from 34 to 76% 9 .As a result, our group developed more www.nature.com/scientificreports/broadly-reactive influenza B HA antigens using a layered consensus approach termed computationally optimized broadly reactive antigen (COBRA), as previously described for the generation of influenza A antigens [11][12][13][14][15][16][17][18][19][20][21][22][23] .
Others have demonstrated cross-lineage protection through stalk-directed antibodies.Chimeric HA (cHA) proteins, made with an IBV stalk and the head of exotic IAVs, were used in sequential vaccinations with different HA heads in order to induce stalk-reactive antibodies 24 .Mice were protected from lethal challenge against B/ YAM and B/VIC viruses, but the mechanism of protection was through non-neutralizing antibodies that engage Fc-mediated effector functions.This method was expanded upon through a "mosaic" HA (mHA) approach in which instead of replacing the entire head domain of IBV with one from IAV, site-directed replacements with sequences from exotic HAs were introduced at the major antigenic sites in the head domain 25 .The B mHA vaccinations were able to prevent mortality in mice and reduce morbidity, but like the cHA vaccinations, the afforded protection was through non-neutralizing antibodies.While these platforms certainly have merit, the reliance on non-neutralizing antibodies that do not prevent infection can be a limitation for use in humans, especially in vulnerable populations where simply ameliorating disease may not be enough.And currently, a reliable correlate of protection for non-neutralizing antibodies in humans has yet to be established.
The COBRA IBV (B-COBRA) HA vaccines elicited protective antibodies against viruses from both lineages.Moreover, these B-COBRA HA vaccines were more effective than wild-type HA in a pre-immune animal model.Immunological imprinting can greatly dictate immune responses to future influenza A viruses 26 , but it is unclear how IBV imprinting affects future immune responses to IBV strains.In this report, we explored how the breath of pre-existing immune responses elicited by infection of influenza B viruses could shape subsequent vaccine-induced antibodies.

Cross-reactive antibodies elicited by influenza B viruses
In an effort to better understand how anti-IBV derived antibodies reacted against other influenza B strains, a ferret antibody landscape experiment was conducted.14 influenza B strains representing a time period from 1940 to 2017 were used to infect ferrets and, 30 days later, collected sera was tested for hemagglutination inhibition (HAI) activity against a panel of influenza B viruses (Table 1).Nine different B/YAM viruses that were used to infect ferrets elicited antibodies with high HAI activity against all B/YAM strains in the panel, but little-to-no HAI activity against B/VIC strains.Ferrets infected with B/VIC viruses had antibodies with high HAI activity against all the B/VIC strains in the panel.In addition, ferrets infected with the 3 older viruses (B/HK/01, B/MY/04, and B/BR/08) had seroprotective HAI titers (≥ 1:40) against B/YAM strains.These same 3 B/VIC viruses elicited HAI activity against B/Lee/40, as did the older strains of B/YAM (isolated from 1988 to 1999).Ferret sera collected from B/Lee/40 infected ferrets had a high HAI titer (1:5120) against the B/Lee/40 virus, but low HAI activity against the B/YAM viruses in the panel and little-to-no HAI activity against the B/VIC strains.

Characterization of IBV HA vaccines
Three initial hemagglutinin amino acid sequences were designed and characterized as vaccine candidates.HA sequences from 318 influenza B viruses isolated from 1940 to 2011 were aligned and the most common amino acid at each residue was determined resulting in an HA consensus sequence named BC1 (Fig. 1A).The same 318 IBV HA sequences were aligned using the multi-layered COBRA methodology and termed BC2.Another layered consensus alignment using 217 IBV HA amino acid sequences isolated from 1999 to 2011 was used to generate BC3.Phylogenetically, BC1 falls within the Yamagata-lineage, whereas BC2 and BC3 are aligned within or near the Victoria-lineage (Fig. 1B).For sequence alignments of these along with vaccine strains for IBV from 1940 to 2021, see Supplementary Fig. S1.Immunological memory was established by infecting one set of ferrets with a B/YAM virus (B/SH/02) and a second set with a B/VIC virus (B/HK/01) prior to vaccination (Fig. 1D).These pre-immune ferrets were then vaccinated with virus-like particles (VLP) that were generated via lipotransfection into a mammalian cell line www.nature.com/scientificreports/(Fig. 1C).The VLP vaccines consisted of one of the B-COBRA HA candidates or a wild-type HA from circulating influenza B viruses included in past commercial vaccines (Fig. 1D).Serum samples collected at day 90 (day of vaccination) following pre-immune infections, as well as sera collected at day 120 (~ 4 weeks post-vaccination), were tested for serological activity.Influenza virus infection is initiated by viral HA binding to sialic acid receptors on the surface of host cells, and as such, anti-HA antibodies to block this initial binding is the best course of action in preventing viral infection 27,28 .The HAI assay mimicks this by measuring the highest dilution of anti-HA antibodies from serum that prevents hemagglutination from influenza virus binding to red blood cells 29 .In this study, we utilize the HAI assay to assess the functional activity of ferret antisera against a panel of 5 Yamagata-lineage viruses isolated from 1988 to 2013 and 5 Victoria-lineage viruses isolated from 1987 to 2019.
Antisera from B/YAM (B/SH/02) infected ferrets that were vaccinated with VLP vaccines expressing wildtype B/YAM HA antigens derived from B/FL/06 or B/PH/13 (Fig. 2A,B) had > twofold increase in HAI activity on average against all 5 B/YAM strains following vaccination.These titers ranged between 1:80-1:320 prior to vaccination and 1:320-1:1280 post-vaccination.Little-to-no HAI activity against the B/VIC panel was observed.These same antigens used to vaccinate ferrets that were previously infected with B/VIC (B/HK/01) (Fig. 2C,D) elicited antibodies with high HAI titers against the entire panel.Sera with HAI activity against the B/YAM strains increased > eightfold from a titer of ~ 1:40 prior to vaccination.HAI titers to the B/VIC strains increased two-to fourfold from a pre-vaccination titer of ~ 1:160.Ferrets vaccinated with VLPs expressing the wild-type B/VIC HA from B/MY/04 or B/CO/17 (Fig. 3) had seroprotective HAI titers (≥ 1:40) with ~two-to fourfold increase against most of the viruses in the panel.These vaccinations in B/YAM pre-immune ferrets elicited lower HAI activity against B/VIC viruses (Fig. 3A,B).B/VIC pre-immune ferrets vaccinated with B/MY/04 had seroprotective antibody titers against all viruses in the panel, albeit they were not statistically significant (Fig. 3C).In contrast, B/VIC pre-immune ferrets vaccinated with B/ CO/17 had higher HAI responses (~ fivefold increase) (Fig. 3D).
Regardless of pre-immunization status, ferrets vaccinated with BC2 had a > sixfold increase in HAI activity against all the viruses in the panel.B/YAM pre-immune ferrets (Fig. 4B) had high HAI activity (~ 1:160) prior to vaccination against the B/YAM viruses and little-to-no HAI activity against the B/VIC viruses.Post-vaccination titers were boosted on average > sixfold (~ 1:1280) against the B/YAM viruses and 15-fold (~ 1:160) against the B/VIC viruses.B/VIC pre-immune ferrets (Fig. 4E) had HAI titers (~ 1:40) against the B/YAM viruses prior to vaccination and higher HAI activity (~ 1:320) against the B/VIC viruses.Post-vaccination titers for these ferrets were boosted sevenfold (~ 1:160) against the B/YAM viruses and tenfold (~ 1:2560) against the B/VIC panel of viruses.Similar HAI titers, albeit lower, were detected in BC3-vaccinated ferrets.No significant increases in HAI activity were observed against the B/YAM viruses in the B/YAM pre-immune ferrets, although all HAI titers exceeded seroprotective levels (Fig. 4C).The B/VIC pre-immune ferrets vaccinated with BC3 only had three ferrets in the group, thus reducing the significance scoring (a fourth point matching the others would have a p-value ≤ 0.01 against all viruses) (Fig. 4F).Sera collected from BC1-vaccinated ferrets had similar, but generally higher HAI titers, compared to sera collected from ferrets vaccinated with wild-type B/YAM HA.B/YAM pre-immune ferrets vaccinated with VLPs expressing BC1 HA (Fig. 4A) had on average a > fivefold increase in HAI activity against the B/YAM strains following vaccination with HAI titers ranging ~ 1:160 prior to vaccination and 1:1280-1:2560 post-vaccination.Sera collected from these same ferrets had little-to-no HAI activity against the B/VIC viruses.The same antigens used in B/VIC pre-immune ferrets (Fig. 4D) elicited antibodies with high HAI titers against the entire panel of B/YAM strains with an increase of > 20-fold from a titer of ~ 1:40 prior to vaccination.These same ferrets had a five to eightfold increase in HAI titers against the B/VIC strains from a titer of ~ 1:160 prior to vaccination.

Neutralization of influenza B viruses
While the HAI assay is still regarded as the "gold standard" for determining vaccine efficacy, microneutralization (MN) assays are often preferred amongst researchers since they are generally less subjective and can also capture other neutralizing antibodies that HAI does not, such as anti-stalk or NA-inhibiting antibodies 30 .Because the majority of neutralizing antibodies are anti-HA, however, MN and HAI generally correlate fairly well 31,32 , particularly when plaque reduction neutralization (PRNT) or focus reduction assay (FRA) are used which include ELISA as a readout 33,34 .In this study, we use FRA to evaluate the ability of serum antibodies to neutralize live virus infection for B/YAM (B/SH/02) pre-immunized ferrets (Fig. 5) and B/VIC (B/HK/01) pre-immunized ferrets (Fig. 6).
Ferrets infected with B/SH/02 virus had neutralizing antibodies against the two B/YAM viruses tested (Fig. 5A,B) with an FRA 50 log 2 titer > 6.8, but had no neutralizing antibodies against the two B/VIC viruses (Fig. 5C,D).VLP vaccinations boosted antibodies against the B/YAM viruses (Fig. 5E,F) to an FRA 50 log 2 titer ~ 10.0 for most vaccination groups.Ferrets vaccinated with either of the three B-COBRA HA candidates or the two older wild-type HA proteins (B/MY/04 and B/FL/06) had antisera that neutralized B/BR/08 infection (Fig. 5G).Only ferrets vaccinated with BC2, BC3, or B/MY/04 HA-expressing VLP vaccines were able to elicit neutralizing antibodies against B/CO/17 (Fig. 5H).For both B/VIC viruses, the highest vaccine-induced fold-change occurred in ferrets vaccinated with BC2 or B/MY/04 VLPs.For the B/YAM viruses, the highest vaccine-induced fold-change occurred in ferrets vaccinated with BC1 and BC2.Ferrets pre-immunized with B/VIC (B/HK/01) had high neutralizing antibody titers against B/VIC viruses prior to vaccination (Fig. 6C,D) as well as cross-lineage neutralizing antibodies to B/MA/12 (Fig. 6A), but not to B/PH/13 (Fig. 6B).Ferrets vaccinated with BC1 HA VLPs or the B/YAM wild-type HA VLPs had the highest vaccine-induced increase in antibodies following the boost against the B/YAM viruses, with an FRA 50 log 2 titer of 7.83-10.26to B/MA/12 (Fig. 6E) and 7.69-8.18to B/PH/13 (Fig. 6F).Ferrets vaccinated with BC2 or BC3 HA VLPs had an FRA 50 log 2 titer of 7.39-7.86against B/MA/12 and 5.74-6.77against B/PH/13, whereas the B/VIC wild-type VLP vaccinated ferrets had titers of 5.82-6.85against B/MA/12 and 5.66-5.97against B/PH/13.All vaccinated ferrets had high post-vaccination titers (FRA 50 log 2 titer > 10) against the B/VIC viruses (Fig. 6G,H), with the highest titers recorded in ferrets vaccinated with BC2 or BC3 VLP vaccines.

Heterologous challenge of pre-immune ferrets
All ferrets were challenged 30 days post-vaccination with a contemporary IBV from the B/VIC lineage (B/CO/17) or B/YAM lineage (B/PH/13) that were heterologous to the pre-immune infection virus (Fig. 7).Neither virus caused severe disease or death.The B/YAM pre-immune ferrets, as well as mock vaccinated ferrets, lost ~ 5% of their original body weight by day 5 following B/CO/17 challenge (Fig. 7A), which was slightly less than naïve, non-vaccinated ferrets (~ 7% average).All vaccinated ferrets had little weight loss and no signs of influenza virus infection over the time of observation.Comparable results were observed in B/VIC pre-immune ferrets that  www.nature.com/scientificreports/were challenged with B/PH/13 virus (Fig. 7B).Overall, all ferrets were protected against the influenza B virus challenge.Ferrets vaccinated with BC2 and BC3 HA antigens had less weight loss, on average, between the two different challenges and even gained weight over the period of observation (Fig. 7B).Little-to-no detectable viral titers were observed in nasal washes collected three days post-infection (data not shown).

Discussion
Immunological imprinting or original antigenic sin following initial exposure to influenza has been well documented since it was first described in 1960 26 .This concept, however, has been less well described for influenza B virus infection, especially in regard to cross-lineage boosting or negative immunological interference.Using our pre-immune IBV ferret model, we further elucidate the impact of IBV induced pre-immunity on vaccine-induced antibody responses.Pre-immunization with a B/VIC virus led to an induction of low levels of cross-reactive antibodies and enabled B/YAM HA vaccinated ferrets to generate seroprotective antisera against B/VIC viruses.The same cross-reactivity was less apparent for B/VIC HA vaccinated ferrets pre-immunized to B/YAM, but some ferrets still generated seroprotective antibody levels.In homologously matched pre-immunization and vaccination regimens, ferrets receiving a B/YAM HA vaccination mounted a low response to B/VIC viruses.Ferrets vaccinated with the B/YAM-like BC1 HA vaccine elicited slightly higher titers than the wild-type B/YAM HA vaccinated ferrets, though the increases were not determined to be statistically significant.In contrast, almost all B/VIC pre-immune ferrets that were vaccinated with B/VIC HA had seroprotective titers to B/YAM viruses.B-COBRA HA vaccinations however elicited more broadly-reactive antibodies in these B/VIC pre-immunized ferrets.Following vaccination, all ferrets had seroprotective antibody titers against all IBVs in the panel and BC2 and BC3 vaccines efficiently elicited high titer antibodies in B/YAM pre-immune ferrets.All BC2 HA vaccinated ferrets had statistically significant increases in HAI titers against the full panel of IBVs and only one ferret in the BC3 group failed to mount a seroprotective HAI response to the B/VIC panel.Ferrets vaccinated with VLPs expressing wild-type vaccinations preferentially boosted antibody titers against lineage-matched viruses and generally did not have seroprotective titers, particularly for ferrets vaccinated with antigens heterologous to the pre-immunity.Both BC2 and BC3 HA vaccines elicited antibodies with broader HAI activity against a panel of IBVs than wild-type HA vaccine comparators.Determining how initial exposure to one strain of influenza B virus impacts subsequent responses to future strains may be a key factor to consider when testing next-generation influenza vaccines, especially in children who may only have been exposed to an IBV strain from one lineage prior to vaccination.Other studies for example have shown that children vaccinated with a trivalent influenza vaccine (TIV) containing a B/YAM antigen elicited cross-reactive antibody responses against B/VIC viruses if they were previously immunized to B/VIC, but not if they were unexposed 35 .Data on B/VIC vaccinations in children with no pre-existing immunity to IBV is scarce, however children vaccinated with B/VIC based vaccines were shown to have low immune responses to B/VIC antigens when these children were primed with B/YAM the previous season 36 .Children immunized with vaccines containing B/VIC the following year had high anti-HA antibody titers to B/VIC, as well as cross-lineage back-boosting to the B/YAM HA antigens 36 .

Figure 1 .
Figure 1.B-COBRA HA antigens for the vaccination and infection of pre-immune ferrets.(A) Publicly available HA sequences from 1940 to 2011 were downloaded and used to design three unique influenza B HA antigens: BC1, a single consensus sequence of 318 sequences during this period; BC2, the same 318 sequences antigenically layered into 10 secondary sequences and then a final consensus sequence; BC3, 217 sequences from 1999 to 2011 layered into 5 secondary sequences and then a final consensus sequence.(B) Phylogenetic analysis of B-COBRA candidates along with current and past IBV vaccine strains.HA1 regions were extracted and aligned using Muscle 3.8.425.The tree was rendered using FastTree 2.1.11.B/YAM lineage shown with yellow font, B/Victoria-lineage with violet font, and B-COBRA candidates in red font.(C) Three plasmids containing genes for HA, NA, or a Gag viral matrix protein were lipotransfected into mammalian 293T cells.After 3-5 days, supernatants were collected, centrifuged, filter-sterilized, and then ultracentrifuged with a 20% glycerol cushion.Pellets were resuspended in a small volume of PBS, and then HA activity and protein concentration were determined.(D) Immunologically naïve ferrets were infected intranasally (day 0) with a B/ YAM (SH/02) or B/VIC (HK/01) virus and then vaccinated intramuscularly once (day 90) with one of three B-COBRA VLP vaccines (BC1, BC2, BC3), one of four VLP vaccines expressing wild-type HA proteins from the B/VIC lineage (MY/04 or CO/17) or B/YAM lineage (FL/06 or PH/13), or a mock vaccination containing PBS and adjuvant.All vaccines were adjuvanted 1:1 with a squalene-based emulsion adjuvant.Blood was collected from all animals at days 0, 90, and 120.All ferrets were challenged with an IBV heterologous to their preimmune infection and nasal washes were collected 3 days post-infection.Created with BioRender.com.

Figure 5 .
Figure 5. FRA neutralizing titers for Yamagata pre-immune ferrets vaccinated a single time with VLPs expressing different HA proteins: BC1, BC2, BC3, MY/04, FL/06, PH/13, CO/17, or Mock.Serum was collected at day 90 pre-vaccination (A-D) and day 120 post-vaccination (E-H) and tested against four IBVs: two from the Yamagata-lineage (A,B,E,F) and two from the Victoria-lineage (C,D,G,H).The dotted lines in the line graphs represent 50% inhibition and 80% inhibition of viral infection by antisera compared to virus-only control wells.

Figure 6 .
Figure 6.FRA neutralizing titers for Victoria pre-immune ferrets vaccinated a single time with VLPs expressing different HA proteins: BC1, BC2, BC3, MY/04, FL/06, PH/13, CO/17, or Mock.Serum was collected at day 90 pre-vaccination (A-D) and day 120 post-vaccination (E-H) and tested against four IBVs: two from the Yamagata-lineage (A,B,E,F) and two from the Victoria-lineage (C,D,G,H).The dotted lines in the line graphs represent 50% inhibition and 80% inhibition of viral infection by antisera compared to virus-only control wells.

Figure 7 .
Figure 7. Average percent original weight of ferrets (n = 4 per group) following challenge with B/ Colorado/06/2017-VIC (left) or B/Phuket/3073/2013-YAM (right).With exception to the naïve (non preimmune and non-vaccinated) group of ferrets shown for comparison, all ferrets were previously infected with IBVs from lineages heterologous to the challenge strain (B/Shanghai/361/2002-YAM or B/Hong Kong/330/2001-VIC).Following these infections to establish pre-immunity, ferrets were allowed to sit for 90 days and then vaccinated a single time with VLPs expressing different HA proteins.Challenges were performed 30 days post-vaccination.Ferrets were monitored for weights and clinical symptoms daily, and nasal washes were collected 3 days post-infection.

Table 1 .
14 groups of ferrets (n = 2/group) were infected with different viruses spanning the IBV history in order to generate ferret reference sera.The earliest IBV (B/Lee/1940) highlighted in blue, 9 viruses from the B/ YAM lineage (1988-2013) highlighted in yellow, and 4 viruses from the B/VIC lineage (2001-2017) highlighted in purple.Serum was collected from each ferret, pooled per infection group, and then tested via HAI assay against 10 of the viruses (shown at the top).