Development of high-yield influenza B virusvaccine virusesJihui Pinga, Tiago J. S. Lopesa,b, Gabriele Neumanna, and Yoshihiro Kawaokaa,b,c,1

aDepartment of Pathobiological Sciences, Influenza Research Institute, School of Veterinary Medicine, University of Wisconsin–Madison, Madison, WI 53711;bDivision of Virology, Department of Microbiology and Immunology, University of Tokyo, Tokyo 108-8639, Japan; and cInternational Research Center forInfectious Diseases, Institute of Medical Science, University of Tokyo, Tokyo 108-8639, Japan

Edited by Michael B. A. Oldstone, The Scripps Research Institute, La Jolla, CA, and approved November 3, 2016 (received for review July 8, 2016)

The burden of human infections with influenza A and B viruses issubstantial, and the impact of influenza B virus infections can exceedthat of influenza A virus infections in some seasons. Over the pastfew decades, viruses of two influenza B virus lineages (Victoria andYamagata) have circulated in humans, and both lineages are nowrepresented in influenza vaccines, as recommended by the WorldHealth Organization. Influenza B virus vaccines for humans havebeen available for more than half a century, yet no systematic effortshave been undertaken to develop high-yield candidates. Therefore,we screened virus libraries possessing random mutations in the six“internal” influenza B viral RNA segments [i.e., those not encodingthe major viral antigens, hemagglutinin (HA) and neuraminidaseNA)] for mutants that confer efficient replication. Candidate virusesthat supported high yield in cell culture were tested with the HA andNA genes of eight different viruses of the Victoria and Yamagatalineages. We identified combinations of mutations that increasedthe titers of candidate vaccine viruses in mammalian cells used forhuman influenza vaccine virus propagation and in embryonatedchicken eggs, the most common propagation system for influenzaviruses. These influenza B virus vaccine backbones can be used forimproved vaccine virus production.

influenza B virus | Yamagata lineage | Victoria lineage | vaccine |high-yield

On the basis of their antigenic properties, influenza virusesare divided into three types (influenza A, B, and C); however,

only type A and B influenza viruses cause human health concerns.Influenza A viruses are further divided into 18 hemagglutinin (HA,the major viral antigen) and 11 neuraminidase (NA, the secondviral antigen) subtypes that are referred to as H1–H18 and N1–N11, respectively. Influenza A viruses are responsible for annualepidemics (caused by antigenic escape variants possessing pointmutations in the antigenic epitopes of HA) and occasional pan-demics. Pandemics are caused by avian or avian/human/swinereassortant influenza viruses that encode an HA protein to whichhumans lack protective immune responses. The epidemiology ofinfluenza B viruses differs from that of influenza A viruses. In-fluenza B viruses primarily circulate in humans and do not causepandemics. Nevertheless, the impact of influenza B virus infec-tions on influenza-related morbidity and mortality is substan-tial and has exceeded that of influenza A viruses in some seasons(1–3). Until 1983, only one influenza B virus lineage was circu-lating in humans. Since then, two lineages (Victoria, named afterB/Victoria/2/1987, and Yamagata, named after B/Yamagata/16/1988) can be distinguished genetically and antigenically on thebasis of their HA (1, 3, 4). Until 2000, one of these two lineagestended to dominate each season; however, since 2001 both in-fluenza B virus lineages have been cocirculating in human pop-ulations each year (5, 6).Until recently, most influenza vaccines were trivalent: that is,

they were comprised of influenza A strains of the H1N1 andH3N2 subtypes, and an influenza B virus strain. Two studiesdemonstrated that the recommended influenza B vaccine strainmatched the dominant strain of the particular influenza season

only half the time (5, 7). On the basis of these findings and thecontinuing cocirculation of Yamagata- and Victoria-lineage viru-ses, in 2012 the World Health Organization (WHO) recommendedincluding influenza B viruses of both lineages in human influenzavaccines. Accordingly, most seasonal influenza vaccines are nowquadrivalent.Many influenza vaccines are generated by combining the HA and

NA viral RNA (vRNA) segments of WHO-recommended vaccinestrains with the remaining six vRNA segments of a “backbone” strain.For live attenuated influenza A and B vaccine viruses, virus back-bones were developed in the 1960s (8). Many inactivated influenza Avirus vaccines are based on the A/Puerto Rico/8/34 (H1N1; PR8)virus backbone, which was selected because of its efficient replicationin embryonated chicken eggs. For inactivated influenza B vaccines,the B/Lee/40, B/Panama/45/90, or wild-type strains have been usedas backbones (www.who.int/influenza/vaccines/virus/recommendations/summary_b_vic_cvv_nh1516.pdf).Recently, we developed a high-yield vaccine backbone of PR8,

which confers increased yield to influenza A candidate vaccineviruses in embryonated chicken eggs, and Madin–Darby caninekidney (MDCK) and African green monkey (Vero) cells (i.e., thethree propagation systems currently used to grow human in-fluenza vaccine viruses) (9). Briefly, we screened virus librariespossessing random mutations in the six “backbone genes” of PR8(i.e., PB2, PB1, PA, NP, M, and NS), systematically tested po-tentially growth-enhancing mutations described in the literature,and assessed PR8 mutants with mutations in the promoter regionsof their vRNAs. Combinations of growth-enhancing mutationswere then tested with the HA and NA vRNAs of influenza A

Significance

The yield of vaccine viruses is important from an economicpoint of view. Even more important, the ability to produce highnumbers of vaccine doses under tight timelines may save manylives during a virus outbreak. Applying an approach that werecently used to develop high-yield influenza A virus vaccinecandidates, we now developed high-yield vaccine candidatesfor both influenza B virus lineages circulating in humans. Thesevaccine virus candidates confer higher yield in commonly usedpropagation systems for influenza vaccine virus production:that is, embryonated chicken eggs, Madin–Darby canine kidneycells, and African green monkey (Vero) cells. Our vaccine can-didates could be used to improve the influenza B virus vaccineproduction process.

Author contributions: J.P., G.N., and Y.K. designed research; J.P. performed research; J.P.,T.J.S.L., G.N., and Y.K. analyzed data; and J.P. and G.N. wrote the paper.

Conflict of interest statement: Y.K. has received speaker’s honoraria from Toyama Chem-ical and Astellas Inc.; grant support from Chugai Pharmaceuticals, Daiichi Sankyo Phar-maceutical, Toyama Chemical, Tauns Laboratories, Inc., and Otsuka Pharmaceutical Co.,Ltd.; and is a founder of FluGen. G.N. is a founder of FluGen.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: kawaokay@svm.vetmed.wisc.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1616530113/-/DCSupplemental.

E8296–E8305 | PNAS | Published online December 5, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1616530113

viruses of different subtypes, resulting in a high-yield PR8 virusbackbone. Here, we used a similar strategy to develop a high-yieldinfluenza B virus backbone for increased vaccine virus yield.

ResultsVirus Library Screens for High-Yield Variants in MDCK Cells. Com-parable to the strategy we used to develop a high-yield influenzaA virus PR8 vaccine backbone, we carried out mutagenesis andscreening approaches to identify mutations associated with highyield of an influenza B virus. We derived the six internal vRNAsegments from the B/Yamagata/1/73 virus, which grows effi-ciently in MDCK cells and was isolated before the Victoria andYamagata lineages separated. A mutagenesis approach based onerror-prone PCR was then used to generate libraries of cDNAspossessing one to four random amino acid changes in the viralproteins (SI Appendix, Fig. S1). These cDNA libraries were thenused to generate virus libraries. We generated six separate li-braries representing each of the internal vRNAs (i.e., PB2, PB1,PA, NP, M, and NS) (SI Appendix, Fig. S1); three libraries forcombinations of the polymerase vRNAs (i.e., PB2+PB1, PB2+PA,PB2+PB1+PA); one library for the polymerase and nucleoprotein(NP) vRNAs (i.e., PB2+PB1+PA+NP) because the PB2, PB1,PA, and NP proteins form the viral replication complex; one li-brary for the PB2 and NS vRNAs (PB2+NS) because the PB2 andNS1 proteins (encoded by the NS vRNA) of influenza A virusesare important determinants of host virulence (10); and one libraryfor the M and NS vRNAs because the M1 protein (encoded by theM vRNA) of influenza A viruses is associated with high-growthproperties (11). Each of these 12 virus libraries was generated withthe HA and NA vRNAs of a representative virus of the Victoria(B/Yokohama/UT-K1A/2011) or Yamagata (B/Yokohama/UT-K31/2012) lineage, respectively, resulting in a total of 24 virus li-braries. Libraries generated with the HA and NA vRNAs of theVictoria-lineage or Yamagata-lineage viruses will be referred to as“Victoria-lineage” or “Yamagata-lineage” libraries, respectively.To select variants with enhanced growth properties, we pas-

saged each library 12 times in MDCK cells. In parallel, we com-bined virus libraries after two passages in MDCK cells, and thenperformed 10 additional passages in MDCK cells (SI Appendix,Fig. S1). We randomly selected more than 700 viral plaques eachfrom the Victoria- and Yamagata-lineage libraries, respectively,resulting in a total of 1,472 individual, plaque-purified viruses (SIAppendix, Fig. S1). The plaque-purified viruses were then ampli-fied in MDCK cells, and their yields were assessed in hemagglu-tination assays (as a surrogate for high HA yield) and comparedwith those of the parental Victoria- and Yamagata-lineage viruses(which possess the Victoria- or Yamagata-lineage lineage HA andNA vRNAs in combination with the six remaining vRNAs ofB/Yamagata/1/73 virus). We identified 29 Yamagata- and 28 Victoria-lineage viruses with hemagglutination titers that were at leasttwofold higher than those of the respective control virus. Thesecandidate viruses were reamplified in MDCK cells, and their high-yield properties were confirmed by assessing hemagglutinationtiters and replication kinetics in MDCK cells (used as anothersurrogate for high HA yield) (SI Appendix, Fig. S2).Next, we sequenced the entire viral genomes of the top eight

candidates of each lineage and found different sets of mutationsfor high-yield candidates of the Yamagata- and Victoria-lineages(we evaluated amino acid changes and changes in the noncodingregions) (SI Appendix, Tables S1 and S2). Seven of the eighthigh-yield candidates isolated from the Yamagata-lineage li-braries possessed G34V and I97N mutations in the M1 matrixprotein, and H58R and R80G mutations in the BM2 ion channelprotein (also encoded by the M gene) (SI Appendix, Table S1),suggesting that these amino acid substitutions may confer effi-cient replication in MDCK cells.All eight high-yield candidates obtained from the Victoria-

lineage libraries encoded a P40S mutation in NP and an M86T

mutation in M1 (SI Appendix, Table S2). In addition, six of theseeight high-yield candidates encoded nucleotide changes in thenoncoding region of the NS vRNA: an additional nucleotideafter position 38 was detected in five viruses [NS-38(+1)g; allnucleotide changes in noncoding regions are shown in italicizedlowercase letters], and an a39g nucleotide replacement was de-tected in one virus (SI Appendix, Table S2). We also identified ag1795a mutation in the noncoding region of the NP segment inthree high-yield candidates (SI Appendix, Table S2).Although the HA and NA vRNAs were not targeted by PCR-

mediated random mutagenesis, we detected several mutations inthe HA and NA proteins (SI Appendix, Tables S1 and S2). Spe-cifically, threonine at position 196 of HA was replaced with variousother amino acids (i.e., alanine, isoleucine, proline, or asparagine)in seven of eight high-yield candidates derived from the Victoria-lineage libraries (SI Appendix, Table S2), suggesting strong selec-tive pressure at this position.

Potential Combinatorial Effects of Mutations. Next, we used reverse-genetics approaches to generate viruses possessing various combi-nations of the mutations found in the top eight high-yield Yamagata-or Victoria-lineage candidates (SI Appendix, Tables S3 and S4). Forexample, the NP-E52K andM1-R77K mutations found in high-yieldcandidate #21 (which replicated to the highest titers inMDCK cells)(SI Appendix, Fig. S2) were combined with the NS1-M117Y/S252Tmutations found in high-yield candidates #23 and #26 (SI Appendix,Fig. S2). Because the internal vRNAs of the Yamagata- andVictoria-lineage libraries are derived from the same virus, we alsocombined mutations found in high-yield Yamagata- and Victoria-lineage candidates (SI Appendix, Tables S3 and S4). The resultingviruses were tested for their hemagglutination titers and viral titersin MDCK cells (SI Appendix, Fig. S3). All high-yield Yamagata-and Victoria-lineage candidates replicated in MDCK cells moreefficiently and had higher hemagglutination titers than the wild-type viruses at one or more time points, and most of these dif-ferences were statistically significant (SI Appendix, Fig. S3).Yamagata-lineage RG(Yam) #8 (encoding NP-P40S, M1-R77K,NS1-K176Q, and NS-a39g mutations) and Victoria-lineageRG(Vic) #2 (encoding NP-P40S/M204T, M1-M86T, and NS-38(+1)g mutations) were selected as lead candidate vaccine backbonesbecause they had the highest titers in their respective groups.

Virus Library Screens for High-Yield Variants in Vero Cells. The idealvaccine virus backbone should confer a high yield in all three prop-agation systems currently used in the commercial production of hu-man influenza vaccines: that is, MDCK cells, Vero cells, andembryonated chicken eggs. In parallel to the development of a high-yield influenza B vaccine backbone in MDCK cells, we therefore alsoplanned to passage all 24 virus libraries (SI Appendix, Fig. S1) in Verocells. Because the titers of influenza B virus libraries were low in Verocells, we first passaged them in coculturedMDCK and Vero cells fivetimes, followed by five passages in Vero cells. In parallel, virus li-braries were combined after two passages in cocultured MDCK andVero cells, passaged three more times in cocultured cells, and thenpassaged five times in Vero cells. A total of 382 individual virusplaques were randomly picked from the various virus libraries andamplified in Vero cells. Based on the results of hemagglutinationassays, we picked the top six candidates from the Yamagata andVictoria lineage, Vero cell-passaged libraries and determined theirfull genomic sequences (SI Appendix, Tables S5 and S6).The high-yield candidates possessed several of the mutations

that were identified after the MDCK cell passages, which mayhave been selected in the MDCK cells during the passages incocultured cells. These mutations include amino acid changes atpositions M1-34/97, BM2-58 or -80, and HA1-196 (compare SIAppendix, Tables S1 and S2 with SI Appendix, Tables S5 and S6).The mutations at positions M1-34/97 and BM2-58 were only de-tected in the Yamagata-lineages libraries, whereas mutations at

Ping et al. PNAS | Published online December 5, 2016 | E8297

MICRO

BIOLO

GY

PNASPL

US

position BM2-80 occurred in both the Yamagata- and Victoria-lineage viruses. The mutation at position HA1-196 predominatedamong viruses of the Victoria lineage, but also occurred in onevirus of the Yamagata lineage. In addition, we also detectedmutations not previously found after passages in MDCK cells,most notably an a2272t nucleotide replacement in the noncodingregion of PA, which was found in high-yield candidates of bothvirus lineages (SI Appendix, Tables S5 and S6).

Selection of High-Yield Yamagata- and Victoria-Lineage Vaccine VirusCandidates. We next tested whether the PA-a2272t mutationfound after the Vero cell passages would enhance the growthproperties of RG(Yam) #8 and/or RG(Vic) #2. The resultingviruses [HY(Yam) and HY(Vic), respectively] displayed higherhemagglutination and virus titers in MDCK cells comparedwith RG(Yam) #8 and RG(Vic) #2, and compared with theparental Yamagata- and Victoria-lineage viruses (Fig. 1); someof these differences were small (although statistically significant).Therefore, HY(Yam) (encoding NP-P40S, M1-R77K, NS1-K176Q,

NS-a39g, PA-a2272t) and HY(Vic) [encoding NP-P40S/M204T,M1-M86T, NS-(38+1)g, PA-a2272t] were selected as lead high-yield candidates.

Evaluation of High-Yield Vaccine Virus Backbones with DifferentInfluenza B Virus HA and NA Genes. HY(Yam) and HY(Vic) weredeveloped with the HA and NA genes of B/Yamagata/UT-K31/2012 and B/Yokohama/UT-K1A/2011, respectively. Becausehigh-yield vaccine virus backbones should have a general growth-enhancing effect, we tested the HY(Yam) and HY(Vic) vaccinevirus backbones with the HA and NA genes of six influenza Bviruses isolated over several decades, including WHO-recom-mended vaccine viruses (Fig. 2). At one or more time pointstested, the HY(Yam) and HY(Vic) vaccine virus backbonesconferred higher viral or hemagglutination titers compared withthe parental viruses, although not all of the differences werestatistically significant. For the viruses tested, HY(Vic) had agreater growth-enhancing effect than HY(Yam).

Exchange of HY(Yam) and HY(Vic) Backbones.Next, we tested whetherthe HY(Yam) and HY(Vic) backbones supported efficient repli-cation of viruses possessing HA and NA genes derived from theother influenza B virus lineage (Fig. 3). High viral and hemag-glutination titers were detected for HY(Yam) viruses encodingVictoria-lineage HA and NA genes, and for HY(Vic) virusesencoding Yamagata-lineage HA and NA genes. Overall, theHY(Yam) vaccine backbone resulted in slightly higher virusor hemagglutination titers than the HY(Vic) vaccine back-bone at several time points, but most of these differenceswere not statistically significant. These data indicate that theHY(Yam) and HY(Vic) vaccine backbones confer efficient rep-lication to viruses of both lineages.

Comparison of Influenza A and B Virus Vaccine Backbones. Chimericinfluenza A viruses possessing the HA and NA genes of an in-fluenza B virus have been generated previously by us (12) andothers (13). Here, we therefore tested our recently developedinfluenza A high-yield vaccine candidate (9) with influenza Bvirus HA and NA genes. The use of a universal backbone forboth influenza A and B viruses would simplify the vaccine pro-duction process. Reassortants between influenza A and B virusesdo not occur naturally, likely because of type-specific viral pack-aging signals located in the 5′ and 3′ terminal regions of influ-enza vRNA segments (14, 15). We, therefore, generated vRNAsegments in which the ectodomains of the influenza A PR8 virusHA and NA proteins were replaced with influenza B viruscounterparts (SI Appendix, Fig. S4). We, then, compared thehemagglutination and viral titers of the following three viruses: awild-type Yamagata-lineage influenza B virus; HY(Yam) with theHA and NA genes of a Yamagata-lineage virus; and high-yieldPR8 virus with type A/B chimeric HA and NA genes of a Yama-gata-lineage virus (Fig. 4 A and B). Similar experiments werecarried out for viruses of the Victoria-lineage (Fig. 4 C and D). Atmost time points tested, the high-yield influenza A or B vaccinevirus backbones conferred significantly increased hemagglutinationand viral titers compared with wild-type viruses. Comparison of theinfluenza A and B vaccine backbones revealed higher virus titerswith the influenza B vaccine backbone but, interestingly, higherhemagglutination titers with the influenza A vaccine backbone.Moreover, for two influenza B viruses representing both lineages,the influenza A vaccine backbone was superior to the influenza Bvaccine backbones with respect to hemagglutination and viral ti-ters in embryonated chicken eggs (Fig. 4 E and F). Other influ-enza B viruses containing the HA and NA from B/Yokohama/UT-K1A/2011, B/Yokohama/P-2922/2005, B/Tokyo/UTE2/2008,or B/Tochigi/UT-T1/2014 did not grow well in embryonatedchicken eggs regardless of the backbone (wild-type or high-yield),

A

B

Fig. 1. Growth kinetics and hemagglutination titers of high-yield Yamagata-and Victoria-lineage viruses. (A) We compared the indicated Yamagata-lineagewild-type virus with the Yamagata-lineage high-yield candidate RG(Yam) #8 (SIAppendix, Table S3) and with RG(Yam) #8 possessing the PA-a2272t mutation;the latter virus was selected as lead candidate HY(Yam). (B) We compared theindicated Victoria-lineage wild-type virus with the Victoria-lineage high-yieldcandidate RG(Vic) #2 (SI Appendix, Table S4) and with RG(Vic) #2 possessing thePA-a2272t mutation; the latter virus was selected as lead candidate HY(Vic). Inboth sets of experiments, MDCK cells were infected in triplicate with the in-dicated viruses at anMOI of 0.001 and incubated at 35 °C. At the indicated timepoints, virus and hemagglutination titers were determined by performingplaque or hemagglutination assays, respectively. The values presented are theaverage of three independent experiments ± SD. P values were calculated byusing the linear mixed model described in Materials and Methods (*P < 0.05;**P < 0.01). Red and blue asterisks indicate the comparison of the respectivevirus withWT virus; beige asterisks indicate the comparison between the virusesdepicted in red and blue. HA titer, hemagglutination titer.

E8298 | www.pnas.org/cgi/doi/10.1073/pnas.1616530113 Ping et al.

suggesting that the HA and NA genes of these human viruses mayrestrict efficient growth in embryonated chicken eggs.

Evaluation of Total Viral Protein Yield and HA Content. Most prep-arations of inactivated influenza vaccines contain 15 μg each ofH1 HA, H3 HA, and type B HA proteins. For vaccine optimi-zation, total viral protein yield and HA content are thereforeimportant parameters, prompting us to compare the total viralprotein yield and HA content of different HY(Yam) andHY(Vic) viruses with their respective wild-type viruses in MDCKcells. Cell-culture supernatants were collected from infected cells,and viruses were concentrated and purified by use of sucrosegradient centrifugation. The total viral protein yield was thendetermined by using the Pierce BCA Protein Assay Kit (ThermoScientific). In parallel, we treated purified virus samples withPNGase F, resulting in HA deglycosylation, which allows easierdetection of HA2 (which in its glycosylated form is similar in sizeto M1). Samples were separated by using SDS/PAGE (Fig. 5 Aand B) and the amounts of HA1, HA2, NP, and M1 were de-termined based on densitometric analysis. The HA content wascalculated by dividing the HA amount (calculated by summing theamounts of HA1 and HA2) by the sum of the amounts of HA1,HA2, NP, and M1, and multiplying this value by the amount oftotal viral protein in the samples analyzed via gel electrophoresis(Fig. 5 A and B). For all viruses tested, total viral protein yield andHA content were significantly higher with the HY(Yam) andHY(Vic) vaccine backbones compared with the wild-type virusesfrom which the HA and NA vRNAs were derived.

Virulence of PR8-HY–Based Vaccine Viruses in Mice. Our experimentalapproach was designed to select mutants with increased replicativeability, which may also increase their virulence in mammals. Toaddress this question, we infected five mice per group with 106 pfuof a wild-type Yamagata-lineage virus, a virus possessing the HAand NA vRNAs of the Yamagata-lineage virus in combinationwith the remaining six genes of wild-type B/Yamagata/1/73 virus(which was used to generate the virus libraries), and a virus possessingthe HA and NA vRNAs of the Yamagata-lineage virus in combina-tion with the HY(Yam) vaccine virus backbone (Fig. 6 A and B). Inparallel, groups of 10 mice were infected with 106 pfu of theviruses described above, and five animals each were killed on days

3 and 6 postinfection to assess lung virus titers (Fig. 6C). Similarly,experiments were carried out with a Victoria-lineage virusesand the HY(Vic) vaccine backbone (Fig. 6 D–F). The wild-typeB/Yamagata/1/73 virus backbone conferred higher pathogenicitythan the B/Massachusetts/2/2012 and B/Brisbane/60/2008backbones, respectively. However, the yield-enhancing mutationsof HY(Yam) and HY(Vic) did not increase mouse virulencefurther; in fact, they had slightly attenuating effects compared withthe B/Yamagata/1/73 backbone.

Genetic Stability of the HY(Yam) and HY(Vic) Vaccine Backbones.Vaccine viruses should be genetically stable so that their desiredproperties are maintained. To test the genetic stabilities ofour high-yield candidates, we performed 10 serial passages of aHY(Yam) virus with the HA and NA vRNAs of B/Yokohama/UT-K31/2012, and of a HY(Vic) virus with the HA and NA vRNAs ofB/Yokohama/UT-K1A/2011 in MDCK cells. After each passage,the genomic sequences of the viruses were determined by Sangersequencing. For the Yamagata-lineage virus, no mutations weredetected. For the Victoria-lineage virus, we did not detect muta-tions in the internal genes that define the high-yield properties ofHY(Vic). However, we did detect a mixed population encodingHA1-196T and -196I after passage 5; after passage 10, only theHA1-196I mutant was detected.Similarly, we passaged several wild-type and high-yield in-

fluenza B viruses of both lineages in embryonated chicken eggs(SI Appendix, Table S7). After 5–10 consecutive passages, no egg-adapting mutations were detected in the internal genes. However,for viruses that possessed a glycosylation site at amino acids 194–196 of HA, mutations arose that resulted in the loss of that gly-cosylation site. This finding is consistent with our earlier finding ofmutations at this glycosylation site in viruses of the Victoria-lineage(SI Appendix, Table S2). Collectively, our data indicate that theyield-enhancing mutations in HY(Yam) and HY(Vic) were ge-netically stable for at least 5–10 consecutive passages in MDCKcells and embryonated chicken eggs.

Contribution of Individual vRNAs to High-Yield Properties of HY(Yam)and HY(Vic). The HY(Yam) and HY(Vic) vaccine backbones possessmutations in several vRNA. To better understand the contributionsof these mutations to the HY(Yam) and HY(Vic) phenotypes, we

98765432

1011

9876543210

23456789

10

23456789

10

012345678

- /02/2012 (HA/NA)B/Yamagata/16/1988 (HA/NA)

9876543210

**

***

**** **

**

**

** ** **

*

**

**

** **

12 24 36 48 12 24 36 48 12 24 36 48

12 24 36 48 12 24 36 48 12 24 36 48

Viru

s yi

eld

(Log

10P

FU/m

l)

Viru

s yi

eld

(Log

10P

FU/m

l)

Viru

s yi

eld

(Log

10P

FU/m

l)

HA

tite

r (Lo

g 2)

HA

tite

r (Lo

g 2)

HA

tite

r (Lo

g 2)Hours post-infection Hours post-infection Hours post-infection

Hours post-infection Hours post-infection Hours post-infection

-T1/2011 (HA/NA)B/Yokohama/P 2922/2005 (HA/NA) B/Massachusetts B/Tokyo/UTE2/2008 (HA/NA) B/Tochigi/UTB/Brisbane/60/2008 (HA/NA)

23456789

10

9876543210

23456789

10

9876543210

10

98765432

1011

9876543210

10

**

** *

**

** ** **

**

****

**

**** **

**

** **

**

**** **

12 24 36 48 12 24 36 48 12 24 36 48

12 24 36 48 12 24 36 48 12 24 36 48

Viru

s yi

eld

(Log

10P

FU/m

l)

Viru

s yi

eld

(Log

10P

FU/m

l)

Viru

s yi

eld

(Log

10P

FU/m

l)

HA

tite

r (Lo

g 2)

HA

tite

r (Lo

g 2)

HA

tite

r (Lo

g 2)

Hours post-infection Hours post-infection Hours post-infection

Hours post-infection Hours post-infectionHours post-infection

Internal genes of WTInternal genes of HY

A B C D E F

Fig. 2. Comparison of wild-type and high-yield viruses possessing different HA and NA vRNAs. Viruses possessing the HA and NA vRNAs of the indicated virusesin combination with the internal vRNA segments of the respective natural wild-type isolate (WT), or of HY(Yam) (A–C) or HY(Vic) (D–F) (i.e., the viruses indicatedby the black graphs possess the eight wild-type vRNA segments of a human influenza B virus isolate). Experiments were carried out as described in the legend toFig. 1. The values presented are the average of three independent experiments ± SD. P values were calculated by using the linear mixed model described in theMethods section (*P < 0.05; **P < 0.01). Red asterisks indicate the comparison of the respective virus with WT virus. HA titer, hemagglutination titer.

Ping et al. PNAS | Published online December 5, 2016 | E8299

MICRO

BIOLO

GY

PNASPL

US

used reverse genetics to generate viruses possessing individualmutant vRNA segments of HY(Yam) or HY(Vic): for example,we generated viruses in which the HA and NA vRNAs of aYamagata-lineage virus were combined with the six remainingvRNAs of wild-type B/Yamagata/1/73 (used to generate virus li-braries), of wild-type B/Yamagata/1/73 [also encoding the NP-P40Smutation found in HY(Yam)], or of HY(Yam) (Fig. 7). Theresulting viruses were tested for their replicative abilities andhemagglutination titers in MDCK cells (Fig. 7). When tested in-dividually, the NP-P40S, M1-R77K, and NS-a39g+NS1-K176Qmutations significantly increased the viral and hemagglutinationtiters of Yamagata-linage viruses at one or more time pointscompared with the reference virus. For Victoria-lineage viruses,each of the mutations tested had a statistically significant growth-enhancing effect at one or more time points. Most of the virusespossessing individual mutations found in HY(Yam) or HY(Vic) didnot replicate as efficiently as the high-yield vaccine candidates,demonstrating that combinations of several mutations are impor-tant for the high-yield properties of HY(Yam) and HY(Vic).

Effect of Individual Mutations in HY(Yam) and HY(Vic) on the Activityof the Viral Replication Complex. The NP-P40S mutation was se-lected individually and in combination with NP-M204T from theYamagata- and Victoria-lineage libraries, respectively. More-over, these mutations increased viral and hemagglutination titerswhen tested without the other growth-enhancing mutations (Fig.7). To assess whether these mutations affect the activity of the

viral replication complex, we transfected 293T and MDCKcells with plasmids expressing the three polymerase subunits ofB/Yamagata/1/73 virus, wild-type or mutant B/Yamagata/1/73 NP,and an influenza B virus-like RNA expressing the luciferase re-porter protein (the NP-M204T mutant was not tested separatelybecause it was selected from the virus libraries only in combinationwith NP-P40S). Interestingly, the NP-P40S and -P40S/M204Tmutations significantly reduced the replicative activity of the viralreplication complex (SI Appendix, Fig. S5 A and B).The PA-a2272t mutation emerged during passages of virus

libraries in Vero cells (SI Appendix, Tables S5 and S6) and sig-nificantly increased the viral and hemagglutination titers of aVictoria-lineage vaccine candidate (Fig. 1). Here, we found that thismutation significantly increased the replication of a virus-like RNAin minireplicon assays (SI Appendix, Fig. S5C). HY(Yam) andHY(Vic) possess NS-a39g and NS-38(+1)g mutations, respectively.We, therefore, introduced these mutations into a virus-like RNAthat expresses luciferase. Both mutations conferred significantly in-creased expression of the reporter protein from the virus-like RNA(SI Appendix, Fig. S5D); this effect was substantially greater for theNS-38(+1)g mutation compared with the NS-a39g mutation.

Effect of Individual Mutations in HY(Yam) and HY(Vic) on the Compositionof Virus-Like Particles. The M1-R77K and M1-M86T mutations inHY(Yam) and HY(Vic) affected viral and hemagglutination titerswhen tested individually (Fig. 7). M1 is the major structural com-ponent of virions and mutations in this protein could affect the

98765432102

3456789

10

A

******

**** **** ****

******

**** **

12 24 36 48 12 24 36 48Hours post-infection Hours post-infection

Viru

s yi

eld

(Log

10P

FU/m

l)

HA

tite

r (Lo

g 2)

B/Yokohama/UT-K31/2012(Yamagata-lineage, WT)HY(Yam) + B/Yokohama/UT-K31/2012(HA+NA)HY(Vic) + B/Yokohama/UT-K31/2012(HA+NA)

23456789

10 9876543210

B

** *****

**** **** ****

12 24 36 48 12 24 36 48Hours post-infection Hours post-infection

Viru

s yi

eld

(Log

10P

FU/m

l)

HA

tite

r (Lo

g 2)

B/Yokohama/P-2922/2005(Yamagata-lineage, WT)HY(Yam) + B/Yokohama/P-2922/2005(HA+NA)HY(Vic) + B/Yokohama/P-2922/2005(HA+NA)

C

234567

89

10

9876543210

1011

******

******

****

****

******** ****

12 24 36 48 12 24 36 48Hours post-infection Hours post-infection

Viru

s yi

eld

(Log

10P

FU/m

l)

HA

tite

r (Lo

g 2)

B/Yokohama/UT-K1A/2011(Victoria-lineage, WT)HY(Yam) + B/Yokohama/UT-K1A/2011(HA+NA)HY(Vic) + B/Yokohama/UT-K1A/2011(HA+NA)

D

23456789

10

9876543210

1011

****

******** ****

****

**** ****

12 24 36 48 12 24 36 48Hours post-infection Hours post-infection

Viru

s yi

eld

(Log

10P

FU/m

l)

HA

tite

r (Lo

g 2)

B/Tochigi/UT-T1/2011(Victoria-lineage, WT)HY(Yam) + B/Tochigi/UT-T1/2011(HA+NA)HY(Vic) + B/Tochigi/UT-T1/2011(HA+NA)

Fig. 3. Exchange of HY(Yam) and HY(Vic) backbones. (A and B) Comparison of the virus and hemagglutination titers of two wild-type Yamagata-lineageviruses with viruses possessing the same HA and NA vRNAs in combination with the internal genes of HY(Yam) or HY(Vic). (C and D) Comparison of the virusand hemagglutination titers of two wild-type Victoria-lineage viruses with viruses possessing the same HA and NA vRNAs in combination with the internalgenes of HY(Yam) or HY(Vic). Experiments were carried out as described in the legend to Fig. 1. The values presented are the average of three independentexperiments ± SD. P values were calculated by using the linear mixed model described inMaterials and Methods (*P < 0.05; **P < 0.01). Red and blue asterisksindicate the comparison of the respective virus with WT virus; beige asterisks indicate the comparison between the viruses depicted in red and blue. HA titer,hemagglutination titer.

E8300 | www.pnas.org/cgi/doi/10.1073/pnas.1616530113 Ping et al.

composition of virions. We, therefore, transfected 293T cells withplasmids for the expression of the B/Yamagata/1/73 HA, NA, NP,BM2, NS2, and wild-type or mutant M1 proteins; expression ofthis set of viral protein results in efficient virus-like particle (VLP)formation and release (16). We then collected cell culture su-pernatant and assessed the VLP incorporation efficiency of viralproteins (SI Appendix, Fig. S6). Both mutations in M1 signifi-cantly increased the amount of viral HA, NP, and M1 proteins inthe culture supernatant, which presumably resulted in the in-creased viral and hemagglutination titers conferred by the M1-R77K and -86T mutations.

Effect of a Mutation in the HY(Yam) NS1 Protein on IFN AntagonistActivity. The NS1 protein is the major influenza viral IFN antag-onist (17, 18). To assess the effect of the HY(Yam) NS1-K176Qmutation on NS1’s ability to interfere with IFN-β synthesis, 293Tcells were transfected with wild-type or mutant NS1 protein ex-pression plasmids and with the reporter plasmid pGL–IFN-β,which encodes the firefly luciferase protein under the control ofthe IFN-β promoter (19). Cells were then infected with Sendaivirus to stimulate IFN-β synthesis, resulting in increased reportergene expression. Wild-type and mutant NS1 were comparable intheir ability to down-regulate IFN-β synthesis (SI Appendix, Fig.S7A). To determine the ability of wild-type and mutant NS1 tointerfere with the expression of IFN-β–stimulated genes, wetransfected 293T cells with wild-type or mutant NS1 protein ex-pression plasmids and the reporter plasmid pISRE-Luc (Prom-ega), which encodes the firefly luciferase protein under the controlof an IFN-regulated promoter. Cells were stimulated with IFN-β24 h later, incubated again for 24 h, and then assayed for luciferaseexpression. The NS1-K176Q protein was slightly less efficient thanwild-type NS1 in suppressing gene synthesis from an IFN-regu-lated promoter (SI Appendix, Fig. S7B), suggesting that othermechanisms account for its growth-enhancing effect.

DiscussionTo date, no systematic efforts have been carried out to develop ahigh-yield influenza B vaccine backbone. Vodeiko et al. (20) com-pared two influenza B viruses that differed in their replicative abilityin embryonated chicken eggs. Reassortment experiments revealedseveral vRNAs that contributed to the phenotypic differences(20); however, the specific amino acids that determined the growthproperties were not identified. Kim et al. (21) found that cold-adaptation of influenza B viruses enhanced their growth properties.Several amino acid changes in HA, NA, and NP (not the mutationsreported here) were responsible for the increased virus titers (21).Le et al. (22) tested reassortants between B/Lee/40 and Yamagata-and Victoria-lineage influenza B viruses isolated in 2002–2007; all14 high-yield candidates possessed the NP vRNA of B/Lee/40 virus,suggesting that this vRNA confers efficient replicative ability.We recently published a more comprehensive strategy to de-velop high-yield influenza A vaccine backbones (9), which weapplied here to influenza B viruses: from virus libraries pos-sessing random mutations in the internal genes, candidates wereselected that improved influenza B virus replication in MDCKand Vero cells. Combinations of mutations resulted in Yamagata-and Victoria-lineage vaccine candidates [encoding the NP-P40S,M1-R77K, NS1-K176Q, NS-a39g, and PA-a2272t mutations forthe Yamagata-lineage vaccine and the NP-P40S/M204T, M1-M86T,NS-(38+1)g, and PA-a2272t mutations for the Victoria-lineagevaccine] with high yield in MDCK and Vero cells, and also inembryonated chicken eggs. Further studies in (semi)industrialsettings would be necessary to determine whether lineage-specific vaccine backbones provide an advantage over a singlebackbone used for viruses of both lineages.We found that the total viral protein yield and HA content

obtained from HY(Yam) and HY(Vic) were substantially higherthan those from wild-type viruses (Fig. 5). High virus and HAyields are important for cost-effective vaccine production. Moreimportantly, increases in vaccine virus yield may be imperative inyears with high-vaccine demand and in years with shortened

Vir u

s yi

eld

(Lo g

10P

FU/m

l)

HA

tite

r (Lo

g 2)23456789

1011

0123456789

10A

******

******

****

******

****** ****

********

B/Yokohama/UT-K31/2012(Yamagata-lineage, WT)

PR8-HY + A/B-chimeric-Yokohama/UT-K31/2012(HA+NA)HY(Yam) + B/Yokohama/UT-K31/2012(HA+NA)

12 24 36 48Hours post-infection

12 24 36 48Hours post-infection

23456789

1011

0123456789

B

******

****

**

******** ****

B/Massachusetts/2/2012(Yamagata-lineage, vaccine virus)

PR8-HY + A/B-chimeric-Massachusetts/2/2012(HA+NA)HY(Yam) + B/Massachusetts/2/2012(HA+NA)

12 24 36 48Hours post-infection

12 24 36 48Hours post-infection

Viru

s yi

eld

( Log

10P

FU/m

l)

HA

tite

r (Lo

g 2)

0123456789

101112

23456789

1011

C

B/Yokohama/UT-K1A/2011(Victoria-lineage, WT)

PR8-HY + A/B-chimeric-Yokohama/UT-K1A/2011(HA+NA)HY(Vic) + B/Yokohama/UT-K1A/2011(HA+NA)

******

****** ****

******

****** ****

********

12 24 36 48Hours post-infection

12 24 36 48Hours post-infection

Viru

s yi

eld

(Log

10P

FU/m

l)

HA

tite

r (L o

g 2)

D

23456789

1011

0123456789

10

****

******

**** ****

******

****** ****

B/Brisbane/60/2008(Victoria-lineage, vaccine virus)

PR8-HY + A/B-chimeric-Brisbane/60/2008(HA+NA)HY(Vic) + B/Brisbane/60/2008(HA+NA)

12 24 36 48Hours post-infection

12 24 36 48Hours post-infection

HA

tite

r (Lo

g 2)

Viru

s yi

eld

(Lo g

10P

FU/m

l)

23456789

0123456789

10

E

******

****

**** ** **

B/Massachusetts/2/2012(Yamagata-lineage, vaccine virus)

PR8-HY + A/B-chimeric-Massachusetts/2/2012(HA+NA)HY(Yam) + B/Massachusetts/2/2012(HA+NA)

Viru

s yi

eld

(Log

10P

FU/m

l)

HA

tite

r (Lo

g 2)

12 24 36 48 60Hours post-infection

12 24 36 48 60Hours post-infection

12 24 36 48 6023456789

0123456789

10

F**** ***

****

**** *** **** ****

12 24 36 48 60

HA

tite

r (L o

g 2)

V ir u

s yi

eld

(Log

10P

FU/m

l)

B/Brisbane/60/2008(Victoria-lineage, vaccine virus)

PR8-HY + A/B-chimeric-Brisbane/60/2008(HA+NA)HY(Vic) + B/Brisbane/60/2008(HA+NA)

Hours post-infection Hours post-infection

Fig. 4. Comparison of high-yield influenza A and B vaccine virus backbones. (A and B) We compared the virus yield and hemagglutination titers of: (i) theindicated wild-type viruses; (ii) viruses possessing the indicated HA and NA vRNAs in combination with the internal genes of HY(Yam); and (iii) virusespossessing the indicated type A/B chimeric HA and NA vRNAs in combination with the internal genes of high-yield influenza A virus. (C and D) Similar ex-periments were carried out for viruses of the Victoria lineage. (E and F) Comparison of the indicated wild-type and hybrid viruses in embryonated chickeneggs. Experiments were carried out as described in the legend to Fig. 1. The values presented are the average of three independent experiments ± SD. Thestatistical significance was determined by using the linear mixed model described in Materials and Methods (A–D), or by two-way ANOVA, followed byTukey’s post hoc test (E and F) (*P < 0.05; **P < 0.01); P values are not shown if the titer of the high-yield vaccine candidate was lower than that of wild-typevirus. Red and blue asterisks indicate the comparison of the respective virus with WT virus; beige asterisks indicate the comparison between the virusesdepicted in red and blue. HA titer, hemagglutination titer.

Ping et al. PNAS | Published online December 5, 2016 | E8301

MICRO

BIOLO

GY

PNASPL

US

vaccine production times. However, some of the observed dif-ferences in virus titers or HA yield were small (although statis-tically significant), and we currently do not know the extent bywhich HY(Yam) and HY(Vic) would increase vaccine virus yieldin industrial vaccine production.Evaluation of influenza B virus sequences revealed that the

amino acid changes in HY(Yam) and HY(Vic) are rare amongnatural influenza B viruses (i.e., the NP-P40S, NP-M204T, and

M1-M86T mutations have each been found in one isolate; theM1-R77K mutation has been reported in 14 isolates; and theNS1-K176Q mutation has not been detected). The 3D structureof an influenza B virus NP protein has been resolved (23), butposition 40 (at which P-to-S mutations were selected fromthe Yamagata- and Victoria-lineage libraries) is part of theN-terminal 71 amino acids for which no structural data could beobtained, suggesting that this region is highly flexible. Sequence

0

1

2

3

4

HA

con

tent

(mg/

liter

)

B/Yok

oham

a/UT-K

31/20

12

Y(Yam

)+B/Y

okoh

ama/U

T-K31

/2012

(HA+N

A)

B/Mas

sach

usett

s/2/20

12

HY(Yam

)+B/M

assa

chus

etts/2

/2012

(HA+N

A)0123456789

Tota

l vira

l pro

tein

(mg/

liter

)100 kD75 kD

50 kD

37 kD

25 kD20 kD15 kD

NP

Deglycosylated HA1

Deglycosylated HA2

M1

HA1

HA2

_ +PNGase F: _ +

B/Yokohama/UT-K31/2012

_ + _ +

B/Massachusetts/2/2012

WT HY(Yam) WT HY(Yam)

A

****

****

B/Yok

oham

a/UT-K

31/20

12

HY(Yam

)+B/Y

okoh

ama/U

T-K31

/2012

(HA+N

A)

B/Mas

sach

usett

s/2/20

12

HY(Yam

)+B/M

assa

chus

etts/2

/2012

(HA+N

A)

NP

Deglycosylated HA1

Deglycosylated HA2

M1

100 kD75 kD

50 kD

37 kD

25 kD20 kD15 kD

HA1

HA2

_ +PNGase F: _ +

B/Yokohama/UT-K1A/2011

_ + _ +

B/Brisbane/60/2008

WT HY(Vic) WT HY(Vic)

B

****

****

Tota

l vira

l pro

tein

(mg/

liter

)

HA

con

tent

(mg/

liter

)

B/Yok

oham

a/UT-K

1A/20

11

HY(Vic)

+B/Y

okoh

ama/U

T-K1A

/2011

(HA+N

A)

B/Bris

bane

/60/20

08

HY(Vic)

+B/B

risba

ne/60

/2008

(HA+N

A)

B/Yok

oham

a/UT-K

1A/20

11

HY(Vic)

+B/Y

okoh

ama/U

T-K1A

/2011

(HA+N

A)

B/Bris

bane

/60/20

08

HY(Vic)

+B/B

risba

ne/60

/2008

(HA+N

A)0123456789

0

1

2

3

4

Fig. 5. Evaluation of the total viral protein yield and HA content of HY(Yam) and HY(Vic) viruses. (A) Comparison of viruses possessing the HA and NA vRNAs ofthe indicated Yamagata-lineage viruses in combination with the internal vRNAs of the same natural wild-type virus (WT) or of HY(Yam). (B) Comparison of virusespossessing the HA and NA vRNAs of the indicated Victoria-lineage viruses in combination with the internal vRNAs of the same wild-type virus (WT) or of HY(Vic).The total viral protein yield of MDCK cell-grown, sucrose gradient-purified virus samples is shown (Left and Center). PNGaseF treatment deglycosylates HA1 andHA2; this treatment was carried out because glycosylated HA2migrates at a similar molecular weight as M1. The HA contents (Right) were calculated based on thetotal viral protein amounts and the relative amounts of HA; for details, see Materials and Methods. The values presented are the average of three independentexperiments ± SD. The statistical significance was assessed by using one-way ANOVA followed by Dunnett’s test, comparing the total viral protein yield and HAcontent of wild-type viruses with that of recombinant high-yield vaccine viruses (*P < 0.05; **P < 0.01).

E8302 | www.pnas.org/cgi/doi/10.1073/pnas.1616530113 Ping et al.

analysis of influenza B virus NP proteins revealed that the prolineat position 40 is highly conserved: only 1 in 3,234 sequences doesnot encode NP-40P (the only exception, B/Tennessee/01/2015,encodes NP-40S, as found in our study). Interestingly, the NP-P40S mutation reduced the activity of the viral replication com-plex in minireplicon assays, suggesting that the yield-enhancingeffect of this mutation is mediated by functions other than rep-lication and transcription. For example, this region in NP couldinteract with viral or cellular proteins during the export of viralribonucleoprotein complexes from the nucleus to the cytoplasm,or during virion assembly.We also identified several mutations in the noncoding re-

gions of vRNA segments that increased virus yield; some ofthese mutations conferred increased levels of replication andtranscription, as measured in minireplicon assays. These muta-tions may, for example, increase the stability of the vRNA oraffect its interaction with the viral polymerase complex. Furtherstudies are needed to decipher the exact mechanistic functionsof these mutations.Collectively, we have developed influenza B vaccine virus

backbones that could increase the titers of seasonal influenzaB vaccines in the propagation systems currently used for humaninfluenza vaccine virus production.

Materials and MethodsCells. MDCK cells were grown in MEM containing 5% (vol/vol) newborn calfserum. Vero cells were maintained in MEM containing 10% (vol/vol) FBS.293T human embryonic kidney cells were grown in DMEM supplementedwith 10% (vol/vol) FBS.

Construction of Plasmids. The sequences of the eight viral RNAs of B/Yama-gata/1/73 virus (available in our group) were used to design gBlocks GeneFragments (Integrated DNA Technologies), which were amplified and joinedby PCR; the resulting viral cDNAs were inserted into the RNA polymerase Ivector pHH21 (24). The vRNAs of the B/Yokohama/UT-K31/2012, B/Yokohama/UT-K1A/2011, B/Yokohama/P-2922/2005, B/Tokyo/UTE2/2008, B/Tochigi/UT-T1/2011, B/Massachusetts/2/2012, and B/Brisbane/60/2008 viruses were extractedfrom virus stocks by using the RNeasy Kit (Qiagen). The viral HA and NA geneswere amplified with gene-specific oligonucleotides by using the One-Step RT-PCR Kit (Invitrogen), and the PCR-products were cloned into thepHH21 vector. The HA and NA genes of B/Yamagata/16/1988 were synthesizedby PCR-amplification of joined gBlocks Gene Fragments, followed by cloninginto pHH21. The type A/B chimeric HA and NA genes of B/Yokohama/UT-K31/2012, B/Yokohama/UT-K1A/2011, B/Massachusetts/2/2012 and B/Brisbane/60/2008 viruses were generated by overlapping PCRs based on the strategy shownin SI Appendix, Fig. S4.

Construction of Plasmid Libraries. We introduced one to four random mu-tations into each of the six internal genes of B/Yamagata/1/73 virus by error-prone PCR using the GeneMorph II Random Mutagenesis Kit. The randomlymutated PCR products were inserted into the pHH21 vector, and the diversityof the resulting plasmid libraries was confirmed by sequence analysis of atleast 24 Escherichia coli colonies for each viral gene.

Virus Rescue and Virus Library Generation. Wild-type viruses and virus librariespossessing randommutations in the internal genes were generatedwith the helpof reverse-genetics approaches (24). Virus libraries were generated by trans-fecting 293T cells with a mutant plasmid library instead of the wild-type con-struct. Forty-eight hours later, supernatants from plasmid-transfected 293T cellswere collected and amplified in MDCK cells to generate virus stock; the titers ofthe virus stocks were determined by using plaque assays in MDCK cells.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 1440

60

80

100

120

Rel

ativ

e bo

dy w

eigh

t (%

)

BA

0 1 2 3 4 5 6 7 8 9 10 11 12 13 140

20

40

60

80

100

Sur

viva

l (%

)

Days post-infection Days post-infection

Day 3 Day 6C

012345678

012345678

****

****

Viru

s tit

er ( L

og10

PFU

/g)

V iru

s tit

er (L

og10

PFU

/ g)

B/Massachusetts/2/2012 B/Yamagata/1/73+B/Massachusetts/2/2012(HA+NA) HY(Yam)+B/Massachusetts/2/2012(HA+NA)PBS infection

D

Rel

ativ

e bo

dy w

eigh

t (%

)

Sur

viva

l (%

)

40

60

80

100

120

0

20

40

60

80

100

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14Days post-infection Days post-infection

Day 3 Day 6F

012345678

012345678**

** ****

Viru

s tit

er (L

og10

PFU

/g)

Viru

s tit

er (L

og10

PFU

/g)

B/Yamagata/1/73+B/Brisbane/60/2008(HA+NA)HY(Vic) + B/Brisbane/60/2008(HA+NA)

B/Brisbane/60/2008

PBS infection

E

Fig. 6. Virulence of HY(Yam) and HY(Vic) viruses in mice. (A–C) Comparison of a wild-type Yamagata-lineage virus (B/Massachusetts/2/2012), a virus pos-sessing the B/Massachusetts/2/2012 HA and NA vRNAs in combination with the remaining vRNAs of B/Yamagata/1/73 (used for virus library generation), and avirus possessing the B/Massachusetts/2/2012 HA and NA vRNAs in combination with the remaining vRNAs of HY(Yam). (D–F) Comparison of a wild-typeVictoria-lineage virus (B/Brisbane/60/2008), a virus possessing the B/Brisbane/60/2008 HA and NA vRNAs in combination with the remaining vRNAs ofB/Yamagata/1/73 (used for virus library generation), and a virus possessing the B/Brisbane/60/2008 HA and NA vRNAs in combination with the remainingvRNAs of HY(Vic). BALB/c mice (five per group) were inoculated intranasally with 106 pfu of the indicated viruses and monitored daily for body weight changes(A and D) and survival (B and E). To assess virus replication in mice, 106 pfu of the indicated viruses were used to infect 10 additional mice. On days 3 and 6postinfection, five mice in each group were killed, and lung virus titers were determined by use of plaque assays in MDCK cells (C and F). Statistical significancewas assessed by using one-way ANOVA followed by Dunnett’s test (*P < 0.05; **P < 0.01).

Ping et al. PNAS | Published online December 5, 2016 | E8303

MICRO

BIOLO

GY

PNASPL

US

Evaluation of Viral Growth Kinetics. Wild-type or recombinant viruses wereinoculated in triplicate into MDCK cells at a multiplicity of infection (MOI)of 0.001. After infection, cells were incubated with MEM/BSA medium with0.6 μg/mL TPCK-trypsin. Supernatants were collected at the indicated timepoints and the virus titers were assessed by means of plaque assays inMDCK cells.

To analyze viral replication in embryonated chicken eggs, we inoculated10-day-old embryonated chicken eggs (four per virus) with 1 × 104 pfu ofvirus and incubated them at 35 °C. The allantoic fluids were collected atthe indicated time points and virus titers were determined by use of plaqueassays in MDCK cells.

The hemagglutination titers of viruses amplified in MDCK cells or em-bryonated chicken eggs were determined by using a hemagglutination assay.Briefly, 50 μL of virus sample was serially diluted twofold in 96-well U-bot-tom microtiter plates (Thermo Scientific) containing 50 μL of PBS per well.Next, 50 μL of 0.5% turkey red blood cells were added to each well, plateswere incubated for 45 min at room temperature, and hemagglutinationtiters were calculated as the reciprocal value of the highest dilution at whichagglutination occurred.

Virus Concentration and Purification. MDCK cells were grown in two 4-LayersEasy-Fill Cell Factories (Thermo Scientific) and infected with wild-type or high-yield influenza B viruses at an MOI of 0.001 when the cells reached ∼95%confluency. Cell-culture supernatants were harvested 48 h later and clarifiedby centrifugation (3,500 rpm in a Beckman SX4750 rotor, 15 min, 4 °C).Viruses were pelleted by ultracentrifugation (18,500 rpm, 90 min at 4 °C in aBeckman Type19 rotor), resuspended in 5 mL of PBS, and loaded onto 20–50% (wt/vol) continuous sucrose gradients which were centrifuged at 25,000rpm for 90 min at 4 °C in a Beckman SW32 rotor. The virus-containing bandwas collected, diluted in PBS, and pelleted again by centrifugation (25,000rpm, 90 min, 4 °C, Beckman SW32 rotor). The virus pellet was resuspended in400 μL of PBS, aliquoted, and stored at −80 °C.

Total Protein Assay. Total protein yield of virus concentrates was determinedby using the Pierce BCA protein assay kit (Thermo Scientific) according to themanufacturer’s instructions.

Deglycosylation of Viral Proteins Using PNGase F. To remove sugar moieties,we denatured 10 μL of virus concentrate. The sample was then incubated at37 °C for 20 h with 2 μL of a one-tenth dilution of PNGase F enzyme (NewEngland Biolabs) in the buffer provided by the manufacturer and withNonidet P-40 at a final concentration of 1%.

SDS/PAGE. We mixed 2 μL of virus concentrate with PBS to a total volume of10 μL, added 2.5 μL of loading dye with 2% (vol/vol) β-mercaptoethanol asreducing agent, and heated the samples to 95 °C for 5 min. Samples werethen loaded onto NuPage 4–12% (wt/vol) Bris-Tris precast gels (Life tech-nology), which were run at 150 V for 120 min using 1× Mes buffer (Bio-Rad),and then stained with SYPRO-Ruby (Sigma). Quantitation of protein amountswas carried out by using ImageJ software (NIH). The HA content was calcu-lated by dividing the HA amount (calculated by summing the amounts of HA1and HA2) by the sum of the amounts of HA1, HA2, NP, and M1, and multi-plying this value by the amount of total viral protein.

Virulence Studies in Mice. Six-week-old female BALB/c mice (Jackson Labo-ratory) were anesthetized with isoflurane and inoculated intranasally with106 pfu of influenza B viruses in a volume of 50 μL. We infected five mice pergroup; this sample size is adequate to detect large effects between groups.Mice were randomized and investigators were not blinded. Body weightchanges and survival were monitored daily for 14 days. To assess virus rep-lication in mice, we infected 10 mice per virus with 106 pfu; on days 3 and 6postinfection, five mice in each group were killed and virus titers in the lungswere determined by use of plaque assays in MDCK cells.

Genetic Stability Testing. To evaluate the genetic stability of the high-yieldvaccine backbones, we passaged viruses possessing the HY(Yam) and HY(Vic)backbones combined with the HA and NA vRNAs of B/Yokohama/UT-K31/2012 and B/Yokohama/UT-K1A/2011, respectively, 10 times in MDCK cells atan MOI of 0.01. Viruses collected after each passage were sequenced bymeans of Sanger sequencing.

Minireplicon Assay. For the minireplicon assay, 293T cells andMDCK cells weretransfected with 0.25 μg each of plasmids expressing the B/Yamagata/1/73PB2, PB1, PA and NP proteins, together with 0.05 μg of pPolI-B/Yamagata/1/73-NS-Luc (which encodes the firefly luciferase gene under the control of thehuman RNA polymerase I promoter) or pPolIC250-B/Yamagata/1/73-NS-Luc(which encodes the firefly luciferase gene under the control of the canineRNA polymerase I promoter), respectively. Cells were cotransfected with0.025 μg of pGL4.74[hRluc/TK] (an internal control to monitor transfectionefficiency; Promega). The transfected cells were incubated at 35 °C for 48 h,lysed, and assayed for luciferase activity by using the dual-luciferase systemdetector kit according to the manufacturer’s protocol (Promega). The fireflyluciferase expression levels were normalized to the Renilla luciferase activity.The data presented are the averages of three independent experiments ± SD.

To investigate the significance of the mutations in the noncoding regionsof the B/Yamagata/1/73 PA and NS1 vRNAs, we transfected cells as describedabove; however, pPolIC250-NP(0)Fluc(0) was replaced with a reporter con-struct in which the firefly luciferase gene was flanked by the wild-type ormutant noncoding regions of the PA or NS vRNAs, respectively. At 48 hposttransfection, luciferase activity was measured as described above.

VLP Budding Assay. For the VLP budding assay, 293T cells were transfected with2 μg each of protein expression plasmids for wild-type or mutant B/Yamagata/1/73 M1, HA, NA, NP, BM2, and NS2. At 48 h posttransfection, culture

A

B

Fig. 7. Growth kinetics and hemagglutination titers of single reassortantviruses. (A) Comparison of the parental virus used for Yamagata-lineagevirus library generation (i.e., B/Yamagata/1/73 with the HA and NA vRNAs ofB/Yokohama/UT-K31/2012) with viruses that also possess an individual vRNAof HY(Yam). (B) Comparison of the parental virus used for Victoria-lineagevirus library generation (i.e., B/Yamagata/1/73 with the HA and NA vRNAs ofB/Yokohama/UT-K1A/2011) with viruses that also possess an individual vRNAof HY(Vic). Experiments were carried out as described in the legend to Fig. 1.Data were obtained from three independent experiments; shown are av-erage titers ± SD. The values presented are the average of three indepen-dent experiments ± SD. Statistical significance was determined by using thelinear mixed model described in Materials and Methods (*P < 0.05; **P <0.01). The color of the asterisks indicates the comparison of the respectivevirus with the comparator virus (depicted in black). HA titer, hemaggluti-nation titer.

E8304 | www.pnas.org/cgi/doi/10.1073/pnas.1616530113 Ping et al.

supernatant was harvested, clarified, loaded on a 20% (wt/vol) sucrosecushion, and ultracentrifuged at 60,000 rpm for 2 h in a Beckman SW 60 Tirotor; the pelleted VLPs were then resuspended in PBS overnight at 4 °C. Inparallel, we lysed the transfected cells with RIPA buffer.

Purified VLPs and cell lysates were separately mixed with 5× loading dyebuffer and fractionated on NuPage 4–12% (wt/vol) Bris-Tris precast gels (LifeTechnology). The proteins were transferred to nitrocellulose membranes byusing an iBlot dry blotting system (Invitrogen). The membranes were thenblocked for 3 h at room temperature with PBS with 0.05% Tween 20 (PBS-T)containing 5% (wt/vol) skimmed milk. Then, the membranes were incubatedwith monoclonal antibodies to B/Brisbane/60/2008 HA (1:1,000; my BioSource,MBS430175), or to influenza B virus NP (1:1,000; Abcam, ab47876) or M1(1:2,000; Abcam, ab82608) protein overnight at 4 °C. After four washes withPBS-T for 10 min each, the membranes were incubated with goat anti-mouse secondary antibodies conjugated with horseradish peroxidase(1:2,000; Life Technology) for 1 h at room temperature. After four washeswith PBS-T for 10 min each, the blots were developed by using lumi-lightWestern blotting substrate (Roche Applied Science) and visualized follow-ing autoradiography. Quantitation of protein amounts was carried out byusing ImageJ software (NIH). To calculate the percentages of HA, NP, and M1protein incorporation into VLPs, the following formula was used: [ratio ofthe amount of protein in the mutant VLP to the total amount of mutantprotein (VLP + cell lysate)/ratio of the amount of protein in the wild-typeVLP to the total amount of wild-type protein (VLP + cell lysate)] × 100.

IFN Antagonist Assays. To assess the IFN-antagonist activity of wild-type andmutant NS1 proteins, 293T cells were transfected with the NS1 protein ex-pression plasmid and the reporter plasmid pGL–IFN-β, which encodes thefirefly luciferase protein under the control of the IFN-β promoter (19).Twenty-four hours posttransfection, we infected cells with Sendai virus at anMOI of 5 for 1 h. Cells were incubated for 24 h and lysed with Glo lysis buffer(Promega); then, we added Steady-Glo assay buffer (Promega) and mea-sured luciferase expression. In another set of experiments, 293T cells weretransfected with wild-type or mutant NS1 protein expression plasmids andthe reporter plasmid pISRE-Luc (Promega), which encodes the firefly lucif-erase protein under the control of an IFN-regulated promoter. Twenty-fourhours later, cells were treated with 104 U/mL of human IFN-β for another24 h, followed by lysis and measurement of luciferase expression levels.

Statistical Analysis. Statistical analyses of the data were accomplished usingthe R software (www.r-project.org), v3.1. To compare multiple groups with

measurements collected independently at several time points, we used two-way ANOVA followed by Tukey’s post hoc test. To compare measurementsfrom multiple groups collected at a single time point, we used one-wayANOVA followed by either Tukey’s or Dunnett’s post hoc test. To comparemultiple groups with dependent measurements (e.g., viral growth curves incell culture for which aliquots were collected from the same culture at dif-ferent time points), we fitted a linear mixed-effects model to the data byusing the R package NLME; the time, virus strains, and interaction betweenthese two factors were considered. We used the R package PHIA to build acontrast matrix for comparing strains in a pairwise fashion at the same timepoint (e.g., group_1 vs. group_2 at 24 h postinfection, group_1 vs. group_3at 24 h postinfection, group_2 vs. group_3 at 24 h postinfection). Compar-isons were performed individually; therefore, the final P values were ad-justed by using Holm’s method to account for multiple comparisons.

Raw data from growth curves were converted to the logarithmical scalebefore being analyzed; results were considered statistically significant forP (or adjusted P values) < 0.05. Variance between groups was assessedby using Levene’s test (which was similar for the groups being compared,with P > 0.05).

Ethics and Biosafety. Our experiments in mice followed the University ofWisconsin–Madison’s Animal Care and Use Protocol. All experiments wereapproved by the Animal Care and Use Committee of the University of Wis-consin–Madison (protocol number V00806), which acknowledged and ac-cepted both the legal and ethical responsibility for the animals, as specifiedin the Fundamental Guidelines for Proper Conduct of Animal Experimentand Related Activities in the Animal Welfare Act and associated AnimalWelfare Regulations and Public Health Service Policy.

The NIH reviewed and approved these experiments, granting an exceptionto the pause on gain-of-function research.

ACKNOWLEDGMENTS. We thank Susan Watson for scientific editing. Thiswork was supported by the National Institute of Allergy and InfectiousDiseases-funded Center for Research on Influenza Pathogenesis (ContractHHSN272201400008C); Scientific Research on Innovative Areas from the Min-istry of Education, Culture, Sports, Science, and Technology of Japan Grants16H06429, 16K21723, and 16H06434; grants from the Strategic Basic Re-search Program of the Japan Science and Technology Agency; and the Lead-ing Advanced Projects for medical innovation from the Japan Agency forMedical Research and Development.

1. Paul Glezen W, Schmier JK, Kuehn CM, Ryan KJ, Oxford J (2013) The burden of in-fluenza B: A structured literature review. Am J Public Health 103(3):e43–e51.

2. Tafalla M, Buijssen M, Geets R, Vonk Noordegraaf-Schouten M (2016) A compre-hensive review of the epidemiology and disease burden of Influenza B in 9 Europeancountries. Hum Vaccin Immunother 12(4):993–1002.

3. van de Sandt CE, Bodewes R, Rimmelzwaan GF, de Vries RD (2015) Influenza B viruses:Not to be discounted. Future Microbiol 10(9):1447–1465.

4. Rota PA, et al. (1990) Cocirculation of two distinct evolutionary lineages of influenzatype B virus since 1983. Virology 175(1):59–68.

5. Belshe RB (2010) The need for quadrivalent vaccine against seasonal influenza.Vaccine 28(Suppl 4):D45–D53.

6. Belshe RB, Toback SL, Yi T, Ambrose CS (2010) Efficacy of live attenuated influenzavaccine in children 6 months to 17 years of age. Influenza Other Respi Viruses 4(3):141–145.

7. Ambrose CS, Levin MJ (2012) The rationale for quadrivalent influenza vaccines. HumVaccin Immunother 8(1):81–88.

8. Maassab HF (1969) Biologic and immunologic characteristics of cold-adapted in-fluenza virus. J Immunol 102(3):728–732.

9. Ping J, et al. (2015) Development of high-yield influenza A virus vaccine viruses. NatCommun 6:8148.

10. Wright PF, Neumann G, Kawaoka Y (2013) Orthomyxoviruses. Fields Virology, edsKnipe DM, et al. (Lippincott Williams & Wilkins, Philadelphia, PA), 6th Ed, vol 1, pp1186–1243.

11. Ramanunninair M, et al. (2013) Molecular signature of high yield (growth) influenza avirus reassortants prepared as candidate vaccine seeds. PLoS One 8(6):e65955.

12. Horimoto T, Iwatsuki-Horimoto K, Hatta M, Kawaoka Y (2004) Influenza A virusespossessing type B hemagglutinin and neuraminidase: Potential as vaccine compo-nents. Microbes Infect 6(6):579–583.

13. Flandorfer A, García-Sastre A, Basler CF, Palese P (2003) Chimeric influenza A viruseswith a functional influenza B virus neuraminidase or hemagglutinin. J Virol 77(17):9116–9123.

14. Fujii Y, Goto H, Watanabe T, Yoshida T, Kawaoka Y (2003) Selective incorporation ofinfluenza virus RNA segments into virions. Proc Natl Acad Sci USA 100(4):2002–2007.

15. Baker SF, et al. (2014) Influenza A and B virus intertypic reassortment through com-patible viral packaging signals. J Virol 88(18):10778–10791.

16. Gómez-Puertas P, et al. (1999) Efficient formation of influenza virus-like particles:Dependence on the expression levels of viral proteins. J Gen Virol 80(Pt 7):1635–1645.

17. Yuan W, Krug RM (2001) Influenza B virus NS1 protein inhibits conjugation of theinterferon (IFN)-induced ubiquitin-like ISG15 protein. EMBO J 20(3):362–371.

18. Dauber B, Heins G, Wolff T (2004) The influenza B virus nonstructural NS1 protein isessential for efficient viral growth and antagonizes beta interferon induction. J Virol78(4):1865–1872.

19. Bale S, et al. (2012) Marburg virus VP35 can both fully coat the backbone and cap theends of dsRNA for interferon antagonism. PLoS Pathog 8(9):e1002916.

20. Vodeiko GM, McInnis J, Chizhikov V, Levandowski RA (2003) Genetic and phenotypicanalysis of reassortants of high growth and low growth strains of influenza B virus.Vaccine 21(25-26):3867–3874.

21. Kim H, Velkov T, Camuglia S, Rockman SP, Tannock GA (2015) Cold adaptationgenerates mutations associated with the growth of influenza B vaccine viruses.Vaccine 33(43):5786–5793.

22. Le J, et al. (2015) Development of high yield reassortants for influenza type B virusesand analysis of their gene compositions. Vaccine 33(7):879–884.

23. Ng AK, et al. (2012) Structural basis for RNA binding and homo-oligomer formationby influenza B virus nucleoprotein. J Virol 86(12):6758–6767.

24. Neumann G, et al. (1999) Generation of influenza A viruses entirely from clonedcDNAs. Proc Natl Acad Sci USA 96(16):9345–9350.

Ping et al. PNAS | Published online December 5, 2016 | E8305

MICRO

BIOLO

GY

PNASPL

US