Influenza A variants with reduced susceptibility to baloxavir isolated from Japanese patients are fit and transmit through respiratory droplets

Abstract

Here we report the isolation of influenza A/H1N1 2009 pandemic (A/H1N1pdm) and A/H3N2 viruses carrying an I38T mutation in the polymerase acidic (PA) protein, which confers reduced susceptibility to baloxavir, from patients before drug treatment and from patients after treatment with this drug in Japan. These variants showed replicative abilities and pathogenicity similar to those of wild-type isolates in hamsters; they also transmitted efficiently between ferrets via respiratory droplets.

Influenza A and B viruses possess eight single-stranded negative-sense RNA segments; each segment associates with a heterotrimeric polymerase complex and nucleoprotein (NP). The polymerase complex consists of the PB2 (polymerase basic 2), PB1 (polymerase basic 1), and PA (polymerase acidic) subunits, which are responsible for transcription and replication of the viral RNA genome. In 2018 and 2019, baloxavir marboxil (BXM), the first in a new class of anti-influenza drugs that target the polymerase complex, was licensed in Japan, the United States, and Hong Kong. This drug efficiently prevents the replication of influenza A and B viruses by inhibiting the cap-dependent endonuclease activity of their PA subunits, which is involved in the generation of capped RNA primers for viral transcription^1,2.

Previous studies have shown that propagation of influenza viruses in the presence of baloxavir acid (BXA), the active form of BXM, in cell culture can result in the generation of variants bearing amino acid changes at position 38 of the PA protein, such as I38T, which confer reduced susceptibility to BXA¹. Importantly, changes at position 38 of PA were detected in 4/182 (2.2%) and 36/370 (9.7%) of influenza patients treated with BXM in Phase II and III studies, respectively³. Notably, in a pediatric study, PA-I38 variants were presented in 18 (23.4%) of 77 BXM-treated children with influenza². Influenza virus surveillance studies in Japan during the 2018–2019 influenza season reported that PA-I38T and I38M variants were isolated from BXM-treated children with influenza A/H3N2^4,5. In addition, an A/H3N2 virus with the PA-I38T substitution was detected in a hospitalized child who had not received BXM⁵, suggesting the possibility of person-to-person transmission of the variant.

Other studies found that recombinant influenza A/H1N1 and A/H3N2 viruses expressing the I38T substitution in their PA protein in the background of the A/WSN/33 (H1N1) or A/Victoria/2/75 (H3N2) strains were greatly attenuated in their replication in cell culture relative to their BXA-sensitive counterparts^1,2, suggesting that the reduction in replicative fitness of influenza A viruses is caused by the I38T mutation. However, the impact of this substitution on the viral fitness of clinical isolates from patients and on viral fitness in vivo is unknown. Furthermore, it remains unclear whether the emergence of this mutation is related to adverse clinical outcomes. During the 2018–2019 influenza season, A/H1N1 2009 pandemic (A/H1N1pdm) and A/H3N2 viruses were the predominant influenza viruses circulating in Japan (https://nesid4g.mhlw.go.jp/Byogentai/Pdf/data95j.pdf). Here, we characterized the in vitro and in vivo properties of influenza A/H1N1pdm and A/H3N2 viruses with reduced susceptibility to BXA isolated from patients in Japan.

In Japan, BXM was approved for the treatment of influenza across all age groups on February 23, 2018. To monitor the susceptibility of A/H1N1pdm and A/H3N2 viruses to BXA in Japan during the winter season after its licensure, we first analyzed respiratory specimens collected from patients prior to drug treatment. To detect viruses possessing an amino acid change at residue 38 of PA that showed reduced susceptibility to BXA, the nucleotide sequences of the PA genes of viruses in pre-treatment clinical specimens from 253 patients, confirmed to be influenza virus-positive by real-time RT-PCR (96 A/H1pdm-positive patients and 157 A/H3-positive patients), were determined by means of Sanger sequencing (Extended Data Fig. 1a,b). All of the clinical specimens from untreated patients with A/H1pdm encoded isoleucine at position 38 of PA (Supplementary Table 1). However, among untreated patients with A/H3, samples (BB233–0 and GR142–0) from two patients (BB233 and GR142) contained variants carrying the PA-I38T mutation (Supplementary Table 2). These cases will be discussed in more detail later.

We next attempted to identify the frequency with which viruses with decreased susceptibility to BXA emerge following treatment. Pre-treatment and post-treatment clinical specimens were obtained from 38 patients with influenza A virus infections (22 A/H1pdm and 16 A/H3), and the PA sequences of the viruses in the samples were determined by Sanger sequencing (Extended Data Fig. 1a,b). All of the viruses in the pre-treatment samples encoded isoleucine at position 38 of PA (Extended Data Fig. 1a,b and Supplementary Tables 1–4). By contrast, when the post-treatment samples were analyzed, variants, including those bearing the PA-I38T, PA-I38T/N/S, PA-I38T/S, and/or PA-I38M substitutions, were detected in 5/22 (22.7%) and 4/16 (25.0%) of the A/H1pdm- and A/H3-positive patients, respectively. Interestingly, all but one of these patients who shed variants were children (aged 0 to 15 years): 4/15 (26.7%) for the A/H1pdm-positive cohort and 4/12 (33.3%) for the A/H3-positive group. For one adult (A/H3) and five pediatric (2 A/H1pdm and 3 A/H3) patients, we were able to obtain only post-BXM treatment samples. Of these, PA-I38T variants were detected in four pediatric patients: 1 for A/H1pdm and 3 for A/H3 (Extended Data Fig. 1a,b and Supplementary Tables 3 and 4). We also performed deep sequencing analysis on all pre-treatment and post-treatment samples obtained from 259 patients (98 A/H1pdm-positive patients and 161 A/H3-positive patients). This analysis detected all PA-I38 variants in the clinical samples that were found by Sanger sequencing (Supplementary Tables 1–4). The analysis did not detect mutations that occurred at position 38 of PA in any of the samples from which PA variants were not found by Sanger sequencing.

Among the two patients (BB233 and GR142) described above from whom variants carrying the PA-I38T mutation were isolated prior to drug treatment, GR142 had contact with an influenza patient who was treated with BXM in their household. The brother of patient GR142 (patient GR125, an 11-year-old boy) experienced onset of an influenza-like illness on January 31, 2019 (Extended Data Fig. 1c). Patient GR125 presented to a clinic on the same day and received BXM treatment. Biphasic fever was observed in this patient. His clinical sample (GR125–0), collected prior to drug treatment, was found to contain A/H3N2 viruses encoding PA-38I. Patient GR142 (a 3-year-old girl) developed influenza-like symptoms on February 7, and was seen by a physician on February 8. Her clinical sample (GR142–0) collected before drug treatment contained A/H3N2 variants carrying the PA-I38T mutation. Specimens were also inoculated into hCK, a Madin-Darby canine kidney (MDCK) cell line that expresses high levels of human-type influenza virus receptors (i.e., α2,6-sialoglycans) and low levels of avian-type virus receptors (i.e., α2,3-sialoglycans), for virus isolation⁶. Sequence analysis of the isolates recovered from samples GR125–0 (a pre-treatment sample from patient GR125) and GR142–0 (a pre-treatment sample from patient GR142) revealed no nucleotide differences in any of the eight gene segments, with the exception of a difference at residue 113 within the codon of amino acid residue 38 of PA. The Influenza Research Database (https://www.fludb.org) analysis by Takashita et al.⁴ showed that the I38T mutation was not detected among 17,227 A/H3N2 viruses until December 2018. In addition, analysis of the PA sequence of A/H3N2 viruses isolated in Japan during the 2017–2018 season and in the United States during the 2016–2017 and 2017–2018 seasons before the licensure of BXM revealed that the mutation was not detected in these A/H3N2 isolates^7,8. Taken together, these epidemiologic and virologic data indicate that person-to-person transmission of PA-I38T variants occurred within this family.

Consistent with previous surveillance studies during the 2018–2019 season in Japan^4,5, our data also demonstrate that seasonal influenza A viruses possessing an I38T mutation in their PA, which confers reduced susceptibility to BXA, can emerge during treatment with this drug and that such variants appear to be transmissible from person to person.

To assess the impact of the PA-I38T substitution on viral replication and the pathogenicity of seasonal influenza A virus clinical isolates, we characterized four influenza A viruses harboring a PA-I38T mutation that were recovered from four patients (KK001, KK003, KK015, and GR117) who received BXM: A/Isumi/UT-KK001–3/2018 (H1N1pdm; MUT-KK001-I38T), A/Isumi/UT-KK003–4/2018 (H1N1pdm; MUT-KK003-I38T), A/Isumi/UT-KK015–3/2019 (H3N2; MUT-KK015-I38T), and A/Tokyo/UT-GR117–5/2019 (H3N2; MUT-GR117-I38T) (Supplementary Table 5). For comparison, we also characterized the wild-type parent viruses encoding PA-38I from the same patients: A/Isumi/UT-KK001–1/2018 (H1N1pdm; WT-KK001-I38), A/Isumi/UT-KK003–0/2018 (H1N1pdm; WT-KK003-I38), A/Isumi/UT-KK015–0/2019 (H3N2; WT-KK015-I38), and A/Tokyo/UT-GR117–0/2019 (H3N2; WT-GR117-I38). These isolates were propagated in hCK cells to prepare virus stocks. Because the MUT-KK015-I38T (H3N2) variant had a mixed viral population encoding either T or I at position 38 of PA, the variant in the culture supernatant was plaque-purified and amplified on hCK cells for stock virus preparation. Sequencing analysis of the stock viruses revealed that the PA-I38T mutation was maintained in all four of the variants [MUT-KK001-I38T (H1N1pdm), MUT-KK003-I38T (H1N1pdm), MUT-KK015-I38T (H3N2), and MUT-GR117-I38T (H3N2)] (Supplementary Table 6). Compared with the wild-type virus [WT-KK001-I38 (H1N1pdm)], the MUT-KK001-I38T (H1N1pdm) variant had a mixed viral population encoding either V or I at position 116 of NA (N2 numbering), in addition to the PA-I38T mutation. The other A/H1N1pdm variant (MUT-KK003-I38T) encoded isoleucine at position 621 of PA, whereas the wild-type virus WT-KK003-I38 (H1N1pdm) possessed a mixed population encoding PA-621S or PA-621I. One A/H3N2 variant (MUT-KK015-I38T) contained glutamic acid at position 390 of HA (H3 numbering); however, a mixed population encoding HA-390E or HA-390K was found in the wild-type virus, WT-KK015-I38 (H3N2).

The susceptibilities of the PA-I38T variant isolates to BXA were first assessed by means of a plaque reduction assay on hCK cells (Supplementary Table 7). We confirmed that the MUT-KK001-I38T (H1N1pdm), MUT-KK003-I38T (H1N1pdm), MUT-KK015-I38T (H3N2), and MUT-GR117-I38T (H3N2) variants exhibited decreased BXA susceptibility of >107-, 67-, 233-, and 71-fold, respectively, compared with their pre-treatment wild-type viruses. The efficiencies of replication of the wild-type and mutant viruses were then compared in a multiple-step growth cycle in hCK cells. The four variant isolates grew to titers comparable to those typically obtained for each wild-type parent virus in hCK cells (Fig 1a). In contrast, previous studies demonstrated that recombinant influenza A viruses bearing an I38T substitution in PA, in the background of the A/WSN/33 (H1N1) or A/Victoria/2/75 (H3N2) strain, were attenuated in cell culture^1,2. Therefore, replication of these variants may be improved by compensatory mutation(s). To examine this possibility, we generated recombinant A/H1N1pdm and A/H3N2 viruses encoding either I or T at position 38 in PA in the background of the consensus sequence of each wild-type virus, and analyzed their growth kinetics in hCK cells (Fig 1b). One A/H3N2 recombinant virus encoding PA-38T (rMUT-KK015-I38T) replicated more rapidly than the wild-type virus. In contrast, two A/H1N1 viruses and another A/H3N2 recombinant virus encoding PA-38T [(rMUT-KK001-I38T (H1N1pdm), rMUT-KK003-I38T (H1N1pdm), and rMUT-GR117-I38T (H3N2)] showed delayed growth kinetics, suggesting that the viral replication fitness of these original clinical isolates may be restored by a concomitant compensatory mutation(s). To gain further insight into the biological fitness of influenza A viruses encoding PA-I38T in vitro, we conducted a growth competition experiment (Fig 1c). Each of the four recombinant viruses encoding PA-38T were mixed with its wild-type counterpart at an equal ratio based on virus titers, and the virus mixture was inoculated into hCK cells. Sequence analysis of individual virus clones (n= 44–50 clones per sample) in the cell culture supernatant revealed that the proportion of the two A/H1N1 recombinant viruses encoding PA-38T was markedly reduced at 48 h post-infection. In contrast, the proportion of the two A/H3N2 recombinant viruses encoding PA-38T increased at 48 h post-infection, although rMUT-GR117-I38T displayed reduced growth in the growth kinetics analysis. Overall, these results suggest that the I38T substitution in PA compromises the replication fitness of these A/H1N1pdm viruses in vitro, and that replication of A/H1N1pdm variants with the mutation could be improved by a compensatory mutation(s). In contrast, the PA-I38T mutation does not have a detrimental effect on the fitness of these A/H3N2 viruses in vitro.

Figure 1. in vitro characterization of four pairs of influenza A/H1N1pdm and A/H3N2 viruses with and without the PA-I38T substitution.

a,b, Growth kinetics of viruses in hCK cells. hCK cells were infected with clinical isolates (a) or with recombinant viruses (b) at a multiplicity of infection (MOI) of 0.001. The supernatants of the infected cells were harvested at the indicated times, and virus titers were determined by means of plaque assays on hCK cells. Data are shown as the mean (± standard deviation, SD) of three independent experiments. Error bars indicate SDs. P-values were calculated by using pairwise comparisons after a linear mixed model analysis (*P < 0.05; **P < 0.01). Asterisks indicate statistically significant differences between mutant and wild-type viruses. See Methods for the details of the statistical analysis. c, Growth competition assays between wild-type and PA-I38T mutant viruses. Recombinant wild-type and PA-I38T mutant viruses were mixed at an equal ratio based on their titers, and the virus mixture was inoculated into hCK cells at an MOI of 0.001. Forty-eight hours later, the virus-containing supernatant of the infected cells was harvested. The resulting viruses were purified by plaque cloning, amplified on hCK cells, and the PA genes were analyzed by RT-PCR and sequenced. The relative proportions of wild-type and PA-I38T mutant viruses in the mixture (input) and the supernatants at 48 h post-infection were determined by use of Sanger sequencing. “n” denotes the number of clones examined.

Next, we evaluated the replication and pathogenicity of the PA-I38T variant isolates in Syrian hamsters, which are highly susceptible to seasonal influenza viruses, including the recent A/H3N2 viruses⁹. Syrian hamsters were intranasally infected with 10⁶ plaque-forming units (PFU) of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T (H1N1pdm), WT-KK003-I38 (H1N1pdm), MUT-KK003-I38T (H1N1pdm), WT-KK015-I38 (H3N2), or MUT-KK015-I38T (H3N2), or with 10^5.2 PFU of WT-GR117-I38 (H3N2) or MUT-GR117-I38T (H3N2). The body weight of mock-infected Syrian hamsters gradually increased over the 6-day experiment (Fig. 2a). Syrian hamsters infected with MUT-KK001-I38T (H1N1pdm) experienced slight weight loss as early as 1 day post-infection, and the animals showed a maximum of 4% weight loss by 2 days post-infection. By contrast, none of the animals infected with WT-KK001-I38 (H1N1pdm) showed any weight loss. A very slight decrease in body weight was observed for most of the animals infected with WT-KK003-I38 (H1N1pdm) or MUT-KK003-I38T (H1N1pdm) at 2 days post-infection, but no substantial differences in weight changes were found between the two groups. The Syrian hamsters infected with the two A/H3N2 virus pairs showed increased body weights. No differences in weight changes were observed between the wild-type virus WT-KK015-I38 (H3N2)- and the MUT-KK015-I38T (H3N2) variant-infected groups. In contrast, the increase in body weight was significantly slower in the MUT-GR117-I38T (H3N2) variant-infected group than in the wild-type virus WT-GR117-I38 (H3N2)-infected group. We observed a statistically significant difference in weight change between the WT-KK001-I38 (H1N1pdm)- and MUT-KK001-I38T (H1N1pdm)-infected animals in an additional experiment with Syrian hamsters (Extended Data Fig. 2a).

Figure 2. in vivo characterization of four pairs of influenza A/H1N1pdm and A/H3N2 viruses with and without the PA-I38T substitution.

a, Body weight changes in Syrian hamsters after viral infection. Syrian hamsters were intranasally inoculated with 10⁶ PFU of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T (H1N1pdm), WT-KK003-I38 (H1N1pdm), MUT-KK003-I38T (H1N1pdm), WT-KK015-I38 (H3N2), or MUT-KK015-I38T (H3N2), or 10^5.2 PFU of WT-GR117-I38 (H3N2) or MUT-GR117-I38T (H3N2), or MEM containing 0.3% BSA (mock). Body weights of virus-infected (n=5) and mock-infected hamsters (n=4) were monitored daily for 6 days. Data are presented as the mean percentages of the starting weight (± standard deviation, SD). P-values were calculated by using pairwise comparisons after a linear mixed model analysis (*P < 0.05; **P < 0.01). Asterisks next to data points depict statistically significant differences between mutant and wild-type viruses. See Methods for more details regarding the statistical analysis. b, Virus replication in infected Syrian hamsters. Syrian hamsters were infected intranasally with 10⁶ PFU of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T (H1N1pdm), WT-KK003-I38 (H1N1pdm), MUT-KK003-I38T (H1N1pdm), WT-KK015-I38 (H3N2), or MUT-KK015-I38T (H3N2), or 10^5.2 PFU of WT-GR117-I38 (H3N2) or MUT-GR117-I38T (H3N2). Three Syrian hamsters per group were euthanized on Days 3 and 6 post-infection for virus titration. Virus titers in nasal turbinates, trachea, and lung were determined by use of a plaque assay on hCK cells. Vertical bars show the mean. The vertical bar is shown only when virus was recovered from all three hamsters. Points indicate data from individual Syrian hamsters. NT, nasal turbinate; R cra/acce, right cranial and accessory lobes; R middle, right middle lobe; R caudal, right caudal lobe; L, left lobe. c, Respiratory droplet transmission among ferrets. Three or five ferrets were infected with 10⁶ PFU of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T (H1N1pdm), WT-GR117-I38 (H3N2), or MUT-GR117-I38T (H3N2). One day later, three or five naïve ferrets (exposed ferrets) were each placed in a cage adjacent to an infected ferret. Nasal washes were collected from infected ferrets on day 1 after inoculation and from exposed ferrets on day 1 after co-housing, and then every other day (for up to 11 days) for virus titration. The lower limit of detection is indicated by the horizontal dashed line.

Two A/H1N1pdm variants (MUT-KK001-I38T and MUT-KK003-I38T) replicated efficiently in the lungs and other respiratory organs of the infected animals, similar to their wild-type counterparts (Fig. 2b). The two A/H3N2 variants (MUT-KK015-I38T and MUT-GR117-I38T) replicated well in the nasal turbinates, trachea, and lungs on Day 3 post-infection, although no virus was recovered from the lungs of one of three MUT-GR117-I38T-infected animals on Day 3 post-infection or from the lungs of all of three MUT-KK015-I38T- or MUT-GR117-I38T-infected animals on Day 6 post-infection (Fig. 2b). For any of the four virus pairs, no substantial difference in viral replication was observed between the mutant and wild-type viruses. We confirmed that the I38T mutation in PA was retained in the viruses recovered from the lungs of MUT-KK001-I38T (H1N1pdm)- and MUT-KK003-I38T (H1N1pdm)-infected animals on Days 3 and 6 post-infection, and from the lungs of MUT-KK015-I38T (H3N2)- and MUT-GR117-I38T (H3N2)-infected animals on Day 3 post-infection. Pathological examination revealed that there were no differences in the histological changes in the lungs of the animals infected with the mutant or the wild-type viruses (Extended Data Fig. 2b).

To further investigate the replication and virulence of the PA-I38T mutant isolates in vivo, we inoculated mice with the mutant viruses. Because mice are not susceptible to recent seasonal A/H3N2 viruses, we analyzed only the A/H1N1pdm viruses. Both variants, MUT-KK001-I38T (H1N1pdm) and MUT-KK003-I38T (H1N1pdm), replicated efficiently in the nasal turbinates and lungs of BALB/c mice or DBA/2 mice (Extended Data Fig. 3), and there was no difference in replication between the mutant and wild-type viruses. In BALB/c mice, MUT-KK001-I38T (H1N1pdm) caused significantly greater weight loss than the wild-type WT-KK001-I38 (H1N1pdm) virus (Extended Data Fig. 4a), consistent with our observations in hamsters infected with these viruses. A similar trend was observed for DBA/2 mice infected with these viruses. BALB/c and DBA/2 mice infected with WT-KK003-I38 (H1N1pdm) or MUT-KK003-I38T (H1N1pdm) exhibited comparable weight loss patterns. Infection of DBA/2 mice with the variants yielded a slightly shorter mean time to death compared with infection of these mice with the wild-type counterparts (Extended Data Fig. 4b).

Ferrets inoculated with MUT-KK001-I38T (H1N1pdm) showed substantial weight loss, similar to infection with the wild-type counterpart (Extended Data Fig. 5a). The maximum mean weight loss was 11.2% and 9.6% for the wild-type and mutant viruses, respectively. In contrast, the animals infected with MUT-GR117-I38T (H3N2) exhibited increased body weights, similar to the wild-type virus. For both of the two virus pairs, no marked differences in viral titers in the nasal washes were observed between the mutant and wild-type viruses on Days 1, 3, and 5 after infection (Fig. 2c).

Overall, these findings indicate that the virulence in hamsters, mice, and/or ferrets of A/H1N1pdm and A/H3N2 variants with reduced susceptibility to BXA is similar to that of the wild-type viruses.

To assess the respiratory droplet transmissibility of the seasonal influenza A viruses with reduced susceptibility to BXA, we set up three or five transmission pairs, each comprising a naïve ferret housed adjacent to an infected ferret one day after infection, as described previously¹⁰. Both WT-KK001-I38 (H1N1pdm) and MUT-KK001-I38T (H1N1pdm) were efficiently transmitted via respiratory droplets to all three exposed ferrets of each group, as clearly shown by the detection of virus in the nasal washes of these animals (Fig. 2c). Two H3N2 viruses, WT-GR117-I38 and MUT-GR117-I38T, were transmitted to all and four of five exposed ferrets, respectively. Neither virus shedding nor seroconversion was detected in one animal exposed to MUT-GR117-I38T (Supplementary Table 8). The exposed animals infected with WT-KK001-I38 (H1N1pdm) or MUT-KK001-I38T (H1N1pdm) exhibited weight loss (maximum mean weight loss of 11.5% and 8.9%, respectively) (Extended Data Fig. 5b). The exposed animals infected with WT-GR117-I38 (H3N2) or MUT-GR117-I38T (H3N2) showed increased body weights. We confirmed that the I38T mutation in PA was retained in the viruses recovered from the nasal washes of MUT-KK001-I38T (H1N1pdm)- and MUT-GR117-I38T (H3N2)-infected animals on Day 5 post-infection, and from the nasal washes of the exposed animals on Day 5 post-exposure. These results indicate that the transmissibility in ferrets of A/H1N1pdm and A/H3N2 variants with reduced susceptibility to BXA is similar to that of the wild-type viruses, suggesting that they have the potential to disseminate.

In summary, our data suggest that influenza A/H1N1pdm and A/H3N2 viruses circulating in humans can rapidly acquire an I38T mutation in their PA, which confers reduced susceptibility to BXA, without a loss of viral fitness (see Supplementary Discussion). It is therefore possible that widespread use of BXM will result in the circulation of influenza A viruses with this mutation. The proper use of this drug and continued close monitoring for the emergence or prevalence of seasonal influenza A virus PA-I38T variants is extremely important.

Methods

Cells.

hCK cells⁶ were maintained in the presence of 2 μg/ml puromycin and 10 μg/ml blasticidin in Eagle’s minimal essential media (MEM) containing 5% newborn calf serum. Human embryonic kidney 293T cells were maintained in Dulbecco’s modified Eagle’s medium containing 10% fetal calf serum. All cells were incubated at 37 °C with 5% CO₂, and regularly tested for mycoplasma contamination by using PCR and were confirmed to be mycoplasma-free.

Antiviral compounds.

Baloxavir acid was purchased from MedChemExpress (Monmouth Junction, NJ).

Clinical specimens.

After informed consent was obtained, respiratory specimens were obtained from patients with influenza-like symptoms who visited clinics in Japan during the 2018–2019 season. For pediatric patients, informed consent was obtained from the parents. The specimens were submitted to the Division of Virology, Department of Microbiology and Immunology, Institute of Medical Science, the University of Tokyo for virus isolation. The research protocol was approved by the Research Ethics Review Committee of the Institute of Medical Science of the University of Tokyo (approval no. 26–42-0822). Samples that were positive by real-time RT-PCR (see below) were used in this study.

Viruses.

Influenza viruses were propagated in hCK cells in MEM containing 1 μg of L-1-Tosylamide-2-phenylethyl chloromethyl ketone (TPCK)-trypsin/ml at 33 °C.

Reverse genetics.

Plasmid-based reverse genetics for influenza virus generation was performed as previously described¹¹. In brief, plasmids encoding the complementary DNAs for the eight viral RNA segments under the control of the human RNA polymerase I promoter and the mouse RNA polymerase I terminator (referred to as PolI plasmids), and plasmids for the expression of the viral PB2, PB1, PA and nucleoprotein proteins derived from a laboratory-adapted influenza A virus strain A/Puerto Rico/8/34 (H1N1), under the control of the chicken β-actin promoter¹², were transfected into 293T cells with the help of a transfection reagent, Trans-IT 293 (Mirus). At 48 h post-transfection, culture supernatants were collected and inoculated to hCK cells for virus propagation. All virus stocks were sequenced to confirm the absence of unwanted mutations.

Real-time RT-PCR.

RNA was extracted from clinical specimens by using the Simply RNA Tissue Kit (Promega). Amplification and detection by real-time PCR were performed with the LightCycler 96 System (Roche). RT-PCR was carried out using the QuantiTect Probe RT-PCR Kit (Qiagen). The probes contained oligonucleotides with the 6-carboxyfluorescein (FAM) or the hexacholoro-6-carboxyfluorescein (HEX) reporter dye at the 5′ end, and the Black Hole Quencher-1 (BHQ-1) quencher dye at the 3′ end. A list of the primers and probes used is provided in Supplementary Table 9.

Virus isolation.

hCK cells grown in 12-well plates were inoculated with 0.1 ml per well of the clinical samples and incubated at 33 °C for at least 30 min. One milliliter of MEM containing 0.3% bovine serum albumin (BSA) and 1 μg/ml TPCK-treated trypsin was then added to cells. The cultures were then incubated for up to 7 days, until cytopathic effects were evident. Cell culture supernatants were harvested, viral RNA was extracted and subjected to RT-PCR, and the viral genes were sequenced (see below).

RT-PCR and sequencing of viral genes.

Viral RNA was extracted from 140 μl of culture supernatant by using the QIAamp Viral RNA Mini kit (Qiagen). Samples were amplified using the QIAGEN OneStep RT-PCR Kit (Qiagen) and specific primers of the PA gene. PCR products were then analyzed by means of 1.5% agarose gel electrophoresis in tris-buffer, and target bands were visualized by staining with GelRed (Biotium). The PCR products were purified and subjected to direct sequencing. We estimated the level of mutation frequencies based on the height of the waves at each position on the sequencing chromatogram. The detection limit for a minor population was 10%–20%. The list of primers used is provided in Supplementary Table 9.

Deep sequencing analysis.

Viral RNA was extracted from samples using the QIAamp Viral RNA Mini kit (Qiagen), according to the manufacturer’s instructions. The PA gene was amplified using the One-Step SuperScript III RT-PCR kit (Invitrogen) and PA segment-specific primers (see Supplementary Table 9 for primer pairs). PCR amplicons were cleaned with 0.45X Agencourt AMPure XP Magnetic Beads (Beckman Coulter) according to the manufacturer’s protocol. The concentration of purified amplicons was measured using the Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen). After samples were normalized to a concentration of 0.2 ng/μl, adapters were added by tagmentation using the Nextera XT DNA library preparation kit (Illumina). Samples were purified by using 0.6X Agencourt AMPure XP Magnetic Beads and fragment size distributions were analyzed on a Bioanalyzer using the High Sensitivity DNA kit (Agilent). After bead-based normalization (Illumina) according to the manufacturer’s protocols, sequence-ready libraries were sequenced in a paired-end run using the MiSeq v2, 300 cycle reagent kit (Illumina).

Computational analysis of deep sequencing data.

The raw FASTQ format data files obtained from the Illumina sequencing samples were first filtered with cutadapt¹³ to remove low-quality reads and adapters. We used the parameter “--match-read-wildcards -e 0.1 -O 6 -m 32”. Next, we used BWA mem¹⁴, to align the reads to the nucleotide sequence of the influenza virus PA gene, using either A/Isumi/UT-KK001–1/2018 (H1N1pdm) or A/Tokyo/GR104–0/2019 (H3N2) viruses. Finally, we used a custom Python script to determine the ratio of wild-type to mutant codons at position 38 of PA. A cutoff of 5% was used for a variant frequency if the sequence coverage was more than 1000 reads at the position where the variant was found, and 10% if the coverage was between 100 and 1000.

Growth competition experiment.

Recombinant wild-type and PA-I38T mutant viruses were mixed at an equal ratio based on their PFU titers, and the virus mixture was inoculated into hCK cells in T75 tissue culture flasks at an MOI of 0.001. Forty-eight hours later, the virus-containing supernatant of the infected cells was harvested. The resulting viruses were purified by plaque cloning, and amplified on hCK cells. For each sample, at least 44 plaques were randomly picked and their PA genes were analyzed by RT-PCR and sequenced, as described above.

Plaque reduction assay.

Confluent hCK cells in 6-well or 12-well plates were infected with a dilution of virus that resulted in 30–80 virus plaques per well. After a 1-h incubation, the viral inoculum was removed and the cells were overlaid with 1% agarose-containing MEM containing 0.3% BSA in the presence of TPCK-treated trypsin and different concentrations of baloxavir acid. The plates were incubated for 2–3 days; then, the agar overlaid with formalin to fix the cells was removed. Plaques were counted, and EC₅₀ values were calculated by using Graphpad Prism (GraphPad Software Inc. La Jolla, CA).

Virus neutralization assay.

Viral neutralization assays were performed by using the methodology outlined in the WHO Manual on Animal Influenza Diagnosis and Surveillance with the following modifications. Sera were treated with receptor-destroying enzyme (RDE; Denka Seiken Co., Ltd) at 37 °C for 18–20 h, followed by RDE inactivation at 56 °C for 30–60 min. Twenty-five microliters of virus (100 tissue culture infectious dose 50) was incubated with 25 μl of two-fold serial dilutions of RDE-treated sera for 30 min at room temperature, and the mixtures were added to confluent hCK cells in 96-well microplates, and incubated for 1 h at 33 °C. After the inoculum was removed, the cells were incubated with MEM containing 0.3% BSA and 1.0 μg/ml TPCK-trypsin at 33 °C for 48–72 h. Viral cytopathic effects were observed under an inverted microscope and virus neutralization titers were calculated as described in the WHO manual.

Animal experiments.

The sample sizes (n=3) for the hamster, mouse, and ferret studies were chosen because they have previously been shown to be sufficient to evaluate a significant difference among groups^15,16,17. No method of randomization was used to determine how the animals were allocated to the experimental groups and processed in this study. The investigator was not blinded to the group allocation during the experiments or when assessing the outcome.

Experimental infection of Syrian hamsters and mice.

Four- to six-week-old female Syrian hamsters and six-week-old female BALB/c and DBA/2 mice (Japan SLC Inc., Shizuoka, Japan) were used in this study. Baseline body weights were measured before infection. Under Ketamine-Xylazine and isoflurane anesthesia for hamsters and mice, respectively, three or five hamsters or five mice per group were intranasally inoculated with 10⁶ PFU (100 μl for hamsters; 50 μl for mice) of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T (H1N1pdm), WT-KK003-I38 (H1N1pdm), MUT-KK003-I38T (H1N1pdm), WT-KK015-I38 (H3N2) or MUT-KK015-I38T (H3N2), or with 10^5.2 PFU (100 μl for hamsters) of WT-GR117-I38 (H3N2) or MUT-GR117-I38T (H3N2). Body weight and survival were monitored daily for 6 and 11 days for hamsters and mice, respectively. For virological examinations, three hamsters or three mice per group were intranasally infected with 10⁶ PFU or 10^5.2 PFU of the viruses; 3 and 6 days post-infection, the animals were euthanized and nasal turbinates, tracheas (only for hamsters), and lungs were collected. The virus titers in the various organs were determined by means of plaque assays in hCK cells. The detection limit was 10^1.3 PFU/g. All experiments with Syrian hamsters and mice were performed in accordance with the University of Tokyo’s Regulations for Animal Care and Use and approved by the Animal Experiment Committee of the Institute of Medical Science, the University of Tokyo.

Pathological examination.

Excised tissues of animal organs were fixed in 4% paraformaldehyde phosphate buffer solution for 48 h and processed for paraffin embedding. The paraffin blocks were cut into 3-μm-thick sections and then mounted on silane-coated glass slides. One section from each tissue sample was stained using a standard hematoxylin and eosin procedure; another was processed for immunohistochemical staining with a rabbit polyclonal antibody for type A influenza nucleoprotein antigen (prepared in the Department of Pathology, National Institute of Infectious Diseases) that reacts comparably with all of the viruses tested in this study¹⁸. Specific antigen-antibody reactions were visualized by means of 3,3’-diaminobenzidine tetrahydrochloride staining using the Dako Envision system (Dako Cytomation).

Ferret transmission study.

Three- to five-month-old female ferrets (Wuxi Sangosho Biotechnology Co., Ltd., Wuxi, China), which were serologically negative by an hemagglutination inhibition assay for currently circulating human influenza viruses, were used in this study. Pairs of ferrets were individually housed in adjacent wireframe cages that prevented direct and indirect contact between animals but allowed spread of influenza virus by respiratory droplets. Three or five animals per group were anesthetized intramuscularly with ketamine and xylazine (5–30 mg and 0.2–6 mg/kg of body weight, respectively), and inoculated intranasally with 10⁶ PFU (500 μl) of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T, WT-GR117-I38 (H3N2), or MUT-GR117-I38T (H3N2) (inoculated ferrets). One day after infection, three or five naive ferrets (exposed ferrets) were each placed in a cage adjacent to an infected ferret (in these cages, infected and exposed ferrets are separated by ~5 cm). The naïve ferrets were exposed to the inoculated ferrets for 16 days. Nasal washes were collected from infected ferrets on day 1 after inoculation and from exposed ferrets on day 1 after co-housing, and then every other day (for up to 11 days) for virological examinations. The virus titers in nasal washes were determined by means of plaque assays in hCK cells. All experiments with ferrets were performed in accordance with the University of Tokyo’s Regulations for Animal Care and Use and were approved by the Animal Experiment Committee of the Institute of Medical Science, the University of Tokyo.

Statistical analysis.

Data are expressed as the mean ± SD. For the analysis of the growth curve data and the body weight data, we transformed the virus titer values and the body weight values to the log₁₀ scale and performed a linear mixed effects analysis. As fixed effects, we used the different groups and the time of the measurement (with an interaction term between those fixed effects). As random effects, we had intercepts for the individual replicates and the animals. We performed all pairwise comparisons using lsmeans¹⁹, and adjusted the p-values using Holm’s method. We considered the differences significant if the p-values were less than 0.05. We used the R statistical package (www.r-project.org), lme4²⁰, and the lsmeans package¹⁹ for the analyses.

Extended Data

Extended Data Fig. 1. Detection of influenza A/H1pdm and A/H3 virus variants carrying a mutation at amino acid position 38 of PA in patients.

a,b, Respiratory tract specimens were obtained from pediatric (aged 0 to 15 years) and adult (16 years or older) patients before and after baloxavir marboxil therapy and from untreated patients with influenza A/H1pdm (a) and A/H3 (b) virus infections in Japan during the 2018–2019 influenza season. The nucleotide sequences of the PA genes of the viruses in the specimens were determined by means of Sanger sequencing. c, Timeline of the clinical course of the two patients (GR125 and GR142) with influenza A/H3N2 virus infection.

Extended Data Fig. 2. Body weight changes and histopathological findings in infected Syrian hamsters.

a, Syrian hamsters were intranasally inoculated with 10⁶ PFU of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T (H1N1pdm), WT-KK003-I38 (H1N1pdm), or MUT-KK003-I38T (H1N1pdm), or MEM containing 0.3% BSA (mock). Body weights of virus-infected (n=3) and mock-infected animals (n=4) were monitored daily for 6 days. Data are presented as the mean percentages of the starting weight ± SD. P-values were calculated by using pairwise comparisons after a linear mixed model analysis (*P < 0.05; **P < 0.01). Asterisks next to data points depict statistically significant differences between mutant and wild-type viruses. See Methods for more details regarding the statistical analysis. b, Representative pathological images of lungs infected with WT-KK001-I38 (H1N1pdm, n=2), MUT-KK001-I38T (H1N1pdm, n=2), WT-KK003-I38 (H1N1pdm, n=2), MUT-KK003-I38T (H1N1pdm, n=2), WT-KK015-I38 (H3N2, n=2), MUT-KK015-I38T (H3N2, n=2), WT-GR117-I38 (H3N2, n=2), or MUT-GR117-I38T (H3N2, n=2) on Day 6 post-infection. Left panels, hematoxylin and eosin (HE) staining. Scale bars, 200 μm. Right panels, immunohistochemistry (IHC) for influenza viral antigen detection. Scale bars, 100 μm.

Extended Data Fig. 3. Virus titers in respiratory organs of infected mice.

Three mice per group were intranasally inoculated with 10⁶ PFU of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T (H1N1pdm), WT-KK003-I38 (H1N1pdm), or MUT-KK003-I38T (H1N1pdm). Three mice per group were euthanized on Days 3 and 6 post-infection for virus titration. Virus titers in the nasal turbinates and lungs were determined by use of a plaque assay on hCK cells. Vertical bars show the mean. Points indicate data from individual mice.

Extended Data Fig. 4. Virulence in mice.

Five mice per group were intranasally inoculated with 10⁶ PFU of WT-KK001-I38 (H1N1pdm), MUT-KK001-I38T (H1N1pdm), WT-KK003-I38 (H1N1pdm), or MUT-KK003-I38T (H1N1pdm), or MEM containing 0.3% BSA (mock). Body weights (a) and survival (b) were monitored daily for 11 days. The values for body weights are presented as the mean percentages of the starting weight ± SD. P-values were calculated by using pairwise comparisons after a linear mixed model analysis (*P < 0.05; **P < 0.01). Asterisks next to data points depict statistically significant differences between mutant and wild-type viruses. See Methods for more details regarding the statistical analysis.

Extended Data Fig. 5. Body weight changes of inoculated and exposed ferrets during the transmission study.

Ferrets were intranasally inoculated with 10⁶ PFU of WT-KK001-I38 (H1N1pdm, n=3), MUT-KK001-I38T (H1N1pdm, n=3), WT-GR117-I38 (H3N2, n=5), or MUT-GR117-I38T (H3N2, n=5). One day later, naïve ferrets (exposed ferrets, n=3, n= 3, n=5, and n=5 for WT-KK001-I38, MUT-KK001-I38T, WT-GR117-I38, and MUT-GR117-I38T, respectively) were each placed in a cage adjacent to an infected ferret. a, Body weights of inoculated ferrets were measured on Days 0, 1, 3, 5, 7, 9, and 11 post-infection. b, Body weights of exposed ferrets were measured on Days 0, 3, 5, 7, 9, 11, and 13 post-exposure. Data are presented as the mean percentages of the starting weight ± SD.

Data availability.

Numerical source data underlying the graphs shown in figures 1 and 2 and Extended Data figures 2, 3, 4, and 5 are associated with each figure. The data that support the findings of this study are available from the corresponding author upon request. Samples were deposited in the Sequence Read Archive of the NCBI, under project accession PRJNA573567.