Global update on the susceptibilities of human influenza viruses to neuraminidase inhibitors and the cap-dependent endonuclease inhibitor baloxavir, 2018–2020

Global analysis of the susceptibility of influenza viruses to neuraminidase (NA) inhibitors (NAIs) and the polymerase acidic (PA) inhibitor (PAI) baloxavir was conducted by five World Health Organization Collaborating Centres for Reference and Research on Influenza during two periods (May 2018–May 2019 and May 2019–May 2020). Combined phenotypic and NA sequence-based analysis revealed that the global frequency of viruses displaying reduced or highly reduced inhibition (RI or HRI) or potential to show RI/HRI by NAIs remained low, 0.5% (165/35045) and 0.6% (159/26010) for the 2018–2019 and 2019–2020 periods, respectively. The most common amino acid substitution was NA-H275Y (N1 numbering) conferring HRI by oseltamivir and peramivir in A(H1N1)pdm09 viruses. Combined phenotypic and PA sequence-based analysis showed that the global frequency of viruses showing reduced susceptibility to baloxavir or carrying substitutions associated with reduced susceptibility was low, 0.5% (72/15906) and 0.1% (18/15692) for the 2018–2019 and 2019–2020 periods, respectively. Most (n = 61) of these viruses had I38→T/F/M/S/L/V PA amino acid substitutions. In Japan, where baloxavir use was highest, the rate was 4.5% (41/919) in the 2018–2019 period and most of the viruses (n = 32) had PA-I38T. Zoonotic viruses isolated from humans (n = 32) in different countries did not contain substitutions in NA associated with NAI RI/HRI phenotypes. One A(H5N6) virus had a dual substitution PA-I38V + PA-E199G, which may reduce susceptibility to baloxavir. Therefore, NAIs and baloxavir remain appropriate choices for the treatment of influenza virus infections, but close monitoring of antiviral susceptibility is warranted.


Introduction
Antivirals known as neuraminidase (NA) inhibitors (NAIs) and polymerase acidic (PA) inhibitor (PAI) have been approved in many countries for controlling influenza infections. The four NAIs in current use are oral oseltamivir, inhaled zanamivir, intravenous peramivir, and inhaled laninamivir (approved only in Japan). The orally administered PAI baloxavir marboxil (baloxavir) is the first in a new class of drugs targeting the cap-dependent endonuclease activity of the influenza PA protein. This antiviral was approved in Japan and the United States in 2018 and is now approved in several countries (Beigel and Hayden, 2021;Ison et al., 2021). Antiviral usage against influenza differs among countries, with the highest usage per capita being reported in Japan (Table S1). Oseltamivir remains the most widely used influenza antiviral and is now available as a generic drug.
The emergence and spread of viruses with reduced susceptibility to antivirals can diminish the usefulness of these drugs for controlling influenza infections. This necessitates continuous monitoring of antiviral susceptibility among circulating viruses. Antiviral testing has become an integral part of influenza virologic surveillance conducted by laboratories of the World Health Organization (WHO) Global Influenza Surveillance and Response System (GISRS). Members of the WHO GISRS Antiviral Working Group (WHO-AVWG) compile and review data collected throughout the year and provide regular reports Hurt et al., 2016;Lackenby et al., 2018;Meijer et al., 2014;Takashita et al., 2015aTakashita et al., , 2020a.
NAI susceptibility is assessed using the MUNANA (4-(methyl-umbelliferyl)-Nacetylneuraminic acid) fluorescence-based NA enzyme inhibition assay, with minor modifications in each laboratory, to determine 50% inhibitory concentrations (IC 50 ). Tokyo WHO Collaborating Centre (CC) uses a chemiluminescent-based assay (NA-XTD ™ Influenza Neuraminidase Assay Kit, Applied Biosystems) to assess a small subset of viruses with low NA activity. As there are no established cut-offs defining drug resistance, and to harmonize phenotypic data across WHO CCs, the WHO-AVWG criteria for reporting NAI susceptibility data are based on comparison to the median IC 50 value of the respective type/subtype/lineage (World Health Organization, 2012). These criteria differ for influenza A and B viruses: normal inhibition (NI) (<10-fold for type A viruses; <5-fold for type B viruses), reduced inhibition (RI) (10-to 100-fold for type A viruses; 5-to 50-fold for type B viruses), and highly reduced inhibition (HRI) (>100-fold for type A viruses; >50-fold for type B viruses). RI or HRI phenotypes of influenza viruses are associated with amino acid substitutions or deletions at conserved residues that form/stabilize the active site of the NA . Overall, the frequency of influenza viruses with RI/HRI in global circulation has been low (<2%). However, during 2007-2009, oseltamivir resistance rose drastically among A(H1N1) viruses, being conferred by the NA-H275Y amino acid substitution (Dharan et al., 2009;Hurt et al., 2009;Meijer et al., 2009). A(H1N1)pdm09 viruses displaced the oseltamivir-resistant A(H1N1) viruses. Since then, oseltamivir-resistant A(H1N1)pdm09 viruses containing NA-H275Y have been detected sporadically, sometimes as clusters of cases but with limited circulation in communities (Hurt et al., 2012;Takashita et al., 2015b). Besides NA-H275Y, numerous other substitutions have been associated with RI/HRI phenotypes, and these have been summarized by the WHO-AVWG (NA marker table, https://cdn.who.int/media/docs/default-source/influenza/ nai-reduced-susceptibility-marker-table-who_table-1.pdf).
With the advances in next-generation sequencing (NGS) technologies, there is increased emphasis on establishing sequence-based virologic surveillance, which can provide a framework for selecting viruses for phenotypic testing. Several years ago, the Atlanta WHO CC implemented the sequence-first initiative, whereby all submitted viruses are subjected to full-genome sequence analysis so that the generated sequences can be screened for various changes, including potential markers of reduced susceptibility to drugs. Other GISRS laboratories are increasing their sequencing capabilities and are sharing their findings via the Global Initiative on Sharing All Influenza Data (GISAID), thereby bolstering surveillance data generation.
Phenotypic testing enables the detection of viruses with reduced susceptibility to drugs and combined with gene sequencing, allows identification of known and new markers. Cell culture-based assays are used to determine the baloxavir susceptibility of influenza viruses by assessing virus replication in the presence of the drug. Two assays, the high-content imaging neutralization test (HINT) and the focus reduction assay (FRA), are currently used by the Atlanta and Tokyo WHO CCs, respectively (Gubareva et al., 2019;. These two assays yield different effective concentrations (EC 50 ) values for a given virus, but the fold changes are similar (Takashita et al., 2020a). Currently, there are no criteria for defining resistance or reduced susceptibility to baloxavir. An arbitrary threshold (cut-off) of a ≥3-fold increase in the median EC 50 is used for reporting viruses with reduced susceptibility to baloxavir (Gubareva et al., 2019;Takashita et al., 2020a). This cut-off should capture >95% of viruses with reduced susceptibility to baloxavir (Ince et al., 2020).
National Influenza Centres (NICs) receive influenza virus-positive clinical specimens collected in their respective countries and perform initial analyses (https://www.who.int/teams/global-influenza-programme/laboratorynetwork/virological-surveillance). Representative numbers of influenza-positive clinical specimens and/or viruses of each genetic type/subtype/lineage are then shipped to a GISRS WHO CC for further characterization. The WHO CCs propagate viruses in MDCK, MDCK-SIAT1, or hCK cells before drug susceptibility testing (Hurt et al., 2012;Takada et al., 2019). This is the seventh WHO-AVWG review of antiviral susceptibility, and it primarily comprises data generated by WHO CCs. It includes influenza antiviral susceptibility data for two consecutive periods, May (week 21) to May (week 20) of the following year as used in previous global reports (thereby covering Southern and subsequent Northern Hemisphere influenza seasons) for 2018-2019 and 2019-2020, with an emphasis on sequence-based analysis. Antiviral susceptibility results for NAIs and baloxavir are analyzed, keeping in mind the changing algorithms for virus testing employed by the WHO CCs. For the first time, the report includes antiviral susceptibility analysis of zoonotic viruses detected in different countries and reported to the WHO. This analysis provides data for evaluating the risk posed by viruses with pandemic potential. The emergence of the COVID-19 pandemic delayed analyses of the data presented in this report.

Analysis of phenotypic and genotypic NAI susceptibility data from WHO CCs
During 2018-2019 and 2019-2020 periods, a total of 19966 and 15582 viruses, respectively were assessed for NAI susceptibility by five WHO CCs using phenotypic and/or NA sequence-based methods (Figs. S1 and S2). Of these respective total numbers, 13536 in 2018-2019 and 9853 in 2019-2020 were tested phenotypically by an NA inhibition assays. The potential NAI susceptibility of remaining viruses 6430 in 2018-2019, 5729 in 2019-2020, with almost equal proportions of viruses collected in the United States and other countries were assessed by Atlanta CC based on NA sequence analysis to ensure that all viruses with previously reported markers were tested phenotypically ( Fig. 1A and Table S2). All these viruses (19966 and 15582, respectively for both periods) were assessed for susceptibility to oseltamivir and zanamivir (Fig. 1B). Three WHO CCs (Atlanta, Melbourne, and Tokyo) also assessed viruses for susceptibility to peramivir and laninamivir using phenotypic and/or NA sequence-based methods. Most viruses originated from the WHO regions of the Western Pacific (WPRO; 55.4% for 2018-2019 and 50.1% for 2019-2020) and the Americas (PAHO; 23.9% for 2018-2019 and 28% for 2019-2020) (Fig. 1B). Only 20.7% of viruses for 2018-2019 and 21.9% for 2019-2020, were from the African (AFRO), Eastern Mediterranean (EMRO), European (EURO), or Southeast Asian regions (SEARO) (Fig. 1B). Based on the WHO GISRS global web-based tool for influenza virologic surveillance (https://www.who.int/initiatives/ global-influenza-surveillance-and-response-system), totals of 685332 influenza viruses were detected globally and reported to FluNet for 2018-2019 and 694536 for 2019-2020, respectively. Therefore, the viruses analyzed for NAI susceptibility by the WHO CCs in this study represent 2.9% of global influenza detections reported to FluNet for 2018-2019 and 2.2% of those reported for 2019-2020.

A(H1N1)pdm09 viruses showing RI/HRI
In 2018-2019 and 2019-2020, 69/10371 (0.7%) and 64/4827 (1.3%) viruses, respectively, exhibited RI/HRI by at least one NAI ( Fig. 2A and B), indicating an increase in the incidence of RI/HRI viruses in 2019-2020. Most viruses with a RI/HRI phenotype contained the NA-H275Y substitution (n = 101, 76%) and showed the expected increases in IC 50 for oseltamivir and peramivir (Table 1 and Fig. 3A). In addition, six viruses with a NA-H275Y/H mixture showed elevated IC 50 for oseltamivir and peramivir (Table 1). These NA-H275Y variants were collected in 15 countries (Table S3). For 71/101 viruses, NA-H275Y was confirmed in the corresponding clinical specimens; no clinical specimens were available for the remaining 30 viruses (Tables 1 and S3). Of the 85 patients with available clinical history, 42 were outpatients and 43 were hospitalized (Table 1). Antiviral treatment history was available for 60/101 patients, including 29 patients who did not receive an NAI before specimen collection (Tables 1 and S3). Immunocompromised status was reported for five patients who shed NA-H275Y viruses (Table 1).
Three viruses with NA-I223K substitution showed RI by oseltamivir, and the substitution was confirmed in the two corresponding clinical specimens available (Tables 1 and S3).
A new substitution at this residue, NA-I223L, conferring a similar effect was also found in one clinical specimen/virus pairing. Two viruses recovered from patients with unknown treatment histories had either NA-I223M or NA-I223R substitution and exhibited RI by oseltamivir and/or peramivir. One virus from India had the dual substitution NA-H275Y + NA-G147R/G and displayed a 1376-fold increase in the IC 50 for oseltamivir and a 746-fold increase in the IC 50 for peramivir (Tables 1 and S3). No treatment history was available for this case, but the same dual substitution was previously reported in connection with NAI treatment . Three viruses had NA-N295S and exhibited RI/HRI by oseltamivir, with the corresponding change being confirmed in the two clinical specimens available.
One virus displayed RI by oseltamivir and zanamivir, which was conferred by the substitution NA-V116A (clinical specimen not available). This substitution was previously reported in A(H5N1) virus (Boltz et al., 2010;Hurt et al., 2007). A virus with NA-R152K showed RI by oseltamivir and zanamivir; the antiviral treatment history of this patient was unknown.
It is worth noting that several NA substitutions that confer RI/HRI, especially by zanamivir, have been linked to virus culture (NA marker table; Little et al., 2015). Among these viruses were those isolates containing NA-Q136R or NA-Q136K substitution or NA-Q136K/Q + NA-D151E/D or D151N/D mixtures (Table 1). In addition, an isolate with NA-D199G substitution displayed RI by oseltamivir, two viruses containing NA-E119K substitution (in one case combined with NA-Q136K) displayed RI by zanamivir, and one virus with NA-E119G substitution exhibited HRI by zanamivir and RI by peramivir and laninamivir (Table 1). These substitutions were not found in the matching clinical specimens.
One virus isolated from a hospitalized peramivir-treated patient displayed RI by oseltamivir and had NA-N245Y substitution in both the isolate and the matching clinical specimen; this substitution was not reported previously (Fig. 3B). Another virus showing RI by oseltamivir was isolated from an immunocompetent patient who had received no NAI treatment; NA-S331R substitution was found in both the isolate and the clinical specimen. A(H3N2) viruses with this change were reported previously and were characterized by a borderline NI/RI phenotype (Takashita et al., 2020a). Indeed, it has been shown that A(H3N2) viruses carrying positively-charged amino acid substitutions at NA positions 329, 331 or 334 can confer an apparent RI phenotype, but this is caused by markedly higher K m s for MUNANA and K i values for NAIs (Hussain et al., 2021). It is possible that antibody pressure is responsible for selection of amino acid substitutions at these three positions (Air et al., 1985). Three viruses exhibiting RI by zanamivir had either NA-D151N/D mixture (in one case combined with NA-V165I) or NA-M241V/M mixtures (Tables 1 and S3). These changes are likely to have arisen during virus culture. These RI/HRI variants were isolated from 14 countries (Table S4), and most displayed RI/HRI by peramivir (Fig. 3C). A new NA-G145R substitution that conferred a borderline NI/RI phenotype for zanamivir was found in two viruses and in one of the two corresponding clinical specimens (Tables 2 and S4). The substitution NA-G243S was identified in a clinical specimen and its respective isolate, conferring RI/HRI by all NAIs except laninamivir. A different substitution at this residue, NA-G243D, conferred RI/HRI by oseltamivir and zanamivir; data not available for the other two NAIs. A virus with dual substitutions, NA-G247D + NA-I361V, displayed RI by zanamivir and HRI by peramivir; both substitutions were present in the clinical specimen. Based on its location, NA-G247D alone could reduce inhibition by NAIs. Numerous viruses collected in different parts of the world had previously reported substitutions (NA-D197N, NA-I221T, NA-H273Y and NA-D432G) and displayed expected changes in inhibition by NAIs.

B/Victoria-lineage viruses showing RI/HRI
The substitution NA-E105K or NA-E105K/E mixture was found in 11 viruses, alone or combined with other changes (i.e., NA-G104R/G mixture, NA-I115 deletion, NA-P139T/P mixture, or K382R substitution), with some viruses displaying RI/HRI by NAIs. Another substitution at this residue, NA-E105G, combined with NA-P139L substitution, conferred RI by peramivir. Notably, the majority of corresponding matching clinical specimens did not have either NA-E105K or NA-E105G. Similarly, substitutions at neighboring residues (i.e., NA-H101L and NA-G108E) that conferred RI/HRI to peramivir were not confirmed in clinical specimens or they were not available. The following substitutions conferred RI/HRI but were not found in the clinical specimens: NA-T146K or P, NA-Q138K, NA-A245T, NA-T460I). It appears that certain cell culture and virus propagation conditions favor the rapid selection of type B virus NA variants, which present challenges for phenotypic testing.

B/Yamagata-lineage viruses showing RI/HRI
For B/Yamagata-lineage viruses, 7/1171 (0.6%) viruses assessed for 2018-2019 and 0/250 (0%) for 2019-2020 exhibited RI/HRI by at least one NAI (Fig. 2A). These seven RI/HRI variants were collected in five countries (Table S4). A virus from Japan with an NA-H273Y substitution showed RI by oseltamivir and HRI by peramivir (Takashita et al., 2020b), and another virus from the United States with a NA-H273Y/H mixture showed RI by peramivir only (Tables 2 and S4). NA-H273Y was detected in the corresponding clinical specimens. Two viruses, one with NA-A200T substitution and one with NA-I221V substitution, displayed RI by peramivir (Fig. 3D). A virus with NA-S249N substitution (not previously reported) showed RI by zanamivir, but no clinical specimen was available. A virus with an NA-D197N substitution showed borderline NI/RI by oseltamivir.

Frequency of NA amino acid substitutions associated with RI/HRI by NAIs in sequence databases
We  (Table S5).
Of the 4499 B/Victoria-lineage sequences analyzed for both seasons, 13 (0.3%) had NA substitutions associated with RI/HRI: five had NA-K360E, four had NA-D197N, two had NA-I221T and two had either NA-Y142H or NA-G145E. For both seasons, a total of 535 NA sequences were analyzed for B/Yamagata-lineage viruses. Only two (0.4%) of the sequences from 2018-2019 had either NA-G145E or NA-S246P substitution associated with RI/HRI.
Overall, a combined analysis using phenotypic and/or sequence-based methods revealed that the global frequency of influenza viruses either displaying RI/HRI or with potential to exhibit RI/HRI by NAIs was low, being 0.5% (165/35045) and 0.6% (159/26010) for 2018-2019 and 2019-2020, respectively.

Combined analysis of genotypic and phenotypic baloxavir susceptibility data
During the periods covered, PA sequence analysis was the primary tool for baloxavir susceptibility assessment, as only two WHO CCs (Atlanta and Tokyo) had tested viruses phenotypically. To this end, PA sequences submitted by WHO CCs and other laboratories to GISAID were screened for amino acid substitutions associated with reduced susceptibility to this antiviral (PA marker table). Although not included in the WHO-AVWG table, residue 34 is part of the PA active site and one of the key residues (i.e., residues 20, 24, 34, 37, and 38) to which baloxavir binds (Omoto et al., 2018). Therefore, it was of interest to screen PA sequences for amino acid changes at this position. When isolates were available, flagged viruses were subjected to phenotypic testing to confirm the drug susceptibility phenotype. Also, phenotypic testing of viruses representing different subtypes/lineages was performed to calculate the median EC 50 for each type/subtype/lineage. Most (9994/15505, 64.5%) PA sequences analyzed for this period were deposited in GISAID by three WHO CCs [Atlanta (n = 8911; 89.2%), Melbourne (n = 460; 4.6%), and Tokyo (n = 623; 6.2%)], with the remaining 5511 sequences (35.5%) submitted by other laboratories worldwide. A total of 72 PA sequences were flagged as containing previously reported (n = 68) or suspected (n = 4) reduced susceptibility markers; the latter were PA-K34R and PA-E23K + PA-K34Q (Table S6). A total of 1218 viruses from 2018-2019 were tested at the Atlanta CC(n = 387; 31.8%) or the Tokyo CC (n = 831; 68.2%). Baloxavir EC 50 s for the flagged viruses were determined and compared to the median EC 50 to calculate a fold-change. The log 10 transformed baloxavir EC 50 fold-change values were used to prepare column-scatter plots, as was done for the NAI phenotypic data (Fig. 4). As expected, viruses without PA markers (n = 1172) showed a <3-fold increase when compared to the respective median EC 50 . PA marker substitutions were present in all viruses showing a ≥3-fold increase. However, five viruses with the markers PA-K34R, PA-I38F/I, PA-I38M/I, PA-I38V/I, or PA-E199G showed a <3-fold increase in EC 50 as compared to the respective median EC 50 s (Table 3). Of these five viruses, two with PA-I38M/I and PA-I38V/I still showed a <3-fold increase in EC 50 and the one with PA-E199G displayed a ≥3-fold increase when compared to the respective PA sequence-matched controls (Table S6).  Table S6). All viruses that showed a ≥3-fold increase in EC 50 had a PA reduced susceptibility marker (PA-I38T, PA-I38L, or PA-E23K), but not all viruses with other markers (PA-L28P, PA-M34I, or PA-I38V) showed a ≥3-fold increase relative to the subtype/lineage-specific median EC 50 or the PA sequence-matched control virus EC 50 (Tables 3 and S6).
Six viruses that contained PA-I38F/S/T were tested and displayed 2.5-to 49.5-fold increases in EC 50 compared to the median EC 50 . Only three viruses with PA-I38V were tested and they displayed 3.0-to 3.7-fold increases by FRA. The presence of valine at residue 38 represents a natural polymorphism and has not been linked to baloxavir treatment. Viruses with PA-E23G, PA-E23K (Takashita et al., 2020c) or PA-E199G showed 6.9-, 7.4-and 2.9fold increases in EC 50, respectively (Table S6); the latter virus showed a 3.7-fold increase when compared to the PA sequence matched control. Two viruses (one each from Japan and the United States) had PA-K34R and showed 1.6-and 4.7-fold increases, respectively, in EC 50 (Table S6). The reduced susceptibility of the United States virus was confirmed, with a 4.1-fold increase in the EC 50 , by comparing it to a PA sequence matched control.
Twenty-one viruses that contained PA-I38T alone showed 64.3-to 614.0-fold increases in EC 50 , and one virus with PA-I38T was not tested (Table 3). Eight viruses with mixtures at this position (PA-I38T/I or PA-I38T/M/I) showed 5.6-to 250.0-fold increases in EC 50 (Tables 3 and S6). Two other viruses with PA-I38M or PA-I38M/I exhibited 23.7-to 91.8-fold increases in EC 50 . All these viruses were collected in Japan, and most were from baloxavir-treated patients <12 years of age (Table S6). The other flagged viruses had markers associated mainly with less common natural substitutions or polymorphisms: PA-I38L or PA-I38L/V/I (n = 2), PA-I38M or PA-I38M/I (n = 3), PA-I38V or PA-I38V/I (n = 3), and PA-L28P (n = 16) (Table S6). In addition, a virus from the Democratic Republic of the Congo had PA-K34R, one from the UK had PA-E23K + PA-K34Q, and one from Chile (Talca) had PA-E199G. Viruses with PA-I38L or PA-K34R showed 4.1-fold increases in EC 50 . All other available viruses showed fold-changes below the cut-off threshold, including two with PA-L28P. Notably, placing clinical specimens in culture led to the enrichment of virus isolates with PA-I38-substituted variants in several instances.

B/Yamagata-lineage viruses
For B/Yamagata-lineage viruses, 0/793 and 0/335 viruses analyzed for 2018-2019 and 2019-2020, respectively, had PA reduced susceptibility markers. (Table S7). The susceptibility of zoonotic influenza viruses to NAIs and baloxavir was assessed based on analysis of NA and PA sequences deposited in GISAID. Overall, sequences were available for 32 zoonotic viruses. Those viruses with an available NA sequence appear to be NAI susceptible, as no NA substitutions previously associated with RI/HRI phenotypes were found in them. A/Jiangsu/32888/2018 (H5N6) contained dual PA-I38V + PA-E199G substitutions. These substitutions alone have little or no effect on in vitro baloxavir susceptibility when present in other subtypes (PA marker table), but their combined effect has not been determined. Thirteen viruses had PA-A37S; this substitution did not alter baloxavir susceptibility in a recombinant A(H1N1) virus (Hashimoto et al., 2021). Of nine A(H1N2)v viruses collected in the United States during these two periods, three were tested for NAI susceptibility and six were tested for baloxavir susceptibility; all were susceptible to NAIs and baloxavir.

Concluding remarks
Several factors are thought likely to contribute to the detection rate of influenza viruses showing reduced susceptibility to antivirals. For example, when A(H3N2) viruses predominate, fewer viruses displaying RI/HRI by oseltamivir are typically detected. Conversely, higher detection rates are observed when A(H1N1)pdm09 viruses circulate widely; this is mainly due to oseltamivir-resistant NA-H275Y viruses. This potentially reflects the inherent abilities of N1 and N2 NAs to accommodate RI/HRI conferring substitutions without significantly affecting virus fitness (Collins et al., 2009).
The detection of A(H1N1)pdm09 viruses in untreated patients is an indicator of community transmission of resistant NA-H275Y viruses (Hurt et al., 2012;Takashita et al., 2015b), and this needs to be closely monitored. Notably, a small cluster of NA-H275Y viruses was detected in the United States in 2020 (Mohan et al., 2021) and the antigenically drifted hemagglutinin of these viruses may facilitate their spread.
Viruses with RI/HRI phenotypes can emerge as a result of natural amino acid polymorphism (sequence variance among circulating viruses). For example, natural polymorphism is the underlying factor in type B viruses displaying a RI/HRI phenotype to peramivir (Leang et al., 2014;Sleeman et al., 2011). Conversely, A(H3N2) viruses with a RI phenotype to zanamivir commonly emerge during virus culture, which promotes the selection of NA substitutions that reduce NA activity towards sialic acid-containing receptors. In addition, propagating viruses with a mixed population in culture can increase the proportion of variants with reduced susceptibility to antivirals; this applies to both NAIs and baloxavir.
It can be challenging to reconcile the outcomes of sequence-based analysis and phenotypic testing as both approaches have their limitations. For simplicity, the detection rate was calculated based on all markers listed in the WHO-AVWG tables, plus newly identified markers. However, not all markers conferred the expected drug phenotype. Another caveat is a border line NI/RI phenotype observed for viruses with certain markers. For example, NA-D197N was associated with NI/RI by zanamivir. Similarly, PA-L28P and PA-I38V conferred no change in baloxavir susceptibility. Such viruses can be counted as drug susceptible or displaying reduced susceptibility depending on the reference EC 50 value used to calculate fold-change. As sequence-based analysis becomes a foundation for virologic surveillance, it would be beneficial to put additional efforts into establishing clear correlates between identified markers and drug susceptibility phenotype. To this end, recombinant viruses could be used to delineate the effect of a particular marker on a drug-susceptibility phenotype, or sequence-matched control viruses (e.g., PA sequence matched viruses) could be used for comparisons. This is especially important, because sequence-based analysis is the primary means of assessing the drug susceptibility of zoonotic influenza viruses as such viruses are often unavailable for phenotypic testing.
Overall, the proportions of viruses displaying reduced susceptibility to at least one antiviral were low (<1%) for both periods. However, during 2018-2019, there was a striking difference between Japan and the United States in the detection rate of PA reduced susceptibility markers. In Japan, the rate was 4.5% (41/919) and most of the detected viruses (n = 32) had PA-I38T, the principal treatment-emergent marker. Notably, most viruses with PA-I38T belonged to the A(H3N2) subtype and were isolated from young children with several patients not being exposed to baloxavir. This indicates the probable transmission of baloxavir-resistant viruses in local communities and/or households. It appears that using baloxavir to treat children <12 years of age may have contributed to the higher rate of baloxavir resistance in Japan in 2018-2019. In 2019, two Japanese medical professional societies issued recommendations to avoid prescribing baloxavir to children <12 years of age. In 2019-2020, the rate of baloxavir resistance detected in viruses collected in Japan was much lower at 0.4% (3/708) and only one virus, which was collected from a baloxavirtreated adult, had PA-I38T.
It is worth noting that in the United States, baloxavir was approved only for treating people aged 12 years or older. The detection rate was low in both periods: 0.3% (23/7024) in 2018-2019 and 0.1% (7/6509) in 2019-2020. None of the analyzed viruses had PA-I38T substitution. Also, samples submitted to the United States national surveillance program are typically collected before treatment is initiated. These factors may explain the low detection rate, which is consistent with the low circulation of PA variants with naturally reduced susceptibility to baloxavir (Gubareva et al., 2019).
Overall, the detection rate of viruses that either showed, or possessed markers predictive of reduced susceptibility to NAIs and/or the PAI baloxavir was low regardless of the usage of these antivirals. However, the data reported here indicate the need to continue the close monitoring and elucidation of factors contributing to reduced susceptibility to antivirals in influenza viruses.   and NAI (labelled on the x-axis: oseltamivir, zanamivir, peramivir and laninamivir). The boxes indicate the 25th-75th percentiles, and the whiskers stretch to the lowest and highest values within 1.5 times the interquartile region (IQR) value from both the 25th and 75th percentile values, respectively (Tukey's definition). The y-axes have been split into three compartments according to the thresholds recommended by the World Health Organization Expert Working Group of GISRS for normal inhibition (NI) (<10-fold for type A viruses; <5-fold for type B viruses), reduced inhibition (RI) (10-to 100-fold for type A viruses; 5to 50-fold for type B viruses), and highly reduced inhibition (HRI) (>100-fold for type A viruses; >50-fold for type B viruses). NA amino acid substitutions are shown for viruses displaying RI or HRI phenotypes that have been sequenced. Viruses showing NI but carrying amino acid substitutions previously associated with RI or HRI by one or more NAI or showing an RI or HRI phenotype for another NAI are indicated in grey in the NI area above 1.5 times the IQR from the 75th percentile border and below the RI threshold value. Full details about these viruses are given in Tables S3 and S4. Amino acid position numbering is specific to A subtype and B type. Most viruses were tested for susceptibility to oseltamivir and zanamivir; only a subset was tested for susceptibility to peramivir and laninamivir. Column-scatter plots of log-transformed 50% effective concentration (EC 50 ) fold-change values for the PAI baloxavir. The phenotypic susceptibility of influenza viruses to baloxavir was determined with cell culture-based assays, focus-reduction assay (FRA) or highcontent imaging neutralization test (HINT). Overall, 1218 and 1110 viruses were tested phenotypically for 2018-2019 and 2019-2020, respectively. Data are presented by virus subtype or lineage [labelled on the x-axis: A(H1N1)pdm09; A(H3N2); B/Victoria-lineage; and B/Yamagata-lineage] and log-transformed EC 50 s on the y-axis. The boxes and whiskers are as defined in Fig. 3. An arbitrary cut-off of ≥3-fold increase from the median EC 50 was used for reporting viruses with reduced susceptibility to baloxavir. Viruses without PA reduced susceptibility markers showed a <3-fold increase in EC 50 , as compared to the respective median EC 50 s for the two periods. PA markers were present in all viruses that showed a ≥3-fold increase in EC 50 , but not all viruses with markers showed ≥3-fold increases. PA amino acid substitutions are shown for viruses displaying reduced susceptibility. Amino acid position numbering is specific to type A and B viruses.  (1) Yes, oseltamivir, laninamivir (3) No (29) Unknown (41) Yes (5) No (62) Unknown (34) 6

7-10
2 I221T (1) I221T/I mix (1) I221T (1) None (1) Hospital (1) Unknown (1) Unknown (1) No (1) Unknown ( (1) H273Y (1) H273Y/H mix (1) Community (1) Unknown (1) No (1) Unknown (1) No (1) Unknown ( The number of viruses for which data were reported is shown in parentheses if it is less than the number in the 'n' column. b Reduced inhibition (RI) and highly reduced inhibition (HRI) fold-change values are displayed underlined and in bold typeface. For type B viruses, normal inhibition (NI) is a <5-fold increase in the NAI IC 50 ; RI is a 5-to 50-fold increase; and HRI is a >50-fold increase (World Health Organization, 2012  Virus and patient characteristics for type A and B influenza viruses (n = 53) containing PA amino acid substitutions of concern and phenotypically tested by WHO CCs for baloxavir susceptibility. Community (19) Hospital (2) Yes, baloxavir (17) Yes, oseltamivir (1) No (3) No (20) Unknown ( The number of viruses for which data were reported is shown in parentheses if the number is less than the number in the 'n' column.
Antiviral Res. Author manuscript; available in PMC 2022 July 05.