A Cluster of Influenza A(H3N2) Viruses in 2024 From Tasmania, Australia, With Reduced Susceptibility to Baloxavir

ABSTRACT

The emergence of viruses with reduced susceptibility to antivirals poses an ongoing risk which could limit their effectiveness. Here we report a cluster of A(H3N2) viruses detected in Tasmania, Australia, with a PA gene mutation A37T that showed mild reductions in susceptibility in vitro to baloxavir (3.47‐ to 16.33‐fold increased EC₅₀) when compared to wild‐type influenza A(H3N2) viruses. When the PA‐A37T and PA‐I38T substitutions were combined and tested in IRINA assay, eightfold higher inhibition was seen compared to PA‐I38T alone. On‐going monitoring of influenza viruses for reduced susceptibility to antivirals remains an important part of seasonal and zoonotic influenza surveillance.

1. Introduction

Antiviral treatment and prophylaxis are important interventions to minimize the morbidity and mortality of influenza infections. However, the emergence of viruses with reduced susceptibility (RS) to antivirals poses an ongoing risk of treatment failure. Widespread emergence of resistance has occurred previously in different drug classes, including oseltamivir, a neuraminidase inhibitor (NAI) and the adamantanes (M2 inhibitors) [1]. The cap‐dependent endonuclease inhibitor, orally delivered antiviral baloxavir marboxil (baloxavir), has been available since 2018. During clinical trials, RS to baloxavir was detected in up to 19.5% of children [2] and 10.9% of adults [3], respectively, mostly associated with PA‐I38T substitution in seasonal influenza A(H3N2) viruses. A PA‐A37T substitution was also detected in one paediatric patient in these trials but has rarely been detected in seasonal influenza viruses since its introduction [2, 4, 5, 6]. We report here the detection in Tasmania, Australia, during the 2024 influenza season, of A(H3N2) viruses bearing PA‐A37T at a frequency of 15.6% (17/109 of Tasmanian viruses analysed). Baloxavir use was unknown but very unlikely as while it has been registered for use in Australia since February 2020, it has not been distributed to date due to significantly higher cost than oseltamivir.

2. Results

2.1. Detection of PA‐A37T Mutants From Tasmania During the 2024 Influenza Season

During the 2024 influenza season, 109 influenza A(H3N2) viruses were received from Tasmania at the WHO Collaborating Centre for Reference and Research on Influenza (WHOCCRRI, Melbourne). Seventeen of these viruses bore PA‐A37T (15.6%), previously reported to show mildly RS to baloxavir [2, 6]. The majority of cases (n = 16) were collected in June–July and clustered around the capital city Hobart (South) with one case detected in Burnie (northwestern region) (Figure 1). A further 1422 A(H3N2) viruses were sequenced from Australian influenza positive samples received in 2024, and none possessed the PA‐A37T substitution. Additionally, one A(H3N2) virus from Queensland had PA‐I38T, a substitution that is known to highly reduce the susceptibility of influenza viruses to baloxavir [2, 3].

FIGURE 1.

Map of Tasmania, Australia, showing individual occurrences of patient isolates containing A(H3N2) PA‐A37T. The cities are indicated, and colours denote the month of specimen collection. Map plotted using R software.

Phylogenetic analyses of A(H3N2) HA and PA genes of influenza sequences from Tasmania and Australia received in the 2024 season was performed (Figure S1). PA‐A37T viruses formed a distinct cluster in both PA and HA trees. These viruses belonged to subclade J.2.2 (Nextclade: https://clades.nextstrain.org/). One virus, A/Tasmania/673/2024, also grouped with these viruses in the HA tree but not in the PA tree as it did not harbour PA‐A37T, even not as minor variant through deep sequencing. No further information was available to establish any epidemiological link between these cases; however, the virus lacking the PA‐A37T was taken on 20 August 2024, at the end of the period when the PA‐A37T viruses were circulating. Globally very few influenza viruses with PA‐A37T have been detected in recent years (Table S1).

2.2. Tasmania Isolates Bearing the PA‐A37T Substitution Show Reduced Susceptibility to Baloxavir In Vitro

Virus isolates of either wild‐type PA‐A37 or PA‐T37 isolated from clinical specimens were tested in vitro for susceptibility to baloxavir acid (BXA) using IRINA assay [7] (all methods in Supporting Information) to determine EC₅₀S of PA‐A37T mutants and wild‐type (WT) viruses (Table 1). Currently, there are no criteria for defining resistance or RS to baloxavir. An arbitrary threshold (cutoff) of a more than threefold increase in EC₅₀S compared to reference EC₅₀ (median) is used for reporting viruses with RS to baloxavir [5]. EC₅₀S for BXA for wild‐type Tasmania 2024 isolates tested ranged from 0.15 to 0.42 nM, similar to results from the Tokyo WHOCCRRI using the same assay, Table S2 (0.37–1.36 nM). EC₅₀S of PA‐A37T isolates were compared to the average EC₅₀ of matched PA wild‐type controls. Fifteen of 17 PA‐A37T isolates showed 3.47‐ to 16.33‐fold greater EC₅₀S to BXA compared to wild‐type viruses (Table 1), except for A/Tasmania/362/2024 (27.89‐fold) and A/Tasmania/395/2024 (47.06‐fold). A subset of the Tasmania PA‐A37T viruses IRINA tested by Tokyo WHOCCRRI showed 4.06‐ to 10.51‐fold greater EC₅₀S to BXA compared to wild‐type viruses, including A/Tasmania/362/2024 and A/Tasmania/395/2024 (Table S2). The same subset of PA‐A37T viruses tested in vitro in focus reduction assay (FRA) [8] showed 8.67‐ to 17.82‐fold greater EC₅₀S to BXA compared to wild‐type viruses (Table S2). The PA‐A37T mutants showed a similar RS phenotype for BXA as PA‐I38M but lower EC₅₀S than PA‐I38T (data not shown) containing viruses (Tables 1 and S2).

TABLE 1.

BXA EC50s of Tasmania wild type and PA‐A37T viruses compared WT controls determined by IRINA assay.

Virus	EC50	95% confidence interval	Fold change a
A/Tasmania/106/2024 b	0.16	0.12–0.21	0.89
A/Tasmania/152/2024 b	0.23	0.18–0.29	1.28
A/Tasmania/425/2024 b	0.18	0.14–0.24	1.00
A/Tasmania/412/2024 b	0.15	0.12–0.19	0.83
A/Tasmania/382/2024 c PA‐C321F	0.26	0.20–0.33	1.44
A/Tasmania/399/2024 c PA‐I428V	0.42	0.35–0.50	2.33
A/Tasmania/413/2024 c PA‐I428V	0.18	0.15–0.23	1.00
A/Tasmania/421/2024 c PA‐I201V	0.41	0.33–0.52	2.28
A/Tasmania/104/2024	1.66	1.27–2.15	9.22
A/Tasmania/181/2024	0.94	0.70–1.24	5.21
A/Tasmania/294/2024	2.65	2.03–3.42	14.72
A/Tasmania/340/2024	1.24	0.91–1.67	6.89
A/Tasmania/327/2024	1.18	0.83–1.64	6.56
A/Tasmania/332/2024	0.62	0.47–0.82	3.47
A/Tasmania/362/2024	5.02	3.47–7.10	27.89
A/Tasmania/418/2024	2.36	1.90–2.90	13.11
A/Tasmania/395/2024	8.47	6.40–11.1	47.06
A/Tasmania/290/2024	1.44	1.03–2.00	8.00
A/Tasmania/685/2024	0.64	0.42–0.96	3.53
A/Tasmania/365/2024	1.15	0.85–1.54	6.39
A/Tasmania/293/2024	2.94	2.40–3.59	16.33
A/Tasmania/138/2024	1.69	1.18–2.40	9.39
A/Tasmania/307/2024	1.65	1.29–2.10	9.17
A/Tasmania/315/2024	0.98	0.71–1.34	5.43
A/Tasmania/316/2024	1.21	0.88–1.63	6.72
A/Louisiana/50/2017 (wild‐type) d	0.16	0.13–0.19	0.89
A/Louisiana/49/2017 (PA‐I38M) d	1.74	1.43–2.11	9.66

Note: Mean BXA EC₅₀S and 95% confidence interval of at least three independent experiments, each performed in duplicate or triplicate in MDCK‐SIAT1 cells. EC₅₀S were calculated using GraphPad PRISM software. Bold: Tasmania viruses containing PA‐A37T. RS to baloxavir is provisionally defined as more than threefold change in EC₅₀ of a test virus compared to the subtype/lineage‐specific median EC₅₀S. Viruses with superscript “c” were not used for calculation of mean EC₅₀S of normal susceptibility viruses for fold change calculations.

^a Fold change calculated by dividing EC₅₀S by mean EC₅₀ value of matched PA wild‐type controls.

^b Denotes matched PA wild‐type controls.

^c Has an extra PA substitution listed that is not present in the other Tasmania virus sequences listed in the table.

^d The CDC Baloxavir Susceptibility Reference Virus Panel (Version 1.1), FR‐1678.

2.3. Reverse Genetics Virus Bearing Both the PA‐A37T and PA‐I38T Substitutions Show Greater Reduced Susceptibility to Baloxavir Compared to Single Mutants

PA‐I38T is the most frequently occurring substitution associated with RS to baloxavir and is the only known clinically relevant reduced susceptibility marker [2, 3, 4, 5, 6]. The potential for PA‐A37T to combine with other important PA substitutions, such as PA‐I38T, and cause greater clinical impact on baloxavir susceptibility is theoretically possible; as has been shown previously for oseltamivir resistance with NA‐S247N + H275Y mutants compared with NA‐H275Y mutants [9]. Reverse genetics (RG) was used to make viruses containing PA‐A37T, PA‐I38T or PA‐A37T + PA‐I38T using the A(H3N2) A/Darwin/6/2021 strain. There was no appreciable difference between 50% tissue culture infectious dose (50% TCID₅₀) titres of RG virus stocks for wild‐type, PA‐A37T, PA‐I38T and the double mutant viruses generated from diluted posttransfection supernatants (Table 2). A side‐by‐side comparison of multicycle replication of these viruses was performed over 4 days (Figure 2A). Single mutant viruses PA‐A37T and PA‐I38T viruses showed on average one log lower infectious titres than wild‐type virus at all time points, and the double mutant showed marginally lower titres than all three viruses, showing up to 1.5 log lower titres compared with wild‐type virus at day 4 after infection.

TABLE 2.

Infectious virus titres and BXA EC₅₀S of Reverse genetics (RG) A/Darwin/6/2021 PA wild‐type or mutant viruses determined by IRINA assay.

Virus	TCID50/mL	Standard deviation	EC50	95% confidence interval	Fold change a
RG A/Darwin/6/2021 (wild‐type)	1.37 x 10^8	8.22 x 10^7	0.14	0.10–0.21	1.00
RG A/Darwin/6/2021 PA‐A37T	1.99 x 10^8	7.10 x 10^7	1.11	0.73–1.65	7.79
RG A/Darwin/6/2021 PA‐I38T	1.12 x 10^8	3.99 x 10^7	15.15	12.08–18.93	106.39
RG A/Darwin/6/2021 PA‐A37T + PA‐I38T	6.86 x 10^7	3.51 x 10^7	121.30	91.9–158.60	851.83

Note: TCID₅₀ titres of RG A/Darwin/6/2021 PA wild‐type and mutant virus stocks. Mean TCID₅₀/mL and standard deviation (SD) of three independent experiments performed on freeze‐thawed virus stocks. Mean BXA EC₅₀ and 95% confidence interval of four independent experiments, each performed in duplicate. Bold: RG Darwin viruses containing PA mutations. RS to baloxavir is provisionally defined as more than threefold change in EC₅₀ of a test virus compared to the subtype/lineage‐specific median EC₅₀S.

^a Fold change calculated by dividing EC₅₀S by mean EC₅₀ of matched PA wild‐type control.

FIGURE 2.

Multicycle growth kinetics and BXA susceptibility of RG Darwin PA wild‐type and mutant viruses. (A) Replication kinetics of viruses was determined in MDCK‐SIAT1 cells. Cells were infected at low MOI and supernatants harvested daily for 4 days after infection. Infectious titres were determined by TCID₅₀ assay. Mean and SD of infectious virus titres (logTCID₅₀/mL) are shown for the time course from two independent experiments, each performed in triplicate. A dotted line represents the lower limit of detection (LLOD). (B) BXA susceptibility of viruses was determined in MDCK‐SIAT1 cells using IRINA assay. RFU values were calculated as a percentage of the virus control (VC). Mean percentage RFU (relative to VC) and SD (upper limit) of four independent experiments, each performed in duplicate are shown. A dashed line indicates 50% RFU (relative to VC). Graphs were plotted using GraphPad PRISM.

IRINA assay for BXA susceptibility (Figure 2B and Table 2) showed that the PA‐A37T, PA‐I38T or double mutant PA‐A37T + PA‐I38T viruses had 7.79‐fold, 106.39‐fold and 851.83‐fold greater EC₅₀S, respectively, compared to their respective wild‐type counterpart. A polymerase reconstitution assay (minigenome assay) was also used to determine BXA susceptibility in vitro of these mutant constructs compared to wild‐type (Figure S2). Polymerase inhibition by BXA showed a similar trend to IRINA results from RG viruses with PA‐WT > PA‐A37T > PA‐I38T > PA‐A37T + PA‐I38T. There was no appreciable difference in polymerase activity between wild‐type and mutant viruses, although variability was seen between experiments. Mean ± SD polymerase activity (as a percentage of WT) for PA‐A37T, PA‐I38T and PA‐A37T + PA‐I38T was 215 ± 112, 131 ± 79 and 92 ± 32, respectively.

2.4. Structural Analysis of PA Mutants Complexed With Baloxavir and ΔΔG Protein Stability Modelling Showed a Positive Free Energy Change of Mutants Compared to Wild‐Type

PA protein structures with the wild‐type sequence complexed with BXA were prepared using two approaches: homology modelling via SwissModeller based on Protein Data Bank entries (6fs6, 8t5v, 2w69 and 6qnw), and direct complex prediction using Chai‐1 [9], a state‐of‐the‐art neural network‐based biomolecular structure predictor that takes protein sequence and ligand SMILES as inputs. 6fs6 and 8t5v are baloxavir‐bound structures based on A(H1N1)pdm09 strains, while 2w69 and 6qnw are A(H3N2) strains in apo state. All structures were prepared with the docked ligand and refined in YASARA to add missing atoms/hydrogens, correct bond orders and optimize ligand docking.

Table 3 summarizes the computed ΔΔG_bind values and their energetic components for the PA‐I38T, PA‐A37T and PA‐A37T + I38T mutants. A positive ΔΔG_bind indicates weaker predicted binding of baloxavir relative to the wild‐type protein. ΔΔG_bind can be decomposed into contributions from changes in potential energy (ΔΔG_pot), reflecting direct protein‐ligand interactions, and solvation energy (ΔΔG_sol), reflecting the energetic cost of transferring the receptor and ligand from solvent into the bound complex. The solvation term is further composed of Coulombic (ΔΔG_sol,coulombic), van der Waals (ΔΔG_sol,VdW) and surface entropic cost (ΔΔG_sol,surface) contributions, corresponding respectively to changes in electrostatic interactions with solvent, nonpolar solute–solvent interactions, and the entropy penalty associated with changes in solvent‐exposed molecular surface area upon binding. The experimental data indicate RS of the PA‐A37T mutant relative to the wild‐type, with a further loss of susceptibility observed in the PA‐A37T + PA‐I38T double mutant compared to PA‐A37T alone. Based on this trend, the binding free energy change upon mutation (ΔΔG_bind) is expected to be positive for both mutants, with ΔΔG_bind(PA‐A37T) < ΔΔG_bind(PA‐A37T + PA‐I38T), reflecting progressively weaker binding to BXA. This trend is consistent with the results of our computational binding free energy calculations across the 2w69‐based homology model and most of the Chai‐1‐predicted models.

TABLE 3.

ΔΔG energy breakdown (units: kcal/mol), comprising potential, solvation (Coulombic and van der Waals) and surface energy components.

Mutations	ΔΔGbind	ΔΔGpot	ΔΔGsol	ΔΔGsol,coulombic	ΔΔGsol,VdW	ΔΔGsol,surface
PA‐I38T	19.60	13.88	5.73	7.44	−0.36	−1.36
PA‐A37T	4.07	−0.95	5.02	5.06	0.24	−0.28
PA‐A37T + PA‐I38T	29.43	17.89	11.54	14.98	−1.63	−1.81

Note: Values are the average of 50 replicates of the energy minimization simulation in YASARA.

Among the single mutants, PA‐I38T showed a large positive ΔΔG_bind, driven primarily by a positive ΔΔG_pot. Structural analysis (Figure 3) reveals that wild‐type Ile(I)38 contributes hydrophobic contacts that stabilize baloxavir's aromatic tail, whereas replacement with Thr(T) reduces the local hydrophobic surface and disrupts these interactions, consistent with previous reports of PA‐I38T‐mediated baloxavir resistance [10]. PA‐A37T also showed a positive ΔΔG_bind, but this was driven mainly by a more positive Coulombic solvation energy term. As residue 37 does not directly contact baloxavir in the modelled structures, this suggests an indirect effect in which the substitution of alanine with the more polar threonine alters the local solvent environment and increases the desolvation penalty upon complex formation. In the PA‐A37T + PA‐I38T double mutant, both effects are observed, with increased potential energy and solvation energy penalties contributing synergistically to a larger positive ΔΔG_bind than PA‐A37T alone. This is consistent with the experimental observation that PA‐A37T reduces baloxavir susceptibility relative to wild‐type, while PA‐A37T + PA‐I38T shows a further reduction in susceptibility compared to PA‐A37T.

FIGURE 3.

PA complexed with baloxavir. (A) PA‐WT, (B) PA‐A37T and (C) PA‐A37T + PA‐I38T structures. PA‐WT complex was predicted by Chai1 based on Tasmania/104 sequence without the PA‐A37T mutation. Mutant complexes were obtained by applying the relevant mutations, followed by an energy minimization via simulation in YASARA.

Among models that reproduced the experimental susceptibility trends, residue 37 was positioned closer to baloxavir and the surrounding binding pocket than in models that did not reproduce the observed trends. Although residue 37 does not directly contact baloxavir, introduction of the PA‐A37T substitution may alter the local electrostatic and solvation environment around the binding site, consistent with the more positive Coulombic solvation contribution to ΔΔG_bind observed for PA‐A37T‐containing mutants. In contrast, models that deviated from experimental trends tended to place residue 37 further from the ligand. These outlier models included the SwissModel homology models based on the A(H1N1)pdm09 crystal templates 6fs6 and 8t5v, which differ from the A(H3N2) wild‐type by 11–12 PA mutations. Such sequence mismatches may compromise the structural accuracy of the binding pocket near residue 37 and contribute to discrepancies in predicted binding energy trends. The experimental susceptibility trends, as determined by IRINA assay, are summarized in Table 2.

3. Discussion

Baloxavir has been shown to be safer and significantly shorten the hospitalisation period compared to oral oseltamivir, which is the standard treatment regimen for influenza [10]; however, significant resistance was detected in clinical trials, particularly in children. In Japan, where baloxavir is used most, the frequency of baloxavir resistance was 3.3% in 2022–2023 [5] and 4.5% for the pre‐COVID period 2018–2019 [4], mostly associated with PA‐I38T. This mutation has clinical impact, as baloxavir‐treated patients infected with PA‐I38T virus show prolonged virus shedding and have delayed symptom alleviation compared to those infected with wild‐type virus [3, 6, 11].

The low‐level reductions in baloxavir susceptibility due to PA‐A37T in previous studies showed reductions of 8.00‐fold in A(H3N2) [2], similar to what we observed in this study, and 5.00‐fold in A(H1N1)pdm09 [6]. These reductions are unlikely to alter the drug's efficacy in vivo, as with majority of non‐PA‐I38X variants [2, 11, 12]. However, the potential for PA‐A37T to combine with other important PA substitutions, such as PA‐I38T, and cause greater clinical impact is theoretically possible and we have shown that a double mutant bearing PA‐A37T + PA‐I38T shows greater RS to baloxavir in vitro compared to PA‐I38T alone. The double mutant virus showed somewhat lower fitness in vitro compared to wild‐type virus but fitness and transmissibility in vivo are yet to be determined.

On‐going monitoring of influenza viruses for RS to antivirals remains an important part of surveillance and pandemic preparedness and should include monitoring for the PA‐A37T substitution along with other PA substitutions known to impact the EC₅₀ of baloxavir against influenza viruses.

Author Contributions

Saira Hussain: conceptualization, investigation, writing – original draft, formal analysis, project administration, supervision, methodology. Charmaine Chia: conceptualization, investigation, methodology, writing – original draft. Ashwin Muraleetharan: investigation, data curation. Yi‐Mo Deng: project administration, formal analysis, methodology, supervision, writing – review and editing. Xiaomin Dong: investigation, writing – review and editing. Presa Chanthalavanh: investigation. Emi Takashita: investigation, methodology, writing – review and editing, data curation. Clyde Dapat: writing – review and editing, investigation, data curation. Jessica E. Miller: data curation. Heidi Peck: investigation. Raphael T. C. Lee: writing – review and editing, data curation. Jan Williamson: resources, writing – review and editing. Louise Cooley: resources. Sebastian Maurer‐Stroh: project administration, supervision, writing – review and editing, conceptualization. Ian G. Barr: writing – review and editing, supervision, project administration, conceptualization.

Funding

The Melbourne WHO CC is funded by the Australian Government Department of Health, Disability and Ageing. The Tokyo WHOCCRRI is supported by Grants‐in‐Aid for Emerging and Re‐emerging Infectious Diseases from the Japanese Ministry of Health, Labor, and Welfare (21HA2003 and 24HA2005) and by the JSPS KAKENHI (18K10036).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Phylogenetic analysis of (A) HA and (B) PA of A(H3N2) Tasmania viruses received at WHOCCRI, Melbourne, during 2014 influenza season.

Figure S2: BXA susceptibility of PA wild‐type and mutant polymerases. BXA susceptibility of polymerase was determined in HEK293T cells using the polymerase reconstitution assay. RLU values were normalized to the three polymerase proteins expression only (3P, with no NP) control. Normalised RLU values were calculated as a percentage of the untreated polymerase control. Mean percentage normalized RLU and SD (upper limit) are shown for four independent experiments each performed in duplicate. Graphs were plotted using GraphPad PRISM.

Table S1: Global frequency estimate of influenza A(H3N2) PA‐A37T viruses by year.

Table S2: BXA EC₅₀S of Tasmania PA‐A37T viruses compared WT controls determined by IRINA and FRA assay by Tokyo WHOCCRRI.

Table S3: Primers for one‐step RT‐PCR used for NGS.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.