A rapid pyrosequencing assay for the molecular detection of influenza viruses with reduced baloxavir susceptibility due to PA/I38X amino acid substitutions

Abstract Baloxavir marboxil is a novel endonuclease inhibitor licensed for treatment of otherwise healthy or high‐risk individuals infected with influenza. Viruses with reduced baloxavir susceptibility due to amino acid substitutions at residue 38 of the PA have been detected in some individuals following treatment. Here, we describe a genotypic pyrosequencing method that can be used to rapidly screen circulating influenza A and B viruses for substitutions in the PA/I38 codon and to quantify mixed viral populations. This method is suitable for surveillance of baloxavir susceptibility and to analyse samples from hospitalised patients undergoing baloxavir treatment to aid in clinical decision making.

obtained from baloxavir-treated patients and the potential transmissibility of these viruses, surveillance is important to monitor for the emergence of PA/I38X variants in the community. Importantly, rapid detection of viruses with reduced antiviral susceptibility in hospitalised patients can aid clinicians in selecting appropriate antiviral drugs and improve patient management.
Point-of-care tests are available for the rapid detection of influenza infection, however, these tests do not have the capacity to provide information on the presence of specific amino acid substitutions.
Therefore, laboratory assays are utilised to determine antiviral susceptibility. Phenotypic assays that directly measure baloxavir susceptibility have been developed [5][6][7] ; however these assays typically require cultured isolates, are slow (3-5 days) and relatively low throughput.
Consequently, rapid genotypic assays which can be performed directly on clinical specimens are required. Pyrosequencing has been previously utilised to detect amino acid substitutions that are known to confer reduced susceptibility to M2 ion channel inhibitors and neuraminidase inhibitors. 8 Here, we outline a pyrosequencing method for the detection of PA/I38X variants in A(H3N2), A(H1N1pdm09) and influenza B viruses and report on the accuracy of sequence analysis and estimated mixture proportions.
Full-length PA nucleotide sequences for all circulating influenza subtypes/types submitted to the Global Initiative on Sharing All Influenza Data (GISAID) database from 2009 to 2018 were downloaded. For each virus type/subtype, nucleotide sequences were aligned using MAFFT and primer sets were designed such that they bound to regions of high similarity (>90% conservation of sequences) 9 (Table 1). RNA was extracted using the QIAamp Viral RNA kit (Qiagen) according to the manufacturer's protocol, and RT-PCR was conducted using the MyTaq One-Step RT-PCR kit (Bioline) and standard thermocycling conditions. 10 The PyroMark vacuum prep workstation, PyroMark ID Q96 and PyroMark gold reagents (Qiagen) were used as previously described. 11 The workflow for identifying PA/I38X variants is depicted in Figure 1, where the sequence of the PA/I38 codon is determined using the "sequence analysis" (SQA) mode of the PyroMarkID Q96. A biotinylated PCR product will yield a pyrogram and a nucleotide sequence for a short region (approximately 15-30 base pairs) that encompasses the single nucleotide polymorphism (SNP) of interest. As a result, the presence of an amino acid substitution can be identified.
As biotin is tagged on the forward primer of the RT-PCR reaction, the codons depicted in Figure 1 are in the reverse complement. It is also important to note that the codon sequence for the A(H1N1pdm09) wild-type PA/I38 was TAT (ATA, forward direction) prior to 2015 but has since changed to AAT (ATT, forward direction). Once the nucleotide sequence for a virus is obtained and an amino acid substitution is detected, the relative proportion of the wild type and variant mixture proportion can be assessed using the "Allele Quantitation (AQ)" mode. The AQ mode will estimate the proportion of the two nucleotides of interest based on the relative height of the pyrogram peak. This additional step is only necessary if detailed mixture analysis is required ( Figure 1).
The specificity of the pyrosequencing primers was tested with to 10 1 RNA copy number/mL. PyroMarkID SQA software will provide a "pass," "check" or "fail" quality score for each sequence.
SQA analysis was successfully performed with a "pass" quality check on all virus types/subtypes until 10 3 RNA copies/mL. A "check" quality score was obtained at 10 2 RNA copies/mL for all virus types; however, a nucleotide sequence was still generated by the PyroMark ID software. For sequences generated with the "check" quality score, pyrograms need to be visually checked for markers of poor quality sequence such as wide peaks, spurious peaks, initial baseline drift or overall low signal. At 10 1 RNA copies/mL, the pyrosequencing method we have described cannot obtain a sequence for all influenza virus types and a "fail" quality score was obtained.   Table 2).
To evaluate the accuracy of the mixture proportion estimates at low RNA levels, a PA/I38T:WT virus mixture (reverse genetics derived A/Perth/261/2009) was prepared and titrated 10-fold from 10 7 to 10 1 RNA copy number/mL. AQ analysis of these samples showed that mixture proportions remained consistent until 10 3 copy number/mL ( Table 3).
The main analysis to determine the accuracy of the assay was conducted with varying mixtures of plasmids rather than viruses to improve the accuracy of mixture preparation. However, when equivalent mixtures of live viruses have been prepared, the assay has performed similarly (data not shown). Laboratories that are establishing this assay may wish to validate the primers and assay on clinical material from patients.
Pyrosequencing has several benefits for the detection of SNPs to infer antiviral susceptibility. It is rapid, high-throughput, easy to analyse and can provide the quantification of wild type and variant mixtures. The relatively high sensitivity of our assay also allows for detection of PA/I38X variants and mixture estimates in clinical specimens with low RNA copy numbers and does not require a cultured isolate. This allows for a turn-around time of F I G U R E 1 Workflow for the identification of PA/I38X variants using pyrosequencing. The PyromarkID system and Pyromark Gold reagents in conjunction with the primers in Table 1  Amino acid substitutions known to confer reduced susceptibility to antiviral compounds can also be determined with other genotypic methods such as qPCR, Sanger sequencing and next-generation sequencing (NGS). Although qPCR is a high-throughput method that is widely established in laboratories and is relatively inexpensive, the variety of PA/I38X substitutions already observed in clinical specimen means that separate assays would be required for each potential substitution and the sequence of the PCR products is not usually obtained so novel changes at PA/I38X would not be identified. 14 Sanger sequencing is of moderate cost and is accessible but requires several days to generate sequences and is relatively inaccurate in assessing viral mixtures. 14 NGS has higher accuracy for viral mixture quantification but requires extensive sample preparation and data analysis.
Pyrosequencing is frequently performed for identifying reduced susceptibility to the neuraminidase inhibitors (NA/H275Y) and the M2 ion channel inhibitors (M2/S31N), although the equipment is not as commonly available in laboratories as real-time PCR thermocyclers. 15