Neuraminidase inhibitors (NAIs) are currently the only licenced antiviral drugs which are effective for the treatment or prophylaxis of seasonal influenza. National and regional antiviral stockpile policies, where they exist, rely primarily on oseltamivir and to a far lesser extent zanamivir, both of which have been approved for use in many countries since 1999–2000. In Japan, two other NAIs, peramivir and laninamivir, have been approved for seasonal use, and favipiravir (T705; Toyama Chemicals), a viral RNA dependent RNA polymerase inhibitor, has recently been approved for pandemic preparedness stockpiling only (http://www.toyama-chemical.co.jp/eng/news/news140324e.html). Peramivir is also approved for use in the Republic of Korea, China, and the United States.

Experience from 2007–2008, when the former seasonal A(H1N1) virus acquired oseltamivir resistance due to an H275Y neuraminidase (NA) amino acid substitution and spread globally within 12 months, has demonstrated that surveillance for NAI-resistant viruses is essential both to guide seasonal clinical management and inform pandemic preparedness strategies (Lackenby et al., 2008; Collins et al., 2009; Dharan et al., 2009; García et al., 2009; Hauge et al., 2009; Hurt et al., 2009; Meijer et al., 2009; Ujike et al., 2010).

The former seasonal A(H1N1) H275Y virus exhibited highly reduced inhibition (HRI) by oseltamivir and peramivir in vitro and was shown to be clinically resistant to oseltamivir (Kawai et al., 2009; Dharan et al., 2010; Matsuzaki et al., 2010; Saito et al., 2010). Additional NA substitutions (R222Q, V234M, D344N and D354G) compensated for the detrimental effect of the H275Y substitution on virus fitness, allowing the virus to spread efficiently (Bloom et al., 2010; Rameix-Welti et al., 2011; Abed et al., 2011; Bouvier et al., 2012).

The World Health Organization (WHO) Global Influenza Surveillance and Response System (GISRS) expert working group on surveillance of influenza antiviral susceptibility (WHO-AVWG) was established in 2011 to provide advice on GISRS surveillance strategies for influenza antiviral susceptibility and to provide practical guidance to WHO-recognized National Influenza Centres (NICs) (WHO, 2012, 2013).

To standardise interpretation and reporting of NAI susceptibility of influenza viruses to individual NAIs, clear definitions were formulated by the WHO-AVWG using 50% inhibitory concentration (IC₅₀; the concentration of drug required to inhibit a standardised amount of NA activity by 50%) fold-change thresholds, compared to the median for viruses from the same type/subtype/lineage showing ‘normal inhibition’ (NI) (WHO, 2012). Those showing ‘reduced inhibition’ (RI) are influenza A viruses that have a 10- to 100-fold increase in IC₅₀, or influenza B viruses with a 5- to 50-fold increase in IC₅₀. Viruses showing HRI are influenza A viruses with > 100-fold increase in IC₅₀ or influenza B viruses with >50-fold increase in IC₅₀ (WHO, 2012).

Recently, we published a global update on the antiviral susceptibility of human influenza viruses collected by NICs between May 2012 and May 2013 as the first of a series of annual reports (Meijer et al., 2014). Only 0.2% (n = 27) of 11,387 viruses tested showed HRI against at least one of the four NAIs, usually oseltamivir, and mainly in A(H1N1)pdm09 viruses (21/27). Despite >99% of circulating viruses being sensitive to all four NAIs during the 2012–2013 period, localised community circulation of influenza viruses with RI or HRI has occurred in recent years, most notably with A(H1N1)pdm09 viruses containing the H275Y NA substitution (Hurt et al., 2012; Garg et al., 2013). Animal models have shown that A(H1N1)pdm09 H275Y viruses with additional NA amino acid substitutions, V241I and N369K, have increased replication and transmission fitness (Butler et al., 2014; Abed et al., 2014). Importantly, >97% of the N1 sequences from circulating A(H1N1)pdm09 viruses in 2012–2013 contained the two NA substitutions V241I and N369 K that improve viral fitness of the variant virus (Meijer et al., 2014). These observations, together with those from the former seasonal A(H1N1) 2007–2008 event, illustrate the potential for the global emergence of fit A(H1N1)pdm09 viruses with HRI by oseltamivir and peramivir.

Here, we analysed the NAI susceptibility data for influenza viruses collected across 113 countries by GISRS laboratories between May 2013 and May 2014 (subsequently referred to as 2013–2014).

NIC within each country receives or collects clinical specimens from the national laboratory network in order to conduct preliminary analyses. NICs also send representative virus isolates to at least one of the five WHO CCs (Atlanta, United States; Beijing, China; London, United Kingdom; Melbourne, Australia, and Tokyo, Japan; http://www.who.int/influenza/gisrs_laboratory/collaborating_centres/list/en/) for more advanced analyses. At the WHO CCs, viruses are in general passaged one or two times in MDCK cells before being subjected to phenotypic antiviral susceptibility testing.

Five WHO CCs provided IC₅₀ and NA amino acid substitution data for virus isolates, notably for those showing RI or HRI by NAIs, recovered from clinical specimens collected between week 21/2013 (20/5/2013) through week 20/2014 (18/5/2014). When available, patient-specific epidemiologic data such as gender, age, geographic location, healthcare setting (community, hospitalised and sentinel/non-sentinel specimen collection), antiviral treatment history and immune status, were included in the analyses. All five WHO CCs tested for oseltamivir and zanamivir susceptibility, and additionally the Atlanta, Melbourne and Tokyo WHO CCs tested for peramivir and laninamivir susceptibility (Supplementary Table 1).

The WHO CCs tested 10,641 viruses from the 2013–2014 period for NAI susceptibility using local adaptations of the fluorescence-based NA enzyme inhibition assay described by Potier et al. (1979) (Supplementary Table 1). The majority of viruses tested were derived from community surveillance specimens, typically collected for influenza diagnosis, and therefore prior to any NAI treatment. However, while antiviral treatment information was not available for many of the specimens, a proportion of viruses were probably derived from patients during or after treatment with NAI, in hospital or community settings. The number of viruses tested was well distributed across the time period but with a small peak during the Southern Hemisphere winter and a prominent peak during the Northern Hemisphere winter (Fig. 1A).

During the 2013–2014 period, B/Yamagata-lineage haemagglutinin (HA) – B/Victoria-lineage NA reassortants were detected worldwide, notably in China. While not all B viruses had HA and NA genes sequenced, to assess their reassortant status, 65 known reassortants were tested and results allocated to the B/Victoria-lineage dataset. Over the 12 months 5152 (48%) A(H1N1)pdm09, 2574 (24%) A(H3N2), 2311 (22%) B/Yamagata-lineage and 604 (6%) B/Victoria-lineage viruses were tested.

By WHO region (http://www.who.int/about/structure/en/), 49% of viruses tested originated from the Western Pacific Region, 38% from the Americas, 8% from Europe, <3% from Africa, <2% from South-East Asia and <1% from the Eastern Mediterranean Region (Fig. 1B).

Due to differences in phenotypic NA inhibition assay methodology between the WHO CCs (Supplementary Table 1) all raw IC₅₀ data were converted into relative fold-change values to facilitate pooled analysis of the data from all five WHO CCs (Meijer et al., 2014). Box-and-whisker plots based on log-transformed IC₅₀ fold-change data were generated using Tukey’s method to display the range of data and outliers (Fig. 2). Additionally, the box-and-whisker plots were constructed with the Y-axis split into three sections indicating the IC₅₀ fold-change range for viruses classified as NI, RI or HRI based on the criteria above (Fig. 2). Among 10,641 viruses tested, 204 viruses (2%) showed RI or HRI to one or more of the NAIs (Fig. 2 and Table 1). The NA genes of all 204 RI/HRI viruses were sequenced, together with those present in clinical specimens from which 67 of the viruses were recovered. Compared to the consensus NA sequences of wild-type viruses showing NI, 198 of the RI/HRI viruses encoded an amino acid substitution in the NA glycoprotein, of the clinical specimens available (n = 67), the same mutations to those observed in the isolate were also present in 63 corresponding clinical specimens (Table 1).

Of 5152 A(H1N1)pdm09 viruses tested, 175 (3%) showed RI or HRI by one or more of the NAIs (Fig. 2 and Table 1), of which the NA amino acid substitution H275Y was present in 169 of those. Twelve of the 169 H275Y variants harboured a mixed population of H275Y variant virus and H275 wild type virus. One H275Y variant with an additional NA I223R substitution was detected in Japan from a hospitalised patient treated with peramivir; this dual substitution has been reported previously by the United States, in a child after prolonged treatment with oseltamivir (Nguyen et al., 2010). The timing of specimen collection and geographic distribution of the 169 H275Y variant viruses are shown in Fig. 3. The H275Y variants were detected between weeks 24/2013 and 15/2014 (Fig. 3A), with 112 (66%) detected in the Western Pacific region, 56 (33%) in the Americas and one (1%) in Europe, from 10 countries in total (Fig. 3B). The NA H275Y substitution conferred 151- to 2212-fold higher oseltamivir IC₅₀ values and 87- to 2045-fold higher peramivir IC₅₀ values compared to wild type viruses, but had little or no effect on zanamivir or laninamivir susceptibility (Table 1). The H275Y/H mixed variants showed 5.4- to 584-fold higher oseltamivir IC₅₀ values and 4.8- to 599-fold higher peramivir IC₅₀ values, depending on the proportion of the H275Y variant in the mixed population. The H275Y/I223R variant showed HRI with very high increases in oseltamivir (10,739-fold) and peramivir (7709-fold) IC₅₀ values and RI by zanamivir and laninamivir, with 18.1- and 22-fold IC₅₀ increases respectively, compared to values for A(H1N1)pdm09 viruses displaying NI by the four NAIs. The NA H275Y substitution was confirmed in 58 clinical specimens for which sequence results were available and the NA I223R substitution was confirmed in the clinical specimen from which the H275Y/I223R variant virus was isolated (Table 1).

Of the H275Y variant viruses detected in 2013–2014 where patient setting information was available, 82% were from non-hospitalised patients (Table 1). In addition, only two of the 99 patients for whom immune status information was available, were reported as being immunocompromised. Based on available antiviral treatment information, 10 of the 12 H275Y/H mixed variants were recovered from patients during or after treatment with oseltamivir or peramivir. Of the remaining H275Y variants recovered, 82% were from patients had not been treated with NAIs before specimen collection, suggesting that there is potential for spontaneous emergence and spread of these resistant viruses in the community. Indeed, a cluster of A(H1N1)pdm09 viruses with the NA H275Y substitution was detected in 39 (29%) of 135 untreated community cases in Sapporo, capital of Hokkaido, Japan (described in detail in Takashita et al., 2014). In the United States, 57 oseltamivir-resistant A(H1N1)pdm09 viruses were detected in 20 states with a majority being collected from patients not exposed to oseltamivir (Okomo-Adhiambo et al., 2015). In China, nine H275Y variant viruses were detected in nine provinces, while the oseltamivir exposure history of four patients was unknown, the remaining five had no exposure to oseltamivir.

Other NA amino acid substitutions (D199E, I223K, I223T, I223R and S247G) were detected in viruses with elevated IC₅₀ values (Fig. 2 and Table 1). The clinical specimens yielding four of these isolates were available and all contained the corresponding substitutions; no matching clinical specimen was available for the viruses carrying NA D199E or NA I223R substitutions. The IC₅₀ fold-changes for virus with the dual substitution NA H275Y/I223R compared to those of viruses with NA H275Y or NA I223R only (Fig. 2), shows clearly the synergistic effect of this dual substitution and induction of RI or HRI for all four NAIs tested (Tan et al., 2013).

Nine (0.3%) A(H3N2) viruses out of 2574 tested showed RI or HRI by oseltamivir or zanamivir (Fig. 2 and Table 1). Three of these had a NA Q136K amino acid substitution (Table 1). Available clinical specimens yielding two of these isolates did not contain the mutation conferring the amino acid substitution prior to cell culture. However, NA Q136K substitution has been reported previously in clinical specimens (Dapat et al., 2010). One A(H3N2) virus from the United States contained a NA E119V substitution which conferred HRI by oseltamivir but had no effect on susceptibility to the other NAIs. The E119V variant virus came from an oseltamivir-treated patient. The NA E119V substitution was detected in the corresponding clinical specimen. NA amino acid substitutions T148K, N329K, S331R and V215I were detected in viruses with IC₅₀ fold-change values that were close to the intersect between NI and RI categories for oseltamivir and/or zanamivir. NA T148K and D151G substitutions can emerge during culturing of A(H3N2) viruses (Tamura et al., 2013).

Of 604 B/Victoria-lineage viruses tested, 12 (2%) showed RI or HRI by one or more of the NAIs (Fig. 2 and Table 1). One B/Victoria-lineage virus from Bangladesh with a NA E117G substitution showed HRI by three of the four NAIs tested with a 2326-fold increase in peramivir IC₅₀, a 1010-fold increase in zanamivir IC₅₀, a 649-fold increase in laninamivir IC₅₀ and RI with a 12-fold increase in oseltamivir IC₅₀ compared to values for B/Victoria-lineage viruses displaying NI by the four NAIs (Table 1). NA amino acid substitutions N151S and D197N were detected in B/Yamagata-lineage HA – B/Victoria-lineage NA reassortants. All other substitutions identified in the isolates that were classified as having RI to zanamivir or peramivir require further analysis to assess whether the amino acid substitutions observed are responsible for the changes in IC₅₀. Three B/Victoria-lineage viruses had IC₅₀ values that were close to the intersect between NI and RI categories for zanamivir or peramivir, but no NA amino acid substitutions were detected on sequencing.

Eight (0.3%) B/Yamagata-lineage viruses out of 2311 tested showed RI or HRI by one or more of the NAIs (Fig. 2 and Table 1). One B/Yamagata-lineage virus from Macau, China contained an NA H273Y substitution which conferred HRI (103-fold increase) by peramivir and RI (7.1-fold increase) by oseltamivir but had no effect on susceptibility to zanamivir and laninamivir (Table 1). The two viruses carrying NA D197N substitution yielded different patterns of NI and RI by all NAIs, around the intersect between NI and RI categories; only for peramivir did both show RI. The NA D197N substitution has been observed before in B/Yamagata-lineage virus and shown to confer RI by zanamivir and peramivir (Oakley et al., 2010). One of the two D197N variant viruses, came from a patient during treatment with zanamivir, but no information was available for the second patient. NA amino acid substitutions N151T and S249N were detected in viruses with elevated IC₅₀ values (Table 1). The clinical specimen yielding the S249N variant also contained the substitution. The clinical specimen yielding the N151T variant did not contain the substitution. Further investigation is needed to assess the role of these substitutions in altering NAI susceptibility. Three B/Yamagata-lineage viruses had IC₅₀ values that were around the intersect between NI and RI categories, but NA-gene sequencing revealed no NA amino acid substitutions.

We screened NA sequences from viruses collected during the 2013–2014 period that had been deposited in the Global Initiative on Sharing All Influenza Data (GISAID) database, at www.gisaid.org, and the National Center for Biotechnology Information Influenza Virus Resource (NCBI-IVR), at www.ncbi.nlm.nih.gov/genomes/FLU/FLU.html, for amino acid substitutions which are known to confer clinical resistance (e.g. H275Y in A(H1N1)pdm09 viruses), or are known to occur clinically. A list of the amino acid substitutions that are relevant for clinical and surveillance interpretation, reviewed and updated in May 2014 by the WHO-AVWG, is available at: http://www.who.int/influenza/gisrs_laboratory/antiviral_susceptibility/nai_overview/en/ (accessed 21 November 2014). A total of 3206 sequences from GISAID and NCBI-IVR databases were analysed for the presence of key NA amino acid substitutions, following curation for quality, sequence length and duplication (Supplementary Table 2). The majority of sequences in the databases were from viruses collected through the GISRS network and characterised via phenotypic analysis in GISRS laboratories. Of 1328 N1 sequences from A(H1N1)pdm09 viruses, 165 (12%) contained NA H275Y substitution, of which 16 showed NA 275Y/H polymorphism and one carried NA H275Y/I223R dual substitutions (Table 2). The H275Y/I223R variant virus had been analysed by NA inhibition assay at a WHO CC and is included in Fig. 2 and Table 1. The 165 H275Y variant viruses were derived from 11 countries. Seven H275Y variant viruses were in the sequence database but not present in the WHO CCs IC₅₀ dataset. Phylogenetic analysis of NA (Fig. 4A) and HA (Fig. 4B) gene sequences showed, in both phylogenetic trees, that the H275Y variant viruses did not emerge from a common source. However, the H275Y variants from China were genetically similar to those from the Hokkaido cluster.

Analysis of the available sequences also identified 25 N1 sequences from A(H1N1)pdm09 viruses containing D199N substitution, of which seven originated from Bulgaria, and four influenza B NA sequences containing D197N substitution (Table 3). Twentyone of the 25 N1 D199N variants and three of the four B NA D197N variants had been analysed by NA inhibition assay at the WHO CCs. The A(H1N1)pdm09 viruses with NA D199N substitution showed NI. The influenza B viruses containing NA D197N substitution showed RI to one or more of the NAIs. These results are consistent with the AVWG table (http://www.who.int/influenza/gisrs_laboratory/antiviral_susceptibility/nai_overview/en/). Two N2 sequences contained E119V substitution, with no evidence for 119E/V mixtures, and both E119V variants were analysed by NA inhibition assay at the WHO CCs. One showed HRI to oseltamivir and is included in Fig. 2 and Table 1 and the other showed NI. In the AVWG table, it is listed that the NA E119V substitution confers 18-fold (RI) to 2057-fold (HRI) increases in oseltamivir IC₅₀.

The WHO-AVWG was able to perform this global analysis on influenza antiviral susceptibility thanks to NICs within the WHO GISRS fulfilling their Terms of Reference by collecting influenza virus positive clinical specimens and sharing a representative proportion of them, or viruses recovered, with the WHO CCs for further detailed characterisation (Kitler et al., 2002).

Based on our current analysis, approximately 98% of all viruses circulating during 2013–2014 were sensitive to all four NAIs and therefore these drugs remain an appropriate choice for the treatment and prophylaxis of influenza virus infections. However, a large community cluster of A(H1N1)pdm09 viruses with the NA H275Y substitution occurred in Hokkaido, Japan between November 2013 and February 2014. In 2011, a widespread community cluster of NA H275Y variant A(H1N1)pdm09 viruses occurred in Newcastle, Australia (Hurt et al., 2012). The latter H275Y variant viruses possessed V241I and N369K substitutions in the NA which partially overcame the detrimental effects of the H275Y substitution on virus fitness (Butler et al., 2014; Abed et al., 2014). Almost all recently circulating A(H1N1)pdm09 viruses possess the NA V241I and N369K substitutions, indicating increased risk of H275Y variant viruses emerging and spreading globally. The Hokkaido cluster viruses carried these two substitutions and shared NA N386K substitution with the H275Y variant viruses detected in China. Therefore, the H275Y variant viruses of the Hokkaido cluster and those of China may be derived from a common ancestor.

Despite low numbers of virus isolates from Africa, South East Asia and East Mediterranean regions being available for analysis, this pooled analysis of IC₅₀ data from the five WHO CCs offers the best opportunity to gain a robust global picture of the incidence of RI/HRI by NAIs and the relatedness of the NAI resistant viruses.

The majority of NA sequences in the GISAID sequence database are deposited by the WHO CCs. There are an increasing number of GISRS NICs with NA sequencing capability, but this is not reflected in the current analysis as many of the submitted sequences are incomplete and do not cover all known reduced NAI susceptibility conferring NA amino acid substitution positions. These laboratories should be encouraged to perform full-length NA-gene sequencing and submit available sequences in a timely manner as it would add significant value to global NAI susceptibility surveillance efforts.

The issues of incomplete clinical and epidemiologic data accompanying clinical specimens or virus isolates being referred to the WHO CCs remains a limitation in determining the significance of RI/HRI virus detection. NICs often cannot obtain this information on all samples, particularly when they are received from non-sentinel surveillance systems. Follow-up of individual samples requires significant, often one-to-one interaction with sub-national laboratories or physicians which is not possible for some NICs. In addition, NAI susceptibility analysis of samples for surveillance purposes may not be performed in a timely manner, making the collection of clinical and epidemiologic data for characterisation of the risk factors for NAI-RI/HRI virus infection difficult. Knowledge of whether the sample originated from patients in the community or hospitals can more easily be obtained. The WHO-AVWG is currently developing the WHO database (FluNet) to collect this basic information, together with H275Y screening data for influenza A(H1N1)pdm09 viruses, from all NICs who perform such testing. This basic information, which can also be supplied with the samples referred to the WHO CCs, does at least facilitate screening for increased incidence of viruses with HRI from untreated patients in the community, such as detected in Japan in 2013–2014 and Australia in 2011.

This is the second global update on influenza NAI susceptibility based on analysing data generated by WHO CCs on samples received from the GISRS laboratories. Overall, the proportion of viruses showing RI/HRI was similar between the two years, approximately 1% in 2012–2013 and 2% in 2013–2014. The slight increase in 2013–2014 is due to several clusters of untreated cases of A(H1N1)pdm09 viruses carrying NA H275Y substitution in China, Japan and the United States. Only 1% of A(H1N1)pdm09 viruses exhibited RI/HRI in 2012–2013, compared with 3% in 2013–2014. Rates of RI/HRI detection in A(H3N2) (0.4% vs 0.3%), B/Victoria- (1% vs 2%) and B/Yamagata- (0.3% both years) lineage viruses have remained very similar for the two seasons analysed to date. Prevalence of RI/HRI viruses has consistently been higher for B/Victoria-lineage over B/Yamagata-lineage viruses. This could be a reflection of the overall number of viruses analysed (more B/Yamagata-lineage in both seasons) but could also be an indication of a slightly higher tendency for B/Victoria-lineage NA to tolerate RI/HRI conferring substitutions. This could merit further investigation, taking into account reassortant events between the two lineages, notably the recent emergence of B/Yamagata-lineage HA – B/Victoria-lineage NA reassortants.