2.1 Virus isolation and whole-genome sequencing

In Japan, tracheal or cloacal swabs collected at poultry farms with suspected HPAIV infections were inoculated into embryonated chicken eggs for virus isolation at the diagnostic laboratories of the livestock health center of the prefecture in which the farm was located. Samples of allantoic fluid that showed HA activity against chicken red blood cells and swab samples from dead wild birds in Niigata, Saitama, Ibaraki, Toyama and Tochigi prefectures were submitted to the National Institute of Animal Health (Tsukuba, Ibaraki, Japan) for diagnosis.

The activities on sampling from wild birds in Russia were approved by the Committee on Biomedical Ethics of the Federal Research Center of Fundamental and Translational Medicine (No. 2017-16). No specific permission was required for sample collection from wild birds killed by local hunters in compliance with the Russian Federation hunting laws. Cloacal swabs were collected at the Novosibirsk region, Omsk region, Sakhalin Island, Yakutia, Primorje and Taimyr peninsula. Samples were inoculated into embryonated chicken eggs for virus isolation at the Federal Research Center of Fundamental and Translational Medicine. Samples of allantoic fluid that showed HA activity against chicken red blood cells were examined for H5 subtype using real-time polymerase chain reaction (AmpliSens Influenza virus А H5N1-FRT PCR kit; AmpliSens, Russia) and exported to the National Institute of Animal Health for further identification and study.

The whole genomes of the isolated viruses were obtained by next-generation sequencing (Miseq, Illumina, San Diego, CA, USA). RNA was extracted from the isolated viruses by using a RNeasy Mini kit (Qiagen, Hilden, Germany). cDNA libraries for next-generation sequencing were prepared by using a NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA). In total, 10 pM of the synthesized cDNA libraries were mixed with 10 pM of the PhiX control (Illumina) and sequenced by using a Miseq Reagent Kit version 2 (Illumina). Consensus sequences were generated by using Workbench software (version 9.5.3; Qiagen) or FluGAS software (version 2.2.5; World Fusion, Tokyo, Japan), as described previously (Mine et al., 2020). The viral sequences determined during the present study have been deposited in the GISAID database (http://platform.gisaid.org); the accession numbers of the representative viruses are listed in Table 1.

2.2 Phylogenetic analysis

In Japan in winter 2020−2021, a total of 110 HPAI reports were made, comprising 52 outbreaks at poultry farms and 58 positive identifications in wild birds or the environment. In the present study, isolates of all 52 of the strains that caused outbreaks at poultry farms and 11 of the strains that were detected in wild birds were used for phylogenetic analysis (Table 1). We downloaded all full-length sequences of the AIVs from the NCBI (Influenza Virus Resource; https://www.ncbi.nlm.nih.gov/genomes/FLU/Database/nph-select.cgi?go = database) and GISAID databases on 15 March 2021. BioEdit version 7.2.5 (Hall, 1999) and MAFFT version 7.490 (Katoh & Standley, 2016) were used to align the GISAID sequences with those of AIVs stored at the National Institute of Animal Health in Japan and at the Federal Research Center of Fundamental and Translational Medicine, Novosibirsk, Russia. After alignment, sequences with ambiguous nucleotide bases were removed. The number of sequences used for the phylogenetic analyses were as follows: polymerase basic protein (PB)2, 104,822 sequences; PB1, 103,084; polymerase acidic protein (PA), 107,834; H5, 12,043; nucleoprotein (NP), 107,840; N8, 5549; matrix protein (MP), 109,445; and non-structural protein (NS), 110,696. Maximum likelihood trees based on the aligned sequences were constructed by using FastTree version 2.1.10 (Price et al., 2010) with the 1000 resampling for the statistical support. Location-annotated maximum clade credibility trees for viruses that were most related to the strains identified in the present study were constructed according to Bayesian stochastic search variable selection by using the Bayesian Evolutionary Analysis by Sampling Tree package (BEAST) version 1.10.4 (Drummond et al., 2012), as described previously (Mine et al., 2019b). An asymmetric substitution model with Bayesian stochastic search variable selection and a strict clock model were used for selected taxons in each tree to calculate Bayes factors. Same number of sequences per discrete trait (location) were selected as described in Table S2. Each calculation was set as 1 × 10⁸ to 1 × 10⁹ steps in length, where the number of steps was determined as that needed to obtain an effective sample size of more than 200. The output log file was visualized by spatial phylogenetic reconstruction of evolutionary dynamics using Data-Driven Documents (SPreaD3) version 0.9.7 (Bielejec et al., 2016). Lines with Bayes factors of 3.0 or more and posterior probabilities of .5 or more are indicated in the visualizations.

3.1 Phylogenetic analysis of Japanese H5N8 isolates from the 2020−2021 winter season

Maximum likelihood analysis using the HA gene of the Japanese H5N8 HPAIVs isolated in winter 2020−2021 revealed that the viruses belonged to two clusters in clade 2.3.4.4b (Figures 1, 2 and S1): one group was primarily composed of H5N8 HPAIVs that circulated in Europe in early 2020 (G1 cluster) and the other was primarily composed of H5N8 HPAIVs that circulated in Europe in late 2020 (G2 cluster). The average nucleotide identity between the Japanese H5N8 HPAIVs in the G1 cluster and those in the G2 cluster was 95.8%.

The G1 cluster was found to contain progeny strains of European viruses that were isolated in Germany, Poland and Czech Republic in January–March 2020, and these progeny strains were found in Japan and Korea in September 2020 to March 2021 (Figure 1). In addition, a time-measured phylogenetic analysis estimated that the Asian H5N8 HPAIVs in the G1 cluster were diverged from the European HPAIVs on 8 August 2020 (95% highest posterior density interval [HPD]: 20 June 2020 to 15 September 2020) (Figure 1). Together, these results indicate that the viruses in the G1 cluster most likely spread across the Eurasian continent in an eastward direction.

Although the Japanese viruses in the G1 cluster clustered together with European isolates, those in the G2 cluster clustered together not only with European isolates but also with isolates from Kazakhstan and the Novosibirsk and Omsk region of Siberia (Figure 2). The strains from Kazakhstan, Novosibirsk and Omsk region were isolated in summer (25 September 2020, 15−22 September 2020 and 13−17 August 2020, respectively), whereas those from Asia and Europe were isolated in autumn and winter. Furthermore, the Asian viruses were estimated to have diverged from the European viruses on 29 July 2020 (95% HPD: 16 June 2020 to 16 September 2020). Together, these results suggest that the viruses in the G2 cluster disseminated from breeding sites to both the eastward to Asia and westward to Europe.

3.2 Identification of H5 HPAIVs in Siberia

Surveillance for AIVs and avian paramyxoviruses among wild birds in Siberia resulted in the isolation of several AIVs. Of these, two AIVs of the H5N2 and H5N8 subtypes (A/common_teal/Chany Lake/213/2020_[H5N2] and A/mallard/Novosibirsk region/3509k/2020_[H5N8], respectively) were isolated from wild birds in southwestern Siberia (Figure 3). Phylogenetic analysis of the HA gene of the H5N2 and H5N8 AIVs placed the viruses in the G2 cluster close to H5N8 HPAIVs that were circulating in Europe in winter 2020−2021 (Figure S1). These AIVs also possessed the multiple basic amino acid sequence REKRRKR in proteolytic cleavage site of their HA genes, which is associated with high pathogenicity in chickens defined by OIE-FAO (https://www.offlu.org/wp-content/uploads/2021/01/Influenza_A_Cleavage_Sites-1.pdf). Although none of the H5N2 HPAIVs in the G2 cluster have been detected in poultry or wild birds in the Novosibirsk region, some H5N8 HPAIVs in the G2 cluster have been detected in poultry in the region (Lewis et al., 2021; Sobolev et al., 2021).

3.3 Gene constellations of H5 HPAIVs isolated in Japan and Russia

Phylogenetic analyses based on all RNA segments of the Japanese H5 HPAIVs revealed that the viruses showed five genotypes (E1, E2, E3, E5 and E7) (Table 1 and Figure 3), in accordance with a previously reported classification (Baek et al., 2021).

The Japanese viruses in the G1 cluster showed four genotypes: E1, E3, E5 and E7. The E1 genotype, which was estimated to be the ancestor of the E3, E5 and E7 genotypes according to the gene constellations and the period of detection (Baek et al., 2021), was initially found in Europe in early 2020. The E3, E5 and E7 genotypes were characterized by NP genes that were genetically distinct from those of the E1 genotype. Furthermore, the E3 genotype possessed genetically distinct PB2 and PB1 genes, and the E5 genotype possessed genetically distinct PB2 and PA genes. Together, these findings suggest that after emergence of the E1 HPAIVs in Asia, they exchanged internal genes with Eurasian AIVs.

The Japanese viruses in the G2 cluster all showed the same genotype, E2, which was the same as that of viruses circulating in Korea in winter 2020−2021 and Europe in late 2020 (Baek et al., 2021). However, the Russian viruses in the G2 cluster isolated from wild birds through the surveillance in the present study showed different genotypes compared with the E2 genotype: A/mallard/Novosibirsk_region/3509k/2020_(H5N8) possessed a genetically distinct PA gene, and A/common_teal/Chany_Lake/213/2020_(H5N2) possessed genetically distinct NA, PB1, PA and NP genes.

In Japan in winter 2020−2021, H5N8 HPAIVs belonging to the E3 genotype caused influenza outbreaks in 13 prefectures, whereas those belonging to the E1, E2, E5 and E7 genotypes caused outbreaks in 2, 3, 1 and 4 prefectures, respectively (Figure 3).

3.4 Phylogeographic analyses based on the internal genes of H5 HPAIVs and AIVs

Our phylogenetic analyses using the full-genome sequences of viruses indicated that the Japanese H5 HPAIVs isolated in winter 2020−2021 emerged via genetic reassortment among various AIVs. We therefore performed several phylogeographic analyses based on the H5 HA gene and newly introduced internal genes to estimate when and where the reassortment events occurred.

Phylogeographic analysis using the H5 HA gene of the viruses in the G1 cluster showed high Bayes factor between Europe and Asia (Figure 4(a)). Phylogeographic analysis of the H5 HA gene of the viruses in the G2 cluster showed moderate Bayes factors between Asia and the southwestern Siberia such as Omsk, Novosibirsk regions and Kazakhstan, and high Bayes factors between southwestern Siberia and Eastern Europe (Figure 4(b)). Phylogeographic analyses based on the N8 NA gene and internal genes of HPAIVs showing the E1 and E2 genotypes showed similar results to those obtained using the H5 HA gene (data not shown).

Phylogeographic analysis using the NP gene of HPAIVs showing the E3, E5 and E7 genotypes revealed high Bayes factors from Siberian region to Asia (Figure 5). Bayes factors from Buryatia/Amur to many areas such as Southwestern Siberia, China/Mongolia, Sakhalin/Primorje and Japan/Korea were more than 3.0 (Table S1). Phylogeographic analysis using the PA gene of HPAIVs showing the E5 genotype revealed that Asia showed high Bayes factors with areas in eastern Siberia such as Yakutia, Amur and Primorje (Figure 6). This analysis of the PA gene suggests that the AIVs disseminated from Asia in the south-north direction (Table S1).

Phylogeographic analysis using the PB2 gene of HPAIVs showing the E5 genotype showed high Bayes factors from Asia to Novosibirsk, and high Bayes factors were found between Europe and Novosibirsk (Figure 7). Phylogeographic analysis using the PB2 gene of HPAIVs showing the E3 genotype revealed high Bayes factors from Japan/Korea to Buryatia/Primorje (Figure 8), and these findings were consistent with those conducted using the PB1 gene of HPAIVs showing the E3 genotype (data not shown).