Australia as a global sink for the genetic diversity of avian influenza A virus

Most of our understanding of the ecology and evolution of avian influenza A virus (AIV) in wild birds is derived from studies conducted in the northern hemisphere on waterfowl, with a substantial bias towards dabbling ducks. However, relevant environmental conditions and patterns of avian migration and reproduction are substantially different in the southern hemisphere. Through the sequencing and analysis of 333 unique AIV genomes collected from wild birds collected over 15 years we show that Australia is a global sink for AIV diversity and not integrally linked with the Eurasian gene pool. Rather, AIV are infrequently introduced to Australia, followed by decades of isolated circulation and eventual extinction. The number of co-circulating viral lineages varies per subtype. AIV haemagglutinin (HA) subtypes that are rarely identified at duck-centric study sites (H8-12) had more detected introductions and contemporary co-circulating lineages in Australia. Combined with a lack of duck migration beyond the Australian-Papuan region, these findings suggest introductions by long-distance migratory shorebirds. In addition, on the available data we found no evidence of directional or consistent patterns in virus movement across the Australian continent. This feature corresponds to patterns of bird movement, whereby waterfowl have nomadic and erratic rainfall-dependant distributions rather than consistent intra-continental migratory routes. Finally, we detected high levels of virus gene segment reassortment, with a high diversity of AIV genome constellations across years and locations. These data, in addition to those from other studies in Africa and South America, clearly show that patterns of AIV dynamics in the Southern Hemisphere are distinct from those in the temperate north.

150 migrate beyond the Australian-Papuan Region [33,34]. Indeed, of the key AIV reservoir 151 avian taxa, only members of the Charadriiformes, notably the waders (families 152 Scolopacidae and Charadriidae), migrate and link Australia with Eurasia and North 153 America [35][36][37]. These species may also be less susceptible to AIV infection than some 154 other species [38]. The ecology of this migratory system has the potential to limit viral gene   (Table S1). We recovered the full AIV genomes from 242 of the samples. In 177 some cases, we recovered partial AIV genomes consisting of gene segments with  [40,41]. Metadata is available in Table S1 and a  215 detailed plot illustrating exact virus sample collection dates and locations can be found in Fig S2.   216 217 Across the data set as a whole, we identified 14 different HA subtypes and all nine 218 different NA subtypes, comprising 58 HA-NA combinations. We did not detect avian HA 219 subtypes H14 and H15, and only a single case each of H13 and H16. The most common 220 subtypes in our data set were H1N1 (n = 14), H3N8 (n = 23), H4N6 (n = 16), H5N3 (n = 221 15), H6N2 (n = 23), H9N2 (n = 14) and H11N9 (n =19) (Fig 2A). These subtypes each 222 comprised 5-10% of the subtype combinations. An analysis of HA-NA linkage by 223 assessing the Pearson's residuals following a Chi-squared test revealed a strong positive 224 association between H1-N1, H3-N8, H4-N6, H8-N4, H11-N9 and H12-N5 ( Fig 2B) 280 Sequence data for this H7 lineage from wild birds has only been available since 2007 due 281 to very limited sampling and sequencing of wild birds in earlier years ( Fig S8). 287 and H10 were each in entirely separate lineages from Australian sequences (Fig S3-S4, 288 S6-S8, S11). This is likely due to limited bird migration involving some shorebird species, 289 between New Zealand and Australia [46].

291
Overall, Australia appears to be a sink for Eurasian AIV diversity. Although we identified 292 multiple viral introductions from Eurasia to Australia, in the entire data set there were only 293 two examples of viruses from Australia being introduced to Eurasia. These comprised the 294 H4(N8) subtype (Fig 3B), and one N7 sequence ( Fig S21). Notably, each of these events 295 involved the detection of Australian lineage viruses in Charadriiformes (gulls and 296 shorebirds) in Japan. Overall, all viral introductions stemmed from Eurasian lineages with 297 the exception of H10 and H12 that showed introductions from North American lineages.
298 For H8 and one H9 lineage, the most closely related reference viruses were sampled in 299 Europe. However, due to possible under-sampling and/or under-representation of viral 300 diversity in wild birds in Asia it cannot be concluded that these lineages were seeded 301 directly from Europe (Fig S9-S10). 332 sampling has been unable to detect the index viruses seeding local lineages (Fig 4).  378 Tasmania) that occurred consistently across all of the gene segments examined (Fig 6).
379 Using the largest Australian NP gene lineage, we found more than 10 potential migration 380 events between Victoria and South Australia and between Victoria and Tasmania, 381 suggesting high levels of connectivity between these sampling locations. We also found 382 evidence of movement between temperate Western Australia and the southeastern states 383 (Victoria, South Australia, Tasmania), and between Queensland and the southeastern 384 states, although this was only detected in the NP segments and in two of the HA/NA 385 subtypes analysed. As only limited sequences were available from tropical Australia 386 (northern Queensland, Northern Territory and northeastern Western Australia), migration 387 events to/from these locations were not well estimated in our analyses. However, for the 388 largest NP lineage, a number of potential migration events between temperate and tropical 389 Australia were observed (Fig S26). Potential migration events were also detected between 390 the sampled tropical locations. Although it is likely that we have underestimated the 391 migration events due to poor temporal and spatial coverage, the migration events had 392 strong Bayes Factor support ( Fig S25). Importantly, these analyses also did not record >1 393 migration event or >10 Bayes Factor between all locations that were included in each tree 394 as a default. For example, despite being included in all eight analyses, we only detected 395 significant migration events (or >10 Bayes Factor) to/from Western Australia in the H6 and 396 N6 lineages, and the two NP lineages (Fig 6, Fig S25).

397
398 Overall, analysis suggested Victoria was consistently a net exporter as most migration 399 events originated from the state. Specifically, Victoria played a role as a net exporter in H5, 400 H7, N6, N8, and both NP lineages. South Australia also played a role as an exporter (H4, 401 H6, H7, N6, and both NP lineages), although we detected both import and export events 402 from this state across most analyses. Temperate Western Australia was a net importer of 403 AIV, although as with South Australia, we detected both importation and exportation 404 events across the analyses. A positive association between Markov rewards and the 405 number of exportation events may also be evidence of sampling bias. For example, in the 406 case of H4, H5 and H7, Victoria had substantially more sequences available as compared 407 to other sampling locations and was identified as a net exporter. In these cases, the high 408 number of exportations relative to importation events may be due to sampling biases (Fig   409 S27). 425 Arrows indicate the direction of the migration event. As NP has more than a single discrete Australian 426 lineage, we have generated two independent maps reflecting the 2 largest Australian lineages of NP (Fig  427 S25). Maps illustrating Bayes Factors, also generated using BSSVS can be found in Fig S26, and Markov 428 rewards also generated in this analysis are presented in Fig S27. (Fig 7).       Figure S28. Diversity of the "internal" segments for each sampled location 1146 Table S1. Metadata associated with viral genomes generated in this study 1147 Table S2. Details of mixed viruses detected in this study