Widespread detection of highly pathogenic H5 influenza viruses in wild birds from the Pacific Flyway of the United States

A novel highly pathogenic avian influenza virus belonging to the H5 clade 2.3.4.4 variant viruses was detected in North America in late 2014. Motivated by the identification of these viruses in domestic poultry in Canada, an intensive study was initiated to conduct highly pathogenic avian influenza surveillance in wild birds in the Pacific Flyway of the United States. A total of 4,729 hunter-harvested wild birds were sampled and highly pathogenic avian influenza virus was detected in 1.3% (n = 63). Three H5 clade 2.3.4.4 subtypes were isolated from wild birds, H5N2, H5N8, and H5N1, representing the wholly Eurasian lineage H5N8 and two novel reassortant viruses. Testing of 150 additional wild birds during avian morbidity and mortality investigations in Washington yielded 10 (6.7%) additional highly pathogenic avian influenza isolates (H5N8 = 3 and H5N2 = 7). The geographically widespread detection of these viruses in apparently healthy wild waterfowl suggest that the H5 clade 2.3.4.4 variant viruses may behave similarly in this taxonomic group whereby many waterfowl species are susceptible to infection but do not demonstrate obvious clinical disease. Despite these findings in wild waterfowl, mortality has been documented for some wild bird species and losses in US domestic poultry during the first half of 2015 were unprecedented.


Results
Wild Bird Sampling. From December 20, 2014 through February 1, 2015 a total of 4,729 oral and cloacal swab were opportunistically collected from hunter-harvested birds. These samples were obtained from 6 states in the Pacific Flyway (Fig. 1, Table 1), with 33% of samples coming from California, 9% from Idaho, 9% from Nevada, 19% from Oregon, 7% from Utah, and 23% from Washington. Sample locations were spread across 45 counties and nearly 100 sample locations. A total of 33 species were sampled, the majority of which were dabbling ducks, known reservoirs for AIVs 6,20 .
There were no HPAIVs detected in the US during a previous wild bird surveillance effort that occurred from 2006-2011 22 (Table 3). Six out of 97 waterfowl samples (6.2%) were icA H5 positive (Canada goose, northern pintail, mallard, American wigeon) and 4 out of 35 raptor samples were icA positive (11.4%). Positive raptor species included peregrine falcon (Falco peregrinus), Cooper's hawk (Accipiter cooperii), and red-tailed hawk (Buteo jamaicensis). No HPAIV positive samples were identified from other water birds and no HPAIV positive samples were identified when only swab samples were submitted. Five of the icA H5 detections were made in waterfowl found dead at the index location from the ongoing morbidity and mortality event 17 . In addition 4 individual raptors found dead in 4 different Washington counties (Whatcom, Skagit, Grays Harbor, and Benton), and a single Canada goose found sick in Jefferson County, WA were icA H5 positive. None of the individual raptors or the single Canada goose that tested icA H5 positive were associated with morbidity or mortality events. H5N2 was detected in 7 individual birds and H5N8 was detected in 3 birds (Table 4). Phylogenetic analyses of the H5 clade 2.3.4.4 viruses detected as part of both the hunter-harvested surveillance effort and the morbidity and mortality surveillance reveal a high degree of relatedness (Figs 2 and 3). The H5N2, H5N8, and H5N1 viruses were co-circulating in wild birds sampled in the Pacific Flyway during the surveillance effort.

Discussion
The goals of this monitoring effort were to determine if Eurasian clade 2.3.4.4 viruses, newly introduced to North America, were present in US Pacific Flyway wild bird populations and if so, to better understand the extent to which they were circulating. The study was motivated by both the rapid spread of these H5 viruses in Asia and Europe and by the detection of these viruses in Canada 8,9,24  .3%) were equally prevalent in hunter-harvested birds. Three hunter-harvested samples also tested positive for the reassortant H5N1 (3/41, 7.3%). H5N2 (n = 7) and H5N8 (n = 3) were detected in samples from sick or dead birds.
The isolation of multiple reassortant viruses in wild birds is not surprising given the well-known occurrence of genetic recombination in influenza viruses, which occurs when more than one virus infects the same host cell. The segmented genome of influenza viruses allows segments from different viruses to be traded and in this case, introduction of the Eurasian H5 viruses into the North American influenza gene pool appears to have generated multiple novel subtypes (Eurasian-American lineage H5N2 and H5N1) 17,25 . Taiwan also identified novel reassortants after H5 clade 2.3.4.4 viruses were initially detected 26 . Although it is not surprising to see reassortant viruses, their emergence in what is believed to be a relatively short time frame is intriguing. de Vries et al. 11 noted that this emergence of new H5 combinations is unprecedented in H5N1 evolutionary history. In North America and in other places such as China and Taiwan 17 , rapid reassortment of clade 2.3.4.4 viruses may be facilitated through encounters with a large and diverse population of LPAIVs in wild birds, although this pattern was not necessarily seen in Europe where the association of this novel virus with wild bird populations does not appear to have spawned similar reassortant events in the same relatively short time frame.
Interestingly, more than half of the H5 detections (63/113 samples, 60%), and 13% (63 of 469) of all influenza A detections from hunter-harvest surveillance samples were Eurasian H5 HPAIVs, suggesting that the H5 clade 2.3.4.4 viruses were a substantial part of the influenza A gene pool in US wild bird populations during the winter of 2014-2015. Wild bird influenza surveillance efforts carried out from 2006-2011 22,23 found a significantly lower prevalence of H5 viruses in winter samples in the two most commonly sampled species (mallards and wigeon) from the Pacific Flyway when compared to this current effort. Samples collected from American wigeons accounted for almost half of all icA H5 detections (31 of 63) in wild waterfowl. American wigeons (n = 777) were the species most frequently infected with icA H5 despite mallards being the most frequently sampled species (1,410 mallard samples collected, 15 HPAIV detections). The high infection prevalence of icA H5 viruses detected during this wild bird sampling effort likely at least partially reflects the introduction of a new virus into a naïve population of birds that had no previous exposure to clade 2.3.4.4 viruses; however, it is unknown if these viruses will continue to circulate at high levels over time. In general, it is thought that HPAIV viruses are not maintained in wild bird populations. This is primarily based on HPAIV having rarely been isolated from wild birds prior to this emergence, even though prior research has found H5N1 outbreaks in domestic poultry to closely associate with known wild bird migration routes 27 . Although this effort documented the widespread occurrence of HPAIV in wild bird populations, the possibility of long-term persistence is unknown. Our understanding of HPAIV dynamics in wild birds in still limited despite a substantial increase in research on influenza over the last decade and this highlights the need for continued, consistent monitoring of HPAIV dynamics in wild birds over time.
Detection of icA H5 HPAIV in apparently healthy wild birds 10 , combined with laboratory studies in avian species that demonstrate low morbidity and mortality, reinforces the likely role of wild waterfowl in the emergence and geographic spread of these viruses 4,5,19 . Wild waterfowl infected with icA H5 viruses are likely capable of moving these viruses short distances, and possibly even long distances, along migration routes; however, the length of time that an infected bird sheds an influenza virus, which is typically up to 10-14 days in experimental infection with other goose/Guangdong/96 lineage isolates in ducks 28,29 , also plays a role in the likelihood of virus introduction via long-distance migration.
Despite  30,31 . The mechanisms responsible for the different scale of commercial poultry outbreaks in the Pacific Flyway compared to the other flyways is largely unknown, although commercial outbreaks were shown to be a combination of both independent/point source introductions along with common source or lateral spread 30 . The latter would be independent of wild bird influenza dynamics. Furthermore, the wholly Eurasian H5N8 virus accounted for almost 40% of HPAIVs in wild birds, but only 4 of the 223 documented domestic poultry outbreaks in the US between December of 2014 and June of 2015 were caused by H5N8 30 . Laboratory studies suggest that these viruses were well-adapted to the waterfowl host, with experimentally infected mallards remaining asymptomatic while easily transmitting the virus to conspecifics. This is in contrast to the less efficient transmission to contacts seen in infected chickens 32  This effort focused on waterfowl because they are considered a primary influenza A virus reservoir, and while these data suggest that some waterfowl species are not negatively impacted by infections with Eurasian H5 clade 2.3.4.4 viruses, it appears that other wild species may suffer severe morbidity and mortality. The first detection of icA reassortant H5N2 in wild birds in North America was obtained during a waterfowl mortality event at Wiser Lake, Whatcom County, Washington; although the cause of the mortality event was attributed to aspergillosis with the icA H5 infection reported as a secondary finding 17 . Concurrent to this waterfowl mortality event, a captive-reared gyrfalcon died from icA H5N8 after being fed wild wigeon from the same area 17 . From December through February, morbidity and mortality testing of wild birds in Washington increased; 150 additional waterfowl, other water birds, and raptors were tested as part of this effort. From these, icA H5 was detected in 6 waterfowl, (5 were ducks from Wiser Lake, the index location and included the northern pintail reported by Ip et al. 17 and one from a Canada goose in a second county). Four raptors from 4 different counties were also positive and represented 11% of all raptors (n = 35) tested. Lesions in the lung and heart consistent with HPAI infection were observed in all four raptors. The cause of death in the five ducks found dead at the index location was determined to be aspergillosis with a HPAI non-lethal co-infection. The cause of death in the single Canada goose is suspected  to be due to HPAI, but the role of HPAI infection was difficult to interpret due to freeze artifact and a co-bacterial infection in the lungs. Other sick and dead Canada geese and raptors testing positive for the icA H5 viruses have been reported in other states, suggesting they may be good targets for morbidity and mortality surveillance to identify presence of icA H5 viruses. This surveillance effort that began immediately upon reports of HPAIV detection in British Columbia allowed early detection of clade 2.3.4.4 viruses in the US wild bird populations. This rapid response to collect and test wild bird samples offered an unprecedented opportunity to better understand the dynamics of a novel virus introduced into a naïve wild bird population with its own endemic and diverse influenza A gene pool. Passive surveillance of apparently healthy hunter-harvested wild birds was used in conjunction with morbidity and mortality surveillance and both methodologies provided valuable data. Continued surveillance and characterization of influenza A in wild birds is suggested in order to monitor virus evolution, to understand risk pathways for introduction, and to assess the emergence of mutations that may be relevant for veterinary and public health 15 .

Methods
Sample Areas. Wild bird sample collection focused on the Pacific Flyway of the US. Waterfowl and water bird migration in North America generally consists of north-south seasonal movements between breeding grounds and wintering areas and while the flyway boundaries are not biologically fixed or sharply defined, the flyways represent the prominent movement pathways of migratory waterfowl 22 . Band-recovery data on waterfowl species of interest show that migratory waterfowl move throughout this flyway. Ten priority sampling areas were chosen in the Pacific Flyway. These areas were chosen based on previous research that identified geographic clusters of low pathogenic AIV in wild birds 23,33 and known high concentration wintering areas for waterfowl in the Pacific Flyway. Sample areas roughly corresponded to a broad watershed scale. Sample sizes were based on 95% confidence in detecting 1 positive bird out of a population of 10,000 or more birds 34 . Expected prevalence was 1% based on previously published values of LPAIV 23 , and sensitivity and specificity of the diagnostic assays was 86.3% and 100% respectively (Janice Pedersen, personal communication). Sample Collection. More than 99% of samples came from hunter-harvested waterfowl and < 1% came from birds removed as part of permitted wildlife damage management activities. Sampling of hunter-harvested birds was voluntary on the part of waterfowl hunters and birds were either swabbed immediately in the field or shortly after if hunters called agency field personnel to arrange sampling. Samples collected at hunter check stations were collected in accordance with the guidelines and regulations set forth by the United States Fish and Wildlife Service (USFWS) and with the permission of participating hunters. For most birds, separate swabs were used to collect an oropharyngeal and cloacal sample. The two swabs were placed in a single uniquely labeled cryovial containing 4 ml of viral transport media (VTM) or 3 ml of brain-heart infusion media (BHI). In some cases, waterfowl carcasses were damaged and in those instances only one swab sample was collected. In 4 cases a single swab was used to first swab the oropharyngeal cavity and then the cloacal cavity before being placed in the VTM. Used sample vials were held at 4 °C for up to 48 hours or were placed in liquid nitrogen vapor and frozen before being shipped overnight to a reference laboratory.
Morbidity and Mortality Sampling. Between December 4, 2014 and Feb 28, 2015 we obtained swab samples or whole carcasses from waterfowl, other water birds, and raptors found sick and dead in Washington State. Collection protocols for wild bird morbidity and mortality events were approved under USFWS permit #MB084762-1. Birds that were swabbed only had a single combined oral and cloacal swab submitted to the U.S. Geological Survey's National Wildlife Health Center for avian influenza testing as described previously (for the live bird surveillance). In cases where whole carcasses were submitted, separate tracheal and cloacal swabs samples were obtained at minimum. Additional tissues tested could include trachea, lung, combined trachea-lung, brain, heart, and others, as identified by pathologists conducting a cause of death necropsy on the submitted carcass and varied by individual case. Tissues were processed and tested as previously described for hunter-harvest surveillance swabs.   Presumptive positive findings were confirmed at the National Veterinary Service Laboratories-USDA-APHIS in Ames, Iowa, which is the US reference laboratory for avian influenza. Samples were initially screened by rRT-PCR utilizing a test which targets the influenza matrix (M) gene using the USDA standard procedure (SOP-AV-0001) 35,36 . Further testing on M-gene presumptive samples was conducted using H5 and H7 subtype rRT-PCR assays as a general surveillance tool 36 . A highly specific H5 IcA rRT-PCR (D. Suarez, personal communication) was also run on M gene positive samples to detect the Eurasian H5 clade 2.3.4.4 gene to distinguish IcA positive samples from those with the North American H5 gene. Virus isolation in embryonated chicken eggs was attempted on M-gene positive samples by standard methods 37,38 . Subtypes were identified by either serological assays (hemagglutination inhibition assay, neuraminidase inhibition assay) on isolates, or by gene sequencing on swab material or isolates using standard methods [39][40][41] . The pathotype classification was inferred from the HA gene proteolytic cleavage site sequence as defined by the World Organization for Animal Health (OIE) 42 . Select viruses were processed for in vivo pathotyping in specific pathogen free chickens at the NVSL in accordance with OIE guidelines 42 .
Analyses. Prevalence and 95% confidence limits were calculated using an exact binomial calculation.
Comparisons between H5 prevalence during a prior wild bird surveillance effort and this current surveillance used a general linearized model with a binary distribution and a logit link function. Odds ratios were calculated using a Tukey-Kramer adjustment. All analyses were run in SAS (SAS v.9.2, Cary, North Carolina, USA). Genetic sequences were assembled using Clustal W. Sequences were only available from a subset of samples. All positions containing gaps and missing data were eliminated. Evolutionary analyses were conducted in MEGA6 43 . The evolutionary history was inferred by using the Maximum Likelihood method based on the Hasegawa-Kishino-Yano (HKY) model 44 . The tree with the highest log likelihood is shown. The percentage of trees in which the associated taxa clustered together is shown next to the branches. Initial trees for the heuristic search were obtained by applying the Neighbor-Joining method to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach. A discrete Gamma distribution was used to model evolutionary rate differences among sites. Analysis of the HA gene involved 85 nucleotide sequences; analysis of the NA gene involved 57 nucleotide sequences.