Avian influenza viruses in wild birds in Canada following incursions of highly pathogenic H5N1 virus from Eurasia in 2021–2022

ABSTRACT Following the detection of novel highly pathogenic avian influenza virus (HPAIV) H5N1 clade 2.3.4.4b in Newfoundland, Canada, in late 2021, avian influenza virus (AIV) surveillance in wild birds was scaled up across Canada. Herein, we present the results of Canada’s Interagency Surveillance Program for Avian Influenza in Wild Birds during the first year (November 2021–November 2022) following the incursions of HPAIV from Eurasia. The key objectives of the surveillance program were to (i) identify the presence, distribution, and spread of HPAIV and other AIVs; (ii) identify wild bird morbidity and mortality associated with HPAIV; (iii) identify the range of wild bird species infected by HPAIV; and (iv) genetically characterize detected AIV. A total of 6,246 sick and dead wild birds were tested, of which 27.4% were HPAIV positive across 12 taxonomic orders and 80 species. Geographically, HPAIV detections occurred in all Canadian provinces and territories, with the highest numbers in the Atlantic and Central Flyways. Temporally, peak detections differed across flyways, though the national peak occurred in April 2022. In an additional 11,295 asymptomatic harvested or live-captured wild birds, 5.2% were HPAIV positive across 3 taxonomic orders and 19 species. Whole-genome sequencing identified HPAIV of Eurasian origin as most prevalent in the Atlantic Flyway, along with multiple reassortants of mixed Eurasian and North American origins distributed across Canada, with moderate structuring at the flyway scale. Wild birds were victims and reservoirs of HPAIV H5N1 2.3.4.4b, underscoring the importance of surveillance encompassing samples from sick and dead, as well as live and harvested birds, to provide insights into the dynamics and potential impacts of the HPAIV H5N1 outbreak. This dramatic shift in the presence and distribution of HPAIV in wild birds in Canada highlights a need for sustained investment in wild bird surveillance and collaboration across interagency partners. IMPORTANCE We present the results of Canada’s Interagency Surveillance Program for Avian Influenza in Wild Birds in the year following the first detection of highly pathogenic avian influenza virus (HPAIV) H5N1 on the continent. The surveillance program tested over 17,000 wild birds, both sick and apparently healthy, which revealed spatiotemporal and taxonomic patterns in HPAIV prevalence and mortality across Canada. The significant shift in the presence and distribution of HPAIV in Canada’s wild birds underscores the need for sustained investment in wild bird surveillance and collaboration across One Health partners.

selected for AIV testing and pooled into a single vial containing appropriate transport medium.Vials were stored at temperatures of at least −20°C until testing and −70°C to −80°C when possible (Document S1, Appendix G).For all wild bird mortality events during the study period, swabs were collected from a subset of birds for laboratory analysis to ascertain the presence or absence of HPAIV.When resources and capacity were available, gross and histologic examination of carcasses with HPAIV-positive swab results was undertaken at the CWHC or provincial or territorial laboratory to determine if lesions consistent with HPAIV were present, if HPAIV was the cause of death, and to help rule out false positives, particularly in the case of new species or new locations.In some cases, tissue samples (brain, lung, and intestine) collected during post-mortem examination were submitted in lieu of swabs for AIV testing.

Live and hunter-harvested bird surveillance
Live wild birds were sampled by the ECCC, the United States Fish and Wildlife Service, and provincial or territorial, and Indigenous or academic partners.Sampling opportuni ties were reviewed periodically to prioritize sample collection and ensure they were in line with surveillance objectives, sample size recommendations (Document S1, Appendix E), capacity, and resources.All live bird sampling was performed in accordance with approved animal use protocols, appropriate federal or provincial wildlife permits where applicable, and appropriate safe work procedures.Samples from harvested birds were provided by permitted and Indigenous harvesters.For the purposes of this study, harvested birds are considered apparently healthy prior to harvest and are therefore categorized with live birds in our analyses.Live and harvested birds were sampled for AIV as described above.

Laboratory analyses
Real-time reverse transcriptase polymerase chain reaction (RT-PCR) testing of swab samples and tissue samples was performed at the diagnostic laboratories of the Canadian Animal Health Surveillance Network (CAHSN), which is a network of federal, provincial, and university animal health laboratories across Canada with the central reference laboratory operating from the National Centre for Foreign Animal Disease, Canadian Food Inspection Agency (NCFAD-CFIA) in Winnipeg, Manitoba.The CAHSN standard protocol for the detection of type A influenza viruses and avian H5 and H7 hemagglutinin subtypes by RT-PCR Assay (version 3, January 2020) was utilized (16,17).In this protocol, a RT-PCR assay based on the use of fluorescent 5′ nuclease oligoprobes (hydrolysis probes) was used for the rapid detection of group A specific Matrix (M1), H5, and H7 hemagglutinin subtype AIV sequences.The matrix assay also employs the use of an exogenous armored RNA-Enterovirus internal control for the verification of the potential inhibition of the RT-PCR or inefficient extraction.The matrix RT-PCR is designed to detect M1 gene sequences of all group A influenza viruses (birds and mammals).The H5NA/EA RT-PCR Assay in the CAHSN protocol is capable of detecting most North American and Eurasian lineage H5 AIVs, including different clades of GsGd lineage Eurasian H5Nx viruses.The H7 2013 RT-PCR has been re-designed to detect H7 subtype viruses belonging to both Eurasian and North American lineage AIVs.
Automated nucleic acid extraction from samples was performed using Magnetic Particle Processors (MagMax, KingFisher, Roche, and others) and appropriate kits, while the manual nucleic acid extraction was done using a Qiagen vacuum manifold and Qiagen Viral RNA MiniKit.The following RT-PCR Systems were used: Applied Biosystems 7500/7500 Fast, Roche Light Cycler 480, BioRad CFX 96, and Strategene MX3005.The interpretation of test results as outlined in the CAHSN standard protocol was followed.
All samples that were positive for Group A specific Matrix RT-PCR and positive, suspect, or non-negative for H5 or H7 were sent to NCFAD-CFIA for confirmatory testing and further genomic characterization.The designation of HPAIV or LPAIV was determined after sequencing the HA cleavage site motif at NCFAD-CFIA, which is the Canadian and World Organization for Animal Health reference center for avian influenza.
We assigned AIV sample status as "confirmed or suspect HPAIV-positive" or "con firmed or suspect LPAIV-positive." The former includes cases confirmed H5 HPAIV-pos itive (highly pathogenic virus of the subtype H5Nx confirmed by NCFAD-CFIA) and samples that were non-negative [i.e., positive (cycle threshold, Ct <36) and suspect (Ct ≥ 36 but <40)] on H5 PCR at the regional laboratory or NCFAD-CFIA, but sequenc ing was not possible due to poor sample quality or in a few cases, testing is not yet completed.Virus isolation was attempted on most H5-positive samples at NCFAD-CFIA.Virus isolation was conducted in 9-to-11-day-old, specific pathogen-free embryonated chicken eggs; in some cases, the samples were negative, and in a few cases, testing is not yet complete.These samples were categorized as suspect H5 HPAIV-positive because there were few LPAIV H5 detections in Canada over the study period (Y.Berhane, personal communication).Confirmed or suspected LPAIV-positive include cases with LPAIV confirmed by NCFAD-CFIA and those that tested non-negative on matrix RT-PCR and negative on H5 RT-PCR at the regional laboratory or NCFAD-CFIA.In the latter cases, additional virus isolation and sequencing were not possible or were not yet completed.

Wild bird surveillance data
Metadata and preliminary matrix, H5, and H7 RT-PCR results for sick and dead birds across Canada were received from surveillance partners.Metadata and diagnostic results associated with live and harvested wild birds were managed internally within ECCC.Confirmatory diagnostic results were compiled by NCFAD-CFIA.Data were regularly merged and structured, and samples were assigned additional identifiers (e.g., taxo nomic family, flyway, and watershed), resulting in a compiled national AIV surveillance data set (12).Exact (Clopper-Pearson) confidence intervals were calculated in R version 4.2.2 (31 October 2022).The data are presented for surveillance conducted between 1 November 2021 and 1 November 2022, inclusive, extracted from the full data set on 15 May 2023.

Influenza virus genome sequencing and assembly
AIV full genome segments were amplified either directly from clinical specimens or isolates as described previously (18).High-throughput sequencing was performed either on an Oxford Nanopore Technologies (ONT) GridION sequencer and R9.4.1 Flow Cell following library (n = 921) construction using the ONT Rapid Barcoding Kit (SQK-RBK004 or SQK-RBK110.96) or an Illumina MiSeq (n = 257) using the Nextera XT Library Prepa ration kit (Illumina) following the manufacturer's protocol.The Hamilton Microlab Star Robot was used for Illumina library preparation prior to sequencing with Illumina MiSeq Reagent Kits (300 cycles or 600 cycles) paired with either Illumina MiSeq V2 or V3 Flow Cells.The raw Nanopore signal data were basecalled and demultiplexed with the latest version of Guppy at the time of sequencing (version 5.0.17-version6.5.7) using the high-accuracy or super-high accuracy models.
Basecalled Nanopore reads were analyzed and assembled with the CFIA-NCFAD/nf-flu version 3.3.6Nextflow workflow (19,20), which ran IRMA (version 1.0.2) for initial genome assembly (21); nucleotide BLAST version 2.14 (22,23) search of IRMA assem bled genome segment sequences against all Orthomyxoviridae sequences from the NCBI FTP site (https://ftp.ncbi.nlm.nih.gov/genomes/Viruses/AllNucleotide/;1,070,105 sequences downloaded 14 June 2023); selection of appropriate reference sequence for each genome segment and H/N subtype prediction based on nucleotide BLAST results; Minimap2 version 2.24 (24) for read mapping to each genome segment reference sequence; Samtools version 1.15 (25,26) and Mosdepth version 0.3.3(27) for read mapping and sequencing coverage statistics; Clair3 version 1.0.2(28) for variant calling; Bcftools version 1.15.1 (26) for variant filtering and depth-masked consensus sequence generation for each genome segment; and MultiQC version 1.12 (29) for bioinformat ics analysis summary report creation.Bcftools generated consensus sequences from Nanopore analysis with nf-flu were used for further analyses.For Illumina sequencing reads, IRMA (version 1.0.2) Influenza genome assembly as part of the CFIA-NCFAD/nf-flu (version 3.3.6)workflow was used to generate the consensus sequences used for further analyses.All viral genome sequences reported in this study will be made available on the GISAID database under the NCFAD account.Details regarding the submission process and timeline are described in a forthcoming manuscript ("Phylodynamics of the H5N1 avian influenza outbreak in North America reveal the emergence of reassortants with increased fitness" A. V. Signore, J. Giacinti

Phylogenetic analyses
Non-reassorting viral segments (HA, NA, and M) were trimmed of regions flanking the open reading frames and concatenated.The geographic origin (either Eurasian or North American) of each genome segment prior to concatenation was assessed by BLAST search similarity against reference sequences defined with segment-specific phylogenies from Alkie et al. (30).Concatenated HPAIV H5N1 sequences were aligned using MAFFT version 7.49 (totaling 4,087 nucleotides in length) (31) and used to build a maximum likelihood phylogenetic tree using IQ-TREE version 2.20 (32).A separate partition was designated for each viral segment, allowing each to have its own model of nucleotide substitution and model-specific parameters as determined by ModelFinder (33).Node support for the resulting tree was assessed by 5,000 ultrafast bootstrap replicates (34).The bootstrap consensus tree was re-rooted on the tip that minimizes the squared deviation of the root-to-tip regression, and sampling dates for each tip were used to timescale the tree under a relaxed molecular clock rate in TreeTime (35).Reconstruction of the internal nodes' host state in the time-calibrated phylogenetic tree was conducted using the TreeTime mugration model.The resulting phylogenetic tree, reassortment pattern, and host taxonomy were visualized using R package ggtree version 3.7.2(36).Segment-specific phylogenies were inferred to identify unique genome constellations (i.e., reassortant genotypes), detailed in a forthcoming manuscript on phylogeography of HPAIV (A.V. Signore, J. Giacinti

Additional data
Information on infected domestic bird premises in Canada was obtained from the Canadian Food Inspection Agency (12).Unusual wild bird mortality event information was obtained from the National Environmental Emergencies Centre situation report (37), which includes information received from surveillance partners across Canada throughout the course of the HPAIV epidemic in Canada.These data were supplemented by additional information from provincial/territorial wildlife agencies and information obtained through a regional collaborative effort to document HPAIV-related mortality estimates in Atlantic Canada (38).
We simplified migratory flyway boundaries according to provincial/territorial divisions in Canada (39), acknowledging that migration does not precisely align with administra tive boundaries and adjacent flyways overlap in some areas.ArcMap Pro version 3.0.0was used for mapping.

HPAIV outbreak timeline
The presumed index case among wild birds of the 2021-2022 outbreak in North America was a first-winter Great Black-backed Gull (Larus marinus; order Charadriiformes) from Newfoundland.This bird was exhibiting neurologic signs including inability to fly, head tilt, ataxia, and depression.This bird and two other birds with similar histories were found alive between 4 November and 26 November 2021, all three died within 24 hours of admission to a wildlife rehabilitation center, were submitted to the CWHC in late December, and confirmed HPAIV positive the same month (Fig. 1).HPAIV was detected within the Atlantic Flyway in geese (Anatidae spp.) and raptors (Accipitriformes, Strigiformes, and Falconiformes) in January (Nova Scotia and Prince Edward Island), February (New Brunswick), and March (Quebec) 2022.A separate incursion of HPAIV was detected in the Pacific Flyway in a Bald Eagle (Haliaeetus leucocephalus) in British Columbia in February 2022 (Fig. 1) (3), but there were no further detections in the province until April 2022.These detections represent bicoastal incursions by early 2022.The first detection in the mid-continental Mississippi Flyway was a Red-tailed Hawk (Buteo jamaicensis) in Ontario, in March 2022.Detections in southern Manitoba (Mississippi Flyway) and Saskatchewan (Central Flyway) were in Snow Geese (Chen caerulescens) and began in late March 2022 (Fig. 1).Presumed index cases in flyways, provinces, and new areas within each province typically were sick or dead or apparently healthy members of the order Anseriformes and sick or dead raptors followed by corvids (Passeriformes; Movie S1).Within each flyway, the subsequent species detected through morbidity and mortality surveillance were often Charadriiformes, specifically gulls (Larus spp.; Movie S1).The first detections in northern Canada occurred in early May in the Yukon Territory in a Canada Goose (Branta canadensis) and a Trumpeter Swan (Cygnus buccinator) and in the latter half of June in the Northwest Territories and Nunavut in Herring Gulls (Larus argentatus).
Multiple notable mortality events associated with HPAIV were reported across the country beginning in March 2022.Mortality was reported in Red-breasted Mergansers (Mergus serrator; <100) in Ontario in mid-March 2022, followed by Green-winged Teal (Anas carolinensis; <100) in early April 2022.While it is plausible that these were birds overwintering in the area of Ontario where they were found, it is more probable that they were early in their migration north from their wintering areas.In the Central Flyway beginning late March 2022, mortality events were reported in migrating Snow Geese in southern Alberta (hundreds) and Manitoba (unreported number) and migrating Snow Geese, Canada Geese, and Ross' Geese (Anser rossii; hundreds; Fig. 1) in Saskatchewan.Notable mortality in raptor species (e.g., eagles, owls, and hawks) was reported in Alberta and Saskatchewan (<100) also beginning in late March 2022 in addition to gulls and corvids in Saskatchewan (<100) (Fig. 1).
In the fall of 2022, notable mortality events were again reported in Geese.Canada Geese and Cackling Geese (Branta hutchinsii) mortalities (hundreds) were reported in September in the Pacific Flyway (British Columbia) and in Snow Geese in mid-Novem ber 2022 in the Atlantic (Quebec) and Pacific (British Columbia) Flyways (hundreds).It is important to note that, in all cases, the reported wild bird mortality numbers will represent only a fraction of the total mortality.

Morbidity and mortality wild bird surveillance component
A total of 6,246 sick and dead wild birds were collected and tested for the presence of AIV across Canada from November 2021 to November 2022 (Fig. 2A).Overall, 1,710 [27.4%; 95% confidence interval (CI): 26.3%-28.5%]were confirmed or suspected positive for HPAIV (Fig. 3).Individuals that tested positive for HPAIV based on swab samples were generally reported to have lesions consistent with HPAIV infection upon gross and histologic examination as determined by a pathologist.However, the proportion of birds that underwent gross and histologic examination varied by province/territory, and this information was not uniformly available.In at least one province/territory, no birds that tested positive for HPAIV underwent gross or histologic examination.In another province/territory, 76% of birds that tested positive for HPAIV underwent gross and histologic examination, of which 97% were interpreted by the pathologist to have lesions consistent with HPAIV infection and cause of death attributed to HPAI.A total of 62 (1.0%; 95% CI: 0.8%-1.3%)sick and dead wild birds were positive for LPAIV.LPAIV was detected most commonly in members of the Charadriiformes, Anseriformes, and Accipitriformes (Fig. 3; Table S1).

Spatial
Sick and dead birds were submitted from all provinces and territories, with relatively fewer submissions and detections in sick and dead bird samples from northern regions of provinces or the territories (i.e., northern Canada) (Fig. 2A and C).Suspected or confirmed HPAIV detections in sick and dead birds occurred across all flyways but were in the highest numbers in the Atlantic (particularly Quebec) and Central Flyways (particu larly Alberta and Saskatchewan) (Fig. 2C and 3).

Temporal
Following the initial incursion in November 2021, carcass submissions and detections began increasing between January and February 2022 (Fig. 4A).At the national scale, the highest number of submissions was in April 2022 (Fig. 4A), but this varied by flyway with submissions peaking earlier in February and March 2022 in the Atlantic Flyway and again in June 2022 (Fig. S1).The highest number of HPAIV detections was also in April 2022 at the national scale.Within flyways, the number of detections in sick and dead birds also peaked in April in the Central and Mississippi Flyways but peaked in May and June in the Pacific and Atlantic Flyways, respectively, with a second small peak in the fall (September in the Mississippi Flyway, September and October in the Central Flyway, and November in the Pacific and Atlantic Flyways) (Fig. S1).
At the national scale and across the Pacific, Central, and Mississippi Flyways, sick and dead geese, primarily Snow Geese and Canada Geese, accounted for the most detections (Fig. 3).A first peak of detections in both Canada Geese and Snow Geese occurred in April 2022 (Fig. S3).A second peak in Canada Geese occurred in September 2022, although detections in this species occurred continuously from January through November 2022 across flyways.The second peak in Snow Geese occurred in November 2022 (Fig. S3); trends in Snow Geese were largely driven by detections in the Central and Atlantic Flyways (Fig. S3).
At the national scale, peaks in morbidity and mortality for dabbling ducks occurred in the spring (April) and fall (September; Fig. 3).In diving ducks and seaducks, peaks corresponded with the breeding season in May and June and were largely driven by Common Eiders in eastern Canada (Fig. S4).
Peaks in morbidity and mortality for raptors and corvids occurred in the spring, in April and May, with a slight increase in detections in the fall for both functional groups (Fig. S5).The majority of HPAIV detections in corvids occurred in the Central, eastern Mississippi, and Atlantic Flyways (Fig. S5).HPAIV detections in raptors occurred across all flyways (Fig. S5); however, the majority of detections in Strigiformes were found in the Pacific and Central Flyways (Fig. 3).

Viral reassortment and phylogenetic analysis
There was substantial genetic diversity in HPAIV, as 341 (24.8%) were wholly Eurasian-ori gin HPAIV, with the remaining genotypes resulting from reassortments between Eurasian and North American virus segments, with evidence of 10 different genome constellations (Table 1; Fig. 7 and 8).Identification of unique genome constellations was based on the combination of its membership in genetically distinct, monophyletic clades within phylogenies inferred for each gene segment individually (A.V. Signore, J. Giacinti MacPhee, G. J. Parsons, H. Eaglesome, S. Gilbert, K. Saboraki, R. Davis, A. Jerao, M. Ginn, C. Soos, Y. Berhane, unpublished data).The majority of reassortants detected involved one to four of the five internal gene segments (i.e., PB2, PB1, PA, NP, and NS) originating from North American LPAIVs, and none involved reassortment of the HA gene.Only a single virus (Pattern 10) in an apparently healthy Blue-winged Teal in Manitoba in August 2022 involved reassortment of the NA and M genes.However, while this H5N6 virus showed reassortment involving all but the HA gene, it was sequenced directly from the swab material from this bird, and virus isolation was unsuccessful.In swabs collected from birds at the same location during the same month, virus isolation yielded either Eurasian H5N1 or North American H4N6 virus.The most common gene segments involved in reassortment were NP, PB2, and PB1, which were involved in eight, seven, and five of the 10 genome constellation patterns, respectively (Table 1).The most common genome constellations detected in Canada in the first year since incursion included the wholly-Eurasian lineage along with Patterns 2, 4, and 5, collectively comprising 93.5% of all sequenced viruses (Table 1).
Detections of the Eurasian-origin virus occurred throughout the study period and in every flyway, but most Eurasian HPAIV in Canada occurred in sick and dead birds in the Atlantic Flyway (Fig. 7 to 9).Only a small proportion of viruses sequenced from live and harvested birds were Eurasian, and all were detected in the Atlantic Flyway in August and September 2022 (Fig. 7).Following the first detection of an H5N1 reassortant in March 2022 in the Atlantic Flyway, the proportion of wholly-Eurasian viruses detected decreased in April, increased every month until July when the majority of detections were Eurasian viruses, and then decreased again to zero by November 2022 (Fig. 7).Charadriiformes and Suliformes hosts made up the greatest proportion of wholly-Eurasian virus detections.The wholly-Eurasian virus, more so than its descendant reassortants, showed phylogenetic structuring by host group, as viruses sequenced from Suliformes tended to be genetically distinct from those sequenced from Charadriiformes (Fig. 8).
The most commonly detected genome constellation overall was Pattern 2 (n = 367), which has the most genome segments in common with the wholly-Eurasian virus.Both Pattern 2 and the Eurasian H5N1 were distributed across Canada; however, Pattern 2 was widely distributed in western Canada, as it was the most common genome constellation found in the Pacific and Central Flyways (Fig. 7 and 9).Pattern 2 (including viruses from the Pacific Flyway referenced as clusters 1, 2, 4, and 5 in (45) was first detected in the Central and Atlantic Flyways in March 2022 and in April 2022 in the Mississippi and Pacific Flyways (Fig. 7).Pattern 2 detections decreased to zero in both the Atlantic and Mississippi Flyways by June 2022; however, they remained prominent in the Central and North American 99 N Am N Am N Am N Am N Am N Am N Am N Am Pacific Flyways (Fig. 7).Unlike the wholly-Eurasian virus, there is limited phylogenetic structure by host, as viruses from each host group form much smaller monophyletic groups than those infected with the Eurasian virus (Fig. 8).Pattern 4 (including viruses from the Pacific Flyway referenced as cluster 3 in (45) was among the most commonly detected genome constellations across Canada (n = 212).This reassortant, which is one of the most divergent genome constellations from the Eurasian virus (Fig. 8), was first detected in April 2022 in adjacent Central and Mississippi Flyways and represented a high proportion of detections particularly in the viruses, the genetic composition of which is described in Table 1.
Central Flyway through November 2022.Pattern 4 was detected in the Atlantic Flyway in a small proportion of samples starting in May but became dominant by October in live and harvested birds, and by November 2022, in sick and dead birds.A few viruses with Pattern 4 were detected in early June 2022 (45), report these occurred in late May 2022) in the Pacific Flyway and then again in November 2022 in the majority of collected samples (Fig. 7).Like the other reassortants, Pattern 4 also showed phyloge netic structuring by flyway and included a large cluster that almost exclusively infected Anseriformes in the Pacific Flyway.
Pattern 5, which was among the most commonly detected genome constellation patterns (n = 272), was found only in the Mississippi and Atlantic Flyways and continued to be detected through November 2022 in high proportions in both flyways, including in live and harvested birds (Table 1; Fig. 7).It was the most commonly found genome constellation in the Mississippi Flyway and peaked in the spring (March to April) and fall (September to October) of 2022 (Fig. 7).In the Atlantic Flyway, it was second only to the wholly Eurasian genotype, peaking in spring (April to June) and in fall (September to November), often outnumbering the wholly Eurasian genotype within those time periods (Fig. 7).
Peaks in diversity of genome constellations detected in a given flyway (e.g., >4 patterns) occurred in April and May in the Central, Mississippi, and Atlantic Flyways, with additional increases in the constellation diversity in the fall (October and November) in the Pacific and Mississippi Flyways (Fig. 7).Peaks in diversity often coincided with peaks in detections (Fig. 7).

DISCUSSION
The incursion of HPAIV H5N1 clade 2.3.4.4b into Canada resulted in unprecedented detections in asymptomatic wild birds and large-scale wild bird mortality, affecting a wide range of species.Based on the authors' collective knowledge, no other infectious disease, including the previous HPAIV incursion in 2014-2015, has caused this magni tude of mortality in such a large diversity of bird species in Canada.While not the manuscript's focus, characteristic lesions were associated with HPAIV-positive RT-PCR results in most wild birds that underwent gross and histopathological examination, supporting the assumption that HPAIV-positive dead birds died as a result of infec tion.However, because we rely heavily on opportunistic reporting of mortality events, the data presented here provide only conservative estimates of the scope and scale of HPAIV-associated wild bird mortality in Canada.The field capacity and resources necessary to complete structured surveys are prohibitive during multiple, concurrent, large-scale mortality events across the country.

Wild birds: victims and reservoirs
In the first year of the 2021-2022 outbreak, H5N1 HPAIV was detected in 1,710 sick or dead wild birds from 80 species across 12 taxonomic orders.Most of the sick and dead HPAIV-positive birds submitted for testing were Anseriformes (primarily geese) and raptors (owls, hawks, and eagles), followed by corvids and Charadriiformes (primarily gulls and murres).However, the largest recorded mortality events, during which only a subset of carcasses get submitted for testing, occurred in Canada Geese and Snow Geese, and during the breeding season in colonial nesting species in eastern Canada, including Northern Gannet, Common Murre, and Common Eider.Where wild bird mortality was notable throughout 2021 with a wide taxonomic distribution, distinct peaks in detections among asymptomatic wild birds were observed in the spring and fall primarily in dabbling ducks.Peaks in AIV prevalence in dabbling ducks during spring and fall have been well-described for LPAIVs (46,47) and correspond with northward and southward wild bird migrations in conjunction with an influx of naive juveniles, respectively.Increased abundance and density of wild migratory birds during migration facilitate viral transmission through close contact and environmental contamination.
Anseriformes and Charadriiformes, in particular, have been recognized as reservoirs for LPAIVs (48,49).Our data from Canada along with other studies (9) indicate that candidate wild bird species may also act as reservoirs for H5N1 HPAIV.Among apparently healthy birds sampled through live or hunter-harvested bird surveillance, dabbling ducks had the highest prevalence of both HPAIV (8.4%; specifically American Black Duck, Mallard, their hybrids, and Northern Pintail) and LPAIV (21.3%).This was unsurprising for LPAIV, as dabbling ducks have one of the highest LPAIV prevalence worldwide (49), and the largest number and diversity of AIV subtypes isolated globally (50).There were also no large-scale mortalities and few HPAIV detections in sick and dead dabbling ducks, corroborating that these species exhibit less morbidity and mortality compared to other Anseriformes (51).Exposure to LPAIVs could provide some level of heterosubtypic cross-protective immunity against HPAIV (52).Captive studies have demonstrated that specific LPAIVs can confer partial cross-protective heterosubtypic immunity to other LPAIVs (53-55), as well as HPAIVs (52,56,57), which can result in reduced viral loads, duration of shedding, and, in the case of HPAIV, reduced morbidity and mortality.With their high prevalence of both HPAIV and LPAIV, together with their large population sizes (at least 25 million dabbling ducks; USFWS 2022), dabbling ducks are likely candidates as reservoirs of HPAIV in Canada.
Conversely among Anseriformes, geese, diving ducks, and sea ducks appeared to be highly susceptible to morbidity and mortality.High frequencies of sick or dead Canada Geese and Snow Geese tested positive for HPAIV, and HPAIV detections in apparently healthy geese were low compared to dabbling ducks, with only one Canada Goose across Canada.Viruses detected included those with (A) wholly Eurasian or North American origins, and (B-D) and 10 reassortants, the genetic composition of which is described in Table 1.Provinces and territories are colored by the predominant migratory bird flyway.
testing positive out of 1,427 tested, and 125 (5.1%) Snow Geese positive out of 2,475 tested.Large mortality events were also reported in Common Eider.Of the 96 Common Eider tested through sick and dead bird surveillance, 60 were HPAIV-positive; however, none tested positive for HPAIV through live bird surveillance.Thus, Canada Geese and Common Eider do not appear to be strong candidates as reservoir species and have the potential to be significantly impacted by this virus if population-level immunity does not develop or is not sustained over time.Conversely, Snow Geese have the potential to play an important role as sources of transmission and spread, particularly given their large population sizes [e.g., the mid-continental Snow Goose population was estimated at over 16.2 million (±1.6M) in 2022; USFWS 2022], their gregarious behavior during migration and breeding, long migration distances including to Arctic breeding grounds (i.e., potential areas of flyway overlap, including with trans-Atlantic migrants), and the significant overlap in ranges and habitats with other waterfowl species including dabbling ducks.
Charadriiformes have been proposed as potential candidates for the AIV spread within and between colonies or foraging sites during the breeding season (58,59), as well as over long distances during migration (1).Gulls, including several species identified as candidates for the HPAIV movement from Europe to Canada in 2021 (1,3), were the first cases detected in Canada and were often detected as index cases as the virus moved north.However, the contribution of Charadriiformes as a reservoir species is less clear than for dabbling ducks.While several species of Charadriiformes, particularly murres, gulls, and terns, were susceptible to HPAIV-related mortality, HPAIV detections through live and hunter-harvested surveillance were generally low.Notably, a self-limiting HPAIV outbreak with low mortality was reported in summer 2022 among Herring Gulls on Kent Island, New Brunswick, Canada (60).Similarly, a large proportion of apparently healthy Common Murre tested positive for HPAIV within a short time frame from a single colony experiencing an active outbreak (A.MacLaughlin, J. Giacinti, I. Rahman, J. Wight, K. Hargan, A. Lang, M. Mallory, G. Robertson, K. Elliot, D. Ojkic, S. Lair, M.Jones, Y. Berhane, G. Gilchrist, L. Wilson, S. Wilhelm, M. Brown, J. Provencher , unpublished data), yet no mortalities were observed based on re-sightings days later (J.Cunningham, ECCC, personal communication).In contrast, Common Murre mortality events were reported at several other colonies in the Atlantic region during the same time period (38), suggesting colony-level differences in virus dynamics and susceptibility.
Beyond gulls and Common Murre, certain seabird species, such as the Northern Gannet, exhibited remarkable susceptibility to HPAIV.Large-scale mortality was observed in a number of colonial nesting seabird species during the breeding season in Atlantic Canada and across the North Atlantic (61).Colonial nesting behavior, characterized by dense populations, a high degree of social interaction, and shared foraging areas, can facilitate extensive transmission among conspecifics, leading to focal and largescale die-offs following the introduction of highly transmissible pathogens like HPAIV (62).However, there is evidence of exposure and survival in some highly susceptible colonial nesting species (61).While an in-depth analysis of the impacts to seabirds in Atlantic Canada is the focus of Avery-Gomm et al. (38) and falls outside the scope of this manuscript, we emphasize the importance of continued targeted AIV surveillance in these populations to understand the interacting mechanisms driving species-and colony-level differences in virus dynamics, transmission, and susceptibility.Ongoing serologic surveillance can also contribute to expanding our knowledge of exposure and survival in these populations, advancing our understanding of heterosubtypic immunity, and enhancing our ability to forecast mass mortality following exposure to H5Nx HPAIVs.
Podicipediformes, or grebes, are diving waterbirds with previously demonstrated susceptibility to HPAIV (63), as corroborated by the current results.However, there is little evidence that Podicipediformes play an important role as LPAIV or HPAIV reservoir hosts, with the caveat that these species are generally not well studied.In our surveillance, Horned Grebes (Podiceps auritus) were the only grebe species for which apparently healthy individuals were sampled.Among these individuals, there were no detections of HPAIV or LPAIV.Although there were relatively few samples (n = 83), they were collected during the same month (June) and in the same flyway (Central) as reported grebe mortality, which was primarily observed in Eared (P.nigricollis) and Western Grebes (Aechmophorus occidentalis).These colonial species have similar habitat preferences to dabbling ducks, especially Eared Grebe, which rely on large shallow ponds with dense vegetation during breeding (64).Therefore, in areas of range overlap, there is an increased likelihood of exposure to HPAIV-contaminated habitat coupled with increased transmission risk related to colonial nesting dynamics.In comparison, there was no reported HPAIV-related mortality in Horned Grebes, which may be less likely to occur in proximity to dabbling ducks, because they are highly territorial during breeding and are more likely to nest in isolation on smaller ponds with open water (65).As Podicipedi formes are identified as priority species for conservation and stewardship in one or more locations in Canada (66) and two species appear in Schedule 1 of the Species at Risk Act (42), a better understanding of factors influencing HPAIV-related mortality events for birds in this order is warranted.
Raptors and corvids have demonstrated pronounced susceptibility to HPAIV during the current and previous HPAIV outbreaks in North America (67).The underlying reason is unknown but is likely related to the route and dose of exposure.The most likely routes of exposure are through scavenging of infected carcasses and, in the case of raptors, through infected prey (68) (45).Infected prey that are displaying signs of weakness or abnormal behavior may be preferentially targeted (69)(70)(71).While there are no samples in the national data set from apparently healthy raptors and corvids, the role that these species play as reservoirs is likely to be minimal given that most species are relatively solitary.They may, however, play a role in subsequent spread to conspecifics at shared roosting or feeding sites, where some species can occur in high numbers, or to offspring during the breeding season (e.g., through parental feeding of infected prey items).
The current outbreak reflects a significant shift in HPAIV dynamics, highlighting the dual role of wild birds as victims and reservoirs of this virus.Based on the data presented, there are differences in species susceptibility between and within wild bird taxonomic orders emphasizing the importance of a species-level approach to data interpretation and conclusions.The observed taxonomic and temporal patterns are also important to interpret in the context of a novel AIV clade, to which Canada's migra tory bird populations were immunologically naive.Although widespread transmission should result in the development of immunity, and consequently reduced infection and mortality, the duration and extent of this immunity remain uncertain.Mortality events may continue to be pronounced in highly susceptible species, as a high case fatality rate may limit transmission and delay population-level immunity.Factors like food scarcity and extreme weather events (72) can further impact the health and resilience of populations, rendering them more susceptible to mortality following HPAIV infection.This is particularly important because many species identified here as highly susceptible to HPAIV share characteristics of relatively long lifespan, low fecundity, and high levels of parental care.Mortality of adults during the breeding season, as seen for the majority of seabirds, sea ducks, grebes, and raptors, would also indirectly impact reproductive success through increased nest failures from reduced hatching success or increased nestlings' mortality.The combination of increased mortality and decreased reproduction can result in significant population-level impacts, particularly for species or populations that are vulnerable or are already experiencing multiple concurrent stressors (e.g., reduced food abundance or quality, increasing industrial or agricultural activity, urban encroachment, and other large-scale environmental changes associated with climate change).Concurrent stressors can also impact the ability of many of these populations to recover from mortality and reproductive failure associated with HPAIV.

Genome constellations
Reassortment is a recurring phenomenon among LPAIVs within wild waterfowl populations (73,74), and H5 clades 2.3.4.4 and 2.3.4.4b have demonstrated a high propensity to reassort with LPAIVs (52,75).Co-circulation of HPAIV and LPAIV among wild bird reservoir species (e.g., dabbling ducks) increases opportunities for co-infections and reassortment (73).This is consistent with observations in the year following the first HPAIV incursion into Canada, where increased detection rates of new genome constel lation patterns coincided with periods of increased LPAIV prevalence and concentra ted wild bird abundance on the landscape during the spring and fall of 2022.In addition to these temporal patterns, there was evidence of geographic structuring of genome constellations at the flyway scale, whereas distinct geographic trends were not evident at the ecoprovincial scale within the Pacific Flyway (45).These temporal and spatial relationships underscore the dynamic interplay between virus prevalence, susceptibility, host reservoir abundance, and movement (i.e., migration timing and pathways) in shaping reassortment dynamics.With the continued circulation and spread of H5N1 and its reassortants across Canada, homotypic and heterotypic immunity in wild bird reservoirs will also likely impact HPAIV dynamics (76), influencing the frequency and diversity of viral reassortants.Therefore, longitudinal surveillance targeted during periods of concentrated reservoir abundance, coordinated at the flyway scale, and incorporating serologic sampling will collectively be needed to track and understand the complex, dynamic, and rapid evolution of this virus.
The majority of reassortants detected in the first year following H5N1 2.3.4.4b incursions resulted from the exchange of internal gene segments with North American LPAIVs, which is consistent with reports from the USA (77).The majority of sequences detected in Canada in the first year of surveillance post-incursion were categorized into four broad genome constellations including wholly Eurasian H5N1 and Patterns 2, 4, and 5.The two most frequently detected genome constellations detected in the USA from December 2021 to April 2022, as similarly observed in Canada, were Eurasian H5N1 [genotype A1 in reference (77)] and Pattern 2 (genotype B2, B3.1, and B4); Pattern 7 (B1.1 and B1.2) was also among the most common patterns along with Pattern 4 (B3.2), and Pattern 5 was not detected in the USA in that time period (77).Interestingly, the persistence of the Eurasian virus was particularly notable in the Atlantic Flyway throughout the first year following the first incursion.Although Eurasian viruses were sporadically detected in other flyways, their presence was transient, and they were quickly outnumbered by its decendant reassortants.It is unclear what ecological, evolutionary, or viral factors drove this persistence in the Atlantic Flyway but could include variation in the prevalence and composition of LPAIVs subtypes circulating within the Atlantic region compared to other flyways, differences in survival with subsequent impact on opportunities for virus reassortment, or species-specific interactions unique to the Atlantic Flyway.However, sampling biases in the composition of species, locations, and timing, in addition to diagnostic considerations (e.g., samples yielding lower RT-PCR Ct values were more likely to result in higher quality sequence data and be included in analyses), mean that observed patterns reflect available sequences and are therefore unlikely to represent the complete diversity and distribution of viruses present in wild bird populations.

Surveillance components and sampling limitations
It is important to note that while live and hunter-harvested (i.e., "active") and sick and dead (i.e., "scanning" or "passive") wild bird surveillance methods can be complementary and contribute data from different subsets of wild birds (i.e., those that do and do not survive infection), each surveillance method has limitations and biases that are critical to understanding to contextualize the results presented here (78).
The majority of sick and dead wild bird carcass submissions are opportunistically submitted by organizations or individuals with a vocational or avocational interest in wildlife disease and, therefore, generally originate from more populous areas of Canada (Fig. 2).Geographic proximity to diagnostic centers and higher human population densities increase the likelihood of carcass detection and submission (79).Biases in species detectability (e.g., size, habitat with dense vegetation vs open parkland) and the likelihood of submission based on social (e.g., perceived as a nuisance vs highly valued) or other factors (e.g., disparate levels of awareness between communities) can also influence which samples are processed through this surveillance component (80).Therefore, the absence of detection through sick and dead bird surveillance does not imply the absence of infection and mortality.
Live and hunter-harvested wild bird surveillance is also opportunistic in that it is carried out in conjunction with existing banding and monitoring programs.By strategi cally targeting wild birds during periods and in areas of high abundance, banding and monitoring programs are well-aligned with locations and time periods expected to have increased AIV prevalence.However, these programs are often conducted over short time frames (days or weeks), only during certain months, and limited to focal areas.Conse quently, this can limit our ability to detect infection, which, for AIV, consists of a relatively short viral shedding period (81), and to track changes in incidence and prevalence within these high-risk areas and time periods.Hunter-harvested surveillance is also dependent on the species, individuals, and age categories that are permitted and more likely to be selected or susceptible to selection.Despite these limitations, the continued integration of both sick and dead as well as live and harvested wild bird surveillance remains crucial to understand HPAIV dynamics in wild birds.

ACKNOWLEDGMENTS
We extend our sincere appreciation to members of the public who reported sick and dead birds, rehabilitation centers and migratory game bird hunters who reported observations and contributed samples, and conservation officers and other staff who responded to calls throughout the course of the outbreak.We would also like to thank the Nunatsiavut Government, NunatuKavut Community Council, Miawpukek First Nation as well as the Nunavik Research Center of Makivvik and the Cree Board of Health and Social Services for their role in either the collection of carcasses or harvested bird samples submitted by local harvesters and conservation officers.These valuable inputs have greatly enriched the depth and quality of our work.This manuscript was prepared on behalf of all partners that contribute to Canada's Interagency Surveillance Program for Avian Influenza Viruses in Wild Birds, including Canadian Wildlife Health Cooperative, and Federal, Provincial, Territorial, Indigenous, and Academic partners involved in wildlife, domestic animal, and human health.See Table S2 for acknowledgment of additional collaborators that have contributed to the collection and curation of these data.

FIG 2
FIG 2 Distribution of (A) sick and dead wild birds submitted and tested, (B) live and hunter-harvested wild birds tested for avian influenza virus, (C) the number of sick and dead wild birds confirmed or suspected to be positive for highly pathogenic avian influenza virus, and (D) the proportion of live and hunter-harvested wild bird samples confirmed or suspected to be positive for HPAIV in Canada between November 2021 and December 2022, based on RT-PCR.Internal boundaries indicate the watershed (i.e., sub-sub-drainage area) where samples were received for testing.Shapefile was downloaded from the National Hydro Network (43) on 19 October 2022 and was clipped to the land boundary.

FIG 3
FIG 3 Number of sick and dead wild birds submitted for testing and suspected or confirmed highly pathogenic avian influenza virus or low pathogenicity avian influenza virus positive in Canada between November 2021 and November 2022.Color shading reflects general administrative migratory flyway routes (orange, Pacific; green, Central; blue, Mississippi; and yellow, Atlantic).

FIG 4
FIG 4 Wild birds tested for avian influenza virus between November 2021 and November 2022 across Canada obtained through (A) morbidity and mortality surveillance and (B) live and hunter-harvested surveillance.

FIG 5
FIG 5 Breakdown of sick and dead wild bird samples that were confirmed or suspected highly pathogenic avian influenza virus positive between November 2021 and November 2022 across Canada stratified by taxonomic order and species functional group.The data are shown as a treemap; the area of each rectangle is proportional to the number of wild bird samples that were confirmed or suspected HPAIV positive.[GA, Galliformes; CO, Columbiformes; PR, Procellariiformes (N = 2); unlabeled Passeriformes; other (N = 4); unlabeled Charadriiformes, Sandpipers and Allies (N = 2)].The data shown here reflect the samples that were tested and therefore do not represent the number of birds that died from HPAIV.

FIG 6
FIG 6 Number of live and hunter-harvested wild birds submitted for testing and suspected or confirmed highly pathogenic avian influenza virus or low pathogenicity avian influenza virus positive in Canada between November 2021 and November 2022.Color reflects migratory flyway routes (orange, Pacific; green, Central; blue, Mississippi; and yellow, Atlantic).

FIG 7
FIG 7 Avian influenza viruses characterized from wild birds sampled during the first year of the outbreak (November 2021-November 2022) in the (A) Pacific Flyway, (B) Central Flyway, (C) Mississippi Flyway, and (D) Atlantic Flyway, stratified by surveillance component and month.AIV detected included wholly Eurasian, wholly North American, and 10 reassortant

FIG 8
FIG 8 Time-calibrated maximum likelihood phylogenetic tree of 1,166 HPAIV H5N1 viruses collected from wild bird hosts between November 2021 and 2022.Tree branches are colored according to the flyway in which the sample was collected.Host taxonomic order is represented by colored bars at tree tips.The geographic origin of genome segments 1-8 (either Eurasian or North American) for all tree tips is represented by red and gray bars (from left to right: PB2, PB1, PA, HA, NP, NA, M, and NS).First H5N1 2.3.4.4b detections following the initial incursions into Canada are annotated.

FIG 9
FIG 9 Distribution of avian influenza viruses characterized from wild birds sampled during the first year of the H5N1 outbreak (November 2021-November 2022)

TABLE 1
The geographic origin of avian influenza virus genome segments (Eurasian, EA or North American, N Am) for genome constellations detected through wild bird surveillance in Canada between November 2021 and November 2022