Main

High pathogenicity avian influenza (HPAI) A(H5N1) clade 2.3.4.4b viruses continue to evolve in North America, most notably through reassortment with endemic avian influenza lineages. In autumn 2024, we tracked a new reassortant, genotype D1.1, through coordinated wildlife surveillance in Canada and the USA and observed its rapid spread across several migratory flyways. Although the emergence of D1.1 coincided with human infections and detections in dairy cattle, viruses from wild birds lacked key mammalian-adaptive substitutions found in human cases, and candidate vaccine viruses retained antigenic cross-reactivity with D1.1 strains.

The introduction of HPAI clade 2.3.4.4b A(H5N1) viruses into North America in late 2021 catalyzed extensive reassortment with endemic low-pathogenicity avian influenza (LPAI) viruses, generating diverse new genotypes (distinct combinations of influenza gene segments) with varied phenotypes1. These reassortant viruses have caused widespread mortality among wild birds and mammals2, with nearly 14,500 wild bird and 650 mammal detections in the USA alone3. Canada has recorded HPAI in animals across 12 taxonomic orders and 80 wild bird species within the first year of introduction4. True mortality impacts are probably underestimated. Since February 2022, HPAI outbreaks have also severely impacted the poultry industry, affecting more than 2,000 US flocks and 186 million birds, and more than 14.5 million birds across 534 flocks in all Canadian provinces3,5. There have been 17 reported human A(H5N1) infections caused by genotype D1.1 viruses, of which 4 were severe and 2 fatal6,7,8,9.

The origins and spread of D1.1 across North America, and whether mammalian‑adaptive mutations in the human cases10,11 came from wild bird viruses or arose de novo in humans, remain unclear. To address these questions, we gathered surveillance data during the fall 2024 migratory season for wild birds in Canada and the USA (September–December 2024) from both passive and active surveillance efforts in wildlife, aiming to better understand the dispersion of these viruses.

In total, 926 wildlife samples collected in the USA and Canada during this period tested positive for HPAI A(H5N1) virus genotype D1.1 by sequencing (Extended Data Fig. 1). Infections were distributed across all four North American wild bird flyways, with the Pacific and Mississippi Flyways accounting for the highest proportions of detections, with most samples coming from geese and ducks. Most infections were detected through passive surveillance (that is, samples that were taken from sick and dead birds), although active surveillance (that is, samples from apparently healthy birds through live capture or hunter harvest) contributed to a larger percentage of the detections among ducks compared to geese.

In Canada, 4,141 samples were collected, with 14.1% positive for influenza A (matrix gene quantitative PCR (qPCR)) and 4.1% for H5 (H5 qPCR) (Extended Data Fig. 2). In the USA, positivity ranged from 10.1% to 34.1% for influenza A and from 2.1% to 17.4% for H5 across several projects. Although sampling efforts were not designed to compare HPAI prevalence across space, H5 positivity seemed to increase geographically from 1.7% in Alaska, to 2.1–4.9% in the midcontinent and 12.3–17.4% in Texas and Louisiana, possibly reflecting viral dissemination throughout wild bird populations. However, variation in species, sampling and isolation methods may bias these prevalence estimates.

To trace the origin and spread of genotype D1.1 viruses across Canada and the USA we performed a discrete phylogeographic reconstruction (Fig. 1a). The median time to the most recent common ancestor of all D1.1 viruses was estimated to be 7 July 2024, with upper and lower bounds (95% HPD) of 28 July 2024 and 17 June 2024, respectively. The inferred ancestral location of D1.1 viruses was estimated to be the northern Pacific flyway, with a posterior probability of 1.0 (Extended Data Fig. 3). The phylogeographic analysis strongly supported the northern Pacific and southern Mississippi flyways as D1.1 hotspots, as these locations were the source of most well-supported location jumps inferred by the analysis (Fig. 1c and Extended Data Fig. 4). The earliest, well-supported location jumps originated from the northern Pacific flyway with inferred spread to the south Pacific, Central and Mississippi flyways as well as Alaska. D1.1 viruses may have then spread northward from the southern Mississippi flyway. Overall, D1.1 spread across Canada and the USA within a single migratory season.

Fig. 1: Surveillance and phylodynamics and A(H5N1) genotype D1.1.
Fig. 1: Surveillance and phylodynamics and A(H5N1) genotype D1.1.
Full size image

a, Time-scaled phylogenetic tree of D1.1 viruses (n = 935) used for discrete phylogeographic reconstruction. Tree tips are colored by administrative flyway of sampling, and branches reflect reconstructed ancestral flyways. Sequence source data are provided. b, Genomic surveillance of sequenced detections of A(H5Nx) HPAI detections in wild birds in Canada and the USA (n = 1,901) from 1 January to 31 December. ‘Other’ denotes the sum of minor circulating genotypes. c, Phylogeographic diffusion of D1.1 across discrete locations. Provinces and states are grouped into northern (-N) and southern (-S) halves of their administrative flyways. Arcs represent inferred D1.1 migrations (Markov jumps), colored black (origin) to gray (destination). Only jumps with Bayes factor >10.0 and posterior probability >0.95 are shown (Extended Data Fig. 3). Map data in c from OpenStreetMap (https://www.openstreetmap.org/).

Source data

By December 2024, D1.1 was the predominant North American genotype, nearly replacing previously circulating genotypes as it spread across the continent (Fig. 1b). Although large genotypic sweeps have been common since these viruses began to dominate the HPAI landscape12, the co-circulation of other H5 genotypes during and after the D1.1 sweep was low.

HPAI candidate vaccine viruses (CVVs) are a key component of the World Health Organization pandemic preparedness strategy13. We selected a subset of D1.1 genotype and co-circulating viruses to evaluate their cross-reactivity against four current HPAI A(H5) clade 2.3.4.4b CVVs (Table 1). All viruses showed moderate-to-high seroreactivity to at least one of the four available clade 2.3.4.4b CVVs.

Table 1 Cross-reactivity of currently circulating bovine and avian influenza A(H5N1) 2.3.4.4b viruses (n = 13) with reference CVVs (n = 4)
Full size table

We further tested a subset of D1.1 viruses from wild birds for receptor binding specificity and found strong binding to avian-like receptors, with little to no affinity for human-like receptors (Extended Data Fig. 5). The D1.1 viruses in wild birds largely retained their avian characteristics, indicating that the current risk of spillover and transmission in humans is probably low.

Potential mammalian adaptation markers, including the HA E190D and Q226H substitutions identified in the human D1.1 case from British Columbia, and the Q226H substitution detected in the Louisiana case10,11, were absent from the D1.1 sequences from wild birds. The one exception was a PB2 E627K substitution detected in A/bobcat/Kansas/W24-1062/2024—a sample that did not originate from a wild bird. The absence of these markers in bird-derived viruses, and the lack of their fixation in human infections, supports the conclusion that these mutations probably arose de novo during infection.

Conclusion

The emergence and rapid spread of HPAI A(H5N1) genotype D1.1 viruses across North American migratory flyways during the 2024 migration season represents a notable shift in clade 2.3.4.4b epidemiology. Several factors may explain the unusually efficient transmission of D1.1. Genomic analyses show that this genotype includes a distinct North American–derived neuraminidase (NA) segment, which may provide antigenic advantages because population immunity to this NA is probably lower than to NA segments circulating widely before 2024. Although principal mammalian adaptation markers were absent in the wild bird samples, some D1.1 viruses carried several mutations with reported phenotypic effects (Supplementary Table 2), which may have contributed to their dominance. Increased densities of immunologically naive juvenile birds during southbound migration may have facilitated rapid viral amplification at staging areas14.

Environmental conditions also probably played some role in the dissemination of genotype D1.1 viruses. Persistent drought and altered habitat use at important breeding and staging areas may have led to increased mixing and density of wild birds, facilitating interspecies viral transmission15. These ecological factors, combined with the novelty of the genotype in wild populations, may have enabled the rapid geographic spread observed.

Although D1.1 viruses circulating in wild birds currently lack principal mammalian adaptation markers, their sustained replication in mammalian hosts, such as US dairy cattle16, increases the evolutionary potential for zoonotic transmission. D1.1 viruses detected in poultry in British Columbia were found to carry the NA-H275Y mutation—a known marker for resistance to oseltamivir17, demonstrating that resistance and adaptation markers can emerge stochastically, even without selection pressure.

Fortunately, CVVs retained antigenic cross-reactivity with D1.1 genotype strains, supporting current pandemic preparedness strategies. Nevertheless, continued real-time surveillance across wild birds, livestock and humans remains essential to detect emergence of antigenic variants or reassortants with altered pathogenicity or host range.

This study has several limitations. Surveillance data vary in geographic coverage, sampling intensity and host representation, which may influence observed HPAI patterns. Further, surveillance alone cannot determine viral fitness or transmission dynamics without supporting laboratory studies, and inherent delays between detection and data availability limit timely assessment of emerging viral features.

In conclusion, the emergence of the D1.1 viruses coincided temporally with the southward autumn migration of North American waterfowl, favoring their dissemination. Their spread highlights the need for an integrated One Health response18 that aligns wildlife surveillance, agricultural biosecurity and public health preparedness.

Methods

All samples were collected in compliance with the US Fish and Wildlife Service, Environment and Climate Change Canada federal or provincial permits, and following approval by Institutional Animal Care and Use Committee(s) from research groups (St. Jude Children’s Research Hospital, University of Georgia Southeast Cooperative Wildlife Disease Study, and The Ohio State University) in the Centers of Excellence for Influenza Research and Response at St. Jude Children’s Research Hospital (Memphis, TN), Environment and Climate Change Canada, the Canadian Wildlife Health Cooperative, Alberta Environment and Parks, The US Geological Survey (USGS), The United States Department of Agriculture (USDA), and the University of California, Davis.

Surveillance and sample collection

Samples were collected from sites in 11 US states (Alaska, Texas, Kansas, Florida, Tennessee, Louisiana, North Carolina, Kentucky, Georgia, Ohio and Wisconsin) and four Canadian provinces (British Columbia, Alberta, Saskatchewan and Ontario) over an 8-month period (May 2024 through December 2024). When available, species, sex, age, sample collection date and location were recorded for each bird sampled. Age was characterized as juvenile or adult based upon wing plumage or, in the case of carcasses, through assessment of gonad development and the size of bursa of Fabricius19. Table 1 includes all H5 detections in wild birds contributed by these groups during the study period that were sequenced successfully; mammalian detections are shown only for reference to document instances where mammalian-adaptive markers were identified, but they do not represent comprehensive mammal surveillance.

Cloacal swabs, oropharyngeal swabs and tissue samples were collected from birds either through active (live capture or hunter harvest) or passive surveillance. Live sampled birds in the USA were trapped according to current guidelines20 and in Canada as described previously4. Hunter-harvested waterfowl were sampled during the fall and winter hunting seasons in both the USA21 and Canada4. Passive surveillance samples were obtained from sick or dead wild birds in the USA and Canada, which included reports and submissions from the public and targeted collections by wildlife health partners. These birds were often collected in response to observed individual or large-scale mortality events and/or clinical signs consistent with avian influenza. Passive surveillance occurred year-round and allowed testing and analysis of nonhunted species.

Sample screening

All samples that tested positive for the influenza A matrix gene by real-time reverse transcription-PCR and were either positive, suspect or non-negative for H5 or H7 subtypes were submitted for confirmatory testing and full genomic characterization. In the USA, initial testing was performed by participating research institutions and in Canada initial testing was performed by laboratories that are part of the Canadian Animal Health Surveillance Network as described previously4. These laboratories screened samples using validated molecular assays for influenza A virus detection and subtyping.

Primary samples were screened using real-time PCR on an ABI 7500 FAST using the Influenza A Matrix primer and probe sequences published as part of the Influenza SARS-CoV-2 multiplex kit22. Samples crossing the cycle threshold by cycle 35 were considered positive and subject to additional screening and characterization. For confirmation or exclusion of H5 subtype, each sample was run in triplicate (Influenza H5b) with primers and probe sequences designed by the US Centers for Disease Control and Prevention, which may be acquired through the international reagent resource (https://www.internationalreagentresource.org/).

Samples that met the criteria for potential H5 or H7 involvement were forwarded to the USDA National Veterinary Services Laboratories (NVSL) in Ames, Iowa, and in Canada samples were forwarded to the National Centre for Foreign Animal Disease in Winnipeg, Manitoba. Both laboratories serve as the national reference laboratory for avian influenza in the USA and Canada, respectively, and are designated as World Organization for Animal Health reference centers.

At NVSL and National Centre for Foreign Animal Disease, confirmatory subtyping, pathogenicity determination and full or partial genomic sequencing were performed. The designation of HPAIV or LPAIV was based on sequencing of the hemagglutinin (HA) cleavage site, which indicates the pathogenic potential of the virus in poultry.

No animals or datapoints were excluded from the analyses except when specimens failed predefined laboratory quality control criteria (for example, insufficient sample material, RNA degradation or failure to meet sequencing coverage or assembly thresholds). All exclusions were determined before downstream analyses and were based solely on technical limitations rather than biological characteristics, host species, geographic location or infection status.

Virus isolation

A(H5N1) influenza viruses were isolated from wild birds in the allantoic cavities of 10-day-old embryonated chicken (Gallus gallus domesticus) eggs at 35 °C for up to 48 h. Viral titers were determined by injecting 0.1 ml of tenfold dilutions of virus into the allantoic cavities of 10-day-old eggs and then calculating the 50% egg infectious dose (EID50) or by inoculating MDCK monolayers and then calculating the 50% tissue culture infectious dose (TCID50) by the method of Reed and Muench23. The lower limit of virus detection was 1.0 log10 TCID50 ml−1 or 1.0 log10 EID50 ml−1.

Illumina sequencing

Viral RNA was extracted using a RNeasy Mini Kit (Qiagen), and cDNA was synthesized using the Superscript IV First-Strand Synthesis System (Invitrogen). The influenza A virus gene segments were amplified using modified universal primers in a multisegment PCR as described24. PCR products were purified using Agencourt AMPure XP beads according to the manufacturer’s protocol (Beckman Coulter). Libraries were prepared using the Nextera XT DNA Library Prep Kit (Illumina) according to the manufacturer’s protocol and sequenced using a MiSeq Reagent Kit v.2 (300 cycles) on a MiSeq System (Illumina). Sequencing reads were then quality trimmed and assembled using CLC Genomics Workbench (v.22.0.1). For reproduction, see completed metadata table (Supplementary Table 1) and Fig. 1a source data which includes sequence data.

Phylogenetic analyses

To assess the phylodynamics of D1.1 viruses, we used Bayesian phylogeographic diffusion in discrete space. To maximize the number of informative nucleotide substitutions available to resolve the phylogenetic history of D1.1 viruses, genome segments were concatenated (in the order of PB2, PB1, PA, HA, NP, NA, M and NS) for all viral genome sequences collected from wildlife hosts in Canada and the USA and aligned using MAFFT v.7.4925. The alignment was first used to estimate a maximum-likelihood phylogenetic tree (Extended Data Fig. 6) with IQ-TREE v.2.2.026, under the best fitting model of nucleotide substitution as determined by ModelFinder27. Node support for the resulting tree was assessed by 5,000 ultrafast bootstrap replicates. The bootstrap consensus tree was assessed for a temporal signal with a linear regression of root-to-tip divergence as a function of sampling time with TempEst v.1.5.327, using the best fitting root based on heuristic residuals mean-squared. Samples with a residual greater than 3 s.d. from the residual mean were removed from subsequent analyses.

As sampling imbalances can influence phylogeographic analyses, sequences were subset based on sampling time and location. Viruses were clustered by collection region (Alaska and the northern and southern halves of the Pacific, Central, Mississippi and Atlantic administrative flyways, for nine regions total) and date (14-day windows), and a neighbor-joining phylogenetic tree was estimated for each cluster using the nj function in the R package ape28. Each neighbor-joining tree was pruned to the most genetically diverse set of 12 tips using the phyloprunr function in the R package PDcalc29. All pruned tips were discarded from subsequent analyses. A time-scaled phylogenetic tree was estimated from this sequence subset using BEAST v.1.10.430 under the best fitting model of nucleotide substitution (as determined by ModelFinder27, a relaxed molecular clock with lognormal distribution and a GMFR Bayesian Skyride tree prior). Sampling locations of each tree tip were added as discrete character traits. Transitions between host locations were reconstructed for all internal nodes of the tree by the asymmetric substitution model and social networks inferred by Bayesian stochastic search variable selection. BEAST xml files were edited manually to log Markov rewards and the complete Markov jump history for each location. Three independent Markov chain Monte Carlo chains (200,000,000 steps, sampled every 20,000) were run. Each Markov chain Monte Carlo chain was assessed for convergence (ESS > 200 as assessed in Tracer v.1.7.2), 10% burn-in discarded, and combined to produce a maximum clade credibility tree using TreeAnnotator28. All location transition events were extracted from the maximum clade credibility tree and the posterior distribution of indicator values from the Bayesian stochastic search variable selection procedure were used to conduct Bayes factor tests to quantify statistical support for location transitions using Spread.gl. Host transitions with Bayes factor < 10.0 or posterior probability < 0.95 were excluded from the dataset.

Dataset delineation

Extended Data Fig. 1 summarizes the total number of samples collected through wildlife surveillance conducted by the groups represented in this study. Phylodynamic analyses (Figs. 1a,c) were performed using both sequences generated from this surveillance and publicly available sequences deposited in GISAID (Global Initiative on Sharing All Influenza Data), which are tabulated in Supplementary Table 3. To assess temporal changes in genotype frequencies in wild birds, an additional dataset was compiled, comprising all sequenced detections of A(H5N1) HPAI viruses in wild birds in Canada and the USA between 1 January and 31 December 2024. These data were aggregated by month and visualized as a bar chart to illustrate potential shifts in genotype proportions over time (Fig. 1b).

Statistics and reproducibility

No statistical methods were used to predetermine sample size. Sample collection was based on the availability of wild birds through ongoing active and passive surveillance programs, including hunter-harvested birds, live-captured birds and morbidity/mortality investigations, and therefore did not permit randomization or blinding. Sampling was opportunistic and designed to maximize host, temporal and geographic coverage rather than to test predefined experimental hypotheses.

All laboratory testing followed standardized and validated protocols for influenza A virus detection, subtyping and whole-genome sequencing. Diagnostic assays included appropriate positive and negative controls, and sequence data were generated using established quality control thresholds. Viral genome assemblies and genotype assignments were verified independently using several analytical pipelines where applicable (see ‘Phylogenetic analyses’ for specific tests).

Descriptive statistics were used to summarize virus detections across host species, locations and time. No exclusion of samples was performed beyond predefined quality control criteria. All analyses are reproducible using the described methods, and primary data and viral genome sequences are available through publicly accessible repositories as described in ‘Data availability.’

Serologic testing

Sera samples from ferrets (Mustela furo) were treated with receptor-destroying enzyme II (Denka Seiken Co., Ltd.) at 37 °C overnight, then heat-inactivated at 56 °C for 45 min. HI titer was determined by incubating twofold serial dilutions of serum sample with 25 µl of 4 HAU in 96-well U-bottom plates (Corning, Inc.). Sera and virus mixtures were incubated at room temperature for 45 min before addition of a 0.5% solution of chicken red blood cells (Rockland Immunochemicals, Inc.) in PBS and subsequent incubation at room temperature for 30 min. The HI titers were recorded as the reciprocal of the highest serum dilution where there was complete inhibition of hemagglutination and reported as endpoint doubling dilution or log2 as indicated in Table 1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.