Spillover of highly pathogenic avian influenza H5N1 virus to dairy cattle

The H5Nx goose/Guangdong lineage of highly pathogenic avian influenza A first emerged in 1996, initially in poultry and later in wild birds, and has since diversified through frequent reassortment into multiple clades. The H5N1 clade 2.3.4.4b has driven widespread global outbreaks since 2021–2022, causing major losses in poultry and repeated spillovers into mammals, with limited human infections reported and a low assessed risk of sustained human-to-human transmission. In North America, the lineage became established in migratory wild birds in late 2021, with extensive spread in the USA by early 2022. Against this backdrop, occasional mammalian infections (cats, foxes, bears, seals, skunks) and rare human cases have been recorded, and in March 2024 the first HPAI H5N1 infection in a ruminant (a juvenile goat) was reported. This study investigates a newly recognized spillover of HPAI H5N1 clade 2.3.4.4b into dairy cattle in the USA, characterizing clinical presentation, tissue tropism, shedding dynamics, and inter- and intra-species transmission, and assessing the genomic features and dispersal of the virus associated with farm outbreaks across multiple states.

Prior to this work, HPAI H5N1 clade 2.3.4.4b had caused extensive mortality in domestic and wild birds across multiple continents and had been detected in a range of mammals, including domestic cats, red foxes, black bears, and marine mammals such as harbour and elephant seals, with neurotropism reported in several species. Human infections associated with avian exposures have totaled 860 globally since 2003, with high case fatality in confirmed cases, though seroepidemiology suggests broader, milder infection. In North America, major wildlife events included outbreaks in harbour seals in 2023 and sporadic cases in mesocarnivores (e.g., skunk) and companion animals (cats). Experimental and observational studies have historically suggested limited susceptibility and transmission of IAVs in cattle, with earlier work from the mid-20th century demonstrating that intramammary inoculation with influenza viruses can lead to viral replication in ruminant udders, and scattered reports linked bovine influenza serology to decreased milk yield. Just before the present report, the first HPAI H5N1 infection in a ruminant (a juvenile goat) was identified in the USA in March 2024 following detection in backyard poultry on the same premises. However, sustained mammal-to-mammal transmission and a clear, documented tropism for bovine mammary tissue with milk shedding had not been demonstrated.

Clinicoepidemiological investigation: From late January to mid-March 2024, a morbidity event in dairy cattle was reported in the Texas Panhandle and surrounding states. The study investigated nine farms (farms 1–9) in Texas (1, 2, 5, 6, 7), New Mexico (4, 8), Kansas (9) and Ohio (3). Farm 3 was affected after movement of apparently healthy lactating cattle from farm 1. Clinical observations included decreased feed intake and rumination, mild respiratory signs, lethargy, dehydration, altered faeces, and abnormal colostrum-like milk, with abrupt drops in milk production per affected animal (20–100%). Proportion clinically affected ranged 3–20%, with twofold increased mortality noted in farms 2 and 3. Concurrent bird and mammal mortality (grackles, pigeons, cats, raccoon) was reported in several farms. Sampling and diagnostics: Across farms 1–9, multiple sample types were collected from cattle (milk, nasal swabs, whole blood, serum; total n=331 for rRT-PCR analysis). Initial unbiased viral metagenomic sequencing was performed on nasal swabs, serum, and buffy coats from 10 affected cows in farm 1; influenza A sequences were detected in one nasal swab. Targeted rRT-PCR assays for IAV matrix and H5 genes confirmed HPAI-H5 in bovine nasal swabs and milk. Milk had the highest viral RNA loads and highest detection frequency. Serology (haemagglutination inhibition) on paired sera (farm 2, n=19) confirmed H5N1 infection. Virus isolation and titration: Infectious virus was isolated from pellets of pooled milk samples from affected cows (farms 1 and 2). Infectious titres in milk from affected cows ranged from 10^4 to 10^8.8 TCID50 per ml. Mammary gland tissues contained high infectious titres (approximately 10^7.3–10^7.8 TCID50 per g/ml-equivalent). Shedding and duration study (farm 3): Longitudinal sampling at ~day 3 (n=15), day 16 (n=12), and day 31 (n=9) post-clinical diagnosis assessed persistence of shedding. rRT-PCR quantified viral RNA in milk, nasal swabs, whole blood, and serum; infectious virus in milk was quantified by endpoint titration (limit of detection ~10^1.05 TCID50/ml). Shedding was also compared between clinically affected and non-clinical animals and assessed in urine and faeces. Pathology and tissue tropism: Necropsies and histopathology were performed. In situ hybridization (ISH) targeting IAV matrix RNA and immunohistochemistry (IHC) localized viral RNA/antigen in tissues. In cows, mammary glands showed neutrophilic and lymphoplasmacytic mastitis with alveolar epithelial involvement; in cats, CNS lesions with viral RNA/antigen in multiple brain regions were observed. Additional bovine tissues (lung, small intestine, supramammary lymph node) were assessed. Genomics and phylogeography: Whole-genome sequencing (concatenated 8 segments) of viruses from cattle, birds, and mammals on affected farms yielded 91 sequences. Genotyping and reassortment analysis identified genotype B3.13 (PA, HA, NA, M Eurasian; NS, PB2, PB1, NP American lineages). Bayesian evolutionary analyses (BEAST) and TreeSort, haplotype networks, and phylogeographical dispersal reconstructions were used to infer ancestry, mutation profiles, and farm-to-farm transmission pathways. Comparative mutational profiling used reference sequences from the 2021–2024 USA outbreak across host species/genotypes.

• Widespread detection and high loads in milk: rRT-PCR detected H5 in 129/192 milk samples, versus 10/47 nasal swabs, 3/25 whole blood, and 1/15 serum (across farms 1–9). Milk consistently had the highest viral RNA loads.
• Infectious virus in milk and mammary tissue: Infectious H5N1 was isolated from milk pellets; milk titres ranged ~10^4 to 10^8.8 TCID50/ml. Mammary gland tissues had high infectious loads (~10^7.3–10^7.8 TCID50/ml-equivalent).
• Tissue distribution and tropism: ISH/IHC revealed a distinct tropism for milk-secreting alveolar epithelial cells in bovine mammary glands, with histologic mastitis. In cats from affected farms, viral RNA/antigen localized to CNS (neurons, glia, endothelium, Purkinje cells), consistent with neurotropism. Sparse viral signal was detected in bovine lung, supramammary lymph node, spleen, heart, colon, liver.
• Shedding dynamics (farm 3): Clinical animals’ milk was positive in 24/25 (higher loads than non-clinical 1/15). Clinical shedding in nasal swabs 6/25 and urine 2/15; no faecal RNA detected. Non-clinical animals showed RNA in nasal swabs 6/19 and urine 4/8, indicating subclinical infection. Over time, milk RNA positivity persisted (day 3: 14/15; day 16: 10/12; day 31: 4/9), but infectious virus in milk was only recovered at day 3; none at days 16 or 31.
• Clinical impact in herds: Affected animals showed 20–100% milk drop, clinical disease lasted 5–14 days, with milk production remaining depressed for ≥4 weeks. Proportion affected was 3–20% per herd; mortality doubled relative to baseline in farms 2 and 3 during events.
• Genotype and reassortment: All sequences formed the reassortant B3.13 genotype (Eurasian PA/HA/NA/M; American NS/PB2/PB1/NP). The genotype emerged via incorporation of American-lineage segments in late 2022–2023, with earliest B3.13 detections in wild birds in late 2023. Cattle-derived sequences carried additional substitutions (e.g., PB2 E362G, D441N, M631L; PA L219I; NS1 S7L) relative to earliest B3.13 sequences, suggesting evolution after spillover.
• Transmission between farms and species: Phylogenetics and dispersal analyses supported spread between geographically distinct farms: farm 2 site 1 to site 2; linkage between farms 7 and 9 (with nearby blackbird detections); and farm 1 (Texas) to farm 3 (Ohio) following movement of 42 apparently healthy cattle, indicating efficient cow-to-cow transmission from subclinically infected animals. Sequencing indicated cattle-to-cat and cattle-to-raccoon transmission on farms practicing raw milk feeding. Multidirectional interspecies transmissions were observed among cattle, birds, and mammals.

This study documents an unprecedented spillover of HPAI H5N1 clade 2.3.4.4b genotype B3.13 into dairy cattle, with clear evidence of mammary gland tropism leading to viral-induced mastitis and abundant viral shedding in milk. The findings address the central question of whether H5N1 can infect cattle and transmit in this new host: molecular, pathological, and epidemiological data collectively show efficient cow-to-cow transmission, including spread by subclinical carriers during interstate transport. The detection of closely related viruses in farm-associated cats and a raccoon, together with evidence of raw milk feeding, supports cattle-to-mammal transmission at the farm interface. Mechanistically, high expression of both α2,3- and α2,6-linked sialic acid receptors on mammary epithelial cells plausibly underlies the strong mammary tropism and milk shedding. Possible routes of entry include respiratory/oral exposure with low-level upper airway replication and transient viraemia leading to mammary infection, and direct intramammary exposure via teat orifice contamination or milking equipment. Persistent RNA detection in milk up to 31 days with loss of recoverable infectious virus after the acute phase suggests prolonged presence of viral genomes without sustained infectivity. The phylogenomic reconstruction places bovine infections within a broader wildlife context, with B3.13 circulating in wild birds and mammals before cattle spillover, and indicates both local and long-range farm-to-farm spread, likely mediated by animal movement, contaminated equipment, vehicles, personnel, and potentially wild birds. The emergence of cattle-associated mutations relative to earliest B3.13 sequences raises the possibility of early adaptation signals in this new mammalian host, warranting continued monitoring. Given three human cases linked to dairy outbreaks in other settings with mild illness, the documented efficient mammal-to-mammal transmission in cattle elevates concern for further adaptation and potential zoonotic risk, underscoring the importance of strengthened biosecurity, surveillance, and risk mitigation in dairy systems.

HPAI H5N1 clade 2.3.4.4b genotype B3.13 has spilled over into dairy cattle in the USA, where it exhibits strong mammary gland tropism, high viral loads and infectious virus in milk during acute illness, and efficient cow-to-cow transmission, including spread via subclinically infected animals. Genomic data demonstrate reassortant origin and farm-to-farm dispersal, with interspecies transmission to cats and a raccoon likely facilitated by raw milk exposure. These findings redefine a non-traditional transmission interface for H5N1 and highlight the need for enhanced farm biosecurity, milk handling precautions, and active surveillance. Future work should include controlled experimental infections comparing respiratory and intramammary routes to define portals of entry, infection dynamics, tissue distribution, and shedding; expanded molecular epidemiology with denser temporal and spatial sampling to resolve transmission networks; functional studies of cattle-associated mutations to assess host adaptation; and comprehensive risk assessments for occupational exposures and milk safety under field-relevant conditions.

Key limitations include: (1) limited number of necropsied cattle and tissue samples, constraining detailed pathogenesis inferences; (2) incomplete epidemiological information for some wildlife detections (e.g., a skunk in New Mexico), hindering definitive linkage to farm outbreaks; (3) unresolved directionality in some farm-to-farm dispersal (e.g., ambiguity between farms 7 and 9); (4) observational design without experimental infection, leaving the initial site of replication and exact routes of transmission unproven; and (5) reliance on available sequences and sampling windows, with a need for additional historical and prospective data to refine evolutionary and transmission timelines.