The switch in receptor specificity from sialic acid (SA) linked to galactose in α2–3 linkages (avian) to α2–6 linkages (human) has been shown to be a critical step for avian A(H5N1) IAV to adapt to mammals^15–17. Molecular analysis of the HA sequence of TX/37 showed that the residues of the receptor-binding site have high homology with other clade 2.3.4.4b avian and bovine isolates (Extended Data Fig. 1a). One amino acid difference outside the receptor-binding site was observed between TX/37 and New Mexico A(H5N1) virus (A/dairy cattle/New Mexico/A240920343–93/2024; cow/NM), which was reported to exhibit a preference for both avian- and human-like SAs¹⁸ (Extended Data Fig. 1b). Here, glycan microarray analysis consistently confirmed an avian α2–3-linked SA specificity (Fig. 1a–d, glycans 4–40) for TX/37 and other representative H5 subtype viruses, in contrast to control human seasonal H3 subtype virus, which exhibited a strict human α2–6-linked SA binding preference (Fig. 1e, glycans 41–63). The only α2–6-linked SA-binding signals were from mixed α2–3- and α2–6-linked SA on the array (Supplementary Table 1, glycans 64–67). This lack of detectable binding to α2–6-linked SA was corroborated by an additional array that focused on the ability of HAs to bind to more representative human- and avian-type glycans, looking at both length and avidity of binding with glycans spotted on the array at different concentrations (Supplementary Table 2). Binding to only α2–3-linked SA was detected, with shorter and biantennary glycans preferentially bound over longer and linear glycans (Extended Data Fig. 2a,b). Increased binding breadth could be attributed to amino acids near the receptor-binding site¹⁹. The HA of all of the recent clade 2.3.4.4b viruses showed a signal for an N-glycolylneuraminic acid-linked glycan, Neu5Gcα2–3Galβ1–4(Fucα1–3)GlcNAc (Fig. 1a–d, glycan 71). Although Neu5Gc-linked glycans are not believed to be receptors for IAVs, they are abundant in bovine tissues, including bovine mammary glands, which were shown to be the preferential site of replication of the A(H5N1) viruses in cows^14,20.

Next we assessed the ability of TX/37 to infect and cause disease in the ferret model⁵. Following intranasal inoculation with a 6 log₁₀[plaque-forming unit (PFU)] dose of TX/37, ferrets rapidly developed severe disease, as evidenced by fever, weight loss, sustained diarrhoea, nasal discharge, laboured breathing and lethargy (Table 1 and Extended Data Fig. 3a–c). All inoculated animals succumbed to infection or necessitated humane euthanasia by day 2–3 post inoculation (p.i.), a more rapid decline than observed with previous 2.3.4.4b viruses isolated from animals and humans, which succumbed to infection between days 4 and 10 p.i. (refs. 7,9–11). TX/37 was detectable in ferret nasal wash (NW) and rectal swab (RS) samples at mean maximum titres of 6.9 and 3.7 log₁₀[PFUs ml⁻¹], respectively (Fig. 2a–c and Extended Data Fig. 4a–c). When subjected to post-mortem necropsy on day 3 p.i., all ferrets had high virus titres in systemic tissues, including the brain and gastrointestinal tract (Fig. 2d). Viraemia was observed in inoculated ferrets as early as 1-day p.i. (10–80 PFUs per millilitre of blood), contributing to the rapid systemic spread and accelerated fatal disease progression observed. The virus was also detected in the eye, conjunctiva and conjunctival washes collected from infected ferrets; however, overt inflammation of the conjunctiva was not noted (Fig. 2d and Extended Data Fig. 5).

Information on the transmissibility of newly emerging IAV in animal models is an important component of public health risk assessments^21,22. Here the capacity of TX/37 to transmit to naive ferrets was evaluated using three experimental set-ups (direct contact, fomite contact and respiratory droplet contact ferret transmission models). Contact of the naive ferret with the donor was established 24 h after inoculation of the donor ferrets, resulting in 30–48 h of total contact time. Efficient transmission was detected between all ferret pairs in the direct contact setting (3/3 contact ferrets). Virus was detected in NWs and RSs collected from contact ferrets at mean maximum titres of 3.9 and 1.6 log₁₀[PFUs ml⁻¹], respectively (Table 1 and Fig. 2a). These contacts exhibited weight loss, fever, sustained diarrhoea, lethargy and laboured breathing, necessitating euthanasia by days 3–4 post contact (p.c.). The similarly rapid disease progression in both the inoculated donor and contact animals indicated that virus-inoculated animals were shedding high quantities of infectious virus into the environment, leading to rapid exposure and infection of contact animals.

Recent work has shown that IAV can remain infectious in unpasteurized milk and on milking equipment materials for several hours^23,24, suggesting that exposure to contaminated surfaces may be contributing to the onward transmission of this virus. To test this, we used a fomite transmission model in which uninfected ferrets were placed in cages that had previously housed a virus-inoculated animal, so that the contact animal was exposed to cage walls, bedding, food and water that had been in direct contact with the inoculated ferret. The set-up was validated using four human seasonal and swine-origin H1 subtype viruses previously shown to be highly transmissible between ferrets placed in direct contact^25,26. Whereas the human seasonal virus exhibited efficient transmission following exposure to fomites, the three swine-origin viruses differed in their ability to transmit in this setting (Extended Data Fig. 6a–h). By contrast, a recently evaluated clade 2.3.4.4b A(H5N1) from Chile (A/Chile/25945/2023 (Chile/25945)) did not transmit in this setting⁷. When TX/37 was evaluated, transmission mediated by fomites occurred between one out of the three ferret pairs, as evidenced by the detection of virus in NW and RS samples and severe illness necessitating euthanasia of this animal on day 9 p.c. (Table 1 and Fig. 2b). The transmission outcome was accompanied by the detection of infectious virus in the air collected from the cage on the day of the first cage swap and in a cage swab sample collected on the day of the second cage swap (Extended Data Fig. 7a–c). However, the virus found on cage surfaces was largely non-infectious, suggesting that high levels of airborne infectious viral particles originating from the contaminated cage (for example, stirred up bedding and dander) could have contributed to the observed transmission.

We then assessed transmission mediated by respiratory droplets, a model in which all forms of direct and indirect contact are eliminated. In this setting, virus transmission was less efficient than in the presence of direct contact; four out of the six contact ferrets began exhibiting signs of infection on day 4 p.c., as evidenced by weight loss, fever and lethargy. Virus was detected in NW and RS samples from contact animals at mean maximum titres of 4.6 log₁₀[PFUs ml⁻¹] and 4.7 log₁₀[PFUs ml⁻¹], respectively. Although the exposure dose is typically lower in this more stringent setting, all contacts with detectable virus in NWs met humane endpoint criteria by day 4–6 p.c. (Table 1 and Fig. 2c). The remaining two contacts showed no signs of disease and remained seronegative to TX/37 at day 21 p.c., indicating that productive transmission did not occur.

Irrespective of the transmission mode assessed, all contact animals with detectable virus in NWs developed severe and fatal disease. Overall, viraemia and high-titre systemic detection of virus were observed in tissue samples from all contacts (Extended Data Fig. 8a,b). Similar to what was observed in samples from inoculated ferrets, no genetic changes at known mammalian adaptation markers were detected in samples collected from the contacts. Detected variants were not consistently enriched in contact animals, suggesting that transmission was not attributed to changes in TX/37 sequence (Supplementary Table 3). Collectively, these studies support the capacity for TX/37 to transmit to serologically naive mammals by a variety of modes, including the airborne route, although the relative efficiency of transmission varied on the basis of the setting.

In addition to its detection in cattle, clade 2.3.4.4b A(H5N1) virus has been reported in multiple other mammals and gallinaceous poultry on affected farms, including cats, raccoons and opossums^2,12,27, indicating interspecies transmission. Although our findings and others support the ability of some wild-type A(H5N1) strains to spread among ferrets in close contact, airborne transmission as seen here with TX/37, has not been observed^{7,9–11,28,29}. To more closely examine the relative ability of this virus to spread through the air, we collected airborne respiratory particles shed by TX/37-infected ferrets, to quantify infectious virus and viral RNA present in the air, using both a cyclone-based sampler (BC 251) and a water condensation sampler (SPOT). The results were compared to a 2023 2.3.4.4b A(H5N1) virus, Chile/25945, that was isolated from a human and did not transmit by the airborne route⁷, as well as human seasonal A(H1N1) and swine-origin A(H1N2)v viruses, previously shown to be highly transmissible among ferrets^25,30. All of the tested viruses were capable of robust replication in the ferret model and were detected in NWs at comparable infectious virus titres (Supplementary Table 4), with high virus levels detected early (days 1–2 p.i.), in agreement with studies supporting the critical timing of virus transmission in this model^31–34. However, infectious virus levels in the air were significantly higher and detected to day 3 p.i. (up to 246 PFUs h⁻¹) for the H1 subtype viruses as compared to Chile/25945. This virus was detectable at low levels (1–16 PFUs h⁻¹) only on day 1 p.i., further supporting its inability to transmit through the airborne route in ferrets⁷. TX/37 detection peaked later as compared to that of the well-adapted H1 viruses. The virus was detected in the air on both day 1 and 2 p.i. with most air samples containing higher titres of infectious virus (1–52 PFUs h⁻¹ on day 1 p.i. and 9–109 PFUs h⁻¹ on day 2 p.i.) relative to the titres observed in the Chile/25945 samples (Fig. 3a–d). These results suggest that TX/37 is shed in the air at levels sufficient to support the airborne transmissible phenotype observed.

To cause a pandemic, an influenza virus must be immunologically distinct from previous circulating strains, be able to productively infect humans, and be able to spread efficiently among humans³⁵. Like many other animal-origin IAVs associated with documented zoonotic human infection, clade 2.3.4.4b A(H5N1) viruses meet two of these criteria at present but lack the capacity to spread person-to-person. However, transmission of the virus in multiple mammalian species affords opportunities for the virus to adapt and become capable of meeting all three pandemic criteria³⁶. Although sustained human-to-human transmission of A(H5N1) has not been detected so far, epidemiologic reports suggest that interspecies transmission from infected dairy cattle to other mammals is taking place².

Here we show that TX/37 was capable of rapidly causing fatal, systemic infection in naive ferrets, as related clade 2.3.4.4b viruses did in infected cats on dairy farms²⁷. However, extrapolation of our findings to the human population at large should proceed with caution as multiple factors may contribute to disease outcome. As humans lack widespread immunity to H5 subtype viruses, investigation of pre-existing IAV immunity (especially due to N1 neuraminidase) may shed light on the mild human cases reported in dairy workers in the USA so far³⁷. In addition, studies presented here were conducted with young male ferrets, and high inoculation doses were tested; viral pathogenicity can be age and sex dependent and modulated by comorbidities, exposure routes and doses^38–40, which all require further evaluation. Unlike other human and bovine isolates associated with outbreaks on dairy farms, TX/37 possesses a mammalian adaptation marker in PB2, E627K, which has been shown to enhance virulence and transmission in ferrets⁴¹. As recent reports suggest that bovine isolates with sequences highly homologous to TX/37 but lacking that substitution in PB2 may also have capacity for lethal infection and limited transmission in ferrets^18,37, further investigation of additional bovine and human isolates is needed to better understand the overall public risk associated with the A(H5N1) viruses that emerged in cattle. In addition, notable differences in the sequences of bovine outbreak-related viruses and other clade 2.3.4.4b strains (for example, the presence of Q591K and D701N instead of E627K in PB2 of Chile/25945; (Supplementary Table 5) require further investigation to fully appreciate the role of strain-specific variability in airborne virus shedding and transmission phenotypes among A(H5N1) viruses circulating at present.

Our work reporting the shedding of diverse IAVs into the air using two distinct airborne particle collection approaches in the ferret model offers an advancement in our understanding of the role that infectious virus shed early after infection plays in onward transmission. The optimization and inclusion of aerobiological sampling in risk assessments can thus enhance our ability to identify strains with increased airborne transmission potential. Despite the severe disease and enhanced transmission exhibited by the TX/37 virus studied here in the ferret model, our results indicate that the virus maintains avian-like receptor-binding specificity and, therefore, would require further adaptation to augment its pandemic potential. Nonetheless, continued monitoring of the evolution of A(H5N1) viruses, especially those isolated from humans, and the animals to which humans are frequently exposed, is a critical component of pandemic preparedness.

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-024-08246-7.