A human isolate of bovine H5N1 is transmissible and lethal in animal models

Abstract

The outbreak of clade 2.3.4.4b highly pathogenic avian influenza viruses of the H5N1 subtype (HPAI H5N1) in dairy cows in the US has so far resulted in spillover infections of at least thirteen farm workers^1–3, who presented with mild respiratory symptoms or conjunctivitis, and one individual with no known animal exposure who was hospitalized but recovered^3,4. Here, we characterized A/Texas/37/2024 (huTX37-H5N1), a virus isolated from the eyes of an infected farm worker who developed conjunctivitis⁵. huTX37-H5N1 replicated efficiently in primary human alveolar epithelial cells, but less efficiently in corneal epithelial cells. Despite causing mild disease in the infected worker, huTX37-H5N1 was lethal in mice and ferrets and spread systemically with high titres in respiratory and non-respiratory organs. Importantly, in four independent experiments in ferrets, huTX37-H5N1 transmitted via respiratory droplets in 17%–33% of transmission pairs and five of six exposed ferrets that became infected died. PB2–631L (encoded by bovine isolates), promoted influenza polymerase activity in human cells, suggesting a role in mammalian adaptation like that of PB2–627K (encoded by huTX37-H5N1). Additionally, bovine HPAI H5N1 viruses were found to be susceptible to polymerase inhibitors both in vitro and in mice. Thus, HPAI H5N1 virus derived from dairy cattle transmits by respiratory droplets in mammals without prior adaptation and causes lethal disease in animal models.

Introduction

Since the highly pathogenic avian influenza (HPAI) H5N1 (clade 2.3.4.4b) outbreak began in US dairy cattle in early 2024, infections have been reported in 214 herds in 14 US states⁶ (accessed on September 20, 2024). Recently, we characterized an HPAI H5N1 virus isolated from the milk of an infected cow and found that it replicated in respiratory and non-respiratory organs and caused fatal disease in mice infected orally^7,8 or intranasally⁷, as well as in intranasally⁷ infected ferrets. Respiratory droplet transmission in ferrets was inefficient; no virus was detected in exposed animals, but one of four exposed ferrets seroconverted⁷.

Viruses closely related to those detected in dairy cattle also have been isolated from poultry on nearby premises, from cats on affected farms, and from house mice^9,10. Moreover, 14 human cases of infection by HPAI H5N1 viruses of the bovine lineage in four US states have been reported as of September 20, 2024, most with mild symptoms (one individual with no known animal exposure was hospitalized but recovered)^1–4. In late March 2024, a dairy farm worker who had been in contact with sick cows presented with conjunctivitis but no fever or respiratory symptoms⁵. The virus isolated from a conjunctival swab (A/Texas/37/2024; referred to as ‘huTX37-H5N1’) is closely related to the viruses isolated from dairy cattle but possesses the PB2-E627K substitution, a known marker of mammalian-adaptation that frequently arises during the replication of avian influenza viruses in mammals^11,12. The US Centers for Disease Control and Prevention (CDC) reported respiratory droplet transmission in one of three ferret transmission pairs infected with huTX37-H5N1¹³, suggesting that this isolate (encoding PB2-E627K) may be more transmissible than the isolate from dairy cattle we previously tested⁷. Therefore, we characterized the huTX37-H5N1 virus in primary human cells grown at the air-liquid interface and in mice and ferrets, the two most used animal models in influenza virus research. We also examined the effects of potential mammalian-adapting mutations on polymerase activity and assessed the antiviral sensitivity of the bovine HPAI H5N1 virus.

Results

Replication in primary human cells.

To better understand how HPAI H5N1 viruses derived from dairy cattle replicate in human cells, we performed growth curve analyses of two different human primary epithelial cell types grown in air-liquid interface (ALI) cultures: alveolar (lower respiratory tract) and corneal (outer surface of the eye). Cells were infected with three different viruses: huTX37-H5N1; a clade 2.3.4.4b HPAI H5N1 virus isolated from the milk of a lactating dairy cow (A/dairy cattle/New Mexico/A240920343–93/2024; referred to as ‘NM93-H5N1’); and an avian H5N1 virus isolated from an infected human (A/Vietnam/1203/2004; referred to as ‘VN1203-H5N1’) at two different multiplicities of infection [0.1 plaque forming units (PFU)/cell and 0.01 PFU/cell]. A comparison of the amino acid differences between the two dairy cattle-derived HPAI H5N1 viruses is provided in Extended Data Table 1. The cells were incubated at 33 °C or 37 °C, and every 24 h for 96 h after infection, apical surfaces were washed to collect secreted viruses. In alveolar epithelial cells, huTX37-H5N1 replicated to high titres at both 33 °C (Fig. 1A) and 37 °C (Fig. 1B). NM93-H5N1 and VN1203-H5N1 replicated to similar levels as huTX37-H5N1 at 37 °C; however, both exhibited 1–2 log lower titres at 33 °C. The PB2–627K substitution (which is expressed in huTX37-H5N1) has been reported to increase viral replication at 33 °C in some mammalian cell types¹⁴. However, since the VN1203-H5N1 virus also possesses PB2–627K, other factors must contribute the higher replicative ability of huTX37-H5N1 in alveolar epithelial cells at 33 °C. All three viruses replicated in the corneal epithelial cells, albeit at lower levels than those observed in respiratory cells (Fig. 1). While VN1203-H5N1 was more consistently detected across all timepoints at both 33 °C and 37 °C in corneal epithelial cells, the highest titres were observed in huTX37-H5N1 infections at both temperatures.

Figure 1. Replication in human epithelial cells.

Primary human alveolar or corneal epithelial cells grown in air-liquid interface cultures were infected with MOIs of 0.01 or 0.1 PFU per cell of huTX37-H5N1 (N=3 biological replicates per condition), NM93-H5N1 (N=2 biological replicates per condition), or VN1203-H5N1 (N=2 biological replicates per condition) and incubated at 33 °C (A) or 37 °C (B). Every 24 h for 96 h, apical surfaces were washed to collect secreted virus, which was titred using plaque assays in MDCK cells. For all conditions, all replicate data points were plotted individually, the median value is indicated by the floating bar, and variability is represented as the range. h, hours; PFU/ml, plaque-forming units per millilitre.

Pathogenicity in mice.

We previously demonstrated that NM93-H5N1 spreads systemically, causing severe disease in mice⁷. To assess the pathogenesis of huTX37-H5N1, we first determined its mouse lethal dose 50 (MLD₅₀). Groups of female BALB/cJ mice were intranasally inoculated with serial dilutions of huTX37-H5N1 (10⁰ to 10⁶ PFU, N=5 animals/dose) and body weights and survival were monitored daily for 13 days. For doses ≥ 10¹ PFU, all mice exhibited rapid body weight loss (Fig. 2A) and succumbed to their infection by day 6 (Fig. 2B). Among mice infected with 10⁰ PFU (i.e., 1 PFU/mouse) two of five showed disease similar to that seen in mice given higher doses, whereas the other three mice showed no obvious signs of disease in the first 6 days; subsequently, all five mice lost weight and died by day 13 post-inoculation. Based on these observations, the MLD₅₀ of huTX37-H5N1 is less than 1 PFU, which is lower than that of NM93-H5N1 (MLD₅₀ = 31.6 PFU)⁷ and VN1203-H5N1 (MLD₅₀ = 2.2 PFU)¹⁵. Therefore, huTX37-H5N1 seems to be more lethal in mice than the cow isolate.

Figure 2. huTX37-H5N1 is lethal in mice and spreads systemically.

(A) and (B) BALB/cJ mice (7 weeks old, N=5 biologically independent animals per dosage) were deeply anaesthetised and intranasally inoculated with 10-fold-serial dilutions of A/Texas/37/2024 (H5N1) in 50 μl of PBS. (A) Body weight and (B) survival were monitored daily for 13 days. In panel (A), points represent mean values and error bars represent the standard deviation. (C) BALB/cJ mice (7 weeks old, N=10 biologically independent animals per virus) were deeply anaesthetised and intranasally inoculated with 10³ PFU of huTX37-H5N1 in 50 μl of PBS. At 3 and 5 days post-infection, five mice were euthanised and tissues were collected for plaque assays in MDCK cells. In panel (C), points represent titres of individual mice, the floating bars show the median titre for each tissue of each inoculation group, and variability is represented by the range. PFU, plaque forming units; d, days; PFU/g, plaque-forming units per gram of tissue; PFU/ml, plaque-forming units per millilitre.

To examine the tissue tropism of huTX37-H5N1 in mice, female BALB/cJ mice (N=10) were intranasally inoculated with 10³ PFU, and tissues (blood, eye, teat, mammary gland, brain, intestine, liver, spleen, kidney, heart, hamstring, latissimus dorsi, nasal turbinate, trachea, and lung) were collected on day 3 or day 5 post-infection (N=5 mice per day) for virus titration using plaque assays in MDCK cells (Fig. 2C). On day 3, virus was detected in all tissues assessed in all mice, with a minimum median value of 10⁴ PFU per gram of tissue (in teat) and a maximum median value of 10^8.6 PFU per gram of tissue (in lung). Overall, the highest viral loads were detected in the respiratory tissues (nasal turbinate, trachea, and lung), but we also detected > 10⁶ PFU per gram of tissue in the spleen and > 10⁵ PFU per ml in the blood. On day 5, virus titres were similar to those observed on day 3 for all tissues except colon, in which no virus was detected, and no statistically significant difference in virus titre in any tissue was observed between day 3 and day 5. We previously showed that NM93-H5N1 (derived from a cow) also spreads systemically in mice, with high virus loads in non-respiratory tissues and substantial levels of virus in the blood. Although we did not compare NM93-H5N1 and huTX37-H5N1 side-by-side in this study, a comparison of NM93-H5N1 tissue titres from our previous study⁷ and the data presented here suggests that huTX37-H5N1 may replicate more efficiently in early infection, as indicated by higher median virus titres and more uniform detection across replicate animals at day 3 (Extended Data Table 2).

Pathogenicity in ferrets.

Previously, we showed that NM93-H5N1 (10⁶ PFU given intranasally) causes severe disease with systemic virus spread, but limited lethality, in ferrets⁷. To evaluate the pathogenicity of the human isolate, we intranasally infected female ferrets (N=8) with 10⁶ PFU and assessed clinical disease, survival, and viral tropism for the same tissues we assessed in mice. All ferrets inoculated with huTX37-H5N1 exhibited body weight loss (Fig. 3A) and decreasing body temperature (after an initial fever) (Fig. 3B); all succumbed to their infection by day 5 (Fig. 3C). Clinical symptoms of disease were generally consistent, with all animals exhibiting a fever on day 1, additional symptoms (i.e., dishevelled fur, reduced activity, and diarrhea) emerging on day 2, and further deterioration (i.e., inability to stand and/or substantial body temperature reduction) prior to humane euthanasia or death (Fig. 3D). As observed in mice, virus was detected in all assessed tissues, with high viral loads in the respiratory tract (median values of 10^8.2, 10⁷, and 10^7.4 PFU/gram of nasal turbinate, trachea, or lung, respectively, on day 3) and blood (median value 10^6.8 PFU/ml on day 3) (Fig. 3E). Virus was also abundant in the liver, with median values of 10^7.7 PFU or 10^8.1 PFU per gram of tissue on day 3 and day 4, respectively, although the significance of this observation is currently unclear. As in mice, huTX37-H5N1 exhibited higher median titres and was more uniformly detected across replicate ferrets in all tissues examined compared to NM93-H5N1 (see ⁷ and Extended Data Table 3). These data indicate that huTX37 is more pathogenic than the closely related NM93-H5N1 virus in the ferret model.

Figure 3. huTX37-H5N1 is lethal in ferrets and spreads systemically.

Ferrets (6–8 months old, N=8 biologically independent animals per virus) were deeply anaesthetised and intranasally inoculated with 10⁶ PFU of huTX37-H5N1 in 500 μl of PBS. Body weights (A), body temperature (B), survival (C), and clinical symptoms (D) were monitored daily; virus titres in various tissues (E) were determined by use of plaque assays in MDCK cells. In panels (A) and (B), each line represents an individual ferret. In panel (D), symptom scores were plotted by adding one point per symptom observed on each day. Some ferrets died between health checks (indicated by a ‘D’ on the figure panels), and others were euthanised after meeting humane endpoint criteria (indicated by an ‘E’ on the figure panels). A ‘D’ or ‘E’ shown in black text indicates the outcome occurred on the same day, while gray text indicates the outcome occurred previously. In panel (E), points represent titres of individual ferrets (note, the triangular points represent the ferret that died before a health check on day 3), the floating bars show the median titre for each tissue at each timepoint, and variability is represented by the range. PFU, plaque forming units; d, days; PFU/g, plaque-forming units per gram of tissue; PFU/ml, plaque-forming units per millilitre.

Transmissibility in ferrets.

We previously reported inefficient respiratory droplet transmission of NM93-H5N1 in ferrets (i.e., one of four exposed ferrets seroconverted without virus detection)⁷. The US CDC has reported limited (33%) respiratory droplet transmission of huTX37-H5N1 (the same virus used in this study) in ferrets¹³. Given that the ability of bovine H5N1 viruses to transmit by respiratory droplets in mammals would increase their pandemic potential, we comprehensively evaluated huTX37-H5N1 respiratory droplet transmission in ferrets, using an experimental system described in detail elsewhere¹⁶. Groups of female ferrets were infected with 10⁶, 10³, 10², or 10¹ PFU of huTX37-H5N1 (N=6 ferrets/dose) and one day later, naïve animals (N=1 naïve ferret per infected ferret) were housed in cages adjacent to the infected animals but separated by about 5 cm to prevent direct contact. Nasal swabs were collected every other day starting from day 1 post-infection or post-exposure and virus titres were quantified by use of plaque assays in MDCK cells (Fig. 4A–D).

Figure 4. huTX37-H5N1 transmits by respiratory droplets in ferrets.

Ferrets (6–8 months old, N=6 animals per virus) were deeply anaesthetised and intranasally inoculated with 10⁶ PFU (A), 10³ PFU (B), 10² PFU (C), or 10¹ PFU (D). One day later, naïve ferrets (N=1 animal per infected animal) were placed in adjacent cages allowing for air flow but no direct contact between animals. Nasal swab samples were collected at the indicated timepoints and tested by using plaque assays in MDCK cells. In each figure panel, from left to right are plots of survival of infected donors, survival of respiratory droplet contacts, virus titres in nasal swabs of infected donors, and virus titres in nasal swabs of respiratory droplet contacts. Dotted lines in the virus titre plots represent the limit of detection. In panel (A), the asterisk indicates a respiratory droplet contact animal from pair 3 for which a low amount of virus was detected in the nasal swab at only the day 3 post-exposure timepoint, no virus was detected in any other tissue from the same animal upon necropsy. PFU, plaque forming units; PFU/ml, plaque-forming units per millilitre; d, days; n.d., not determined. Panel (E) shows virus titres in tissues of respiratory droplet contacts collected at the time of euthanasia or within 14 h of death. Each dot represents the titre of an individual ferret.

Ferrets infected with 10⁶ PFU exhibited body temperature and weight changes consistent with our earlier observations (Extended Data Fig. 1A and 1B; compare to Fig. 3A and 3B), and all six died by day 4 post-infection (Fig. 4A and Extended Data Fig. 1C). Virus was detected in the nasal swabs of all infected ferrets on days 1 and 3 post-infection (Fig. 4A) and had spread systemically in all animals prior to death (Extended Data Fig. 1D). Importantly, we identified > 10⁴ PFU of virus in the nasal swabs of two of the six contact ferrets (Fig. 4A), both of which exhibited clinical symptoms and either died or were euthanised by day 6 post-exposure. A third ferret (from pair 3, indicated by the asterisk in the figure), which was euthanised at day 4 post-exposure due to excessive body weight loss, yielded a single plaque in the undiluted well of the plaque assay, and attempts to amplify virus from the nasal swab sample in embryonated chicken eggs were unsuccessful. Like the infected ferrets, virus spread systemically in two of the three virus-positive contacts (i.e., the contacts from pairs 5 and 6; Fig. 4E), whereas no virus was detected in any respiratory or non-respiratory tissue sample from the virus-positive contact animal from pair 3 except for the day 3 nasal swab. Among the three surviving contact ferrets, none seroconverted. Therefore, huTX37-H5N1 transmitted by respiratory droplets from at least 33% of ferrets inoculated with 10⁶ PFU, which is consistent with the data reported by the US CDC.

All ferrets inoculated with lower doses (10³ PFU, 10² PFU, or 10¹ PFU) exhibited symptoms similar to those observed for the 10⁶ PFU inoculation group (Extended Data Fig. 1A and 1B), and all succumbed by day 6 post-infection (Fig. 4B, 4C, and 4D, respectively), showing systemic virus spread (Extended Data Fig. 1D). Despite the 5-log difference between the highest (10⁶ PFU) and lowest (10¹ PFU) doses tested, there was no significant difference in the survival curves of animals infected with any of the doses (Extended Data Fig. 1C), although some animals infected with 10² or 10¹ PFU exhibited a delay of 1–2 days in their time to death. In addition, although virus was detected in the nasal swabs of infected ferrets at only one timepoint for most animals infected with the lower doses (i.e., day 3 post-infection for ferrets infected with 10³, 10², or 10¹ PFU; note, in one ferret inoculated with 10² PFU, virus was detected on day 3 and day 5 post-infection, Fig. 4C), and respiratory droplet transmission occurred in one of six contacts (16.7%) for all three doses, as indicated by virus detection in the nasal swabs and lethal disease characterized by systemic virus spread in the contact animals (Fig. 4B–E). Further, among the contact ferrets that survived, one from the 10² PFU infection group (contact #6) exhibited evidence of seroconversion, with a hemagglutination inhibition titre of 40 (none of the other surviving contact ferrets in the 10³ or 10¹ PFU inoculation groups seroconverted). Therefore, huTX37-H5N1 transmitted via respiratory droplets in 17%–33% of transmission pairs after inoculation with 10³, 10², or 10¹ PFU.

Together, our experiments reveal that the huTX37-H5N1 virus: (1) is highly lethal in ferrets, with all infected animals, whether directly infected or infected via respiratory droplets, exhibiting fatal disease characterized by systemic virus spread and high virus levels in respiratory tissues, liver, and blood; (2) transmits by respiratory droplets despite the rapid course of lethal disease; and (3) appears to be transmitted via respiratory droplets slightly more efficiently than the cow NM93-H5N1 isolate⁷, although a side-by-side comparison was not performed.

Deep sequencing of ferret nasal swabs.

To determine if amino acid sequence variants arose after huTX37-H5N1 replication in the ferret upper respiratory tract and which (if any) were transmitted, we performed deep sequencing of RNA extracted from the huTX37-H5N1 virus stock (i.e., inoculum) and all available nasal swab samples of infected and exposed ferrets from which virus was recovered (we were not able to sequence virus from the exposed ferret from pair 3 of the 10⁶ PFU dosage infection). Focusing on non-synonymous mutations detected in at least 3% of the sequence reads, we identified several polymorphisms in the huTX37-H5N1 virus stock, including PA-655L/F, NA-454G/D, and NS1–164P/S (Supplementary Tables 1 and 2). In the ferret nasal swabs (infected and exposed animals), there was a wide range of polymorphism ratios at these positions, with no dominant pattern (Supplementary Tables 1 and 2; note, exposed ferrets are indicated by light orange in Supplementary Table 2). Besides the polymorphisms already present in the virus stocks (indicated by asterisks in Supplementary Table 2), there were few amino acid substitutions overall, none were found in more than one sample, and only two samples obtained from different ferrets had multiple substitutions (Supplementary Table 2, sample names: ‘10^2pfu_I2_d3’ and ‘10^3pfu_E5_d5’). While these data suggest a lack of strong selective pressure against huTX37-H5N1 during replication and transmission in ferrets, the limited time frame for replication prior to the animals becoming moribund could have restricted the emergence of ferret- or mammalian-adapting substitutions.

Mutations affecting polymerase activity.

Tissue tropism studies in mice (Fig. 2C) and ferrets (Fig. 2C, Fig. 3E, and Extended Data Fig. 1D) suggest that huTX37-H5N1 replicates more efficiently in both respiratory and non-respiratory tissues than a bovine H5N1 virus isolated from a lactating dairy cow⁷. One likely contributing factor is the PB2-E627K substitution—a known marker of mammalian adaptation that frequently arises during the replication of avian influenza viruses in mammals and promotes replicative ability^11,12—which is present in huTX37-H5N1 but not HPAI H5N1 viruses isolated from dairy cattle. Dairy cattle HPAI H5N1 viruses encode PB2–631L (rather than PB2–631M, which is highly conserved among influenza A viruses), which allows influenza viruses to replicate in genetically modified chickens that do not express functional ANP32A¹⁷ (the host factor implicated in PB2–627-conferred host restriction), suggesting that the amino acid at PB2–631 may function like that of PB2–627. However, whether PB2–631 contributes to influenza polymerase activity and the effects of different amino acid substitutions at this locus are unknown. To test this, we carried out minireplicon assays with plasmids derived from a clade 2.3.4.4b cow H5N1 virus (A/dairy cattle/Texas/24–008749-001/2024; referred to as ‘TX001-H5N1’), which expresses PB2–627E and PB2–631L, and mutant PB2 plasmids expressing either PB2–627K or PB2–631M (Fig. 5A). As expected, the PB2-E627K variant (expressed in huTX37-H5N1 but not cow H5N1 isolates) strongly enhanced polymerase activity in human (Fig. 5B), but not avian (Fig. 5C), cells. The PB2-L631M substitution greatly reduced polymerase activity in human cells (Fig. 5B) but had only a minor effect in avian cells (Fig. 5C), suggesting that PB2–631L (expressed in all cow H5N1 isolates but not in huTX37-H5N1) confers much higher replicative ability than PB2–631M in human cells.

Figure 5. Effects of mutations on viral polymerase activity.

Panel (A) indicates the PB2 amino acids at positions 627 and 631 for each of the TX001-H5N1 polymerase complexes used in the mini-replicon assays. The amino acids encoded by the huTX37-H5N1 virus are shown for comparison. Human 293T (B) or avian DF-1 (C) cells were transfected with plasmids expressing TX001-H5N1 polymerase complex proteins and a plasmid expressing a reporter vRNA encoding the firefly luciferase gene. After incubating cells for 24 h at the temperatures indicated on the figure panels, luciferase activity was measured. Polymerase activity was calculated by standardizing firefly luciferase activity (vRNA) to Renilla luciferase activity expressed from a control plasmid and is shown as relative luciferase activity. The plots show the mean and standard deviation of 8 data points collected over 3 replicate experiments for each PB2 protein. Groups were compared by using non-parametric two-tailed Mann-Whitney tests with a Holm-Sidak adjustment for multiple comparisons. Asterisks (***) on the panels indicate that p < 0.001 for comparisons between the wild-type TX001-H5N1 polymerase complex (with PB2–627E and PB2–631L) and a given mutant. In human cells (panel B), the polymerase activity of the TX001-H5N1 PB2-E627K mutant was significantly increased at 37 °C (p = 0.000311) and at 33 °C (p = 0.000311), and the polymerase activity of the TX001-H5N1 PB2-L631M mutant was significantly decreased at 37 °C (p = 0.000311) and at 33 °C (p = 0.000311). In avian cells (panel C), the polymerase activity of the TX001-H5N1 PB2-E627K mutant was not significantly different at 39 °C (p = 0.382284) or 37 °C (p = 0.102880), and the polymerase activity of the TX001-H5N1 PB2-L631M mutant was significantly decreased at 39 °C (p = 0.000622) and at 37 °C (p = 0.000622).

Receptor binding.

Our earlier report⁷ showed that NM93-H5N1 preferentially binds to α2,3-linked sialic acids (which are abundant in the gastrointestinal tract of avian species) but also binds to α2,6-linked sialic acids (which predominate in the upper respiratory tract of humans). The ability to bind to α2,6-linked sialic acids is a feature of influenza viruses that spread by respiratory droplets in humans and is a key factor in the risk assessment of viruses with pandemic potential. In the same assay used to assess NM93-H5N1 receptor binding, the huTX37-H5N1 virus showed preferential binding to α2,3-linked sialic acids, but also bound to α2,6-linked sialic acids (Extended Data Fig. 2). As expected, seasonal human Isumi-H1N1 preferentially bound to α2,6-linked sialic acids and the VN1203-H5N1 virus preferentially bound to α2,3-linked sialic acids. Our data contrast with that of others^18–20, who found little-to-no binding of recombinant HA proteins^18,20 or virions¹⁹ of bovine HPAI H5N1 to α2,6-linked sialic acids.

Antiviral susceptibility.

Lastly, we examined the susceptibility of bovine H5N1 viruses to available antiviral drugs, which are an essential countermeasure against influenza. The neuraminidase (NA) T438I substitution, which is known to reduce inhibition of HPAI H5N1 viruses by zanamivir and peramivir in vitro²¹, was identified in one of the publicly available NA sequences derived from a cow H5N1 virus²². TX001-H5N1 carrying the T438I substitution exhibited normal antiviral sensitivity to oseltamivir carboxylate and reduced sensitivity to other NA inhibitors (i.e., zanamivir, laninamivir, and peramivir) in vitro (Extended Data Table 4). The wild-type TX001-H5N1 virus was sensitive to zanamivir in vitro (Extended Data Table 4), and intranasal treatment of mice delayed body weight loss and death and significantly reduced virus titres in lung, nasal turbinate and brain, but did not protect against lethality (Fig. 6; left panels). Minimal protection against body weight loss, viral replication in tissues, or survival was observed in mice orally treated with oseltamivir phosphate (Fig. 7; left panels); similar observations have been made for mice infected with a highly pathogenic H7N9 virus and treated with neuraminidase inhibitors²³. As expected, in mice infected with TX001-H5N1-T438I, intranasal zanamivir (Fig. 6; right panels) or oral oseltamivir phosphate (Fig. 7; right panels) treatment had minimal effects on mouse body weight loss or viral replication in the nasal turbinate, lung, or brain (the only significant difference was observed in lung titres of zanamivir-treated mice at day 3 post-infection) and had no effect on survival. In contrast, the polymerase inhibitors favipiravir and baloxavir marboxil were highly effective against both wild-type TX001-H5N1 and TX001-H5N1-T438I in vitro (Extended Data Table 5) and in vivo (Fig. 7).

Figure 6. Zanamivir has minimal effects on the outcome of TX001-H5N1 infection in mice.

Female BALB/c mice (N=15 per treatment condition) were deeply anaesthetised and intranasally infected with 10 PFU of wild-type A/dairy cattle/Texas/24–008749-001/2024 (H5N1) or a mutant virus expressing neuraminidase T438I. At 1 h post-infection, mice were treated with the indicated compounds daily for 5 days. (A) Body weights and survival were monitored daily for 14 days (N=5 mice per treatment condition). In the plots showing weight loss, each point represents the mean value, and the error bars represent the standard deviation. (B) Virus titres in the lung, nasal turbinate (NT), or brain were determined by using plaque assays in MDCK cells at day 3 (N=5 mice per treatment condition) and day 5 (N=5 mice per treatment condition) post-infection. Each point represents an individual titre value, bars represent the mean titre, and the error is represented by standard deviation. Titres were compared by using a non-parametric two-tailed Mann-Whitney test without adjustments for multiple comparisons. Asterisks on the panels indicate that p < 0.05 (*) or p < 0.01 (**). In mice infected with the wild-type (WT) virus, significantly reduced titres were observed in the lung (p = 0.0079), nasal turbinate (p = 0.0238), and brain (p = 0.0159) tissues of zanamivir-treated mice at day 3 post-infection, and in the lung (p = 0.0079), nasal turbinate (p = 0.0079), and brain (p = 0.0079) tissues of zanamivir-treated mice at day 5 post-infection. In mice infected with the NA-T438I mutant virus, significantly reduced titres were observed only in the lung tissues (p = 0.0079) of zanamivir-treated mice at day 3 post-infection. d, days; PFU/g, plaque forming units per gram of tissue.

Figure 7. Bovine H5N1 is susceptible to polymerase inhibitors.

Female BALB/c mice (N=15 per treatment condition) were deeply anaesthetised and intranasally infected with 10 PFU of wild-type A/dairy cattle/Texas/24–008749-001/2024 (H5N1) or a mutant virus expressing neuraminidase T438I. At 1 h post-infection, mice were treated with the indicated compounds daily for 5 days. (A) Body weights and survival were monitored daily for 14 days (N=5 mice per treatment condition). In the plots showing weight loss, each point represents the mean value, and the error bars represent the standard deviation. (B) Virus titres in the lung, nasal turbinate (NT), or brain were determined by use of plaque assays in MDCK cells at day 3 (N=5 mice per treatment condition) and day 5 (N=5 mice per treatment condition) post-infection. Each point represents an individual titre value, bars represent the mean titre, and the error is represented by standard deviation. Titres were compared by using a non-parametric one-sided Kruskal–Wallis test with the Dunn’s multiple comparisons procedure. Asterisks on the panels indicate that p < 0.05 (*) or p < 0.01 (**). WT, wild-type; d, days; PFU/g, plaque-forming units per gram of tissue.

Discussion

Recently, we reported that A/dairy cattle/New Mexico/A240920343–93/2024 (H5N1, clade 2.3.4.4b, genotype B3.13; NM93-H5N1) does not transmit efficiently among ferrets via respiratory droplets⁷. The US CDC reported respiratory droplet transmission of huTX37-H5N1 in 1 of 3 transmission pairs¹³, suggesting that the human isolate may transmit more efficiently than dairy cow isolates via respiratory droplets. Here, we carried out ferret transmission studies with four different inoculation doses (i.e., 10⁶, 10³, 10², and 10¹ PFU) of huTX37-H5N1 virus. In each study, respiratory droplet transmission was detected in at least two (10⁶ and 10² PFU inoculation doses) or one (10³ and 10¹ PFU inoculation doses) transmission pairs with high virus load and lethality in five of six exposed animals that became infected. These findings demonstrate that HPAI H5 viruses from the dairy cattle outbreak can transmit among mammals via respiratory droplets (albeit with limited efficiency), and that the risk of respiratory droplet transmission may be higher with human than with dairy cattle isolates.

Historically, HPAI H5 viruses have not been uniformly lethal to ferrets. However, recent reports show that clade 2.3.4.4b HPAI H5N1 viruses may have a greater propensity to cause lethal disease^24–26. Pulit-Penaloza et al.²⁵ reported 100% lethality in ferrets intranasally inoculated with 10⁶ egg infectious dose 50 (EID₅₀) of a clade 2.3.4.4b virus isolated from an infected human in Chile (A/Chile/25945/2023). Belser et al.²⁴ showed that intraocular inoculation of ferrets with the same virus at a high dose (10⁶ PFU) or lower dose (10³ PFU) resulted in 100% or 66.7% lethality, respectively. Restori et al.²⁶ comprehensively examined a clade 2.3.4.4b mink isolate (A/mink/Spain/3691–8_22VIR10586–10/2022) and found 75%–100% lethality in ferrets inoculated with 10⁶, 10³, 10², and 10¹ EID₅₀, and 50% lethality in ferrets inoculated with 10⁰ EID₅₀. Here, we show 100% lethality in ferrets intranasally inoculated with four different doses of huTX37-H5N1 (10⁶, 10³, 10², and 10¹ PFU) within 3–6 days of inoculation. Moreover, five of six exposed ferrets succumbed to the infection (the exposed ferret that survived exhibited evidence of transmission only by seroconversion), even though the amount of transmitted virus should have been low. A News Report from the CDC¹³ also described 100% lethality after inoculation of ferrets with huTX37-H5N1, but the inoculation dose was not reported. In contrast, we recently found that infection of ferrets with 10⁶ PFU of NM93-H5N1 killed only 1 of 4 infected ferrets⁷. The huTX37-H5N1 virus possesses mammalian-adapting mutations in PB2, but its high lethality in ferrets cannot be explained by these residue(s) alone because other HPAI H5 viruses encoding mammalian-adapting residue(s) in PB2 (such as A/Vietnam/1203/2004) are not uniformly lethal in ferrets^7,15.

The PB2-E627K substitution is a well-established marker of mammalian-adaptation, conferring efficient replication to avian influenza viruses in mammals^11,12. It is encoded by huTX37-H5N1 but not by any of the HPAI H5 dairy cattle isolates. Interestingly, the HPAI H5 dairy cattle isolates encode PB2–631L, whereas huTX37-H5N1 encodes PB2–631M, which is highly conserved among influenza A viruses. PB2–631L increased the pathogenicity and neurovirulence of a mouse-adapted H10N7 virus²⁷, facilitates influenza virus replication in chickens with gene-edited ANP32A¹⁷, the host factor essential for host restriction conferred by PB2–627²⁸, and is located close to residue 627 in the three-dimensional structure²⁹. As expected, introduction of the PB2-E627K substitution into an HPAI H5 dairy cattle isolate polymerase complex increased viral polymerase activity in minireplicon assays in mammalian but not in avian cells. Replacement of PB2–631L (encoded by dairy cattle isolates) with PB2–631M (commonly found at this position) greatly reduced viral polymerase activity in minireplicon assays in mammalian (but not avian) cells, demonstrating that the PB2-M631L substitution facilitates avian influenza virus replication in mammals. Thus, in addition to known the mammalian-adapting substitutions (e.g., PB2-E627K or -D701N), the PB2-M631L substitution may also confer efficient replication in mammals by increasing viral polymerase activity in mammalian cells. It is interesting to note that huTX37-H5N1 is phylogenetically slightly different from the dairy cattle isolates^8,30,31. We speculate that a common ancestor evolved into isolates encoding PB2–627K/631M (such as huTX37-H5N1) or PB2–627E/631L (i.e., the dairy cattle isolates), both of which facilitate avian influenza virus replication in mammalian cells (Fig. 8).

Figure 8. Schematic representation of the possible evolution of dairy cow influenza lineages.

PB2–627 and PB2–631 amino acids are shown for the avian progenitor virus, the huTX37-H5N1 virus, and a representative dairy cattle isolate (NM93-H5N1).

Phylogenetically, the huTX37-H5N1 virus is separated by a long branch length from viruses originating in cattle and is not representative of the H5N1 viruses of the cattle outbreak, although huTX37-H5N1 and cattle viruses likely have a common avian virus ancestor^8,30,31. huTX37-H5N1 encodes PB2–627K and PB2–631M, whereas NM93-H5N1 and TX001-H5N1 (representative dairy cattle isolates of the cattle outbreak) encode PB2–627E and PB2–631L (Extended Data Table 1 and Fig. 8). PB2-E627K is well-established as a mammalian-adapting mutation^11,12, and our data indicate that PB2-M631L is also a mammalian-adapting mutation (Fig. 5). A/Canada_goose/Wyoming/24–003692-001-original/2024 (‘cgWY001-H5N1’) of the B3.13 subclade was isolated in January 2024³⁰, that is, around the time when the jump from the avian reservoir into cattle occurred³⁰. It has the closest sequences to those of the H5N1 viruses of the dairy cattle lineage. cgWY001-H5N1 encodes avian-type PB2–627E and PB2–631M (the latter substitution is found in most influenza A viruses). Although it is possible that two substitutions (PB2–627E to PB2–627K and PB2–631L to PB2–631M) occurred during the generation of huTX37-H5N1, it is more likely that only the PB2–627E to PB2–627K substitution occurred in the progenitor virus, which possessed PB2–627E and PB2–631M. Therefore, we propose that huTX37-H5N1 and the rest of the bovine H5N1 viruses identified to date could have independently acquired key mammalian-adapting substitutions (i.e., PB2-E627K in the case of huTX37-H5N1 and PB2-M631L for the bovine isolates). This hypothesis is consistent with a recent finding³² that PB2-E627K arose after replication of a current European wild bird H5N1 isolate in experimentally infected lactating dairy cattle³².

Among the dairy cattle HPAI H5N1 viruses isolated from humans so far, only huTX37-H5N1 has the PB2–627K substitution, which is known to enhance viral polymerase activity^33,34 and pathogenicity in mammals¹¹. Therefore, the PB2–627K substitution likely plays a major role in the extreme pathogenicity of huTX37-H5N1 in ferrets. Whether the high replicative ability afforded by the PB2–627K substitution also contributes to huTX37-H5N1 transmissibility is unknown. However, the US CDC has reported that a second human isolate associated with the dairy cattle outbreak [A/Michigan/90/2024 (H5N1)], which encodes PB2–627E and PB2–631L (like most bovine H5N1 viruses, including NM93-H5N1) also transmits by respiratory droplets in ferrets to an extent similar to huTX37-H5N1³⁵. Thus, respiratory droplet transmission may occur irrespective of the presence or absence of PB2–627K. Our data also show that huTX37-H5N1 transmission is not dependent on the amount of virus used for infection of donor animals, suggesting that regardless of the inoculum dose, the virus titres in donor animals reached a level sufficient for transmission. Additional respiratory droplet transmission studies in ferrets, along with aerosol sampling studies may shed light on the viral characteristics that are important for respiratory droplet transmission in mammals.

huTX37-H5N1 is highly lethal in mice and ferrets, yet the individual infected with this virus did not develop any of the severe symptoms frequently associated with HPAI H5 virus infections in mammals. There are several possible explanations for this. Heightened public awareness and testing may have resulted in the detection of mild cases that went unnoticed in the past. In fact, asymptomatic or unreported human HPAI H5 infections have been identified, albeit at low numbers^36–39. All thirteen farm workers infected with dairy cow HPAI H5 virus developed conjunctivitis, perhaps suggesting ocular exposure. In the present study, we found inefficient replication of HPAI H5 virus in human corneal cells, although we emphasize that additional studies are needed (e.g., testing replicative ability of other human isolates of the cow HPAI H5 lineage and/or in different ocular cell types such as conjunctival cells or cells from additional donors) to fully clarify the ability of clade 2.3.4.4b HPAI H5N1 viruses to infect and grow in ocular cells. Previous studies have shown that ocular inoculation of ferrets with influenza viruses delays the onset of disease, reduces clinical signs, and results in less efficient transmission when compared with respiratory inoculation^40,41, indicating that the route of infection affects the severity of infection and disease progression. Although speculative, ocular infection with a low dose of virus may result in localized conjunctivitis without severe disease in humans. In addition, but also speculative, multiple exposures of people to seasonal human influenza viruses may provide low levels of protection against currently circulating HPAI H5N1 viruses.

In our recent study⁷ and confirmed here, we detected binding of HPAI H5N1 viruses of the bovine lineage (i.e., NM93-H5N1 and huTX37-H5N1) to α2,6-linked sialylglycopolymers (i.e., human-type receptors) in solid-phase binding assays. Yet, several other studies have reported little or no α2,6 binding of recombinant HA proteins from huTX37-H5N1¹⁸ or another dairy cattle-derived HPAI H5N1 virus²⁰, or of (live or inactivated) dairy cattle HPAI H5N1 virions¹⁹ in various assay formats. Technical differences in the assays may account for these different results. It is important to note that substantial evidence supports the idea that recognition of α2,6-linked sialic acids is a key determinant of respiratory droplet transmissibility of influenza viruses in the ferret model. We and others have shown that a shift toward α2,6 receptor binding specificity is a prerequisite for respiratory droplet transmission of H5N1 viruses in ferrets^16,42, and mutations promoting α2,6 receptor binding specificity are imperative for respiratory droplet transmission of the 1918 H1N1 influenza virus in ferrets⁴³. Our data (presented herein and elsewhere⁷) and data reported by the US CDC¹³ indicate that both NM93-H5N1 and huTX37-H5N1 are capable of respiratory droplet transmission in ferrets, which aligns with our finding that these viruses bind to α2,6-linked sialic acids.

Multiple antiviral drugs are available for treating humans infected with influenza virus, including NA inhibitors (i.e., oseltamivir phosphate, zanamivir, laninamivir, and peramivir) and polymerase inhibitors (i.e., favipiravir and baloxavir marboxil). Our data in mice suggest that NA inhibitors may not have good efficacy against dairy cattle HPAI H5N1 viruses, including those without established markers of antiviral resistance. However, polymerase inhibitors (favipiravir and baloxavir marboxil) were highly effective in the mouse model against both a wild-type dairy cattle HPAI H5N1 virus and a mutant carrying a mutation known to reduce NA inhibitor sensitivity. The doses of oseltamivir phosphate, favipiravir, and baloxavir marboxil used in our mouse studies were several-fold higher than those used in humans, according to the pharmacokinetics in these animals^44–46. We used these doses based on previous reports on the efficacy of influenza drugs against HPAI viruses^23,47–49. Therefore, favipiravir and baloxavir marboxil may be potential treatment options for humans with dairy cattle HPAI H5N1 infections.

In summary, our data suggest that the huTX37-H5N1 isolate may be able to bind to and replicate in cells of the upper respiratory tract in humans, that it is extremely pathogenic in animals, and that it transmits by the respiratory route in ferrets. Based in these observations, every effort should be made to contain HPAI H5N1 outbreaks in dairy cattle to limit the possibility of further human infections.

Methods

Ethics Statement.

All animal experiments and procedures were approved by the Institutional Care and Use Committees of the University of Wisconsin-Madison School of Veterinary Medicine (protocol # V006426-A04) or the University of Tokyo (protocol # 19–72). Animals were acclimated to the ambient conditions of the facilities (25–28 °C and 35%–45% humidity) prior to the start of experiments, allowed access to food and water ad libitum, kept on a 12 h on/off light cycle, and given enrichment. Humane endpoints for euthanasia included ≥ 35% body weight loss or inability to remain upright (University of Wisconsin—Madison) or ≥ 25% body weight loss with no possibility of recovery (University of Tokyo).

Biosafety.

Experiments with HPAI H5N1 viruses were carried out in Biosafety Level 3 (BSL-3) containment laboratories at the Influenza Research Institute at the University of Wisconsin-Madison, which is approved by the Federal Select Agent Program for studies with these viruses, or at the University of Tokyo. Specifically, all experiments other than antiviral susceptibility tests were performed at the University of Wisconsin—Madison (ferret experiments were performed under BSL-3-Ag containment), and antiviral susceptibility testing experiments, including the generation of the TX001-H5N1 NA-T438I mutant, were performed at the University of Tokyo. Funding for this study came in part from the NIAID Centers of Excellence for Influenza Research and Response (CEIRR, Contract Number 75N93021C00014). All experiments were approved by Institutional Biosafety Committees (IBCs) of the University of Wisconsin—Madison or the University of Tokyo and all animal experiments were approved by the University of Wisconsin—Madison or the University of Tokyo Institutional Animal Care and Use Committees. The NIAID grant for the studies conducted was reviewed by the University of Wisconsin-Madison Dual Use Research of Concern (DURC) Subcommittee in accordance with the United States Government September 2014 DURC Policy and determined to not meet the criteria of DURC. The University of Wisconsin-Madison Institutional Contact for Dual Use Research reviewed this manuscript and confirmed that the studies described herein do not meet the criteria of DURC.

Cells.

MDCK (Madin-Darby canine kidney) and 293T (human embryonic kidney) cell lines (obtained from the ATCC) were grown at 37 °C and 5% CO₂ in Eagle’s minimum essential medium (MEM) containing 5% newborn calf serum or Dulbecco’s modified Eagle medium (DMEM) containing 10% fetal bovine serum, respectively. DF-1 (chicken fibroblast) cells were grown at 39 °C and 5% CO₂ in DMEM containing 10% fetal bovine serum. No authentication was performed for any of the cell lines used, and all were monitored regularly for mycoplasma contamination. Mature primary human epithelial cells (alveolar and corneal) grown in air-liquid interface (ALI) culture were purchased from MatTek Life Sciences and cultured according to the manufacturer’s recommendations. Basolateral medium was changed every day. ALI cultures were used within 48 h of receipt and were not authenticated or tested for mycoplasma contamination. All alveolar or ocular cells used for all infections were derived from the same donor.

Viruses.

A/Texas/37/2024 (H5N1; huTX37-H5N1), which was isolated from the eye of a human with conjunctivitis⁵ was provided by the US CDC. The virus stock provided by the CDC was fully sequenced, amplified in MDCK cells, and re-sequenced. A/dairy cattle/New Mexico/A240920343–93/2024 (H5N1; NM93-H5N1) was isolated from the milk of a lactating dairy cow provided by the Texas A&M Veterinary Medical Diagnostic Laboratory and was described previously^7,8. A/dairy cattle/Texas/24–008749-001/2024 (H5N1; TX001-H5N1) was generated by reverse genetics, as described below. All three bovine H5N1 viruses are in the same clade as other publicly available cow H5N1 virus sequences⁸; differences in their consensus sequences are provided in Extended Data Table 1. Viruses used as controls included a highly pathogenic H5N1 avian influenza virus (A/Vietnam/1203/2004; VN1203-H5N1)¹⁴ and a human H1N1 influenza virus (A/Isumi/UT-KK001–01/2018; Isumi-H1N1)¹⁵.

Plasmids and reverse genetics.

To generate TX001-H5N1 virus, cDNAs encoding viral RNAs, based on published sequences (EPI_ISL_19014384), were synthesized by using a commercial vendor (Integrated DNA Technologies) and cloned into RNA polymerase I-based plasmids, followed by influenza virus generation as previously described⁵⁰. T438I was introduced into the pPolI plasmid encoding the NA genome segment by PCR amplification with primers with the desired mutation. The reverse genetics-generated TX001-H5N1 virus was sequenced to confirm the absence of unwanted mutations. To prepare plasmids for viral protein expression, the open reading frames of the PB2, PB1, PA and NP genes from TX001-H5N1 were amplified by PCR by using gene-specific primers. The PCR products were cloned into the pCAGGS vector. Mutations in the PB2 gene were generated by PCR amplification of the respective plasmid with primers possessing the desired mutations (primer sequences are available upon request). All constructs were sequenced to confirm the absence of unwanted mutations.

ALI infection.

Prior to virus inoculation, EpiCorneal and EpiAlveolar cells were washed three times with 400 μl of TEER buffer (PBS with Ca²⁺ and Mg²⁺). Tissue inserts were inoculated at multiplicities of infection of 0.1 or 0.01 PFU per cell of the indicated viruses in 100 μl of TEER buffer for 1 h at 33 °C or 37 °C. Inoculum was removed and tissue inserts were washed twice with TEER buffer. The inserts were incubated in ALI culture at 33 °C or 37 °C. Samples were collected at 24, 48, 72 and 96 h post-infection from the apical surface. The basolateral medium was changed daily. Apical sample collection was performed by adding 500 μl of TEER buffer to the apical surface, followed by incubation for 30 min at 33 °C or 37 °C and removal of the medium from the apical surface. Titres of viruses released into the TEER buffer were determined by performing plaque assays in MDCK cells.

Mouse lethal dose 50.

Seven-week-old female BALB/cJ mice (Jackson Laboratory) were anaesthetised by injection of ketamine and dexmedetomidine (45–75 mg/kg ketamine + 0.25–1 mg/kg dexmedetomidine) i.p. and intranasally infected with serial dilutions (10⁰, 10¹, 10^2, 10^3, 10⁴, 10⁵, or 10⁶ PFU) of huTX37-H5N1 in 50 μl of phosphate-buffered saline (PBS) (5 mice per dose). Dexmedetomidine was reversed by i.p. injection with atipamezole (0.1–1 mg/kg). Mice were monitored daily for body weight changes and survival and were euthanised if they lost ≥35% of their initial body weight.

Tissue tropism in mice.

Seven-week-old female BALB/cJ mice were anaesthetised and intranasally infected with 10³ PFU of huTX37-H5N1 as just described. Groups of 5 mice were euthanised at days 3 and 5 post-infection, and tissues were collected in the following order: whole blood (via cardiac puncture), eye, teat, mammary gland, hamstring, latissimus dorsi, brain, colon, liver, spleen, kidney, heart, nasal turbinate, trachea, and lung. Tissue dissection instruments were disinfected between each tissue to prevent cross-contamination. Whole blood was collected without anticoagulant, immediately frozen on dry ice after collection, and later thawed and used directly for plaque assays in MDCK cells. Other tissues were directly frozen at –80 °C without buffer; later, they were thawed, mixed with 1 ml of MEM medium containing 0.3% bovine serum albumin (BSA), homogenised with a TissueLyser II (Qiagen) (30-Hz oscillation frequency for 3 min), clarified by centrifugation (14,000 rpm for 10 minutes), and the resultant supernatants were used for plaque assays in MDCK cells.

Tissue tropism in ferrets.

Six- to eight-month-old female ferrets (Triple F Farms) were confirmed to be serologically negative for the following four influenza viruses prior to use for infection experiments: A/Hong Kong/4/2022 (H3N2), A/Wisconsin/588/2019 (H1N1), B/Washington/02/2019, and A/Astrakhan/3212/2020 (H5N8). Ferrets were anaesthetised by injection with ketamine and dexmedetomidine (4–5 mg/kg and 10–40 μg/kg of body weight, respectively) i.m. and infected intranasally with 10⁶ PFU of huTX37-H5N1 in 500 μl of PBS (N=8 ferrets). Body weights, temperatures, and other clinical symptoms were monitored daily and animals were euthanised upon reaching humane endpoints (≥ 35% body weight loss or inability to remain upright); some animals died between health checks, which were performed at least twice per day, separated by about 10–14 h. Tissues were collected in the following order to prevent cross-contamination of virus from respiratory organs: eye, teat, mammary gland, hamstring, latissimus dorsi, brain, whole blood (collected from the heart), colon, liver, spleen, kidney, heart, nasal turbinate, trachea, and lung. Blood was collected and processed as described for mice (see above), and other tissues were directly frozen at –80 °C without buffer. After thawing, blood was used directly for plaque assays in MDCK cells. Other thawed tissues were homogenised in 1 ml of MEM medium containing 0.3% BSA with a multi-bead homogeniser (Yasui Kikai Corporation, Japan; six cycles of ON for 6 seconds and OFF for 4 seconds), centrifuged at 14,000 rpm for 10 min, and then used for plaque assays in MDCK cells.

Respiratory droplet transmission.

Serologically screened female ferrets (6–8 months old) were infected intranasally with 10⁶, 10³, 10², or 10¹ PFU of huTX37-H5N1 as just described (6 ferrets per dose). One day post-infection, naïve ferrets (1 per infected animal) were placed in adjacent cages in an isolator rack; infected and naïve ferrets were separated by about 5 cm to prevent direct contact. Transmission was assessed under controlled temperature (20–25 °C) and relative humidity (38.4% ± 8.8%), with airflow from the front to the back of the isolator rack, perpendicular to the direction of respiratory droplet transmission between animals. Nasal swab samples were collected under anaesthesia starting from day 1 post-infection or post-exposure and every other day thereafter. Swabs were soaked in PBS prior to insertion into the nasal cavity, and vortexed with 1.0 ml of MEM containing 50 U/ml penicillin and 50 μg/ml streptomycin for 1 minute. Swab samples were stored at 4 °C prior to plaque assays in MDCK cells. At 21 days post-infection or post-exposure, blood was collected under anaesthesia from surviving ferrets, placed in separator tubes, centrifuged at 2,000 × g for 10 minutes, and serum was frozen at −80 °C.

Plaque assays.

Plaque assays were carried out according to standard methods. Prior to inoculation with serial dilutions of virus, MDCK cells were washed with 1X MEM containing 0.3% BSA. Inoculated cells were incubated at 37 °C for 1 h, then washed with 1X PBS and overlaid with 1X MEM containing 0.3% BSA, 1% low melting point agarose, and 0.6 μg/ml L-(tosylamido-2-phenyl) ethyl chloromethyl ketone (TPCK)-treated trypsin. Overlaid cells were incubated for 2–3 days at 37 °C and 5% CO₂ and then fixed with 10% formalin. After the overlay was removed, the plates were rinsed with water and air-dried, and then virus plaques were counted under fluorescent light.

Virus growth in embryonated chicken eggs.

Nasal swab samples were inoculated into 10-day-old embryonated chicken eggs as previously described⁵¹. After two days, eggs were placed at 4 °C overnight, and then allantoic fluids were collected and subjected to haemagglutination assays according to standard methods⁵¹.

Haemagglutination inhibition assay.

Ferret sera were treated with receptor destroying enzyme (Denka Seiken Co., Ltd., Tokyo, Japan) at 37 °C for 18–20 h, heat inactivated (56 °C for 50 minutes), and then adsorbed with turkey red blood cells (room temperature for 1 h with gentle shaking). Two-fold serial dilutions of treated sera were added to 96-well V-bottom plates, mixed with 4 hemagglutination (HA) units of huTX37-H5N1, and incubated for 30 minutes at room temperature. Then, turkey red blood cells (0.5%) were added, and the plate was further incubated at room temperature for 1 hr. The hemagglutination inhibition (HI) titre was read as the reciprocal of the last dilution of serum that completely prevented hemagglutination.

Deep sequencing.

Total RNA was extracted from aliquots of inoculum used to infect ferrets or from ferret nasal swab samples with the QIAamp Viral RNA Mini kit (Qiagen) or the MagMAX-96 viral RNA isolation kit (Invitrogen) according to the respective manufacturer’s instructions. Single-reaction amplification of viral genomes was performed by using pooled primers (Supplementary Table 3) as previously described⁸. Amplicon libraries were fragmented using an Illumina DNA Prep Kit (Illumina) and indexed using Illumina DNA/RNA UD Indexes (Illumina), according to manufacturer’s instructions. DNA libraries were sequenced on an Illumina MiSeq system using MiSeq Reagent Kit v3 (600-cycle) cartridges (Illumina). Initial trimming and filtering of reads was performed using Local Run Manager Version 3.0.0 (Illumina). To generate consensus, Illumina MiSeq demultiplexed reads were processed and assembled de novo by using the Iterative Refinement Meta-assembler (IRMA v. 1.0.2)⁵². The consensus sequences of all influenza viral genes were obtained from the secondary assembly. Default IRMA FLU parameters were used, with the following exceptions: LABEL was used for read sorting, the residual assembly factor was set to 400 for the secondary assembly, and reference elongation was prevented. All sequences were then analysed by using version v1.0.1 of a snakemake workflow⁵³. iVar (1.4.2)⁵⁴ was used to call variants using the inoculum consensus sequences generated by IRMA as a reference. Reads were trimmed using Trimmomatic⁵⁵ within the snakemake workflow specified above. Trimmed reads were aligned to the reference sequence using bwa (version 0.7.17-r1188)⁵⁶. Samtools (version 1.19.2)⁵⁷ was used to generate sam and bam files, and variants were called using iVar (1.4.2) specifying a quality threshold of 20 and those occurring in at least 3% of reads were reported.

Solid-phase binding assay.

Influenza binding to sialylglycopolymers was determined as previously described⁷. Briefly, serial dilutions (2.5, 0.625, 0.156, 0.039, 0.01, 0.002, and 0.001 μg/ml) of the sodium salts of sialylglycopolymers (Yamasa Corporation Co. Ltd)—Neu5Acα2,3Galβ1,4GlcNAcβ1-poly-Glu (α2,3SA) and Neu5Acα2,6Galβ1,4GlcNAcβ1-poly-Glu (α2,6SA)—in PBS were adhered to microtitre plates (Nunc; 4 °C overnight), non-specific binding was blocked with 4% BSA in PBS (1 h at room temperature), and then 32 hemagglutination (HA) units of virus were added (4°C overnight). Virus binding was detected by a broadly reactive human monoclonal CR9114 antibody (HumImmu; 1:1000 dilution, catalog no. A90001; 1 h at 4 °C), horseradish peroxidase (HRP)-conjugated anti-human IgG (Abcam; 1:4000 dilution, catalog no. ab6858; 1 h at 4 °C), and o-phenylenediamine (Sigma) in PBS containing 0.03% H₂O₂ (10 minutes at room temperature). All washes (between binding steps) were carried out with cold PBS. Absorbance was measured at 450 nm using an optical plate reader (BioTek). All data in each replicate experiment represent a single technical replicate per condition.

Minireplicon assays.

A minigenome assay based on the dual-luciferase system was performed as previously reported^58,59. Briefly, 293T and DF-1 cells, incubated at 37 and 39 °C, respectively, were transfected with viral protein expression plasmids for PB2, PB1, PA, and NP or their mutants (0.5 μg of each), with a plasmid expressing a reporter vRNA encoding the firefly luciferase gene under the control of the human or chicken RNA polymerase I promoter (pPolI-Fluc or pPolIGG-Fluc, respectively; 0.05 μg of each), and pCAGGS-Rluc (0.05 μg), expressing Renilla luciferase as a transfection control. Luciferase activities in the transfected 293T and DF-1 cells were measured using a Dual-Luciferase Reporter Assay System (Promega) 24 h after transfection. Polymerase activity was calculated by standardization of the firefly luciferase activity to the Renilla luciferase activity. Wild-type polymerase activity was set to 100%.

Antiviral susceptibility testing.

Virus dilutions (equivalent to 800 to 1200 fluorescence units) were mixed with oseltamivir carboxylate, zanamivir, laninamivir, or peramivir (0.01 nM-1 mM) in 33 mM 2-[N-morpholino]ethanesulfonic acid (pH 6.0) with 4 mM CaCl₂ and incubated for 30 min at 37 °C. Then, 2’-(4-Methylumbelliferyl)-α-D-N-acetylneuraminic acid sodium salt hydrate (MUNANA, Sigma; final concentration of 0.1 mM) was added and reactions were incubated for 1 h at 37 °C, then stopped by addition of sodium hydroxide in 80% ethanol. Fluorescence was measured at an excitation wavelength of 360 nm and an emission wavelength of 465 nm. Interpretations of antiviral susceptibility to NA inhibitors were made based on published criteria⁶⁰.

To determine polymerase inhibitor susceptibility, MDCK cells in 6-well plates were infected with dilutions of influenza viruses sufficient to result in approximately 50 plaques per well. Infections were carried out for 1 h at 37 °C, and then the inoculum was removed, and cells were overlaid with MEM containing 0.3% bovine serum albumin, 1% low melting point agarose, 1 μg/ml TPCK-treated trypsin, and different concentrations of baloxavir acid or favipiravir. After 48 h at 37 °C, the cells were fixed with 10% formalin, overlays were removed, plates were dried, and plaques were counted under fluorescent light.

For antiviral susceptibility testing in mice, seven-week-old female BALB/c mice (Japan SLC, Inc.; N=15 per group) were anaesthetised under isoflurane and intranasally inoculated with 10 PFU of the indicated viruses. At 1 h post-inoculation, mice were treated with antiviral compounds or vehicle controls as described previously²³. In one experiment, mice were treated with zanamivir (8 mg per kg per 20 μl) or saline (20 μl) administered intranasally once daily for 5 days. In another experiment, seven-week-old female BALB/c mice (N=15 per group) were treated with oseltamivir phosphate (40 mg per kg per 200 μl), favipiravir (150 mg per kg per 200 μl), baloxavir marboxil (50 mg per kg per 200 μl), or methylcellulose (200 μl) administered orally twice daily for 5 days. For some of the mice (N=5 per treatment), body weights and survival were monitored daily for 14 days. For the remaining mice, tissues were collected on days 3 and 5 post-infection for virus titration (5 per time point).

Statistics and reproducibility.

All animals were randomly allocated to experimental groups. No blinding was performed in any experiment. Sample sizes were based on our previous work. All statistical analyses (including basic summary statistics, comparisons, and IC₅₀ calculations) were performed with GraphPad Prism software, version 9.5.1. All graphs were generated in GraphPad Prism. Survival curves were compared by using the log-rank Mantel-Cox test. In some experiments, virus titres or relative luciferase values were log₁₀-transformed and compared by using non-parametric, two-tailed Mann-Whitney tests with or without the Holm-Sidak adjustment for multiple comparisons, or a Kruskal–Wallis test with the Dunn’s multiple comparisons procedure. Experiments with primary cells in ALI culture were performed once with three biological replicates (huTX37-H5N1) or two biological replicates (NM93-H5N1 and VN1203-H5N1). Individual ALI cultures were randomly assigned to experimental groups. All experiments with animals were performed once. Three independent minireplicon experiments were performed (two with three biological replicates and one with two biological replicates) and the data from all three were combined for analysis and presentation. Three independent receptor binding experiments were performed, each with a single biological replicate for all conditions, and all three experiments are presented. In vitro antiviral susceptibility experiments were performed once.

Extended Data

Extended Data Figure 1. Clinical data and virus titres associated with ferrets used to assess respiratory droplet transmission.

For infected ferrets (donors) shown in Fig. 4 (N=6 biologically independent animals per dose), daily body weights (A), body temperatures (B), and survival (C) are given. In panels (A) and (B), data points represent mean values and error bars represent the standard deviation. Survival curves in panel (C) were compared by using a log-rank Mantel-Cox test and the p-value is reported in the figure panel. Panel (D) shows virus titres in tissues of the same animals collected at the time of euthanasia or within 14 h of death. Each dot represents the titre of an individual ferret. °C, degrees Celsius; d, days; PFU, plaque forming units; PFU/g, plaque forming units per gram of tissue; PFU/ml, plaque forming units per millilitre.

Extended Data Figure 2. huTX37-H5N1 receptor binding activity.

Four-fold serial dilutions of α2,3 and α2,6 sialylglycopolymers adhered to microtitre plates were incubated with 32 hemagglutination (HA) units of the indicated viruses or PBS (negative control). After washing, virus binding was detected by an anti-HA human monoclonal antibody (CR9114) and an HRP-conjugated secondary antibody. The absorbance values for each condition with each virus or PBS are shown. Each dot represents a single biologically independent replicate value. Three independent replicate experiments are shown.

Extended Data Table 1.

Amino acid differences among the cow HPAI H5N1 viruses used in this study.

	Amino Acid Difference
	HA*	PB2					PB1					PA								NA		NS1
	336 (348)	102	362	441	627	631	11	384	392	586	722	13	128	142	219	336	348	497	613	71	217	7	40	77	229
TX001-H5N1	S	N	G	N	E	L	K	S	I	K	A	I	H	K	I	L	I	R	E	S	K	L	Q	L	E
NM93-H5N1	N	N	G	N	E	L	Q	P	I	K	A	V	H	K	I	L	I	K	K	N	K	L	Q	R	K
huTX37-H5N1	S	N	E	N	K	M	K	S	V	K	A	I	H	E	L	L	I	K	E	N	K	L	R	L	E
cgWY001- H5N1	S	S	E	D	E	M	K	S	I	N	G	I	N	K	L	M	F	K	E	N	R	S	Q	L	E

^*number in parentheses indicates amino acid positions of the full-length HA protein

Extended Data Table 2.

Comparison of tissue titres in huTX37-H5N1- and NM93-H5N1-infected BALB/cJ mice. The data shown for NM93-H5N1-infected mice were previously reported⁷.

	huTX37-H5N1		NM93-H5N1		huTX37-H5N1		NM93-H5N1
	Day 3		Day 3		Day 5		Day 6
Tissue	Median (log10)	# Positive	Median (log10)	# Positive	Median (log10)	# Positive	Median (log10)	# Positive
Blood	5.34	5/5	1.85	4/5	6.41	5/5	3.19	5/5
Eye	4.10	5/5	0.00	2/10	3.54	5/5	0.00	4/10
Teat	4.00	5/5	0.00	0/10	4.00	5/5	3.28	5/10
Mammary gland	5.19	5/5	2.82	5/10	5.38	5/5	5.31	10/10
Brain	4.20	5/5	0.00	2/10	3.70	5/5	5.93	9/10
Colon	4.34	5/5	0.00	2/10	0.00	5/5	0.00	3/10
Liver	4.62	5/5	0.00	2/10	4.12	5/5	2.34	7/10
Spleen	6.20	5/5	3.43	8/10	5.75	5/5	2.54	6/10
Kidney	4.70	5/5	1.82	6/10	4.06	5/5	5.00	9/10
Heart	4.55	5/5	0.00	3/10	5.12	5/5	5.71	10/10
Hamstring	4.52	5/5	0.00	0/5	3.40	5/5	2.57	3/5
Latissimus dorsi	4.36	5/5	0.00	1/5	4.41	5/5	0.00	2/5
Nasal turbinate	6.00	5/5	3.15	8/10	6.54	5/5	5.36	10/10
Trachea	7.97	5/5	6.79	10/10	7.37	5/5	5.87	10/10
Lung	8.63	5/5	6.55	10/10	8.16	5/5	6.88	10/10

Extended Data Table 3.

Comparison of tissue titres in huTX37-H5N1- and NM93-H5N1-infected ferrets. The data shown for NM93-H5N1-infected ferrets were previously reported⁷. n.d., not determined.

	huTX37-H5N1		NM93-H5N1		huTX37-H5N1		NM93-H5N1
	Day 3		Day 3		Day 4**		Day 6
Tissue	Median (log10)	# Positive	Median (log10)	# Positive	Median (log10)	# Positive	Median (log10)	# Positive
Blood	6.85	3/4*	n.d.	n.d.	n.d.	n.d.	0.00	0/4
Eye	5.25	4/4	0.00	1/4	5.89	3/3	0.00	3/8
Teat	4.64	4/4	0.00	0/4	5.70	3/3	0.00	0/8
Mammary gland	4.43	4/4	0.00	0/4	4.82	3/3	0.00	1/8
Brain	5.08	4/4	0.00	0/4	5.39	3/3	2.36	4/8
Colon	6.02	4/4	2.77	3/4	6.41	3/3	1.03	4/8
Liver	7.74	4/4	0.00	1/4	8.19	3/3	0.00	2/8
Spleen	6.57	4/4	3.09	2/4	7.45	3/3	0.00	1/8
Kidney	6.61	4/4	0.00	0/4	7.26	3/3	0.00	0/8
Heart	6.05	4/4	0.71	2/4	6.55	3/3	0.00	0/8
Hamstring	4.62	4/4	n.d.	n.d.	5.38	3/3	0.00	0/4
Latissimus dorsi	4.35	4/4	n.d.	n.d.	5.49	3/3	0.00	0/4
Nasal turbinate	8.18	4/4	6.83	3/4	8.31	3/3	6.11	8/8
Trachea	7.04	4/4	4.56	4/4	7.08	3/3	4.48	8/8
Lung	7.35	4/4	4.72	4/4	7.41	3/3	4.52	7/8

^*No blood was collected from 1 of 4 ferrets infected with huTX37-H5N1 on day 3 post-infection.

^**Titres are shown for 3 ferrets that died on day 4 post-infection. No data is shown for the ferret that died on day 5 post-infection.

Extended Data Table 4. In vitro inhibitory activity of neuraminidase inhibitors.

Inhibitory concentration 50 (IC₅₀) values of neuraminidase inhibitors are shown, with standard deviations, for wild-type TX001-H5N1, a TX001-H5N1 mutant possessing a T438I substitution in the NA protein and wild-type VN1203-H5N1, which has been shown to reduce inhibition of HPAI H5N1 viruses by zanamivir and peramivir in vitro and has been identified in an HPAI H5N1 isolated from a dairy cow.

	IC50 (nM)
	Oseltamivir carboxylate	Zanamivir	Laninamivir	Peramivir
TX001-H5N1 wild-type *	23.0 ± 3.0	1.15 ± 0.05	0.65 ± 0.05	0.195 ± 0.015
TX001-H5N1 NA T438I	63.5 ± 1.5	73.6 ± 10.8	6.5 ± 2.5	3.3 ± 0.7
VN1203-H5N1	0.45 ± 0.07	1.2 ± 0.1	0.41 ± 0.03	0.24 ± 0.03

^*Each drug was mixed with influenza viruses diluted to give equivalent NA activity (800–1200 fluorescence units of 4-methylumbelliferone).

Extended Data Table 5. In vitro inhibitory activity of polymerase inhibitors.

Inhibitory concentration 50 (IC₅₀) values of polymerase inhibitors are shown, with standard deviations, for wild-type TX001-H5N1, the TX001-H5N1 NA T438I mutant and wild-type VN1203-H5N1.

	IC50
	Favipiravir (μg/ml)	Baloxavir acid (nM)
TX001-H5N1 wild-type *	3.2 ± 0.6	0.61 ± 0.12
TX001-H5N1 NA T438I	2.4 ± 0.7	1.38 ± 0.42
VN1203-H5N1	1.65 ± 0.15	1.25 ± 0.75

^*For each drug, MDCK cells in 6-well plates were infected with dilutions of influenza viruses sufficient to result in approximately 50 plaques per well and virus titres were quantified by using plaque assays.

Supplementary Material

Additional Information

Supplementary Information is available for this paper.

Data availability.

All source data underlying animal, receptor binding, and mini-replicon experiments described herein are available in the online version of the paper. Deep sequencing data have been deposited to the Sequence Read Archive (SRA) under bioproject accession number PRJNA1163435.