Browse the corpus

H5N1 influenza virus stability and transmission risk in raw milk and cheese. Highly pathogenic avian influenza H5N1 viruses have recently spread to dairy cattle, with high levels of virus detected in milk from affected animals, raising concern about the risk posed by unpasteurized dairy products consumed by humans. Here we evaluated H5N1 virus persistence in raw-milk cheeses (n = 3 per condition) made with milk acidified to pH 6.6, 5.8 and 5.0 before cheese making and validated our findings in raw-milk cheeses (n = 4) inadvertently produced with naturally contaminated raw milk. The pH values tested (6.6, 5.8 and 5.0) reflect the pH range encountered in raw-milk cheeses at the marketplace. We observed pH-dependent virus survival, with infectious virus persisting through the cheese-making process and up to 120 days of aging in cheeses made with raw milk at pH levels of 6.6 and 5.8, whereas at pH 5.0, the virus did not survive the cheese-making process. Notably, while ferrets (Mustela furo) fed H5N1 virus-contaminated raw milk (n = 4) became infected, those fed raw-milk cheese (n = 4) or cheese suspension (n = 4) did not. These results demonstrate that the H5N1 virus can remain infectious for extended periods in raw-milk cheeses under specific conditions, underscoring the potential public health risks associated with consuming raw-milk cheese produced from contaminated milk and highlighting the need for additional mitigation measures in cheese production to prevent human exposure to the virus.

The spillover of highly pathogenic avian influenza (HPAI) H5N1 virus clade 2.3.4.4b—specifically genotype B3.13 and, more recently, genotype D1.1—into dairy cows1,2, along with its continued circulation in dairy cattle in the USA3, poses major animal and public health concerns. The tropism of the virus for the mammary gland milk-secreting epithelial cells, leading to severe viral mastitis1,4,5 and shedding of high levels of infectious virus in milk (up to 8.8 log10 TCID50 ml−1, where TCID50 is the 50% tissue culture infectious dose), poses a risk of exposure to other animals (including dairy cows, pets, birds and wildlife) and potentially humans1,6–9. Although recent studies have shown that pasteurization and various thermization conditions effectively inactivate HPAI H5N1 virus in milk10–13, consumption of raw milk and raw-milk cheeses, which are not subjected to thermal treatment, remains popular worldwide, including in the USA, and represents a unique exposure risk to infectious virus.

have shown that pasteurization and various thermization conditions effectively inactivate HPAI H5N1 virus in milk10–13, consumption of raw milk and raw-milk cheeses, which are not subjected to thermal treatment, remains popular worldwide, including in the USA, and represents a unique exposure risk to infectious virus. A survey conducted between 2016 and 2019 in the USA by the US Food and Drug Administration (FDA) showed that 4.4% of US adults reported consuming raw milk at least once a year, with 1.6% consuming it frequently14 and 1.6% of the population consuming raw-milk cheeses15. These findings illustrate the high risk to public health, as HPAI H5N1 virus has been shown to persist in raw milk under refrigeration (4 °C) for at least 8 weeks11,13 and many states in which the virus has been detected in dairy cattle allow commercialization of raw-milk dairy products. Notably, contact with and/or consumption of raw milk from affected cows has been recently reported as a source of human infections16. Importantly, in California, one of the states with the largest number of herds affected by HPAI H5N1 virus3, voluntary recalls of several raw-milk dairy products were issued in November and December 2024 after retail raw milk tested positive for H5N117.

cows has been recently reported as a source of human infections16. Importantly, in California, one of the states with the largest number of herds affected by HPAI H5N1 virus3, voluntary recalls of several raw-milk dairy products were issued in November and December 2024 after retail raw milk tested positive for H5N117. A variety of cheeses can be made from raw milk, including Cheddar, Colby and other semisoft and soft-ripened cheeses. Cheddar is a hard, ripened cheese whose production begins with the addition of rennet – containing the protease chymosin, which cleaves casein and curdles the milk – and mesophilic starter cultures, typically strains of Lactococcus lactis, which acidify the milk18. The curd is then cut, cooked to about 40 °C for 30 min, and the whey is drained from the curd18. The remaining curds are then cheddared, milled, salted, molded and pressed, and the resulting cheese can be ripened or aged for months to years18. Finished cheddar cheese typically has a pH ranging from 5.1 to 5.4, which is essential for its texture, microbial stability and flavor development. However, during aging, proteolysis and microbial activity may cause further pH changes18,19. The water activity (Aw) of cheddar cheese typically ranges from 0.91–0.94. This moderate Aw level inhibits the growth of most pathogenic bacteria and molds, thereby contributing to the cheese’s shelf life and ripening process20.

owever, during aging, proteolysis and microbial activity may cause further pH changes18,19. The water activity (Aw) of cheddar cheese typically ranges from 0.91–0.94. This moderate Aw level inhibits the growth of most pathogenic bacteria and molds, thereby contributing to the cheese’s shelf life and ripening process20. To enhance the safety of raw-milk cheeses, the US Code of Federal Regulation (21 CFR Part 133 – Cheeses and Related Cheese Products) requires that cheeses produced from unpasteurized (raw) milk must undergo a curing process for a minimum of 60 days at temperatures no lower than 35 °F (1.67 °C) to inactivate bacterial pathogens. However, research has not yet determined whether the HPAI H5N1 virus is inactivated through the intricate cheese-making process and the mandatory aging period of raw-milk cheeses. In this study, we investigated the survival of HPAI H5N1 virus in raw-milk cheeses using two approaches. First, we developed a mini-cheese model to examine viral stability during cheese production and aging under varying pH conditions, using raw milk spiked with H5N1 virus. In addition, we analyzed viral stability in company-made raw-milk cheeses inadvertently made with naturally contaminated milk. Finally, we evaluated the infectivity of HPAI-contaminated aged cheese in ferrets (Mustela furo) following voluntary ingestion.

To evaluate the effect of pH on HPAI H5N1 virus stability using the raw-milk mini-cheese model, we selected three pH levels (6.6, 5.8 and 5.0), which were chosen based on a survey of 273 commercial cheese samples used to establish a categorization framework for critical physicochemical parameters of cheeses21. Cheeses were made by direct acidification (by adding lactic acid) of raw milk spiked with HPAI H5N1 TX2/24 virus (clade 2.3.4.4b genotype B3.13) (Fig. 1). Milk, whey and curd samples were collected during cheese production to quantify viral RNA (real-time reverse-transcriptase PCR (rRT–PCR)) and infectious virus (titrations in embryonated chicken eggs (ECEs)). Viral RNA loads in milk before (7.0 ± 0.42–7.16 ± 0.09 log10 copy number ml−1) and after (6.6 ± 0.51–7.04 ± 0.19 log10 copy number ml−1) pH adjustment and after heat treatment at 34 °C for 1 h (6.78 ± 0.22–7.01 ± 0.3 log10 copy number ml−1) were comparable in all three pH conditions but were about 100-fold lower (4.93 ± 0.66–5.96 ± 0.22 log10 copy number ml−1) in whey (Fig. 2a). Notably, lower viral RNA loads (4.93 ± 0.66–5.85 ± 0.14 log10 copy number ml−1) were detected in whey and curd samples in the pH 5.0 cheese group compared with corresponding samples from the pH 6.6 (5.88 ± 0.28–6.38 ± 0.33 log10 copy number ml−1) and 5.8 (5.87 ± 0.19–6.23 ± 0.02 log10 copy number ml−1) groups (Fig. 2a).Fig. 1Mini-cheese model to assess the stability of H5N1 influenza in raw-milk cheese.a, An illustration depicting the cheese-making process using the mini-cheese model used here. Raw normal milk was spiked with HPAI H5N1 virus and the milk was acidified (pH 6.6, 5.8 and 5.0) by adding lactic acid (50% solution). Milk was then prewarmed at 34 °C for 1 h before adding rennet and calcium chloride for milk coagulation at 35 °C for 65 min. The milk coagulum was cut using a sterile knife, and whey was released by applying increment heat (35–43 °C) and adding sodium chloride solution. The whey was drained from the curd using a sterile cheese cloth, and 5 g portions of the curd were weighed and p laced into each well of a 12-well tissue culture plate for molding.

lk coagulum was cut using a sterile knife, and whey was released by applying increment heat (35–43 °C) and adding sodium chloride solution. The whey was drained from the curd using a sterile cheese cloth, and 5 g portions of the curd were weighed and p laced into each well of a 12-well tissue culture plate for molding. After overnight pressing the 5-g mini-cheese samples were aged at 4 °C. Panel a created with BioRender. b–g, Photographs taken during the cheese-making process showing cut coagulum (b), curd after whey separation (c), 5-g curd portions in the 12-well plate (d), molding and pressing (e) and molded (f) mini-cheese blocks (g).Fig. 2Long-term stability of HPAI H5N1 virus in raw-milk cheese.a,b, HPAI H5N1 viral RNA (a) and infectious virus (b) loads in HPAI-spiked raw milk, whey and curd during the cheese-making process in cheeses made with milk at pH 6.6, 5.8 and 5.0 as determined by rRT–PCR and virus titration in ECEs, respectively (n = 3). c,d, Changes in pH (c) and water activity (d) of raw-milk cheese during a 60-day aging period at 4 °C (n = 3). e,f, HPAI H5N1 viral RNA (e) and infectious virus (f) loads in raw-milk cheese during a 120-day aging period as determined by rRT–PCR and virus titration in ECEs, respectively (pH 6.6, n = 3; pH 5.8, n = 3 (D0–D77), n = 2 (D84–D120); pH 5.0, n = 3 (D0–D60), n = 2 (D63–D77), n = 1 (D84–D91)). Two-way ANOVA followed by Tukey’s multiple-comparisons test. Data were obtained from three independent experiments.Source data

ng period as determined by rRT–PCR and virus titration in ECEs, respectively (pH 6.6, n = 3; pH 5.8, n = 3 (D0–D77), n = 2 (D84–D120); pH 5.0, n = 3 (D0–D60), n = 2 (D63–D77), n = 1 (D84–D91)). Two-way ANOVA followed by Tukey’s multiple-comparisons test. Data were obtained from three independent experiments.Source data a, An illustration depicting the cheese-making process using the mini-cheese model used here. Raw normal milk was spiked with HPAI H5N1 virus and the milk was acidified (pH 6.6, 5.8 and 5.0) by adding lactic acid (50% solution). Milk was then prewarmed at 34 °C for 1 h before adding rennet and calcium chloride for milk coagulation at 35 °C for 65 min. The milk coagulum was cut using a sterile knife, and whey was released by applying increment heat (35–43 °C) and adding sodium chloride solution. The whey was drained from the curd using a sterile cheese cloth, and 5 g portions of the curd were weighed and p laced into each well of a 12-well tissue culture plate for molding. After overnight pressing the 5-g mini-cheese samples were aged at 4 °C. Panel a created with BioRender. b–g, Photographs taken during the cheese-making process showing cut coagulum (b), curd after whey separation (c), 5-g curd portions in the 12-well plate (d), molding and pressing (e) and molded (f) mini-cheese blocks (g).

overnight pressing the 5-g mini-cheese samples were aged at 4 °C. Panel a created with BioRender. b–g, Photographs taken during the cheese-making process showing cut coagulum (b), curd after whey separation (c), 5-g curd portions in the 12-well plate (d), molding and pressing (e) and molded (f) mini-cheese blocks (g). a,b, HPAI H5N1 viral RNA (a) and infectious virus (b) loads in HPAI-spiked raw milk, whey and curd during the cheese-making process in cheeses made with milk at pH 6.6, 5.8 and 5.0 as determined by rRT–PCR and virus titration in ECEs, respectively (n = 3). c,d, Changes in pH (c) and water activity (d) of raw-milk cheese during a 60-day aging period at 4 °C (n = 3). e,f, HPAI H5N1 viral RNA (e) and infectious virus (f) loads in raw-milk cheese during a 120-day aging period as determined by rRT–PCR and virus titration in ECEs, respectively (pH 6.6, n = 3; pH 5.8, n = 3 (D0–D77), n = 2 (D84–D120); pH 5.0, n = 3 (D0–D60), n = 2 (D63–D77), n = 1 (D84–D91)). Two-way ANOVA followed by Tukey’s multiple-comparisons test. Data were obtained from three independent experiments. Source data

a,b, HPAI H5N1 viral RNA (a) and infectious virus (b) loads in HPAI-spiked raw milk, whey and curd during the cheese-making process in cheeses made with milk at pH 6.6, 5.8 and 5.0 as determined by rRT–PCR and virus titration in ECEs, respectively (n = 3). c,d, Changes in pH (c) and water activity (d) of raw-milk cheese during a 60-day aging period at 4 °C (n = 3). e,f, HPAI H5N1 viral RNA (e) and infectious virus (f) loads in raw-milk cheese during a 120-day aging period as determined by rRT–PCR and virus titration in ECEs, respectively (pH 6.6, n = 3; pH 5.8, n = 3 (D0–D77), n = 2 (D84–D120); pH 5.0, n = 3 (D0–D60), n = 2 (D63–D77), n = 1 (D84–D91)). Two-way ANOVA followed by Tukey’s multiple-comparisons test. Data were obtained from three independent experiments. Source data Infectious virus titers in HPAI H5N1 virus-spiked raw milk ranged between 7.34 ± 0.51 and 7.68 ± 0.89 log10 50% embryo infectious dose per milliliter (EID50 ml−1) (Fig. 2b). After adjustment of the milk pH at setting, infectious virus levels remained stable in the pH 6.6 (8.02 ± 0.57 log10 EID50 ml−1) and 5.8 (7.06 ± 1.48 log10 EID50 ml−1) cheese groups, whereas a significant decrease in infectious virus titers (4.5 ± 0.0–4.92 ± 0.72 log10 EID50 ml−1; P < 0.01) was observed in the pH 5.0 cheese group on samples collected after pH adjustment and after incubation at 34 °C for 1 h (Fig. 2b). Infectious virus was also detected in whey samples (3.68 ± 0.06–5.86 ± 0.63 log10 EID50 ml−1) collected from the pH 6.6 and 5.8 cheese groups after coagulation (whey 1), salting (whey 2) and whey draining (whey 3) and in the cheese curd (Fig. 2b). Viral titers were about 15- to 30-fold lower (3.68 ± 0.06–4.93 ± 0.49 log10 EID50 ml−1) in whey 1–3 and curd samples collected from the pH 5.8 cheese group when compared with pH 6.6 (5.21 ± 0.73–5.86 ± 0.63 log10 EID50 ml−1) (Fig. 2b). Notably, no infectious virus was detected in whey and curd samples collected from pH 5.0 cheese group (Fig. 2b). These results demonstrate a pH-dependent stability of HPAI H5N1 virus, with direct acidification of contaminated raw milk to pH 5.0 before cheese making leading to rapid virus inactivation during raw-milk cheese production.

ctious virus was detected in whey and curd samples collected from pH 5.0 cheese group (Fig. 2b). These results demonstrate a pH-dependent stability of HPAI H5N1 virus, with direct acidification of contaminated raw milk to pH 5.0 before cheese making leading to rapid virus inactivation during raw-milk cheese production. Next, we evaluated the stability of HPAI H5N1 virus during cheese aging or curing at 4 °C over a 120-day period. Cheese samples were collected daily between days 1 and 7 and then on days 14, 21, 35, 42, 49, 56, 60, 63, 70, 77, 84, 91, 98, 105, 113 and 120 of aging. The cheese pH and water activity (Aw) were monitored for up to 60 days, and viral stability in the cheese matrix was quantified for up to 120 days. The mean pH of the curd in the pH 6.6 and 5.8 cheese groups was significantly lower (5.78 ± 0.18 and 5.36 ± 0.2, respectively) than the milk pH at setting (Fig. 2c). During the first week of aging, the mean pH of the 6.6 and 5.8 cheese groups remained constant, but then gradually increased (by ~1.4 units) in the subsequent weeks until day 60 of aging (Fig. 2c). By contrast, the mean pH of the pH 5.0 cheese group remained relatively stable, only slightly increasing by 0.35 units from 4.74 ± 0.15 to 5.09 ± 0.23 over the 60 days of aging in which the cheese pH was monitored (Fig. 2c). The mean Aw in the curd of the pH 6.6, 5.8 and 5.0 cheese groups was comparable (0.99 ± 0.01, 0.98 ± 0.02 and 0.97 ± 0.01, respectively) (Fig. 2d). The Aw in all three cheese groups markedly decreased in the first 14 days of aging (Fig. 2d). This trend continued, although at a slower rate, until day 60 of aging (Fig. 2d).

mean Aw in the curd of the pH 6.6, 5.8 and 5.0 cheese groups was comparable (0.99 ± 0.01, 0.98 ± 0.02 and 0.97 ± 0.01, respectively) (Fig. 2d). The Aw in all three cheese groups markedly decreased in the first 14 days of aging (Fig. 2d). This trend continued, although at a slower rate, until day 60 of aging (Fig. 2d). The HPAI H5N1 viral RNA remained stable in raw-milk cheeses in the pH 6.6 and 5.8 groups until day 56 of aging, followed by a reduction of approximately 1.3 log and 0.43 log on day 91 in the pH 6.6 and 5.8 groups, respectively (Fig. 2e). By day 120 of aging, there was an approximate 2-log reduction in viral RNA load in both the pH 6.6 and 5.8 groups. While viral RNA levels were comparable between the pH 6.6 and 5.8 cheese groups, a reduction of approximately 1 log was observed in the pH 5.0 group (Fig. 2e). Consistently high amounts of infectious virus were recovered from raw-milk cheeses in the pH 6.6 and 5.8 groups during the first 7 days of aging (5.81 ± 0.64 and 4.38 ± 0.67 log10 EID50 g−1, respectively) (Fig. 2f). Notably, virus titers decreased by only approximately 2 logs over the subsequent 60 days of aging, resulting in viral yields of 3.99 ± 1.29 and 2.58 ± 0.14 log10 EID50 g−1 in the pH 6.6 and 5.8 cheese groups, respectively (Fig. 2f). After day 60 (the minimum aging period required by the FDA for raw-milk cheeses) the infectious virus titers decreased gradually, reaching 2.88 ± 0.65 and 1.81 ± 2.56 log10 EID50 g−1 in the pH 6.6 and 5.8 cheese groups, respectively, on day 120 of aging. Of note, viral titers recovered from the pH 5.8 cheese group were 1.0 to 2.0 logs lower than those in the pH 6.6 group throughout the aging period. Importantly, no infectious virus was recovered from raw-milk cheeses at the pH 5.0 group (Fig. 2f). Virus infectivity in allantoic fluid collected from ECEs inoculated with the cheese sample homogenates was confirmed by hemagglutination (HA) assay (Extended Data Table 1) throughout the 120-day aging period and by rRT–PCR targeting the influenza A M gene on samples from days 60, 91 and 120 (Extended Data Fig. 1). The decimal reduction times, D values, of H5N1 virus in the raw-milk cheeses in the pH 6.6 and 5.8 groups were estimated to be 29.2 and 48.3 days, respectively, indicating long-term stability of the virus in raw-milk cheese under these experimental conditions.

on samples from days 60, 91 and 120 (Extended Data Fig. 1). The decimal reduction times, D values, of H5N1 virus in the raw-milk cheeses in the pH 6.6 and 5.8 groups were estimated to be 29.2 and 48.3 days, respectively, indicating long-term stability of the virus in raw-milk cheese under these experimental conditions. We also investigated the HPAI H5N1 virus stability in raw-milk cheeses from a raw-milk dairy that inadvertently made cheddar cheese with HPAI H5N1 contaminated raw milk, following an H5N1 influenza virus outbreak in the dairy cattle at the farm. We received four company-made raw-milk cheese blocks (2 pounds per block) on day 24 of aging. Upon arrival, these cheese samples were tested for the presence of HPAI H5N1 virus by rRT–PCR and virus titrations in ECEs, and the pH (5.37 ± 0.06) and Aw (0.94 ± 0.01) on each sample were recorded (Fig. 3a,b). After testing and confirmation that the cheese blocks were indeed positive for HPAI H5N1 virus genotype B3.13 (Fig. 4), we continued the aging process at 4 °C up to 120 days. Individual samples (1 g) were collected from each company-made raw-milk cheese block on days 29, 32, 35, 38, 42, 45, 49, 52, 56, 60, 64, 68, 72, 75, 79, 82, 86, 90, 95, 98, 105, 113 and 120 of aging. In addition, the pH and Aw of the cheeses were recorded until day 60. The mean pH of the company-made cheeses, which initiated at 5.37 ± 0.06 on day 24, remained relatively stable until day 60 of aging (5.34 ± 0.2) (Fig. 3a). As expected, there was a slight reduction in Aw during the aging process, which initiated at 0.94 ± 0.01 on day 24 and was determined to be 0.93 ± 0.01 on day 60 of aging (Fig. 3b). The rRT–PCR analysis revealed the presence of high viral RNA loads (5.82 ± 0.37 log10 copy number g−1) in the company-made cheese samples on day 24, which remained stable throughout the 120-day aging period (Fig. 3c). Similarly, virus titrations in ECEs showed mean viral titers of 4.21 ± 0.48 log10 EID50 g−1 of infectious virus in the company-made cheeses on day 24, with slight variations in the viral loads being observed throughout the aging period. Notably, on day 120 of aging, 3.6 ± 0.89 log10 EID50 g−1 of HPAI H5N1 virus were recovered from the company-made raw-milk cheese samples (Fig. 3d). The presence of HPAI H5N1 virus in the allantoic fluids collected from ECEs inoculated with the cheese sample homogenates was confirmed by HA and rRT–PCR (Extended Data Fig. 2).

ly, on day 120 of aging, 3.6 ± 0.89 log10 EID50 g−1 of HPAI H5N1 virus were recovered from the company-made raw-milk cheese samples (Fig. 3d). The presence of HPAI H5N1 virus in the allantoic fluids collected from ECEs inoculated with the cheese sample homogenates was confirmed by HA and rRT–PCR (Extended Data Fig. 2). These findings confirm and validate the data obtained in our laboratory-scale mini-cheese model. Most importantly, these results provide compelling evidence that HPAI H5N1 virus is stable in raw-milk cheese, surviving throughout the minimum required aging period (60 days) and at least up to 120 days of aging in raw-milk cheeses.Fig. 3HPAI H5N1 virus stability in company-made raw-milk cheeses.a,b, Changes in pH (a) and water activity (b) of company-made raw-milk cheese samples during the 60-day aging period. c, HPAI H5N1 viral RNA loads in company-made raw-milk cheese samples from day 24 (day of sample receipt) to day 120 of aging as determined by rRT–PCR (n = 4). d, HPAI H5N1 infectious virus loads in company-made raw-milk cheese samples from day 24 (day of sample receipt) to day 120 of aging as determined by virus titration in ECEs (n = 4). The data are presented as individual cheese sample readouts (a–d) and as the mean ± standard error (c and d) of four (n = 4) company-made raw-milk cheese samples.Source dataFig. 4Phylogenetic analysis of H5N1 virus sequences recovered from company-made raw-milk cheeses reveals that the virus belongs to genotype B3.13.a, Phylogenetic analysis based on concatenated whole-genome sequences of HPAI H5N1 virus sequences from genotypes A3, B3.13 and D1.1 obtained from different hosts and from company-made raw-milk cheese samples studied here. b, Phylogenetic analysis based on concatenated whole-genome sequences of HPAI H5N1 virus sequences from genotype B3.13 from the four company-made raw-milk cheese samples and other avian and mammalian hosts.

ypes A3, B3.13 and D1.1 obtained from different hosts and from company-made raw-milk cheese samples studied here. b, Phylogenetic analysis based on concatenated whole-genome sequences of HPAI H5N1 virus sequences from genotype B3.13 from the four company-made raw-milk cheese samples and other avian and mammalian hosts. a,b, Changes in pH (a) and water activity (b) of company-made raw-milk cheese samples during the 60-day aging period. c, HPAI H5N1 viral RNA loads in company-made raw-milk cheese samples from day 24 (day of sample receipt) to day 120 of aging as determined by rRT–PCR (n = 4). d, HPAI H5N1 infectious virus loads in company-made raw-milk cheese samples from day 24 (day of sample receipt) to day 120 of aging as determined by virus titration in ECEs (n = 4). The data are presented as individual cheese sample readouts (a–d) and as the mean ± standard error (c and d) of four (n = 4) company-made raw-milk cheese samples. Source data a, Phylogenetic analysis based on concatenated whole-genome sequences of HPAI H5N1 virus sequences from genotypes A3, B3.13 and D1.1 obtained from different hosts and from company-made raw-milk cheese samples studied here. b, Phylogenetic analysis based on concatenated whole-genome sequences of HPAI H5N1 virus sequences from genotype B3.13 from the four company-made raw-milk cheese samples and other avian and mammalian hosts.

ypes A3, B3.13 and D1.1 obtained from different hosts and from company-made raw-milk cheese samples studied here. b, Phylogenetic analysis based on concatenated whole-genome sequences of HPAI H5N1 virus sequences from genotype B3.13 from the four company-made raw-milk cheese samples and other avian and mammalian hosts. To evaluate the effect of acidification on HPAI H5N1 virus viability in raw milk, we spiked raw milk with a recombinant HPAI H5N1 virus expressing NanoLuc luciferase (rTX2/24-NLuc), performed acidification of the contaminated raw milk and then assessed virus infectivity in cell culture using a luciferase reporter assay. HPAI-spiked raw milk or phosphate-buffered saline (PBS) were acidified to pH 6.0, 5.5, 5.0 or 4.5 using a 50% lactic acid solution and incubated at 4 °C for 1 h. Nonacidified milk (pH 6.78) and PBS (pH 7.4) served as controls. All samples were inoculated into bovine uterine epithelial cells (Cal-1) and incubated for 4 h at 37 °C. Luminescence was measured to assess viral entry and replication in inoculated cells. Control HPAI H5N1-spiked raw milk with unadjusted pH (pH 6.78) showed a 116.6 ± 27.7-fold increase in luminescence compared with mock-inoculated cell controls. Acidification of the milk to pH 6.0 modestly reduced viral infectivity (90.75 ± 28.9-fold), while further acidification to pH 5.5, 5.0 and 4.5 significantly reduced infectivity (P ≤ 0.0001), as indicated by negligible luminescence detected in milk samples at these pH levels (1.76 ± 0.3, 1.83 ± 0.93 and 1.94 ± 0.8, fold luminescence expression over control mock-infected cells, respectively) (Fig. 5a). A similar trend was observed in PBS. Nonacidified PBS (pH 7.4) showed higher infectivity (862 ± 134.76-fold increase in luminescence) as compared with the nonacidified milk (116.6 ± 27.7-fold increase in luminescence), which decreased significantly at pH 6.0 (244.63 ± 101.34-fold; P ≤ 0.0001). As expected, virus infectivity was markedly reduced at pH 5.5 (5.67-fold; P ≤ 0.01) and nearly abolished at pH 5.0 and 4.5 (≤1-fold over control) (Fig. 5b). These results demonstrate that acidification of raw milk to pH 5.5 or lower leads to rapid loss of infectivity of HPAI H5N1 virus, probably through inhibition of viral entry into cells.Fig. 5pH-dependent infectivity inhibition of HPAI H5N1 virus in raw milk.a,b, Raw normal milk (a) or PBS (b) spiked with HPAI H5N1 rTX2/24-NLuc virus were acidified to pH 6.0, 5.5, 5.0 and 4.5 and incubated at 4 °C for 1 h.

id loss of infectivity of HPAI H5N1 virus, probably through inhibition of viral entry into cells.Fig. 5pH-dependent infectivity inhibition of HPAI H5N1 virus in raw milk.a,b, Raw normal milk (a) or PBS (b) spiked with HPAI H5N1 rTX2/24-NLuc virus were acidified to pH 6.0, 5.5, 5.0 and 4.5 and incubated at 4 °C for 1 h. Cal-1 cells were inoculated with 0.2 ml milk or PBS samples (at each pH level) and inoculated at 4 °C for 1 h (virus adsorption) and then transferred and incubated at 37 °C for 4 h. Cell lysates were collected, and luciferase activity was measured using a luminometer. The luciferase activity was expressed as fold change normalized to mock cell control. One-way ANOVA followed by Tukey’s multiple-comparisons test. Data represent observations (dots) from four replicates (n = 4) overlaid with the mean (horizontal line) and ±s.e.m. (whiskers) from four independent experiments.Source data

The luciferase activity was expressed as fold change normalized to mock cell control. One-way ANOVA followed by Tukey’s multiple-comparisons test. Data represent observations (dots) from four replicates (n = 4) overlaid with the mean (horizontal line) and ±s.e.m. (whiskers) from four independent experiments.Source data a,b, Raw normal milk (a) or PBS (b) spiked with HPAI H5N1 rTX2/24-NLuc virus were acidified to pH 6.0, 5.5, 5.0 and 4.5 and incubated at 4 °C for 1 h. Cal-1 cells were inoculated with 0.2 ml milk or PBS samples (at each pH level) and inoculated at 4 °C for 1 h (virus adsorption) and then transferred and incubated at 37 °C for 4 h. Cell lysates were collected, and luciferase activity was measured using a luminometer. The luciferase activity was expressed as fold change normalized to mock cell control. One-way ANOVA followed by Tukey’s multiple-comparisons test. Data represent observations (dots) from four replicates (n = 4) overlaid with the mean (horizontal line) and ±s.e.m. (whiskers) from four independent experiments. Source data

a,b, Raw normal milk (a) or PBS (b) spiked with HPAI H5N1 rTX2/24-NLuc virus were acidified to pH 6.0, 5.5, 5.0 and 4.5 and incubated at 4 °C for 1 h. Cal-1 cells were inoculated with 0.2 ml milk or PBS samples (at each pH level) and inoculated at 4 °C for 1 h (virus adsorption) and then transferred and incubated at 37 °C for 4 h. Cell lysates were collected, and luciferase activity was measured using a luminometer. The luciferase activity was expressed as fold change normalized to mock cell control. One-way ANOVA followed by Tukey’s multiple-comparisons test. Data represent observations (dots) from four replicates (n = 4) overlaid with the mean (horizontal line) and ±s.e.m. (whiskers) from four independent experiments. Source data Next, we assessed the potential infectivity of HPAI H5N1 virus following ingestion of contaminated raw milk and raw-milk cheeses using a ferret model of infection. For this, we fed ferrets with the company-made raw-milk cheese contaminated with HPAI H5N1 virus (5 g of solid cheese, n = 4; or 5 ml of a 10% cheese suspension in PBS, n = 4) or with HPAI-spiked raw milk (5 ml, n = 4). Controls in the study included ferrets fed with HPAI H5N1-negative store-bought cheese (5 g of solid cheese, n = 4) and ferrets directly inoculated orally (P/O, n = 2) or intranasally (I/N, n = 2) with HPAI-spiked milk (Fig. 6a). The virus titers in the HPAI-spiked milk were adjusted to 5 × 104 EID50 to match the virus titers obtained from the company-made raw-milk cheese samples at 60 days of aging. Animals in the groups that were fed HPAI H5N1 virus-contaminated cheese or milk were allowed to ingest their portions voluntarily and received repeated exposures to contaminated products on days 0, 3, 4 and 5, whereas animals in the P/O and I/N groups were inoculated only on day 0. Infectious virus titers in milk and cheese samples fed to ferrets were determined by back titration in ECEs (Extended Data Table 2).Fig. 6Infectivity of HPAI H5N1 virus following oral exposure of ferrets to contaminated raw milk and raw-milk cheese.a, Experimental design. b–d, Body temperature (b), body weight (c) and survival curve (d) of ferrets inoculated or fed with HPAI-spiked milk or contaminated raw-milk cheese. e–g, HPAI H5N1 viral RNA loads quantified by rRT–PCR in nasal (e) and oral (f) secretions and feces (g) collected on days 0, 1, 3, 5, 7, 10 and 14 post-exposure. h–j, Infectious viral loads in nasal (h) and oral (i) secretions and feces (j) determined by virus titrations. Virus titers were determined using endpoint dilutions and expressed as TCID50 ml−1. The limit of detection for infectious virus titration was 101.05 TCID50 ml−1. k, Virus neutralizing (VN) antibody titers in serum of ferrets. Data indicate mean ± s.e.m. of two animals per group per time point for H5N1-spiked milk P/O and I/N groups and four animals per groups per time point for rest of the four groups. Susp., suspension.

on for infectious virus titration was 101.05 TCID50 ml−1. k, Virus neutralizing (VN) antibody titers in serum of ferrets. Data indicate mean ± s.e.m. of two animals per group per time point for H5N1-spiked milk P/O and I/N groups and four animals per groups per time point for rest of the four groups. Susp., suspension. Panel a created with BioRender.Source data a, Experimental design. b–d, Body temperature (b), body weight (c) and survival curve (d) of ferrets inoculated or fed with HPAI-spiked milk or contaminated raw-milk cheese. e–g, HPAI H5N1 viral RNA loads quantified by rRT–PCR in nasal (e) and oral (f) secretions and feces (g) collected on days 0, 1, 3, 5, 7, 10 and 14 post-exposure. h–j, Infectious viral loads in nasal (h) and oral (i) secretions and feces (j) determined by virus titrations. Virus titers were determined using endpoint dilutions and expressed as TCID50 ml−1. The limit of detection for infectious virus titration was 101.05 TCID50 ml−1. k, Virus neutralizing (VN) antibody titers in serum of ferrets. Data indicate mean ± s.e.m. of two animals per group per time point for H5N1-spiked milk P/O and I/N groups and four animals per groups per time point for rest of the four groups. Susp., suspension. Panel a created with BioRender. Source data

a, Experimental design. b–d, Body temperature (b), body weight (c) and survival curve (d) of ferrets inoculated or fed with HPAI-spiked milk or contaminated raw-milk cheese. e–g, HPAI H5N1 viral RNA loads quantified by rRT–PCR in nasal (e) and oral (f) secretions and feces (g) collected on days 0, 1, 3, 5, 7, 10 and 14 post-exposure. h–j, Infectious viral loads in nasal (h) and oral (i) secretions and feces (j) determined by virus titrations. Virus titers were determined using endpoint dilutions and expressed as TCID50 ml−1. The limit of detection for infectious virus titration was 101.05 TCID50 ml−1. k, Virus neutralizing (VN) antibody titers in serum of ferrets. Data indicate mean ± s.e.m. of two animals per group per time point for H5N1-spiked milk P/O and I/N groups and four animals per groups per time point for rest of the four groups. Susp., suspension. Panel a created with BioRender. Source data Following exposure (direct inoculation or feeding), clinical parameters, including clinical signs, body temperature (using a subcutaneous transponder), and body weight were monitored daily. Ferrets in both P/O- and I/N-inoculated groups presented with high fever (40–40.6 °C), depression and lack of appetite on day 1 post-exposure (p.e.), which continued on days 2 and 3 p.e. (Fig. 6b). Animals from both inoculated groups lost 16.5–23.3% body weight in 4 days and were euthanized at the humane endpoint on day 4 p.e. (Fig. 6c,d). Ferrets fed with control or contaminated cheese (both solid cheese and cheese suspension) presented normal body temperature and gained weight throughout the experimental period (Fig. 6b,c). Two ferrets fed HPAI H5N1 virus-spiked raw milk developed fever (39.9–40.6 °C) on days 5 and 8 p.e. and lost body weight, reaching humane endpoints and being euthanized on days 9 and 13 p.e. (Fig. 6b–d).

eese suspension) presented normal body temperature and gained weight throughout the experimental period (Fig. 6b,c). Two ferrets fed HPAI H5N1 virus-spiked raw milk developed fever (39.9–40.6 °C) on days 5 and 8 p.e. and lost body weight, reaching humane endpoints and being euthanized on days 9 and 13 p.e. (Fig. 6b–d). To assess virus replication and shedding dynamics, nasal (NS), oropharyngeal (OPS) and rectal (RS) swabs were collected to quantify viral RNA (rRT–PCR) and infectious virus (titration in Cal-1 cells). Ferrets inoculated with HPAI H5N1 virus-spiked milk orally and intranasally presented high levels of viral RNA in NS, OPS and RS on days 1 and 3 p.e., with higher viral RNA loads detected on day 3 p.e. (Fig. 6e–g). Viral RNA loads in RS were lower than those detected in NS and OPS. Among the two ferrets that were fed with HPAI H5N1 virus-spiked milk and became clinical, one had (F#14) viral RNA in RS on days 3 and 5 p.e. and in NS and OPS on days 5 and 7 p.e., whereas the second animal (F#13) had detectable viral RNA on day 7 in NS, OPS and RS and continued shedding virus until day 10 p.e. No viral RNA was detected in any of the ferrets that were fed with control (solid) or contaminated cheese samples (solid or suspension). These results were confirmed by virus titrations, which revealed virus replication and shedding through respiratory and oral secretions, and feces only in the animals in the directly inoculated groups or in the two animals in the HPAI H5N1 virus-spiked raw-milk group (Fig. 6h–j) that became clinically affected. Importantly, no infectious virus was detected in any of the ferrets fed with control or contaminated cheese samples. None of the animals presented neutralizing antibodies to HPAI H5N1 by the end of the experiment (Fig. 6k).

We also investigated the HPAI H5N1 virus stability in raw-milk cheeses from a raw-milk dairy that inadvertently made cheddar cheese with HPAI H5N1 contaminated raw milk, following an H5N1 influenza virus outbreak in the dairy cattle at the farm. We received four company-made raw-milk cheese blocks (2 pounds per block) on day 24 of aging. Upon arrival, these cheese samples were tested for the presence of HPAI H5N1 virus by rRT–PCR and virus titrations in ECEs, and the pH (5.37 ± 0.06) and Aw (0.94 ± 0.01) on each sample were recorded (Fig. 3a,b). After testing and confirmation that the cheese blocks were indeed positive for HPAI H5N1 virus genotype B3.13 (Fig. 4), we continued the aging process at 4 °C up to 120 days. Individual samples (1 g) were collected from each company-made raw-milk cheese block on days 29, 32, 35, 38, 42, 45, 49, 52, 56, 60, 64, 68, 72, 75, 79, 82, 86, 90, 95, 98, 105, 113 and 120 of aging. In addition, the pH and Aw of the cheeses were recorded until day 60. The mean pH of the company-made cheeses, which initiated at 5.37 ± 0.06 on day 24, remained relatively stable until day 60 of aging (5.34 ± 0.2) (Fig. 3a). As expected, there was a slight reduction in Aw during the aging process, which initiated at 0.94 ± 0.01 on day 24 and was determined to be 0.93 ± 0.01 on day 60 of aging (Fig. 3b). The rRT–PCR analysis revealed the presence of high viral RNA loads (5.82 ± 0.37 log10 copy number g−1) in the company-made cheese samples on day 24, which remained stable throughout the 120-day aging period (Fig. 3c). Similarly, virus titrations in ECEs showed mean viral titers of 4.21 ± 0.48 log10 EID50 g−1 of infectious virus in the company-made cheeses on day 24, with slight variations in the viral loads being observed throughout the aging period. Notably, on day 120 of aging, 3.6 ± 0.89 log10 EID50 g−1 of HPAI H5N1 virus were recovered from the company-made raw-milk cheese samples (Fig. 3d). The presence of HPAI H5N1 virus in the allantoic fluids collected from ECEs inoculated with the cheese sample homogenates was confirmed by HA and rRT–PCR (Extended Data Fig. 2).

To evaluate the effect of acidification on HPAI H5N1 virus viability in raw milk, we spiked raw milk with a recombinant HPAI H5N1 virus expressing NanoLuc luciferase (rTX2/24-NLuc), performed acidification of the contaminated raw milk and then assessed virus infectivity in cell culture using a luciferase reporter assay. HPAI-spiked raw milk or phosphate-buffered saline (PBS) were acidified to pH 6.0, 5.5, 5.0 or 4.5 using a 50% lactic acid solution and incubated at 4 °C for 1 h. Nonacidified milk (pH 6.78) and PBS (pH 7.4) served as controls. All samples were inoculated into bovine uterine epithelial cells (Cal-1) and incubated for 4 h at 37 °C. Luminescence was measured to assess viral entry and replication in inoculated cells. Control HPAI H5N1-spiked raw milk with unadjusted pH (pH 6.78) showed a 116.6 ± 27.7-fold increase in luminescence compared with mock-inoculated cell controls. Acidification of the milk to pH 6.0 modestly reduced viral infectivity (90.75 ± 28.9-fold), while further acidification to pH 5.5, 5.0 and 4.5 significantly reduced infectivity (P ≤ 0.0001), as indicated by negligible luminescence detected in milk samples at these pH levels (1.76 ± 0.3, 1.83 ± 0.93 and 1.94 ± 0.8, fold luminescence expression over control mock-infected cells, respectively) (Fig. 5a). A similar trend was observed in PBS. Nonacidified PBS (pH 7.4) showed higher infectivity (862 ± 134.76-fold increase in luminescence) as compared with the nonacidified milk (116.6 ± 27.7-fold increase in luminescence), which decreased significantly at pH 6.0 (244.63 ± 101.34-fold; P ≤ 0.0001). As expected, virus infectivity was markedly reduced at pH 5.5 (5.67-fold; P ≤ 0.01) and nearly abolished at pH 5.0 and 4.5 (≤1-fold over control) (Fig. 5b). These results demonstrate that acidification of raw milk to pH 5.5 or lower leads to rapid loss of infectivity of HPAI H5N1 virus, probably through inhibition of viral entry into cells.Fig. 5pH-dependent infectivity inhibition of HPAI H5N1 virus in raw milk.a,b, Raw normal milk (a) or PBS (b) spiked with HPAI H5N1 rTX2/24-NLuc virus were acidified to pH 6.0, 5.5, 5.0 and 4.5 and incubated at 4 °C for 1 h.

Next, we assessed the potential infectivity of HPAI H5N1 virus following ingestion of contaminated raw milk and raw-milk cheeses using a ferret model of infection. For this, we fed ferrets with the company-made raw-milk cheese contaminated with HPAI H5N1 virus (5 g of solid cheese, n = 4; or 5 ml of a 10% cheese suspension in PBS, n = 4) or with HPAI-spiked raw milk (5 ml, n = 4). Controls in the study included ferrets fed with HPAI H5N1-negative store-bought cheese (5 g of solid cheese, n = 4) and ferrets directly inoculated orally (P/O, n = 2) or intranasally (I/N, n = 2) with HPAI-spiked milk (Fig. 6a). The virus titers in the HPAI-spiked milk were adjusted to 5 × 104 EID50 to match the virus titers obtained from the company-made raw-milk cheese samples at 60 days of aging. Animals in the groups that were fed HPAI H5N1 virus-contaminated cheese or milk were allowed to ingest their portions voluntarily and received repeated exposures to contaminated products on days 0, 3, 4 and 5, whereas animals in the P/O and I/N groups were inoculated only on day 0. Infectious virus titers in milk and cheese samples fed to ferrets were determined by back titration in ECEs (Extended Data Table 2).Fig. 6Infectivity of HPAI H5N1 virus following oral exposure of ferrets to contaminated raw milk and raw-milk cheese.a, Experimental design. b–d, Body temperature (b), body weight (c) and survival curve (d) of ferrets inoculated or fed with HPAI-spiked milk or contaminated raw-milk cheese. e–g, HPAI H5N1 viral RNA loads quantified by rRT–PCR in nasal (e) and oral (f) secretions and feces (g) collected on days 0, 1, 3, 5, 7, 10 and 14 post-exposure. h–j, Infectious viral loads in nasal (h) and oral (i) secretions and feces (j) determined by virus titrations. Virus titers were determined using endpoint dilutions and expressed as TCID50 ml−1. The limit of detection for infectious virus titration was 101.05 TCID50 ml−1. k, Virus neutralizing (VN) antibody titers in serum of ferrets. Data indicate mean ± s.e.m. of two animals per group per time point for H5N1-spiked milk P/O and I/N groups and four animals per groups per time point for rest of the four groups. Susp., suspension.

Our study demonstrates that HPAI H5N1 virus exhibits remarkable stability throughout the cheese-making process, with slow decay rates (D values 29.2–48.3 days) and infectious virus persisting for up to 120 days of aging. Notably, we found that preprocessing milk acidification to pH 5.0 (with lactic acid) effectively inactivated the virus, leading to no infectious virus detection in cheese curd immediately after cheese production and subsequently in raw-milk cheese throughout the 120-day aging period. Our recent findings indicate that thermization (subpasteurization heat treatment) of raw milk at temperatures above 54 °C successfully inactivates the virus within 15 min (ref. 13), offering an alternative safety measure for cheese production.

ion and subsequently in raw-milk cheese throughout the 120-day aging period. Our recent findings indicate that thermization (subpasteurization heat treatment) of raw milk at temperatures above 54 °C successfully inactivates the virus within 15 min (ref. 13), offering an alternative safety measure for cheese production. The mini-cheese and the company-made cheddar cheese in our study were not subjected to drastic thermal treatment during cheese making (maximum temperature of 43 °C for 30 min); thus, HPAI H5N1 virus was stable throughout the cheese-making process. An early study using laboratory cheddar cheese demonstrated the stability of poliovirus during cheddar cheese making and up to day 7 of aging, while vesicular stomatitis virus and influenza A virus (IAV) (not HPAI H5N1) were inactivated during cheese making22. The differences in the stability of influenza A could be due to the different strain or subtype of virus used, or to differences in the sensitivity of the cell culture system (primary monkey kidney cells) used by Cliver22 and ECEs used by us here. Importantly, in the same study, the authors demonstrated that heat treatment of milk (60 °C for 30 s) during cheese making led to a one-million-fold reduction in poliovirus infectivity and complete inactivation of vesicular stomatitis virus and IAV22. Given the sensitivity of HPAI H5N1 virus to thermal treatment of milk—including to subpasteurization conditions13—it is not surprising that cheese making of cheeses that involve heating (for example, 53.5 °C for 50 min for Grana- or Parmigiano Reggiano-type cheeses) has been shown to result in inactivation of the HPAI H5N1 virus23. However, our results using both the mini-cheese model and company-made cheddar cheese clearly demonstrate the stability of HPAI H5N1 virus during the production of raw-milk cheeses that are not subjected to heat at temperatures that are known to inactivate the virus13.

o result in inactivation of the HPAI H5N1 virus23. However, our results using both the mini-cheese model and company-made cheddar cheese clearly demonstrate the stability of HPAI H5N1 virus during the production of raw-milk cheeses that are not subjected to heat at temperatures that are known to inactivate the virus13. It is important to note that our study included cheeses that were made with spiked milk (mini-cheeses) and with milk from naturally infected cows (company-made cheese). Although we cannot directly compare HPAI H5N1 virus decay during the cheese-making process nor at the early stages of aging (due to the lack of samples from the company-made cheese that was only received on day 24 of aging), the stability of HPAI H5N1 virus over the 120-day aging period was similar in these two cheeses, with 3–4.2 log10 EID50 of the virus being detected in both by the end of the aging period. Similar to our data in raw clinical HPAI H5N1 virus-contaminated milk or in raw milk spiked with HPAI H5N1 virus13, the results in raw-milk cheeses here demonstrate the long-term stability of the HPAI H5N1 virus in both spiked and naturally infected raw-milk cheeses stored at refrigeration temperature (4 °C).

g period. Similar to our data in raw clinical HPAI H5N1 virus-contaminated milk or in raw milk spiked with HPAI H5N1 virus13, the results in raw-milk cheeses here demonstrate the long-term stability of the HPAI H5N1 virus in both spiked and naturally infected raw-milk cheeses stored at refrigeration temperature (4 °C). IAVs are sensitive to acidic environments similar to those found in cellular endosomes (37 °C, pH ~5.5)24. This sensitivity is primarily driven by pH-dependent conformational changes in the viral HA protein24,25. Under normal infection conditions, HA transitions to its irreversible postfusion conformation within the endosome, facilitating the crucial membrane fusion between the viral envelope and the endosomal membrane, thereby enabling viral uncoating and cellular entry24. However, if this conformational change occurs prematurely outside the cellular environment, the HA protein loses its ability to bind to host cell receptors, resulting in viral inactivation. This mechanism makes IAV particularly vulnerable to acidic conditions in the external environment. Our results from the raw-milk cheese group subjected to preprocessing milk acidification to pH 5.0 provide strong support for this mechanism. In addition, using a nano-luciferase reporter H5N1 virus (rTX2/24-NLuc) in viral entry assays, we demonstrated that acidification of milk to pH 5.5 or lower significantly reduced the luminescence signals in the rTX2/24-NLuc infected bovine uterine epithelial cells, confirming the role of pH in viral inactivation, resulting in impaired viral entry and infectivity. A recent study also demonstrated complete inactivation of HPAI H5N1 virus in yogurt prepared with spiked milk at pH 5.0 (ref. 26).

ed the luminescence signals in the rTX2/24-NLuc infected bovine uterine epithelial cells, confirming the role of pH in viral inactivation, resulting in impaired viral entry and infectivity. A recent study also demonstrated complete inactivation of HPAI H5N1 virus in yogurt prepared with spiked milk at pH 5.0 (ref. 26). The oral infectivity study in ferrets demonstrated that voluntary consumption of HPAI H5N1 virus-contaminated raw milk can result in infection, corroborating previous findings in a mouse model11. By contrast, ferrets that ingested contaminated cheese (company-made) did not become infected. These contrasting findings—showing that repeated ingestion of contaminated milk led to HPAI H5N1 virus infection in ferrets, while repeated ingestion of contaminated cheese did not result in productive infection—may be explained by several factors, including: (1) the potential higher oral infectious dose of HPAI H5N1 virus in solid versus liquid matrix, (2) the fact that ferrets are more likely to swallow small pieces of cheese whole without sustained mastication, potentially limiting viral contact and exposure to the oropharynx, (3) or yet by the small number of animals in our oral infectivity studies. These results, however, largely corroborate current epidemiological data indicating that most human HPAI H5N1 virus cases confirmed so far are linked to exposure to an infected animal6. It is, however, important to note that human infections with HPAI H5N1 virus have been linked to exposure to and/or ingestion of contaminated raw milk from affected cows16. Additionally, various domestic cats have been infected via ingestion of contaminated raw milk (fluid) or pet food (solid)1,8,27–29, underscoring the risk of oral exposure to both public and animal health. Additional studies addressing species-specific differences in sialic acid receptor expression and/or distribution in humans versus ferrets or in other susceptible animal species (for example cats, wild carnivores, cattle and so on) would be important, for example, to understand potential differences in oral susceptibility and the risk of foodborne influenza A infections.

ences in sialic acid receptor expression and/or distribution in humans versus ferrets or in other susceptible animal species (for example cats, wild carnivores, cattle and so on) would be important, for example, to understand potential differences in oral susceptibility and the risk of foodborne influenza A infections. The lack of infection in ferrets fed with the cheese suspension, can probably be attributed to the lower infectious HPAI H5N1 virus titer in these samples (~1–2 logs lower) compared with the spiked milk (Extended Data Table 2). Given the long-term stability of H5N1 in raw milk and raw-milk cheeses, and its continuous circulation in dairy cattle—which could lead to emergence of new variants (for example, genotype D.1.1 in dairy cattle30)—it is crucial to determine the minimum oral infectious dose of HPAI H5N1 virus in these products to accurately assess the potential risk to human health. Our study has implications for public health, food safety and regulatory policies. First, the current regulation requiring 60-day aging of raw-milk cheese before marketing proves insufficient to achieve HPAI H5N1 virus inactivation and guarantee cheese safety. This concern may extend to other raw-milk products such as yogurt and whey, as the virus can persist for up to 56 days in raw milk under refrigeration13. Therefore, implementing additional mitigation steps, such as testing of raw-milk bulk tanks or using milk pasteurization, thermization or acidification before cheese making, becomes crucial to ensure food safety.

All work involving handling and propagation of HPAI H5N1 virus, cheese making and processing, and virus isolation in ECEs was performed following strict biosafety measures in the Animal Health Diagnostic Center research BSL-3 laboratories at the College of Veterinary Medicine, Cornell University. The animal study procedures were reviewed and approved by the Institutional Animal Care and Use Committee at Cornell University (IACUC approval number 2024-0094). All relevant ethical regulations were followed. Human embryonic kidney cells HEK293T (ATCC CRL-3216) and bovine uterine epithelial cells (Cal-1, developed in house at the Virology Laboratory at the Cornell Animal Health Diagnostic Center) were cultured in Modified Eagle Medium (MEM) supplemented with 1% L-glutamine and 10% fetal bovine serum and containing penicillin–streptomycin (Thermo Fisher Scientific; 10 U ml−1 and 100 µg ml−1, respectively) at 37 °C with 5% CO2.

ed in house at the Virology Laboratory at the Cornell Animal Health Diagnostic Center) were cultured in Modified Eagle Medium (MEM) supplemented with 1% L-glutamine and 10% fetal bovine serum and containing penicillin–streptomycin (Thermo Fisher Scientific; 10 U ml−1 and 100 µg ml−1, respectively) at 37 °C with 5% CO2. The HPAI H5N1 virus isolate TX2/24 (A/Cattle/Texas/063224-24-1/2024, genotype B3.13, GISAID accession number: EPI_ISL_19155861) obtained from pooled milk samples of HPAI H5N1 virus infected dairy cows in Texas, USA1, was used to spike the raw milk (target titer 107 EID50 ml−1) used for the raw-milk cheese studies. The virus stock (passage 3) was propagated in 10-day-old ECEs and titrated in ECEs (EID50) and Cal-1 cells (TCID50). Virus stocks were subjected to whole-genome sequencing to determine the integrity of the HPAI H5N1 TX2/24 virus sequences. The recombinant HPAI H5N1 TX2/24 expressing NanoLuc luciferase reporter (rTX2/24-NLuc) virus was generated using a reverse genetics system as described below.

ECEs (EID50) and Cal-1 cells (TCID50). Virus stocks were subjected to whole-genome sequencing to determine the integrity of the HPAI H5N1 TX2/24 virus sequences. The recombinant HPAI H5N1 TX2/24 expressing NanoLuc luciferase reporter (rTX2/24-NLuc) virus was generated using a reverse genetics system as described below. A reverse genetics system for the bovine HPAI H5N1 virus based on the isolate A/Cattle/Texas/06322424-1/2024 (TX2/24) obtained from milk from infected dairy cows1 was established in our laboratory and used as the backbone to generate a recombinant virus expressing the Nanoluc luciferase reporter gene (rTX2/24-NLuc). In brief, full-length genome sequences of PB1, PB2, PA, HA, NA, NP and M gene segments of TX2/24 strain (H5N1 clade 2.3.4.4b, genotype B3.13, GISAID accession number: EPI_ISL_19155861) were synthesized commercially (Twist Bioscience) and cloned into the dual promoter influenza reverse genetics plasmid pHW200031 (kindly provided by Dr. Richard Webby at St. Jude Children’s Research Hospital) using the BsmBI (New England Biolabs) restriction sites. To generate the NLuc reporter virus, the NS segment of the rTX2/24 recombinant virus was modified to encode a fusion protein (NS-NLuc) in a single nonoverlapping transcript. The NLuc was cloned at the C terminal of NS1. The NS1 and NEP open reading frames were separated by the porcine teschovirus 1 2A autoproteolytic cleavage site. The NS-NLuc gene segment was synthesized (Twist Bioscience) and cloned into pHW2000 vector using the BsmBI sites. The pHW2000 plasmids containing seven TX2/24 gene segments (PB1, PB2, PA, HA, NA, NP and M) and the modified NS segment encoding NS-NLuc fusion were cotransfected into a coculture of HEK293T and Cal-1 (bovine uterine epithelial cells) using Lipofectamine 3000 reagents (Thermo Fisher Scientific). Cell culture supernatant was collected after 96 h and used to infect newly seeded Cal-1 cells. Both cell lysate and culture supernatant were collected after 72–96 h to prepare the seed stock for the rTX2/24-NLuc virus. The working stock of the virus was prepared by inoculation in 10-day-old ECEs via the allantoic cavity route, and the infected allantoic fluid was collected after 48 h. Viruses from the initial rescue (passage 0, P0) and from P1 and P2 were sequenced to confirm the integrity of the sequences and absence of unwanted mutations.

king stock of the virus was prepared by inoculation in 10-day-old ECEs via the allantoic cavity route, and the infected allantoic fluid was collected after 48 h. Viruses from the initial rescue (passage 0, P0) and from P1 and P2 were sequenced to confirm the integrity of the sequences and absence of unwanted mutations. The TCID50 of the virus stock was determined in Cal-1 cells using end-point dilutions and the Spearman and Karber’s method and expressed as TCID50 ml−1. The sequence-verified stock of the rTX2/24-NLuc virus was used in the pH-dependent inhibition assay of the HPAI H5N1 virus. Raw normal bulk tank milk used for cheese making in our BSL-3 laboratory was obtained from the Teaching Dairy at the College of Veterinary Medicine at Cornell University. The physical, chemical and microbiological properties and parameters of the milk samples were analyzed before cheese making in a third-party laboratory (DairyOne) (Extended Data Table 3).

ese making in our BSL-3 laboratory was obtained from the Teaching Dairy at the College of Veterinary Medicine at Cornell University. The physical, chemical and microbiological properties and parameters of the milk samples were analyzed before cheese making in a third-party laboratory (DairyOne) (Extended Data Table 3). To assess the stability of HPAI H5N1 virus in raw-milk cheese, we developed a 5-g mini-cheese model and implemented this model in our BSL3 laboratory. For this, 600 ml of raw bulk tank milk was placed in a sterile glass bottle and spiked with the HPAI H5N1 virus isolate TX2/24 virus to obtain a target titer of about 107 EID50 or 106 TCID50 per milliliter of milk. Next, the pH of the HPAI H5N1 virus-spiked milk was measured using a pH meter (HI2002-01 edge Dedicated pH, Scientific Equipment Source) and then adjusted to three different conditions, pH 6.6, 5.8 and 5.0 using a 50% lactic acid solution (Sigma-Aldrich) to investigate the effect of pH on HPAI H5N1 virus stability in cheese. After pH adjustment, the milk was heated at 34 °C for 1 h in a water bath. Samples (1-ml aliquots) of raw milk were collected after warming and stored at −80 °C for virus titrations. After incubation, we confirmed that the milk had reached 32 °C, and then we proceeded with the coagulation step, by adding 1.875 ml of 32% calcium chloride (GetCulture) and 78 µl of rennet (Chy-Max, Chr. Hansen A/S). The milk bottle was then inverted 25 times for mixing and transferred into a sterile empty pipette tip (1,250 µl) box. The box was placed into a zip lock bag, sealed and then incubated in a water bath at 35 °C for 65 min for coagulum formation. After incubation, we confirmed that the milk temperature reached 35 °C. The coagulated curd (coagulum) was then cut with a sanitized knife, and the tip box was returned to the water bath at 35 °C for 10 min to heal the coagulum and release the whey from the curd. The coagulum and whey were then heated gradually over the course of 30 min in two phases until reaching 43 °C. In the first phase, the temperature of the water bath was increased from 35 °C to 39 °C over a 15-min period (1 °C increase every 2 min; then the temperature was held at 39 °C for the remainder of the 15 min). Similarly, in the second phase, the temperature of the water bath was increased from 39 °C to 43 °C over a 15-min period (1 °C increase every 2 min; then the temperature was held at 43 °C for the remainder of the 15 min).

°C increase every 2 min; then the temperature was held at 39 °C for the remainder of the 15 min). Similarly, in the second phase, the temperature of the water bath was increased from 39 °C to 43 °C over a 15-min period (1 °C increase every 2 min; then the temperature was held at 43 °C for the remainder of the 15 min). The curd temperature was confirmed to be around 39–40 °C. The curd and whey were then incubated at 40 °C for another 30 min. After incubation, we removed the curd and whey from the water bath, removed 60 ml of whey from the container and added 60 ml of sodium chloride solution (0.16 g ml−1) (salting). The box with curd and whey was returned to the water bath after salting and incubated at 41 °C for 20 min. After incubation, the whey was drained from the curd using a sterile cheese cloth and allowed to flow off the curd by gravity. The curd was then cut into small pieces, weighed into 5 g portions and molded into individual wells of a sterile 12-well tissue culture plate. After all the curd was weighed and poured into individual wells of the 12-well tissue culture plate, each mini-cheese was pressed by placing a sterile cap of a 15-ml conical tube (top-side down) on top of each well. The lid of the 12-well plate was placed on the plate, and each plate containing the mini-cheeses was flipped upside down to allow effecting draining of any additional whey into a tray. A 0.5 kg weight was placed on top of the 12-well plates containing the mini-cheeses, the plates were placed in a secondary container with locked lid and the mini-cheeses were pressed for 16 h at 4 °C. The 15-ml conical tube caps were removed after pressing, excess whey was drained and the 12-well plates containing the mini-cheeses were aged at 4 °C.

op of the 12-well plates containing the mini-cheeses, the plates were placed in a secondary container with locked lid and the mini-cheeses were pressed for 16 h at 4 °C. The 15-ml conical tube caps were removed after pressing, excess whey was drained and the 12-well plates containing the mini-cheeses were aged at 4 °C. As part of routine field investigations, the FDA identified company-made raw-milk cheeses suspected of being HPAI H5N1 virus positive, as both the raw milk and raw-milk cheese from the dairy tested positive for the virus. These were cheddar-type raw-milk cheeses molded in 18-kg blocks that were on day 24 of aging. A total of four 1-kg cheese samples were collected by the FDA from the 18-kg cheese blocks and submitted to our laboratory for testing and aging. Upon arrival, 1-g samples were collected from each field cheese block (day 24) and processed for virological assessments as described below. Aging was performed at 4 °C as described for the mini-cheeses in ‘Development of a mini-cheese model and production of raw milk mini-cheese’.

our laboratory for testing and aging. Upon arrival, 1-g samples were collected from each field cheese block (day 24) and processed for virological assessments as described below. Aging was performed at 4 °C as described for the mini-cheeses in ‘Development of a mini-cheese model and production of raw milk mini-cheese’. To assess the stability of HPAI H5N1 virus during the cheese-making process and aging, multiple samples were collected from the raw-milk mini cheeses and the company-made raw-milk cheese studies. During the mini-cheese-making process, milk samples (1 ml) were collected immediately after HPAI H5N1 virus spiking, after pH adjustment and after heating at 34 °C for 1 h. In addition, whey samples were collected at different steps of coagulation and whey separation, including after coagulum formation and incubation at 35 °C for 65 min (whey 1), after heating the curd at 40 °C for 30 min (whey 2) and after salting of the curd by addition of sodium chloride solution and heating at 41 °C for 20 min (whey 3). Furthermore, a 1-g sample of the curd was collected after draining of the whey was completed. All samples were stored at −80 °C for further virological assessments.

ating the curd at 40 °C for 30 min (whey 2) and after salting of the curd by addition of sodium chloride solution and heating at 41 °C for 20 min (whey 3). Furthermore, a 1-g sample of the curd was collected after draining of the whey was completed. All samples were stored at −80 °C for further virological assessments. To assess HPAI H5N1 stability during raw-milk cheese aging, 1-g samples of mini-cheeses or of the company-made raw-milk cheeses were collected throughout the 120-day aging period. Samples from the mini-cheeses were collected daily from days 1 to 7 and then on days 14, 21, 28, 35, 42, 49, 56, 60, 63, 70, 77, 84, 91, 98, 105, 113 and 120 of aging. While cheese from all three experiments was available at pH 6.6 throughout the 120-day aging period, only two and one experiments remained available at pH 5.8 and pH 5.0, respectively, after day 77 of aging. Samples (4 × 1 g) from each of the company-made raw-milk cheeses were collected twice per week until day 98 (days 29, 32, 35, 38, 42, 45, 48, 52, 56, 60, 64, 68, 72, 75, 79, 82, 86, 90, 95 and 98) and then once per week (days 105, 113 and 120) during the 120-day aging period. One 1-g sample of each cheese was transferred to a sterile 7-oz whirl-pak homogenizer blender filter bag and homogenized with 10 ml sterile PBS (10% w/v suspension). The cheese suspension was clarified by centrifuging at 3,000g for 10 min in a refrigerated centrifuge, and 1-ml aliquots of the cheese homogenate were collected in sterile screw-capped vials and stored at −80 °C for virological assessments.

blender filter bag and homogenized with 10 ml sterile PBS (10% w/v suspension). The cheese suspension was clarified by centrifuging at 3,000g for 10 min in a refrigerated centrifuge, and 1-ml aliquots of the cheese homogenate were collected in sterile screw-capped vials and stored at −80 °C for virological assessments. The pH of raw milk used to produce the mini cheeses and following pH adjustment for cheese production was measured by using a pH meter (HI2002-01 edge Dedicated pH, Scientific Equipment Source), and the pH of the mini cheeses and of the company-made raw-milk cheeses was measured using a solid pH probe and the same pH meter (HI14140, Digital flat tip pH electrode for Edge, Scientific Equipment Source). The pH measurements for mini cheeses were taken on days 0, 7, 14, 21, 28, 35, 42, 49, 56 and 60 of aging and for company-made raw-milk cheeses on days 24, 29, 32, 35, 38, 42, 45, 49, 52, 56, 60, 64, 68, 72, 75, 79, 82, 86, 90, 95, 98, 105, 113 and 120.

gital flat tip pH electrode for Edge, Scientific Equipment Source). The pH measurements for mini cheeses were taken on days 0, 7, 14, 21, 28, 35, 42, 49, 56 and 60 of aging and for company-made raw-milk cheeses on days 24, 29, 32, 35, 38, 42, 45, 49, 52, 56, 60, 64, 68, 72, 75, 79, 82, 86, 90, 95, 98, 105, 113 and 120. The water activity cheese affects the growth of most pathogenic bacteria and molds, contributing to the cheese shelf life and aging potential20. The water activity (Aw) in the mini cheeses and the company-made raw-milk cheese was measured using the Aqualab 4TE Water Activity Meter (Aqualab). In brief, 3–4 g of each cheese was chopped or minced into small pieces and transferred to Aqualab Sample Cup Bottoms (Aqualab). The cup was then transferred to the water activity meter, and the Aw was determined. The Aw measurements were taken on the same days of sample collection and pH measurements until day 60 of aging as described in ‘Sample collection and processing’.

to small pieces and transferred to Aqualab Sample Cup Bottoms (Aqualab). The cup was then transferred to the water activity meter, and the Aw was determined. The Aw measurements were taken on the same days of sample collection and pH measurements until day 60 of aging as described in ‘Sample collection and processing’. Viral RNA was extracted from milk, cheese homogenates, allantoic fluid (200 µl) from ECEs inoculated with cheese samples, or ferret swab samples using the IndiMag Pathogen kit (INDICAL Bioscience) on the IndiMag 48s automated nucleic acid extractor (INDICAL Bioscience), following the manufacturer’s instructions. rRT–PCR was performed using the Path-ID Multiplex One-Step RT-PCR Kit (Thermo Fisher). The following primers and probes targeting the M gene of IAV were synthesized commercially (IDT DNA) and used: M+25: AGA TGA GTC TTC TAA CCG AGG TCG, M−124: TGC AAA AAC ATC TTC AAG TCT CTG, M+64 Probe: FAM-TCA GGC CCC CTC AAA GCC GA-BHQ1. The following thermal profile was used: 15 min at 48 °C, 10 min at 95 °C, 40 cycles of 15 s at 95 °C for denaturation and 60 s at 60 °C. A standard curve was prepared using RNA extracted from HPAI H5N1 virus TX2/24 spiked milk samples. Serial tenfold dilutions of the stock virus (2 × 107 log TCID50 ml−1) were prepared in raw-milk samples for RNA extraction followed by RT–PCR. The Ct values were used to estimate the viral RNA copy number in the tested samples using the relative quantification method.

ed from HPAI H5N1 virus TX2/24 spiked milk samples. Serial tenfold dilutions of the stock virus (2 × 107 log TCID50 ml−1) were prepared in raw-milk samples for RNA extraction followed by RT–PCR. The Ct values were used to estimate the viral RNA copy number in the tested samples using the relative quantification method. For virus titration in eggs, 10-day-old ECEs were used. Serial tenfold dilutions of milk, whey, cheese homogenate or ferret swab samples were prepared in PBS supplemented with antibiotic–antimycotic (anti-anti 100X, Thermo Fisher Scientific). The ECEs were candled to mark the air sac on the shell, then sanitized with 70% ethanol, and a hole was drilled in the eggshell. One-hundred microliters of the diluted milk or homogenized cheese samples was injected in triplicate into the allantoic cavity route and sealed with glue. Eggs were candled daily for 4 days, and dead embryos were chilled overnight before collection of allantoic fluid. After 4 days of inoculation, all eggs were chilled for 24 h, and allantoic fluid was collected. All allantoic fluid samples were tested by HA assay using 0.5% turkey red blood cells (RBC). Finally, the 50% embryo infectious dose (EID50) was calculated using the Reed and Muench method.

ore collection of allantoic fluid. After 4 days of inoculation, all eggs were chilled for 24 h, and allantoic fluid was collected. All allantoic fluid samples were tested by HA assay using 0.5% turkey red blood cells (RBC). Finally, the 50% embryo infectious dose (EID50) was calculated using the Reed and Muench method. The presence of IAV in the allantoic fluid of ECE inoculated with select cheese homogenate samples (days 60, 91 and 120 of aging for mini-cheeses or days 24, 42, 45, 60, 72, 82, 90, 105 and 120 of aging for company-made raw-milk cheeses) was confirmed by rRT–PCR as in ‘RNA extraction and RT–PCR’. In addition, influenza A whole-genome sequencing was performed on company-made raw-milk cheese samples and allantoic fluids collected on day 24 of aging to confirm the presence of H5N1 virus. For the HA assay, 100 µl allantoic fluid was taken into the wells of the first column of a 96-well U-bottom plate. Fifty microliters of PBS was added to the wells of the second and third columns. Two twofold dilutions (1:2 to 1:8) of the allantoic fluids were prepared in PBS. Then, 50 µl of 0.5% turkey RBC was added to each well and incubated at room temperature for 30 min. Lack of RBC button formation indicates a positive HA reaction.

y microliters of PBS was added to the wells of the second and third columns. Two twofold dilutions (1:2 to 1:8) of the allantoic fluids were prepared in PBS. Then, 50 µl of 0.5% turkey RBC was added to each well and incubated at room temperature for 30 min. Lack of RBC button formation indicates a positive HA reaction. The whole-genome sequence of influenza A present in the company-made raw-milk cheese samples was determined using targeted influenza A sequencing at the Virology Laboratory at the Cornell Animal Health Diagnostic Center as previously described1. Sequence and phylogenetic analyses were performed as previously described1. The complete genome sequences were deposited in the Global Initiative on Sharing All Influenza Data (GISAID database (accession numbers EPI_ISL_20155217–EPI_ISL_20155220). The dataset consisted of HPAI H5N1 clade 2.3.4.4b genomes from samples collected between January and February 2025 in the American continent, downloaded from the GISAID Epiflu32, and complete genomes from the present study that include four genomes obtained from company-made cheese samples. Phylogenomic analyses were performed on complete genomes, formed by concatenation of all gene segments, using the Augur v21.0.1 tool kit33 procedures implemented in Nextstrain34. In brief, multiple sequence alignments were performed using MAFFT (v7.515)35; maximum likelihood phylogenetic trees were inferred using IQ-TREE (v1.6.12)36. Discrete trait analysis was performed using TreeTime (v0.9.4)37, and the resultant dataset was visualized through Auspice (https://auspice.us/).

d in Nextstrain34. In brief, multiple sequence alignments were performed using MAFFT (v7.515)35; maximum likelihood phylogenetic trees were inferred using IQ-TREE (v1.6.12)36. Discrete trait analysis was performed using TreeTime (v0.9.4)37, and the resultant dataset was visualized through Auspice (https://auspice.us/). The thermal inactivation kinetics of HPAI H5N1 in raw-milk cheese were calculated on the basis of the decimal reduction time (D value). The D value is defined as the time required at a specific temperature to achieve a one-logarithm reduction in viral titer. This was determined by plotting the logarithm (base 10) of the infectious viral titers against time for cheese made with milk at pH 6.6 and 5.8. The D value was then calculated as the negative inverse of the slope of the resulting plot, with the line of best fit for the survivor curves established through regression analysis.

s determined by plotting the logarithm (base 10) of the infectious viral titers against time for cheese made with milk at pH 6.6 and 5.8. The D value was then calculated as the negative inverse of the slope of the resulting plot, with the line of best fit for the survivor curves established through regression analysis. Cal-1 cells were seeded in a 24-well plate (2.5 × 105 cells ml−1) and incubated at 37 °C for 24 h. Raw normal bulk tank milk or PBS was spiked with the recombinant HPAI H5N1 TX2/24-NLuc virus to obtain a target titer of about 107 EID50 or 106 TCID50 per milliliter of milk. The pH of the raw bulk tank milk or PBS was left unadjusted at pH 6.78 or adjusted to pH 6.0, 5.5, 5.0 and 4.5 and incubated at 4 °C for 1 h. Cal-1 cells were washed with sterile PBS, and 200 µl of the milk or PBS samples was added to each well. Cells inoculated with raw normal milk or control PBS were kept as a control. The plate was transferred to a refrigerator and incubated for 1 h (virus adsorption) at 4 °C. After incubation, cells were washed twice with sterile PBS, and 500 µl of complete growth medium (MEM 10% fetal bovine serum) was added to each well. The plate was then incubated at 37 °C for 4 h. After incubation, cell supernatant was aspirated, and 50 µl of 1× Passive Lysis Buffer (Promega) was added to each well and incubated at room temperature for 10 min. After one cycle of freeze–thaw, 10 µl of cell lysate was transferred to each well of a luminometer plate to which 50 µl of luciferase assay reagent (Nano-Glo, Promega) was added. Luminescence was measured using a luminometer plate reader (BioTek Synergy LX Multimode Reader). Virus infectivity was determined by normalizing the luminescence (relative light units) of HPAI H5N1-virus infected to mock-infected control cells and expressed as fold changes of the mock cell control.

Promega) was added. Luminescence was measured using a luminometer plate reader (BioTek Synergy LX Multimode Reader). Virus infectivity was determined by normalizing the luminescence (relative light units) of HPAI H5N1-virus infected to mock-infected control cells and expressed as fold changes of the mock cell control. Twenty 8–9-week-old influenza negative ferrets (Mustela furo) (10 females and 10 males) were obtained from Tripple F Farms. All animals were housed individually in Horsfall HEPA-filtered cages in the animal biosafety level 3 facility at the East Campus Research Facility at Cornell University. To test the infectivity of HPAI H5N1 after oral exposure through ingestion, we fed ferrets voluntarily with HPAI H5N1 virus-spiked raw milk (5 ml, n = 4) and company-made raw-milk cheese inadvertently produced with naturally contaminated milk (5 g solid cheese or 5 ml 10% cheese suspension, n = 4 per group). In addition, ferrets were directly inoculated P/O (n = 2) or I/N (n = 2) with HPAI H5N1 virus-spiked milk. For P/O, animals were sedated. Holding the animals in an upright position, 1 ml of the HPAI-spiked milk was poured slowly into the oropharynx. Animals were held in an upright position until they swallowed the liquid. For I/N, animals were sedated and held in an upright position as for P/O innoculation. Then, 1 ml of the HPAI-spiked milk was slowly added drop by drop into the nostrils (0.5 ml into each nostril). For voluntary feeding, milk and cheese (solid and suspension) samples were provided in small cups. The virus titers in the HPAI H5N1 virus-spiked milk were adjusted to 5 × 104 EID50 to match the virus titers obtained from the company-made raw-milk cheese samples after the 60-day aging period. Ferrets fed with company-made store bought cheese (5 g, n = 4) were kept as negative controls. For the direct oral and intranasal routes, a single inoculation was performed, whereas the voluntary feeding groups received four repeated feedings on days 0, 3, 4, and 5 p.e. After inoculation/feeding, clinical parameters, including body temperature (using a transponder), body weight and clinical signs, were monitored daily. NS, OPS and RSwere collected on days 0, 1, 3, 5, 7, 10 and 14 p.e./feeding. Upon collection, swabs were placed in sterile tubes containing 1 ml of viral transport medium (VTM Corning) and stored at −80 °C until processed for further analyses.

(using a transponder), body weight and clinical signs, were monitored daily. NS, OPS and RSwere collected on days 0, 1, 3, 5, 7, 10 and 14 p.e./feeding. Upon collection, swabs were placed in sterile tubes containing 1 ml of viral transport medium (VTM Corning) and stored at −80 °C until processed for further analyses. Blood was collected from the cranial vena cava using a 1 ml sterile syringe with a 25 G × 1″ needle and transferred into serum separator tubes on days 0, 7 and 14 p.e. The blood tubes were centrifuged at 1,200g for 10 min, and serum was aliquoted and stored at −20 °C until further analysis. Animals that reached the humane endpoint or survived until day 14 p.e. were humanely euthanized. The study procedures were reviewed and approved by the Institutional Animal Care and Use Committee at Cornell University (IACUC approval number 2024-0094).

nd serum was aliquoted and stored at −20 °C until further analysis. Animals that reached the humane endpoint or survived until day 14 p.e. were humanely euthanized. The study procedures were reviewed and approved by the Institutional Animal Care and Use Committee at Cornell University (IACUC approval number 2024-0094). Neutralizing antibodies in serum against HPAI H5N1 virus was assessed by virus neutralization assay using a recombinant TX2/24 virus expressing miniGFP2 (rTX2/24-miniGFP2) as previously described38. For this, serial twofold serum dilutions (1:8 to 1:1,024) of each serum sample were prepared in MEM and incubated with 200 TCID50 of rTX2/24-miniGFP2 for 1 h at 37 °C. Next, 100 µl of a cell suspension of Cal-1 cells was added to each well of a 96-well plate and incubated at 37 °C for 48 h. Plates were visualized using a fluorescence microscope (Hybrid microscope ECHO Revolve 3K) to determine neutralizing antibody titers, expressed as the reciprocal of the highest serum dilution capable of completely inhibiting HPAI H5N1 virus replication based on the expression of miniGFP2 by the rTX2/24-miniGFP2 virus. Known positive and negative control serum samples were used in the assay.

ECHO Revolve 3K) to determine neutralizing antibody titers, expressed as the reciprocal of the highest serum dilution capable of completely inhibiting HPAI H5N1 virus replication based on the expression of miniGFP2 by the rTX2/24-miniGFP2 virus. Known positive and negative control serum samples were used in the assay. For thermal stability of HPAI H5N1 virus in cheese, three independent experiments were performed for each pH tested. Bulk tank raw milk was used to reduce the effect of sample variation and resemble real-world cheese-making practices. For company-made cheese, four cheese blocks were tested at each time point. For animal experimentations, two to four animals were used per group per condition and both male and female animals were used equally. Animals were randomly allocated to different experimental groups. The effect of sex was not evaluated separately in our study. We used robust testing methods established and validated in our laboratory (RT–PCR, cell culture and embryonated eggs, virus titration and virus neutralization assays). For experiments with two factors (for example, time and test/treatment) a two-way analysis of variance (ANOVA) followed by Tukey’s multiple-comparisons test was used (P < 0.05 was considered statistically significant). No data were excluded from the analysis. Graphs were prepared using GraphPad Prism 10 software. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

All work involving handling and propagation of HPAI H5N1 virus, cheese making and processing, and virus isolation in ECEs was performed following strict biosafety measures in the Animal Health Diagnostic Center research BSL-3 laboratories at the College of Veterinary Medicine, Cornell University. The animal study procedures were reviewed and approved by the Institutional Animal Care and Use Committee at Cornell University (IACUC approval number 2024-0094). All relevant ethical regulations were followed.

Human embryonic kidney cells HEK293T (ATCC CRL-3216) and bovine uterine epithelial cells (Cal-1, developed in house at the Virology Laboratory at the Cornell Animal Health Diagnostic Center) were cultured in Modified Eagle Medium (MEM) supplemented with 1% L-glutamine and 10% fetal bovine serum and containing penicillin–streptomycin (Thermo Fisher Scientific; 10 U ml−1 and 100 µg ml−1, respectively) at 37 °C with 5% CO2.

The HPAI H5N1 virus isolate TX2/24 (A/Cattle/Texas/063224-24-1/2024, genotype B3.13, GISAID accession number: EPI_ISL_19155861) obtained from pooled milk samples of HPAI H5N1 virus infected dairy cows in Texas, USA1, was used to spike the raw milk (target titer 107 EID50 ml−1) used for the raw-milk cheese studies. The virus stock (passage 3) was propagated in 10-day-old ECEs and titrated in ECEs (EID50) and Cal-1 cells (TCID50). Virus stocks were subjected to whole-genome sequencing to determine the integrity of the HPAI H5N1 TX2/24 virus sequences. The recombinant HPAI H5N1 TX2/24 expressing NanoLuc luciferase reporter (rTX2/24-NLuc) virus was generated using a reverse genetics system as described below.

A reverse genetics system for the bovine HPAI H5N1 virus based on the isolate A/Cattle/Texas/06322424-1/2024 (TX2/24) obtained from milk from infected dairy cows1 was established in our laboratory and used as the backbone to generate a recombinant virus expressing the Nanoluc luciferase reporter gene (rTX2/24-NLuc). In brief, full-length genome sequences of PB1, PB2, PA, HA, NA, NP and M gene segments of TX2/24 strain (H5N1 clade 2.3.4.4b, genotype B3.13, GISAID accession number: EPI_ISL_19155861) were synthesized commercially (Twist Bioscience) and cloned into the dual promoter influenza reverse genetics plasmid pHW200031 (kindly provided by Dr. Richard Webby at St. Jude Children’s Research Hospital) using the BsmBI (New England Biolabs) restriction sites. To generate the NLuc reporter virus, the NS segment of the rTX2/24 recombinant virus was modified to encode a fusion protein (NS-NLuc) in a single nonoverlapping transcript. The NLuc was cloned at the C terminal of NS1. The NS1 and NEP open reading frames were separated by the porcine teschovirus 1 2A autoproteolytic cleavage site. The NS-NLuc gene segment was synthesized (Twist Bioscience) and cloned into pHW2000 vector using the BsmBI sites. The pHW2000 plasmids containing seven TX2/24 gene segments (PB1, PB2, PA, HA, NA, NP and M) and the modified NS segment encoding NS-NLuc fusion were cotransfected into a coculture of HEK293T and Cal-1 (bovine uterine epithelial cells) using Lipofectamine 3000 reagents (Thermo Fisher Scientific). Cell culture supernatant was collected after 96 h and used to infect newly seeded Cal-1 cells. Both cell lysate and culture supernatant were collected after 72–96 h to prepare the seed stock for the rTX2/24-NLuc virus. The working stock of the virus was prepared by inoculation in 10-day-old ECEs via the allantoic cavity route, and the infected allantoic fluid was collected after 48 h. Viruses from the initial rescue (passage 0, P0) and from P1 and P2 were sequenced to confirm the integrity of the sequences and absence of unwanted mutations.

king stock of the virus was prepared by inoculation in 10-day-old ECEs via the allantoic cavity route, and the infected allantoic fluid was collected after 48 h. Viruses from the initial rescue (passage 0, P0) and from P1 and P2 were sequenced to confirm the integrity of the sequences and absence of unwanted mutations. The TCID50 of the virus stock was determined in Cal-1 cells using end-point dilutions and the Spearman and Karber’s method and expressed as TCID50 ml−1. The sequence-verified stock of the rTX2/24-NLuc virus was used in the pH-dependent inhibition assay of the HPAI H5N1 virus.

Raw normal bulk tank milk used for cheese making in our BSL-3 laboratory was obtained from the Teaching Dairy at the College of Veterinary Medicine at Cornell University. The physical, chemical and microbiological properties and parameters of the milk samples were analyzed before cheese making in a third-party laboratory (DairyOne) (Extended Data Table 3).

To assess the stability of HPAI H5N1 virus in raw-milk cheese, we developed a 5-g mini-cheese model and implemented this model in our BSL3 laboratory. For this, 600 ml of raw bulk tank milk was placed in a sterile glass bottle and spiked with the HPAI H5N1 virus isolate TX2/24 virus to obtain a target titer of about 107 EID50 or 106 TCID50 per milliliter of milk. Next, the pH of the HPAI H5N1 virus-spiked milk was measured using a pH meter (HI2002-01 edge Dedicated pH, Scientific Equipment Source) and then adjusted to three different conditions, pH 6.6, 5.8 and 5.0 using a 50% lactic acid solution (Sigma-Aldrich) to investigate the effect of pH on HPAI H5N1 virus stability in cheese. After pH adjustment, the milk was heated at 34 °C for 1 h in a water bath. Samples (1-ml aliquots) of raw milk were collected after warming and stored at −80 °C for virus titrations. After incubation, we confirmed that the milk had reached 32 °C, and then we proceeded with the coagulation step, by adding 1.875 ml of 32% calcium chloride (GetCulture) and 78 µl of rennet (Chy-Max, Chr. Hansen A/S). The milk bottle was then inverted 25 times for mixing and transferred into a sterile empty pipette tip (1,250 µl) box. The box was placed into a zip lock bag, sealed and then incubated in a water bath at 35 °C for 65 min for coagulum formation. After incubation, we confirmed that the milk temperature reached 35 °C. The coagulated curd (coagulum) was then cut with a sanitized knife, and the tip box was returned to the water bath at 35 °C for 10 min to heal the coagulum and release the whey from the curd. The coagulum and whey were then heated gradually over the course of 30 min in two phases until reaching 43 °C. In the first phase, the temperature of the water bath was increased from 35 °C to 39 °C over a 15-min period (1 °C increase every 2 min; then the temperature was held at 39 °C for the remainder of the 15 min). Similarly, in the second phase, the temperature of the water bath was increased from 39 °C to 43 °C over a 15-min period (1 °C increase every 2 min; then the temperature was held at 43 °C for the remainder of the 15 min).

As part of routine field investigations, the FDA identified company-made raw-milk cheeses suspected of being HPAI H5N1 virus positive, as both the raw milk and raw-milk cheese from the dairy tested positive for the virus. These were cheddar-type raw-milk cheeses molded in 18-kg blocks that were on day 24 of aging. A total of four 1-kg cheese samples were collected by the FDA from the 18-kg cheese blocks and submitted to our laboratory for testing and aging. Upon arrival, 1-g samples were collected from each field cheese block (day 24) and processed for virological assessments as described below. Aging was performed at 4 °C as described for the mini-cheeses in ‘Development of a mini-cheese model and production of raw milk mini-cheese’.

To assess the stability of HPAI H5N1 virus during the cheese-making process and aging, multiple samples were collected from the raw-milk mini cheeses and the company-made raw-milk cheese studies. During the mini-cheese-making process, milk samples (1 ml) were collected immediately after HPAI H5N1 virus spiking, after pH adjustment and after heating at 34 °C for 1 h. In addition, whey samples were collected at different steps of coagulation and whey separation, including after coagulum formation and incubation at 35 °C for 65 min (whey 1), after heating the curd at 40 °C for 30 min (whey 2) and after salting of the curd by addition of sodium chloride solution and heating at 41 °C for 20 min (whey 3). Furthermore, a 1-g sample of the curd was collected after draining of the whey was completed. All samples were stored at −80 °C for further virological assessments.

The pH of raw milk used to produce the mini cheeses and following pH adjustment for cheese production was measured by using a pH meter (HI2002-01 edge Dedicated pH, Scientific Equipment Source), and the pH of the mini cheeses and of the company-made raw-milk cheeses was measured using a solid pH probe and the same pH meter (HI14140, Digital flat tip pH electrode for Edge, Scientific Equipment Source). The pH measurements for mini cheeses were taken on days 0, 7, 14, 21, 28, 35, 42, 49, 56 and 60 of aging and for company-made raw-milk cheeses on days 24, 29, 32, 35, 38, 42, 45, 49, 52, 56, 60, 64, 68, 72, 75, 79, 82, 86, 90, 95, 98, 105, 113 and 120.

The water activity cheese affects the growth of most pathogenic bacteria and molds, contributing to the cheese shelf life and aging potential20. The water activity (Aw) in the mini cheeses and the company-made raw-milk cheese was measured using the Aqualab 4TE Water Activity Meter (Aqualab). In brief, 3–4 g of each cheese was chopped or minced into small pieces and transferred to Aqualab Sample Cup Bottoms (Aqualab). The cup was then transferred to the water activity meter, and the Aw was determined. The Aw measurements were taken on the same days of sample collection and pH measurements until day 60 of aging as described in ‘Sample collection and processing’.

Viral RNA was extracted from milk, cheese homogenates, allantoic fluid (200 µl) from ECEs inoculated with cheese samples, or ferret swab samples using the IndiMag Pathogen kit (INDICAL Bioscience) on the IndiMag 48s automated nucleic acid extractor (INDICAL Bioscience), following the manufacturer’s instructions. rRT–PCR was performed using the Path-ID Multiplex One-Step RT-PCR Kit (Thermo Fisher). The following primers and probes targeting the M gene of IAV were synthesized commercially (IDT DNA) and used: M+25: AGA TGA GTC TTC TAA CCG AGG TCG, M−124: TGC AAA AAC ATC TTC AAG TCT CTG, M+64 Probe: FAM-TCA GGC CCC CTC AAA GCC GA-BHQ1. The following thermal profile was used: 15 min at 48 °C, 10 min at 95 °C, 40 cycles of 15 s at 95 °C for denaturation and 60 s at 60 °C. A standard curve was prepared using RNA extracted from HPAI H5N1 virus TX2/24 spiked milk samples. Serial tenfold dilutions of the stock virus (2 × 107 log TCID50 ml−1) were prepared in raw-milk samples for RNA extraction followed by RT–PCR. The Ct values were used to estimate the viral RNA copy number in the tested samples using the relative quantification method.

For virus titration in eggs, 10-day-old ECEs were used. Serial tenfold dilutions of milk, whey, cheese homogenate or ferret swab samples were prepared in PBS supplemented with antibiotic–antimycotic (anti-anti 100X, Thermo Fisher Scientific). The ECEs were candled to mark the air sac on the shell, then sanitized with 70% ethanol, and a hole was drilled in the eggshell. One-hundred microliters of the diluted milk or homogenized cheese samples was injected in triplicate into the allantoic cavity route and sealed with glue. Eggs were candled daily for 4 days, and dead embryos were chilled overnight before collection of allantoic fluid. After 4 days of inoculation, all eggs were chilled for 24 h, and allantoic fluid was collected. All allantoic fluid samples were tested by HA assay using 0.5% turkey red blood cells (RBC). Finally, the 50% embryo infectious dose (EID50) was calculated using the Reed and Muench method. The presence of IAV in the allantoic fluid of ECE inoculated with select cheese homogenate samples (days 60, 91 and 120 of aging for mini-cheeses or days 24, 42, 45, 60, 72, 82, 90, 105 and 120 of aging for company-made raw-milk cheeses) was confirmed by rRT–PCR as in ‘RNA extraction and RT–PCR’. In addition, influenza A whole-genome sequencing was performed on company-made raw-milk cheese samples and allantoic fluids collected on day 24 of aging to confirm the presence of H5N1 virus.

For the HA assay, 100 µl allantoic fluid was taken into the wells of the first column of a 96-well U-bottom plate. Fifty microliters of PBS was added to the wells of the second and third columns. Two twofold dilutions (1:2 to 1:8) of the allantoic fluids were prepared in PBS. Then, 50 µl of 0.5% turkey RBC was added to each well and incubated at room temperature for 30 min. Lack of RBC button formation indicates a positive HA reaction.

The whole-genome sequence of influenza A present in the company-made raw-milk cheese samples was determined using targeted influenza A sequencing at the Virology Laboratory at the Cornell Animal Health Diagnostic Center as previously described1. Sequence and phylogenetic analyses were performed as previously described1. The complete genome sequences were deposited in the Global Initiative on Sharing All Influenza Data (GISAID database (accession numbers EPI_ISL_20155217–EPI_ISL_20155220).

The dataset consisted of HPAI H5N1 clade 2.3.4.4b genomes from samples collected between January and February 2025 in the American continent, downloaded from the GISAID Epiflu32, and complete genomes from the present study that include four genomes obtained from company-made cheese samples. Phylogenomic analyses were performed on complete genomes, formed by concatenation of all gene segments, using the Augur v21.0.1 tool kit33 procedures implemented in Nextstrain34. In brief, multiple sequence alignments were performed using MAFFT (v7.515)35; maximum likelihood phylogenetic trees were inferred using IQ-TREE (v1.6.12)36. Discrete trait analysis was performed using TreeTime (v0.9.4)37, and the resultant dataset was visualized through Auspice (https://auspice.us/).

The thermal inactivation kinetics of HPAI H5N1 in raw-milk cheese were calculated on the basis of the decimal reduction time (D value). The D value is defined as the time required at a specific temperature to achieve a one-logarithm reduction in viral titer. This was determined by plotting the logarithm (base 10) of the infectious viral titers against time for cheese made with milk at pH 6.6 and 5.8. The D value was then calculated as the negative inverse of the slope of the resulting plot, with the line of best fit for the survivor curves established through regression analysis.

Cal-1 cells were seeded in a 24-well plate (2.5 × 105 cells ml−1) and incubated at 37 °C for 24 h. Raw normal bulk tank milk or PBS was spiked with the recombinant HPAI H5N1 TX2/24-NLuc virus to obtain a target titer of about 107 EID50 or 106 TCID50 per milliliter of milk. The pH of the raw bulk tank milk or PBS was left unadjusted at pH 6.78 or adjusted to pH 6.0, 5.5, 5.0 and 4.5 and incubated at 4 °C for 1 h. Cal-1 cells were washed with sterile PBS, and 200 µl of the milk or PBS samples was added to each well. Cells inoculated with raw normal milk or control PBS were kept as a control. The plate was transferred to a refrigerator and incubated for 1 h (virus adsorption) at 4 °C. After incubation, cells were washed twice with sterile PBS, and 500 µl of complete growth medium (MEM 10% fetal bovine serum) was added to each well. The plate was then incubated at 37 °C for 4 h. After incubation, cell supernatant was aspirated, and 50 µl of 1× Passive Lysis Buffer (Promega) was added to each well and incubated at room temperature for 10 min. After one cycle of freeze–thaw, 10 µl of cell lysate was transferred to each well of a luminometer plate to which 50 µl of luciferase assay reagent (Nano-Glo, Promega) was added. Luminescence was measured using a luminometer plate reader (BioTek Synergy LX Multimode Reader). Virus infectivity was determined by normalizing the luminescence (relative light units) of HPAI H5N1-virus infected to mock-infected control cells and expressed as fold changes of the mock cell control.

Twenty 8–9-week-old influenza negative ferrets (Mustela furo) (10 females and 10 males) were obtained from Tripple F Farms. All animals were housed individually in Horsfall HEPA-filtered cages in the animal biosafety level 3 facility at the East Campus Research Facility at Cornell University. To test the infectivity of HPAI H5N1 after oral exposure through ingestion, we fed ferrets voluntarily with HPAI H5N1 virus-spiked raw milk (5 ml, n = 4) and company-made raw-milk cheese inadvertently produced with naturally contaminated milk (5 g solid cheese or 5 ml 10% cheese suspension, n = 4 per group). In addition, ferrets were directly inoculated P/O (n = 2) or I/N (n = 2) with HPAI H5N1 virus-spiked milk. For P/O, animals were sedated. Holding the animals in an upright position, 1 ml of the HPAI-spiked milk was poured slowly into the oropharynx. Animals were held in an upright position until they swallowed the liquid. For I/N, animals were sedated and held in an upright position as for P/O innoculation. Then, 1 ml of the HPAI-spiked milk was slowly added drop by drop into the nostrils (0.5 ml into each nostril). For voluntary feeding, milk and cheese (solid and suspension) samples were provided in small cups. The virus titers in the HPAI H5N1 virus-spiked milk were adjusted to 5 × 104 EID50 to match the virus titers obtained from the company-made raw-milk cheese samples after the 60-day aging period. Ferrets fed with company-made store bought cheese (5 g, n = 4) were kept as negative controls. For the direct oral and intranasal routes, a single inoculation was performed, whereas the voluntary feeding groups received four repeated feedings on days 0, 3, 4, and 5 p.e. After inoculation/feeding, clinical parameters, including body temperature (using a transponder), body weight and clinical signs, were monitored daily. NS, OPS and RSwere collected on days 0, 1, 3, 5, 7, 10 and 14 p.e./feeding. Upon collection, swabs were placed in sterile tubes containing 1 ml of viral transport medium (VTM Corning) and stored at −80 °C until processed for further analyses.

Neutralizing antibodies in serum against HPAI H5N1 virus was assessed by virus neutralization assay using a recombinant TX2/24 virus expressing miniGFP2 (rTX2/24-miniGFP2) as previously described38. For this, serial twofold serum dilutions (1:8 to 1:1,024) of each serum sample were prepared in MEM and incubated with 200 TCID50 of rTX2/24-miniGFP2 for 1 h at 37 °C. Next, 100 µl of a cell suspension of Cal-1 cells was added to each well of a 96-well plate and incubated at 37 °C for 48 h. Plates were visualized using a fluorescence microscope (Hybrid microscope ECHO Revolve 3K) to determine neutralizing antibody titers, expressed as the reciprocal of the highest serum dilution capable of completely inhibiting HPAI H5N1 virus replication based on the expression of miniGFP2 by the rTX2/24-miniGFP2 virus. Known positive and negative control serum samples were used in the assay.

For thermal stability of HPAI H5N1 virus in cheese, three independent experiments were performed for each pH tested. Bulk tank raw milk was used to reduce the effect of sample variation and resemble real-world cheese-making practices. For company-made cheese, four cheese blocks were tested at each time point. For animal experimentations, two to four animals were used per group per condition and both male and female animals were used equally. Animals were randomly allocated to different experimental groups. The effect of sex was not evaluated separately in our study. We used robust testing methods established and validated in our laboratory (RT–PCR, cell culture and embryonated eggs, virus titration and virus neutralization assays). For experiments with two factors (for example, time and test/treatment) a two-way analysis of variance (ANOVA) followed by Tukey’s multiple-comparisons test was used (P < 0.05 was considered statistically significant). No data were excluded from the analysis. Graphs were prepared using GraphPad Prism 10 software.

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 10.1038/s41591-025-04010-0.

Source Data Fig. 2Statistical source data. Source Data Fig. 3Statistical source data. Source Data Fig. 5Statistical source data. Source Data Fig. 6Statistical source data. Source Data Extended Data Fig. 1Statistical source data. Source Data Extended Data Fig. 2Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data.