In late March 2024, highly pathogenic avian influenza virus (HPAI) of the H5N1 subtype was for the first time detected in nasal swabs and milk of dairy cows, increasing concern that HPAI A(H5N1) viruses may enter the human food chain. The Texas A&M Veterinary Medical Diagnostic Laboratory obtained cow’s milk samples from an affected herd in New Mexico, from which eight HPAI A(H5N1) viruses were isolated (Table S1; for details, see the Supplementary Appendix, available with the full text of this letter at NEJM.org).

We compared the genetic origin of these HPAI A(H5N1) milk virus isolates with the sequences publicly available at the time of our analysis (Fig. S1 in the Supplementary Appendix). The cow viruses form a single clade encompassing many smaller clades of viruses isolated from cats, raccoons, chickens, and wild birds. The phylogeny is consistent with a single introduction into cows. The viruses isolated in our study (labeled in Fig. S1) fall within the clade of publicly available cow virus sequences, including that from a human isolate, A/Texas/37/2024 (Fig. S1). Further assessment of the cow virus sequences and all avian influenza A virus sequences collected in the Americas since the start of 2020 identified a reassortment event for NP and PB2 segments that occurred immediately before the introduction of HPAI A(H5N1) viruses into cows (Fig. S2), consistent with findings reported by Anderson and colleagues.1

Studies involving foot-and-mouth disease virus revealed that heat inactivation of virus-positive milk samples required higher temperature or longer incubation times (or both) than heat inactivation of virus spiked into milk,2,3 presumably because fat globules and casein micelles may partly protect viruses in virus-positive milk samples. Accordingly, we tested heat inactivation of four HPAI A(H5N1) virus–positive milk samples (NM#93, NM#115, KS#3, and KS#6). Undiluted milk samples were incubated in a polymerase-chain-reaction (PCR) thermocycler at 63°C for 5, 10, 20, or 30 minutes or at 72°C for 5, 10, 15, 20, or 30 seconds (Table 1; see also the Supplementary Appendix). Control samples were left untreated. Heat treatment at 63°C reduced the virus titers below the detection limit of the TCID50 (50% tissue-culture infectious dose) assay (1.5 log10/ml). Heat treatment at 72°C was performed, with the default settings of the PCR thermocycler (i.e., preheated lid at 105°C) or with a metal lid (heated to 72°C) covering the PCR block (see the Supplementary Appendix for details). After heat treatment, samples were inoculated into embryonated chicken eggs or Madin–Darby canine kidney (MDCK) cells for virus detection. Under these conditions, heat treatment for 15 or 20 seconds reduced virus titers by more than 4.5 log units but did not completely inactivate the virus (Table 1). We emphasize that the conditions used in our laboratory study are not identical to the large-scale industrial treatment of raw milk.

The stability of HPAI A(H5N1) virus in cow’s milk stored at 4°C is another important question. For milk sample NM#93, we detected a decline of only two log units over 5 weeks. HPAI A(H5N1) virus may therefore remain infectious for several weeks in raw milk kept at 4°C.

To further assess the risk that HPAI A(H5N1)–positive milk poses to animals and humans, we orally inoculated BALB/cJ mice with 50 μl (3×106 pfu) of sample NM#93. The animals showed signs of illness starting on day 1, including ruffled fur and lethargy. All the animals survived until day 4, when they were euthanized to determine virus titers in multiple organs (Figure S3 in the Supplementary Appendix). We detected high virus titers in the respiratory organs (which suggests that infection may have occurred through the pharynx) and moderate virus titers in several other organs, findings consistent with the systemic infections typically caused by HPAI H5 viruses in mammals. Detection of virus in the mammary glands of two mice was consistent with the high virus load in the milk of lactating cows, even though these mice were not lactating. Collectively, our data indicate that HPAI A(H5N1) virus in untreated milk can infect susceptible animals that consume it. In summary, HPAI H5–positive milk poses a risk when consumed untreated, but heat inactivation under the laboratory conditions used here reduces HPAI H5 virus titers by more than 4.5 log units. However, bench-top experiments do not recapitulate commercial pasteurization processes.