Evidence of H5N1 Nasal Exposure to Five Mice
The very brief ‘correspondence’ by Guan and colleagues in the New England Journal of Medicine, from my perspective as a medical microbiologist and microbial risk analyst, fails to provide definitive data that supports the claims by the authors and their followers in the media.
This blog will address what we know from this ‘correspondence’ in reference to full studies published in the peer-reviewed literature. It will also introduce questions about extrapolating data on transmission and infection in animal models to humans, as well as what is uncertain or unknown or assumed.
Let’s examine the lines of evidence from the brief ‘correspondence’.
1. Researchers anesthetized five laboratory mice (Balb/cJ) and inoculated one dose-level (50 μl containing 3 million pfu) of a sample of raw milk from New Mexico cows (NM#93) infected with influenza A H5N1 (H5N1 hereafter), also termed highly pathogenic avian influenza. The raw milk sample was inoculated in the back of the throats of the five anesthetized mice with a micropipette. Researchers reported that these five mice were noted to have swallowed, but also to have had milk in their nasal cavities. The only signs of illness noted were ruffled fur and lethargy.
2. Take a look at evidence of viral detection reported for these 5 mice, euthanized at day 4 after inoculation (Figure S3 below). No evidence of viral detection was reported in intestine or feces as would have been expected if H5N1 caused intestinal flu. The highest titers were reported in lung, trachea, and nasal turbinates (the networks of bones, vessels, and tissue within nasal passageways, markedly different in mice and humans). This pattern is consistent with transmission via the respiratory system.
3. The authors’ Table 1 reported results on heat inactivation for 4 milk samples positive for virus (NM#93, NM#115, KS#3, and KS#6) and controls not subject to heat. A further claim is that over five weeks, antiviral activity of refrigerated raw milk samples was demonstrated, but the report of a 2-log viral titer decrease by five weeks was not accompanied by raw data or data on rates of inactivation over the 5-week refrigeration period.
4. Now take a look the authors’ Table S1 below. The authors report that the cow milk samples obtained from an affected herd in New Mexico were tested by quantitative RT-PCR (qPCR) for two viral RNAs, the influenza virus matrix (M) and hemagglutinin (HA). Note that determining infectivity of PCR-positive raw milk samples requires additional testing, typically in animal models or tissue culture cells. Milk samples were inoculated into embryonated chicken eggs and Madin-Darby canine kidney (MDCK) cells to test for infectivity. Note that this table does not address whether virus amplification was confirmed in one or both of these assays. Three of 11 samples provided estimates of viral titer above the limit of detection, and three samples failed tests for infectivity. Eight samples reportedly amplified intact infective virus.
5. The authors also reported detection of virus in mammary glands of 2 of 5 mice and conclude raw milk positive for H5N1 poses a risk when consumed untreated.
Now, let’s examine other lines of evidence that influence interpretation of the results of the ‘correspondence’.
Regarding points 1 and 2 above, a 2023 study (Effects of Extended-Release Buprenorphine on Mouse Models of Influenza) states the following caution:
‘Analgesia is rarely used’ in mouse models of influenza because of ‘concerns that [such] treatment might have confounding effects on primary study parameters such as the course of infection and/or the serological response to infection.’
We do not know if anesthesia or problems with inoculating technique or just that mice are obligate nose-breathers might have contributed to spreading inoculum into the nasal cavities of the five mice. What is apparent is that respiratory infection, not GI infection, developed.
This ‘correspondence’ does NOT demonstrate oral infectivity of H5N1 in the five mice.
Similarly, a recent review entitled Animal Models for Influenza Research: Strengths and Weaknesses noted limited utility for mouse models in the study of influenza A virus transmission, due to differences between mouse and human immune systems, and differences in toll-like receptors, defensins, and antibody classes (IgGs). Also, replication of Influenza A viruses occurs in the lower respiratory tract in mice, unlike replication predominantly in the upper respiratory tract of humans and ferrets. These researchers note that experiments with small groups of animals (5 is a small group) are subject to misinterpretation of results that may not sufficiently characterize transmission and pathogenicity in the animal model. The 5-mouse ‘correspondence’ seems grossly insufficient to permit any extrapolations to humans, in my view, without definitive data on multiple exposure pathways.
Regarding points 3 and 4, antiviral activity was demonstrated for raw milk, though raw data and the timeline for the 2-log reduction in viral titer in raw milk samples was not provided. It is possible that for the 8 of 11 samples below the detection limit in Table S1, natural antiviral activity could eliminate infectivity in the egg and canine tissue culture assays. We do not know if any of the remaining 10 raw milk samples in Table S1 with titers of 0.2 million pfu/mL (#119), 500 pfu/mL (#90), or the 8 samples below the limit of detection would cause respiratory infections in mice that were observed with inoculation of the sample with the highest titer (60 million pfu/mL, #93) in the 5 mice for this ‘correspondence’.
Point 5 raises more questions than it answers.
The observation of viral RNA in mammary glands of 2 out of 5 inoculated mice is puzzling. Were mammary tissues assayed by PCR and not for infectivity? If so, the ‘correspondence’ does not document presence of infective virus in mouse mammary tissue.
The conclusion that raw milk positive for H5N1 poses a risk when inhaled by mice has no bearing on making scientific inferences for extrapolation to as yet only ‘theoretical’ risk to humans who consume or inhale raw milk from affected cows.
At present, the only transmission pathways demonstrated for the current H5N1 in humans are respiratory (e.g., 455 total deaths among 862 human cases in China, Egypt, Indonesia, and Vietnam from 2003-2009) and ocular (2 mild cases of conjunctivitis in US dairy workers in 2024). Remember that influenza A is NOT among the 7 classes of viruses that infect the gastrointestinal system and cause intestinal flu (i) adenovirus; ii) astrovirus; iii) enterovirus; iv) hepatovirus; v) norovirus; vi) rotavirus; and viii) sapovirus).
Feel free to comment or to raise more questions about the ‘correspondence’ and about how data from animal models can be reliably extrapolated to assess risks for humans.