The detection of highly pathogenic A/Goose/Guangdong/1/96 (GsGD) A(H5N1) influenza viruses in Southern China in 1996 has been followed by two and a half decades of virus evolution and geographic expansion and contraction, resulting in significant economic loss and the depopulation of billions of poultry1. Over time, viruses of the GsGD H5-lineage have evolved into several transient, phylogenetically distinct hemagglutinin (HA) clades2. Genotype turnover driven by virus reassortment has been more sporadic but has occurred frequently in clade 2.3.4.4 viruses specifically. It is most apparent in the generation of A(H5Nx) viruses possessing different neuraminidases (NA), most frequently N1, N6, or N8. Although there have been expansions and contractions of GsGD H5-lineages over time, the clade 2.3.4.4b viruses have expanded since 2020. In the 4 months prior to February 2022, 30 countries or territories across Asia, Europe, and Africa reported the detection of viruses of this clade in birds3. Most of these 2.3.4.4b outbreaks had been caused by A(H5N1) viruses, with the noticeable exception of the outbreaks in China, where the dominant form was A(H5N6), which had also been associated with human infections4.
In December 2021, A(H5N1) viruses were detected in poultry and a gull in Eastern Canada. The virus responsible was closely related to 2.3.4.4b viruses identified in Europe in the spring of that year5. This was only the second time that a GsGD H5-lineage virus was detected in birds in the Americas; the first such occurrence was in 2014, when the infected birds had crossed the Bering Strait and entered the Pacific flyway. After the initial detection in 2021 in Newfoundland, infected wild birds were reported in the Atlantic states of the United States6 and soon thereafter in several other states. The purpose of this study was to understand the course of the genetic and phenotypic evolution of the 2.3.4.4b viruses as they spread throughout North America.
An initial risk assessment was conducted using the gold-standard ferret (Mustela furo) model of influenza to determine the pathogenic and transmission properties of the North American viruses in mammals. Chickens were also used to assess avian transmission at the avian–mammalian interface. Both animal models were inoculated with two early isolates, A/American wigeon/South Carolina/22-000345-001/2021 (Wigeon/SC/21) and A/bald eagle/Florida/W22-134/2022 (Eagle/FL/22), which were collected in December 2021 and February 2022, respectively (Supplementary Table 1). Both viruses replicated robustly in inoculated chickens and were able to transmit to naïve contact chickens (Supplementary Fig. 1). However, neither virus transmitted from inoculated ferrets to naïve direct-contact ferrets. There was also no observed transmission from inoculated chickens to ferrets. Surprisingly, there was a dramatic difference in the severity of the disease resulting from direct infection with these two viruses in ferrets. Consistent with a recent report7, Wigeon/SC/21 infection resulted in mild disease and all of the ferrets survived (Fig. 1A). Only one of three animals exhibited weight loss (Supplementary Fig. 2A), and virus shedding from the upper respiratory tract was detected only until 5 days post inoculation (dpi) (Fig. 1B). In contrast, Eagle/FL/22 infection resulted in rapid weight loss, lethargy, and severe neurologic signs, including ataxia and hindlimb paralysis. By 7 dpi, all animals inoculated with Eagle/FL/22 had reached humane endpoints and were euthanized (Fig. 1A).
The differential pathogenicity of the two viruses was reflected in the higher nasal wash titers for Eagle/FL/22 (5.8 log10TCID50/mL) than for Wigeon/SC/21 (4.3 log10TCID50/mL) (P > 0.0001) (Fig. 1B). Ferrets inoculated with Eagle/FL/22 also had higher viral loads in turbinate, tracheal, and lung samples, and virus was observed in brain tissue at 3 dpi (mean titer, 6.2 log10TCID50/mL) and 5 dpi (mean titer, 8.2 log10TCID50/mL) (Fig. 1C), whereas Wigeon/SC/21 infection led to virus being detected primarily in turbinate and tracheal tissue, with no virus being detected in brain tissue and only trace amounts being found in lung samples (Fig. 1C).
To begin to understand the molecular determinants of the pathogenicity of Eagle/FL/22, we sequenced it and selected A(H5N1) viruses and/or samples collected from wild birds. These data revealed that reassortment event(s) among H5 clade 2.3.4.4b viruses and North American wild bird influenza viruses had occurred soon after introduction of the A(H5N1) viruses into North America (Fig. 2). Whereas Wigeon/SC/21 had the same genotype as the first viruses detected in Canada and those detected in Europe in the spring of 2021, Eagle/FL/22 had undergone reassortment across three internal genes, acquiring NAm-lineage polymerase basic 2 (PB2), polymerase basic 1 (PB1), and nucleoprotein (NP) genes (Fig. 2A). In total, we identified four different viral genotypes among the 58 viruses sequenced (Fig. 2B).
All viruses maintained the parental Eurasian-origin (EA) HA, NA, matrix (M), and nonstructural (NS) gene segments, but they had different combinations of polymerase and NP gene segments of EA or NAm lineages. In all cases except for PB2, for which two phylogenetically distinct genes were detected (Fig. 2A), the NAm gene segments were monophyletic, suggesting that a single or minimal number of reassortment events had occurred, with the resulting viruses spreading geographically. The NAm-lineage proteins had no markers that had previously been identified as being associated with increased virulence in mammalian hosts8, and all of the viruses remained antigenically homogeneous, as determined by hemagglutination inhibition (HAI) assays (Supplementary Table 2).
To assess further the correlation of virulence with genotype, we used the ferret model to pathotype four additional A(H5N1) viruses collected later in the outbreak that represented the genotypic diversity observed. These viruses were A/Fancy Chicken/Newfoundland/FAV-0033/2021 (Ck/NL/21; native EA constellation); A/Red-shouldered hawk/North Carolina/W22-121/2022 (Hawk/NC/22), with the same genotype as Eagle/FL/22; A/Lesser scaup/Georgia/W22-145E/2022 (Scaup/GA/22), with PB2, PB1, polymerase acidic (PA), and NP of NAm lineage; and A/Bald eagle/North Carolina/W22-140/2022 (Eagle/NC/22), with PB2 and NP of NAm lineage (Supplementary Table S1). In general, there was a positive correlation between the number of NAm gene segments in the virus and the clinical signs and outcomes, including the time of onset and severity of weight loss and the number of animals that ultimately succumbed to disease (Fig. 3A), with the least number of clinical signs being observed with Ck/NL/21, which has no NAm genes, and the most significant signs being produced by Scaup/GA/22, which possesses four NAm genes (Supplementary Fig. 2 and Supplementary Table 3). Peak mean viral titers in nasal wash samples were 6.0, 5.4, 3.8, and 2.9 log10TCID50/mL for Scaup/GA/22, Hawk/NC/22, Eagle/NC/22, and Ck/NL/21, respectively (Fig. 3B). All EA/NAm viruses were detected in multiple organs confirming systemic viral spread; only the native constellation virus, Ck/NL/21, was not detected outside the respiratory tract (Fig. 3C). Additionally, infection with the EA/NAm viruses Eagle/FL/22, Hawk/NC/22 and Scaup/GA/22 resulted in temperature increase and respiratory and neurologic symptoms (Supplementary Table 3).
Virulence trends were also reflected in the histopathologic examination of the upper respiratory tract (URT), lower respiratory tract (LRT), and extrapulmonary tissues of ferrets. The most virulent virus, Scaup/GA/22, produced severe necrotizing lesions throughout the URT (Supplementary Fig. 3, A and B) and LRT (Supplementary Fig. 4, panels A and B) that correlated with robust viral antigen staining. Histopathology was less severe in ferrets infected with Hawk/NC/22, which produced fewer necrotizing, well limited multifocal clusters of neuroepithelium in URT (Supplementary Fig. 3, C and D) demarcated LRT lesions (Supplementary Fig. 4, panels C and D), and this corresponded to less extensive viral antigen staining. Staining for Scaup/GA/22 antigens was abundant throughout the CNS of ferrets infected with that virus (Supplementary Fig. 5, A and B), whereas staining for Hawk/NC/22 antigens was limited to the olfactory bulb and olfactory cortex of infected animals (Supplementary Fig. 5, C and D). Eagle/NC/22 produced even less severe histopathology, with only a few small foci of virus-positive cells being observed in the URT olfactory neuroepithelium and no lesions or virus antigen staining extending into the LRT (Supplementary Fig. 3, E and F, and Supplementary Fig. 4, E and F). Finally, in ferrets infected with Ck/NL/21, both the URT and LRT respiratory epithelia were devoid of lesions or cells staining for viral antigens, with the exception of a single small focus of virus antigen–positive cells in the olfactory neuroepithelium (Supplementary Fig. 3, G and H, and Supplementary Fig. 4, G and H).
To investigate whether the pathogenicity in ferrets was mirrored in other influenza mammalian models, we determined the virulence of each virus in BALB/c mice. Viruses that caused 100% lethality in ferrets and those that had acquired increasing numbers of NAm gene segments, including Scaup/GA/22, Eagle/FL/22, and Hawk/NC/22, had the lowest 50% lethal doses (LD50s) in mice (i.e., less virus was required for lethality) at 3.8, 2.2, and 4.6 log10EID50/mL, respectively. Additionally, these viruses induced neurologic symptoms in mice, along with weight loss and virus replication in the lungs and brains. Viruses causing partial or no lethality in ferrets, including Eagle/NC/2, Ck/NL/21, and Wigeon/SC/21, had the highest mouse LD50s at >6 log10EID50/mL (Supplementary Fig. 6 and 7).
Several influenza virus characteristics are known to support mammalian infection and spread, including the receptor binding properties, the pH of HA activation, and the polymerase activity9-15. Wigeon/SC/21 and Eagle/FL/22 bound strongly to sialic acid receptors with a 3′SLN linkage to the underlying sugar, as preferred by avian viruses, but poorly to 6′SLN-linked sialic acid receptors, as preferred by human viruses. The relative strength of binding was similar among all viruses examined (Fig. 4, A and B). The pH of HA activation and inactivation segregates with host adaptation12, and the pH of HA activation for A(H5N1) viruses in avian species is 5.6–6.013. Consistent with this, both Wigeon/SC/21 and Eagle/FL/22 had a relatively high pH of HA activation of 5.8 by syncytium assay (Fig. 4C and Supplementary Fig. 8). These values were slightly higher than the pH of HA activation of the control human virus A/California/04/2009 (CA/04) (H1N1)pdm09 (pH 5.6) and substantially higher than the pH of HA activation of most human-adapted influenza viruses14. In contrast, Wigeon/SC/21 and Eagle/FL/22 had virus inactivation pH values of 4.4, substantially lower than that of the CA/04 (H1N1)pdm09 virus (pH 5.4) (Fig. 4C). A virus inactivation pH that is lower than the HA activation pH is rare but has been observed with some swine H1 and H3 isolates and a bat A(H9N2) isolate15. The polymerase activities of Wigeon/SC/21 and Eagle/FL/22 were similar over a range of temperatures (Fig. 4D). Whereas the polymerase activities that were measured in the reporter gene assay with transiently expressed proteins showed no significant differences, we did detect differences in the replication rates of whole virus in MDCK and Calu-3 cells, with the differences between viruses being more pronounced in the latter cell type (Fig. 4E). At 48 hpi, cell cultures infected with Scaup/GA/22, Eagle/FL/22, and Hawk/NC/22 had the highest viral loads, followed by those infected with Eagle/NC/22, and then those infected with Ck/NL/21 and Wigeon/SC/21.
Protective antibody immunity to influenza virus targets the HA and NA and is an important consideration when assessing the risk posed by zoonotic influenza viruses11. Using a set of 48 human serum samples obtained from blood donors aged 18 to 46 years, we looked for evidence of cross-reactive HA and NA antibodies. As expected, we found no evidence of neutralizing HA antibodies against any of the A(H5N1) viruses. In contrast, an enzyme-linked lectin assay (ELLA) targeting antibodies to the N1 protein, which is far more conserved across influenza A viruses in general, revealed the geometric mean titers (GMTs) of antibodies against both Wigeon/SC/21 and Eagle/FL/22 to be equivalent to those of antibodies against the seasonal CA/04 (H1N1)pdm09 NA (fig. 4F). The NA proteins of the A(H5N1) viruses of clade 2.3.4.4b and those of A(H1N1)pdm09 viruses have 89.6% amino acid identity, with considerable conservation at some antigenic sites16,17.
To evaluate whether the available antiviral therapies would be effective against the A(H5N1) viruses, we examined the phenotypic susceptibility of these viruses to antivirals from two US FDA–approved classes: the NA inhibitors oseltamivir and zanamivir and the active metabolite of the PA endonuclease inhibitor baloxavir marboxil. All six viruses tested (both reassortant and native constellations) had IC50 and EC50 values comparable to those of drug-susceptible human A(H1N1)pdm09 influenza reference viruses (Supplementary Table 4). Additionally, genotypic M gene analysis revealed the absence of amino acid changes associated with reduced susceptibility to the adamantane class of drugs.
With the spread of A(H5N1) viruses throughout the United States and the detection of an infection in a human18, the increased virulence of the reassortant viruses is of considerable concern. Except for pockets of endemic clade activity in South and Southeast Asia, clade 2.3.4.4b viruses have predominated over other A(H5Nx) clades over the past 18 months. Clade 2.3.4.4b viruses have become entrenched in Asia, in Europe, and probably in parts of Africa. The latest WHO update (covering September 2021 to February 2022) reported 26 cases of 2.3.4.4b infection in humans, comprising 25 cases of A(H5N6) infection in China and one case of A(H5N1) in the United Kingdom, demonstrating the zoonotic transmission potential of these viruses3. From a public health perspective, the increased pathogenicity of the reassortant A(H5N1) viruses is of significant concern. However, this is tempered by the avian virus–like characteristics of the viruses with respect to their receptor binding preference and their pH of HA activation. These characteristics probably need to change to enable sustained human-to-human transmission, although only a few amino acid changes among various influenza proteins are needed to switch these properties during adaptation in mammals19,20. To date, a single case of human A(H5N1) infection in an individual involved in poultry culling has been detected in North America during the current outbreak18. This individual underwent isolation and oseltamivir treatment and recovered after experiencing only minor symptoms. Zoonotic transmission is likely to continue if this virus remains present in North American wild birds, and vigilance must be maintained as these birds start their southward migrations. The A(H5Nx) 2.3.4.4 viruses of 2014–2015 did disappear from North America for reasons that are unclear21. However, previous experience is unlikely to predict future events in this situation. The 2014–2015 viruses also reassorted soon after detection on the continent, but this reassortment was not associated with changes in mammalian pathogenicity22. The recently prevalent 2.3.4.4b viruses are also changing the dynamics of disease in Europe, with the potential for transition from epizootic to enzootic status23. Our data highlight how quickly things can change in a natural system, and the potential for further A(H5Nx) reassortment and phenotypic diversification will only increase as the unprecedented global distribution of these viruses broadens.