Receptor basis of unusual tissue tropism of avian influenza H5N1 clade 2.3.4.4b virus in cattle

Abstract

The unusual tropism of H5N1 clade 2.3.4.4b for cattle mammary glands, causing necrotizing mastitis without major respiratory involvement, raises critical questions about its underlying mechanisms. We conducted glycomics and linkage-specific lectin histochemistry to characterize sialic acid (SA) receptor diversity and anatomical distribution, and virus binding assays and high-resolution electron microscopy (EM) to visualize virus-receptor interactions. Cattle mammary gland exhibited an abundance of N- and O-linked SA glycans, showing a stronger binding affinity to clade 2.3.4.4b H5 than clade 2.2 H5. In contrast, the cattle trachea contained only O-linked but not N-linked SAs and showed no detectable H5 binding, indicating limited compatible influenza A virus (IAV) receptor availability in the tracheal epithelium. EM of virus-bound tissues further validated the receptor basis of H5N1 infection in cattle. As H5N1 continues infecting unusual hosts, as evidenced by the first case in sheep, our study offers a methodological framework for assessing IAV susceptibility.

Receptor-based insights explain necrotizing mastitis in cattle caused by H5N1 clade 2.3.4.4b.

INTRODUCTION

On 25 March 2024, a multistate outbreak of the highly pathogenic avian influenza (HPAI) A (H5N1) clade 2.3.4.4b virus genotype B3.13 was reported in dairy cows in the United States (US), marking the first documented occurrence of HPAI virus outbreaks in cattle. As of 21 April 2025, 1021 dairy herds were affected with H5N1 infection across 17 states, with California reporting the highest number of cases (1). Infected lactating cows were found to shed high levels of virus through their milk, substantially elevating the risk to raw milk consumers (2). Although most US adults drink pasteurized milk, 4.4% reported raw milk consumption in the past year and 1.6% at least monthly (3). As of 19 March 2025, 70 US H5N1 cases have been confirmed—41 linked to infected dairy cattle and 26 to infected poultry (4). In response, the Food and Drug Administration reinforced the pasteurization guidance and the US Department of Agriculture (USDA) launched the National Milk Testing Strategy (5, 6).

Recently, a new spillover event involving the H5N1 D1.1 genotype was identified in cattle in Nevada through the USDA’s silo testing program (7). This detection highlights the continued spillover of HPAI H5N1 infections from wild birds to dairy cattle, underscoring the risk of viral adaptation and potential reassortment. The D1.1 genotype is particularly concerning, as it has been linked to severe human infections in previous outbreaks, including a fatal case in Louisiana (8) and a severe pediatric infection in British Columbia (9). The presence of D1.1 in cattle suggests the virus’s evolving adaptation to mammalian hosts and raises serious concerns about the risks to farm workers and the broader public through direct contact and raw milk consumption.

HPAI H5N1 infections in domestic poultry usually cause rapid disease progression with high mortality and respiratory, neurological, and gastrointestinal signs (10). H5N1 clade 2.3.4.4b viruses have also infected many mammalian hosts, often with systemic disease and pronounced neurotropism, especially in red foxes and cats (11). In contrast, infections in cattle present a distinct profile, with necrotizing mastitis as the predominant feature and little to no respiratory involvement. Experimental oronasal inoculation of calves with the H5N1 clade 2.3.4.4b virus B3.13 genotype resulted in moderate nasal replication and shedding without severe clinical manifestations or transmission to sentinel calves, underscoring the virus’s limited respiratory involvement. However, infection of dairy cows in the same study led to severe acute mammary gland pathology, characterized by high fever, a marked reduction in milk production, and marked physical decline. Notably, the virus reached extraordinarily high titers in milk (up to 10⁹ median tissue culture infectious dose/ml) without causing systemic infection, suggesting that the mammary gland is the primary site of viral replication (12). Although rarely fatal, necrotizing mastitis causes fibrosis, reduced milk yield, and often culling, amplifying economic losses. This unusual tissue tropism, manifesting as necrotizing mastitis in cattle without concurrent respiratory illness, starkly contrasts the known respiratory tropism of HPAI H5N1 viruses in most mammalian species, including humans. Recent receptor-binding studies using recombinant clade 2.3.4.4b H5 hemagglutinins (HAs) have demonstrated preferential binding to bovine mammary gland tissue with minimal binding to tracheal epithelium, supporting mammary gland receptor availability (13). However, the detailed glycan composition, receptor class, and structural basis underlying this tissue specificity remain incompletely defined. Understanding the receptor basis for this unique clinical outcome is critical for disease management and zoonotic risk mitigation.

Notably, in a 2008 study of experimental infection of calves with an earlier clade virus (H5N1 A/cat/Germany/R606/2006), calves exhibited no clinical signs and minimal viral shedding through nasal swabs (14). Experimental infection of mice and ferrets with a bovine-derived HPAI H5N1 virus (A/dairy cattle/New Mexico/A240920343-93/2024) revealed robust replication in both the upper and lower respiratory tracts, along with systemic dissemination to nonrespiratory organs, including the eyes, brain, colon, liver, spleen, kidney, heart, and even the mammary glands. Notably, a similar broad tissue tropism was observed in mice and ferrets infected with an earlier clade H5N1 isolate (A/Vietnam/1203/2004, VN1203-H5N1) (15). These findings suggest that while systemic spread to multiple organs, including the mammary glands, is not unique to the bovine H5N1 virus, the exclusive mammary tropism with minimal or no respiratory involvement remains a distinctive feature of H5N1 infections in cattle and is not observed in other species.

A key distinction in receptor tropism among influenza viruses is their preference for specific sialic acid (SA) linkages. Influenza A and B viruses recognize sialylated glycans presented on multiple scaffold types, including N-linked glycans, O-linked glycans, and glycosphingolipids, with receptor usage varying by strain and host adaptation (16). In contrast, influenza C and D viruses use 9-O-acetylated SAs as receptors (17). It is also important to distinguish between the N-acetylneuraminic acid (Neu5Ac) and N-glycolylneuraminic acid (Neu5Gc) forms of SA. Influenza A viruses (IAVs) preferentially recognize Neu5Ac-containing glycans, whereas Neu5Gc-containing glycans are generally not considered functional receptors for IAVs. Commonly used plant lectins can bind both Neu5Ac and Neu5Gc, including modified or acetylated variants, which should be considered when interpreting lectin-based receptor mapping studies. The HA of avian influenza viruses (AIVs) predominantly recognizes α2,3-linked sialylated glycans across multiple glycan classes, whereas human influenza viruses preferentially recognize α2,6-linked sialylated glycans (18). Experimental studies have demonstrated that clade 2.3.4.4b H5N1 viruses exhibit the characteristic receptor binding preference of AIVs, showing a strong affinity for α2-3 receptors while binding poorly to α2-6 receptors (19). Glycan microarray and structural analyses have further shown that contemporary dairy cow–associated clade 2.3.4.4b H5 HAs display expanded α2,3-glycan binding breadth due to increased receptor-binding site flexibility, providing a molecular basis for broadened tissue tropism (20). Recent studies using lectin histochemistry (LHC) have identified α2-3 SA receptors in the cattle trachea, indicating the presence of receptors in bovine respiratory tissues, but not by itself, demonstrating productive viral binding or replication (21). However, both natural and experimental infection studies consistently point to the mammary gland as the primary site of H5N1 infection in cattle, with minimal respiratory tract involvement (12) (22).

LHC is widely used to study influenza receptors (23–27), but linkage-specific lectins detect only terminal residues and cannot distinguish between N- and O-linked SA receptors (28). Prior studies have described influenza receptors in bovine respiratory and mammary tissues using LHC (13, 21, 29–31), but this approach did not resolve whether the receptors are N- or O-linked. This limitation is important because such differences may strongly influence virus tissue tropism. Glycomics, defined as mass spectrometry (MS)–based structural characterization of released glycans, enables direct determination of glycan class, composition, and linkage context beyond lectin-based detection. Recent glycan array studies have shown that contemporary clade 2.3.4.4b H5 viruses exhibit expanded and promiscuous receptor binding, including recognition of mucin-like O-glycans, further broadening their glycan engagement profile (32). Glycomic analyses of bovine milk glycans and mammary tissue glycans have been reported previously, primarily in the context of nutrition and lactation biology (33, 34), but have not been applied in the context of influenza virus susceptibility. To our knowledge, this is the first study to combine MS-based glycomic profiling with wild-type and influenza pseudovirus binding assays, LHC, and ultrastructural imaging to define bovine mammary glycan receptor architecture in the context of influenza virus tropism. Building on this foundation, we used a multifaceted approach combining glycomics, virus binding assays (VBAs), high-resolution electron microscopy, and linkage-specific LHC to provide a detailed receptor basis for the unusual mammary tropism of avian influenza H5N1 clade 2.3.4.4b virus in cattle.

RESULTS

Cattle mammary glands exhibit abundant IAV-compatible N-linked SA glycans

MS-based glycomic profiling of cattle mammary gland luminal epithelial cells revealed a rich and diverse array of both O-linked and N-linked glycans, offering key insights into the receptor landscape that may underlie viral binding and tissue tropism. The analysis showed that O-linked glycans were overwhelmingly dominated by sialylated structures, with 94.26% classified as sialylated O-glycans, which were further subdivided into Neu5Ac- and Neu5Gc-containing species. A small fraction of undecorated O-glycans and minor populations of fucosylated and mixed O-glycans were also present (Fig. 1A, blue bar chart and donut plot).

Fig. 1. Glycan profiling, virus binding, and SA receptor distribution in bovine mammary gland.

(A) MS-based analysis of N- and O-linked glycans in bovine mammary gland tissue, highlighting the prevalence of sialylated glycans. Pie charts indicate the proportion of glycan subtypes, including N-linked glycans containing SA. (B) Binding of pS H5 clade 2.3.4.4b, pS H5 clade 2.2, and H1 to bovine mammary gland sections (60×). Mean fluorescence intensity quantification of different viruses binding to mammary gland tissue shown as a bar graph. (C) Abundant coexpression of duck α2,3-Gal (MAA II) and human α2,6-Gal (SNA) SA receptors in cattle mammary gland. (D) Minimal expression of chicken α2,3-Gal (MAA I) SA in cattle mammary glands, shown at 20× and 60×. (E) SEM images showing abundant surface binding of pS H5 clade 2.3.4.4b (pseudocolored pink) to mammary gland epithelium compared to mock control. H&E annotations: (1) secretory alveoli, (2) glandular epithelium, (3) intralobular duct, and (4) interlobular duct. SEM scale bars, 10 μm.

In contrast, N-linked glycans exhibited even greater structural complexity, with a broad distribution that included complex (31.22%), hybrid (30.10%), and paucimannose-type structures (17.14%), along with smaller proportions of mixed and fucosylated glycans. These categories are color-coded in shades of pink and beige in the bar and donut charts to reflect their relative abundance and structural class. Notably, ~31.22% of these N-linked glycans contained terminal SAs, which were further separated into Neu5Ac- and Neu5Gc-containing glycans (24.15 and 7.07%, respectively).

Because SA linkage type (α2-3 versus α2-6) cannot be resolved by our MS-based glycomics workflow, linkage-specific distribution was assessed using LHC. The abundant codisplay of α2-3 and α2-6 SA receptors in mammary tissue suggests that the bovine mammary epithelium harbors a glycan profile capable of supporting infection by a broad range of IAVs. To determine the anatomical localization and accessibility of these receptors within the bovine mammary gland, we performed LHC using linkage-specific probes. This analysis was essential to evaluate whether these receptors are positioned in epithelial regions potentially exposed to virus entry via the teat canal.

LHC using Maackia amurensis lectin II (MAA II; binds α2,3-Gal-β1,3-N-acetylglucosamine (GlcNAc) SA preferentially bound by avian influenza viruses from ducks) (35) and Sambucus nigra lectin (SNA; binds α2-6-Gal SA, preferred by human influenza viruses) revealed widespread display of both receptor types throughout the mammary gland epithelium (Fig. 1C). Merged images at both 20× and 60× magnifications demonstrated colocalization of α2-3 and α2-6 SA receptors within the luminal epithelium, glandular ducts, and intralobular alveoli, suggesting that these receptors are accessible to virus if introduced intramammarily.

Minimal staining was observed with MAA I, which binds a distinct SA-α2,3-Gal-β1,4-N-acetylgalactosamine (GalNAc) structure (35) preferentially recognized by chicken-adapted avian influenza viruses (Fig. 1D), indicating that bovine mammary tissue predominantly presents SA motifs associated with duck-adapted rather than chicken-adapted virus binding. Scanning electron microscopy (SEM) further confirmed receptor accessibility by showing abundant surface binding of pseudotyped H5 clade 2.3.4.4b virus to the mammary epithelium (Fig. 1E). These data demonstrate that both α2-3 and α2-6 SA receptors are present and anatomically exposed in the bovine mammary gland, providing a plausible route for viral entry and tissue infection by HPAI H5N1. VBAs were conducted using pseudoviruses expressing HA of highly pathogenic avian influenza virus (HPAIV) H5N1 clade 2.2, 2.3.4.4b, or human H1N1 virus. These pseudovirus binding assays preserved a native-like HA conformation, providing physiologically relevant interactions with host SA receptors. Notably, the luminal cells of the mammary gland demonstrated strong binding to pseudovirus (pS) H5 clade 2.3.4.4b compared with pS H5 parent clade 2.2 and pS H1 (Fig. 1B).

Cattle trachea lacks IAV-compatible N-linked SA receptors

Cattle tracheal epithelium displayed a glycan profile markedly different from the mammary gland, particularly in its capacity to support IAV binding. Glycomic analysis revealed a moderately diverse population of O-linked glycans, with 57.84% classified as sialylated O-glycans, which were further subdivided into Neu5Ac- (31.46%) and Neu5Gc- (27.07%) containing species (Fig. 2A). Undecorated O-glycans made up 34.99% of the O-linked pool, whereas fucosylated and mixed structures were present at much lower levels.

Fig. 2. Glycan profiling, virus binding, and SA receptor distribution in bovine tracheal tissue.

(A) MS-based analysis of N- and O-linked glycans in bovine trachea showed the absence of N-linked glycans containing SAs. Pie charts show the relative proportions of O-glycan and N-glycan subtypes. (B) Minimal binding of pS H5 clade 2.3.4.4b and H5 clade 2.2 to tracheal luminal epithelium. Arrows indicate virus-binding sites. Pseudovirus control is shown as a negative control. (C) Lack of avian α2,3-linked (MAA II) and mammalian α2,6-linked (SNA) SA receptors on cattle tracheal luminal cells. Arrows indicate the tracheal luminal region of interest (60×). (D) Minimal expression of chicken α2,3-Gal (MAA I) SA in cattle tracheal luminal cells. (E) SEM images showing minimal luminal binding to pS H5 clade 2.3.4.4b. No viral binding is observed on the visible tracheal cilia. H&E annotations: (1) cilia, (2) basal cells, (3) pseudostratified columnar epithelium, and (4) goblet cells. SEM scale bars, 10 μm.

In contrast, the N-linked glycan pool in the trachea showed no detectable sialylation, a notable finding given the known role of sialylated N-glycans in mediating influenza virus attachment. Most N-glycans were complex (58.34%) and hybrid (29.93%) types, with small proportions of undecorated and fucosylated glycans. However, the absence of terminal SA residues either in α2-3 or α2-6 linkages suggests a fundamental limitation for IAVs to bind to the cattle trachea and establish infection.

LHC was used to assess whether the bovine tracheal epithelium displays accessible SA receptors capable of facilitating IAV binding following upper respiratory exposure. Minimal staining with MAA II (α2-3 SA) and SNA (α2-6 SA) was observed in the trachea, with no detectable binding to the mucosal lining of the epithelium (Fig. 2C). At both 20× and 60× magnification, staining appeared weak and scattered, lacking the widespread surface accessibility noted in the mammary gland (Fig. 1C). Further staining with MAA I also showed limited signal in the trachea (Fig. 2D). These findings were consistent with glycomic profiling, which showed a complete absence of N-linked sialylated glycans. This reinforces that the tracheal epithelium lacks high-affinity SA receptors typically exploited by IAVs.

SEM and VBA revealed minimal to no binding of pseudotyped H5N1 clade 2.3.4.4b virus to the tracheal surface (Fig. 2, B and E). No viral particles were visible in the SEM, and the mock control appeared indistinguishable, confirming that the tracheal epithelium is not permissive to virus attachment. To further validate these findings with authentic virus, we performed complementary binding experiments using wild-type HPAI H5N1 clade 2.3.4.4b virus (A/dairy cattle/Texas/24-008749-002/2024; B3.13 genotype) under biosafety level 3 (BSL-3) containment. Consistent with the pseudotyped virus results, wild-type virus showed strong binding to bovine mammary gland luminal epithelial structures, whereas binding to tracheal luminal epithelium was minimal. Note that occasional staining was observed in deeper tracheal wall regions, including muscular layers; however, binding at the luminal epithelial surface, the anatomically relevant interface for inhaled virus exposure, remained minimal. Detection using anti-H5 HA immunohistochemistry with horseradish peroxidase (HRP)– 3,3′-diaminobenzidine (DAB) chromogenic readout confirmed preferential attachment to mammary gland tissue (fig. S3).

We also performed LHC on the nasal cavity and lungs to build a complete picture of SA distribution across the bovine respiratory tract. The nasal mucosa showed strong display of α2-3 SA receptors, particularly along the surface epithelium that would encounter inhaled virus (fig. S1A). In contrast, α2-6 SA receptors were only minimally present, primarily restricted to goblet cells and submucosal glands. The VBA showed that the mucosal lining of the nasal cavity displayed moderate binding to pS H5 clade 2.3.4.4b (fig. S1B). Both α2-3 and α2-6 SA receptors were detected in the lungs, with α2-6 predominating. This pattern differs from the commonly described human distribution, where α2-6 linkages are enriched in the upper respiratory tract and α2-3 linkages are more prominent in the lower lung (fig. S1A).

Distribution of SA receptors in the gastrointestinal tract and brain of cattle

The cattle gastrointestinal tract and brain tissues showed codisplay of both α2,3-linked and α2,6-linked SA receptors by LHC. However, in several parts of the intestines, particularly in the jejunum and colon, the villi mucosal lining showed predominant α2,6-linked SA signal with relatively lower levels of α2,3-linked SA receptors in the lamina propria (fig. S4). Human H1 pseudoviruses demonstrated moderate binding, whereas minimal attachment of H5 pseudoviruses to these intestinal tissues was observed, consistent with the abundant α2,6-linked SA receptors (fig. S5). In the central nervous system (CNS), including the midbrain, brainstem, cerebellum, and cerebrum, moderate codisplay of both α2,3 and α2,6 SA receptors was observed (fig. S6). Correspondingly, pseudovirus binding was detected across these brain regions, but it was scattered and not extensive, suggesting limited viral engagement within the CNS (fig. S7). Codisplay was also observed in other organs such as the salivary gland (fig. S1A).

DISCUSSION

Since early 2024, the US has witnessed two independent spillover outbreaks of HPAI H5N1 viruses involving dairy cattle caused by genotypes B3.13 and D1.1 of clade 2.3.4.4b (7, 36). These outbreaks represent the first documented instances of extensive, clinically apparent disease in dairy cattle caused by H5N1 viruses. Although current cases are geographically restricted, the ongoing global circulation of H5N1 clade 2.3.4.4b in wild birds and poultry raises growing concerns about the potential for similar outbreaks in other regions. Affected cows developed severe necrotizing mastitis with high viral titers in milk, marking a profound and unexpected shift in the host range and tissue tropism of H5N1. This organ-specific presentation is unique among known mammalian influenza virus infections and signals a previously unrecognized adaptation pathway with profound implications for animal health and zoonotic risk.

Using a multiplatform approach combining glycomics, linkage-specific LHC, pseudovirus binding assays, and high-resolution electron microscopy, our study shows that the bovine mammary gland harbors an abundant and diverse repertoire of both N- and O-linked sialylated glycans receptor motifs compatible with avian influenza virus HAs. Notably, we demonstrate that these receptors are anatomically accessible within the luminal epithelium and intralobular ducts of the udder, supporting efficient viral attachment at the teat canal interface and consistency with mammary-dominant infection reported in vivo. In stark contrast, the bovine tracheal epithelium lacks the sialylated N-linked glycans necessary for productive H5N1 binding, which may explain the limited respiratory involvement observed in natural and experimental infections. The predominance of receptor-compatible glycans in the mammary gland and their absence in key respiratory tissues offer a molecular basis for this unusual tissue specificity.

Similar to previous reports (13, 21, 29, 37), our lectin-based analysis showed strong SNA and MAA lectin signals consistent with abundant α2,3- and α2,6-linked SAs in the cattle mammary gland (mean fluorescence intensity quantification in fig. S2). A recent study reported that both the respiratory and mammary glands of HPAI H5N1–infected dairy cattle are rich in SA receptors, particularly the avian influenza virus–specific α2,3-Gal SA (21). However, lectin binding reflects only the presence of SA linkages and does not necessarily indicate productive viral binding, which depends on the underlying glycan structure and its accessibility. In our study, the receptor and accessibility landscape aligns with the observed disease pattern—severe, necrotizing mastitis confined to the mammary gland and minimal respiratory involvement—underscoring the fact that accessibility and structural compatibility of SA receptors are critical determinants of productive infection. Although both Neu5Ac- and Neu5Gc-containing sialylated glycans were identified in the glycomic profiles of tracheal and mammary tissues, IAVs preferentially recognize Neu5Ac-containing receptors, and Neu5Gc-bearing glycans are generally not considered functionally competent receptors for IAV binding (38, 39). Accordingly, lectin-detected sialylation and total SA abundance should not be interpreted as uniformly indicative of influenza A receptor compatibility (21).

Glycans can be covalently linked to proteins and are broadly classified as N-linked or O-linked based on the site of attachment. N-linked glycans are attached to the nitrogen atom of asparagine residues within the consensus Asn-X-Ser/Thr motif and typically feature a conserved core of GlcNAc and mannose (40). In contrast, O-linked glycans are attached to the oxygen atom of serine or threonine residues and exhibit greater structural variability depending on the initiating sugar and subsequent elongation. Sialylated N-linked glycans are well-established contributors to IAV attachment and HA-glycan interactions (41). In humans, α2,6 sialylation outside the brain is largely mediated by a specific sialyltransferase with strong preference for complex N-glycan substrates, which frequently associates α2,6 receptor motifs with N-linked glycans in airway tissues (42). Our findings are consistent with an important role for N-linked sialylated glycans in bovine mammary tissue, although detailed species-specific receptor scaffolds in cattle remain undefined. In contrast, O-linked glycans are often presented as dense mucin-type structures that can spatially restrict access to sialylated binding sites and act as a barrier to pathogen attachment (43). The predominance of O-linked, but not N-linked, sialylated glycans in the bovine trachea therefore provides a structural explanation for the minimal IAV binding observed in this tissue. Although LHC with linkage-specific lectins provides valuable spatial context for SA distribution, it does not distinguish between N- and O-linked core glycan structures (28); thus, combining LHC with glycomic analysis is essential to accurately assess the biochemical compatibility of sialylated glycans with IAV binding.

We found that both α2,3- and α2,6-linked sialylated N-glycans were abundantly displayed and anatomically exposed on the luminal surface of the mammary epithelium, conditions highly conducive for IAV binding. In contrast, the tracheal mucosal epithelium lacked sialylated N-glycans and showed minimal pseudovirus attachment, helping to explain the lack of upper respiratory tract infection. Our pseudovirus binding assays provided supportive data in line with these findings, but should be viewed as secondary evidence, given that natural and experimental infections have already established the mammary gland as the primary site of productive infection. Although sialylated receptors were also detected in the intestines and brain, viral binding in these tissues was limited and scattered. In bovine brain tissues, lectin staining revealed only moderate and regionally variable display of both α2,3- and α2,6-linked SA receptors, which likely accounts for the limited virus binding observed. This finding is consistent with prior experimental infection studies in cattle, where H5N1 viral antigen was not detected in the cerebral cortex (12). In contrast, our prior analysis of H5N1-infected cat brains demonstrated markedly stronger and more extensive receptor staining, particularly in the hippocampus, together with clear viral nucleoprotein detection, consistent with higher neurotropism in that species (44). The comparatively lower receptor abundance and distribution in cattle brain may therefore contribute to reduced CNS involvement. Continued viral circulation and adaptation could potentially alter tissue tropism over time, but present findings support limited CNS susceptibility in cattle. While the mammary gland currently is the primary site of severe infection in cattle, other organs may emerge as potential sites for viral entry, replication, or persistence as the virus adapts. Notably, HPAI H5N1 has caused severe gastrointestinal (45) and neurological symptoms (11, 44) in other mammalian hosts.

Our findings not only clarify the mechanistic basis for the unusual clinical presentation of HPAI H5N1 infection in cattle but also raise important questions about current approaches used to assess host susceptibility. By demonstrating the critical role of glycan architecture, particularly the distinction between N- and O-linked sialylation, we underscore the limitations of the long-standing reliance on LHC alone as a proxy for viral receptor analysis. In this study, lectin staining in nasal and lung tissues was interpreted together with functional VBAs. Detailed N- and O-linked glycomic profiling was limited to tracheal and mammary tissues, where the primary pathogenesis differences were observed. As the virus continues to cross species barriers, the need for more structurally resolved and anatomically contextual receptor profiling becomes clear. This is especially timely in light of the recent confirmation of H5N1 infection in a sheep in the UK, the first reported case in this species, detected through milk testing following an outbreak among captive birds on the same premises (46). This event reinforces the urgency of re-evaluating how we define and investigate IAV susceptibility in emerging hosts. Although infections in dairy cattle have thus far been confined to the US, global spread is conceivable given ongoing wild bird transmission. A key unknown is whether susceptibility varies across cattle breeds or among other bovine species, such as buffalo (Bubalus bubalis), which are predominant in many parts of the world. Our study offers a practical and proactive framework to assess susceptibility in diverse mammalian hosts, helping to identify high-risk species before wider transmission occurs, an approach with broad implications for understanding cross-species spread and preparing for future influenza outbreaks.

MATERIALS AND METHODS

Animal tissues

Formalin-fixed, paraffin-embedded (FFPE) healthy bovine tissue sections were acquired from Zyagen in San Diego, California, USA. These included lung (BP-601), cerebellum (BP-202), cerebrum (BP-210), stomach (BP-302), ileum (BP-309), jejunum (BP-308), duodenum (BP-307), and colon (BP-311). The Zyagen tissues were collected from beef cattle breeds used for meat production. Tissues were also collected at a bovine necropsy at the North Dakota Veterinary Diagnostic Laboratory (NDSU VDL) in May 2024. The NDSU VDL cattle tissue samples were collected from lactating Red Angus, lactating Black Angus, and a nonlactating beef crossbred animal. At the necropsy, a comprehensive set of tissues, including the trachea, lung, salivary gland, pancreas, midbrain, brainstem, and intestines, was fixed in 10% neutral buffered formalin. The fixed tissues underwent routine histological processing and sectioning for hematoxylin and eosin (H&E) staining and immunohistochemical analysis.

Influenza pseudovirus production

The third-generation lentiviral plasmids obtained from BEI resources (catalog no. NR-53816) were used to produce H5 and H1 pseudotyped viruses. The influenza A H5N1 HA sequences of clade 2.2 and clade 2.3.4.4b were retrieved from National Center for Biotechnology Information (NCBI; OQ967076.1 and KP209302.1), and expression constructs were designed using SnapGene software. The constructs were synthesized by GenScript, USA, and cloned into a pHCMV vector under the control of a CMV promoter. H1 expression plasmid pHCMV_FLUAV_HA was a gift from F. Zhang (Addgene plasmid no. 207273; http://n2t.net/addgene:207273; RRID: Addgene_207273, NCBI: NC_002017.1). Lentiviral helper plasmids, reporter plasmids, and plasmids encoding H5 or H1 were cotransfected into human embryonic kidney (HEK) 293T cells using Fugene6 transfection reagent (Promega, USA, catalog no. E2691). Pseudoviruses were collected 48 hours posttransfection, and titers were determined by infecting the HEK-293T cells and measuring the luciferase expression using BrightGlo reagent (Promega, USA, catalog no. E2650). The relative luminescence units (RLU) were measured using BioTek Cytation 7 Cell Imaging Multimode Reader. The pseudoviruses with infectivity titers between 10⁴ and 10⁵ RLU were used for the VBA.

Pseudovirus binding assay

A VBA was conducted using influenza A pseudoviruses (H5 clade 2.2, H5 clade 2.3.4.4b, and H1 subtype). The deparaffinized and antigen-retrieved tissue sections were incubated with 250 μl of influenza pseudoviruses at 37°C for 2 hours. The RLU of H5 clade 2.3.4.4b, H5 clade 2.2, and H1 pseudoviruses used in the VBA were between 10⁴ and 10⁵. Furthermore, the sections were blocked with inactivated goat serum (Thermo Fisher Scientific, catalog no. PCN5000) at a 1:40 dilution overnight. The sections treated with H5 were immunostained using a primary mouse monoclonal anti–influenza virus H5 HA protein (VN04-8), A/Vietnam/1203/2004 (H5N1), from BEI resources (NR-2733), while those treated with H1 were stained with anti–influenza A H1N1 HA antibody [C102 (IV.C102)] (Abcam, catalog no. ab128412), both at a 1:100 dilution for 90 min. Post–primary antibody incubation, the sections were washed and incubated with a secondary goat anti-mouse immunoglobulin G H&L (Alexa Fluor 647) antibody (Abcam, catalog no. ab150115) at a 1:800 dilution for 35 min. After three tris-buffered saline (TBS) washes, the sections were mounted in ProLong Gold antifade mountant with 4′,6-diamidino-2-phenylindole (DAPI). Following a 24-hour curing period at room temperature, the sections were imaged using a Nikon AXR Confocal Microscope System. Negative control staining, omitting the virus and primary antibody, showed minimal background.

Wild-type H5N1 clade 2.3.4.4b VBA

Virus binding with wild-type H5N1 clade 2.3.4.4b virus was performed in BSL-3 containment at Regional Biocontainment Laboratory, University of Pittsburgh, using the strain A/dairy cattle/Texas/24-008749-002/2024 (HPAI H5N1 B3.13) obtained from United States Department of Agriculture (USDA) Animal and Plant Health Inspection Service (APHIS), National Veterinary Services Laboratories (NVSL). Deparaffinized, antigen-retrieved FFPE tissue sections (mammary gland and trachea) were incubated with virus for 2 hours at 37°C, followed by washing. Virus attachment was detected using primary mouse monoclonal anti–influenza A H5 HA antibody [VN04-8; A/Vietnam/1203/2004 (H5N1); BEI Resources, NR-2733]. Signal was developed using the Dako EnVision+ HRP polymer detection system, Anti-Mouse (Agilent/Dako, catalog no.K4001), and Dako Liquid DAB+ Substrate Chromogen System (catalog no. K3468) according to the manufacturer’s instructions. Slides were mounted using Dako Glycergel mounting medium. Bright-field images were acquired using EVOS XL Core Configured Cell Imager available within the BSL-3 suite.

Lectin histochemistry

LHC was carried out on paraffin-embedded tissue sections as previously described (23). Briefly, 5-μm-thick sections were first deparaffinized in xylene and then rehydrated through a graded series of ethanol treatments. The rehydrated sections were presoaked in TBS for 5 min, followed by antigen retrieval using eBioscience immunohistochemistry (IHC) Antigen Retrieval Solution—Low pH (10×) (Thermo Fisher Scientific, catalog no. 00-4955-58), as per the manufacturer’s instructions. Sections were then blocked using goat serum (Thermo Fisher Scientific, catalog no. PCN5000) diluted 1:40 with TBS and further blocked using a Streptavidin/Biotin Blocking Kit (Vector Laboratories, SP2002), following the manufacturer’s guidelines. The sections were incubated with SNA, Fluorescein-labeled (Vector Laboratories, FL13012); MAA II, Unconjugated (Vector Laboratories, L-1260-2); and MAA I, Fluorescein (Vector Laboratories, FL13112). SNA binds specifically to SA linked to galactose via an α-2,6 linkage (α-2,6-Gal SA), while MAA II and MAA I bind to SA via an α-2,3 linkage (α-2,3-Gal SA). MAA I prefers distinct glycan structures compared to MAA II. Each lectin was used at a concentration of 10 μg/ml and incubated overnight at 4°C. Following three TBS washes, sections were treated with Streptavidin, Alexa Fluor 594 Conjugate (Thermo Fisher Scientific, catalog no. S32356) for 2 hours at 4°C to stain the lectins. Following three washes with TBS, sections were mounted using ProLong Gold antifade mountant with DAPI (Thermo Fisher Scientific, catalog no. P36931). After curing for 24 hours at room temperature, the sections were imaged with a Nikon AXR Confocal Microscope System. Negative control staining, performed by omitting the lectins, showed minimal background (fig. S8).

Fluorescence intensity quantification

Fluorescence quantification of LHC and IHC images was performed using ImageJ (v1.53). For each tissue section, fluorescence intensity was measured at five distinct sites of identical size within the image, and the mean fluorescence intensity was calculated. In tracheal and mammary gland tissues, the region of interest was specifically focused on the mucosal lining, as this area is most likely to interact with the virus. This approach ensured consistent quantification across all analyzed tissue types. Statistical analysis of mean fluorescence intensity values was conducted using one-way analysis of variance (ANOVA), followed by Tukey’s post hoc test to determine significant differences between groups.

Glycomics sample preparation and analysis

Tracheal tissue from cattle was obtained from Lampire Biologicals, while fresh mammary gland tissue was sourced from a local meat processing plant. For both tissue types, samples were collected from three distinct sites and processed for MS analysis. Samples were processed using the Protifi S-trap protocol (https://files.protifi.com/protocols/s-trap-96-well-plate-long-1-4.pdf) with modifications for glycomics (47). Samples were homogenized in 100 μl of lysis buffer [5% SDS (w/v; Sigma-Aldrich, catalog no. 75746) and 50 mM triethylammonium bicarbonate (TEAB) (Sigma-Aldrich, catalog no. T7408), pH 7.1] using probe sonication and subsequently denatured with 10 mM dithiothreitol (Sigma-Aldrich, catalog no. D0632) to facilitate glycan processing. Denatured proteins were acidified by addition of 1.2% phosphoric acid (Sigma-Aldrich, catalog no. 49685), diluted with 1.5 ml of S-trap binding buffer [90% methanol (Thermo Fisher Scientific, catalog no. A456) and 100 mM TEAB, pH 7.1], and loaded onto S-trap (Protifi, catalog no. P002-96MINI). S-trap was centrifuged (Eppendorf, Centrifuge 5810 R) at 1500g for 2 min, and wells were washed with an additional 200 μl of binding buffer three times. N-glycans were enzymatically cleaved from proteins using 1 μl of PNGase F (New England Biolabs, catalog no. P0710S) in 20 μl of 10 mM ammonium bicarbonate (Sigma-Aldrich, catalog no.A6141) and incubated at 37°C overnight. Released N-glycans were eluted from S-trap with 60 μl of 0.1% trifluoroacetic acid (TFA; Sigma-Aldrich, catalog no. T6508) by centrifuging at 1500g for 2 min. Eluant was then purified using hypercarb solid-phase extraction (Thermo Fisher Scientific, catalog no. 03250409), and desalted N-glycans were eluted with 50% acetonitrile (Sigma-Aldrich, catalog no. 34851) with 0.1% TFA. Following N-glycan removal, O-glycans were released through reductive β-elimination by adding 40 μl of 50 mM KOH (Sigma-Aldrich, catalog no. P4494) with 1 M NaBH₄ (Sigma-Aldrich, catalog no. 71320) overnight at 56°C. Following incubation, O-glycans were eluted with 0.1% TFA and purified as described above for N-glycans using hypercarb. N- and O-glycans were dried to completion by speedvac (Eppendorf, Vacufuge plus) and resuspended in 30 μl of Milli-Q water. The purified N- and O-glycans were analyzed using porous graphitized carbon (Thermo Fisher Scientific, catalog no. 35003-101030) liquid chromatography coupled with tandem MS on a Thermo Fisher Scientific TSQ Altis Mass Spectrometer (model no. TSQ03-10002) coupled to a Thermo Fisher Scientific Vanquish liquid chromatography system (model no. VH-P10-A-02) for glycan profiling and structural characterization (48, 49). Data analysis was performed using Skyline software and manual interpretation (50).

Electron microscopy

The deparaffinized and antigen-retrieved tissue sections were incubated with 250 μl of influenza pseudoviruses at 37°C for 2 hours. Postincubation, the slides were fixed with Karnovsky’s Fixative (Electron Microscopy Sciences, catalog no. 15720). Samples were characterized using a SEM (JEOL 7800F) operated at an accelerating voltage of 5 kV. Before imaging, specimens were mounted on aluminum stubs using conductive carbon tape and sputter-coated with a thin layer (~1 nm) of gold/palladium alloy to minimize charging effects and enhance surface conductivity. The scale bar on the SEM images was set at 10 μm.

Supplementary Materials

This PDF file includes:

Figs. S1 to S8