Beyond H5N1: influenza A virus infection in bovine udder organoids

ABSTRACT

Influenza A viruses (IAVs) have recently expanded their host range to bovine species, raising critical questions about tissue tropism, viral replication capacity, and impacts on dairy production and zoonotic risk. Extensive shedding of highly pathogenic avian influenza virus (HPAIV) H5N1 in milk from naturally infected cattle in the USA highlights the bovine mammary gland as a relevant site of viral replication. To investigate these dynamics, we established and characterized bovine mammary gland organoids (MGOs) derived from primary epithelial cells and maintained in chemically defined media. This efficient, scalable model reconstructs structural and functional hallmarks of the bovine udder, including luminal–basal polarity, lineage-specific cytokeratin patterns, and lactation-associated differentiation. We then used this system to study low and highly pathogenic avian, as well as swine and human IAVs. Viruses from diverse host species productively infected MGOs, largely independent of culturing conditions. Mammalian origin H1N1 strains matched bovine H5N1 in efficiency, but low pathogenic avian H9N2 showed restricted replication, indicating subtype-specific differences in mammary tropism. This suitability study develops a MGO platform that provides a tractable basis for detailed molecular investigation of mammary infection biology and influenza pathogenesis in cattle, to inform strategies that protect both animal and public health.

Introduction

Influenza A viruses (IAVs) possess remarkable host plasticity and have repeatedly emerged in novel mammalian hosts, including sporadic avian-to-human events and sustained enzootic transmission in pigs [1]. The 2009 H1N1 pandemic positioned pigs as a major interspecies interface, because their respiratory receptor duality supports coinfection with avian- and human-adapted IAVs, creating conditions for reassortment and antigenic innovation [2]. Beyond swine, equines, canines, and marine mammals experience recurring IAV outbreaks [3], even when classical avian-adapted strains fail to establish long-term lineages in these hosts. Accumulating reports of highly pathogenic avian influenza virus (HPAIV) in diverse terrestrial [4] and marine mammals [5] indicate increasing selective pressure on viral and host factors that govern virulence, pathogenicity, and transmission to new host populations. In avian hosts, acquisition of a multibasic cleavage site in haemagglutinin (HA) is a key event that expands viral tropism from mucosal surfaces to systemic infection, resulting in severe and frequently fatal disease [6]. Since the emergence of the goose/Guangdong (gs/GD) lineage in 1996, wild bird migration has driven global dissemination, and HPAIV H5 is now widely regarded as panzootic in birds [7]. Recent high-mortality outbreaks of HPAIV H5N1 in commercial and wild bird populations across North America and Europe have caused mass culling of poultry and further underscore the intensity and geographic scale of ongoing activity [8,9].

The expansion of the HPAIV H5N1 host range to dairy cattle represents a host jump of substantial veterinary and public health relevance, raising concerns about cross-species transmission and adaptation beyond the classical avian reservoir [10]. Previously, cattle were primarily recognized as host species for influenza D viruses [11]. High viral loads of HPAIV H5N1 in raw milk and mastitis-like disease in naturally infected cattle in the USA point to an unexpected and underexplored tissue niche: the bovine mammary gland [1,12]. Although the respiratory tract and for avian influenza viruses also the intestinal tract, are canonical sites of influenza virus replication, these field observations in bovine hosts indicate that non-respiratory epithelia, particularly the mammary epithelium, can support extensive IAV replication with implications for animal health, milk quality, and zoonotic exposure through milk and dairy products [1,13]. Experimental H5N1 infections in heifers have reproduced mastitis-like changes, extensive virus shedding, and local inflammation [14]. The study did not identify systemic spread and extra-mammary dissemination. Collectively, these findings challenge previous notions about IAV tissue tropism in cattle. Additionally, susceptibility of bovine mammary tissue to influenza viruses was reported as early as the 1950s. Experimental studies demonstrated viral propagation in infected ruminants [15], and field observations associated IAV seropositivity with reduced milk yield in dairy cattle [16], further questioning the idea that this tissue tropism is exclusive to H5N1.

The bovine mammary gland is a specialized secretory organ composed of distinct epithelial lineages organized into a complex glandular architecture. Luminal epithelial cells synthesize and secrete milk components under the control of lactogenic hormones, while basal and myoepithelial cells provide structural support and contractile function required for milk ejection [17]. The gland undergoes dynamic structural and functional remodelling across developmental stages and lactation, reflecting tightly regulated hormonal cues and epithelial differentiation programs [1,18]. Several studies have examined sialic acid receptor distribution in cattle, demonstrating receptor duality in the mammary gland that could, in principle, support binding of both avian- and human-type viruses [19–21]. However, contemporary clade 2.3.4.4b H5N1 viruses largely retain avian-type receptor binding preference [22–25], and the molecular determinants that enable productive infection of bovine mammary tissue remain poorly defined. Progress in understanding tissue-specific IAV tropism in cattle has been constrained by the lack of physiologically relevant bovine mammary in vitro models. Organoid systems have advanced infection research in human mucosal tissues by providing scalable, polarized, and functionally differentiated epithelial structures [26]. However, the development of mammary models for virological questions remains at an early stage [27], and in vivo infection studies require housing large ruminants under biosafety level 3 (BSL-3) conditions, limiting experimental throughput and global accessibility.

To address this gap, this study describes the development, cultivation, and characterization of a continuously cultured bovine mammary gland organoid (MGO) model. Morphological and molecular analyses reveal key features of mammary epithelial plasticity and heterogeneity, mirroring the physiological organization of the bovine udder. This platform was used to explore IAV susceptibility, replication competence, and functional consequences in mammary epithelium across a panel of phylogenetically and phenotypically diverse strains. Alongside contemporary HPAIV H5N1, this study included low-pathogenic avian influenza virus (LPAIV) subtypes H7N7 and H9N2, as well as mammalian-adapted strains: swine-origin H1N1 and human pandemic H1N1, which are well established in non-avian hosts and associated with documented interspecies transmission and substantial public health impact [28]. Testing these diverse viruses in bovine MGOs enables comparative analysis of strain-specific infection dynamics and their impact on lactation-associated gene expression within a controllable microenvironment. This work establishes bovine MGOs as a tractable platform for studying influenza ecology and supporting infection research in line with 3R principles and biosecurity requirements. This complementary in vitro model enables controlled, tissue-specific ex vivo investigations with direct implications for animal health, milk production, and possibly zoonotic exposure.

Materials and methods

Cells & viruses

Madin-Darby canine kidney 2 (MDCK.2) cell line was routinely cultured in Dulbecco's Modified Eagle Medium (DMEM) (Gibco Thermo Fisher, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS) (Capricorn Scientific, Ebsdorfergrund, Germany) and 100 U/mL penicillin and $10\mu \mathrm{g}/\mathrm{mL}$ streptomycin (Sigma-Aldrich Merck, Burlington, MA, USA). The L-WRN cell line was kindly provided by Prof. Guntram Graßl (Institute of Medical Microbiology and Hospital Hygiene, MHH, Hannover, Germany) and cultured in DMEM + GlutaMAX™ (Gibco Thermo Fisher) containing 10% FBS. Cell lines were originally obtained from American Type Culture Collection (ATCC) and routinely tested for mycoplasma contamination. Serum pools were routinely tested for bovine viral diarrhea virus, bovine herpesvirus-1, and bovine parainfluenzavirus contamination through virus isolation and were heat-inactivated at $60{}^{\circ }\mathrm{C}$ for 30 min.

This study used previously characterized influenza A viruses including LPAIV strains of subtypes H7N7 (A/duck/Potsdam/15/80) and H9N2 (A/chicken/Saudi Arabia/CP7/98), as well as H1N1 subtypes of swine (A/swine/Bad Griesbach/IDT 5640/06) and human (A/Hamburg/1580/09) origin, hereafter referred to as H1N1 06 and H1N1 09, respectively [29]. Cell culture supernatant containing HPAIV H5N1 strain (A/cattle/Texas/063224-24-1/2024; clade 2.3.4.4b, genotype US B3.13) [30] was kindly provided by Timm Harder (National Reference Laboratory for Avian Influenza, Institute of Diagnostic Virology, Friedrich-Loeffler-Institut, Greifswald – Insel Riems, Germany) with permission of Diego Diel (Cornell University, College of Veterinary Medicine, Ithaka, NY, USA) and stocks were propagated in MDCK.2 cells under certified BSL-3 containment. Viral supernatants were clarified by centrifugation, aliquoted, and stored at $-80{}^{\circ }\mathrm{C}$.

Isolation of bovine mammary epithelial cells

Fresh udder samples were obtained from two dairy cows euthanized at the university's cattle clinic due to a lack of prospects of recovery from multiple arthritis and tendonitis (animal A) or a humerus fracture (animal B). Ethical approval was not required as udder samples were obtained from animals euthanized for clinical reasons unrelated to this study, and tissue was collected as part of routine post-mortem procedures under applicable institutional regulations. Parenchymal tissue was washed with phosphate-buffered saline (PBS), minced, and digested overnight in enzymatic digestion medium (Table S1). The suspension was filtered through a $100\mu \mathrm{m}$ cell strainer and centrifuged at $800\times \mathrm{g}$ for 5 min. The resulting cell pellet was cryopreserved in FBS containing 10% DMSO and stored at $-80{}^{\circ }\mathrm{C}$. Tissue samples were screened for the same pathogens as the serum pools by RT-qPCR.

Culture of mammary gland organoids

Cryopreserved primary cells were thawed in DMEM/F12 + GlutaMAX™ containing 10% FBS and cell pellets were resuspended in cold sphere formation medium (SFM) (Table S1). Cells were embedded in matrigel (Corning, Corning, NY, USA) domes in 24-well plates and cultured in SFM with medium changes three times per week. Organoids were passaged every 10–14 days by enzymatic dissociation with trypsin-EDTA, followed by mechanical shearing and re-embedding in fresh matrigel. The growth of the forming structures was documented at regular intervals using a Nikon Eclipse Ti-S inverted microscope (Nikon Corporation, Tokyo, Japan) and quantified in ImageJ.

For differentiation experiments, cultures were transferred from SFM to recombinant differentiation medium (RDM) from day 1 to day 14. Lactogenic stimulation was induced by incubation in lactation medium (LM) (Table S1) for 2–4 days, followed by withdrawal to assess hormonal reversibility. Gene expression changes were evaluated by RT-qPCR of milk-associated transcripts from purified cellular RNA.

Influenza A virus infection study

Viral growth kinetics of the individual viral strains were determined using both expanding and differentiated organoids. Cell counts for multiplicity of infection (MOI) calculation were performed on trypsin-dissociated organoids one day prior to infection as a proxy for cell numbers in domes used the following day. Organoids were released from matrigel and infected with influenza viruses at an MOI of 0.1 for 1 h at $37{}^{\circ }\mathrm{C}$. Following infection, organoids were washed five times with PBS, re-embedded in matrigel, and cultured in minimal mammary medium (MMM) (Table S1). For low-pathogenic IAV strains, $1\mu \mathrm{g}/\mathrm{mL}$ L-(tosylamido-2-phenyl) ethyl chloromethyl ketone (TPCK) treated trypsin (Thermo Fisher) was used throughout. Samples were collected at 2, 24, 48, and 72 h post infection (hpi). Supernatants were stored at $-80{}^{\circ }\mathrm{C}$ for virus titration, while cell pellets were lysed for RNA extraction. Infections were performed in triplicate for each strain, donor, and time point.

RNA isolation and reverse transcription quantitative PCR

Cellular RNA was prepared using RNeasy Mini (Qiagen, Hilden, Germany) or NucleoSpin RNA kits (Macherey-Nagel, Düren, Germany) with on-column DNase digestion. Influenza A virus genome copies were analyzed by reverse transcription quantitative PCR (RT-qPCR) on CFX96 platform (Bio-Rad) using Probe RT-qPCR kit (Qiagen) and previously described oligonucleotides targeting the nucleoprotein (NP) gene [31]. Gene expression of lactation markers was analyzed using SYBR Green RT-qPCR kit (Qiagen) using specific oligonucleotides from previous studies (Table S2).

Virus titration

Infectious viral particles were quantified using end-point dilution assay [32]. Confluent MDCK.2 cells in 96-well plates were washed with PBS, and ten-fold serial dilutions of supernatants were prepared in DMEM containing antibiotics and $2\mu \mathrm{g}/\mathrm{mL}$ acetylated trypsin (Sigma-Aldrich Merck) (Except for HPAIV). A volume of $100\mu \mathrm{L}$ of each dilution was added to quadruplicate wells. After 48–72 h, individual wells were examined for cytopathic effects and TCID₅₀ values were calculated using the Spearman-Kaerber method. Productive replication was defined as a ≥10-fold increase in infectious titre between 2 and 48 hpi. Strains failing to meet this threshold were classified as showing restricted or abortive infection.

Histology and immunostaining

Organoids were extracted from the hydrogel, washed, and fixed in 4% PFA (Sigma-Aldrich Merck) for 45 min at $4{}^{\circ }\mathrm{C}$. Subsequently, the organoids were embedded in 4% agarose (Biozym, Hessisch Oldendorf, Germany) and stored in 70% ethanol (Carl Roth, Karlsruhe, Germany) until they were further processed, sectioned, and stained, following previously described procedures [33,34]. Antibodies used for molecular analysis are listed in Table S3. Olympus CKX53 microscope was used for documentation of histology slides with cellSens software (version 4.4, Evident Corporation, Tokyo, Japan).

The protocol from Dekkers et al. [35] was adapted for whole-mount 3D staining of infected cells using an IAV NP-directed antibody. The procedure for Lectin staining is described elsewhere [36]. Fluorescence microscopy images were captured using a Nikon Eclipse Ti-S inverted fluorescence microscope and processed with Nikon NIS-Elements AR software (version 5.21.03, Nikon Corporation).

Statistics

Organoid cultures were generated from two individual donor animals, and three independent experiments per animal were analyzed. Visualizations and statistical reports were generated using GraphPad Software (version 10.4.1, GraphPad Software, Boston, MA, USA). The significance level was set to $\alpha =0.05$. Expressional data were log₂ transformed and subjected to two-way ANOVA with Šidák's multiple comparison test. For analyses of viral infection kinetics, IAV genome copies and titres were log₁₀-transformed and analyzed using a repeated-measures (RM) one-way ANOVA with Dunnett's test for H5N1 and RM two-way ANOVA with the same post hoc test for the other strains. The impact of lactogenic hormones on infection was assessed using two-way ANOVA with Šidák's multiple comparison test. Individual p-values are indicated in the figures together with the corresponding statistical tests. Donor- and strain-wise statistics are reported in Tables S4 and S5, respectively.

Results

Establishment of MGOs

The isolation protocol, adapted from a bronchial epithelial cell workflow [37], was successfully applied to mammary tissue from two euthanized dairy cows with distinct physiological backgrounds (Animal A: third lactation, early pregnancy; Animal B: first lactation, no pregnancy). Thawed cryovials contained 30–60% viable cells, with debris likely underestimating cell viability. Primary cells formed spheroid structures within two weeks of seeding, and subsequent passaging generated large organoids from single cells within 14 days (Figures 1A and S1A).

Figure 1.

Culture of bovine mammary gland organoids from primary udder epithelial cells. (A) Bright field images of MGOs after different days of development (4–14 days) under both sphere formation medium (SFM, blue) and recombinant differentiation medium (RDM, magenta) conditions derived from animal A. Scale bar: 500$\mu \mathrm{m}$. (B–D) Quantitative analysis of emerging structures: equivalent diameter, organoid size, and formation efficacy (structures/image) were determined from six domes per time point and condition. Data are presented as mean with individual data points, and the number of quantified structures is indicated (B, C), or as mean ± SD, with the number of quantified images indicated (D). SD: Standard deviation.

Subsequent optimization focussed on scalability and morphological uniformity to support infection assays. Refinements in medium formulation reduced non-uniform structural subtypes, and favoured a homogeneous spheroid morphology, consistent with maintenance of multipotent progenitor populations (Figure S1B-D). Under final conditions, passaging ratios exceeded 1:24 (initially 1:4) with an optimal seeding density of $1.25-2.0\times {10}^5$ cells/mL, yielding 2500–4000 cells per dome. Organoids began to form central lumina from day 7 onwards, indicating acquisition of apical-basal polarity (Figures 1A and S1A).

Derivation of lactogenic MGOs

To create a hormonally responsive mammary model, we developed a differentiation protocol that promotes partial lactogenic differentiation. The resulting RDM formulation was designed to drive lactogenic commitment and produced smaller, more compact structures (Figures 1A and S1A). Quantitative assessment showed a distribution of sizes with predominant large structures ($>200\mu \mathrm{m}$) under SFM and only smaller ($<200\mu \mathrm{m}$) but a slightly higher number of organoids under RDM condition (Figure 1B–D). Phenotypic characterization showed KRT-positive cells in organoids grown in both SFM and RDM. Basal KRT14-positive cells predominated under both conditions, whereas KRT18-positive cells were more prominent in RDM, indicating a bilayered epithelial lineage organization (Figure 2A). Vimentin expression increased under RDM, consistent with partial epithelial–mesenchymal plasticity during differentiation, without disrupting the overall organoid architecture (Figure 2A). Sambucus nigra agglutinin (SNA) and Maackia amurensis lectin II (MAL-II) were used to map the distribution of terminal α2,6- and α2,3-linked sialic acids, respectively. MGOs abundantly displayed glycans detectable by both SNA and MAL-II, with the latter showing a stronger signal on the abluminal side (Figure S5). The observed distribution remained consistent across media conditions.

Figure 2.

Phenotypic characterization, lactogenic differentiation protocol and induction of milk-associated genes in bovine mammary gland organoids. (A) Histological characterization of mammary gland organoids (MGOs) derived from animal A cultured for 14 days in sphere formation medium (SFM) and recombinant differentiation medium (RDM). Organoids were stained with H&E, and analyzed for the presence of KRT14, KRT18, and Vimentin. Scale bars: $20\mu \mathrm{m}$. (B) Schematic representation of the differentiation scheme. Organoids were expanded in SFM from day 0. On day 1, cultures were maintained in SFM (blue) or switched to RDM (magenta). On day 14, lactation medium (LM, green) was added for 4 days (+LM) or omitted (−LM). Samples were collected on day 18. (C) Relative normalized mRNA expression of lactation markers CSN2, LTF, and PAEP in organoids derived from animal A under indicated conditions. Data are shown as log₂ relative normalized expression compared to the SFM–LM condition and represent mean ± SEM of triplicates. Statistical significance was assessed using two-way ANOVA with Šidák's multiple comparisons test. H&E: Hematoxylin and Eosin; KRT14: Cytokeratin 14 (basal/myoepithelial marker); KRT18: Cytokeratin 18 (luminal epithelial marker); CSN2: β-casein; LTF: Lactotransferrin; PAEP: Progestagen-associated endometrial protein; SEM: Standard error of mean.

Although organoids formed multilayered epithelia in SFM, they displayed only a moderate increase in β-casein (CSN2) mRNA following LM exposure, indicating limited functional secretory differentiation (Figures 2C and S2B). Partial lactogenic differentiation was achieved using a protocol involving initial seeding in SFM, subsequent culture in RDM for 13 days, and final culture in LM for 4 days (Figure 2B). Quantitative analysis of lactation markers revealed increased CSN2 expression by nine logs (512-fold) relative to SFM without LM, while RDM alone induced intermediate CSN2 levels (Figures 2C and S2B). Lactotransferrin (LTF) showed a more modest induction (2–3 logs) in differentiated organoids relative to SFM, while progestagen-associated endometrial protein (PAEP) expression remained below baseline in all conditions (Figures 2C and S2B). These differential responses indicate non-uniform regulation of milk-associated genes during in vitro lactogenic differentiation. Withdrawal of LM from RDM-cultured organoids reduced CSN2 expression, consistent with an involution-like phenotype and underscoring the hormonal plasticity of the system. Organoids from Animal A showed slightly higher induction of lactogenic markers and stronger mesenchymal plasticity but expanded somewhat less efficiently than those from Animal B (Figures 2 and S2; Table S4c). Overall, the described protocol is optimized for MGO culture establishment and differentiation, accounting for distinct physiological backgrounds. Luminal-basal polarized MGOs were successfully generated from unpurified epithelial suspensions, with transcriptional evidence of hormone-responsive secretory activity suitable for downstream infection studies.

Viral replication of bovine HPAIV H5N1 in bovine MGOs

To assess the replication capacity of bovine HPAIV H5N1 in mammary tissue in vitro, organoids from two donor cattle (animal A and animal B), cultured in SFM or RDM, were infected with HPAIV H5N1 (A/cattle/Texas/063224-24-1/2024) at an MOI of 0.1 (Animal A: Figure 3; Animal B: Figure S3). Across both donors and media, the initial infection showed $\sim {10}^4$ normalized IAV genome copies per reaction at 2 hpi and increased steeply over time (Figures 3A and S3A). Viral RNA increased by roughly four orders of magnitude within 24 h and peaked near ${10}^8-{10}^9$ copies per reaction at 48 hpi in SFM and RDM. Analysis of infectious virus production closely mirrored the kinetics of genome replication (Figures 3B and S3B). In organoids from Animals A and B, titres were close to the detection limit at 2 hpi and increased over time to reach $\ge {10}^5$ TCID₅₀/mL by 48 hpi in both SFM and RDM, confirming efficient production and release of progeny virions. Immunofluorescence staining at 48 hpi corroborated these findings: NP antigen was detected in most cells within individual organoids under both media conditions and in both donors, whereas mock-infected organoids remained NP negative and preserved a compact morphology (Figures 3C and S3C). We then tested whether differentiation status or lactogenic hormone supplementation modulated H5N1 replication. Comparison of SFM versus RDM showed highly similar viral RNA trajectories, endpoint titres, and NP antigen distributions for both animals, indicating that the degree of lactogenic differentiation did not significantly alter the permissiveness to H5N1 (Figures 3A–C and S3A–C). Likewise, pre-exposure to lactation medium (+LM) or culture without LM (−LM) had little impact on virus output: in organoids from both donors, titres at 48 hpi reached approximately ${10}^5-{10}^6$ TCID₅₀/mL in +LM and −LM conditions (Figures 3D and S3D).

Figure 3.

Viral infection kinetics of bovine HPAIV H5N1 in expanding and differentiated mammary gland organoids. (A–C) Permissiveness of cultured organoids from animal A to contemporary H5N1. Organoids cultured in either sphere formation medium (SFM) or recombinant differentiation medium (RDM) were infected with bovine HPAIV H5N1 at an MOI of 0.1. (A) Viral RNA copies (log₁₀ GAPDH-normalized IAV copies/reaction) were quantified at 2, 24, and 48 hpi. (B) Infectious titres of the supernatants at 2, 24, and 48 hpi were determined on MDCK.2 cells. Statistical significance was assessed by using RM one-way ANOVA with Dunnett's multiple comparisons test, comparing each time point to 2 hpi within each medium. (C) Immunofluorescence images show NP-positive cells (red) and nuclei (blue) at 48 hpi in H5N1-infected organoids. Scale bar as indicated. (D, E) Differentiated organoids were infected with bovine HPAIV H5N1 under lactation medium (+LM) or without lactation medium (−LM). (D) Corresponding infectious virus titres (log₁₀ TCID₅₀/mL) and (E) relative normalized CSN2 mRNA expression were quantified. Statistical significance was assessed by using two-way ANOVA with Šidák's multiple comparisons test. Data are presented as mean ± SEM of triplicates. CSN2: β-casein; LOD: Limit of detection; SEM: Standard error of mean.

Finally, we analyzed the effect of H5N1 infection on the key milk protein marker CSN2. In organoids from Animals A and B, CSN2 mRNA levels at 48 hpi remained close to mock-infected controls under both +LM and −LM conditions, with no significant differences between infected and mock groups despite high viral titres (Figure 3E and S3E). Collectively, these results show that bovine HPAIV H5N1 efficiently infects MGOs from multiple donors, achieving high viral RNA loads and infectious titres across different media and hormone conditions, while leaving CSN2 transcript levels largely unchanged in this system. Together, these data suggest that the model captures key aspects of in vivo permissiveness of the bovine mammary gland to H5N1.

Viral replication of LPAIV, swine- and human-origin IAV strains in bovine MGOs

To determine whether bovine MGOs support replication of influenza A viruses beyond H5N1, organoids from two donor cows were infected with LPAIV H9N2 and H7N7, swine origin H1N1 06, and human pandemic H1N1 09 under SFM and RDM conditions, respectively (Figures 4 and S4). All influenza viruses showed a marked time-dependent increase in normalized viral IAV genome copies over several orders of magnitude (Figures 4A and S4A). For the LPAIV strains, peak genome copies were at least 5-fold lower than those of the H1N1 subtypes. H9N2 showed a more restricted temporal progression, resulting in a significantly reduced viral genome replication (Figures 4A and S4A; Table S5). Infection with the H1N1 subtypes produced a maximum of $>{10}^8$ normalized IAV genome copies/reaction, similar to the level of H5N1 replication. In LPAIV H9N2-infected organoids, virion quantification remained near the detection threshold at all time points, consistent with an abortive infection in this model. Conversely, LPAIV H7N7 infection resulted in titres of ${10}^4$ TCID₅₀/mL in both SFM and RDM conditions (Figures 4B and S4B). Swine H1N1 06 and human H1N1 09 reached infectious virus titres comparable to H5N1 (Figures 4B and S4B; Table S5). Immunofluorescence analysis at 48 hpi supported these differences: organoids infected with H7N7, H1N1 06, and H1N1 09 showed extensive NP positive regions, whereas H9N2 infected organoids displayed only sparse or absent NP staining, and mock controls remained NP negative (Figures 4C and S4C).

Figure 4.

Viral infection kinetics of influenza A viruses in expanding and differentiated mammary gland organoids. (A–C) Permissiveness of cultured organoids from animal A to four IAV strains (H9N2, H7N7, H1N1 06, H1N1 09). Organoids cultured in either SFM or RDM were infected with each strain at an MOI of 0.1. (A) Viral RNA copies (log₁₀ GAPDH-normalized IAV copies/reaction) were quantified at 2, 24, 48, and 72 hpi. (B) Infectious titres of the supernatants at 2, 24, 48, and 72 hpi were determined on MDCK.2 cells. Statistical significance was assessed using two-way ANOVA, with Dunnett's test comparing each time point to 2 hpi within each strain and medium. (C) Immunofluorescence images show NP-positive cells (red) and nuclei (blue) at 48 hpi in IAV-infected organoids. Scale bar as indicated. (D, E) Differentiated organoids were infected with IAV strains (H7N7, H1N1 06, H1N1 09) under lactation medium (+LM, left) or without lactation medium (−LM, right). (D) Corresponding infectious virus titres (log₁₀ TCID₅₀/mL) and (E) relative normalized CSN2 mRNA expression were quantified. Statistical significance was assessed by using two-way ANOVA with Šidák's multiple comparisons test. Data are presented as mean ± SEM of triplicates. CSN2: β-casein; LOD: Limit of detection; SEM: Standard error of mean.

We next assessed whether lactogenic conditions modulate infectious output and CSN2 expression for these strains. In organoids from both donors, H7N7, H1N1 06, and H1N1 09 reached similar endpoint titres under +LM and −LM, with the highest titres (up to ${10}^5-{10}^6$ TCID₅₀/mL) observed for the H1N1 strains under −LM conditions (Figures 4D and S4D). Analysis of CSN2 mRNA at 72 hpi showed that infection under lactogenic conditions (+LM) had little effect on relative CSN2 expression. Under non-lactogenic conditions (−LM), infection with H1N1 06 and H1N1 09 led to a pronounced reduction in CSN2 expression, and H7N7 caused a milder decrease, although this effect was not consistently observed across both animals (Figures 4E and S4E).

LPAIV infection in the absence of exogenous trypsin (Figure S6) demonstrated that both H7N7 and H9N2 require trypsin for productive replication, with H7N7 titres significantly reduced under trypsin-absent conditions, while H9N2 remained restricted regardless of trypsin availability. This strongly suggests that the restricted replication of H9N2 reflects fundamental susceptibility rather than trypsin dependence. Mock-infected organoids showed no significant difference in CSN2 expression with or without TPCK-trypsin at 2 or 72 hpi (Figure S7), confirming that trypsin does not impair organoid differentiation. Collectively, these data demonstrate that bovine MGOs from independent donors support robust replication of avian H7N7, swine origin H1N1 06, and human pandemic H1N1 09.

Discussion

Three-dimensional reconstruction of mucosal interfaces has transformed animal-sparing infection research through precision-cut organ slices, air-liquid interface (ALI) cultures, and organoids, which are frequently used in veterinary virology [37–39]. Explant cultures have been applied to the mammary gland, capturing both parenchymal and stromal elements of tissue architecture. Yet this model can only reliably address infection of ductular structures, as functional alveoli were involuted ex vivo, and thus not accessible to infection, preventing conclusions about the true secretory compartment [12]. Primary cell–derived models are more accessible but often lack structural and functional fidelity, as conventional 2D cultures do not recapitulate the three-dimensional architecture, lineage hierarchy, or functional differentiation that define the native gland [21,27]. The complexity of the udder's epithelial organization implies the necessity of extensive characterization and suggests high variability dependent on the cultured cell population and its dynamics. Against this backdrop, we established a bovine MGO model that supports epithelial reconstruction and preserves luminal-basal-specific signatures and hormone-responsive CSN2 expression through a selection-based approach. Because most MGO research focuses on mouse and human tissues and neglects infection biology, we wanted to pivot the bovine MGOs toward infection research by exploring IAV tropism and its functional consequences in the secretory epithelium in vitro.

The initial expansion medium favoured a dense, basal-like phenotype and generated morphologically uniform spheroids from unpurified epithelial suspensions. Minor changes in the SFM composition shifted the organoid morphology toward cystic, luminal-like structures, highlighting the sensitivity of mammary epithelia to niche cues (Figure S1C). Under optimized conditions, passaging ratios exceeded 1:24, and organoids formed bilayered, lumenized structures within 14 days, recapitulating early hallmarks of mammary morphogenesis. The three-tier optimization strategy: growth, homogeneity, lactogenic differentiation resulted in the RDM composition, when combined with lactation medium, induced strong CSN2 expression while maintaining robust cell expansion. This highlights CSN2 as a particularly sensitive marker of lactatogenic differentiation in bovine mammary epithelium in vitro. Bovine MGOs acquired lactogenic characteristics primarily through differentiation cues, with physiological hormones acting as amplifiers rather than initiators of CSN2 expression, unlike murine MGOs [40]. Withdrawal of lactation medium induced a reversible decrease in CSN2, consistent with hormonally regulated involution. Technical constraints limit protein analyses, as organoid polarity and matrix viscosity restrict apical protein release and sampling, making transcript quantification the most reliable functional readout. Because the milk proteins LTF and PAEP were differentially regulated, the differentiation trajectory of the luminal epithelial cells may be incomplete and thus may not represent full lactogenic competence or reflect averaging of subpopulation-specific regulation. Occasional branching in early passages (Figure S1B) suggests latent, intricate morphogenetic potential, although more complex architectures will likely require advanced culture systems such as composite hydrogels or feeder-cell supported formats [41,42].

Functionally, the MGOs combined luminal–basal polarity with robust susceptibility to diverse IAV lineages. The system supported productive replication of HPAIV H5N1, LPAIV H7N7, swine-origin H1N1 06, and human pandemic H1N1 09, whereas LPAIV H9N2 showed restricted infection. Peak H5N1 titres were similar to those in infected udder tissue but were lower than in naturally infected cow milk [43]. This likely results from apical-in-polarity and basolateral sampling, which underestimates apical shedding. Interestingly, H5N1 did not outperform swine or human H1N1, suggesting that the bovine udder is generally permissive to IAVs, including mammalian-adapted lineages. Given its unique origins and receptors, this suggests a non-exclusive HPAIV niche in the udder, and mammary infection may not be a hallmark of H5N1. Notably, exogenous trypsin was used to activate H1N1 HA, thereby potentially enabling replication in cells that would otherwise restrict in vivo infection. This limits direct comparison with HPAIV H5N1, which replicates well in MGOs without trypsin due to its multibasic cleavage site. Receptor analysis of our MGOs indicates abundant display of both α2,3 and α2,6 linked sialic acids, providing a mechanistic link to the observed broad susceptibility. This suggests that receptor shifts might not be necessary for short-term replication in this specific niche. While ex vivo studies were inconclusive for parenchymal tissue, the described model helps to bridge this gap and supports the existence of a broad IAV niche in the secretory compartment, fulfilling an initial requirement for potential reassortment. Recent investigations in other bovine in vitro models argue similarly that the niche is beneficial for IAV in general, and provide first evidence for IAV coinfections in cow mammary cells [21]. Whether the MGOs function only as IAV amplification reactors or also meet additional criteria for genomic reassortment and genetic innovation can now be experimentally assessed by double infection with mammalian and avian IAVs.

In contrast to mammalian-origin IAVs, LPAIVs showed divergent replication patterns, with H9N2 being highly restricted and H7N7 showing moderate replication. Interestingly, H7N7 retains avian-type α2,3 sialic acid receptor preference, consistent with the predominant α2,3 receptor distribution reported in bovine mammary epithelium [19,20], which may explain its moderate but productive replication. In contrast, H9N2 binds both α2,3 and α2,6 linked sialic acids [44] and showed highly restricted replication, supporting the idea that α2,3-linked sialic acid recognition facilitates productive infection, as observed for H7N7 and H5N1 [22,23]. However, this doesn't fully explain the divergence in replication in MGOs, since both avian-type and mammalian-adapted viruses can infect this model, and both linkage isomers were detected by lectin staining. Hence, α2,3 linkage preference is not the sole determinant of productive infection. Instead, bovine tissues abundantly display N-glycolylneuraminic acid (Neu5Gc) in addition to N-acetylneuraminic acid (Neu5Ac) [45], thus the stereochemistry of the terminal glycosidic bond does not equate to entry efficacy, and analysis of complex glycoconjugates is needed to assess this distinction on the level of receptor binding. Moreover, post-entry restrictions, including differences in polymerase activity, intracellular trafficking, or innate immune sensing in bovine mammary epithelial cells, may additionally contribute to the divergent outcomes, though direct evidence for these mechanisms awaits further investigation. Focussing only on the HA, recombinant binding studies on udder slices indicated a specific attachment of H5 compared to other IAV subtypes [46], not reflecting true functional susceptibility. Although LPAIV infection was facilitated by exogenous trypsin, our study demonstrated that the mammary gland tissue provides a specialized environment for IAV replication, extending beyond H5N1 to selected LPAIVs and to swine and human-origin IAVs. This may extend to other ruminants, as H5N1 was detected in a sheep in the UK [47] and pathogenic outcomes observed in goats infected with HPAIV resembled those in cattle [48]. Consequently, the emergence of H5N1 in the udder may not represent a distinct viral case but rather a distinct tissue context. This is especially true, as other viruses, such as tick-borne encephalitis virus, have previously been detected in milk products in the EU [49,50]. Yet, the public health concerns regarding the enzootic HPAIV infection in dairy cows remain substantial.

A key strength of the model is the ability to link infection to lactation biology. The bulk transcriptional analysis captures the overall functional impact of infection at the tissue level. The finding that infection was largely independent of apparent CSN2 expression implies that entry and replication factors are broadly present across maturation stages in the bovine mammary epithelium [51]. Differentiated MGOs exhibited strong hormone-responsive CSN2 expression, and infection kinetics were not facilitated by lactogenic hormones, suggesting no direct dependence on milk protein synthesis. These findings are consistent with recent experimental evidence showing sustained mammary susceptibility during the dry-off period in cows [52]. Since both studies do not capture the full spectrum of mammary physiological states, a possible role of the lactational state on IAV susceptibility and replication efficiency remains to be further investigated. Furthermore, cell-type-specific effects on viral susceptibility and lactogenic gene expression cannot be excluded. Future application of single-cell RNA sequencing or RNA in situ hybridization targeting CSN2 and IAV NP would provide higher spatial resolution to infer IAV susceptibility in mammary cell populations and resolve differential responses. Notably, infection did not have a homogeneous impact on CSN2 expression. The differential effect on CSN2 expression between H5N1 and H1N1 strains under non-lactogenic conditions may reflect strain-specific differences in innate immune activation, as type I interferon signalling has been shown to intersect with lactogenic differentiation programs and modulate CSN2 expression in mammary epithelial cells [53]. With hormonal support, prolactin-driven STAT5 signalling may counteract virus-induced transcriptional suppression [54], though this interpretation remains hypothetical.

The potential for extramammary dissemination of bovine H5N1 remains debated. Bovine respiratory ALI cultures show productive infection with bovine-adapted HPAIV, Eurasian H5N1, and the IAV panel used in this study [29,55]. The respiratory replication hierarchy of diverse IAVs was consistent with our udder model, with HPAIV and swine strains achieving similar endpoint titres [29,55]. This is consistent with dual receptor presentation reported for bovine respiratory and mammary tissues [19]. Notably, LPAIV H7N7 was more efficient than the human-origin strain in the respiratory context. Nevertheless, current genetic analyses indicate that bovine H5N1 viruses retain avian-type receptor specificity despite the extensive scale of the enzootic event in the USA [22–24]. Our research contributes to a conceptualization in which the bovine udder may be less a site of rapid adaptation and possibly a site of potential genomic recombination, although this hypothesis requires solid scientific substantiation. The recent demonstration of porcine infection with bovine H5N1 [56] and susceptibility of human nasal ALIs [57] emphasize that tissue tropism and host range must be evaluated across multiple compartments and species. Geographically, the mammary involvement of the Eurasian H5N1 lineage has been demonstrated in vivo [43], and recent surveillance in the Netherlands has detected H5N1 antibodies in dairy cattle [58], suggesting that this phenomenon may not be restricted to the United States.

Limitations of this study include the absence of stromal, endothelial, and immune components, as well as systemic mediators that shape in vivo responses. In the intact mammary gland, the dense vascular network supports high metabolic activity, facilitates viral spread, and supports immune cell recruitment. In vivo, severe infection has been associated with extensive epithelial necrosis accompanied by infiltration of degenerate neutrophils and accumulation of cellular debris, while the basal lamina and myoepithelial layer remain largely intact [14,43]. Such features of tissue-level pathology cannot be reproduced in MGOs thus far, as no morphological changes were evident. Further studies are needed to evaluate cytopathic effects and tissue damage, incorporating quantitative assessments of viability. Since MGOs are epithelia-only, viral spread and inflammatory responses are likely underrepresented because vascular dissemination and key early effector cells, such as neutrophils and macrophages, are absent. Individual contributions to pathogenesis can now be experimentally deconvoluted, with neutrophils and macrophages as promising initial targets for co-cultivation. Further incorporation of vascular or stromal elements, and in vivo validation of mammary reconstruction using xenografts will be important next steps to enhance the model's physiological relevance. In conclusion, this study establishes a scalable bovine mammary gland organoid model that recapitulates key epithelial lineages and hormone-responsive lactation, suitable for studying epithelial susceptibility and early virus–host interactions.

Funding Statement

This study was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)-398066876/GRK 2485/2.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental Material

Supplemental data for this article can be accessed online at http://dx.doi.org/10.1080/22221751.2026.2698239.