Comprehensive pathological, immunohistochemical, and apoptotic analysis of the highly pathogenic avian influenza a virus H5N1 clade 2.3.4.4b in poultry in Taiwan Authors: Zhi-Jie Yi, Yao-I Huang, Yu-Ju Lin, Yu-Wei Tsai, Ming-Chu Cheng, Ya-Mei Chen Published: 2026 ABSTRACT AbstractSince 2021, the highly pathogenic avian influenza (HPAI) virus H5N1 clade 2.3.4.4b has become the predominant circulating strain in Taiwan and has caused substantial losses in the poultry industry. However, its pathogenicity and tissue tropism across different poultry species remain insufficiently characterized. This study investigated gross lesions, histopathological changes, viral antigen distribution, and apoptosis-associated signals in chickens, ducks, geese, and quails naturally infected with HPAI virus H5N1 clade 2.3.4.4b in Taiwan. Samples were collected from 41 laboratory-confirmed H5N1-positive farms between October 2023 and February 2025, including 23 chicken farms, 4 duck farms, 10 goose farms, and 4 quail farms. Gross lesions were systematically evaluated. Furthermore, tissues including the brain, heart, liver, spleen, lung, kidney, and pancreas were examined using histopathology, immunohistochemistry for influenza A virus nucleoprotein and cleaved caspase-3, and terminal deoxynucleotidyl transferase dUTP nick end labeling assay. Semiquantitative scoring systems were applied to compare the severity of histopathological lesions, viral antigen distribution, and apoptotic activity across tissues and species. Results showed that chickens exhibited systemic infection with widespread necrosis and viral antigen distribution. Ducks displayed milder lesions but a more coordinated apoptotic response. Geese showed intermediate pathological features, with moderate correlation between lesion severity and viral antigen distribution. Quails displayed a unique host response, including prominent splenic lesions and distinct associations among viral antigen distribution, apoptosis, and tissue damage. Interspecies comparisons further identified significant differences in splenic and renal lesion scores, renal viral antigen scores, and apoptosis-associated scores in selected organs. Overall, these findings indicate that HPAI virus H5N1 clade 2.3.4.4b induces species-dependent pathological patterns and apoptotic responses in poultry. This field-based study provides pathological evidence supporting host-specific differences in tissue tropism and disease outcome under natural infection conditions.Keywords: Apoptosis, Highly pathogenic avian influenza virus H5N1 clade 2.3.4.4b, Histopathology, Immunohistochemistry Introduction Avian influenza is caused by influenza A viruses belonging to the family Orthomyxoviridae and the genus Alphainfluenzavirus. Influenza A viruses are classified into distinct HxNy subtypes according to the combinations of the surface glycoproteins hemagglutinin (HA) and neuraminidase (NA). To date, 18 HA (H1–H18) and 11 NA (N1–N11) subtypes have been identified (Imperia et al., 2023; Azeem and Yoon, 2025). Since the isolation of the A/goose/Guangdong/1/96-like (GsGd) highly pathogenic avian influenza (HPAI) H5 virus in Guangdong, China, in 1996, this lineage—commonly referred to as the GsGd virus—has been found to infect chickens, domestic ducks, and wild waterfowl under both natural and experimental conditions (Kwon and Swayne, 2010). Beginning in 2021, HPAI H5N1 clade 2.3.4.4b became dominant in wild-bird populations, replacing the previously prevalent H5N8 subtype observed during the 2020–2021 autumn–winter seasons. This shift has led to worldwide epizootics that continue to the present (Pohlmann et al., 2022; Kwon et al., 2023). The H5N1 outbreaks have caused not only deaths in previously unaffected wild-bird species but also substantial losses in poultry (Capelastegui and Goldhill, 2025), indicating differences in pathogenicity and epidemiology compared with earlier strains. Since its emergence, the GsGd lineage has undergone extensive antigenic diversification caused by mutations in the HA protein, resulting in multiple distinct sublineages. Previous studies have shown that clade 2.3.4.4c predominated in Taiwan from 2015 to 2019 and became endemic during that period (Li et al., 2024). According to surveillance reports from the Veterinary Research Institute, Ministry of Agriculture in Taiwan, clade 2.3.4.4b was first detected in Taiwan in 2021. The H5N2 subtype was first identified in 2021. Thereafter, H5N1 was initially detected in 2022, and it became the predominant subtype found in subsequent cases. Official national records documented 50 HPAI-affected poultry farms in 2023, 41 in 2024, and 23 in 2025, with 887,752, 573,552, and 748,946 birds culled, respectively. H5N1 was the predominant detected subtype during this period and accounted for all farm outbreaks listed in the 2024 and 2025 national summaries. These recurring outbreaks across terrestrial poultry, waterfowl, and quails provided the epidemiological context for the cases analyzed in this study (Animal and Plant Health Inspection Agency, 2024, 2025a,b; Veterinary Research Institute, 2025). Regarding the pathogenesis, HPAI virus can infect endothelial cells, leading to increased vascular permeability, viremia, and systemic organ dysfunction. Necrotizing inflammation is the characteristic lesion associated with HPAI (Landmann et al., 2021). Previous studies have revealed that the number of infected endothelial cells is correlated with disease progression and mortality. However, clinical severity and host susceptibility are also affected by viral subtypes, host species, and age (Capua and Alexander, 2009). Experimental infections with HPAI virus are associated with mortality rates of up to 100% in chickens and 60%–80% in ducks (Wahlgren, 2011; Cha et al., 2025), thereby emphasizing the significant impact of H5N1 on the poultry industry. Therefore, whether the pathogenicity and tissue tropism of H5N1 clade 2.3.4.4b differ among species compared with previous reports should be validated. Apoptosis is a type of programmed cell death that functions to eliminate unnecessary or damaged cells. However, excessive apoptosis can result in tissue injury (Zaid and Ali, 2016). In viral infections, apoptosis may reflect virus-induced cellular damage or can be a host defense mechanism (Ito et al., 2002). The HPAI H5N1 virus can induce apoptosis via multiple pathways, including both caspase-dependent and caspase-independent mechanisms (Yang et al., 2009). The HA protein of H5N1 has been identified as a key factor involved in the induction of apoptosis (Ueda et al., 2010). In addition, some studies using chicken tracheal organ cultures infected with HPAI H5N1 and H7N1 viruses have shown that the number of apoptotic cells observed at 8 h postinfection and the interferon titer levels at 24 h postinfection were higher in H5N1 viruses than in H7N1 viruses. Based on these findings, increased interferon levels may promote apoptosis while inhibiting viral replication (Petersen et al., 2013). Therefore, the role of apoptosis in viral infection may vary depending on viral subtype and host species, indicating a bidirectional relationship in which viral replication can induce apoptosis. Meanwhile, excessive apoptosis may limit viral dissemination within the host. Understanding the pathogenic mechanisms of HPAI H5N1 clade 2.3.4.4b across different poultry species is important for risk assessment and the development of effective control strategies. Tissue tropism, patterns of viral distribution, and apoptotic responses vary among host species and viral strains. However, comprehensive analyses of this emerging clade in Taiwanese poultry remain limited. This study aimed to elucidate the association between viral distribution and host responses and to further advance the understanding of its pathogenic mechanisms by conducting a comprehensive pathological, immunohistochemical, and apoptotic analysis of poultry naturally infected with the HPAI H5N1 clade 2.3.4.4b in Taiwan. Materials and methods Collected samples This study included samples collected between October 2023 and February 2025. All samples were obtained from poultry farms that were laboratory-confirmed as HPAI-positive by the Veterinary Research Institute of the Ministry of Agriculture in Taiwan. Diagnostic specimens were initially screened by reverse-transcription PCR targeting the influenza A matrix (M) gene. M-gene-positive specimens were submitted for viral gene sequencing; available sequence data, together with subtype-specific diagnostic results, were used for subtype and clade assignment. Farms investigated after reported illness, increased mortality, a complaint, or another field abnormality were classified as passive-surveillance cases. Farms detected through routine active surveillance were classified as active-surveillance cases. For each farm, approximately 10 live poultry exhibiting evident clinical signs were targeted for necropsy and organ sampling. Gross lesions were systematically recorded. Organ samples, including the cerebrum and cerebellum, heart, liver, spleen, lungs, kidneys, and pancreas, were subsequently collected for further analyses. In addition, a 46-day-old specific pathogen-free chicken was used as a negative control for histopathological and immunohistochemical analyses (IACUC-114-073). Histopathological examination The collected tissues were fixed in 10% neutral buffered formalin for <48 h. The fixed tissues were routinely processed, paraffin-embedded, and cut into 4-μm-thick sections. Tissue sections were subsequently stained with hematoxylin and eosin (HE) for histopathological examination. Necrotizing inflammatory lesions consistent with the characteristic features of HPAI were recorded as lesion scores. The lesion score criteria were modified from a previous study (Landmann et al., 2021). Based on the extent of necrosis or necrotizing inflammation, lesions were graded as follows: score 0, no lesions; score 1, lesions involving 1%–50% of the section; score 2, lesions involving 50%–75% of the section; and score 3, lesions involving 75%–100% of the section. Immunohistochemistry (IHC) To identify viral tropism in different tissues, a monoclonal antibody (WF-4) against the nucleoprotein of influenza A virus, provided by the Veterinary Research Institute of the Ministry of Agriculture, was used (Tu et al., 2019). Apoptotic activity was assessed using a polyclonal antibody against cleaved caspase-3 (p17/p19; cat. no. 19677-1-AP; Proteintech, Rosemont, IL, USA). To assess spatial correspondence between viral antigen and apoptosis-associated signals, the same anatomical regions and corresponding cell populations were compared across serial sections stained for influenza A virus nucleoprotein, cleaved caspase-3, and TUNEL. After an overnight dry incubation of 4-μm-thick tissue sections at 37°C, the sections were dewaxed in xylene and rehydrated via a graded ethanol series. For the WF-4 antibody, the sections were incubated with proteinase K (cat. 03115828001; 100 μg/mL; Roche Diagnostics, Germany) at 55°C for 15 min. Meanwhile, antigen retrieval for the caspase-3 (p17/p19) antibody was performed using Trilogy (cat. 920P-06; Cell Marque, Rocklin, CA, the USA) at 115°C for 30 min. The sections were then washed with Tris-Buffered Saline with Tween-20 for 1 min and incubated with UltraVision hydrogen peroxide block (cat. TL-125-QHL; Epredia, Kalamazoo, MI, the USA) for 10 min to block endogenous peroxidase activity. Each tissue section was incubated with primary antibodies diluted 1:2,000 at 4°C for 16–18 h. The sections were later incubated with HRP Polymer Quanto (cat. TL-125-QHL; Epredia, Kalamazoo, MI, the USA) for 10 min to catalyze chromogenic development. Subsequently, the sections were incubated with 3,3′-diaminobenzidine (cat. TA-125-QHDX; Epredia, Kalamazoo, MI, the USA) for 5 min. Finally, the sections were counterstained with hematoxylin for 30 s. Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay To detect DNA fragmentation associated with late-stage apoptosis, the TUNEL assay was performed using a commercial kit (cat. 11684817910; Roche Diagnostics, Germany) according to the manufacturer’s instructions. The sections were then incubated with 3,3′-diaminobenzidine for 5 min. Finally, they were counterstained with hematoxylin for 30 s. Semiquantitative scoring system To evaluate viral distribution, the presence of viral antigens was assessed using a virus scoring system and recorded as the virus score (Landmann et al., 2021). Briefly, the distribution of viral antigens was graded on a scale of 0–3 according to the extent of immunoreactive areas and the number of positive staining clusters. The scoring criteria were defined as follows: A score of 0 indicated no detectable immunoreactivity. A score of 1 was assigned based on limited immunoreactive areas, defined as < 10% in the heart; < 5% in the brain, lung, liver, and pancreas; and ≤ 2% in the spleen and kidney. A score of 2 represented moderate involvement with organ-specific criteria: 5%–20% immunoreactive area and ≤ 5 clusters per low-power field (LPF) in the brain; ≤ 5% immunoreactive area or ≥ 2 clusters per LPF in the kidney; 5%–24% immunoreactive area or > 1 cluster per LPF in the pancreas; < 10% immunoreactive area with 2–10 immunoreactive cells per high-power field (HPF) in the heart; < 5% immunoreactive area with > 3 immunoreactive cells per HPF in the lung; 5%–14% immunoreactive area or 3–10 immunoreactive cells per HPF in the liver; and 2%–14% immunoreactive area or 16–40 immunoreactive cells per HPF in the spleen. A score of 3 indicated extensive involvement: > 20% immunoreactive area and/or > 5 clusters per LPF in the brain; > 5% immunoreactive area or 5% area with coalescing clusters in the kidney; ≥ 25% immunoreactive area in the pancreas; ≥ 10% immunoreactive area and/or > 10 immunoreactive cells per HPF in the heart; ≥ 5% immunoreactive area in the lung; ≥ 15% immunoreactive area and/or > 10 immunoreactive cells per HPF in the liver; and ≥ 15% immunoreactive area or > 40 immunoreactive cells per HPF in the spleen. Caspase-3 and TUNEL staining were evaluated using a previously established scoring system and recorded as the caspase-3 score and TUNEL score, respectively (Anis et al., 2013). According to the extent of immunoreactive areas, positive signals were categorized as follows: score 0, no positive signal; score 1, positive signals involving 1%–50% of the tissue section; score 2, positive signals involving 51%–75% of the section; and score 3, positive signals involving 76%–100% of the section. Statistical analysis Gross lesions were evaluated on an individual-animal basis. The lesion, virus, caspase-3, and TUNEL scores were assessed on a per-farm basis, with the final score for each farm case defined by the most severe lesion observed among the examined sections. Within-species differences in lesion, virus, caspase-3, and TUNEL scores were assessed using the nonparametric Friedman test, followed by the Dunn’s post-hoc test for pairwise comparisons. Interspecies differences in lesion, virus, caspase-3, and TUNEL scores among chickens, ducks, geese, and quails were evaluated using the nonparametric Kruskal–Wallis test, followed by the Dunn’s post-hoc test for pairwise comparisons. Spearman’s rank correlation analysis was additionally performed to assess the association between histopathological and immunohistochemical scores. All statistical analyses were conducted using GraphPad Prism version 10.1.2 (GraphPad Software, San Diego, CA, the USA). A P value of ≤ 0.05 indicated statistically significant differences. Results Gross lesions This study included 41 farms, comprising 23 chicken farms, 4 duck farms, 10 goose farms, and 4 quail farms. All study farms were confirmed as H5N1 clade 2.3.4.4b. Overall, 38 of 41 farms (92.7%) were passive-surveillance cases and 3 (7.3%) were active-surveillance cases. Clinical presentation and surveillance source varied among farms (Supplemental Table S1). Recorded presentations included increased mortality or sudden death, depression, neurological signs such as torticollis, inability to stand, opisthotonus or paddling, respiratory signs or increased oral/nasal secretions, cyanosis or hyperemia of unfeathered skin, and diarrhea or green feces. Several farms had no specific clinical history recorded. Due to field sampling limitations and variations in availability, the number of birds collected per farm ranged from 3 to 15, with an average of 10 birds per farm (Supplemental Table S1). A total of 204 chickens, 29 ducks, 89 geese, and 45 quails were examined for gross lesions. The frequencies of individual gross lesions among the poultry species (Fig. 1). Common lesions included brain congestion, tracheal reddening, hydropericardium, heart collapse, epicardial or endocardium petechiae, rubbery lung, spleen enlargement, spleen necrosis, pancreas reddening, pancreas necrosis, kidney enlargement, cecal tonsil reddening, proventriculus–gizzard junction hemorrhage, and muscular and/or subcutaneous hemorrhage (Fig. 2). The presence of gross lesions in different poultry species naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b, including chickens (n = 204), ducks (n = 29), geese (n = 89), and quails (n = 45), was evaluated. The numbers in the heat map represent the percentage of cases in which each lesion was observed. Gross lesions, including brain congestion (A), pericardial effusion (B), hemorrhage at the proventriculus–gizzard junction (C), spleen enlargement (D), pancreatic necrosis (E), and renal enlargement (F), observed in chickens naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b. Overall, chickens were more likely to develop lesions compared with the other species, and these lesions were significantly nonspecific. The predominant gross lesions included renal enlargement, pericardial effusion, and pulmonary congestion/edema. Classical lesions previously described in HPAI, such as hemorrhage at the proventriculus–gizzard junction and shank hyperemia, were infrequently observed in the present study. Ducks, geese, and quails had lower frequencies of gross lesions, with some lesions not observed. In ducks, the main lesions included cerebral congestion, pulmonary congestion/edema, spleen enlargement, and petechial hemorrhages on the endocardium and epicardium. In geese, the most common lesions included cerebral congestion, pancreatic necrosis, and petechial hemorrhages on the endocardium and epicardium. The quails did not exhibit evident gross lesions. Histopathological lesions The lesion scores were used to represent the severity of HPAI-associated tissue damage. Detailed scoring criteria are provided in Supplemental Table S2. Histopathological lesions were evaluated in seven routinely examined organs, including the brain, heart, liver, spleen, lungs, kidneys, and pancreas (Fig. 3). The major lesions included necrotizing encephalitis with perivascular cuffing, necrotizing myocarditis accompanied by lymphocytic infiltration, necrotizing pneumonia or bronchopneumonia, necrotizing hepatitis with lymphocytic infiltration, necrotizing splenitis with lymphoid depletion, necrotizing tubulointerstitial nephritis with hyaline degeneration, and necrotizing pancreatitis. In addition, sporadic lesions were identified in several other tissues, including the proventriculus–gizzard junction, proventriculus, skeletal muscle, bursa of Fabricius, and cecal tonsils. Lesion scores in poultry naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b. Viral antigen distribution score The virus score was used to evaluate AI antigen localization (Fig. 4). Supplemental Table S3 shows the scoring details. As the WF-4 monoclonal antibody targets the nucleoprotein, only parenchymal cells exhibiting a clearly defined nuclear immunoreactivity were included in the assessment. The cell types assessed were neurons, glial cells, cardiomyocytes, bronchial and parabronchial epithelial cells, hepatocytes, splenic reticular cells, renal tubular epithelial cells, and pancreatic acinar cells. Meanwhile, cells of uncertain classification were excluded. Virus scores in poultry naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b. Apoptosis distribution Apoptosis was evaluated using two methods, including IHC and TUNEL assay. Execution-phase apoptosis (hereafter referred to as the caspase-3 score) was assessed based on the distribution of activated caspase-3 (Fig. 5). Because activated caspase-3 initially presented in the cytoplasm and subsequently translocated to the nucleus, parenchymal cells showing positive signals in either the cytoplasm or nucleus were included in the evaluation. Late-stage apoptosis (hereafter referred to as the TUNEL score) was examined using the TUNEL assay, which detects DNA fragmentation within the nucleus (Fig. 6). Therefore, only parenchymal cells with nuclear positivity were investigated. The parenchymal cell types assessed were similar to those used for the virus score. Supplemental Tables S4 and S5 show the scoring details for caspase-3 and TUNEL results, respectively. Caspase-3 scores in poultry naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b. Terminal deoxynucleotidyl transferase dUTP nick end labeling assay scores in poultry naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b. Within-species analysis In order to evaluate tissue tropism, Friedman’s test was performed to analyze lesion scores, viral antigen distribution, and apoptosis scores among different tissues within each avian species. Because the samples were derived from field cases of natural infection, not all target organs were available for every individual. In chickens and geese, farms with incomplete organ data were excluded from the analysis because sufficient case numbers remained after exclusion, resulting in final sample sizes of 16 chicken farms and 8 goose farms for the Friedman analysis. In contrast, ducks and quails had only 4 available farms, and exclusion of incomplete farms would have markedly reduced the sample size. Therefore, organs with insufficient paired observations (fewer than three samples) were excluded to retain all four farms in the Friedman analysis, which requires complete paired observations across all compared organs (Fig. 7). Semiquantitative scoring results of histopathological lesions, viral antigen distribution, caspase-3 and Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay in chickens (A–D), ducks (E–H), geese (I–L), and quails (M–P) naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b. Statistical analysis was performed using the Friedman test, followed by the Dunn’s post-hoc test. * indicates P < 0.05. In chickens, the lesion scores differed significantly among the organs (χ²(6) = 15.57, P = 0.0163). However, no pairwise comparisons were significant after Dunn’s post-hoc test. Based on the mean ranks, the spleen had the greatest lesion severity, followed by the pancreas. The virus scores did not differ significantly (χ²(6) = 6.566, P = 0.3628). The spleen had the highest ranking, followed by the liver. The caspase-3 scores significantly differed (χ²(6) = 27.88, P < 0.0001), with only the brain and kidney having remarkable differences after Dunn’s correction (adjusted P = 0.0452). The mean ranks indicated that the brain had the highest caspase-3 distribution, followed by the spleen and lungs. Meanwhile, the kidneys had the lowest distribution. The TUNEL scores also differed significantly (χ²(6) = 15.38, P = 0.0175). However, significant pairwise differences were not identified. The heart had the highest mean rank, followed by the liver and spleen. In ducks, the lungs and kidneys were excluded due to limited sampling. The organs did not significantly differ in terms of lesion scores (χ²(4) = 7.61, P = 0.1300), virus scores (χ²(4) = 5.91, P = 0.2400), or caspase-3 scores (χ²(4) = 5.89, P = 0.2160). Based on the mean ranks, the brain had the highest lesion scores, followed by the liver and pancreas. The brain also had the highest virus scores, followed by the liver and spleen. The spleen had the highest caspase-3 scores, followed by the brain. The overall TUNEL scores significantly differed (χ²(4) = 11.17, P = 0.0033). However, the pairwise comparisons showed no significant results. The brain had the highest ranking, followed by the heart and liver. In geese, the lesion scores differed significantly among the organs (χ²(6) = 15.64, P = 0.0158). However, significant pairwise differences were not identified. The pancreas had the greatest lesion severity, followed by the brain and spleen. The virus scores differed significantly (χ²(6) = 30.39, P < 0.0001), with a remarkable difference only between the liver and kidneys after the Dunn’s test (adjusted P = 0.0454). Based on the mean ranks, the liver had the highest viral antigen distribution, followed by the spleen and brain. The kidney had the lowest distribution. The caspase-3 scores also differed significantly (χ²(6) = 13.77, P = 0.0323). However, significant pairwise differences were not detected. The spleen had the highest ranking, followed by the liver and lungs. The TUNEL scores differed significantly (χ²(6) = 14.82, P = 0.0217). Nevertheless, pairwise differences were not identified. The brain and liver had the highest mean ranks, followed by the pancreas. In quails, the heart, lungs, and kidneys were excluded due to insufficient sampling. The organs did not significantly differ in terms of lesion scores (χ²(3) = 5.23, P = 0.1806), virus scores (χ²(3) = 4.17, P = 0.3333), or TUNEL scores (χ²(3) = 7.20, P = 0.1250). Based on the mean ranks, the spleen had the highest lesion scores, followed by the pancreas and liver. The brain had the highest virus scores, followed by the spleen and pancreas. The brain and liver had the highest TUNEL scores. Overall, the caspase-3 scores significantly differed (χ²(3) = 9.77, P = 0.0052). However, no significant pairwise differences were identified. The liver had the highest ranking, followed by the pancreas and brain. Interspecies analysis Interspecies comparisons were performed using all available scoring data and analyzed with the Kruskal–Wallis test. Because the samples were derived from field cases of natural infection, not all target organs were available for every individual. Accordingly, the final numbers of samples included in the analysis were as follows: brain (chickens, n = 22; ducks, n = 4; geese, n = 10; quails, n = 4), heart (chickens, n = 20; ducks, n = 4; geese, n = 10; quails, n = 3), liver (chickens, n = 18; ducks, n = 4; geese, n = 10; quails, n = 4), spleen (chickens, n = 21; ducks, n = 4; geese, n = 10; quails, n = 4), lung (chickens, n = 21; ducks, n = 3; geese, n = 10; quails, n = 2), kidney (chickens, n = 22; ducks, n = 3; geese, n = 9; quails, n = 2), and pancreas (chickens, n = 21; ducks, n = 4; geese, n = 10; quails, n = 4). Based on the interspecies comparisons, the lesion scores (Fig. 8) for the spleen significantly differed (P = 0.0137). Furthermore, the Dunn’s post-hoc test showed that quails had significantly higher splenic lesion scores than ducks (P = 0.0071). Significant differences were also observed in the kidneys (P = 0.0138), with geese having significantly higher renal lesion scores than chickens (P = 0.0171). These findings indicate that quails had more severe splenic lesions than ducks, and geese had more severe renal lesions than chickens. For the virus scores, a significant difference was observed in the kidneys (P = 0.0008). The Dunn’s post-hoc test showed that chickens had significantly higher renal virus scores than geese (P = 0.0032), indicating a more extensive viral antigen distribution in the kidneys of chickens than in those of geese. Comparison of lesion (A), virus (B), caspase-3 (C), and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) (D) scores in poultry species naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b. Statistical analysis was performed using the Kruskal–Wallis test, followed by the Dunn’s post-hoc test. * indicates P < 0.05, and ** demonstrates P < 0.01. The caspase-3 scores for the liver significantly differed (P = 0.0142), with quails having significantly higher scores than ducks (P = 0.0084). Significant differences were also found for the spleen (P = 0.0336). The Dunn’s post-hoc test showed that chickens and geese had significantly higher splenic caspase-3 scores than quails (P = 0.0405 and 0.0424, respectively). In the kidneys, significant differences were observed (P = 0.0092), with quails having significantly higher caspase-3 scores than chickens, ducks, and geese (P = 0.0328, 0.0482, and 0.005, respectively). Based on these results, the execution-phase apoptotic signals after infection were more extensive in the liver and kidneys of quails. However, they were less evident in the spleen. The TUNEL scores significantly differed in the liver (P = 0.0352), with chickens having higher scores than ducks (P = 0.0428). Significant differences were also found for the lungs (P = 0.0191), with chickens having higher scores than ducks (P = 0.0374). According to these findings, late-stage apoptotic signals in the liver and lungs were less pronounced in ducks after infection with the HPAI H5N1 clade 2.3.4.4b. Correlation analysis among scoring systems within each poultry species In chickens, the overall correlations among the scores were low (Fig. 9). Lesion scores were not significantly correlated with virus scores (r = 0.14, P = 0.0879) or caspase-3 scores (r = 0.15, P = 0.0672), but showed a weak positive correlation with TUNEL scores (r = 0.23, P = 0.0052). Virus scores exhibited a weak positive correlation with caspase-3 scores (r = 0.43, P < 0.0001), whereas no correlation was observed between virus scores and TUNEL scores (r = 0.08, P = 0.3689). In addition, caspase-3 scores were not correlated with TUNEL scores (r = −0.01, P = 0.9269). Overall, viral antigen distribution was weakly associated with execution-phase apoptosis, whereas lesion severity was weakly associated with late-stage apoptosis but not with viral antigen distribution or execution-phase apoptosis. Spearman correlation analysis of lesion, virus, caspase-3, and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) scores in chickens (A), ducks (B), geese (C), and quails (D) naturally infected with the highly pathogenic avian influenza A viruses H5N1 clade 2.3.4.4b. The heat map displays correlation coefficients (r values), with the blue color indicating positive correlations and the red color denoting negative correlations. In ducks, the correlations ranged from low to moderate. Lesion scores tended to show a weak positive correlation with virus scores, although this association did not reach statistical significance (r = 0.38, P = 0.0528). Lesion scores were not significantly correlated with caspase-3 scores (r = 0.23, P = 0.2552) or TUNEL scores (r = 0.21, P = 0.3112). Virus scores showed weak positive correlations with both caspase-3 scores (r = 0.42, P = 0.0313) and TUNEL scores (r = 0.43, P = 0.0292). No significant correlation was observed between caspase-3 and TUNEL scores (r = −0.03, P = 0.8999). Overall, viral antigen distribution showed consistently weak positive associations with apoptotic markers, whereas lesion severity was not significantly associated with apoptosis or viral antigen distribution. In geese, the correlations ranged from low to moderate. Lesion scores showed a moderate positive correlation with virus scores (r = 0.54, P < 0.0001), but were not significantly correlated with caspase-3 scores (r = 0.06, P = 0.6057) or TUNEL scores (r = 0.18, P = 0.1487). Virus scores showed weak positive correlations with both caspase-3 scores (r = 0.27, P = 0.0277) and TUNEL scores (r = 0.45, P = 0.0001). No significant correlation was observed between caspase-3 and TUNEL scores (r = −0.14, P = 0.2590). Overall, viral antigen distribution was moderately associated with lesion severity and weakly associated with both execution-phase and late-stage apoptosis. In quails, the correlations among the evaluated scores ranged from weak to moderate, with several negative associations observed. Lesion scores tended to show a weak positive correlation with virus scores (r = 0.32, P = 0.1357) and weak negative correlations with caspase-3 scores (r = −0.24, P = 0.2700) and TUNEL scores (r = −0.36, P = 0.0891); however, none of these associations reached statistical significance. Virus scores showed a statistically significant weak negative correlation with caspase-3 scores (r = −0.42, P = 0.0459), but were not significantly correlated with TUNEL scores (r = 0.14, P = 0.5329). No significant correlation was observed between caspase-3 and TUNEL scores (r = −0.01, P = 0.9679). Overall, lesion severity was not significantly associated with viral antigen distribution, caspase-3 scores, or TUNEL scores. The only statistically significant association was an inverse correlation between viral antigen and caspase-3 scores. Discussion This study descriptively characterized gross lesions and applied semiquantitative scoring systems to investigate the relationships among histopathological lesions, viral antigen distribution, and apoptotic signals in chickens, ducks, geese, and quails naturally infected with HPAI virus H5N1 clade 2.3.4.4b. By integrating histopathology with immunohistochemistry and TUNEL analysis, this study aimed to determine whether tissue necrosis observed in HPAI infection was associated with viral distribution and apoptosis across different avian species. In contrast to most studies based on experimental infections, this study utilized naturally infected individuals, thereby more closely reflecting real-world epidemiological conditions. Overall, three key features were identified. First, there were significant interspecies differences in tissue tropism and target organs. Second, species-specific relationships were observed among viral antigen distribution, tissue lesions, and apoptosis. Third, cell death mechanisms may be a key factor underlying host susceptibility and viral adaptation. In chickens, the findings generally support the systemic nature of HPAI, with a predilection for highly vascularized organs such as the spleen and lungs (Schat et al., 2012; Gobbo et al., 2022; Graziosi et al., 2024). However, the incidence of classical gross lesions, including splenic necrosis and cyanosis of the shanks, was lower than that reported in previous studies (Lean et al., 2022; Kim et al., 2025), indicating a greater variability under natural infection conditions. This variation may reflect differences in infection timing, host age, farm conditions, and unmeasured concomitant diseases. Notably, no correlation was observed between histopathological lesions and viral antigen distribution in chickens, indicating that tissue damage is not solely caused by direct viral cytopathic effects. Rather, immune-mediated injury may play a more prominent role. Previous studies have shown that chickens infected with HPAI virus develop delayed yet intense cytokine responses, including significant increases in interleukin-6 and interleukin-8 levels (Horman et al., 2018; Huo et al., 2018), leading to endothelial damage and tissue necrosis. In addition, the downregulation of STAT3 signaling may exacerbate inflammatory imbalance (Kuchipudi et al., 2014). Regarding cell death, the viral antigen distribution had a weak positive correlation with caspase-3 scores. However, viral antigen distribution was not significantly correlated with TUNEL scores, suggesting that it was associated with execution-phase caspase activation but not with TUNEL-detectable DNA fragmentation. Nevertheless, the absence of a significant correlation does not demonstrate that late-stage apoptosis was incomplete. In contrast, lesion scores showed a weak positive correlation with TUNEL scores, indicating that TUNEL-detectable DNA fragmentation was weakly associated with lesion severity. This phenomenon may reflect virus–host interactions. The NS1 protein of HPAI virus can inhibit interferon signaling and activate the PI3K-Akt pathway (Ehrhardt et al., 2007; Zhirnov and Klenk, 2007), thereby delaying apoptosis to facilitate viral replication. However, excessive immune activation may lead to ATP depletion and mitochondrial dysfunction, shifting cell death toward necrosis or regulated necrosis (Sachet et al., 2017), which may partially explain the discordance among viral antigen distribution, tissue lesions, and apoptotic markers. In contrast, ducks exhibited a distinct pathological response. Histopathological lesions tended to show a weak positive association with viral antigen distribution, although this relationship did not reach statistical significance (P = 0.0528). Viral antigen had weak positive correlations with both caspase-3 and TUNEL scores, indicating associations with execution-phase caspase activation and DNA fragmentation. However, these correlations do not by themselves demonstrate a complete or sequential apoptotic process. This pattern is consistent with the relatively robust innate immune capacity of ducks. Ducks and geese possess the RIG-I gene, which enables the rapid recognition of viral RNA and the activation of type I interferon responses (Saito et al., 2018; Gaide et al., 2022), thereby suppressing viral replication. In addition, ducks exhibit an earlier and more moderate cytokine response (Cornelissen et al., 2013), which may prevent excessive inflammation-induced tissue damage. At the tissue level, this study confirmed previous observations of evident neurotropism in ducks (James et al., 2023; Graziosi et al., 2024). The brain exhibited the highest levels of viral antigen and TUNEL signals, indicating that neural tissue is a primary target. This may be related to the limited regenerative capacity of neurons and the slower yet more prolonged viral replication kinetics in ducks (Schat et al., 2012; Graziosi et al., 2024). Geese presented with an intermediate pathological profile between chickens and ducks. Histopathological lesions had a moderate positive correlation with viral antigen distribution, suggesting a partial contribution of direct viral cytopathic effects. Viral antigen also had weak positive correlations with both TUNEL and caspase-3 scores, indicating associations with DNA fragmentation and execution-phase caspase activation. Nevertheless, caspase-3 and TUNEL scores were not significantly correlated, suggesting that these apoptotic markers may have differed in their spatial or temporal distributions. Apoptosis-inducing factor mediates caspase-independent apoptosis (Sevrioukova, 2011), and it can induce DNA fragmentation. Influenza A H1N1 virus and goose parvovirus can exploit this pathway to trigger cell death (Qu et al., 2017; Yan et al., 2021). However, the correlations observed in the present study do not establish the involvement of caspase-independent apoptosis in geese. Therefore, the potential contribution of caspase-independent pathways to HPAI-associated cell death requires further molecular validation. Quails exhibited a distinct pathological pattern Lesion scores showed a weak positive trend with viral antigen scores, suggesting that increased viral antigen levels may be accompanied by more severe lesions. However, this association did not reach statistical significance; therefore, the relationship between viral accumulation and tissue damage cannot be clearly established. In contrast, lesion scores showed weak negative trends with both cleaved caspase-3 and TUNEL scores, although neither association was statistically significant. These findings should therefore not be interpreted as evidence that increased lesion severity is associated with reduced apoptosis. Notably, viral antigen scores were significantly but weakly negatively correlated with cleaved caspase-3 scores. Previous studies have shown that quail may lack obvious lesions during the early stages of infection but can rapidly induce an interferon-gamma response (Uno et al., 2013), followed by extensive necrosis at later stages (Bertran et al., 2019). Thus, the inverse relationship between viral antigen and cleaved caspase-3 scores may reflect differences in the stage of infection or in their distribution at the tissue and cellular levels. Nevertheless, owing to the cross-sectional design of this study, these findings cannot establish that caspase-3 activation suppresses viral replication. Furthermore, individual variation under natural infection conditions may have influenced the relationships among viral antigen expression, cell death markers, and lesion severity. Differences in receptor distribution and viral adaptation may also contribute to interspecies variation. The distribution of SAα2,3-gal and SAα2,6-gal receptors differs among species, and it may influence viral entry efficiency (Wan and Perez, 2006). In addition, viral binding preferences for specific sialic acid structures, such as Sialyl-Lewis X and 3′-sialyl-LacNAc (Guo et al., 2017; Gaide et al., 2022), may further affect viral replication and tissue distribution across hosts. These receptor-level differences, along with immune responses and cell death mechanisms, shape species-specific pathological outcomes. This study also emphasizes the importance of natural infection models. Unlike experimental infections, naturally infected individuals present with uncontrolled variability in viral dose, infection timing, and unmeasured concomitant conditions, which may reduce correlations among pathological and molecular parameters. This likely explains discrepancies between the current findings and those obtained from experimental studies, such as the lack of association between lesions and viral antigen in chickens. Notably, the predominance of passive surveillance means that the study population was enriched for farms with clinically apparent disease or increased mortality. Therefore, the frequencies and severities of lesions and apoptosis-associated findings observed in this study may not be representative of all H5N1-infected poultry farms in Taiwan. Although this study performed a comprehensive pathological analysis of poultry naturally infected with the H5N1 virus in Taiwan, several limitations should be considered. First, because the samples were derived from field cases of natural infection, the timing of infection, initial viral dose, and transmission routes could not be precisely controlled. The complexity of field conditions may result in greater variability in lesion severity and viral distribution compared with controlled experimental settings. H9N2 reassortants containing Y280-like gene segments can replicate at low titers in the brains of experimentally infected chickens without causing obvious clinical signs. Thus, H9N2 coinfection could potentially confound the interpretation of neural injury and apoptosis markers. Low-pathogenic H9N2 was detected on an offshore island in Taiwan in 2023, but not on the main island. In this study, no H9 subtype virus was identified among 41 farms based on subtype testing and sequence analysis of M gene-positive samples. However, low-level coinfection below the detection limit cannot be completely excluded because H9-specific quantitative or Y280 lineage-specific assays were not performed. Second, most sampling was conducted after notification by local animal health authorities, and complete sets of target organs were not available for all individuals during field necropsy. Thus, some data, such as lung and kidney scores in ducks and quails, were excluded from some statistical analyses due to insufficient sample size. In addition, the number of samples varied among species, with the proportion of chickens being significantly higher than that of ducks and quails, which might have affected statistical power in interspecies comparisons. Finally, clinical histories were heterogeneous and were unavailable for some farms, and other infectious agents were not investigated using a standardized panel. Coinfections with other pathogens cannot be ruled out, although the necrotizing pattern with generally limited inflammation suggests that bacterial infection was unlikely to be the main source of the characteristic lesions. Nevertheless, future studies incorporating quantitative molecular analyses, such as RT-qPCR for viral load, and a more detailed characterization of cell death pathways could further validate the pathogenic mechanisms of H5N1 infection in different poultry species. In conclusion, this study is the first to compare natural cases of infection in four major poultry species in Taiwan using a consistent pathological framework. This study reveals distinct tissue tropisms and relationships among viral antigen distribution, apoptosis, and tissue damage associated with H5N1 clade 2.3.4.4b across avian species. Although chickens exhibited systemic infection, viral antigen distribution did not correspond with lesion severity, suggesting that tissue damage may not result solely from direct viral effects and that host immune responses may also play an important role. Ducks generally developed milder lesions, showed weak positive correlations between viral antigen distribution and both apoptotic markers, and exhibited pronounced neurotropism. In contrast, geese showed the clearest association between viral antigen distribution and lesion severity, suggesting that direct virus-induced tissue damage may contribute more substantially to disease pathogenesis. Despite the absence of obvious gross lesions, quail exhibited prominent splenic damage and a distinctive negative correlation between viral antigen and cleaved caspase-3 scores, indicating that assessments based solely on gross lesions may underestimate the extent of infection. Therefore, disease manifestations following H5N1 infection in poultry may be jointly shaped by host species, viral tissue tropism, and cell death responses. Funding This project was supported by the Animal and Plant Health Inspection Agency, Ministry of Agriculture, Taiwan (113AS-5.5.5-VP-01 and 114AS-5.5.4-VP-01). Author contributions Zhi-Jie Yi: Investigation, Methodology, Analysis, Data Curation, Writing – Original Draft Yao-I Huang: Methodology Yu-Ju Lin: Methodology Yu-Wei Tsai: Investigation, Methodology, Validation Ming-Chu Cheng: Validation Ya-Mei Chen: Conceptualization, Validation, Writing – Review & Editing Disclosures The authors declare no conflicts of interest. Footnotes Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.psj.2026.107614. Supplementary materials References Associated Data Supplementary Materials