A flavonoid extract from Bupleurum aerial parts given at recovery onset attenuates residual liver and lung injury in yellow-feathered broilers after acute heat exposure

Abstract

Most broiler heat stress studies focus on the exposure phase, whereas the first hours after heat withdrawal remain understudied. We tested whether a flavonoid extract from Bupleurum aerial parts (BAPF), administered at recovery onset, could reduce residual liver and lung injury in yellow-feathered broilers exposed to 35 °C for 6 h and then cooled to 24 °C over the next 6 h. The extract improved liver and lung histology, lowered serum injury markers and circulating interleukin-1β and interleukin-18, and improved hepatic redox indices, with the most consistent responses at 46 to 68 mg/kg body weight per day. Exploratory composite analysis also prioritized the 46 to 68 mg/kg body weight per day range. In a separate heat and rewarming model using the chicken fibroblast cell line DF-1, the extract added at recovery onset lowered reactive oxygen species, increased nuclear abundance of nuclear factor erythroid 2-related factor 2, and reduced immunoblot signals for NLR family pyrin domain containing 3, cleaved caspase-1, cleaved caspase-3, and cleaved gasdermin E. An ML385 pretreatment arm attenuated several of these changes, supporting involvement of this pathway. Overall, the findings suggest that early recovery after heat exposure may provide a post-heat intervention window and warrant further evaluation of BAPF initiated at recovery onset to limit residual organ injury in yellow-feathered broilers.

Introduction

Heat stress is a major constraint in broiler production. Recent reviews summarize oxidative stress and broader physiological consequences of heat exposure in poultry (Oke et al., 2024), and nutritional approaches to mitigation have also been reviewed (Olayiwola and Adedokun, 2025). However, most intervention studies have focused on the exposure phase rather than the first hours after heat withdrawal. In the present study, early recovery was treated as a distinct phase after heat exposure because oxidative disruption can persist for hours after heat withdrawal (Yang et al., 2010). Yellow-feathered broilers were studied at 80 d of age because this stage approximates the late growing period relevant to production for this breed type and may differ physiologically from younger fast growing commercial broilers commonly used in studies of heat stress. The challenge at 35 °C for 6 h, followed by 6 h of programmed cooling, was designed to impose an acute but nonlethal heat load and to isolate the early recovery period after heat withdrawal, rather than to model chronic or cyclic heat exposure.

Phytochemical work has identified anti-inflammatory constituents in the aerial parts of Bupleurum scorzonerifolium, providing a basis for evaluating a flavonoid extract from Bupleurum aerial parts (BAPF) in a heat injury model (Li et al., 2024). Studies in yellow-feathered broilers have documented altered feeding behavior and welfare indicators related to heat stress (Feng et al., 2024; Uea-Anuwong et al., 2025), whereas studies in broilers more broadly have reported impaired growth and physiological responses under thermal load (Beckford et al., 2020; Kim et al., 2025). In broilers, heat exposure disrupts hepatic redox homeostasis and, in chronic models, has been linked to hepatic inflammatory signaling and injury to the pulmonary blood-air barrier (Liu et al., 2022; Tang et al., 2022; Wu et al., 2023). Recent broiler nutrition studies have mainly tested supplements delivered before or during chronic or cyclic heat challenge, including phlorotannin, resveratrol, rutin, and puerarin, rather than interventions initiated only at recovery onset (Zhao et al., 2024; Ding et al., 2025; Ma et al., 2025; Xu et al., 2025). Reviews have also summarized phytogenic strategies for heat-stress mitigation in broilers (Oni et al., 2024).

We established an acute heat exposure and programmed recovery model in yellow-feathered broilers at 80 d of age to test whether treatment initiated at recovery onset could limit residual liver and lung injury across physiological, histological, biochemical, and selected molecular endpoints. An exploratory composite score was used only to rank doses within the present study.

To complement the in vivo study, we used the chicken embryonic fibroblast cell line DF-1 as a simplified avian heat and rewarming model to examine nuclear factor erythroid 2-related factor 2 (NRF2) responses. Avian cell studies have supported oxidative stress readouts linked to NRF2 in DF-1 fibroblasts and chicken hepatocytes (Gao et al., 2023; Yang et al., 2024). To align the cell model with the in vivo treatment window, BAPF exposure in DF-1 cells began at recovery onset after the heat phase, whereas ML385 was retained as a pretreatment arm for pathway interrogation.

Materials and methods

Broilers, housing, and experimental design

All animal procedures were reviewed and approved by the Institutional Animal Care and Use Committee of Shanxi Agricultural University (approval no. SXAU-EAW-2025SD0237). In vivo experiments were conducted in an accredited poultry facility (license no.: SYXK (Shanxi) 2025-0010) in October 2025. Study planning followed the PREPARE guidelines, reporting adhered to ARRIVE 2.0, and animal care and use were conducted in accordance with the Guide for the Care and Use of Agricultural Animals in Research and Teaching, 4th ed. (Smith et al., 2018; FASS, 2020; Percie du Sert et al., 2020a, b).

Humane endpoints were prespecified and included severe respiratory distress, inability to stand, loss of righting reflex or moribund state, marked feed or water refusal, severe dehydration, rapid body weight loss > 15%, or other irreversible severe suffering. Broilers meeting any criterion would be humanely euthanized and excluded from further procedures. No broilers met these criteria. Broilers selected for terminal sampling or reaching a humane endpoint were euthanized by cervical dislocation performed by trained personnel. Death was confirmed by the absence of respiration and heartbeat and by fixed and dilated pupils. No survival surgery or other invasive operative procedure requiring anesthesia was performed in this study.

A total of 144 healthy yellow-feathered broilers at 80 d of age were obtained from a single flock in Shanxi Province, North China. Broilers at 80 d of age were selected to represent a late growing stage of yellow-feathered broilers in this production context, at which residual organ injury after acute heat exposure may be directly relevant to market age management. The cohort was mixed sex, with 70 males and 74 females, and body weight variation was less than 10%.

Upon arrival, broilers were acclimated for 7 days under thermoneutral conditions (24 ± 1 °C) with ad libitum access to feed and water. During acclimation and throughout the experimental period, all broilers received the same basal diet; ingredient formulation and calculated nutrient levels are provided in Additional file 1 (Supplementary Tables S1 and S2). A commercial 5% premix for Three-yellow chickens (T511; product standard no. Q/320481ZCK004-2023; Liyang Zhengchang Feed Technology Co., Ltd., Liyang, Jiangsu, China) was included at the recommended 5% (wt/wt). Batch traceability details are provided in Additional file 2.

The cage was the experimental unit for growth performance, feed intake, and mean body weight per cage. The 8 groups were the thermoneutral control (CON), heat stress model (HS), HS plus carbasalate calcium (HS+CCA) at 100 mg/kg body weight per day [carbasalate calcium soluble powder, 50% formulation; item no. KaBPLG; Weiqianfang brand, China; veterinary drug approval no. 120642313], and HS plus the flavonoid extract from Bupleurum aerial parts (HS+BAPF) at 20, 30, 46, 68, or 102 mg/kg body weight per day. Broilers in the CON group were maintained under thermoneutral conditions and received vehicle (purified water) gavage only. All groups except CON were subjected to a single acute heat challenge followed by programmed cooling. Oral treatments began at recovery onset on Day 0 and were repeated once daily for the next 3 days. Group codes remained masked until completion of laboratory measurements and statistical analysis.

Acute heat challenge, recovery protocol, and oral administration

The acute heat challenge was applied in an environmental chamber with programmable temperature and humidity control (air conditioning unit plus humidity controlled humidifier HR-30-K; Shanghai Yilijie Industrial Co., Ltd., Shanghai, China). Except for the CON group, all cages were placed in the chamber simultaneously. After a 4-h thermoneutral baseline, ambient temperature was raised from 24 °C to 35 °C, maintained for 6 h, and then gradually reduced from 35 °C to 24 °C over the next 6 h. This acute challenge was chosen to generate a clear but survivable heat load and to preserve the ability to evaluate the first hours after heat withdrawal as a defined intervention window at recovery onset. The experimental design and sampling timeline are summarized in Fig. 1.

Fig. 1.

Experimental design and sampling timeline for the broiler heat exposure and recovery study. A total of 144 yellow-feathered broilers (80 d old) were assigned by randomization stratified by body weight to CON, HS, HS+CCA, and HS+BAPF at 20, 30, 46, 68, or 102 mg/kg·d. All groups except CON underwent a 4-h thermoneutral baseline, 6 h at 35 °C, and a 6-h programmed return to 24 °C. Treatments began at recovery onset and continued once daily for 3 additional days. Sampling was performed on Day 0 (end of recovery) and Day 4. Abbreviations: CON, thermoneutral control; HS, heat stress model; CCA, carbasalate calcium (50% formulation); BAPF, flavonoid extract from Bupleurum aerial parts.

Water was provided ad libitum, and broilers were handled to minimize additional physical stress. Temperature and relative humidity inside the chamber were continuously monitored at 3 fixed positions (air conditioner outlet, near the humidifier, and near the door) using a Bluetooth temperature humidity data logger (TempU, model K3; Shenzhen Youkong Electrical Co., Ltd., Shenzhen, China) at 1 min intervals. Chamber records are summarized in Additional file 1.

Time was indexed from the start of the protocol. Baseline refers to the 4-h thermoneutral preheat period; T10 to T12 h refers to the end of the heat phase and early cooling window; and T14 to T16 h refers to late recovery after return to thermoneutral conditions. Panting with the beak open was recorded as an index of severe heat dissipation behavior by counting breaths for 15 s and multiplying by 4. These data are shown descriptively because only a subset of cages was observed at each time point (n = 2 per group).

Within 1 h after recovery began, broilers received the assigned oral treatments (Day 0). The CON and HS groups received purified water, the BAPF groups received 20, 30, 46, 68, or 102 mg/kg·d BAPF in purified water, and the HS+CCA group received carbasalate calcium (50% formulation) at 100 mg/kg·d. Treatments then continued once daily for 3 additional days under thermoneutral conditions (24 ± 1°C) without further heat challenge. On Day 0 (end of recovery; ∼5 to 6 h after gavage), blood was collected from the wing vein for early serum endpoints. Unless otherwise stated, the prespecified broiler from each cage was sampled again on Day 4 after an overnight fast for terminal tissue and serum measurements.

BAPF preparation and quality control

The flavonoid extract was prepared from dried stems and leaves of Bupleurum scorzonerifolium Willd. sourced from Shanxi Lvzhijin Pharmaceutical Co., Ltd. (Jincheng, Shanxi, China). No voucher specimen was assigned at procurement. Species identity was therefore supported by supplier traceability, expert morphological authentication, and batch-level chemical quality control.

Batch consistency across 6 independent production batches (C1 to C6) was assessed by HPLC fingerprint similarity (Fig. 2C). Marker compound quantification is summarized in Table 1 and Supplementary Table S29 (Additional file 3). Chemical consistency was further evaluated by HPLC using a C18 column (250 × 4.6 mm, 5 μm), with a gradient of solvent A (0.4% phosphoric acid in water) and solvent B (methanol), UV detection at 257 nm, a flow rate of 1.0 mL/min, a column temperature of 30 °C, and an injection volume of 5 μL. Peak identity was verified with a mixed reference standard containing chlorogenic acid, rutin, and quercetin (Fig. 2B). Representative fingerprints and peak assignments are shown in Fig. 2. For batch release, acceptance criteria were chlorogenic acid ≥1.0% (wt/wt) and combined rutin plus quercetin ≥4.5% (wt/wt). Only batches meeting these criteria were used for in vivo and in vitro experiments.

Fig. 2.

BAPF extraction workflow and quality control. (A) Schematic of BAPF preparation from Bupleurum aerial parts. (B) HPLC chromatogram of mixed reference standards showing chlorogenic acid (peak 1), rutin (peak 2), and quercetin (peak 3). (C) HPLC fingerprints of 6 independent production batches (C1 to C6), normalized and stacked to highlight shared retention time profiles. Quantitative marker contents are presented in Table 1 and Supplementary Table S29 (Additional file 3). Abbreviations: BAPF, flavonoid extract from Bupleurum aerial parts; HPLC, high performance liquid chromatography; QC, quality control.

Table 1.

Batch release criteria for BAPF marker compounds.

Index	Acceptance criteria	Range across 6 batches
Chlorogenic acid	≥1.0% (wt/wt)	1.1% to 1.7%
Rutin + quercetin, sum	≥4.5% (wt/wt)	4.6% to 5.2%

Values are ranges across 6 independent production batches (C1 to C6). Acceptance criteria indicate release thresholds (wt/wt). Rutin and quercetin are reported as summed content. Abbreviations: BAPF, flavonoid extract from Bupleurum aerial parts; wt/wt, weight/weight.

Leg band coding and sample collection

Broilers were individually identified by leg bands coded by color and were assigned in advance within each cage to specific sampling endpoints before the experiment. This coding maintained a fixed sampling schedule and kept laboratory analyses blinded. On Day 0, at the end of the 6 h recovery period (∼5 to 6 h after gavage), blood was collected from the wing vein. Samples were allowed to clot, centrifuged at 3,500 × g for 15 min at 4 °C, and the serum was stored at −80°C until assay.

Histopathological examination

After fixation for 24 to 48 h, liver and lung samples were dehydrated through graded ethanol, cleared in xylene, embedded in paraffin, sectioned (∼4 μm), deparaffinized, rehydrated, and stained with hematoxylin and eosin (H&E). Liver slides were examined under a light microscope (Olympus BX53, Olympus, Tokyo, Japan) for hepatic cord arrangement, hepatocellular degeneration and necrosis, inflammatory cell infiltration, sinusoidal congestion, and bile duct lesions (Tang et al., 2022). For the lung, bronchial and alveolar structures, interstitial edema, capillary congestion, and inflammatory cell infiltration were evaluated (Wu et al., 2023). Lesions were scored semiquantitatively. Liver injury was graded with an adapted Suzuki score (0 to 4 each for sinusoidal congestion, hepatocyte cytoplasmic vacuolization, and parenchymal necrosis; total score 0 to 12); the full rubric is provided in Table S23A (Additional file 4).

Lung injury was assessed with a prespecified rubric based on edema, congestion, structural disruption, and inflammatory cell infiltration. The scoring criteria are provided in Supplementary Table S28 in Additional file 4. Three broilers per group were scored under blinded conditions. Six non-overlapping fields were randomly selected per bird, and field scores were averaged to yield one value per bird.

Lung wet to dry ratio

To evaluate pulmonary edema, the right lung was excised immediately after euthanasia as the standardized sample. Attached vessels and bronchi were trimmed, and surface moisture was removed with filter paper before the wet weight was recorded. The tissue was then dried in a forced air oven (DHG-9053A; Shanghai Yiheng Scientific Instrument Co., Ltd., Shanghai, China) at 60 °C for 72 h to obtain the dry weight. The lung wet-to-dry ratio was calculated as wet weight divided by dry weight. Wet and dry weights were measured on a calibrated analytical balance (LE204E; METTLER TOLEDO Co., USA; readability 0.1 mg), and drying continued until constant mass was confirmed (Δ < 1 mg on 2 consecutive measurements). Individual wet and dry weights, together with derived ratios, are provided in Additional file 4 (Supplementary Table S3).

Oxidative stress and serum biochemical assays

Serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), and lactate dehydrogenase (LDH) activities were measured with commercial enzymatic/colorimetric kits (AKAM006M, AKAM019M, and AKCO003M; Beijing Boxbio Science & Technology Co., Ltd., Beijing, China) on a Spark 10 M multimode microplate reader (Tecan, Männedorf, Switzerland; SparkControl v2.3) and are reported as units per liter. Hepatic malondialdehyde (MDA), superoxide dismutase (SOD), and glutathione peroxidase (GPx) were measured in Day 4 liver homogenates using kits from the same manufacturer (AKFA013M, AKAO001M, and AKPR014M). Values were normalized to protein content and are reported as nanomoles per milligram protein for MDA and units per milligram protein for SOD and GPx. Standard curves were generated from standards supplied with the kits, and enzyme activities were reported according to the manufacturers’ instructions. For serum ALT and AST, each sample was assayed in duplicate wells on 2 plates (P1 and P2) run on the same day, and the final per bird value was the mean of the 2 plate means. Serum LDH was assayed in duplicate wells on a single analytical plate, and the mean of the duplicate wells was used per bird. For hepatic MDA, SOD, and GPx, each biological sample was assayed in duplicate wells on a single analytical plate, with quality control material run in parallel for each plate; technical replicates were averaged within each sample and were not treated as independent observations. Raw OD and absorbance values, duplicate well data, per bird processed values, group summaries, and concise quality control information for these assays are provided in Additional file 5 (Supplementary Tables S8 to S17).

Serum cytokines by ELISA

Chicken interleukin-1β (IL-1β) and interleukin-18 (IL-18) in serum were quantified with commercial ELISA kits, including Chicken IL-1β Quantitative Detection Kit (ELISA; cat. no. YPJV1888; UpingBio Technology Co., Ltd., Hangzhou, Zhejiang, China) and Chicken IL-18 Detection Kit (cat. no. YPY31629; UpingBio Technology Co., Ltd., Hangzhou, Zhejiang, China), according to the manufacturers’ protocols. Kit details and raw OD450 and quality-control data are provided in Additional file 5 (Supplementary Tables S13 to S15). Standard curves were generated from serially diluted standards, and appropriately diluted samples were incubated in antibody-coated plates with the corresponding detection reagents. Absorbance was read at 450 nm, and concentrations were calculated from the standard curves. For the competitive IL-1β assay, optical density was inversely related to analyte concentration. All samples were assayed in duplicate, and the mean per bird was used for analysis.

Western blot analysis

Frozen liver samples stored at −80 °C were homogenized in RIPA lysis buffer (cat. no. G2002; Servicebio, Wuhan, China) supplemented with a protease inhibitor cocktail (cat. no. G2006) and phosphatase inhibitor (cat. no. G2007) (Servicebio, Wuhan, China), lysed on ice for 30 min, and centrifuged at 12,000 × g for 15 min at 4 °C. Supernatants were collected, and protein concentration was determined using a BCA protein assay kit (cat. no. PK10026; Proteintech, Rosemont, IL, USA). Equal protein amounts (30 to 40 μg per lane; matched within each blot) were mixed with loading buffer, boiled for 5 min, separated by SDS-PAGE, and transferred to PVDF membranes. Membranes were blocked with 5% non-fat milk in TBST for 1 h at room temperature.

For the representative immunoblots, 1 lane corresponded to 1 biological replicate (n = 3 broilers per group; CON, HS, HS+BAPF30, and HS+BAPF68). Membranes were incubated overnight at 4 °C with primary antibodies against NRF2 (cat. no. 16396-1-AP), HO-1 (cat. no. 10701-1-AP), NLRP3 (cat. no. 68102-1-Ig), cleaved caspase-3 (cat. no. 25128-1-AP), GSDME and cleaved GSDME (cat. no. 31363-1-AP), and β-actin (cat. no. 66009-1-Ig) (Proteintech, Rosemont, IL, USA). Antibodies were selected based on manufacturer information and preliminary optimization for chicken samples. Primary antibodies were diluted at 1:1,000, and HRP-conjugated secondary antibodies were diluted at 1:5,000. After TBST washes, membranes were incubated for 1 h at room temperature with HRP-conjugated goat anti-rabbit IgG(H + L) (cat. no. SA00001-2) or HRP-conjugated goat anti-mouse IgG(H + L) (cat. no. SA00001-1) secondary antibodies (Proteintech, Rosemont, IL, USA). Bands were visualized using an enhanced chemiluminescence substrate (SignalBright Plus; cat. no. PK10012; Proteintech, Rosemont, IL, USA), imaged on a Tanon 5200 chemiluminescence imaging system (Tanon Science & Technology Co., Ltd., Shanghai, China), quantified in ImageJ, normalized to β-actin, and expressed relative to the CON group.

Quantitative reverse transcription PCR

Total RNA was extracted from RNAlater preserved liver tissues using the RNAprep Pure Tissue Kit (cat. no. GDP431; TIANGEN Biotech, Beijing, China). RNA purity and integrity were checked by agarose gel electrophoresis and the OD260:OD280 ratio. One microgram of total RNA was reverse transcribed into cDNA using FastKing gDNA Dispelling RT SuperMix (cat. no. 4992227; TIANGEN Biotech, Beijing, China). NFE2L2 (NRF2), HMOX1 (HO-1), TLR4, NLRP3, CASP3, CASP7, GSDME, and IL18 were quantified using SuperReal PreMix Plus (SYBR Green) (cat. no. 4992214; TIANGEN Biotech, Beijing, China) on a ViiA 7 Real Time PCR System (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA; cat. no. 4453534). RELA and IL1B were also assessed in exploratory analyses. qPCR reporting followed MIQE 2.0 recommendations (Bustin et al., 2025). GAPDH was used as the reference gene. Cycling conditions were 95 °C for 30 s, followed by 40 cycles of 95 °C for 5 s and 60 °C for 30 s, with melting curve analysis to confirm product specificity. Relative gene expression was calculated using the 2^−ΔΔCt method, with the mean value of the CON group set to 1 for normalization (Livak and Schmittgen, 2001). Primer sequences are provided in Additional file 7. Quantitative PCR comparisons focused on CON, HS, HS+BAPF30, and HS+BAPF68 (n = 6 broilers per group; 1 bird per cage).

DF-1 cell heat stress model and in vitro treatments

To examine NRF2 responses in a simplified avian heat and rewarming model, the chicken embryonic fibroblast cell line DF-1 (UMNSAH/DF-1, ATCC CRL-3586; Shanghai Fuheng Biotechnology, Shanghai, China) was cultured in Dulbecco's modified Eagle medium (DMEM), high glucose (Gibco, Thermo Fisher Scientific, Waltham, MA, USA; cat. no. 11965092) supplemented with 10% fetal bovine serum (cat. no. 10099141C) and 1% penicillin and streptomycin (cat. no. 15140122). Cells were maintained at 39 °C in a humidified 5% CO2 incubator to reflect avian physiological temperature and were passaged every 2 to 3 days. For experiments, cells were seeded in 6 well or 96 well plates and grown to 70% to 80% confluence before treatment.

Cells were assigned to CON, HS, HS+BAPF62.5, HS+BAPF125, and HS+BAPF125+ML385 (2 μM). Control cells were maintained at 39 °C. Cells in the heat challenge groups were exposed to 42 °C for 2 h and then rewarmed at 39 °C for 2 h. In the BAPF groups, the extract was added when recovery began and remained present throughout the 39 °C rewarming phase, with matched vehicle exposure across groups. In the inhibitor group, ML385 was added 1 h before heat exposure, and final DMSO concentration was kept below 0.1% and matched across groups.

Working concentrations were selected from a preliminary screening study using Cell Counting Kit-8 (CCK-8) and MTT assays after 48 h exposure to 0 to 2,000 μg/mL BAPF. Each concentration was tested in quadruplicate wells, with matched blank wells without cells used to correct for extract absorbance (Corrected OD = OD_test −OD_blank). Viability was normalized to vehicle; concentrations ≤62.5 μg/mL were noncytotoxic in both assays, whereas 125 μg/mL caused only a mild decrease after 48 h. Accordingly, 62.5 and 125 μg/mL were selected as lower and higher working concentrations for the heat stress and rewarming experiments and subsequent mechanistic analyses (Additional file 7, Supplementary Table S25).

Nuclear and cytosolic fractions were prepared after rewarming to assess NRF2 localization. The ML385 arm was included in the immunoblot and fractionation assays shown in Fig. 8E and G. Fractionation used a Nuclear and Cytoplasmic Protein Extraction Kit (cat. no. P0027; Beyotime Biotechnology, Shanghai, China), with Lamin B1 (cat. no. 66095-1-Ig) and β-actin (cat. no. 66009-1-Ig) as nuclear and cytosolic markers, respectively.

Fig. 8.

Chicken fibroblast heat and rewarming assays support NRF2 responses. (A) DF-1 heat and rewarming protocol. Cells underwent heat stress at 42 °C for 2 h, followed by rewarming at 39 °C for 2 h. BAPF (62.5 or 125 μg/mL) was added at recovery onset and remained present during rewarming. Where indicated, ML385 (2 μM) was added 1 h before heat exposure (DMSO < 0.1%). (B) In vitro group allocation. CON, HS, and both BAPF dose groups were used for ROS quantification and imaging, whereas the ML385 arm was included in immunoblot and fractionation assays. (C) Intracellular ROS quantified by DCFH-DA flow cytometry as mean fluorescence intensity (MFI). Data are mean ± SEM from 3 independent experiments. Statistics for C: one-way ANOVA with Dunnett test versus HS. (D) Hoechst 33342 and Annexin V-PE staining for phosphatidylserine externalization (representative fields; scale bar = 50 μm). (E) Immunoblots of NLRP3, GSDME, cleaved GSDME, cleaved caspase-1, caspase-3, cleaved caspase-3, and HO-1, with β-actin as loading control (CON, HS, and HS+BAPF 125 μg/mL with or without ML385 (2 μM), as indicated; representative blots; n = 3). (F) Validation of transient FLAG-chGSDME-HA overexpression in HEK293T and DF-1 cells using anti-FLAG; GAPDH served as loading control. NC indicates untransfected, OE indicates transient overexpression, and 62.5 or 125 indicates BAPF (μg/mL) in OE cells. (G) NRF2 abundance in cytosolic and nuclear fractions under the indicated BAPF concentrations in the presence of ML385 (2 μM); β-actin and Lamin B1 served as fraction markers (representative blots; n = 3). The 2 BAPF concentrations were selected from preliminary CCK-8 and MTT screening, in which 125 μg/mL caused only a mild viability decrease (Additional file 7: Supplementary Table S25). For C, symbols denote adjusted P values versus HS: ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001 and ****P ≤ 0.0001. Panel C source data (biological replicate MFI values) are provided in Additional file 7 (Supplementary Table S26). Uncropped blots are provided in Additional file 10. Abbreviations: DF-1, chicken embryonic fibroblast cell line; ML385, nuclear factor erythroid 2-related factor 2 inhibitor; DMSO, dimethyl sulfoxide; ROS, reactive oxygen species; DCFH-DA, 2′,7′-dichlorodihydrofluorescein diacetate; MFI, mean fluorescence intensity; HO-1, heme oxygenase 1; NRF2, nuclear factor erythroid 2-related factor 2; HEK293T, human embryonic kidney 293T cells; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; NC, untransfected control; OE, transient overexpression; CCK-8, Cell Counting Kit-8; CON, control; HS, heat stress; BAPF, flavonoid extract from Bupleurum aerial parts; SEM, standard error of the mean.

To improve detection of chicken GSDME processing, HEK293T cells (ATCC CRL-3216) and DF-1 cells (UMNSAH/DF-1, ATCC CRL-3586) were transiently transfected with a dual-tagged FLAG-chGSDME-HA construct 24 h before heat stress. A summary table of the transfection conditions and uncropped immunoblot source images is provided in Additional file 10. Overexpression was verified by anti-FLAG immunoblotting (Fig. 8F and Additional file 10), and cleavage was inferred from FLAG-positive N-terminal fragments together with reduced full-length HA-tagged chGSDME, consistent with prior work (Chen et al., 2025).

Intracellular ROS was quantified by flow cytometry after staining with the Reactive Oxygen Species Assay Kit containing DCFH-DA (cat. no. S0033; Beyotime Biotechnology, Shanghai, China). Cells were incubated with probe for 20 min at 39 °C, washed, and analyzed immediately on a NovoCyte 2060R flow cytometer (Agilent Technologies, Santa Clara, CA, USA; part no. 2010004AA). Flow cytometry data were analyzed in NovoExpress using the gating strategy provided in Additional file 7. After subtraction of background from samples without probe acquired under identical settings, ROS levels were expressed as mean fluorescence intensity. Phosphatidylserine externalization was additionally visualized after rewarming using Hoechst 33342 live-cell staining solution (cat. no. C1027) and Annexin V-PE apoptosis detection kit (cat. no. C1065) (Beyotime Biotechnology, Shanghai, China), and representative fields are shown in Fig. 8D.

For NRF2, HO-1, NLRP3, cleaved caspase-3, and cleaved GSDME analysis, cells were washed with ice-cold PBS after treatment, lysed in inhibitor-supplemented RIPA buffer on ice for 30 min, and centrifuged at 12,000 × g for 15 min at 4 °C. Supernatants were used for protein quantification and Western blotting as described above. Target proteins included NRF2, HO-1, NLRP3, cleaved caspase-1 (p20), cleaved caspase-3, and GSDME (full length and cleaved forms), with β-actin as the loading control. Immunoblot signals were quantified by densitometry in ImageJ, normalized to β-actin for total lysates or to Lamin B1 and β-actin for nuclear and cytosolic fractions, respectively, and expressed relative to the indicated reference group. These normalized immunoblot signals, together with ROS mean fluorescence intensity and Annexin V imaging, were used to evaluate antioxidant responses and pyroptosis-related signaling. Representative DF-1 blots are shown. All in vitro experiments were repeated independently 3 times, with 3 parallel wells per treatment in each experiment.

Exploratory composite score and dose range analysis

To rank doses within this study, we calculated an exploratory composite efficacy score (ICOMP) for the 5 BAPF dose groups (20, 30, 46, 68, and 102 mg/kg·d). ICOMP integrated 12 endpoints covering residual hyperthermia during recovery, lung wet-to-dry ratio, serum injury markers, hepatic redox indices, circulating cytokines, and liver and lung histopathology scores. Higher values indicate better overall recovery. ICOMP was prespecified as an exploratory index for dose ranking and was not used as a confirmatory primary endpoint. Analyses of individual endpoints remained the basis for biological interpretation, whereas ICOMP was used only to summarize the consistency of responses across heterogeneous recovery related outcomes within this study.

For endpoints in which higher values indicate worse outcomes (all endpoints except SOD and GPx; including cloacal temperature, lung wet-to-dry ratio, ALT, AST, LDH, MDA, IL-1β, IL-18, and histopathology scores), the normalized inhibition at dose d was defined as:

$$\begin{array}{c}\hfill I_i\left(d\right)=\frac{X_{i,\mathrm{HS}}-X_i\left(d\right)}{X_{i,\mathrm{HS}}-X_{i,\mathrm{CON}}}\times 100\%\end{array}$$ (Eq. 1)

For endpoints in which higher values indicate better antioxidant capacity (SOD and GPx), the normalized improvement at dose d was defined as:

$$\begin{array}{c}\hfill I_i\left(d\right)=\frac{X_i\left(d\right)-X_{i,\mathrm{HS}}}{X_{i,\mathrm{CON}}-X_{i,\mathrm{HS}}}\times 100\%\end{array}$$ (Eq. 2)

In the equations above, Xi,HS and Xi,CON denote the HS and CON group means, respectively, whereas Xi(d) denotes the individual value at dose d for the given endpoint. Each Ii(d) value was bounded to 0 to 100%, and the composite efficacy score was calculated as the unweighted mean across available normalized endpoints.

$$\begin{array}{c}\hfill \mathrm{ICOMP}\left(d\right)=\frac{I_1\left(d\right)+I_2\left(d\right)+\cdots +I_m\left(d\right)}{m}\end{array}$$ (Eq. 3)

The relation between dose and response was modeled with a 4-parameter logistic (4PL) function with variable slope, where X = log₁₀(dose, mg/kg·d) and Y = ICOMP(d):

$$\begin{array}{c}\hfill Y=\mathrm{Bottom}+\frac{\mathrm{Top}-\mathrm{Bottom}}{1+{10}^{(\mathrm{Log}E{D}_{50}-X)\times \mathrm{HillSlope}}}\end{array}$$ (Eq. 4)

A 4-parameter logistic model with variable slope was fitted to ICOMP against log₁₀ dose. ED₅₀, ED₉₀, and ED₉₅ were derived from this exploratory model and used only for dose prioritization within this study. To assess whether the prioritized range was driven by any single endpoint domain, sensitivity analyses were performed by recalculating ICOMP after excluding selected endpoint categories and refitting the dose response model.

Statistical analysis

Statistical analyses were performed using SPSS 26.0 (IBM, Armonk, NY, USA) and GraphPad Prism (GraphPad Software, San Diego, CA, USA). Normality and homogeneity of variance were assessed with the Shapiro-Wilk and Levene’s tests, respectively. For single-factor comparisons, one-way ANOVA was used when assumptions were met, followed by Tukey multiple comparisons for all pairwise contrasts or Dunnett test with HS as the prespecified reference. Otherwise, the Kruskal-Wallis test followed by Dunn multiple comparisons was applied. For variables measured across multiple time points, repeated-measures ANOVA was used with time as the within-subject factor. The individual bird was the subject for cloacal temperature, whereas the cage was the experimental unit for variables measured at the cage level.

Panting data recorded at the cage level were summarized descriptively and plotted to illustrate overall trends (n = 2 cages per group per time point; subset) rather than used for formal inference at each time point. The relation between dose and response and the corresponding ED values were estimated by nonlinear regression using the 4-parameter logistic model described above. Allocation was stratified by body weight. Data are presented as mean ± standard error of the mean (SEM) unless otherwise stated. The cage (n = 6 per group) was the experimental unit for growth performance and feed intake measured per cage. For most other in vivo assays, 1 sampled bird per cage (n = 6 per group) was used, whereas histopathology was assessed in 3 broilers per group. For Fig. 9, exploratory indices were calculated per individual using available endpoints; each dot in Fig. 9B represents 1 ICOMP value per bird (n = 6 broilers per group). For in vitro assays, technical replicates were averaged within each independent experiment, and the independent experiment served as the experimental unit. Multiplicity was controlled within each endpoint using familywise error rate-adjusted post hoc procedures. Environmental monitoring data were used descriptively for quality control because measurements collected every minute are autocorrelated. P < 0.05 was used as the decision threshold, and nominal P values are reported. For qPCR, inferential statistics were performed on ΔCt values (Ct_target −Ct_GAPDH), whereas 2^−ΔΔCt values are shown for visualization. In the DF-1 cytotoxicity screen, treatment versus vehicle comparisons were evaluated by one-way ANOVA with Dunnett test. Graphs were generated in GraphPad Prism 9.5.1.

Fig. 9.

Exploratory composite efficacy scoring and analysis of dose response for BAPF. (A) Heatmap of normalized endpoint inhibition or improvement across up to 12 endpoints, with CON and HS as anchors (0% = HS; 100% = CON). SOD and GPx were treated as benefit type endpoints; all others were treated as injury type endpoints. (B) 4-parameter logistic regression of per bird ICOMP versus log10 BAPF dose (20, 30, 46, 68, and 102 mg/kg·d). Each dot represents 1 per bird ICOMP value (n = 6 broilers per group; 1 bird per cage). ED₅₀, ED₉₀, and ED₉₅ were derived from the fitted curve and should be interpreted only as exploratory dose indices within this study. The HS+CCA group is shown as a benchmark comparator but was excluded from curve fitting. (C) Relative endpoint contributions to ICOMP at each BAPF dose. Source data and sensitivity analyses are provided in Additional file 9. Abbreviations: ICOMP, composite efficacy score; BAPF, flavonoid extract from Bupleurum aerial parts; SOD, superoxide dismutase; GPx, glutathione peroxidase; CON, thermoneutral control; HS, heat stress model; CCA, carbasalate calcium; 4PL, 4-parameter logistic; ED₅₀, ED₉₀, and ED₉₅, effective doses corresponding to 50%, 90%, and 95% of the fitted maximal response.

Results

HPLC fingerprints show consistent batch quality of the BAPF preparation

Before efficacy testing, we confirmed batch consistency of the preparation. Mixed reference standards identified chlorogenic acid, rutin, and quercetin as characteristic peaks (Fig. 2B). Across 6 production batches (C1 to C6), the overall fingerprint profiles remained highly similar (Fig. 2C). Quantitative marker analysis was consistent with the fingerprint results. Chlorogenic acid ranged from 1.1% to 1.7%, and the combined rutin plus quercetin content ranged from 4.6% to 5.2%, all within the predefined acceptance criteria (Table 1; Supplementary Table S29, Additional file 3). Overall, all 6 batches met the prespecified release criteria.

Chamber conditions and physiological responses during heat exposure and recovery

The programmed heat exposure and recovery protocol matched the intended temperature and humidity profile and imposed a clear heat load (Fig. 3A). Cloacal temperature remained elevated in HS broilers through the end of heat exposure and into early recovery. Panting with the beak open increased during heat loading and eased during recovery; however, these behavioral observations were interpreted descriptively because only a subset of cages was observed at each time point.

Fig. 3.

Heat load verification and physiological responses during heat exposure and recovery. (A) Chamber temperature (°C) and relative humidity (%) recorded every minute throughout the challenge and recovery protocol. (B) Feed intake per cage (%) relative to each cage baseline over the heat exposure and recovery period. (C) Lung wet-to-dry ratio on Day 4 after an overnight fast. (D) Residual recovery hyperthermia, calculated as the mean difference from baseline cloacal temperature (°C), with bootstrap 95% confidence intervals. (E) Initial body weight at allocation. (F) Panting with the beak open across observation times during heat exposure and recovery, shown descriptively as a heat map. Data in B and C are mean ± SEM; dots in C and E indicate individual broilers. In C, symbols denote adjusted P values versus HS: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001. Panting was recorded descriptively in a subset of cages (n = 2 per group per time point). Raw chamber logs, feed intake data, exact P values, and source data are provided in Additional file 1. Abbreviations: CON, thermoneutral control; HS, heat stress model; HS+CCA, HS plus carbasalate calcium (50% formulation, 100 mg/kg·d); BAPF, flavonoid extract from Bupleurum aerial parts; SEM, standard error of the mean; CI, confidence interval.

Feed intake per cage, expressed as a percentage of baseline, declined after heat exposure and showed a numerical recovery pattern in the CCA and BAPF groups over the observation period. Because feed intake was measured at the cage level over a short recovery window, this endpoint was used as supportive physiological context rather than as a primary efficacy conclusion. Exact P values and cage level feed intake source data are provided in Additional file 1; no efficacy conclusion was based on this supportive endpoint from this short recovery window.

BAPF improves liver and lung histopathology during recovery

Liver and lung sections from CON broilers showed largely preserved architecture, with only occasional mild background changes. In contrast, acute heat exposure followed by programmed recovery produced clear liver and lung lesions in HS broilers (Fig. 4). Livers from HS broilers showed disorganized hepatic cords, ballooning degeneration, nuclear pyknosis and karyolysis, inflammatory infiltration, and sinusoidal congestion. Lungs from HS broilers showed thickened alveolar septa, edema, capillary congestion, and inflammatory cell accumulation around bronchi and within the interstitium; accordingly, both liver and lung injury scores were higher in HS than in CON.

Fig. 4.

BAPF alleviates liver and lung histopathological injury after heat exposure and recovery. (A) Representative H&E stained liver sections from CON, HS, HS+CCA, and HS+BAPF (20, 30, 46, 68, and 102 mg/kg·d) collected on Day 4 after an overnight fast. (B) Representative H&E stained lung sections from the same groups and time point. Scale bar = 100 μm. (C) Liver injury score (adapted Suzuki score, 0 to 12). (D) Lung injury score (0 to 20). Histological scoring was performed in 3 broilers per group. Six non-overlapping fields per bird were scored under blinded conditions and averaged to yield 1 value per bird. Dots indicate individual broilers; bars show mean ± SEM. Statistics: Kruskal-Wallis test with Dunn multiple comparisons versus HS. ns, P > 0.05; ****P ≤ 0.0001. Raw scoring data and rubrics are provided in Additional file 4. Abbreviations: H&E, hematoxylin and eosin; CON, thermoneutral control; HS, heat stress model; CCA, carbasalate calcium; BAPF, flavonoid extract from Bupleurum aerial parts; SEM, standard error of the mean.

These lesions were attenuated across doses. Within the lower dose range, improvement was more evident at 30 mg/kg·d than at 20 mg/kg·d. Responses were most consistent at 46 and 68 mg/kg·d, whereas 102 mg/kg·d also improved the phenotype without a clear added benefit. Likewise, lungs from broilers receiving 46 or 68 mg/kg·d BAPF showed better preserved alveolar structures, less interstitial edema, and fewer inflammatory cells than HS. The 102 mg/kg·d group also improved, but the response was not clearly greater than that at 46 or 68 mg/kg·d. The HS+CCA group showed a similar overall pattern.

Serum biochemistry, hepatic redox status, and cytokine responses during recovery

Serum biochemical markers were measured on Day 0 (end of recovery), and hepatic redox status was assessed on Day 4 after an overnight fast. Relative to CON, HS increased serum ALT, AST, and LDH (U/L) and hepatic MDA (nmol/mg protein), whereas hepatic SOD and GPx activities (U/mg protein) were lower (Fig. 5A-D, G, and H).

Fig. 5.

Serum biochemistry, hepatic redox indices, and circulating cytokines after heat exposure and recovery. Serum ALT (A), AST (B), and LDH (C), hepatic MDA (D), circulating IL-18 (E) and IL-1β (F), and hepatic SOD (G) and GPx (H) were measured in CON, HS, HS+CCA, and graded BAPF groups (20, 30, 46, 68, and 102 mg/kg·d). Serum endpoints were assessed on Day 0 at the end of recovery, and hepatic redox endpoints on Day 4 after an overnight fast. Dots show individual broilers with mean ± SEM. vs. HS: ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Exact P values and source data are in Additional file 5. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; LDH, lactate dehydrogenase; MDA, malondialdehyde; SOD, superoxide dismutase; GPx, glutathione peroxidase; IL-18, interleukin-18; IL-1β, interleukin-1β; CON, thermoneutral control; HS, heat stress model; CCA, carbasalate calcium; BAPF, flavonoid extract from Bupleurum aerial parts; SEM, standard error of the mean.

Compared with HS, BAPF attenuated these changes. Hepatic MDA was lower at 46 and 68 mg/kg·d (P = 0.003 and P < 0.001, respectively), whereas hepatic SOD and GPx activities were higher at the same doses (SOD: P = 0.004 and P < 0.001; GPx: P = 0.039 and P < 0.001, respectively; Fig. 5D, G, H). Serum ALT and AST were likewise lower at 46 and 68 mg/kg·d than in HS and moved toward control values (both ALT comparisons, P < 0.001; both AST comparisons, P < 0.001; Fig. 5A, B). Serum lactate dehydrogenase activity also decreased relative to HS, although the response was not strictly monotonic. Improvement was evident across the intermediate to high doses, with a slight plateau or rebound at the highest dose. The HS+CCA group showed a comparable overall pattern of benefit (Fig. 5).

BAPF lowers circulating interleukin-18 and interleukin-1β during recovery

Circulating IL-18 and IL-1β concentrations were measured by ELISA on Day 0 at the end of recovery. Both cytokines were higher in HS than in CON (Fig. 5E, F). Compared with HS, circulating IL-18 was lower at 46 and 68 mg/kg·d (P = 0.003 and P < 0.001, respectively), and IL-1β was lower at the same doses (P = 0.001 and P < 0.001, respectively), with the lowest levels generally observed at 68 mg/kg·d (Fig. 5E, F). For IL-1β, the 102 mg/kg·d group showed a modest rebound but remained below HS. HS+CCA showed lower circulating IL-1β than HS, whereas IL-18 was numerically lower but did not reach adjusted significance after multiple comparison correction. Raw ELISA data, including OD450 readings, calculated concentrations, and quality control metrics, are provided in Additional file 5 (Supplementary Tables S13 to S15).

BAPF shifts hepatic molecular readouts toward the control pattern on Day 4

To determine whether these phenotypic improvements were accompanied by hepatic molecular changes, Day 4 RT-qPCR and immunoblotting focused on 30 and 68 mg/kg·d, representing a lower-response dose and a higher dose near the prioritized range. Because 46 mg/kg·d was not included in this molecular subset, these data were interpreted as evidence supporting the pathway rather than as a complete mechanistic dose response analysis across the full 46 to 68 mg/kg·d range.

Relative to CON, HS shifted hepatic readouts toward lower NRF2 and HO-1 signaling together with higher expression of markers linked to injury, including TLR4, NLRP3, CASP3, CASP7, GSDME, and IL18. On the ΔCt-based analysis, hepatic NLRP3 expression was lower in the 30 and 68 mg/kg·d groups than in HS (P = 0.001 and P < 0.001, respectively), and the 68 mg/kg·d group showed the clearest overall shift back toward the CON pattern (Fig. 6). Immunoblotting showed concordant directional changes in NRF2, HO-1, NLRP3, cleaved caspase-3, and cleaved GSDME, and the 68 mg/kg·d group had lower CASP3 transcript abundance than HS on the ΔCt-based analysis (P < 0.001), consistent with the reduced cleavage-associated protein pattern seen in liver (Fig. 6, Fig. 7). Because RELA mRNA did not differ among groups and HMOX1 mRNA and HO-1 protein were not fully concordant, we interpret upstream priming and HO-1 involvement cautiously.

Fig. 6.

Hepatic transcriptional responses linked to antioxidant defense and signaling related to the inflammasome. Reverse transcription quantitative PCR analysis of NFE2L2 and HMOX1 mRNA, together with TLR4, NLRP3, CASP3, CASP7, GSDME, and IL18 mRNA, in liver collected on Day 4 after an overnight fast. Expression values are shown for visualization only; statistics were performed on ΔCt values. Groups shown: CON, HS, HS+BAPF30, and HS+BAPF68 (n = 6 broilers per group; 1 bird per cage). Dot plots show individual values with mean ± SEM. Statistics: one-way ANOVA followed by Dunnett multiple comparisons versus HS. Symbols denote adjusted P values versus HS: ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; and ****P ≤ 0.0001.Exact P values and supporting qPCR tables are provided in Additional file 7. Abbreviations: NFE2L2, nuclear factor erythroid 2-related factor 2; HMOX1, heme oxygenase 1; TLR4, toll-like receptor 4; NLRP3, NLR family pyrin domain containing 3; CASP3, caspase-3; CASP7, caspase-7; GSDME, gasdermin E; IL18, interleukin-18; ΔCt, delta cycle threshold; CON, thermoneutral control; HS, heat stress model; BAPF, flavonoid extract from Bupleurum aerial parts; SEM, standard error of the mean.

Fig. 7.

Liver immunoblots and densitometric analysis. Representative immunoblots of NRF2, HO-1, and proteins related to pyroptosis in liver collected on Day 4 after an overnight fast. Uncropped blots are provided in Additional file 6. (A) NLRP3, HO-1, and NRF2 immunoblots with β-actin as loading control. (B) Cleaved caspase-3 and cleaved GSDME (GSDME-NT) immunoblots with β-actin. (C-G) Densitometric quantification normalized to β-actin: NLRP3 (C), HO-1 (D), NRF2 (E), cleaved GSDME (F), and cleaved caspase-3 (G). Three broilers per group were randomly selected from different cages (1 bird per cage), and each lane represents 1 biological replicate (n = 3 broilers per group). Dot plots show individual values with mean ± SEM. Densitometric values and exact P values are provided in Additional file 8 (Supplementary Table S27). Groups shown: CON, HS, HS+BAPF30, and HS+BAPF68. Statistics: one-way ANOVA followed by Dunnett multiple comparisons versus HS. Symbols denote adjusted P values versus HS: ns, P > 0.05; *P ≤ 0.05; ***P ≤ 0.001 ****P ≤ 0.0001.. Abbreviations: NRF2, nuclear factor erythroid 2-related factor 2; HO-1, heme oxygenase 1; NLRP3, NLR family pyrin domain containing 3; GSDME-NT, N-terminal cleaved gasdermin E; CON, thermoneutral control; HS, heat stress model; BAPF, flavonoid extract from Bupleurum aerial parts; SEM, standard error of the mean.

In vitro heat and rewarming assays support NRF2 responses

When BAPF was initiated at recovery onset, DF-1 cells showed lower intracellular ROS after rewarming, with the clearer reduction at 125 μg/mL, and Annexin V imaging showed a parallel reduction in phosphatidylserine externalization (Fig. 8 C and D). In the 125 μg/mL arm, representative immunoblots showed lower signals for NLRP3, cleaved caspase-1, cleaved caspase-3, and cleaved GSDME than in HS, whereas fractionation assays showed greater nuclear NRF2 abundance after rewarming. ML385 attenuated the BAPF-associated increase in nuclear NRF2 and weakened the accompanying shift in injury-associated blot patterns, supporting NRF2 pathway involvement in this simplified rewarming model. Because nuclear NRF2 estimates from fractionation are semiquantitative and model dependent, they were interpreted together with ROS fluorescence, total injury related immunoblot signals, and the ML385 response, rather than as standalone proof of NRF2 dependence.

Composite dose-response analysis

For dose prioritization within this study, composite efficacy scores for individual broilers (ICOMP) were fitted to a 4-parameter logistic model (Fig. 9B). The exploratory 4PL fit yielded an ED₅₀ of about 36 mg/kg·d and suggested limited incremental benefit at the highest dose. The composite efficacy score integrated Day 0 thermoregulatory, serum injury, and cytokine endpoints with Day 4 organ-level outcomes. The fitted curve identified 46 to 68 mg/kg·d as the range prioritized for subsequent study, with limited additional gain at 102 mg/kg·d. The HS+CCA group is shown only as a benchmark comparator and was excluded from curve fitting. Sensitivity analyses using alternative composite definitions yielded similar ED₅₀ estimates, about 35.4 to 36.4 mg/kg·d, suggesting that the prioritized dose range was not driven by a single endpoint. Because ICOMP was exploratory and unweighted, the prioritized range should be interpreted as a dose range for subsequent testing rather than as a definitive therapeutic dose.

Discussion

The main finding of this study is that early recovery after heat withdrawal remains a period of ongoing injury rather than a simple return to baseline in yellow-feathered broilers. Despite removal of the external heat load, residual hyperthermia, pulmonary edema, histopathological injury, serum enzyme elevations, and inflammatory cytokine responses persisted into recovery. This pattern is consistent with previous evidence that hepatic mitochondrial dysfunction, ROS production, and lipid peroxidation can persist after acute heat stress is removed in broilers (Yang et al., 2010). However, the present study extends that work by testing an intervention initiated at recovery onset rather than only characterizing injury after stress.

Previous broiler studies have linked heat stress to hepatic oxidative injury, inflammatory signaling, apoptosis, and pulmonary blood-air barrier disruption, mostly in chronic or cyclic models, or models with treatment during exposure (Liu et al., 2022; Tang et al., 2022; Wu et al., 2023; Iliopoulou et al., 2025; Xu et al., 2025). In contrast, the present model separated the heat exposure phase from the intervention phase by initiating BAPF only after heat withdrawal. This design therefore supports the concept that the first hours of recovery constitute a measurable post-heat treatment window, with early serum and thermoregulatory responses detected during the first 6 h after heat withdrawal and residual organ-level injury still detectable on Day 4.

Across endpoints, the most consistent in vivo responses were observed at 46 to 68 mg/kg·d. This range was supported by improvements in individual endpoints, including histopathology, serum injury markers, hepatic redox indices, and circulating cytokines, and was further prioritized by the exploratory 4PL dose response analysis of ICOMP. The fitted model yielded an ED₅₀ of about 36 mg/kg·d, whereas sensitivity analyses using alternative composite definitions gave similar ED₅₀ estimates, supporting the robustness of the dose ranking result. Because ICOMP was exploratory and unweighted, the prioritized range should be interpreted as a dose range for subsequent testing rather than as a definitive therapeutic dose.

Mechanistically, the present results support a response during recovery in which BAPF enhances antioxidant defense associated with NRF2 after heat withdrawal. Heat exposure increased residual oxidative stress and inflammatory signals related to injury, whereas BAPF reduced ROS accumulation, increased nuclear NRF2 abundance in DF-1 cells, and shifted hepatic and cellular readouts toward higher NRF2/HO-1 and lower NLRP3, cleaved caspase-1, cleaved caspase-3, and cleaved GSDME signals. ML385 attenuated the increase in nuclear NRF2 observed with BAPF and weakened the accompanying reduction in protein signals related to injury, supporting NRF2 involvement in this simplified rewarming model. Nevertheless, because ML385 was used as a pretreatment arm and the DF-1 system does not reproduce in vivo absorption, metabolism, tissue microenvironment, programmed cooling and rewarming kinetics, or whole-organ recovery dynamics, these findings should be interpreted as evidence supporting the pathway rather than as definitive proof that NRF2 activation alone caused the organ protection in vivo.

This study has limitations. BAPF is a multicomponent extract, and the active constituents and pharmacokinetic basis of the prioritized dose range remain to be defined. Several mechanistic endpoints were measured only in selected doses or smaller subsets, and the molecular subset did not include 46 mg/kg·d; therefore, the molecular data should be interpreted as evidence supporting the pathway rather than as a complete dose response mechanism. In addition, the DF-1 model cannot fully reproduce in vivo recovery pharmacology, tissue interactions, or repeated heat exposure closer to production conditions. Future studies should test the prioritized dose range under repeated or field-relevant heat challenge and use orthogonal assays to further define NRF2-dependent and pyroptosis-related mechanisms. These proposed relationships are summarized in Fig. 10.

Fig. 10.

Working model of tissue injury during recovery after heat exposure and the proposed action of BAPF. Solid lines indicate supported links. Broken lines indicate inferred links. (A) Recovery after acute heat exposure is proposed to sustain oxidative stress and promote liver injury linked to the inflammasome and GSDME. In lung, comparable molecular involvement is inferred from phenotypic data rather than directly demonstrated. Supported hepatic readouts include changes involving NLRP3, markers in the IL-1β and IL-18 axis, and caspase-3 mediated GSDME cleavage. TLR4 and NF-κB priming cues are shown as contextual rather than demonstrated. (B) BAPF is proposed to enhance antioxidant defense via NRF2 nuclear enrichment, which may lower ROS and attenuate signals of injury. Mechanistic nodes are supported primarily by hepatic and DF-1 data, whereas lung benefit is supported mainly by physiological and histological endpoints. Abbreviations: BAPF, flavonoid extract from Bupleurum aerial parts; NRF2, nuclear factor erythroid 2-related factor 2; ROS, reactive oxygen species; NLRP3, NLR family pyrin domain containing 3; GSDME, gasdermin E; IL-1β, interleukin-1β; IL-18, interleukin-18; TLR4, toll-like receptor 4; NF-κB, nuclear factor kappa B; DF-1, chicken embryonic fibroblast cell line.

Funding

This work was supported by the National Natural Science Foundation of China [grant number 81872962]; the Shanxi Province Basic Research Program [grant number 202503021211078]; and the Research and Development Project on Social Development of Shanxi Province [grant number 201903D321009].

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used Gemini to improve the language and readability of the manuscript. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

CRediT authorship contribution statement

Zehui Wang: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation. Zhenyu Li: Validation, Methodology, Investigation, Data curation. Yuguang Du: Writing – review & editing, Resources, Methodology. Ke Li: Writing – review & editing, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization.

Disclosures

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Ke Li reports financial support was provided by National Natural Science Foundation of China. Ke Li reports financial support was provided by Shanxi Province Basic Research Program. Ke Li reports financial support was provided by Research and Development Project on Social Development of Shanxi Province. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix. Supplementary materials

ADDITIONAL FILE CAPTIONS

Additional file 1. Husbandry, allocation, environmental monitoring, thermoregulation, behavior, and cage-level feed-intake source data.

Additional file 2. Commercial premix specification extract and batch traceability record for T511 used in the basal diet.

Additional file 3. BAPF batch-quality summary and raw HPLC chromatogram support.

Additional file 4. Histopathology scoring rubrics and raw scoring data, together with lung wet-to-dry ratio source data.

Additional file 5. Serum biochemistry, hepatic oxidative-stress assays, and serum cytokine assay source data with concise quality control.

Additional file 6. Uncropped liver immunoblot source images supporting Fig. 7.

Additional file 7. Primer sequences, qPCR source data and quality control, DF-1 cytotoxicity screening, and flow-cytometry ROS source data.

Additional file 8. Western blot densitometry data supporting Fig. 7.

Additional file 9. Source data and sensitivity analyses for the exploratory composite efficacy score shown in Fig. 9.

Additional file 10. Transfection conditions and uncropped immunoblot source images supporting Fig. 8.