The AC was collected from Guangxi Province, China, authenticated by researcher Hailong Liu (Hainan Academy of Agricultural Sciences), and deposited in the veterinary medicine lab of Hebei Agricultural University with voucher specimen number (HEBAU-20201105). The AC was dried, cleaned, and then cut into 2 cm long pieces. The AC pieces were soaked for 1 h in 10 × (v/w) distilled water and then boiled for 2 h and filtered. Distilled water was mixed five times (v/w) to the residue and boiled for another 1 h and also filtered. The two filtrates were collected and merged, and then concentrated to 3 g/ml (crude drugs) to prepare ACD. The ACD was stored at 4°C for experimental use.

An Acquity UPLC H-CLASS (Waters Corporation, MA, USA) was used to further characterize the main components of ACD. Chromatographic separation of ACD was performed on the HSS T3 column (2.1 mm × 100 mm, 1.8 μm, Waters Corporation, MA, USA). The mobile phase was composed of A (water with 0.1% formic acid) and B (acetonitrile) with a linear gradient elution: 0–9 min, 98–90% A; 9–11 min, 90% A; 11–12 min, 90–86% A; 12–15 min, 86–85% A; 15–17 min, 85% A; 17–18 min, 85–83% A; 18–19 min, 83% A; 19–20 min, 83–98% A. The column temperature was maintained at 35°C and with a flow rate of 0.3 ml/min. The detection wavelengths were set at 270 nm, and the injection volume was set at 10 μl. Peaks were identified through the comparison of the retention time with the reference compound.

In this study, forty Jing Fen laying hens aged 90 days were used. After 2 weeks for acclimatization, the laying hens were randomly divided into five groups (8 animals per treatment), namely, control group (basal diet), model group (HELP diet), ACD low dose (ACD-L, HELP diet with 0.5 g ACD/hen per day), ACD medium dose (ACD-M, HELP diet with 1 g ACD/hen per day), and ACD high dose groups (ACD-H, HELP diet with 2 g ACD/hen per day). The ACD was administrated in drinking water for 48 consecutive days. Laying hens were subjected to 1 h water deprivation before administration, which could guarantee the ACD in water was completely taken by laying hens. The FLHS in laying hens was induced by the HELP diet as reported in previous studies (14, 18), and the detailed composition of the basal diet and HELP diet is shown in Table 1.

At the end of the experiment, the hens were fasted for 12 h and anesthetized with intravenous injection of sodium pentobarbital (50 mg/kg). Blood samples were collected from the brachial vein into vacuum tubes. The serum was obtained by centrifugation at 4°C of 4,000 g for 10 min and then stored at −20°C for further analysis. The liver and cecal contents of laying hens were carefully removed, collected, and snap-frozen in liquid nitrogen, and stored at −80°C for metabolomics and gut microbiota analysis, respectively. Meanwhile, a part of liver tissue was fixed in 4% formalin for the observations of pathological changes.

Blood biochemistry indexes in serum such as triglycerides (TG), total cholesterol (TCH), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), alanine aminotransferase (ALT), and aspartate aminotransferase (AST) were analyzed by the commercial kits provided by Shanghai Jiang Lai bio-technology Co., Ltd (Shanghai, China). Under the standard protocol, the liver tissue was made and stained with HE, and oil red O to observe the pathologic changes. The pathological slides were examined using a 13395H2X microscope (Leica Co., Ltd, Wetzlar, Germany).

Before metabolomics analysis, the liver tissues were thawed at room temperature. Next, 80 mg liver tissue were placed in polyethylene tubes, and 1 ml of cold methanol/acetonitrile/H₂O (2:2:1, v/v/v) was added, and the mixture was adequately vortexed and homogenized for 1 min. The homogenate was ultrasonically extracted (30 min/once, twice, 4°C), incubated (1 h at −20°C), and then centrifuged (14,000 g, 20 min, 4°C). The supernatants were collected and then dried by nitrogen to obtain a lyophilized powder. For instrumental analysis, lyophilized samples were reconstituted by dissolving in 100 μl acetonitrile/water (1:1, v/v), and an aliquot of 2 μl was injected for further analysis. To monitor the stability of the system and evaluate the reliability of experimental data, three quality control (QC) samples were prepared by pooling 10 μl of each liver sample and analyzed with the same procedure.

Metabolomics analysis was performed with an Agilent 1290 UPLC (Agilent Technologies Inc., CA, USA) coupled with Triple TOF 6600 system (AB SCIEX, Foster City, USA). Mass-grade acetonitrile, methanol, and formic acid were purchased from Merck (Darmstadt, Germany). Deionized water (18 MΩ) was prepared with a Direct-Q®3 system (Millipore, USA). Hydrophilic interaction liquid chromatographic column (HILIC, 3.0 × 100 mm, 1.8 μm, Agilent Technologies Inc., CA, USA) was used for chromatographic separation. The column was maintained at 25°C and eluted at a flow rate of 0.5 ml/min. The mobile phase consisted of solvent A (25 mM ammonium acetate and 25 mM ammonium hydroxide in water) and solvent B (acetonitrile). The optimized gradient program was shown in Supplementary Table S1. The autosampler remained at 4°C. For mass data acquisition, the electrospray ionization (ESI) source of Triple TOF 6600 was operated in positive (ESI+) and negative ion (ESI–) modes. The main parameters of ESI source and mass spectrometry were used as follows: source temperature, 600°C; ion source gas1, 60 psi; ion source gas2, 60 psi; curtain gas, 30 psi; ion sapary voltage floating, ± 5,000 V (in both ESI modes), MS scan m/z range, 60–1,000 Da; MS scan accumulation time, 0.2 s per spectra. The MS/MS scan is acquired using information-dependent acquisition (IDA) with high sensitivity mode. The settings were as follows: product ion scan m/z range, 25–1,000 Da; product ion scan accumulation time, 0.05 s per spectra; collision energy, 35 ± 15 eV; declustering potential, ± 60 V (in both ESI modes); exclude isotopes within 4 Da, the maximum number of candidate ions to monitor per cycle, 10.

For metabolomics analysis, the raw mass spectrometry data were converted to common data format (.mzXML) using Proteowizard msConvert. Then, peak extraction and alignment were carried out by the XCMS program. Obtained data sets were normalized and Pareto-scaled, and imported into SIMCA-P V13.0 (Umetrics AB, Ume, Sweden) to perform unsupervised principal component analysis (PCA) and supervised partial least-squares discriminant analysis (PLS-DA) to screen and identify different metabolites. The quality and capability of PCA and PLS-DA models were described by cumulative R2 and Q2 (R2X: the fraction of the variables, R2Y: the goodness of the fit, Q2: prediction ability of the model). A permutation test with 200 permutations was performed to evaluate the robustness of the PLS-DA models. The variable importance in the projection (VIP) value from PLS-DA models was calculated to indicate the contribution of the variable to the classification. Potential metabolites with VIP > 1 were further applied to Student's t-test at the univariate level to measure its significance. Metabolites with VIP > 1 and P < 0.05 were selected and considered as potential biomarkers.

Accuracy m/z value of potential metabolites was searched against metabolite databases such as HMDB, METLIN, and Mass Bank (mass accuracy tolerance: 25 ppm). For the confirmation of the structure identification, the MS/MS spectra of the potential metabolites were compared with an in-house database established by available authentic standards. Metabolic pathway interpretation was conducted by MetaboAnalyst and Kyoto Encyclopedia of Genes and Genomes database.

Total genome DNA from cecal content was extracted by using cetyltrimethylammonium bromide/sodium dodecyl sulfate (CATB/SDS) method, and the quality of DNA was monitored on 1% agarose gels. 16S rRNA genes were amplified using 341F/806R primer with the barcode (341F: 5′-CCTAYGGGRBGCASCAG-3′, 806R: 5′-GGACTACNNGGGTATCTAAT-3′). PCR reactions were performed in 30 μl reactions with 15 μl of Phusion® High-Fidelity PCR Master Mix (NEB, USA). The PCR conditions were 98°C for 1 min, followed by 30 cycles of 98°C for 10 s, 50°C for 30 s, 72°C for 60 s, and then 72°C for 5 min. PCR products were purified with AxyPrep DNA Gel Extraction Kit (Axygen, USA). According to the manufacturer's instruction, sequencing libraries were generated using NEB Next® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA), and the index codes were added. With the application of Qubit@ 2.0 Fluorometer (Thermo Scientific, USA) and Agilent Bioanalyzer 2100 system (Agilent, USA), the quality of sequencing libraries was assessed. Finally, the library was sequenced on an Illumina HiSeq 2500 platform.

By using FLASH (http://ccb.jhu.edu/software/FLASH/), the raw paired-end reads from the original DNA fragments were merged. According to the unique barcodes, paired-end reads were assigned to each sample. The cluster of operational taxonomic units (OTUs) of the valid reads was performed by the UPARSE software package (19). Sequences with ≥97% similarity were assigned to the same OTUs. Based on the abundance of OTUs, the alpha diversity (Chao 1 and Ace) and beta diversity were analyzed with Perl scripts. Cluster analysis of cecal samples was preceded by weighted-Unifrac distance-based principal coordinate analysis (PCoA) using the QIIME software package (http://qiime.org/). Pearson's correlation test was used to assess the corrections between the liver metabolites and bacterial compositions, and the correlation heatmap was generated with ggplots package of R software.

Data are represented as mean ± SD. Comparisons of statistical significance among groups were made by one-way ANOVA with Fisher's least significant difference (LSD) test using the Statistical Package for Social Science program (SPSS 16.0, Chicago, IL, USA). A p-value < 0.05 was considered statistically significant.

The UPLC chromatogram of mixed standard solution and ACD are shown in Figure 1. Under the retention time of 20 min, chromatographic peaks of the active ingredient in ACD were clearly separated. Four phytochemicals, including abrine, hypaphorine, vicenin-2, and schaftoside, were identified, and the content of them in ACD was 3.49%, 0.81%, 6.13%, and 18.57%, respectively.

Pathologic examination results of the liver stained by HE and oil red O are shown in Figure 2. In the control group, clear and normal liver cell architecture was observed. Serious fatty degeneration with lipid droplet accumulation and necrosis were found in laying hens with the HELP diet in the model group. In comparison with the model group, HELP diet-induced lipid accumulation in the liver from the laying hens with ACD treatment was decreased at different degrees, indicating the fatty degeneration was ameliorated by ACD. However, compared with the control group, the fat vacuoles, liver cell swelling, and cytoplasm rarefaction could be observed in ACD treatment groups (pointed by an arrow, Figure 2), suggesting the liver injury was not recovered back to normal state.

As shown in Figure 3, compared with the control, AST, ALT, TG, LDL-C, and TCH increased significantly, while HDL-C decreased in laying hens fed with HELP diet in the model group (P < 0.01). Higher serum HDL-C and lower ALT and AST were detected in the ACD-L treatment group in comparison with the model (P < 0.01). Meanwhile, in ACD-M and ACD-H groups, it was found that the levels of AST, ALT, TG, LDL-C, and TCH were much lower and HDL-C were higher than those in the model (P < 0.01). In comparison to the control group, the levels of TCH, TG, LDL-C, AST, and ALT of laying hens in three ACD treatment groups were significantly higher (P < 0.01). HDL-C was markedly reduced in ACD-L and ACD-M groups than that in the control group (P < 0.01). However, no significant difference was observed for HDL-C between the control and ACD-H groups. The laying hens in ACD-M and ACD-H exhibited lower TG, TCH, ALT, LDL-C, and AST than those in the ACD-L group (P < 0.01). There was no difference in HDL-C between ACD-L and ACD-M groups. When compared with the ACD-M group, TCH, TG, LDL-C, ALT, and AST decreased significantly and HDL-C increased significantly in the ACD-H group, which showed that a high dose of ACD had a better reversing effect on biochemical blood indicators than low and medium dose ACD in laying hens fed with HELP diet. Due to the regulation effects of high dose ACD on biochemical indicators, the laying hens in the ACD-H group were used for further metabolomic analysis.

To explore metabolic pathway changes related to ACD treatment, the metabolomics method was used to examine metabolite alterations in the liver. The datasets of the liver metabolomics study in both ESI+ and ESI- modes were provided in Supplementary File S1. Data quality is crucial for metabolomic study. In this study, the stability of the analytical method and instrument was monitored by the pooled QC samples. Results of PCA analysis on all liver samples are shown in Supplementary Figure S1. In PCA score plots, all of the QC samples were tightly clustered, which indicated the method and analytical system were robust with good stability and repeatability. Meanwhile, the separation among control, model, and ACD-H groups in both ESI+ and ESI- modes were evaluated by the unsupervised PCA. A slight but not significant separation trend among the laying hens in control, model, and ACD-H groups was observed in the PCA score plots (Supplementary Figure S1), suggesting the changes of the metabolomic profile of the liver from laying hens under different treatment.

Supervised PLS-DA was employed to identify the potential biomarker. The PLS-DA score plots are shown in Figure 4. A clear separation between control and model groups were observed in PLS-DA score plots in ESI+ and ESI- modes (Figures 4A,C), which indicated that significant metabolic disturbance occurred in the laying hens with FLHS. Meanwhile, the PLS-DA score plots showed that the liver samples in the ACD-H group were located far from those in the model (Figures 4B,D), revealing the metabolic disorders in the liver were improved by ACD treatment. The permutation test was performed to avoid the overfitting of the PLS-DA models. Validation plots of the PLS-DA permutation test showed that all the regression lines of Q2 (cum) points had a negative intercept (Supplementary Figure S2), which indicated the PLS-DA models were robust without overfitting.

Through the VIP and P-value analysis, 17 liver metabolites were identified as potential biomarkers associated with ACD treatment. Compound name, adduct, formula, fold change, and pathways of the metabolites are shown in Table 2. Compared with the control laying hens, the concentrations of 11 metabolites such as kynurenine, 5-hydroxyindoleacetate, arachidonic acid, histidine, tyrosine, and stearoylcarnitine were remarkably reduced, and 6 metabolites including lactate, LysoPE (16:0), betaine, and O-phosphoethanolamine were increased in the model group. Notably, ACD treatment reversed the abnormal liver metabolite changes in different degrees, such as the reduction of O-phosphoethanolamine and the increase of 1-methylhistamine and tyrosine. Hierarchical cluster analysis was carried out to fully and intuitively display the overview of the liver metabolites. The dendrogram in the vertical axis in Figure 5A showed that the liver samples were gathered into two clusters. Interestingly, six liver samples in the ACD-H group were clustered together with those in control, indicating the potential therapeutic effect of ACD on FLHS. As shown in the horizontal axis in Figure 5A, liver metabolites with similar abundance patterns were clustered together. For example, the metabolites such as LysoPE (16:0), O-phosphoethanolamine, lactate, linoleoyl ethanolamide, and betaine were clustered into one group.

Pathway analysis of the potential liver metabolites was performed to visualize the affected metabolic pathways (Figure 5B). The pathway with an impact value above 0.10 was selected out. The results revealed that arachidonic acid metabolism, histidine metabolism, phenylalanine, tyrosine and tryptophan biosynthesis, ubiquinone and other terpenoid-quinone biosyntheses, tyrosine metabolism, and tryptophan metabolism were associated with ACD treatment.

Gut microbiota dysbiosis is related to FLHS. In order to investigate the effects of ACD on gut microbiota composition, 16S rRNA gene sequencing was performed in this study. The summary of the sequencing data and rarefaction analysis are shown in Supplementary Table S2 and Supplementary Figure S3, respectively. These results indicated that high-quality sequencing data were obtained for further analysis. The relationship of the community structure of gut microbiota was investigated by the method of weighted-Unifrac distance-based PCoA, which revealed distinct clustering of rumen content samples in control, model, and ACD-H groups (Figure 6A). Notably, the samples in the ACD-H group were separated from the model, indicating ACD against the dysbiosis of microbiota community induced by HELP in laying hens. Next, Alpha diversities, including the Ace and Chao 1 index, were used to evaluate the changes in microbial communities (Figures 6B,D). Consistently, ACD treatment significantly increased Ace (P < 0.01) and Chao 1 (P < 0.05), which suggested ACD increased richness of the gut microbiota.

The relative abundance of gut bacterial composition at the phylum level is shown in Figure 6C. Compared with the control, the relative abundance of Bacteroidetes decreased, while Proteobacteria increased in the model group. Of note, the reduction of Bacteroidetes and the increase of Proteobacteria were inversed by ACD treatment at the phylum level (Figure 6C). At the genus level, according to the relative abundance and statistical difference, key microbial genera related to ACD treatment were selected (Table 3). When compared with the control group, abundance elevation of Desulfovibrio, Erysipelatoclostridium, and Faecalibacterium and decline of Phascolarctobacterium were observed in the laying hens fed with HELP in the model group (P < 0.01). Remarkable reduction of Barnesiella, Desulfovibrio, Erysipelatoclostridium, Faecalibacterium, Negativibacillus, and Phascolarctobacterium and increase of Butyricicoccus, Prevotellaceae Ga6A1 group, Ruminococcaceae UCG-005, and Ruminococcaceae UCG-013 were found in the ACD-H group compared with those in the model (P < 0.01 and P < 0.05, Table 3).

In this study, Spearman's correlation coefficients were calculated to investigate the correction between liver metabolites and gut microbiota. In Figure 7, the hot color (e.g., red) indicated the positive correlations between specific liver metabolites and certain bacterial genera, while the cool color (e.g., green) denoted the negative correlations. Interestingly, a clear correlation between liver metabolites and microbes in genus level in cecal content was observed. For example, Kynurenine and 5-Hydroxyindoleacetate showed positive corrections with Barnesiella, Negativibacillus, and Phascolarctobacterium. Relative abundance of genera Faecalibacterium and Phascolarctobacterium was positively correlated with arachidonic acid, threonate, and tauroursodeoxycholic acid. Histidine and tyrosine were positively correlated with Butyricicoccus, Prevotellaceae Ga6A1 group, and Ruminococcaceae UCG-005, but negatively with Desulfovibrio and Faecalibacterium.