Single-nucleus RNA sequencing reveals regulatory mechanisms of bile acids on lipid metabolism in Litopenaeus vannamei

Abstract

During lipid metabolism, bile acids are involved in the emulsification and absorption of lipids, and the liver serves as the primary organ responsible for mediating their regulatory functions. Crustaceans lack de novo bile acid synthesis and may differ substantially from vertebrates in bile acid metabolism. To investigate the role of bile acids in the lipid metabolism of crustaceans, this study divided 480 Litopenaeus vannamei (initial weight: 0.640 ± 0.003 g) into three groups. Each group included four replicate tanks, with 40 shrimp per tank. Fed the following diets for an 8-week feeding trial, respectively: high fishmeal diet (HF; 25% fish meal), low fishmeal diet (LF; 12.5% fish meal), or low fishmeal diet supplemented with 400 mg/kg bile acid (LFB). The final body weight, weight gain rate and average daily gain of the LFB group were significantly higher than those of the LF and HF groups (P < 0.05). Hepatopancreatic histomorphological and biochemical analyses revealed that the bile acid supplementation alleviated the accumulation of collagen fibers in the hepatopancreas and reduced the activity of glutamic-pyruvic transaminase (GPT) (P < 0.001). Single-nucleus RNA sequencing (snRNA-Seq) enabled the construction of the first comprehensive hepatopancreatic cell atlas in invertebrates, with 11 major cell types identified via transcriptomic profiling. Bile acid administration significantly increased F cell abundance while reducing the number of R cell and B cell populations (P < 0.05). Intercellular communication analysis demonstrated that bile acid supplementation decreased signal enrichment in neural cell adhesion molecule (NCAM) and collagen pathways, with differential functional pathways between the LFB group and LF group predominantly enriched in non-alcoholic fatty liver disease (NAFLD) pathways. Notably, Maribacter and Tamlana emerged as significantly differentiated genera in the bile acid treatment group, exhibiting functional associations with the mitigation of NAFLD and the catalysis of short-chain fatty acid degradation. Pseudotime trajectory analysis further uncovered potential hepatopancreatic cell differentiation pathways: E cells, as progenitor cells, differentiate into R cells, which subsequently bifurcate into B cell and F cell lineages. Collectively, bile acids alleviate hepatopancreatic fibrosis and inflammation by inhibiting cell junction-related signaling pathways, while concurrently regulating lipid metabolism through the enhancement of F cell proportions.

1. Introduction

Bile acids are amphipathic steroid compounds derived from cholesterol metabolism. They facilitate lipid digestion and absorption by forming mixed micelles and play a central role in regulating cholesterol homeostasis and metabolic signaling pathways (Chiang, 2013). The mechanism by which vertebrates reabsorb biliary secretions through intestinal pathways and transport them back to the liver is termed enterohepatic circulation (Dowling, 1972). Through this enterohepatic circulation, vertebrates accomplish the synthesis, absorption, and metabolism of bile acids, and the regulatory mechanisms underlying these metabolic processes have been extensively elucidated (Frisch and Alstrup, 2018). In contrast, invertebrates lack complete structures for bile acid secretion and storage (e.g., gallbladder) and are deficient in key enzymes for de novo bile acid synthesis from cholesterol, as well as core regulatory components for bile acid homeostasis, such as the farnesoid X receptor (FXR) (Allen, 1976). Consequently, bile acid metabolism in invertebrates exhibits significant divergence from vertebrates and remains poorly characterized.

The Pacific white shrimp (Litopenaeus vannamei) has become one of the most productive and economically valuable crustacean species in global aquaculture due to its remarkable environmental adaptability and high farming profitability (Liao and Chien, 2011). Studies have demonstrated that fish meal serves as a primary protein source in shrimp feed. The deficiency of fish meal can lead to detrimental effects in shrimp, including lipid accumulation, immunosuppression, and reduced growth performance (Shi et al., 2023). Moreover, the low fishmeal diet would disrupt bile acid homeostasis in L. vannamei (Chen et al., 2024). This phenomenon occurs because L. vannamei, as an invertebrate crustacean, lacks the capacity for de novo bile acid synthesis and must obtain bile acids through dietary intake (Yoon et al., 2024). Animal protein sources, particularly fish meal, are essential for crustaceans to acquire bile acids. However, the persistent high prices of fish meal in recent years have hindered the resolution of fish meal deficiency issues. Several studies demonstrated that bile acids supplementation promoted lipid metabolism and improved hepatopancreatic health in shrimp (Su et al., 2021). Studies also have shown that bile acid supplementation in low fish meal diets can enhance the antioxidant capacity of both juvenile and adult L. vannamei, while improving non-specific immune responses and intestinal health (Li et al., 2023a, 2022a). Nevertheless, the cellular and molecular mechanisms underlying bile acid perception, metabolism, and transport in shrimp remain largely unexplored.

The liver is a critical site for bile acid metabolism. In vertebrate hepatocytes, cholesterol is converted into primary bile acids (e.g., cholic acid and chenodeoxycholic acid) through the classical pathway (mediated cholesterol 7α-hydroxylase [CYP7A1]) and the alternative pathway (mediated by sterol 27-hydroxylase [CYP27A1]) (Phelps et al., 2019). Studies have shown that fibroblast growth factor 4 (FGF4), a direct target gene of FXR, inhibits bile acid synthesis by downregulating key enzymes such as CYP7A1 and sterol 12α-hydroxylase (CYP8B1) through paracrine signaling (Song et al., 2025). The hepatopancreas, a multifunctional organ in L. vannamei responsible for digestion, metabolism, and immunity, plays a central role in bile acid metabolism. However, the mechanisms by which invertebrate hepatopancreas participates in bile acid metabolism remain poorly understood, with only limited studies suggesting the existence of a potential enterohepatic axis in invertebrates (Bao et al., 2024).

Single-cell RNA sequencing (scRNA-Seq) technology have revolutionized the ability to unravel the heterogeneity and complexity of RNA transcripts within individual cells (Jovic et al., 2022). Single-cell transcriptomic technologies have gradually been implemented in aquatic research; however, current investigations predominantly focus on building hemolymph immune cell atlases (Daniels et al., 2023). Cui et al. (2022) constructed the first single-cell atlas of L. vannamei hemolymph, identifying three hemocyte clusters (transglutaminase + cells, C-type Lectin + cells, and crustin-like antimicrobial peptide + cells) and elucidating their functional properties, potential differentiation trajectories, and correspondence with morphological subtypes (Cui et al., 2022). Yang et al. (2022) further identified a novel macrophage subpopulation (monocyte-like hemocyte 2) and revealed that this lineage may represent an invertebrate evolutionary homolog of vertebrate monocytes . These studies provide valuable insights into crustacean innate immunity and highlight the utility of single-cell transcriptomics in advancing non-model organism research. However, research on hepatopancreatic cell typing in crustaceans remains scarce, and the evolutionary trajectories of its subpopulations are still debated. The single-cell atlas of zebrafish (Danio rerio) hepatocytes offers a paradigm for such studies. Huang et al. (2024) identified hepatocyte subtypes as potential targets for NAFLD by examining hepatocyte heterogeneity and marker gene expression . However, crustaceans lack homologous cell marker genes, necessitating the establishment of an independent cell classification and functional annotation system. And the specific roles of hepatopancreatic subpopulations in bile acid metabolism and transport, as well as their intercellular interaction mechanisms, remain poorly understood.

To address these gaps, this study used L. vannamei as a single-cell transcriptomic model to identify bile acid metabolism-related cell subpopulations and elucidate the reprogramming mechanisms of key metabolic pathways in response to fishmeal replacement and bile acid supplementation. These findings will establish the cellular basis of bile acid metabolism in invertebrates and provide critical insights for developing sustainable low fishmeal aquafeeds.

2. Materials and methods

2.1. Animal ethics statement

The acquisition of experimental shrimp was authorized by Guangdong Ocean University's Animal Ethics Review Committee (permission number: GOU-IACUC-20240-A0412; approval date: May 30, 2024), and all studies were carried out in compliance with its regulations.

2.2. Diet preparation

The positive control diet (high fishmeal diet, HF) contained 25% fish meal in total, while the negative control diet (low fishmeal diet, LF) was formulated by replacing 12.5% of fish meal with an equal amount of de-phenolized cottonseed protein. A dosage of 400 mg/kg bile acids was supplemented to the negative control diet to produce the third diet (LFB), resulting in three experimental diets prepared for subsequent aquaculture trials. Ingredients and nutrient levels of experimental diets appear in Table 1. Amino acid crystals were pre-coated with carboxymethyl cellulose to minimize leaching losses. Feed processing was performed via the following sequential steps: ingredients were first ground to pass through an 80-μm mesh sieve, then precisely weighed and homogenized in an industrial mixer (M-256, South China University of Technology, Guangzhou, Guangdong, China). Subsequently, feed pellets with a diameter of 1.0 to 1.5 mm were produced using a specialized extruder (School of Chemical Engineering, South China University of Technology, Guangzhou, Guangdong, China). Post-extrusion, the pellets were subjected to heat treatment at 90 °C for 60 min, followed by ambient air-drying for 48 h, and finally stored at −20 °C until subsequent use.

Table 1.

Ingredients and nutrient levels of experimental diets (%, dry matter basis).¹

Items	HF	LF	LFB
Ingredients
Fermented soybean meal	4.69	4.69	4.69
Yeast	3.13	3.13	3.13
Monocalcium phosphate	1.75	1.75	1.75
Fish soluble	3.50	0.00	0.00
Shrimp meal	6.25	6.25	6.25
Squid paste	0.87	0.87	0.87
Chilean fishmeal	12.50	15.00	15.00
Gambian fishmeal	12.50	0.00	0.00
Chicken meal	12.50	12.50	12.50
Dephenol cottonseed protein	0.00	13.50	13.50
Soybean meal	18.28	18.28	18.28
Wheat flour	17.50	17.50	17.50
Soybean phospholipid	1.75	1.75	1.75
Vitamin and mineralpremix2	4.78	4.78	4.78
Total	100.00	100.00	100.00
Nutrient levels3
Crude protein	43.25	41.08	42.70
Crude lipid	7.26	7.21	7.22
Dry matter	93.22	93.26	93.52
GE, kcal/kg	4338.00	4305.00	4311.00
OM	84.10	84.70	84.70

GE = gross energy; OM = organic matter.

¹HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

²The premix was formulated to contain the following (per kg of premix): vitamin A 0.046 KIU, vitamin D₃ 31,202.04 KIU, vitamin E 0.182 KIU, arsenic 0.022 mg, cadmium 1.302 mg, lead 0.029 mg, mercury 0.001 mg, chromium 0.065 mg, fluorine 0.183 mg, iron 0.260 g, manganese 0.042 g, iodine 0.003 g, cobalt 0.001 g, selenium 0.001 g, copper 0.036 g, zinc 0.111 g, vitamin K 0.028 g, thiamin-B₁, 0.044 g, riboflavin-B₂, 0.041 g, pyridoxine-B₆, 0.052 g, folic acid-B₉, 0.026 g, vitamin B₁₂ 0.001 g, vitamin C 0.528 g, calcium 0.005 g, phosphorus 0.003 g, calcium/phosphorus index 1.571.

³The nutrient levels were determined values except for OM.

2.3. Animal and tissue preparation

Juvenile L. vannamei (initial weight 0.640 ± 0.003 g) were obtained from Yuehai Co., Ltd. (Zhanjiang, Guangdong, China). Following acclimation, 480 shrimp were randomly distributed into 12 tanks (0.3 m³ capacity). Each group was assigned four replicate tanks, with 40 shrimp reared in each tank. Three experimental diets were administered for 8 weeks through four daily feedings (08:00, 11:00, 17:00, and 21:00) to apparent satiation. Residual feed was quantitatively retrieved via siphonage 60 min post-feeding for consumption calculations. The recirculating aquaculture system maintained critical water parameters within optimal ranges: temperature 28 to 30 °C, dissolved oxygen > 6.8 mg/L, pH 7.8 to 8.2, and total ammonia nitrogen < 0.03 mg/L. Continuous aeration and biological filtration ensured water quality stability throughout the 8-week trial.

After an 8-week feeding trial, before bulk weighing, shrimp underwent a 12-h fasting period. From each tank, 28 shrimp were allocated for analyses: 5 for proximate composition, 7 for enzyme assays, 4 for hepatopancreatic histology, 4 for gut microbiota, and 8 for single-cell transcriptome. The sample collection methods described above are as follows: all sampling procedures were performed on ice and with sterile equipment; samples for proximate composition analysis, enzyme activity assays, and single-cell transcriptome sequencing were immediately snap-frozen in liquid nitrogen, while during dissection of hepatopancreas sections, care was taken to preserve tissue integrity and avoid stretching or distortion. Two hours after feeding, intestinal content samples were aseptically collected, snap-frozen in liquid nitrogen, and stored at −80 °C for subsequent genomic DNA extraction and microbial community analysis. Hemolymph was collected via aspiration from the fourth abdominal segment using a sterile 1 mL syringe and immediately centrifuged (3000 × g, 4 °C, 5 min) to isolate hemocytes, with the subsequent supernatant stored at −80 °C. Proximate analyses of diets and whole-body samples followed Horwitz and Latimer (2005) protocols. Moisture content was determined gravimetrically through drying at 105 °C to constant mass (method 934.01). Crude protein quantification employed a Primacs100 analyzer (Skalar Analytical B.V., Breda, North Brabant, the Netherlands) with 6.25 as the nitrogen-protein conversion factor (method 992.15). Crude lipid extraction utilized petroleum ether in an XT15 automated system (Ankom Technology, Macedon, NY, USA) (method 920.39). The gross energy of diets was determined using an oxygen bomb calorimeter (C2000, IKA Works GmbH, Staufen, Baden-Württemberg, Germany) (method 984.10). The organic matter content was calculated from the dry matter (method 934.01) and ash (method 942.05) content, representing all organic constituents after removal of moisture and inorganic mineral matter, with triplicate measurements per sample.

2.4. Growth performance

The initial body weight and final body weight of L. vannamei were measured before the start of the feeding trial and after the completion of the eight-week feeding experiment, respectively. The growth performance was calculated as follows:

Weight gain rate (WGR, %) = 100 × (Final body weight - Initial body weight)/Initial body weight;

Survival rate (SR, %) = 100 × Final number of shrimp/Initial number of shrimp;

Feed conversion ratio (FCR) = Feed consumed/(Final body weight - Initial body weight);

Average daily gain (ADG, g/d) = (Final body weight - Initial body weight)/Experimental days;

Average daily feed intake (ADFI, g/d) = Feed consumed/(Experimental days × Number of animals).

2.5. Biochemical indices of hepatopancreas

Hepatopancreas samples (0.2 g) were homogenized with nine volumes of ice-cold normal saline (0.9% NaCl). The homogenate was subjected to centrifugation at 3500 × g (10 min, 4 °C), with collected supernatants used for enzymatic profiling: triglyceride (TG), total cholesterol (T-CHO), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), glutamic oxaloacetic transaminase (GOT), glutamic pyruvic transaminase (GPT), and glucose (GLU). All biochemical determinations were performed using standardized commercial assay kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, Jiangsu, China) following manufacturer protocols.

2.6. Hepatic histological analysis

Three hepatopancreatic tissues per tank were fixed in Bouin's solution for 24 h and subsequently stored in 70% ethanol. Tissue processing included sequential dehydration through a graded ethanol series, followed by paraffin embedding. Serial sections (5 μm thick) were subjected to four histological staining protocols: hematoxylin-eosin (H&E) staining for evaluating cellular morphology, periodic acid-Schiff (PAS) staining for polysaccharide detection, Masson's trichrome staining for collagen visualization, and Sirius red staining for fibrosis assessment. All stained sections were observed under a research-grade microscope (ECLIPSE 90i, Nikon Corporation, Tokyo, Japan) equipped with digital imaging functionality, and images were captured for subsequent analysis.

2.7. Gut microbiota sequencing and analysis

Intestinal microbial profiling was performed by Gene Denovo Biotechnology (Guangzhou, Guangdong, China) using Illumina-based sequencing. The V3–V4 hypervariable regions of bacterial 16S rRNA genes were amplified in triplicate with primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 806R (5′-GGACTACHVGGGTATCTAAT-3′). PCR products were purified via 2% (w/v) agarose gel electrophoresis, extracted using an AxyPrep DNA Gel Extraction Kit (DNA Gel Extraction Kit, AxyPrep DNA Gel Extraction Kit, Axygen Biosciences, Union City, CA, USA), and quantified on an ABI StepOnePlus Real-Time PCR System (Real-time PCR System, StepOnePlus, Thermo Fisher Scientific Inc., Waltham, MA, USA).

Raw sequencing reads were processed through a bioinformatic pipeline: FLASH v1.2.11 merged paired-end reads (Magoc and Salzberg, 2011), followed by demultiplexing based on unique barcodes. Operational taxonomic units (OTUs) were grouped at a 97% sequence similarity threshold via the UPARSE pipeline, with representative sequences taxonomically classified via the RDP classifier (v2.13) against the SILVA 138 database. Microbial α-diversity was evaluated using QIIME2 to calculate Shannon diversity and Chao1 richness indices, with rarefaction curves generated based on observed OTUs. Biomarker identification used linear discriminant analysis effect size (LEfSe) analysis (linear discriminant analysis [LDA] score > 2.0; P < 0.05). Beta diversity was assessed through MUSCLE-aligned sequences and UniFrac distance matrices (weighted/unweighted) with weighted/unweighted UniFrac distance matrix computation (GUniFrac R package v1.7). Community structure patterns were visualized by Bray-Curtis-based principal coordinates analysis (PCoA) based on Bray–Curtis dissimilarity (ggplot2 v3.4.0).

2.8. Hepatopancreatic snRNA-Seq preparation

The collection process of hepatopancreas samples for single nuclei RNA sequencing analysis referred to the method described by (Xie et al., 2024). Frozen hepatopancreas tissues (200 mg pooled from 4 specimens) were homogenized in 0.5 mL ice-cold lysis buffer (250 mmol/L sucrose, 5 mmol/L CaCl₂, 3 mmol/L Mg(OAc)₂, 10 mmol/L Tris–HCl [pH 7.4], 1 mmol/L dithiothreitol [DTT], 0.1 mmol/L ethylenediaminetetraacetic acid [EDTA], 1 × protease inhibitor, 1 U/μL RiboLock RNase inhibitor) using a 2-mL Dounce homogenizer (Wheaton Industries, Inc., Millville, NJ, USA). Sequential mechanical dissociation was performed under cryogenic conditions: 15 gentle strokes with a loose-clearance pestle (Type A), followed by 10 controlled strokes with a tight-clearance pestle (Type B). The resulting homogenate was diluted with 700 μL chilled nuclei isolation buffer (0.04% w/v bovine serum albumin [BSA], 0.2 U/μL RiboLock RNase inhibitor, 500 mmol/L mannitol, 0.1 mmol/L phenylmethanesulfonyl fluoride [PMSF]) and filtered through a 70-μm nylon cell strainer. Nuclei purification was accomplished via discontinuous iodixanol density gradient centrifugation at 800 × g for 20 min at 4 °C. The pelleted nuclei were re-filtered through a 40 μm cell strainer and resuspended in 100 μL nuclei isolation buffer. Nuclear integrity and concentration were assessed using the 0.4% (w/v) Trypan blue exclusion assay, with counting performed on a Neubauer hemocytometer under a phase-contrast microscope (Nikon Eclipse Ti2, 40 × , Nikon Corporation, Tokyo, Japan).

2.9. Single-nuclei library generation and sequencing

Nuclear suspensions were processed using the 10 × Genomics Chromium Controller (v3.1 chemistry) to generate gel bead-in-emulsions (GEMs) following manufacturer specifications. Single-cell libraries were prepared with the Chromium Next GEM Single Cell 3′ Reagent Kit v3.1 (10 × Genomics, Pleasanton, CA, USA), which incorporates the following sequence elements: (1) Illumina R1 (read 1 primer), (2) 16 nt 10 × barcode, (3) 10 nt unique molecular identifier (UMI), and (4) poly-dT primer for mRNA capture. Post-GEM dissolution, reverse transcription was performed to generate barcoded cDNA from polyadenylated transcripts, followed by PCR amplification to achieve sufficient material for library construction. The library preparation workflow included sequential addition of Illumina adapter sequences: R1 incorporation during GEM incubation, and P5/P7 adapters with sample indices and R2 sequences during post-cDNA processing. Final library construction involved end repair, A-tailing, adapter ligation, and amplification steps. Sequencing was conducted using standard Illumina paired-end chemistry with read 1 and read 2 primer sites.

2.10. Single-nuclei RNA-Seq bioinformatics pipeline

Raw sequencing data were processed using Cell Ranger (v6.0.0, 10 × Genomics) with alignment to the L. vannamei reference genome (NCBI Taxonomy ID: 27706). Read quantification required ≥ 50% exon overlap for UMI counting. Cell barcode validation was performed using the EmptyDrops algorithm (Lun et al., 2019). Post-quality control, nuclei-gene matrices for HF, LF, and LFB groups were analyzed in Seurat (v3.1.1) with the following filtering criteria (Butler et al., 2018). Exclusion of nuclei with UMI counts ≥ 18,000, mitochondrial gene content > 10%, or gene detection outside the 460 to 3800 range. Data normalization employed the LogNormalize method, followed by batch effect correction through canonical correlation analysis (CCA) (González et al., 2008).

Dimensionality reduction was performed using principal component analysis (PCA) on a 1% subsampled dataset to establish a null distribution of gene expression patterns. Cell clustering utilized a graph-based approach with PCA-derived distances, implemented at a resolution parameter of 0.2, yielding 13 distinct clusters. Clusters were visualized in two dimensions with t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP).

Differential gene expressions (DGEs) analysis employed the Wilcoxon rank-sum test with stringent criteria: (1) adjusted P-value <0.01 for intra-cluster comparisons or < 0.05 for inter-group (HF, LF, and LFB) comparisons; (2) minimum 1.28-fold change in expression; and (3) detection in > 25% of cells within the target cluster or group. Functional annotation of DGEs was performed using ClusterProfiler (v3.18.1) for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. Statistical significance was determined using cumulative hypergeometric distribution tests with a false discovery rate (FDR) threshold of ≤ 0.05.

2.11. Single-nuclei trajectory inference analysis

Pseudotemporal ordering of hepatocytes was performed using Monocle v2.10.1. The analysis involved dimensionality reduction to a single latent space, followed by cell ordering and trajectory reconstruction in reduced dimensional space. Differential gene expression analysis along the pseudotime trajectory identified key regulatory genes with a stringent FDR threshold of < 1 × 10⁻⁵. Gene clustering was performed based on expression patterns, functional annotations, and shared regulatory networks.

Branch-dependent gene expression analysis employed Monocle's Barcode Enabled Antigen Mapping (BEAM) algorithm, which implements a statistical framework comparing two negative binomial generalized linear models (GLMs) to identify branch point-specific gene expression signatures. The BEAM analysis enabled detection of genes exhibiting significant expression changes at trajectory bifurcation points, providing insights into lineage commitment decisions.

2.12. Intercellular communication network analysis

Cell–cell communication patterns were inferred using CellChat (v1.6.1), which employs a probabilistic framework to reconstruct ligand-receptor interaction networks across distinct cell populations. The algorithm calculates interaction probabilities based on gene expression profiles of ligand-receptor pairs, with statistical significance determined at P < 0.05. Network properties, including interaction diversity and communication strength, were quantified and visualized through chord diagrams and circos plots, respectively. These analyses enabled systematic characterization of signaling pathway activity and cell-type specific communication patterns within the tissue microenvironment.

2.13. Immunohistochemistry

Paraffin-embedded L. vannamei hepatopancreas sections from three groups underwent dewaxing in eco-friendly solutions I-III (10 min each), dehydration via anhydrous ethanol gradients (I-III, 5 min each), and hydration with distilled water. Heat-induced epitope retrieval (HIER) was performed processed in EDTA (pH 8.0) with 30-min 95 °C exposure (buffer evaporation controlled to prevent drying), followed by cooling and three phosphate buffered saline (PBS) (pH 7.4) washes (5 min each). Endogenous peroxidase activity was blocked with 3% H₂O₂ (25 min, room temperature [RT], dark), then rinsed thrice with PBS. Sections were blocked with 3% serum (goat serum for goat-derived primary antibodies; BSA for others) for 30 min (RT), incubated overnight at 4 °C with PBS-diluted primary antibodies, washed, and treated with HRP-conjugated secondary antibodies (50 min, RT). 3,3′-Diaminobenzidine (DAB) substrate was applied for color development (monitored microscopically), halted by water rinsing. Nuclear counterstaining used hematoxylin (3 min), differentiation, and bluing. Sections were dehydrated (75%–85%-anhydrous ethanol I-II), cleared in xylene via tert-butanol, and mounted with neutral resin. Morphology and immunoreactivity were analyzed using a Nikon Ni–U microscope (Nikon Corporation, Tokyo, Japan). Primary antibody details are in Table S1.

2.14. Statistical analysis

Statistical analysis was performed with SPSS v.21 (IBM Corporation, Chicago, IL, USA). Data are expressed as mean and standard error of the mean (SEM), after verifying normality and homogeneity, using one-way ANOVA for comparisons. Duncan's multiple comparisons were performed between the different groups when differences were significant (P < 0.05).

Data were analyzed using a one-way analysis of variance, equivalently formulated as the following general linear model (regression):

$$Y_j={\beta }_0+{\beta }_2{D}_2ⱼ+{\beta }_3{D}_3ⱼ+\dots +{\beta }_k{D}_{kⱼ}+{\epsilon }_j\text{,}$$

Where Y_j denotes the j-th observation; β₀ represents the mean of the reference group (e.g., Group 1); D_iⱼ (i = 2, 3, …, k) are dummy variables (indicator variables) encoding group membership (D_iⱼ = 1 if the observation belongs to the i-th group, 0 otherwise); β_i (i = 2, 3, …, k) represents the difference between the mean of the i-th group and the mean of the reference group; ε_j represents the independently and identically distributed random error term, following ε_j ∼ N (0, σ²). The global F-test of this model was used to test whether all group means are equal (H₀: β₂ = β₃ = … = β_k = 0).

The one-way ANOVA mathematical model is expressed as:

$$Y_{ij}=\mu +{\alpha }_i+{\epsilon }_{ij}\text{,}$$

Where Y_ij denotes the j-th observation in the i-th group (i = 1, 2, …, k); μ represents the grand mean (overall population mean); α_i denotes the treatment effect of the i-th group (α_i = μ_i - μ), subject to the constraint Σ (α_i · n_i) = 0; ε_ij represents the independently and identically distributed random error term, following ε_ij ∼ N (0, σ²). This model was used to test the null hypothesis H₀: μ₁ = μ₂ = … = μ_k.

3. Results

3.1. Growth performance and body composition of shrimp fed different diets

The effects of dietary bile acid supplementation in low fishmeal diets on growth performance, feed intake and body composition of L. vannamei are presented in Table 2 and Table 3. There were no differences in the SR, IBW and FCR among groups (P > 0.05). However, FBW and WGR in the LFB group were significantly higher than those in both HF and LF groups (P < 0.05).

Table 2.

Effects of supplementing bile acids in low fishmeal diet on the growth and feed intake data of L. vannamei.

Items1	IBW, g	FBW, g	FCR	WGR, %	ADG, g/d	ADFI, g/d	SR, %
HF	0.65	7.74b	1.50	1104.06b	0.13b	0.19	96.25
LF	0.65	7.96b	1.40	1157.90b	0.13b	0.19	93.12
LFB	0.65	8.97a	1.32	1287.60a	0.15a	0.20	95.63
SEM	0.001	0.185	0.029	28.031	0.003	0.003	0.014
P-value	0.422	0.011	0.072	0.024	0.003	0.745	0.734

IBW = initial body weight; FBW = final body weight; FCR = feed conversion ratio; WGR = weight gain rate; ADG = average daily gain; ADFI = average daily feed intake; SR = survival rate.

In the same column, values with different superscript letters indicate significant differences (P < 0.05). Data are presented as the mean and standard error of the mean (SEM) (n = 4).

¹HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

Table 3.

Effects of supplementing bile acids in low fishmeal diet on the body composition (%) of L. vannamei.

Items1	Moisture	Crude protein	Crude lipid
HF	75.25	14.53	2.65b
LF	75.50	14.64	3.20ab
LFB	74.25	14.72	3.69a
SEM	12.913	0.733	1.380
P-value	0.794	0.637	0.003

In the same column, values with different superscript letters indicate significant differences (P < 0.05). Data are presented as the mean and standard error of the mean (SEM) (n = 4).

¹HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

The whole-body moisture content and crude protein levels showed no significant variations across the three experimental groups (P > 0.05). Notably, the LFB group exhibited significantly higher whole-body crude lipid content than the HF group (P = 0.003), while showed no difference between LFB and LF groups (P > 0.05).

3.2. Hepatopancreatic biochemical parameters and liver function in shrimp

No significant differences were observed in LDL-C content and GOT activity among the HF, LF, and LFB groups (P > 0.05) (Table 4). The TG content in the LFB group was significantly lower than that in the HF and LF groups (P < 0.001). The T-CHO content in the LFB group was significantly lower than that in the HF group (P = 0.032), but the difference compared to the LF group was not statistically significant (P > 0.05). Both the LFB and LF groups exhibited significantly higher HDL-C levels compared to the HF group (P = 0.018). Glutamic pyruvic transaminase activity in the LFB group was significantly lower than that in the HF and LF groups (P = 0.001). Glucose content in the LFB group was significantly higher than that in the LF group but significantly lower than that in the HF group (P < 0.001).

Table 4.

Effects of supplementing bile acids in low fishmeal diet on the biochemical parameters of hepatopancreas in L. vannamei.

Items1	TG, mmol/g prot	T-CHO, mmol/g prot	LDL-C, mmol/g prot	HDL-C, mmol/g prot	GPT, U/g prot	GOT, U/g prot	GLU, mmol/g prot
HF	0.07a	1.02a	0.37	1.49b	6.76a	4.47	0.31a
LF	0.08a	0.95ab	0.32	2.10a	9.38a	4.32	0.19c
LFB	0.03b	0.75b	0.37	2.09a	3.38b	3.47	0.25b
SEM	0.008	0.048	0.019	0.118	0.920	0.337	0.017
P-value	<0.001	0.032	0.478	0.018	0.001	0.478	<0.001

TG = triglyceride; T-CHO = total cholesterol; HDL-C = high density lipoprotein cholesterol; LDL-C = low density lipoprotein cholesterol; GPT = glutamic pyruvic transaminase; GOT = glutamic oxaloacetic transaminase; GLU = glucose.

In the same column, values with different superscript letters indicate significant differences (P < 0.05). Data are presented as the mean and standard error of the mean (SEM) (n = 4).

¹HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

3.3. Hepatopancreatic histology of shrimp

A low fishmeal diet induces pathological alterations in the hepatopancreas of L. vannamei, while bile acid supplementation restored the normal hepatic structure. The H&E staining revealed severe deformation of hepatocyte structure and dilation of hepatic tubules in the LF group compared to the HF group (Fig. 1C). After bile acid supplementation, the hepatocyte structure was restored to normal, with a significant increase in the number of F cells (P < 0.001) and a significant decrease in the number of B cells (P = 0.003) compared to the LF group (Fig. 1F). Periodic acid-Schiff staining reflects the extent of glycogen accumulation in tissues. The results showed that compared with the HF group, glycogen accumulation was observed in the LF group. Bile acid supplementation significantly reversed this trend (P = 0.007) (Fig. 1B, E). Sirius red staining indicates the degree of collagen fiber deposition in tissues. The results demonstrated that compared with the HF group, the LF group exhibited hepatocellular fibrosis, while bile acid administration significantly reduced collagen fiber accumulation in the hepatopancreas (P = 0.018) (Fig. 1A, D).

Fig. 1.

Effects of supplementing bile acids in low fishmeal diet on the health of hepatopancreas of L. vannamei. (A) Sirius red staining of the hepatopancreas. (B) Periodic acid-schiff (PAS) staining of the hepatopancreas. (C) Hematoxylin and eosin (H&E) staining of the hepatopancreas. (D) The positive ratio of Sirius red staining. (E) The positive ratio of PAS staining. (F) The number of F/E cells. HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid. Data column with different letters indicate significant differences (P < 0.05). Data are presented as the mean and standard error of the mean (SEM) (n = 4).

3.4. Intestinal microbiota analysis

Analysis of 16S rRNA sequences produced 2,746,235 raw tags across shrimp intestinal samples (avg. 114,426/sample). Post-processing generated 5,633,076 effective Tags, averaging 100,265 per sample (Table S2). More than 87.62% of the high-quality effective Tags were used in the OTUs clustering process (Table S3). No significant differences were observed in α-diversity indices (Shannon, Simpson, ACE, and Chao1 indexes) between groups (P > 0.05), but bile acid supplementation significantly increased species richness, as evidenced by higher Sobs index in the LFB group compared to HF and LF groups (P = 0.024). (Table 5). Specifically, the shrimp-fed LF group displayed the highest number of core OTUs, while the shrimp-fed HF group had 18 core OTUs and LFB had 11 (Fig. 2A). Principal component analysis demonstrated clear intergroup separation between the three groups, where 62.06% of the variance was explained by the two primary coordinates. (Fig. 2B).

Table 5.

Intestinal bacterial diversity and community richness of L. vannamei fed different diets.

Items1	Sobs index	Shannon index	Simpson index	Chao1 index	ACE index
HF	278.75b	3.08	0.77	362.33	372.23
LF	262.75b	3.01	0.75	325.06	338.37
LFB	281.25a	3.05	0.72	350.92	356.47
SEM	5.571	0.098	0.019	9.723	10.516
P-value	0.024	0.970	0.592	0.302	0.463

In the same column, values with different superscript letters indicate significant differences (P < 0.05). Data are presented as the mean and standard error of the mean (SEM) (n = 4).

¹HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

Fig. 2.

Effects of supplementing bile acids in low fishmeal diet on microbiota species composition of L. vannamei. (A) Venn diagram. (B) Principal coordinates analysis. (C) Intestinal microbiota composition at the phylum level. (D) Intestinal microbiome composition at the genus level. (E and F) Intestinal microbiome composition at the phylum level (top 5) and genus level (top 10). HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid. Data column with different letters indicate significant differences (P < 0.05). Data are presented as the mean and standard error of the mean (SEM) (n = 4). PCo = principal Co-ordinate.

Illustrated in Fig. 2C, in the phylum, the intestinal microbiota of each group consisted mainly of the following bacteria (top5): Proteobacteria (HF: 68.32%; LF: 88.65%; LFB: 78.64%), Bacteroidota (HF: 26.78%; LF: 6.63%; LFB: 16.05%), Firmicutes (HF: 0.75%; LF: 3.23%; LFB: 2.38%), Verrucomicrobiota (HF: 2.60%; LF: 1.19%; LFB: 2.56%), Actinobacteriota (HF: 0.75%; LF: 0.12%; LFB: 0.11%). The relative abundance of Proteobacteria in the LF group was significantly higher than that in both the HF group and LFB group (P = 0.028), while there was no significant difference in the abundance of Proteobacteria between the HF group and the LFB group; the relative abundance of Bacteroidota in the LF was significantly lower than that in the HF group (P = 0.050), while there was no significant difference in the abundance of Bacteroidota between the HF group and the LFB group (Fig. 2E). As for the genus level, the intestinal microbiota consisted mainly of the following bacteria (top5): Vibrio (HF: 59.06%; LF: 46.10%; LFB: 81.00%), Tenacibaculum (HF: 19.15%; LF: 0.43%; LFB: 1.60%), Photobacterium (HF: 0.63%; LF: 15.45%; LFB: 3.62%), Shewanella (HF: 0.95%; LF: 12.05%; LFB: 1.77%), and Acinetobacter (HF: 0.77%; LF: 11.30%; LFB: 2.43%) (Fig. 2D); at the genus level, the relative abundance of Tenacibaculum was significantly lower in the LF and LFB groups than in the HF group (P = 0.002); and the relative abundance of Shewanella was significantly higher in the LF group than in the HF and LFB groups (P = 0.030) (Fig. 2F).

The diversity of microbial taxa among the three groups was evaluated using the LDA with a score greater than 2 and the LEfSe package (Fig. 3). The analyses showed significant differences in Bacteroidota in the HF group, and Proteobacteria in the LF group at the phylum level. At the genus level, there are differences between d_Flavobacterium, g_Tenacibaculum, y_Fusibacter, a8_Motilimonas in the HF group; a_Porphyromonas, i_Empedobacter, b0_Shewanella, a3_Acetobacterales, u_Lachnospirales, p_Exiguobacterium in the LF group; e_Maribacter, f_Tamlana, s_Streptococcus, b4_Alkalimarinus in the LFB group.

Fig. 3.

Linear discriminant analysis effect size (LEfSe) analysis of the intestinal microbiota. (A) LEfSe analysis bar plot. (B) LEfSe analysis cladogram. HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid. P-value <0.05, linear discriminant analysis (LDA) score ≥2.

3.5. Identification of cell types of the hepatopancreas in L. vannamei by using snRNA-Seq

The overall experimental procedure is shown in Fig. 4A. Hepatopancreas cells collected from three groups were performed snRNA-Seq analysis to investigate their response to dietary fishmeal levels and dietary supplementation of bile acids. A total of 11,231, 11,621 and 7312 cells from the liver derived from HF, LF, and LFB groups were isolated and sequenced. These cells exhibited a median of 6602, 7302, and 6125 UMI counts, and the median number of the genes per cell was 1150, 1360, and 1054, respectively.

Fig. 4.

Single-nucleus RNA sequencing (snRNA-Seq) analysis overview for L. vannamei. (A) Schematic depicting the snRNA-Seq workflow applied to L. vannamei cells. (B) t-distributed stochastic neighbor embedding (t-SNE) visualization delineating the 12 cellular clusters identified in the snRNA-Seq dataset. (C, D, and E) t-SNE visualizations of the 12 major cell types specific to the (C) HF, (D) LF, and (E) LFB groups, respectively. HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

After quality control, the cells underwent t-distributed stochastic neighbor embedding (t-SNE) dimensionality reduction and unsupervised clustering analysis (resolution = 0.15), resulting in the identification of 12 distinct cell populations (clusters 0-11). The distribution of cells across all groups is illustrated in Fig. 4B, while the respective cell distributions for the HF, LF, and LFB groups are shown in Fig. 4C, D, and E, respectively. The 12 major liver cell types were annotated according to their potential functions implied by the marker genes (Fig. 5), which including F cell (Amy2, Apod, CTSC, and LCP2), R cell (Slc37a2, ACAC, and SCD5), B cell (Gs2, vha-10, and LOC113800236), endothelial cell (lgfbp7 and SPON1), E cell (Dsp1 and elF-5A), granulocytes (PEN-3 and LOC113801825), M cell (ChAT and ADCY5), phagocytes (PRDX5, Nlrp3, and ALF), immature R cell (NAAT1 and Picot), keratinocytes (Clca4a and Cd63), myoepithelial cell (Plg and LanB2) and peritrophic matrix (MCFD2 and FSTL4). The full names of the abbreviations of all genes can be found in Supplementary file Abbreviations and Full Names of Genes. Table S4 F cell, R cell, B cell, E cell, M cell, and immature R cell are belonging to the hepatocytes.

Fig. 5.

Heatmap representing normalized expression levels of marker genes across the 12 cell types. Expression magnitude is encoded by color intensity, while the proportion of expressing cells is denoted by dot size.

The cell counts and proportions of the HF, LF, and LFB groups were presented in Fig. 6A and B, with the proportions detailed in panel B: F cell (39.09%, 38.41%, and 64.82%), R cell (22.73%, 29.02%, and 13.43%), B cell (9.59%, 12.74%, and 7.37%), endothelial cell (8.87%, 6.96%, and 4.54%), E cell (4.88%, 4.01%, and 2.61%), granulocytes (4.08%, 2.25%, and 2.38%), M cell (3.45%, 2.28%, and 2.08%), phagocytes (2.55%, 2.68%, and 1.31%), immature R cell (1.79%, 0.69%, and 0.22%), keratinocytes (1.55%, 0.35%, and 0.81%), myoepithelial cell (0.96%, 0.51%, and 0.31%) and peritrophic matrix (0.46%, 0.09%, and 0.11%), respectively. As shown in Fig. 6C, statistical analysis revealed that compared to the HF and LF groups, the LFB group exhibited a significant upregulation in F cell abundance, alongside significant downregulations in the proportions of R cells, B cells, and E cells (P < 0.001).

Fig. 6.

Characterization of hepatocyte subtypes. (A) Absolute counts of each hepatocyte subtype across the three experimental groups. (B) Proportional representation of each subtype within the groups. (C) Heat map of the number of hepatopancreas cells in three groups. (D-F) Gene Ontology (GO) enrichment analysis of the upregulated genes in F, R, and B cells. (G-I) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of the upregulated genes in F, R, and B cells. HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid; ns: P > 0.05; ∗∗∗: P < 0.001.

The up-regulated genes could reveal the different functions of main hepatocyte subtypes. As shown in Fig. 6D–I, the KEGG analysis of the up-regulated gene in F cell mainly enriched in the nutritional digestion and absorption related pathways (pancreatic secretion, protein digestion and absorption, and lysosome etc.), in R cell and B cell are mainly enriched in the oxidative phosphorlylation. The GO enrichment analysis identified extracellular region-associated functions as the primary enrichment targets for upregulated genes in F cells, hydrolase activity, and peptidase activity, acting on L-amino acid peptidase. In R cells, the upregulated genes were mainly enriched in oxidation-reduction processes, including the oxidation-reduction of small molecules and oxidoreductase activity. In contrast, the upregulated genes in B cells were predominantly enriched in functions associated with the cytoplasm and organelle membranes.

3.6. Cellular heterogeneity analysis

This study employed Bulk RNA-Seq and snRNA-Seq to conduct a joint functional analysis of DEGs, aiming to identify key pathways. Differentially expressed genes from the HF vs. LF and LF vs. HF comparisons were analyzed (Fig. 7A–B), and the top 20 enriched pathways were screened. Among these, fatty acid degradation, steroid hormone biosynthesis, oxidative phosphorylation, and NAFLD were the primary pathway of interest in this study. The KEGG functional enrichment analysis of DEGs in major cell types from the low-fishmeal and bile acid-supplemented groups revealed distinct pathway enrichments (Fig. 7C–F): F cells were primarily enriched in “other glycan degradation,” R cells in “sphingolipid metabolism,” B cells in “pancreatic secretion,” and E-cells in “fat digestion and absorption."

Fig. 7.

Comparison of the differences among the subtypes of hepatopancreas cells. (A and B) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of differentially expressed genes in bulk RNA sequencing samples (HF vs. LF/LF vs. LFB). (C–F) KEGG enrichment analysis of the differentially expressed genes in F, R, B and E cells. (G) Quantification of gene expression magnitude and cellular prevalence across four key lipid metabolic pathways (fat digestion and absorption, cholesterol metabolism, bile secretion, and primary bile acid biosynthesis). HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

The expression profiles across different hepatopancreatic cell subtypes revealed that the percentage of genes involved in fat digestion and absorption, cholesterol metabolism, and bile secretion remained unaffected by bile acid supplementation (Fig. 7G). However, in the pathway of primary bile acid biosynthesis, F cells and M cells exhibited significantly lower expression levels compared to other cell subtypes.

Furthermore, within F-cells, the expression of primary bile acid biosynthesis genes was higher in the LF group than in the LFB group. Single-cell gene set enrichment analysis (scGSEA) synthesizes latent data representation and gene set enrichment scoring to identify coordinated gene activity at single-cell resolution. The experimental results demonstrated that bile acid supplementation upregulated gene expression in the steroid hormone biosynthesis pathway across F, R, and B cells. Except for F-cells, bile acid supplementation upregulated gene expression in the steroid biosynthesis pathway across R cells and B cells (Fig. S1 A–F). The results can be attributed to reduced synthesis of non-hormonal steroid compounds in F cells. This mechanistic insight explains why the LFB group exhibited lower expression of primary bile acid biosynthesis genes in F cells compared to the LF group.

3.7. The impact of bile acid supplementation on intercellular communication in hepatocytes of L. vannamei

Intercellular communication between the remaining six major cell subtypes (endothelial cells, granulocytes, phagocytes, keratinocytes, myoepithelial cells, and peritrophic matrix) and hepatopancreatic cells are illustrated in Fig. 8A and B. Among these, myoepithelial cells, endothelial cells, phagocytes, and keratinocytes exhibited stronger interactions with hepatocytes. Hepatocyte-myoepithelial cell interactions were the most frequent, while interactions with endothelial cells and phagocytes showed the highest intensity. Compared to LFB, the LF group displayed more differential interactions between keratinocytes and hepatocytes, along with significantly greater intensity differences in hepatocyte-endothelial cell and hepatocyte–phagocyte interactions (Fig. 8C and D). Comparative analysis of differential signaling pathways between the two groups identified four key shared pathways: neural cell adhesion molecule (NCAM), collagen, laminin (LAMININ), and protease-activated receptors (PARs) (Fig. 8E). Notably, LFB showed significantly stronger enrichment in the contactin (CNTN) pathway than LF, whereas LF exhibited higher enrichment levels in NCAM and collagen pathways.

Fig. 8.

Effects of supplementing bile acids in low fishmeal diet on cell–cell communication. (A and B) The number and strength of interactions among hepatopancreas cells and other cell populations. (C and D) The intergroup differences in the number and strength of the interactions among hepatopancreas cells and other cell populations (HF vs. LF, LF vs. LFB). (E) Total signaling pathway information flow across HF, LF, and LFB groups. (F and G) Interaction metrics (quantity and intensity) between hepatopancreas subpopulations. (H and I) Comparative interaction analysis of hepatopancreas subpopulations (HF vs. LF, LF vs. LFB). (J) Cells with increased signal transduction of the neural cell adhesion molecule (NCAM) and protease-activated receptors (PARs) pathways in the LFB group. (K) The overall signal patterns of the LF group and the LFB group. H0–H11 respectively represent F cells, R cells, B cells, endothelial cells, E cells, granulocytes, M cells, phagocytes, immature R cells, keratinocytes, myoepithelial cells, and peritrophic matrix. HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

Further investigation of inter-subtype communication within hepatopancreatic cells (F cells, R cells, B cells, E cells, M cells, and immature R cells), revealed higher interaction frequencies among R cells, E cells, B cells, and immature R cells, with B cells demonstrating the highest interaction frequency overall (Fig. 8F). The strongest interaction intensities were observed between F cells and B cells, as well as R cell–R cell pairs (Fig. 8G). Comparative analysis between LF and LFB groups highlighted (Fig. 8H and I): (1) the most pronounced difference in interaction frequency between M cells and R cells, favoring LFB; (2) the largest difference in interaction intensity for R cell–R cell pairs, with LF exceeding LFB; (3) the second-largest intensity difference in F cell–R cell interactions, with LFB surpassing LF. The stronger PARs enrichment in LFB was primarily driven by differential ligand-receptor pairs in R cell–R cell, B cell–R cell, and M cell–R cell interactions (Fig. 8J). Both groups shared two core pathways: NCAM and PARs. In LF, R cells displayed stronger NCAM ligand-receptor signals than B- and F cells but weaker PARs signals. Importantly, F cell–R cell interactions in the PARs pathway were significantly stronger in LFB than LF (Fig. 8K).

3.8. Pseudotime trajectory analysis of hepatocyte subpopulations and the impact of bile acid supplementation on cell cycle dynamics

To track the dynamic differentiation process of various cell subtypes, cells were arranged along a trajectory using Monocle 2. The trajectory was divided into three stages: before branch point 1 (Branch 1) and two branches (Branch 2 and Branch 3). Cells from the LFB group were primarily located in Branch 1, while samples from the HF and LF groups were mainly distributed in Branches 2 and 3. Branch 1 was predominantly composed of F cells, Branch 2 mainly consisted of M cells and E cells, and Branch 3 was primarily composed of B cells. R cells and immature R cells were mainly distributed around the branch points (Fig. 9A–C). Further analysis using Monocle 3 allowed the arrangement of cell subpopulations along a pseudotime trajectory, and the cell positions were compared with the differentiation paths displayed on the pseudotime axis trajectory map. The results revealed that in the hepatocytes of L. vannamei, subcellular differentiation began with E cells as the starting point, differentiated into R cells, and then further branched into B-cells and F-cells (Fig. 9D and E). Fig. S2 reflects the gene clusters that determine the differentiation fate in all cell subsets. Cell fate-determining genes were divided into five clusters, with Gene Cluster 1 highly expressed in cells of Branch 1 and Gene Cluster 2 highly expressed in cells of Branch 3. The KEGG enrichment analysis of these gene clusters showed that Gene Cluster 1 was primarily enriched in functions related to nutrient digestion and absorption (lysosome, pancreatic secretion, protein digestion, and absorption) (Fig. 9J), while Gene Cluster 2 was mainly enriched in endocytosis (Fig. 9K).

Fig. 9.

Developmental trajectory of hepatopancreas cells subtypes. (A) Temporal dynamics of differentiation trajectories. (B) Sample-specific differentiation progression. (C) Lineage development across distinct cell types. (D) Dimensionality reduction distribution diagram of cell differentiation trajectories based on Monocle3. (E) Pseudotime value trajectory diagram of cells. (F) The cell cycles of different groups. (G) The cell cycles of cells subtypes. (H) The cell cycle stages of hepatopancreatic cell subpopulations in each group. (I) Cluster 1: Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment for cell fate regulators. (J) Cluster 2: KEGG pathway enrichment for cell fate regulators. HF = high fishmeal diet; LF = low fishmeal diet; LFB = low fishmeal diet supplemented with 400 mg/kg bile acid.

The cell division cycles of samples and their subpopulations across different groups were statistically analyzed. The experimental results indicated that, compared to the LF group, more cells in the LFB group were in the S phase of the interphase, while fewer cells were in the G1 phase, suggesting that bile acid supplementation might promote hepatocyte DNA division (Fig. 9F). A higher proportion of R cells were in the mitotic phase compared to F cells and B cells (Fig. 9G). Comparing the distribution trends of subcellular division cycles across different groups revealed that none of the peritrophic membrane cells in the LF group were in the mitotic phase (M phase). However, after bile acid supplementation, these cells began to enter the mitotic phase. Additionally, more F cells in the LFB group were in the S phase of the interphase compared to the LF group, consistent with the aforementioned result that the number of F cells in the LFB group was higher than in the LF group (Fig. 9H).

3.9. Immunohistochemical validation of the proteins expressed by mark genes

Based on single-cell transcriptomic results, key proteins choline acetyltransferase, acyl-Coa synthetase short chain family member 1 (ACSS1), heat shock protein 60 (HSP60), macrophage colony-stimulating factor (M-CSF), matrix metallopeptidase 2 (MMP2), NOD-like receptor family pyrin domain containing 2 (NLRP2) and proliferating cell nuclear antigen (PCNA) were selected for immunohistochemical validation. The positive expression rate of choline acetyltransferase in the LFB group was significantly higher than that in the LF group (P = 0.041), with positive cells predominantly localized in subpopulations surrounding the hepatopancreatic tubules (Fig. S3A and H). The positive expression rate of ACSS1 in the LFB group was significantly higher than that in both the LF and HF groups (P = 0.001), with positive cells mainly concentrated in the F cell subpopulation (Fig. S1B and I). The positive expression rate of HSP60 in the LFB group was significantly higher than that in the LF group (P = 0.019), with positive cells primarily enriched in the R-cell subpopulation (Fig. S3C and J). The positive expression rate of M-CSF in the LFB group was significantly higher than that in the LF and HF groups (P < 0.001), with positive cells predominantly localized in the R cell and B cell subpopulations surrounding the hepatopancreatic tubules (Fig. S3D and K). The positive expression rates of MMP2 and PCNA were significantly higher in the LFB group than in the LF group (P = 0.001; P = 0.012), with positive cells mainly enriched in the R cell subpopulation (Fig. S3E and G, L, and N). In contrast, the expression rate of NLRP2 did not differ significantly between the two groups (P = 0.059) (Fig. S3F and M).

4. Discussion

Research on marine invertebrates at the cellular level has been hindered by the lack of suitable model organisms. A critical knowledge gap exists in the cellular study of the hepatopancreas—an organ central to lipid regulation—due to the absence of a traditional hepatopancreas in Drosophila, the primary invertebrate model organism. Current investigations into invertebrate metabolism and digestion focus on the midgut and fat body, which fail to provide a reference gene library for cell identification in decapods such as L. vannamei, where the hepatopancreas serves as the main digestive organ (Hung et al., 2020). However, with the emergence and widespread application of single-cell transcriptomic technologies, it has become possible to shift research in non-model organisms from tissue-level to cell-level heterogeneity. Previous studies have utilized single-cell transcriptomics to explore cellular response mechanisms to external stressors and construct cellular atlases in aquatic non-model species. Examples include analyzing hepatocyte differentiation trends and cell–cell interaction mechanisms in largemouth bass (Micropterus salmoides) under high-fat diet-induced liver injury (Xie et al., 2024), investigating cooperative mechanisms among hemocyte subpopulations in white spot syndrome virus (WSSV)-infected L. vannamei (Cui et al., 2024), and mapping hepatopancreas/hemocyte atlases in ammonia-stressed Penaeus monodon to further elucidate innate immunity mechanisms (Li et al., 2022b). This study aims to apply single-cell transcriptomic technology using L. vannamei as a model to address this gap in invertebrate hepatopancreas research by constructing the first cellular atlas of the hepatopancreas in marine invertebrates and investigating its cellular mechanisms in response to nutritional stress.

Invertebrates lack the capacity for de novo bile acid synthesis, yet bile acids are widely utilized as aquafeed additives to alleviate metabolic disorders. This study first validated the application effect of bile acids in L. vannamei. The study replaced 50% of fishmeal with cottonseed meal concentrate (CMC) in low-fishmeal diets, and this substitution level (50%) did not significantly affect the growth performance of the shrimp due to the high protein content and reduced antinutritional factors of CMC, which is consistent with the effect of CMC substitution in Penaeus monodon (Jiang et al., 2021). However, when the substitution ratio exceeded 60%, growth performance declined, suggesting that plant-based proteins may lead to a relative deficiency of bile acids. Notably, supplementation with 0.04% bile acids (LFB group) significantly improved growth performance—consistent with the effects observed by using 0.06%—indicating that bile acids are a critical factor in overcoming the limitations of low-fishmeal diets (Li et al., 2023b). The study further investigated the hepatopancreatic cellular and molecular mechanisms underlying this phenotype.

The key mechanism by which bile acids alleviate hepatopancreatic injury may lie in their ability to modulate cellular composition and inhibit the fibrotic process. Through hepatopancreatic cellular atlas analysis, we identified 12 major subpopulations, among which R, B, and F cells were recognized as key regulatory cell types. Previous studies have preliminarily explored the functions of these cell types, confirming that all three are involved in nutrient digestion and absorption. Specifically, in situ hybridization has verified that digestive enzymes (e.g., amylase, chitinase, cellulase, and trypsin) are localized to the F cells of the Penaeus monodon hepatopancreas (Lehnert and Johnson, 2002), suggesting that F cells mediate both extracellular and intracellular digestion (Ruiz et al., 2020). B cell vacuole size reflects digestive stages, with small vacuoles marking early digestion and large vacuoles indicating completion. Unlike other decapods, L. vannamei R-cells span the entire hepatopancreatic tubule (Caceci et al., 1988). Prior work indicates R cells exhibit heightened sensitivity to starvation and dietary changes, implying roles in nutrient absorption and storage. This study further investigated the functional heterogeneity of these cell types by analyzing enriched upregulated genes in each cell population. The analysis revealed that R cells and B cells possess redox-related functions, while B-cells showed enrichment in cytoplasmic/membrane pathways, validating their critical role in transmembrane transport. Notably, experimental results demonstrated that bile acid supplementation significantly altered the proportions of key cell types: it increased the abundance of F cells responsible for digestive enzyme secretion while reducing the populations of R cells and B cells associated with nutrient storage and oxidative stress. This finding aligns with quantitative trends observed in H&E staining, suggesting that bile acids may improve hepatopancreatic health by enhancing digestive and absorptive functions while reducing metabolic burden.

To gain deeper insights into how bile acids regulate the cellular microenvironment, the study employed the CellChat tool to analyze intercellular communication networks. CellChat is a computational tool that quantifies and analyzes cell–cell communication networks from snRNA-Seq data, enabling systematic exploration of global cellular communication patterns (Jin et al., 2021). Following bile acid supplementation, signaling pathways associated with inflammation and fibrosis—collagen and NCAM—exhibited attenuated activity. Excessive collagen deposition drives organ fibrosis (e.g., hepatic or pulmonary fibrosis) (Karsdal et al., 2020). The NCAM regulates cell–cell and cell–matrix interactions, as well as intracellular signaling, via cis- and trans-interactions with homophilic or heterophilic receptors (Walmod, 2004), thereby modulating critical pathways including protein kinase B (PKB/Akt) and nuclear factor-kappa B (NF-κB) Ditlevsen, 2003, Krushel, 1999. These findings suggest that bile acids may suppress the abnormal immune responses and hepatopancreatic fibrosis induced by low-fishmeal diets in L. vannamei. This hypothesis was confirmed by Sirius red staining: the LF group indeed exhibited higher levels of collagen deposition and fibrosis compared to the LFB group. These results collectively demonstrate that bile acids alleviate hepatopancreatic injury by remodeling intercellular communication networks and suppressing inflammatory and fibrotic processes.

The regulatory effects of bile acids were also evident at both transcriptomic and microbiomic levels, revealing their role in coordinating the “gut–liver axis.” Bulk RNA-Seq analysis revealed that the low-fishmeal diet (LF group) activated pathways related to fatty acid degradation and bile acid biosynthesis, likely representing a compensatory response. Non-alcoholic fatty liver disease is typically associated with steatosis and mild inflammation (Powell et al., 2021) and characterized by gut microbiota dysbiosis, which exhibits bidirectional interactions with bile acids (Mouzaki et al., 2016). Conversely, bile acid supplementation (LFB group) shifted the metabolic pattern toward NAFLD and oxidative phosphorylation pathways, suggesting that bile acids help restore normal energy metabolism. More importantly, by integrating snRNA-Seq and gut 16s sequencing data, the study uncovered evidence of bile acid-mediated coordination of the “gut–liver axis”: the LFB group uniquely exhibited enrichment of Maribacter (Flavobacteriaceae family), a catalase- and oxidase-positive genus (Nedashkovskaya et al., 2004). Catalase mitigates oxidative injury by degrading H₂O₂. Excessive reactive oxygen species (ROS) accumulation directly induces oxidative stress, a hallmark of NAFLD. Oxidative phosphorylation generates ROS (e.g., O₂^-, H₂O₂), which, if not scavenged, causes hepatopancreatic damage. Following bile acid supplementation, the reduced abundance of R cells (involved in oxidative phosphorylation) in shrimp hepatopancreas, coupled with the upregulation of Maribacter—a catalase-active genus—in the intestine, may collectively demonstrate a mechanism countering peroxidative damage. Another differential genus, Tamlana (Flavobacteriaceae), is esterase lipase (C8)-positive (Jung et al., 2019), catalyzing short-chain fatty acid degradation. Meanwhile, the population of F cells (responsible for nutrient digestion/absorption) in the hepatopancreas showed an upregulation trend. This indicates that bile acids may systematically improve the organism's lipid metabolism and oxidative homeostasis by simultaneously regulating both gut microbiota function and hepatopancreatic cell states.

Traditional morphological approaches have classified hepatopancreatic cells in crustaceans. Five cell types (E, F, R, B, and M) were identified in the hepatopancreatic tubules of the freshwater prawn Macrobrachium carcinus, with each type hypothesized to perform distinct functions (Ruiz et al., 2020). Two predominant hypotheses regarding cell lineage have been widely circulated: the first, termed the “dual-lineage hypothesis”, proposes that E cells give rise to two cell lineages—R cells and F/B cells—wherein B cells are defined as mature F cells that store synthesized digestive enzymes in central vacuoles (von Apáthy, 1908); the second hypothesis, referred to as the “tri-lineage hypothesis”, suggests that R cells, F cells, and B cells are independent lineages derived directly from E cells. This hypothesis was inferred based on differences in metal accumulation patterns among distinct cell subpopulations (Vogt, 2019). Pseudotime analysis captures cells in transient differentiation states to infer lineage relationships (Ji and Ji, 2016). This study employed Monocle 2 and Monocle3 to delineate cellular differentiation processes, which, unlike traditional lineage tracing, constructs trajectories using genes associated with differentiation states and pseudotime, followed by dimensional reduction to map cells onto a trajectory tree (Van den Berge et al., 2020). The study identified a novel lineage distinct from existing hypotheses: E cells first differentiate into R cells, which subsequently bifurcate into F and B cells. Furthermore, the higher proportion of R cells in the mitotic phase compared to F- and B cells supports this model. However, whether this lineage applies broadly to decapods or is specific to L. vannamei requires validation. It is particularly noteworthy that bile acid supplementation promoted the differentiation of R cells into F cells. This discovery not only provides new insights into the cell biology of the crustacean hepatopancreas but also explains, from a cellular origin perspective, why F cell populations increase after bile acid supplementation—it may optimize the cellular composition of the hepatopancreas by modulating cell fate determination, driving it toward a more efficient digestive and absorptive phenotype.

5. Conclusion

In summary, this study leverages single-cell transcriptomics to elucidate the mechanisms by which bile acids regulate lipid metabolism in marine invertebrates. This process is mediated through modulation of cellular composition and intercellular communication, effectively alleviating hepatopancreatic injury induced by low-fishmeal diets. Furthermore, the study uncovered a novel hepatopancreatic cell differentiation atlas, filling a critical gap in visualizing cellular lineage dynamics in marine invertebrates and providing a valuable reference for hepatopancreatic cell research in other decapods.

Credit Author Statement

Yuxuan Ma: Writing – original draft, Methodology, Data curation. Kangyuan Qu: Investigation, Formal analysis. Beiping Tan: Supervision, Project administration, Conceptualization. Shiwei Xie: Writing – review & editing, Supervision, Funding acquisition.

Declaration of competing interest

We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, and there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the content of this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: