This study was carried out in accordance with the recommendations of Care and Use of Laboratory Animals in China, Animal Ethical and Welfare Committee of China Experimental Animal Society. The protocol was approved by the Animal Ethical and Welfare Committee of Sun Yat-sen University (Guangzhou, China).

Four practical diets were formulated, a 25% FM diet was used as a control (diet A), and three other diets were prepared by replacing 10% of FM with soy protein concentrate. Low FM diets were supplemented with 0, 0.75, and 1.5% S. limacinum (diet B, C and D, respectively). S. limacinum was provided by Kingdomway (Xiamen, China), which contained 19.4% DHA. Formulation and proximate composition of the diets are presented in Table 1, and fatty acid compositions of the diets are showed in Supplementary Table 1. Crystalline amino acids were mixed and precoated with 1% carboxymethyl cellulose to prevent leaching loss. The extra supplemented vitamins and minerals were dissolved in 200 mL water and then mixed with the dry ingredients. The 1.2 mm diameter pellets were cold-extruded using a pelletizer (Institute of Chemical Engineering, South China University of Technology, Guangdong, China), heated in an electric oven at 90°C for 60 min, then air-dried to approximately 10% moisture, and stored at -20°C until used.

Juvenile P. monodon were obtained from the Shenzhen experimental base of South China Sea Fisheries Research Institute (Shenzhen, China). Prior to the experiment, shrimp were acclimated for 3 weeks and fed the commercial diet (25% FM). At the beginning of the experiment, 480 shrimp with average weight 1.01 ± 0.01 g were distributed randomly into 16 cylindrical fiberglass tanks (500 L), at a density of 30 shrimp in each tank. Each diet was randomly assigned to four tanks. All shrimp were fed to apparent satiation three times daily at 8:00, 13:00, and 18:00 for 8 weeks. Any uneaten feed was collected by siphoning and used to calculate feed intake.

During the experimental period, water temperatures ranged from 26 to 30°C, salinity was approximately 30‰, pH 7.7–8.0, ammonia nitrogen was lower than 0.05 mg L^–1, and dissolved oxygen was higher than 6.5 mg L^–1. Water quality parameters were measure by following the instructions of the detection kit (Sangpu, Beijing, China). Natural light-dark cycle was used during the trial.

After the 8 weeks feeding trial, shrimp from each tank were weighed, counted, and sampled 24 h after the last feeding. Eight shrimp per tank were used to collect hemolymph samples, and hemolymph were taken from the pericardial cavity using 1-ml syringe and then centrifuged (8,000 rpm) at 4°C for 10 min. Supernatant was pooled and stored at −80°C until analysis of the enzymatic activity and metabolomics. Hepatopancreas and intestine from two shrimp per tank were immediately flash-frozen in liquid nitrogen and stored at −80°C prior to RNA extraction.

Moisture, crude protein, and crude lipid contents in the diets were determined using standard methods (AOAC, 1995). Moisture was determined by drying at 105°C until constant weight. Crude protein was measured by the Kjeldahl method after acid digestion using an Auto Kjeldahl System (1030-Auto-analyzer; Tecator, Hoganos, Sweden). Crude lipid was determined by the ether-extraction method using a Soxtec System HT (Soxtec System HT6; Tecator, Sweden). Fatty acid compositions in diets were measured by China National Analytical Center (Guangzhou, China). Lipids were extracted by a mixture of chloroform and methanol (2:1, v/v), the lipid phase was evaporated, and fatty acids were then saponificated by potassium hydrate. Fatty acids methyl esters were separated and quanitified by a gas chromatograph (GC 7820A, Agilent, United States) equipmented with a HP-88 column (30 cm × 0.25 mm).

Lipid-related metabolits and immune-related parameters in hemolymph were determined in this study. Total cholesterol (TC) contents were determined by cholesterol oxidase-peroxidase method (Allain et al., 1974). The contents of glutathione (GSH) were measured by its ability to react with DTNB (dithiobis-2-nitrobenzoic acid), and malondialdehyde (MDA) was measured by thiobarbituric acid method (Janero, 1990). The activities of superoxide dismutase (SOD) were determined by a hydroxylamine method (Masayasu and Hiroshi, 1979). All these parameters were determined following the instructions of the detection kit (Nanjing Jiancheng Bioengineering Institute, China), and they were measured by monitoring the absorbance changes in a microplate reader (TECAN infinite M200, Switzerland).

Hematoxylin/eosin (H&E) staining and transmission eletron microscopy (TEM) were employed to evaluate the intestinal damage of shrimp. The same section of middle intestine from two shrimp per tank were fixed in Bouin’s solution for 12 h and then fixed in 70% ethanol. All fixed samples were dehydrated in a graded series of ethyl alcohol and embedded in paraffin. Sections (4 μm thickness) were stained with H&E and observed under a light microscope (Nikon Ni-U, Japan).

Another two intestinal samples per tank were used for TEM examination. As described by Zhang et al. (2012), samples were firstly fixed in 2.5% glutaraldehyde solution and then fixed in 1% OsO4 for 1 h, dehydrated in a graded series of ethyl alcohol, and embedded with resin. Resin blocks were sectioned using a diamond knife. Ultrathin sections (∼90 nm) from each sample were cut and placed on copper grids. Sections were stained with saturated uranyl acetate solution for 30 min, rinsed with distilled water and poststained with lead citrate for 30 min. Ultrathin sections were then screened with a TEM (JEOL JEM-1400, Japan) at 150 kV.

Total RNA from hepatopancreas and intestine was extracted with RNAiso Plus reagent according to the manufacturer’s instructions (TaKaRa, Japan). Agarose gel electrophoresis and spectrophotometric analysis (A260:A280 nm) were used to assess RNA quality and concentration. cDNA was synthesized using a PrimeScriptTM RT reagent kit with gDNA Eraser (Takara, Japan), according to the manufacturer’s instructions.

Real-time PCR for the target genes were performed using a SYBR^® Premix Ex Taq^TM II (Takara, Japan) and quantified on the LightCycler 480 (Roche Applied Science, Basel Switzerland). As described by manufacturer’s introduction, 400 nM of forward- and reverse-specific primers, 10 ng of cDNA template, and nuclease-free water to a final volume of 10 μL were used. Then we used the following program: denaturation step at 95°C for 1 min, 40 amplification cycles of 5 s denaturation at 95°C, 15 s annealing at 60°C, 20 s extension at 72°C, a melt-curve analysis, and, finally, cooling to 4°C. The efficiency of primers for each gene was previously evaluated by serial dilutions to ensure that it was close to 100% (Supplementary Table 2).

Normalization of real-time qPCR data was done by geometric averaging of multiple internal reference genes (elongation factor 1α and β-actin) (Vandesompele et al., 2002). Three independent duplicates were conducted for each of the data.

In total, 48 shrimp hemolymph samples were analyzed by UPLC-QTOF-MS in this research to investigate the impact of dietary S. limacinum on the lipid profiles of shrimp with 12 shrimp for each group. Briefly, 400 μL acetonitrile/methanol (9:1) was added to 100 μL hemolymph, and the mixture was vortexed for 30 s. The mixture was centrifuged at 13,000 g for 10 min at 4°C, and 200 μL supernatant was collected for UPLC-QTOF-MS analysis. Equal amounts of all samples were pooled as a QC sample for LC-MS system conditioning and quality control.

Ultra-performance liquid chromatography coupled with quadrupole time of flight mass spectrometry was performed on an ACQUITY UPLC system (Waters, Manchester, United Kingdom) coupled with a G2-Si HDMS QTOF mass spectrometer (Waters, Manchester, United Kingdom). Chromatographic separation was carried out at 40°C on an ACQUITY HSS T3 1-class column (2.1 mm × 100 mm, 1.8 μm, Waters). Mobile phase A consisted of 10 nM ammonium formate in water and 0.1% formic acid, and mobile phase B consisted of 10 nM ammonium formate in IPA/ACN (70/30) and 0.1% formic acid. The flow rate was 0.4 mL+/min. From the start to 0.5 min, B was held at 2%, linearly increased to 40% during the next 1 min, increased to 45% during the next 3.5 min, increased to 55% during the next 3 min, increased to 85% during the next 6 min, increased to 99% during the next 6 min, kept constant for 3 min, decreased to 2% in 0.1 min, and then kept constant for 2.9 min. Both positive and negative modes were performed and operated in continuum mode with an acquisition time of 0.3 s per scan. The scan range was set at 50–1500 m/z. The capillary was set at 2.5and 2 kV in positive ion mode and negative ion mode, respectively. Sampling cone voltages were set at 40 V in both modes. The desolvation temperature and gas flow were 350°C and 700 L/h, the source temperature was set at 120°C. Before analyzing the sample sequence, five QC samples were run, and, during the analysis of the sample sequence, one QC sample was run after every six injections.

The acquired mass data were imported to Waters Masslynx v4.1 software for peak detection and alignment. All of the data were normalized to the summed total ion intensity per chromatogram. We detected 4025 and 3233 features in positive and negative modes, respectively. All the features with CVs >30% were removed. The data were imported to SIMCA-P (Version 13.0, Umetrics, Sweden), where multivariate analysis were performe, mainly included principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA). The quality of the models is described by the R₂ and Q₂ values. R₂ is defined as the proportion of variance in the data explained by the models and indicates goodness of fit. Q₂ is defined as the proportion of variance in the data predicted by the model and indicates predictability, calculated by a cross-validation procedure. A default seven-round cross-validation in SIMCA-P was performed throughout to determine the optimal number of principal components and to avoid model overfitting.

Significantly different metabolites were selected on the basis of the combination of a statistically significant threshold of variable influence on projection (VIP) values obtained from the OPLS-DA model, P values from one way ANOVA analysis on the normalized peak areas, and the maximum fold change, where metabolites with VIP values larger than 1.0, P values less than 0.05, and fold change larger than 1.5 were included, respectively.

Metabolite identification was performed using Progenesis QI (Waters, Non-linear Dynamics, Newcastle, United Kingdom). LipidMaps, Chemspider and Human Metabolome databased (HMDB) were used for MS1 identification, and theoretical fragments were used for MS/MS identification. A score (total 60) of identified metabolites obtained from QI larger than 40 was regarded as acceptable, and anything else was regarded as uncertain.

Heatmap and cluster analyses were conducted using the pheatmap package based on R software (version 3.7.2).

Total bacterial DNA in the intestine was extracted with a PowerFecal^® DNA Isolation kit. DNA samples were amplificated (Sutton et al., 2013) and then quantified using Quant-iT^TM dsDNA HS Reagent. High-throughput sequencing analysis of bacterial rRNA was performed by Illumina HiSeq 2500 at Biomarker Tchnologies Corporation (Beijing, China). The splicing and filtering of the raw data were performed using FLASH v1.2.7 and Trimmornatic v0.33, and cleaning of chimera were performed using UCHIME v4.2. OTUs were obtained by QIIME (version 1.8.0), Alpha diversity was evaluated by Mothur (version v.1.30). We used binary jaccard to estimate β diversity (QIIME).

The parameters were calculated as follows:

Percentage weight gain (WG, %) = 100 × (final body weight - initial body weight)/initial body weight

Feed efficiency (FE) = (final body weight - initial body weight)/feed consumed

Survival (%) = 100 × (final amount of shrimp)/(initial amount of shrimp)

The results are presented as the means ± SEM (standard error of the mean). All the data were statistically analyzed by SPSS 19.0 (SPSS, Chicago, IL, United States). The data were first tested for homogeneity; if the data had similar variances, then a one-way ANOVA was used to test the main effect of dietary treatment. When there were significant differences (P < 0.05), the group means were further compared using Duncan’s multiple range test. When data did not have a homogeneous viaration, the non-parameter Kruskal–Wallis test was applied and followed by all pairwise multiple comparisons if the results of Kruskal–Wallis test shown significant difference (P < 0.05).

The final body weight (FBW), WG, and FE were higher in shrimp fed diet A than those fed diet B and D (P < 0.05; Figure 1), and there were no differences between group A and C. Survival was similar between all the groups (P > 0.05).

Hemolymph TC content in shrimp fed diet B was higher than those fed diets containing S. limacinum (P < 0.05; Figure 2). S. limacinum supplementation significantly increased hemolymph MDA content of shrimp (P < 0.05). The GSH content in hepatopancreas was higher in group C than the other groups (P < 0.05). The SOD activities in hemolymph and hepatopancreas showed no significant differences between the four groups, the MDA contents in hepatopancreas showed a similar trend (P > 0.05).

Low FM diet suppressed mRNA expression of immune deficiency (IMD) and endoplasmic reticulum protein57 (ERP57) in the hepatopancreas (P < 0.05), and supplementation of S. limacinum had no effect on their expression (Figure 3). However, S. limacinum supplementation alleviated the low expression of SOD and inhibitor of apoptosis proteins (IAP) induced by a low FM diet (P < 0.05). Shrimp fed diets containing S. limacinum obtained higher catalase (CAT), myeloid differentiation primary response gene 88 (MyD88), and autophagy-related protein 8 (ATG8) mRNA expression compared to those fed other diets (P < 0.05). Shrimp fed diet C obtained the highest mRNA expression of heat shock protein 70 (HSP70), Relish, X-box binding protein 1 (XBP1), defender against cell death 1 (DAD1), and ubiquitin conjugated enzyme 2 (UCE2) between all the groups. The mRNA level of Tube in group D was higher than that of other groups.

Shrimp fed diet B obtained lower Relish and higher XBP1 expression than those fed other diets (P < 0.05; Figure 4). Superoxide dismutase, extracellular signal-regulated kinase (ERK), Toll, and IAP mRNA levels in the intestine were lowest in shrimp fed diet D (P < 0.05). Low dietary FM downregulated the mRNA expression of Tube (P < 0.05). There were no differences in the expression of HSP70 and tumor necrosis factor receptor associated factor 6 (TRAF6) among all the treatment groups (P > 0.05).

Intestinal H&E staining showed that intestinal morphology were adversely affected by diet B (Figure 5), the intestinal fold height was lower in shrimp fed diet B compared with those fed diets A and C (Figure 6). TEM analysis also demonstrated that the intestine suffered more serious damage in shrimp fed diet B (Figure 7), especially the endoplasmic reticulum (ER), which showed obvious swelling in group B. The ER back to normal after dietary supplementation with S. limacinum, and S. limacinum supplementation increased the microvilli length of shrimp (P < 0.05).

In this study, a non-targeted metabolomics strategy was applied. PCA was applied to evaluate the separation between the four groups, Figures 8A,B show a clear cluster in ESI+ (R₂X = 0.982, Q₂ = 0.911) and ESI- (R₂X = 0.775, Q₂ = 0.654) modes. A supervised multivariate data analysis method, OPLS-DA, was established for testing differences between features that clearly distinguished the different groups in ESI+ (R₂X = 0.906, R₂Y = 0.62, Q₂ = 0.497) and ESI- (R₂X = 0.791, R₂Y = 0.97, Q₂ = 0.923) modes (Figures 8C,D), showed a goodness of fit (R₂) and predictive ability (Q₂) in this model.

Features with a VIP value larger than 1.0, P value less than 0.05 and fold change larger than 1.2 were regarded as significantly different features. In total, 218 significant features were found to satisfy the criterion (119 in ESI+ and 99 in ESI-). Among them, 41 in ESI+ and 54 in ESI-were identified, 20 in ESI+ and 39 in ESI- were regarded as reliable. Therefore, 59 significantly different metabolites were identified in this experiment and are presented in Table 2.

A heatmap based on significantly different metabolites showed clear differences between shrimp fed diet A and other diets (Figure 9). Most of these metabolites were lipids, especially phospholipids, including phosphatidylcholine (PC), phosphatidyl ethanolamine (PE), Phosphatidyl serine (PS), sphingolipid and fatty acids. Shrimp fed high FM showed higher levels of lowly unsaturated LysoPCs (LPC) and highly unsaturated fatty acids [HUFAs; DHA, eicosapentaenoic acid (EPA), and arachidonic acid (ARA)] in the hemolymph, than other groups (P < 0.05). Meanwhile, S. limacinum supplementation in low FM diets could not alleviate this situation. The proportion of 2′-deoxyinosine and inosine were lower in shrimp fed low FM diets than those fed a high FM diet, and a similar tendency was found in the proportions of L-tryptophan and cholic acid (P < 0.05). S. limacinum supplementation increased the content of lysoPE (LPE; 22:6) and 6′-hydroxysiphonaxanthin in hemolymph. There were no clear trends found in the effects of different diets on unsaturated degree of phospholipids.

A total of 3,027,778 high quality sequences were produced in this study, with an average of 189,236 sequences per sample. The coverage of each group exceeded 99%, which indicated that the 16S rRNA sequences identified in each group represented the majority of bacteria in the samples analyzed in this study. As shown in Table 3, the Shannon index was higher in shrimp fed a high FM diet compared to those fed diet B and D, which indicated the higher bacterial diversity in group A. There were no differences in OTUs numbers and community richness (Chao1 and ACE) between all the treatments. The analysis of PCoA showed that first two principal components explained 35.12% of the variations (Supplementary Figure 1).

The relative abundance of gut microbiota in each group at phylum level is shown in Supplementary Figure 2. The dominant phyla in the four groups were Proteobacteria, Bacteroidetes, Actinobacteria, Cyanobacteria, and Tenericutes. The relative abundance of Proteobacteria were exceeded 90% in all the groups. The relative abundance of Bacteroidetes, Actinobacteria, and Tenericutes were highest in shrimp fed diet A, while the relative abundance of Proteobacteria showed an opposite trend. The relative abundance of Cyanobacteria were highest in shrimp fed diet B and lowest in shrimp fed diet C.

The relative abundance of gut microbiota in each group at genus level is shown in Supplementary Figure 3. The dominant genera in four groups were Vibrio, Pseudoalteromonas, Photobacterium, Halocynthiibacter, and Shimia. The relative abundance of Vibrio was lower in shrimp fed the high FM diet, compared with those fed low FM diets (P < 0.05). The relative abundance of Silicimonas, Hahella, and Erythrobacter were higher in shrimp fed the high FM diet compared to those fed low FM diets (P < 0.05; Figure 10). The relative abundance of Thalassotalea and Tenacibaculum were higher in group B compared with that of the other groups (P < 0.05), while the relative abundance of Maribacter showed an opposite tendency. S. limacinum supplementation increased the relative abundance of Sphingomonas in the intestine.