Metformin (97%) and compound-c [98.59%, C₂₄H₂₅N₅O; 6-[4-(2-piperidin-1-ylethoxy) phenyl]-3-pyridine-4-ylpyrazolo [1, 5-a] pyrimidine] used in this study were purchased from Sigma-Aldrich. The ingredients and nutrient content of the experimental diets are shown in Table 1. The levels of nutrients were designed to meet the requirements of common carp (Cyprinus carpio) according to the National Research Council [NRC] (2011). Fish meal (Con) used as the dietary protein source was the control group, 40% soybean meal (Dou), 300 mg kg^–1 metformin (Met) (Silva et al., 2015), and 0.2 mg kg^–1 compound-c (c-Com) (Kim et al., 2011), which added to soybean meal were experimental groups. Amino acids were balanced with coated amino acids. All diets were isonitrogenous and isolipidic. All the feed ingredients were stirred evenly in the mixer. After adding lipid and water, the mixture was homogenized and extruded through a pelletizer with a 2 mm mold. The diets were air-dried at room temperature and stored at −20°C until feeding.

The juvenile mirror carps (Cyprinus carpio Songpu) used in this study were purchased from Heilongjiang River Fishery Research Institute of the Chinese Academy of Fishery Sciences. Before the trial, the fish were acclimatized to the recirculating aquaculture system for 2 weeks. Then, 348 fish with a mean initial weight of 9.51 (±0.06) g were randomly assigned to 12 aquariums (500 L water) with an initial stocking density of 29 fish per aquarium. Each feed was randomly assigned to three aquariums. In the feeding trial, each diet was fed at 5% of the bodyweight of fish twice per day for the 59 days. The water temperature and pH were 26 (±2)°C and 7.0 (±0.3), respectively, dissolved oxygen was higher than 5 mg/l. The water quality was tested using YSI Professional Plus, and one-third of water was changed a week thrice to remove solid substances and reduce ammonia concentration.

At the beginning and end of the trial, fish in each aquarium were counted and weighed. After that, 18 fish per diet were anesthetized by MS-222 (100 mg/L), the intestine was quickly removed on ice, dissected longitudinally, rinsed in cold physiological saline (0.86% NaCl solution, pH = 7.2), and the intestinal mucosa was scraped with a glass slide. The intestinal mucosas of three fishes were pooled into one sample, frozen in liquid nitrogen, and stored at −80°C. Samples were collected on the 7th (short-term) and 59th day (long-term). The following growth parameters were calculated after the feeding trial: weight gain rate (WGR), specific growth rate (SGR), feed conversion ratio (FCR), and survival rate (SR).

Metabolite extraction and liquid chromatograph-mass spectrometer analysis were performed by Shanghai Majorbio Bio-pharm Technology Co., Ltd. (Shanghai, China). Metabolite data and profilings were obtained and analyzed on the free online platform of Majorbio Cloud Platform.¹ After identification and quantification, metabolites were annotated into HMDB (human metabolome database) and KEGG (Kyoto encyclopedia of genes and genomes), respectively. Differential metabolites in the mucosa of different treatments were identified using univariate analysis and multivariate statistics. Univariate analysis was performed by the methods of t-test and folded change (FC) analysis (P < 0.05; FC ≥ 1 or ≤1). Multivariate statistics were performed by Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) with variable importance for the projection (VIP) values (≥1.0). HMDB pie chart showed the superclass of differential metabolites between groups, and KEGG was used to help in understanding the functions of metabolites. Growth performance indicators were analyzed by one-way ANOVA using SPSS IBM Statistics 25. Duncan’s analysis was used for multiple comparisons.

Compared with the Con group, the WGR and SGR of fish fed the soybean meal, metformin, and compound-c diets were significantly decreased (P < 0.05). FCR of fish in the experimental groups was significantly increased than those fish in the Con group (P < 0.05). There was no significantly difference in SR of fish among groups (P > 0.05, in Table 2).

The metabolite peaks extracted from all samples and QC samples were analyzed by principal component analysis (Figure 1). It was found that QC samples clustered together, indicating that this experiment had good repeatability, a stable system, and reliable experimental results. Con on 7th day, Dou on 7th day, and Met on 7th day were dispersed among samples, while c-Com on 7th day, Con on 59th day, Dou on 59th day, Met on 59th day, and c-Com on 59th day groups had more concentrated sample points.

The identification and quantification results of metabolites showed that a total of 1,332 metabolites had been obtained, of which 1,157 and 363 were annotated into HMDB (Human Metabolome Database) and KEGG (Kyoto Encyclopedia of Genes and Genomes), respectively. Score plots of the (O) PLS-DA performed to verify the differentiated metabolites between the two groups and multivariate analysis supervision are shown in Figures 2–6. R²X (cum) and R²Y (cum) represent cumulative interpretation rates. Q² indicates the prediction ability of the model. In all comparisons, R²Y was close to 1, indicating that the model is reliable. In Dou vs. Con, Met vs. Con, c-Com vs. Con, and c-Com on 59th day, vs. Dou on 59th day groups, Q² > 0.5 indicated the model’s prediction ability was good. The pie chart (Figure 7) showed the HMDB superclass classification of the different metabolites between every two groups. As shown in the figure, Lipids and lipid-like molecules (purple) accounted for the largest proportion of all groups, respectively. The differential metabolites selected above were enriched into the KEGG database for analysis to obtain differential metabolic pathways. The differential metabolic pathways were sorted according to their significance (P-value) (Figure 8), and the differential metabolites involved in the differential metabolic pathways were emphatically analyzed (Supplementary Table 1).

Sphingosine affected intestinal mucosal cells through apoptosis and necroptosis. Sphingolipids and their metabolites are important active molecules, which participate in many important signal transduction processes such as cell growth, differentiation, senescence, and programmed cell death, including many bioactive metabolites ceramide, sphinganine, sphingosine, and sphingosine-1-phosphate [SM(d18:0/14:0)]. Ceramide can be converted to sphingosine. Activated sphingosine kinase catalyzes sphingosine phosphorylation to sphingosine-1-phosphate. Both ceramide and sphingosine can act as second messengers, promoting cell death, cell senescence, and growth stagnation, while sphingosine-1-phosphate regulates signaling pathways that stimulate cell growth and inhibit programmed cell death (Petra et al., 2020). Previous studies have shown that sphingosine induces apoptosis, which depends on the activation of Caspases-3, -7, and -8 (Chang et al., 2003). Several previous studies showed that the pro-apoptotic gene caspases-3, -7, and -8 were up-regulated by high-dose soybean meal (Bakke-McKellep et al., 2007; Chen et al., 2018; Duan et al., 2019). In this study, soybean meal significantly affected the contents of sphingosine and SM(d18:0/14:0) metabolites. Adding metformin was the opposite of that of the soybean meal group, and compound-c was similar to the soybean meal group. Metformin activated AMPK through inhibiting mitochondrial complex I, which led to the inhibition of mTOR signaling, autophagy, and apoptosis activation, decreased ROS generation, DNA damage, and inflammatory response (Lv and Guo, 2020). Compound-c is the primary reagent used as an AMPK inhibitor (Dasgupta and Seibel, 2018). In previous studies, compound C induced apoptosis, but AMPK was not involved in this effect (Jin et al., 2009). Consistent with the studies mentioned above, these results indicate soybean meal-induced intestinal mucosal cell apoptosis with AMPK independence.

Phosphatidylcholine, also known as lecithin, is an important component of the cell membrane. PC and LysoPC were involved in choline metabolism in cancer in this study. This study has shown that PC and LysoPC were down-regulated by soybean meal. However, PC and LysoPC in the Met group did not change significantly compared with the Dou group. These suggested that lipid metabolism, especially PC and LysoPC metabolism in the intestinal mucosa of carp may be affected through AMPK signaling pathway.

Further more, taurine, glycocholic acid, and betaine were changed by soybean meal whether activator of inhibitor was added or not. According to Valenzuela et al. (2021), tolerance to soybean meal is associated with immune and lipid metabolism-related pathways in fish. The results of this experiment are consistent with it. Acted as an activator of AMPK, metformin administration increases the bile acid pool in the intestine, which may affect glucagon-like-peptide-1 secretion and cholesterol levels (Lv and Guo, 2020). Therefore, metformin down-regulated taurocholic acid of cholesterol metabolism with consistent function. The result indicated that soybean meal could affect lipid metabolism of intestinal mucosa of carp unrelated to AMPK signaling pathway.

Aminoacyl-tRNA synthetases (aaRS) participate in one of the most important steps of life activities–protein biosynthesis. Hence, in the intestinal mucosa, the level of amino acids enriched in the pathways can reflect the availability of amino acids. This experiment found that amino acids were down-regulated in experimental groups, enriched in protein digestion and absorption pathways, ABC transporters, central carbon metabolism in cancer, aminoacyl-tRNA biosynthesis, so soybean meal significantly reduced the amino acids content and the utilization in the carp intestines. This is the same as the results of Zhou in carps (Zhou, 2021), and the difference with the conclusion from Wei et al. (2017) that increased the utilization of amino acids in fish muscles, which the different types of fish should cause a difference. In contrast, metformin up-regulated amino acids in the result. This maybe related with that metformin activates AMPK, which leads to the inhibition of mTOR signaling, and as a result, protein synthesis is disturbed (Lv and Guo, 2020).

In addition to that, arginine plays a vital role in regulating animals’ nutritional metabolism, growth, and development (Rhoads and Wu, 2009). Arginine stimulates protein synthesis, cell proliferation, DNA synthesis, cell protection, and migration (Rhoads et al., 2004; Corl et al., 2008; Tan et al., 2010; Tan et al., 2015). Activation of the mTOR signaling pathway is an important mechanism for arginine to alleviate cell stress injury (Tan et al., 2010). Compared with Con, enriched in the mTOR signaling pathway, arginine was significantly down-regulated in experimental groups. Zhang et al. (2013) found that anti-nutritional factors in soybean decreased TOR gene expression in foregut and midgut and increased 4E-BP gene expression in the hindgut of Jian carp. This is consistent with this experiment. These suggested that soybean meal caused intestinal mucosa injury, which may be related to the down-regulation of arginine, which does not activate the mTOR signaling pathway.

In addition to lipid and protein, sugar is also one of the essential energy materials in fish, which can be directly used to power the body to reduce fat and protein consumption in fish. This experiment suggested that the soybean meal group significantly raised sugar metabolites on the 59th day. This was the same as Zhou’s research in carp and the result of Wei in turbot (Scophthalmus maximus L) that glucose increased in the liver metabolites. Compared with the soybean meal group, isocitrate was significantly up-regulated and enriched in the citrate cycle in Met on 7th day, indicating regulating glucose metabolism. This result was consistent with the property that metformin can regulate the AMPK signaling pathway and further increase glucose catabolism by activating the AMPK signaling pathway (Lv and Guo, 2020). However, this phenomenon did not last for 59 days. It was speculated that soybean meal had an effect on glucose metabolism in the intestinal mucosa; as a result, there was no significant change in glucose metabolism when Met was added. Therefore, adding an activator could not enhance glucose metabolism on the 59th day. Additionally, metabolites in purine metabolic pathways were significantly increased in Dou on 59th day vs. Con on 59th day and c-Com on 7th day vs. Dou on 7th day, and not changed between Met and Dou, which indicated that both soybean meal and compound-c up-regulated the nucleotide metabolism of intestinal mucosal cells. We speculated that the mechanism of soybean meal on intestinal nucleoside metabolism might not be through the AMPK signaling pathway.

Adenosine triphosphate box transporter (ATP-binding Cassette transporter, ABC transporter) is involved in many important physiological processes, such as nutrient intake, cellular detoxification, lipid homeostasis, signal transduction, disease virus defense, and antigen presentation (Wilkens, 2015), whose primary main function is to utilize the energy generated by ATP (adenosine triphosphate) hydrolysis to transport the substrate across the membrane against the concentration gradient (Borths et al., 2002; Locher, 2004; Kathawala et al., 2015). This experiment found down-regulated amino acids in experimental groups enriched in ABC transporters. At the same time, compared with the Con group, ADP and AMP of Dou on 59th day, Met on 7th day and 59th day and c-Com on 7th day, and 59th day were significantly increased and enriched in FoxO and AMPK signaling pathways. The content of ADP, AMP, and ATP in intestinal mucosa reflects the level of energy metabolism. Metformin acts as an AMPK activator by inhibiting mitochondrial complex I (Batandier et al., 2006; Viollet et al., 2012). Mitochondrial complex I is critical for electron transport. As a result, the production of ATP was reduced, and the intracellular concentration of ADP increased. Consequently, the cellular levels of AMP increases and ultimately activate AMPK (Viollet et al., 2012; Pryor and Cabreiro, 2015). Members of the class O of forkhead box transcription factors (FoxO) have essential roles in metabolism, cellular proliferation, stress tolerance, and probably lifespan (Weigel et al., 1989). The AKT/PKB-mediated phosphorylation of FoxO inhibits FoxO function by promoting its interaction with 14-3-3 proteins and its relocalization from the nucleus to the cytosol (Van der Horst and Burgering, 2007). Lee et al. (2017) suggested that FoxO transcription factors could feasibly act downstream of AMPK to mediate the induction of antioxidant systems in response to H₂. This study suggested that soybean meal can improve ADP and AMP content then activate the AMPK and FoxO pathway.

The soybean meal can cause intestinal oxidative damage, which has been shown in grass carp (Duan et al., 2019; Zhang et al., 2020), turbot (Chen et al., 2018), and Scophthalmus maximus L. (Gu et al., 2021). In this study, enriched in glutathione metabolism, oxidized glutathione was up-regulated in the soybean meal group on the 7th day, and the amount of oxidized glutathione in the Met group was significantly higher than that in the Dou group at 7th day. These suggested that soybean meal might cause oxidative damage to the intestinal mucosa in short-time treatment, and metformin supplementation aggravated the injury. However, in a previous study, metformin suppresses mitochondrial complex I, thereby preventing the generation of reactive oxygen species (ROS) and further decreasing DNA damage (Algire et al., 2012). The results of this experiment are inconsistent with this and require further validation.