Sixteen 30-day-old male weaned piglets with similar weights (22.9 ± 0.2 kg) were randomly divided into four groups: a normal temperature group (24 ± 2°C, NT) and a cold exposure (4 ± 2°C) 2-h group (CS2), 6-h group (CS6), and 12-h group (CS12). The piglets in all groups were reared in artificial intelligence climate chambers for 7 days with a 12/12-h light/dark cycle (light from 8:00 a.m. to 8:00 p.m.). The temperature was maintained at 24 ± 2°C, and the relative humidity was 40–50%. Then, the piglets in the cold exposure groups were exposed to 4 ± 2°C conditions for 2, 6, or 12 h in artificial intelligence climate chambers, while the normal temperature control group was kept at 24°C. All piglets were fed the same commercial pellet feed and water, which were provided ad libitum. At the end of the experiment, all piglets were anesthetized and immediately euthanized. Liver samples (3–5 g) were then frozen with liquid nitrogen and stored in a -80 °C freezer until analysis by LC–MS metabonomics.

Liver samples (50 mg) were added to 1,000 μl of the extraction solution (methanol: acetonitrile: water = 2: 2: 1 (V/V), including 2 μg/ml of the internal standard L-2-chlorophenylalanine), vortexed, homogenized, and centrifuged at 13,800 ×g for 10 min at 4°C. Then, 400 μl of the supernatant was placed in an EP tube for vacuum drying, 200 μL of 50% acetonitrile was added for further dissolution, and then, the samples were vortexed and centrifuged at 16,200 ×g for 15 min at 4°C. Then, 75 μL of the supernatant was added to the sampler bottle (Dunn et al., 2011).

The supernatant in the sampler bottle was analyzed using a 1,290 Infinity series UHPLC System equipped with an UPLC BEH Amide column (2.1 × 100 mm, 1.7 μm, Waters) (Agilent Technologies) with a mobile phase comprising 25 mmol L⁻¹ ammonium acetate and 25 mmol L⁻¹ ammonia hydroxide in water (pH = 9.75) (A) and acetonitrile (B). The analysis was carried out with an elution gradient as follows: 0 ∼ 0.5 min, 95% B; 0.5 ∼ 7.0 min, 95 ∼ 65% B; 7.0 ∼ 8.0 min, 65 ∼ 40% B; 8.0 ∼ 9.0 min, 40% B; 9.0 ∼ 1 min, 40 ∼ 95% B; and 9.1 ∼ 12.0 min, 95% B. The column temperature was 25°C, and the injection volume was 2 μL (positive) or 2 μL (negative).

A triple TOF 6600 mass spectrometer (AB Sciex) was used to acquire MS/MS spectra on an information-dependent basis (IDA). The acquisition software (Analyst TF1.7, AB Sciex) continuously evaluates the full-scan survey MS data as it collects and triggers the acquisition of MS/MS spectra depending on preselected criteria. The ESI source conditions were as follows: Gas 1 was 60 psi, Gas 2 was 60 psi, curtain gas was 35 psi, source temperature was 600°C, declustering potential was 60 V, and ion spray voltage floating (ISVF) was 5000 V or -4000 V in positive or negative modes, respectively.

MS raw data files were converted to the mzXML format by ProteoWizard and processed by the R package XCMS (version 3.2). The process includes peak deconvolution, alignment, and integration. MinFrac and cut off were set as 0.5 and 0.3, respectively. The in-house MS2 database (based on HMDB, PubChem, KEGG, and METLIN) was applied for metabolite identification.

To obtain an overview of the chemical changes in the piglet liver during different cold exposure durations, the data were analyzed by unsupervised PCA. The PCA scores are shown in Figure 1. The data showed one principal component (PC) (PC1), two PCs (PC1 and PC2), and three PCs (PC1, PC2, and PC3), which accounted for 37.0, 59.2,, and 70.3% of the variation, respectively.

To further verify the significant difference between the control and cold exposure groups, we used the supervised OPLS-DA multivariate method for each of the two comparisons. The OPLS-DA score plots indicate the good fitness and high predictability of the model, with high statistical values of R²Y and Q². The parameters of R²Y and Q²Y for CS2-NT, CS6-NT, and CS12-NT were 0.991 and 0.27, 0.997 and 0.379, and 0.999 and 0.482, respectively (Figures 2A,C,E). The permutation test results showed that all Q² points from left to right were lower than the original red Q² points on the right side, indicating that the model was robust and reliable without overfitting (Figures 2B,D,F). The control and cold exposure groups could be clearly differentiated, suggesting that the metabolic profiles changed significantly and could be exploited in the subsequent analysis.

According to the supervised manipulations, potential variables were chosen as differentially abundant metabolites based on VIP >1, fold change >2, and p < 0.05 ((Thévenot et al., 2015)). There were 10 differentially abundant metabolites (6 upregulated and 4 downregulated) in the CS2 group, 4 upregulated differentially abundant metabolites in the CS6 group, and 10 differentially abundant metabolites (6 upregulated and 4 downregulated) in the CS12 group compared with the control group (NT) (Table 1 and Figure 3).

Among the NT and cold exposure groups, there were 8 known differentially abundant metabolites, which were primarily important intermediate metabolites for body metabolism.

The metabolic pathways of differentially abundant metabolites were explored using the KEGG database. The pathway enrichment analysis of metabolites was performed via the R package clusterProfiler. The different metabolic pathways between the NT and CS groups are presented in Figure 4, and network maps of metabolites and metabolic pathways between the NT and CS groups are shown in Figure 5.

Metabolic pathway analysis is very important for metabonomics and useful for identifying the metabolic signals and pathways of metabolites and exploring the related metabolites and genes (Li et al., 2019). There were eight main differentially abundant metabolites involved in three KEGG pathway classifications (metabolism, organismal systems, and environmental information processing) between the NT group and the cold exposure groups. Compared with the NT group, the differentially abundant metabolites of the CS2 group participated in 20 metabolic or biosynthetic pathways in the body; those of the CS6 group participated in 17 metabolic or biosynthetic pathways, and those of the CS12 group participated in 11 metabolic or biosynthetic pathways.

Metabolic pathways and enrichment analyses revealed that most of the altered pathways were related to body metabolism, such as 2-oxocarboxylic acid metabolism, alanine, aspartate and glutamate metabolism, biosynthesis of amino acids, citrate cycle (TCA cycle), glycerophospholipid metabolism, glyoxylate and dicarboxylate metabolism, phenylalanine metabolism, purine metabolism, steroid biosynthesis, terpenoid backbone biosynthesis, and tyrosine metabolism. Cold stress affects the body’s ability to adjust its metabolic functions to adapt to abnormal environmental conditions. Changes in body metabolism were also observed in chickens, pigs, fish, and other animals and plants under cold stress conditions (Lu, et al., 2018; Pan et al., 2018; Nie, et al., 2018; Song et al., 2019; Wang, et al., 2018; Zhou, et al., 2018).

In the present study, compared with the NT group, the most affected metabolic pathway among the three cold exposure groups was associated with amino acid metabolism, which indicated greater body amino acid mobilization and requirements. A study on tilapia also showed that cold stress significantly changed the amino acid levels and classes connected with the metabolism (Li et al., 2020). Zhou et al. (2011) reported that the concentration of five free amino acids in the hepatopancreas increased directly with decreasing temperature. Amino acid metabolism is closely related to a variety of proteins in the body, and it is conducive to maintaining the protein integrity and function and enhancing the adaptation of the body to cold environments. An early study on protein–concentration and environment–temperature interactions reported that pigs showed an increased feed intake at environmental temperatures below 16.5°C and a slight reduction in weight gain (Seymour et al., 1964). Henry (1985) reported that exposure to a cold environment increases the energy requirement of pigs for thermoregulation by stimulating the food intake, which also increases the daily amino acid and protein intake and improves the body amino acid and protein balance during cold stress.

In the current study, the 2-h and 6-h cold exposure groups shared 11 similar pathways (except two pathways), which revealed that these pathways are strongly correlated with short-term cold stress and that the pigs under 2-h and 6-h cold conditions had similar metabolic processes. Most of the pathways were connected with isocitrate (2-h exposure; downregulation) and citric acid (6-h exposure; upregulation), which are important for body energy metabolism. The downregulation of isocitrate and the altered pathways probably contributed to energy consumption under sudden and 2-h cold stimulation. With the extension of cold adaptation and the upregulation of citric acid, the body adapts to meet the needs of energy metabolism to some extent. The altered metabolic pathways of the 12-h cold exposure group were almost different from those of the 2-h and 6-h groups, and most of the pathways were related to epinephrine (tyrosine metabolism, adrenergic signaling in cardiomyocytes, regulation of lipolysis in adipocytes, renin secretion, cAMP signaling pathway, and neuroactive ligand–receptor interaction). Epinephrine is a hormone secreted mainly by the adrenal medulla. Its main function is to increase the cardiac output and blood glucose levels. Studies have shown that cold stress promotes its secretion to adjust the body metabolism (Nachankar et al., 2005). Epinephrine is also related to increased heat production. The upregulation of epinephrine during 12 h of cold exposure altered the relevant pathway and contributed to the adjustment of energy metabolism. Moreover, high-quality and adequate diets are beneficial for improving body energy and protein metabolism and minimizing the risk of diseases under cold exposure, especially for long-term cold exposure (da Silva et al., 2019).

Cold stress can change the body metabolism and impair the immune system (LaVoy et al., 2011). Under cold conditions, more effective heat conservation mechanisms can be developed in response to cold stimuli by adaptive training compared to those under acute or initial cold exposure (Castellani and Young, 2016).