The C57BL/6 mice (8 weeks, male) were used in this study. Sixteen mice were divided into two groups: In the control group (n = 8), the mice were subjected to gavage with room-temperature water (20–25°C). In the cold stress group (n = 8), the mice were subjected to gavage with cold water (0–4°C).

Regular drinking water and food were available to the mice at all times. Additionally, each mouse was given a total of 1 ml cold or room-temperature water. Prior to use the water was autoclaved and handled using sterile vessels. The maximum dose given by gavage was 200 μl at a time, and we gave it five times. This treatment had lasted for 2 weeks. Gut function tests and behavioral tests were carried out after gavage for 2 weeks. The protocol was approved by the Animal Research Ethics Board of the Northwest University (Approve number: 20210204).

All the mice were transferred into separate clean cages, and the feces were numbered and collected for 4 h between 8 am, and 12 am. All the feces during the procedure were collected and placed in tubes, then tubes were weighed and the wet weight were recorded, dried overnight at 60°C, then reweighed, and the dry weight were recorded. At last the water content was calculated. The criteria for Bristol score were evaluated as described in the previous study (Koh et al., 2010).

The mice were killed by cervical dislocation. Then, the mice were dissected and the whole intestines were removed from the abdomen. Several segments were taken from the ileum and colon, respectively, and rinsed with cold saline, fixed in 4% polyformaldehyde. The ileum and colon samples were prepared for histological examination of lesions. The resected segments were opened lengthwise, embedded in paraffin, and sectioned. Hematoxylin–eosin staining of the ileum and colon was performed. Intestinal inflammation was assessed by observing ulcer formation, epithelial cell changes (goblet cells), lymphocyte infiltration, and lymph node formation, according to a previous report (Zielinska et al., 2017). DAI score is evaluated according the stool consistency, blood and weight loss as previously described (Kim et al., 2012).

In brief, 0.6 g of carmine was dissolved in 0.5% hydroxymethyl cellulose solution to obtain a 6% carmine solution. A 200 μl aliquot of the 6% carmine solution was administered to the mice via gavage. The animals were then placed in separate cages on a white sheet of paper (to facilitate recognition of red feces). Gut peristalsis activity was defined as the time between gavage and the first red bolus excreted (Zielinska et al., 2017).

In this study, we randomly selected 6 mice from the control group and 6 mice from the cold stress group, to perform the metabolic tests. Twelve mice were transferred into metabolic cages (Columbus Comprehensive Lab Animal Monitoring System, Columbus, OH, USA) and housed for 24 h with food and water provided ad libitum. After 24 h of adaptation, the oxygen volume (VO2), carbon dioxide volume (VCO2), respiratory exchange ratio (RER; VCO2/VO2), heat production, food intake, and drink intake were automatically measured during the following 24 h by the system.

The immunofluorescence was performed on paraffin section from the colon tissue as described (Dinallo et al., 2019). The slides were incubated overnight at 4°C with the primary antibodies E-cadherin (1:1000; BD Biosciences) and ZO-2 (1:1000; Cell Signaling Technology), after washing three times with PBS, slides were incubated for 1 h at room temperature with specific secondary antibodies coupled with Alexa Fluor Dyes (1:300; Servicebio). Coverslips were mounted on glass slides using ProLong Gold antifade reagent with DAPI to counterstain the DNA. Lastly, the images were acquired on fluorescence microscope.

The tail suspension test (TST) was used to evaluate depression- like behaviors. As described previously (Steru et al., 1985), the mice were admitted to suspend for 6 min. All sessions were video-recorded. The time spent struggling in TST (The mice have obvious struggling movements) within a 6-min session was recorded and evaluated as behaviors for survival. The behaviors with reduced time for struggling were regarded as depression-like behaviors. Additionally, anxiety-like behaviors were examined using Open Field Test (OFT) as described previously (Walsh and Cummins, 1976). Their locomotor activity were monitored for 15 min. The time spent in the central area were recorded as indicators of exploratory behavior. The behaviors with reduced time in the center were regarded as anxiety-like behaviors.

Feces samples were collected between 8 am and 11 am. The feces samples collected were stored at −80°C. The feces DNA extraction, polymerase chain reaction, library construction, and sequencing were performed as previously described (Sun et al., 2019a). Diversity indices were calculated using QIIME2. Principal coordinate analysis was performed using Calypso online tools. The relative abundances, Spearman correlation coefficients, and heatmap were calculated and compared using T packages. Spearman’s rank correlations at the genus level were calculated as our previous study (Sun et al., 2019b).

The liquid chromatography-mass spectrometry (LC–MS) was employed to analyze the fecal metabolome. We used the ACQUITY UPLC system (Waters Corporation, Milford, MA, United States) coupled with an AB SCIEX Triple TOF 6600 System (AB SCIEX, Framingham, MA, United States) in positive- and negative-ion modes as described previously (Su et al., 2020; Yan et al., 2020).

The TRIzol reagent (Invitrogen, Carlsbad, CA, United States) was used to extract the RNA, then assess the RNA integrity by Bioanalyzer 2,100 (Agilent, CA, United States). Poly(A) RNA was purified and fragmented into small pieces, and reverse-transcribed to synthesize cDNA, treated with U-labeled double-stranded DNAs with the heat-labile enzyme UDG, then amplified. Finally, the products were sequenced on an Illumina NovaSeq 6,000 platform (LC-Biotechnology Co., Ltd., Hangzhou, China).

We used the Fastp software for quality control, used HISAT2 to map the reads, and used StringTie to assemble them. The different mRNAs were selected as the following criteria: mRNAs with fold change of >2 or < 0.5 and p < 0.05 in a parametric F-test comparing nested linear models.

All data in gut function and behavioral tests were showed as the mean ± SEM. Statistical analysis was performed by using Student’s t-test for comparisons between cold stress group and control group. Statistical analysis for bioinformatics data of RNA-seq, gut microbiota, and metabolites was performed using the R package vegan. The p value less than 0.05 was regarded as statistical significance.

A total of 15 mice (8 in control group and 7 in cold stress group) were able to complete the experiment, as one mouse died from choking during gavage. There were no significant differences in heat production, food intake, drink intake, or respiratory exchange rate between the two groups (Supplementary Figures S1A–D). No changes were observed in the jejunum (Supplementary Figure S1E) or colon length (Supplementary Figure S1F) in the cold stress group. Assessment of gastrointestinal motility by carmine red administration showed that gastrointestinal propulsion was impaired in the cold stress group (Supplementary Figure S1G). In the cold stress group, lower fecal output (t = 2.493, p = 0.027; Figure 1A), lower fecal moisture percentage (t = 4.815, p < 0.001; Figure 1B), and lower Bristol scores (t = 3.015, p = 0.010; Figure 1C) were observed. These results suggest that cold stress disturbs the intestinal function. Altered fecal properties, bloody stools, and weight loss are the most important characteristics in IBD, then we used the DAI scores to evaluate the severity of the colitis. We found that the DAI scores in the cold stress group were significantly higher than in the control group (t = −4.861, p < 0.001; Figure 1D). Correspondingly, hematoxylin–eosin staining showed sparse intestinal villus, and edema in the intestinal villi of the jejunum (Figure 1E), a significant reduction in goblet cells, an enhanced inflammation response, and an elevated histological score in the colon (t = −2.193, p = 0.047; Figure 1F). As to the gut barrier, we found that the expression of E-cadherin and ZO-2 was lowered by immunofluorescence (Figure 1G), compared to that in the control group.

To further identify the changes in gene expression in the intestinal tissue after cold stress, we randomly selected out 6 mice from the two groups and investigated the gene expression profiles of the intestinal tissue using RNA sequencing (Figure 2A). A total of 432 genes with differential expression between groups were identified, with 330 genes upregulated and 102 genes downregulated in the cold stress group when compared with the control group (Figure 2B). Gene Ontology analysis showed that all the genes most strongly enriched in immune function processes, including innate immune response and adaptive immune response, such as the B cell activation and B cell receptor complex, immunoglobulin production, immunoglobulin complex, complement activation, antigen binding, and defense response to a bacterium (Figure 2C). In accordance with these findings, KEGG pathway analysis also showed that the genes were enriched in immune activation in the colon, including the primary immunodeficiency, B cell receptor signaling, intestinal immune network for IgA production, NK cell-mediated cytotoxicity pathways, leukocyte transendothelial migration, Th1 and Th2 cell differentiation, Th17 cell differentiation, cytokine–cytokine receptor interaction, and chemokine and Fc gamma R-mediated phagocytosis, all of which are closely related to the immune response in the colon (Figure 2D).

To further reveal the effects of cold stress on the colon, we used gene set enrichment analysis to analyze the genes which revealed substantial upregulation of genes involved in the B and T cell receptor signaling pathway, leukocyte transendothelial migration, FC epsilon RI signaling pathway, chemokine signaling pathway, and cytokine–cytokine receptor interaction (Figure 2E). These findings suggested that cold stress triggers an IBD-like phenotype in mice.

In general, IBD is considered to occur when the immune system overreacts to the resident gut microbiota, inducing a chain of inflammatory events that can destroy the gut barrier (Eisenstein, 2016). These findings prompted us to further investigate whether changes in microbiota or bacterial metabolites from the feces under cold stress stimulation may regulate the gut barrier and gut–brain interactions.

Eight individual fecal samples from the control group and seven individual fecal samples from the cold stress group were collected and sequenced. The principal component plots with unweighted UniFrac distances showed a clear separation between the cold stress and control groups (Figure 3A), which suggested that cold stress led to a significant alteration in the gut microbiota composition. The Shannon index showed no difference between the two groups (Figure 3B). However, at the operational taxonomic unit (OTU) level, the microbiota were significantly differed (Figure 3C). The abundances of OTU198 (Lachnospiraceae_unclassified), OTU337 (Clostridiales_Incertae_unclassified), OTU156 (Muribaculaceae_unclassified), OTU254 (Erysipelotrichaceae_unclassified), OTU88 (Duncaniella), OTU334 (Muribaculaceae_unclassified), OTU13 (Paramuribaculum), OTU433 (Proteobacteria_unclassified), OTU215 (Lachnospiraceae_unclassified), OTU258 (Firmicutes_unclassified), OTU115 (Paramuribaculum), OTU272 (Mailhella), and OTU34 (Akkermansia) were significantly increased in the cold stress group. Additionally, the abundances of OTU246 (Anaerotruncus), OTU15 (Alistipes), and OTU209 (Clostridiales_unclassified) decreased under cold stress (Figure 3D). The detailed information has been shown in Supplementary Table S1. Linear discriminant analysis of effect size further showed that the bacteria with increased abundance in the cold stress group mainly belong to Proteobacteria (Supplementary Figure S2). To determine the pattern of bacteria, we constructed their networks in the two group respectively, we found that the network of the cold stress-treated mice had a simpler property (nodes/edges = 81/176) than the control group (nodes/edges = 80/234), indicating that cold stress may induce vulnerability to environmental stress in the gut microbiota (Figure 3E).

As a microbe–host bridge, some metabolites of the intestinal flora can affect host physiology by entering the bloodstream. Therefore, we analyzed fecal metabolites using LC–MS. We found that the metabolic data clusters of the control and cold stress groups were separated from each other in both positive- and negative-ion modes by partial least-squares discriminant analysis (Figures 4A,B). The heatmap also showed that cold stress led to significant alterations in fecal metabolite levels (Figure 4C); 1,179 metabolites were upregulated, and 1896 metabolites were downregulated with significant changes (Figure 4D). The most strongly impacted metabolic pathways included cocaine addiction, dopaminergic synapse, amphetamine addiction, and alcoholism addiction (Figure 4E), all of which were accompanied by a significant reduction in levels of dopamine, l-dopamine, l-tyrosine, and homovanillic acid (Figure 4F). As one of the most important neurotransmitters, decreased dopamine levels may contribute to anxiety-like and depression-like behaviors in the cold stress mice. Additionally, patients with IBD present some similar symptoms to patients with IBS, including anxiety and depression. Therefore, these behaviors were further evaluated.

In the tail suspended test, the struggling time was significantly decreased in the cold stress group (t = 2.347, p = 0.035; Figure 5A), suggesting an increase in depression-like behaviors in mice exposed to cold stress. Furthermore, in the open field test, the center time (t = 2.451, p = 0.029) was reduced in the cold stress group (Figure 5B), implying a tendency of decreased exploratory behavior and increased anxiety-like behavior in the cold stress group.

Finally, to explore the functional significance of the metabolite perturbations in the gut microbiota of the cold stress-treated group, the 97 annotated metabolites with significant differences were selected, and their Spearman correlation coefficients with different bacteria were calculated. Significant correlations were observed between the gut microbiota and metabolites (Supplementary Figure S3A), and also observed between metabolites and gut function (Supplementary Figure S3B).