The subjects were 12 healthy males, aged 19–26 years, 163–186 cm tall and 54–95 Kg in weight. The subjects did not have cardiovascular system diseases, hemorrhoids, infectious diseases and skin diseases, etc., did not take alcohol, antibiotics and other drugs 7 days before the experiments, and maintained a good sleep. Before participating in the experiment, the subjects have understood the process and risks of this experiment, and signed the informed consent form. All experiments were approved and performed following the guidelines of the Ethical Committee of Naval Medical University (Approval No. AF-HEC-018).

The experiment was carried out in submarine environment simulation cabin (SESC). The living conditions are normal pressure and room temperature in the confined environment, and the effective volume is about 200 m³. The errors of temperature and relative humidity in the room are less than ±0.5°C and ± 5 %RH. During the experiment, the subjects were forbidden to smoke and drink, eat normally every day. The subjects could use computers, fitness equipment, etc., but there was no external network, and the submarine environment simulation cabin was in a state of information isolation. The entire experiment lasted 14 days and nights. The flow chart of experimental was shown in Figure 1A.

Faecal samples of subjects were collected at 18:00–22:00 on day 1 and day 14. The faecal samples of one subject were added into sterilized 5-mL Eppendorf tubes on ice and stored at −80°C. Urine samples of subjects were collected at 7:30 am on day 1 and day 14. The Urine samples of one subject were added into 20-mL sampling tube on ice and stored at −80°C.

Faecal samples of subjects were prepare for microbiota analysis. Faecal DNA was extracted using an E.Z.N.A. stool DNA Kit (Omega Bio-tek, USA) according to the manufacturer’s instructions. The V3-V4 regions of the bacterial 16S rRNA genes were amplified using universal primers 338F 5′-ACTCCTACGGGAGGCAGCA-3 and 806R 5′-GGACTACHVGGGTWTCTAAT-3′. PCR amplification of the 16S rRNA gene was carried out in triplicate as follow: template DNA 10 ng, 2.5 mM d NTPs 2 μl, forward and reverse primer (5 μM) 0.8 μl respectively, TransStart fast Pfu polymerase 0.4 μl, 5 × Fast Pfu Buffer 4 μl, and ddH₂O in a final volume of 20 μl. PCR amplification program was as follows: an initial activation step with 95°C for 3 min, followed by 27 cycles at 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s and a final extension at 72°C for 10 min (Dennis et al., 2013). The quality of PCR products was quantified using QuantiFluor™-ST system (Promega, USA) according to the standard protocols. Then, purified PCR products were sequenced on an Illumina MiSeq platform (Illumina, USA) at Majorbio Bio-Pharm Technology Co., Ltd., Shanghai, China.

Microbiota analysis of subjects were performed by Majorbio Cloud¹ provided by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). All raw reads were demultiplexed and quality-filtered using QIIME (version 1.9.1) with the following criteria: (1) The 300 bp reads were truncated at any site receiving an average quality score < 20 over a 10 bp sliding window abandoned the truncated reads that were < 50 bp; (2) Sequences that reads containing ambiguous characters, or two nucleotide mismatches in primer matching were removed. An operational taxonomic units (OTUs) were set as at least 97% identified sequences. The invalid sequences were cut-off using UPARSE (version 11²). RDP Classifier (version 2.13³) was used to analyses the taxonomy of each 16S rRNA gene sequence, and against the SILVA (version 138) 16S rRNA database using a confidence threshold of 70% (Quast et al., 2013).

The RBC counting in urine was determined by automated urine analysis workstation IQ200 (Beckman Coulter, Brea, California, America). The UA content in urine was determined by automatic biochemical analyser Cobas C702 (Roche, Basel, Swiss).

The subjects of health conditions were investigated with a questionnaire described by Li et al. (2016). The questionnaire was carried out before and after the experiment, including sleep quality and time, constipation, anxiety, memory loss, dizziness, anorexia, and facial acne. In addition to sleep time, each other health problem is scored as 5 points, and the total score of each item after accumulation was used to represent the health status of the subject.

The α-diversity analysis of different groups was detected by Welch’s t test. Differential relative abundance of taxa was detected by Kruskal-Wallis-H test with corrected p value, and the multiple test correction was fdr. Linear regression of gut microbiota based on Bray_Curtis distance was performed at the genus level to assess the impact of health indicators on gut microbiota. Enterotypes analysis of gut microbiota based on JDS distance was performed at the genus level. LEfSe analysis of gut microbiota was performed by using multi group comparison strategy: All gain all (more strict). The correlations between the relative abundance of species and health indicators were calculated by Spearman’s rank correlation coefficient and visualized by heatmap. Statistical analysis of physical and chemical indexes of urine was performed with ANOVA (SPSS 24.0). A p-value of <0.05, 0.01 or 0.001 was considered statistically significant.

16 s rDNA high-throughput sequencing was used to detect the gut microbiota composition of the subjects during the CE experiment. We obtained 1,177,407 sequences, which were clustered into 641 operational taxonomic units comprising 12 phyla, 89 families, and 233 genera. Based on the abundance of flora at the genus level, the Jensen-Shannon distance was calculated, and clustering was performed by partitioning around medodids to obtain the best clustering K value. Principal coordinate analysis was used to determine the enterotype of the gut microbiota (Wu et al., 2011). As shown in Figure 1B, the gut microbiota enterotype analysis showed that the 12 subjects could be clustered into two enterotypes. Type 1 was the Blautia type (Bla, comprising subjects 4, 7, 10, and 12), and type 2 was the Bifidobacterium type (Bi, comprising subjects 1, 2, 3, 5, 6, 8, 9, and 11). The subjects of both groups showed a high degree of overlap in terms of gut microbiota type, but the flora characteristics were significantly different. Therefore, subsequent gut microbiota analysis was carried out based on the enterotype. α-diversity analysis of the enterotype groups showed that the α-diversity of the Bi group significantly increased in the CE (p < 0.05), whereas that of the Bla group did not significantly change (Figures 1C,D). Horizontal comparison showed that there were significant differences in the α-diversity of the gut microbiota of the two enterotype groups during the experimental process, which was consistent with the results of the subjects’ gut microbiota typing (Figures 1E,F).

As shown in Figure 2, the effect of the CE on the gut microbiota of different enterotypes was analyzed at the phylum and genus levels. The gut microbiota of subjects in the Bi enterotype group were dominated by Firmicutes, Actinobacteriota, and Proteobacteria; these three phyla accounted for >99% of the total abundance, with Firmicutes and Actinobacteriota amounting to 54 and 37%, respectively. In the Bla group, Firmicutes, Actinobacteriota, and Bacteroidota were the main phyla and accounted for >99% of the total population; Firmicutes and Bacteroidota amounted to 41 and 25%, respectively (Figure 2A). At the genus level, Blautia, Bifidobacterium, Streptococcus, and Subdoligranulum were the main components in the Bla group. However, compared with the Bla group, Bifidobacterium was more abundant in the Bi group (Figure 2B). The composition of the gut microbiota was further analyzed using a bubble abundance map (Figure 2C), which showed the top 30 most abundant genera; the sizes of the circles represented differences in the relative abundances of the gut microbiota. Most of the gut microbiota belonged to the phylum Firmicutes, with the compositions of the Bla and Bi groups being significantly different. The predominant genera in the two groups were different, with Bifidobacterium, Collinsella, and other bacteria belonging to the Actinobacteriota phyla being higher in the Bi group than in the Bla group.

Figure 2D shows a Venn analysis of gut microbiota in the CE. We found 113 core microbiota species in common across the four groups, with the number of microbiotas in the Bi group being less than that in the Bla group at the genus level. After 14 days of the CE experiment, the number of microbiotas in the Bi group increased; however, there was no significant change in the Bla group. Source Tracker analysis showed that 31.12% of the microbiota in the Bi_T2 group came from the Bi_T1 group. However, the origins of the majority of the microbiota (62.14%) were unclear, which may reflect the effect of the CE (Figure 2E). Only 11.8% of the microbiota in the Bi_T2 group was derived from Bi_T1, and the origins of most of the microbiota (83.04%) remained unclear (Figure 2F).

There were significant differences in the gut microbiota of the two enterotypes in the CE (Figure 3). Differences in the dominant phyla and genera (Figures 3A,B) of the gut microbiota were analyzed. At the phylum level, there were significant changes in Actinobacteriota and Proteobacteria across the four groups (p < 0.05). Analysis at the genus level showed that in the Bi group, the relative abundances of Bifidobacterium, Ruminococcus_gnavus_group, and Klebsiella were significantly higher than in the Bla group (p < 0.05, p < 0.01). However, the abundances of Dorea, Ruminococcus_torques_group, Fusicatenibacter, Coprococcus, CAG-352, UCG-002, and other genera were significantly lower in the Bi group than in the Bla group (p < 0.05, p < 0.01). We performed non-metric multidimensional scaling (NMDS) analysis of the gut microbiota of the subjects and verified the differences in the compositions of the Bi and Bla groups in the CE. After living in the CE for 14 days, the composition of the gut microbiota of the Bi group underwent significant changes, whereas that of the Bla group experienced only small changes (Figure 3C). As shown in Figure 3D, linear discriminant analysis (LEfsE; linear discriminant analysis threshold = 2) showed that Bifidobacterium, Enterococcus, and other genera were significantly enriched in the Bi_T1 group, and that Klebsiella, Ruminococcus_gnavus_group, and other genera were significantly enriched in the Bi_T2 group. Dorea, Fusicatenibacter, UCG-002, and other genera were significantly enriched in the Bla_T1 group, and Ruminococcus and Ruminococcus_torques_group were significantly enriched in the Bla_T2 group.

The relationships between the gut microbiota of subjects in a CE were investigated through correlation network analysis. Genus-level correlation network analysis revealed significant interactions between the different genera of gut microbiota (a green connection line represents a positive correlation, and red represents a negative correlation; the thickness of the line represents the magnitude of the correlation coefficient; Figure 4). We documented close interactions between CAG-352, Coprococcus, Ruminococcus, and other genera, with the relationship becoming stronger after the CE experiments.

The urine physicochemistry and health indicators of the subjects during the CE experiment were evaluated. As shown in Figure 5, there were significant differences in the urine physicochemical (UA content, red blood cell (RBC) content, sleep time (S), and other health indicators (H)) between the Bi and Bla groups after living in a CE. There was no difference in other physicochemistry indexes in the urine of the subjects during the experiment (Supplementary Table 1). In the Bi group, the UA content decreased significantly after 14 days in the CE (p < 0.05; Figure 5A). There was no significant difference in the RBC content and S between the Bi_T1 and Bi_T2 groups (Figures 5B,C). However, compared with the Bla_T1 group, subjects in Bla_T2 group had a significantly higher UA and RBC content (Figures 5D,E). The UA content increased from 1800 ± 200 mg/l to 3,500 ± 200 mg/l; the UA content of subject no. 1 increased most drastically, from 1899.4 mg/l to 5330.00 mg/l. The S of the subjects in the Bla group significantly decreased from 8 to 6 h (Figure 5F); these subjects also generally found it difficult to fall asleep. Subjects were assessed for other health indicators as well, namely anorexia, vertigo, acne, memory, anxiety, and constipation. The results are shown in Figure 5G. The subjects in the Bi group were generally in good health, with only a few showing abnormal health conditions. In the Bla group, except for subject no. 5, all had 3–4 abnormal health conditions. Constipation (8 subjects) and anorexia (7 subjects) were common in these subjects, whereas acne was documented less frequently (only 2 subjects). All of the subjects underwent a detailed health investigation before the experiment, which revealed none of the above-mentioned health problems. This indicated that the Bi group subjects adapted better to the CE, whereas the Bla group subjects were easily affected by the CE.

Figure 6 shows linear regression analysis of the gut microbiota and UA content, RBC content, S, and other health conditions of the subjects in different groups. There were no significant correlations between the β-diversity of the gut microbiota and health indicators at the beginning of the experiment (day 1, T1 group) (Figures 6A–C). However, the UA content, RBC content, and health indicators were significantly negatively correlated with β-diversity after the experiment (day 14, T2 group, p < 0.001) (Figures 6D–F). Although S was positively correlated with β-diversity, this was not significant (Figure 6G, p = 0.0708). The RDA analysis was consistent with the ordinal regression results. The UA content, RBC content, and health indicators were significantly correlated with the composition of gut microbiota (p < 0.05), and UA content was positively correlated with RBC content and other health indicators. However, there was no significant correlation between S and gut microbiota (p = 0.078) (Figure 6H). The UA and RBC content in the Bla group may be correlated with the changes in Ruminococcus_torques_group, Collinsella, and other bacteria.

The Spearman correlation coefficient was used to analyze the correlations between gut microbiota and indexes such as the UA content, RBC content, and health indicators (Figure 7; Boix-Amorós et al., 2016). During the CE experiment, the correlation between the composition of the gut microbiota and health indicators of the subjects changed significantly. The relative abundances of most of the genera were not significantly correlated with health indicators at the beginning of the experiment; only those of Ruminococcus_gnavus_group and Subdoligranulum were negatively correlated with the UA content (Figure 7A). As shown in Figure 7B, after living for 14 days in the CE, the relative abundances of UCG-002, Ruminococcus, Fusicatenibacter, Collinsella, CAG352, and other genera were significantly positively correlated with the UA content, whereas those of Klebsiella, Ruminococcus_gnavus_group, and Streptococcus were significantly negatively correlated with the UA content. The relative abundance of Klebsiella was significantly positively correlated with S, whereas the abundances of CAG-352, UCG-002, Ruminococcus, Alistipes, and other genera were significantly negatively correlated. The relative abundances of Bifidobacterium, Klebsiella, and other genera were significantly negatively correlated with the RBC content in urine, whereas those of Ruminococcus_torques_group and norank_f_lachnospiraceae showed significantly positive correlations. The relative abundances of Ruminococcus, CAG-352, Coprococcus, and other genera were significantly positively correlated with health indicators, whereas those of Klebsiella and Ruminococcus_gnavus_group were significantly negatively correlated. The Spearman correlation analysis showed that the changes in CAG-352, Ruminococcus, UCG-002, Klebsiella, Ruminococcus_torques_group, and other genera were consistently correlated with the health indicators of subjects in different groups. The relative abundances of these genera significantly impacted the structure of the gut microbiota.

Multivariate association with linear models (MaAsLin) analysis is a method that investigates the correlations between environmental factors (such as clinical data, physicochemical indicators) and the relative abundance of microbial community species using linear models (Torres et al., 2020). MaAsLin analysis verified that the relative abundances of UCG-002, CAG-352, Klebsiella, Bifidobacterium, Coprococcus, and other genera were significantly correlated with the health conditions of the subjects (Figures 7C–J). The CE experiment showed that, compared with the Bla group, the Bi group was more adaptable to such environments. This difference may be related to the higher relative abundances of Bifidobacterium, Klebsiella, and other genera and lower relative abundances of UCG-002 and CAG-352 in the Bi group. The high relative abundances of Ruminococcus_torques_group, Ruminococcus, UGG-002, Coprococcus, and other genera in the gut microbiota may make the subjects in the Bla group vulnerable to CEs, thereby increasing their UA content, insomnia, anxiety, and other health abnormalities.