Recovery Dynamics of Intestinal Bacterial Communities of CCl4-Treated Mice with or without Mesenchymal Stem Cell Transplantation over Different Time Points

Liver injury has caused significant illness in humans worldwide. The dynamics of intestinal bacterial communities associated with natural recovery and therapy for CCl4-treated liver injury remain poorly understood. This study was designed to determine the recovery dynamics of intestinal bacterial communities in CCl4-treated mice with or without mesenchymal stem cell transplantation (i.e., MSC and CCl4 groups) at 48 h, 1 week (w), and 2 w. MSCs significantly improved the histopathology, survival rate, and intestinal structural integrity in the treated mice. The gut bacterial communities were determined with significant changes in both the MSC and CCl4 groups over time, with the greatest difference between the MSC and CCl4 groups at 48 h. The liver injury dysbiosis ratio experienced a decrease in the MSC groups and a rise in the CCl4 groups over time, suggesting the mice in the MSC group at 48 h and the CCl4 group at two weeks were at the least gut microbial dysbiosis status among the corresponding cohorts. Multiple OTUs and functional categories were associated with each of the bacterial communities in the MSC and CCl4 groups over time. Among these gut phylotypes, OTU1352_S24-7 was determined as the vital member in MSC-treated mice at 48 h, while OTU453_S24-7, OTU1213_Ruminococcaceae, and OTU841_Ruminococcus were determined as the vital phylotypes in CCl4-treated mice at two weeks. The relevant findings could assist the diagnosis of the microbial dysbiosis status of intestinal bacterial communities in the CCl4-treated cohorts with or without MSC transplantation.


Introduction
Liver injury is a severe liver condition and has caused significant illness in human worldwide [1,2]. This condition has been associated with the changes of intestinal microbiota [3,4]. The intestinal microbiota is involved in the maintenance of the intestinal barrier by several mechanisms such as preventing colonization by pathogenic bacteria and by cooperating with the intestinal epithelium to produce mucin 2 [5]. Recent findings suggest that the disruption of the intestinal barrier is a prerequisite for liver injury [6]. Some probiotics were effective for the prevention of this condition [7,8]; however, the effective therapies and the corresponding mechanisms remain poorly understood.
The application of mesenchymal stem cells (MSCs) on liver injury repair has attracted increasing attention in recent years. The transplantation of MSCs has been found to alleviate injuries in multiple organs [8][9][10][11][12][13][14], including liver injury [15,16]. Among the MSCs, bone marrow MSCs (BM-MSCs) could improve the clinical indices of liver function in the patients with liver injury caused by hepatitis B [17] and attenuate hepatic ischemia-reperfusion injury in mice [16,18].
Our previous study has provided important insights on the changes of survival rate, liver biochemistry parameters, histology, and intestinal microbiota between the CCl 4 -treated mice with or without MSC therapy [19]. In the current study, we aimed to (1) determine the recovery dynamics of intestinal bacterial communities of CCl 4 -treated mice with or without MSC transplantation over different time points and (2) investigate the vital phylotypes in the least dysbiotic intestinal bacterial community in CCl 4 -treated mice with or without MSC transplantation over the three time points.

Materials and Methods
2.1. Animal Information. Enhanced green fluorescent protein (GFP) transgenic C57BL/6 mice were purchased from Nanjing Biomedical Research Institute of Nanjing University. Then, female mice were backcrossed to C57BL/6 male mice to produce younger male C57BL/6 mice for isolation and culture of MSCs.
The 6-8-week-old male C67BL/6 mice were purchased from Shanghai SLAC Laboratory Animal Co., Ltd. For the induction of acute liver injury (ALI) with CCl 4 . Animals were allowed access to food and water and housed under specific pathogen-free conditions. All animal experimental procedures were conducted according to a protocol approved by the Ethics Committee of the First Affiliated Hospital of Zhejiang University.

Isolation and Culture of Mouse MSCs.
MSCs were isolated and cultured as previously described [20]. Briefly, two-to three-week-old wild-type C57BL/6 male mice were sacrificed for obtaining the humeri, tibiae, and femurs, the marrow of which was flushed out thoroughly with 3 ml αminimal essential medium until the bones became pale. The compact bones were chopped into pieces and transferred to collagenase II digestion solution in 15 ml tubes, before being incubated at 37°C for 1.5 h with continuous rotation. The enzyme-treated bone chips were suspended in 7.5 ml C57BL/6-MSC special complete MEM and incubated at 37°C in a 5% CO 2 incubator. Afterwards, nonadherent cells were removed and the complete MEM was replaced, before harvesting the adherent MSCs using 0.25% trypsin-EDTA and resuspending in fresh complete MEM. Purified MSCs were characterized by inducing osteogenic and adipogenic differentiation and analyzing surface marker expression by flow cytometry (Supplementary Figure S1) [19].
2.3. Inducing Acute Liver Injury in Mice with CCl 4 . Sixty 6-8week-old male C67BL/6J mice were intraperitoneally administered with 3 ml/kg CCl 4 dissolved in olive oil (v/v, 50%) to induce ALI, while six mice in the negative control (NC) group received olive oil. Six hours after the CCl 4 administration, mice with ALI were randomly divided into the MSC group (n = 30) and the CCl 4 group (n = 30). A 0.1 ml aliquot of PBS containing 5 × 10 5 MSCs was injected into the tail vein of each mouse in the MSC group, while each mouse in the CCl 4 group received an injection of 0.1 ml PBS. Mice in the NC group were injected with PBS (n = 6). Eighteen mice were randomly selected from the MSC group and anesthetized at 48 h (M48, n = 6), 1 w (M1W, n = 6), and 2 w (M2W, n = 6). Likewise, 18 mice were randomly selected from the CCl 4 group and anesthetized at 48 h (C48, n = 6), 1 w (C1W, n = 6), and 2 w (C2W, n = 6). The NC group was anesthetized at 48 h.

Tissue Collection and
Histopathology. The anesthetized mice were sacrificed for collecting the liver, small intestinal segments, and cecum and colon contents. The liver and small intestinal segments were processed in standard histological methods, while the ileum samples were processed and observed under transmission electron microscopy by a previous study [19].
2.5. Molecular Experiments for Illumina Sequencing. DNA was extracted from the cecum and colon contents by using QIAamp DNA stoolMini Kit (Qiagen Inc., USA), before being amplified by fusion dual barcoded primers 319F/806R targeting the V3-V4 regions of bacterial 16SrRNA gene by Dong et al. [19]. The PCR products were purified, quality checked, and subjected for sequencing on an Illumina Miseq sequencer (Illumina Inc. USA) using 2 × 300 bp chemistry.

Intestinal Flora Analyses and Statistical Analyses
2.6.1. Processing of Sequencing Data. Quality filtering, dereplication, chimera filtering, and taxonomy assignment procedures were performed in QIIME software version 1.9.0 as described by Dong et al. [19]. Operational taxonomic units (OTUs) were clustered based on sequence identity threshold ≥ 97%. Similarity percentage (SIMPER) analysis was used to compare the dissimilarities of gut bacterial communities between the MSC group and CCl 4 group at three time points after overall transformation of the dataset in square-root.
Partition Around Medoid (PAM) clustering analysis was performed to cluster all the gut bacterial communities in the CCl 4 and MSC groups at three time points, after determining the optimal numbers of clusters by using an average silhouette method [21].
2.6.3. Dysbiosis Ratio in the Intestinal Bacterial Communities in the MSC or CCl 4 Groups. Microbial dysbiosis ratio has been investigated in multiple disease studies to evaluate the microbial dysbiosis status of bacterial community [22][23][24][25]. In the present study, the dysbiosis ratio in the gut bacterial communities, i.e., liver injury dysbiosis ratio (LIDR), was defined as the abundance ratio of OTUs associated with the CCl 4 group and OTUs associated with the NC group at 48 h. A LEfSe analysis was performed to determine the OTUs associated with the CCl 4 group or NC group. LIDRs of C48 and M48 were compared with a t-test, after appropriate data transformation.

Changes of Gut Bacterial Communities of the MSC
Groups or CCl 4 Groups over Time. The LIDRs in M48, M1W, and M2W were compared with one-way ANOVA after appropriate data transformation. t-tests were performed for the pairwise comparisons, with Bonferroni's correction for adjusting the P values. The LIDRs in C48, C1W, and C2W were compared with the same approaches.
PERMANOVA was used to compare the intestinal bacterial composition between M48, M1W, and M2W, as well as those between M48 and M1W and between M1W and M2W. The same technique was applied for the comparisons of CCl 4 groups over time.
One-way analysis of variance (ANOVA) was used to compare the alpha diversity, i.e., richness (observed species), diversity (Shannon index), and evenness (Pielou index), in the MSC groups at 48 h, 1 w, and 2 w. t-tests were used for the pairwise comparisons, with Bonferroni's correction for the correcting the P values. The same approaches were applied for comparisons of alpha diversity of gut bacterial communities of CCl 4 groups at three time points.
LEfSe analysis was applied to determine the OTUs associated with each of the gut bacterial communities of MSC groups at three time points. The same analysis was carried out for determining the OTUs associated with each of the gut bacterial communities in CCl 4 groups at three time points.

Changes of Bacterial Networks and Gatekeepers in the MSC or CCl 4 Groups over Time.
Co-occurrence network (CoNet) analysis was carried out to investigate the cooccurrence and coexclusion of OTUs in the MSC groups at 48 h, 1 w, and 2 w. The top 10 OTUs with most correlations at each time point in MSC group were determined. The detailed procedures were performed as described by Wagner Mackenzie et al. [26]. Briefly, Spearman, Pearson, Bray Curtis, Mutual Information, and Kullback-Leibler dissimilarities were chosen to calculate the ensemble inference, with the top 1000 positive and negative correlations recorded. The method-specific P values were computed by permutation procedure, followed by a bootstrap step to merge the P values into one final P value. The same technique was performed for CCl 4 groups over time.
Gatekeepers were regarded as the OTUs interacting with different parts of the bacterial network that holds together the bacterial community [27]. In the current study, fragmentation was carried out to determine the gatekeeper(s) of the bacterial networks in each of the MSC groups and CCl 4 groups at different time points. The detailed manipulations were performed as described by Wagner Mackenzie et al. [26]. A total of 10, 000 randomly constructed networks with identical node and edge distributions to the original network was used to create a null distribution of fragmentation scores. Statistical significance was defined as the number of times a fragmentation score greater than that resulting from the removal of the phylotype observed within the null distribution.
2.6.6. Changes of Functional Categories in Gut Microbiota of the MSC or CCl 4 Groups at Three Time Points. Functional profiles of bacterial communities of MSC groups at three time points were predicted by Tax4fun based in R software [28]. LEfSe analysis was used to determine the functional categories associated with each of the MSC groups at three time points. The same approaches were applied to determine the functional categories associated with each of the CCl 4 groups at three time points.

Protective Effects of MSCs against CCl 4 -Induced Liver
Injury. MSC transplantation dramatically increased the survival rate of CCl 4 -treated mice from 45.5% to 77.3% as described by Dong et al. [19], though there was no significant difference in body weight between the MSC and CCl 4 groups. The liver and ileum of the MSC transplanted mice have experienced an overall improvement compared with those of the CCl 4 -treated mice (Figures 1(a)-1(c)).
Three and five were determined to be the two optimal numbers for clustering with the highest silhouette scores (Supplementary Figure S2). Three clusters could cluster the bacterial communities of the MSC and CCl 4 groups compared with five clusters, so three was chosen for PAM analysis. PAM clustering analysis showed three clusters, Cluster_1 (including most C48 and one C1W), Cluster_3 (including all M48 and one C48), and Cluster_2 (including the remaining ones) ( Figure 2). All these results suggest that the largest difference in the gut bacterial communities between the CCl 4 and MSC groups was at 48 h, but not at 1 w and 2 w.

Changes of LIDRs in the MSC or CCl 4 Groups over Time.
LIDRs in the intestinal bacterial communities of all groups were calculated after determining the OTUs associated with the CCl 4 or NC groups (Supplementary Figure S3). LIDR was significantly lower in C48 (0:15 ± 0:11 SE) than in M48 (12:9 ± 4:37 SE) at baseline (t-test, P < 0:001).
A total of 73 OTUs were associated with M48, M1W, or M2W according to LEfSe results. Twenty-five out of the 73 OTUs could distinguish M48 from M1W and M2W, over half of which were assigned to Clostridiales and S24-7 (Figure 3). A group of 25 OTUs could differentiate M1W from M48 and M2W, over half of which were from Oscillospira, Lachnospiraceae, and S24-7. The remaining 23 OTUs were more associated with M2W and dominated by OTUs assigned to S24-7.
By contrast, 11 OTUs were associated with C48, among which OTUs assigned to Bacteroides and Prevotella accounted for over half of the phylotypes (Figure 4). Fourteen OTUs from 11 taxa were associated with C1W, with OTU1383_Bacteroides, OTU446_Enterobacteriaceae, and OTU519_Mucispirillum as the three phylotypes with the largest LDA scores (over 4.0). A total of 34 OTUs were associated with C2W, almost twothirds of which were from Clostridiales (12 OTUs), Lachnospiraceae (five OTUs), and Oscillospira (five OTUs) (Figure 4).  (Table 1), with OTUs assigned to S24-7 being determined in the top 10 OTUs in each of the three networks. Likewise, the top 10 OTUs with most correlations in C48, C1W, and C2W were largely distinct (Table 2), and OTUs assigned to S24-7 were also determined in the top 10 OTUs in each of the three networks. Multiple OTUs were determined as gatekeepers in the bacterial networks of M48, M1W, M2W, C48, C1W, or C2W (Table 3). Among them, OTU1352_S24-7, i.e., a gatekeeper in the M48 network, was also associated with M48 by LEfSe analysis (Supplementary Figure S4A). Likewise, OTU453_S24-7, OTU1213_Ruminococcaceae, and OTU841_Ruminococcus, i.e., gatekeepers in C2W, were also associated with C2W determined by LEfSe analysis (Supplementary Figure S4B).  (42). Glycogen phosphorylase, heterodisulfide reductase subunit A, and hexosaminidase were the three functional categories most associated with M48, M1W, and M2W ( Figure 5(a)), respectively. By contrast, outer membrane usher protein, long-chain acyl-CoA synthetase, and methylaccepting chemotaxis protein were the three functional categories most associated with C48, C1W, and C2W, respectively ( Figure 5(b)).

Discussion
The gut microbiota in CCl 4 -treated mice has been reported in different studies [29][30][31]. Our previous study provided sufficient findings about the differences between the intestinal bacterial communities in the CCl 4 -treated mice with and without MSC therapy [19]. This study was designed to determine the recovery dynamics of intestinal bacterial communities of CCl 4 -treated mice with or without mesenchymal stem cell transplantation over different time points, which was rarely reported to our limited knowledge.
PERMANOVA, SIMPER, and PAM clustering analyses have been used in a variety of studies [32][33][34]. In the current study, the differences in bacterial communities between the

BioMed Research International
MSC and CCl 4 groups at designated times (i.e., 48 h, 1 w, and 2 w) were determined by the three analyses, and the relevant results showed the largest difference between the MSC and CCl 4 groups occurred at 48 h. As the most obvious difference of survival percentages of mice between the MSC and CCl 4 groups occurred from 48 h to 1 w, it implied that the change of intestinal bacterial communities was associated with this obvious difference.
PERMANOVA results for MSC groups' comparison suggested the significant change in the gut bacterial communities of MSC groups from 48 h to 1 w, while no such change was found between 1 w and 2 w. By contrast, PERMANOVA results for CCl 4 groups' comparison suggested the significant changes in gut bacterial communities over the two time periods, i.e., 48 h to 1 w and 1 w to 2 w. An overall difference was determined in the richness of the gut bacterial communities of MSC groups over time, while overall differences were determined in diversity and evenness of C48, C1W, and C2W. These results suggest the recovery mechanisms in gut bacterial communities in the MSC and CCl 4 groups were different.
The dysbiosis ratio has been investigated in multiple disease studies [22][23][24][25]. In this study, LIDR was used to evaluate the dysbiosis status of gut bacterial communities in the MSC and CCl 4 groups and was significantly lower in C48 than in M48 at baseline. LIDR was decreasing in MSC groups over time, while this ratio experienced an increase in CCl 4 groups during the same period, which could partly explain why the difference of gut bacterial communities in the MSC and CCl 4 groups were largest at 48 h. These results implied that MSCs-treated mice experienced the mildest intestinal microbial dysbiosis at 48 h, while the CCl 4 -treated mice had the mildest intestinal microbial dysbiosis at two weeks.
LEfSe results showed different phylotypes were associated with each of the MSC and CCl 4 groups. Multiple S24-7 phylotypes were associated with all the MCS and CCl 4 groups, suggesting different S24-7 phylotypes could play different roles in each of the six groups. This is consistent with the previous findings, which hold different views on the roles of Bacteroidales S24-7 in the gut microbiota [35,36]. Multiple OTUs assigned to Clostridiales were associated with M48, among which OTU375_ Clostridiales was most associated with M48. Likewise, Clostridiales was determined with varied effects on health [37,38]. OTU186_Bacteroides was most associated with C2W in this study. Bacteroides species were determined as opportunistic pathogens and probiotics in different studies [39,40].  The top 10 OTUs with most correlations in the networks of M48, M1W, and M2W were largely different, so as for the CCl 4 groups, suggesting the gut bacterial networks changed in both the MSC and CCl 4 groups at different time points. Multiple distinct gatekeepers were determined in the networks of the MSC and CCl 4 groups over time. Among them, OTU1352_S24-7 (i.e., a gatekeeper in M48 network) was also associated with M48 by LEfSe analysis, while OTU453_S24-7, OTU1213_Ruminococcaceae, and OTU841_Ruminococcus (i.e., three gatekeepers in C2W) were also associated with C2W, suggesting these phylotypes could play a key role in maintaining the gut bacterial communities of M48 and C2W (i.e., the cohorts with the highest LIDR in the MSC groups and CCl 4 groups, respectively). As mentioned above, S24-7 was determined with varied effects on health. As for the beneficial effect, the enriched S24-7 in the gut microbiota of mice fed with Lactobacillus reuteri ATG-F4 was associated with the improvements of psychological status of a murine model [41]. Ruminococcaceae was believed as a vital member in maintaining the gut health [42], while some Ruminococcus species were determined as normal members in gut microbiota [43].
Multiple functional categories were associated with gut microbiota of the MSC and CCl 4 groups over time in the current study. Among them, glycogen phosphorylase, heterodisulfide reductase subunit A, and hexosaminidase were most associated with M48, M1W, and M2W, respectively, while outer membrane usher protein, long-chain acyl-CoA synthetase, and methyl-accepting chemotaxis protein were most associated with C48, C1W, and C2W, respectively, suggesting these functional categories could play important roles in the changes of gut bacterial communities of the MSC and CCl 4 groups at different stages. Glycogen phosphorylase normally is located in the brain, liver, and skeletal muscle tissue. Heterodisulfide reductase played an important role in the energy-conserving metabolisms of bacteria and archaea [44,  . The serum beta-hexosaminidase level was associated with reticuloendothelial function of the patients with viral hepatitis [46]. β-Barrel assembly machine-mediated folding of outer membrane usher protein could be selectively disrupted by nitazoxanide [47]. Long-chain acyl-CoA synthetase in fatty acid metabolism was involved in multiple liver diseases [48]. Methyl-accepting chemotaxis protein was associated with the changes of hepatic inflammation [49].

Conclusion
In conclusion, the results suggest that the intestinal microbiota, the corresponding networks, and functional categories changed during the recovery dynamics of the MSC or CCl 4 group. MSC-treated mice were determined with the mildest intestinal microbial dysbiosis at 48 h, with OTU1352_S24-7 as the vital gut phylotype. CCl 4 -treated mice experienced 11 BioMed Research International the least intestinal microbial dysbiosis status at two weeks, with OTU453_S24-7, OTU1213_Ruminococcaceae, and OTU841_Ruminococcus as the vital gut phylotypes. These findings may assist in monitoring the dysbiosis status of intestinal bacterial communities of CCl 4 -treated mice with or without MSC transplantation over different time points.

Data Availability
The raw sequencing data were deposited in NCBI under Bio-Project accession no. PRJNA660814 and will be available once the manuscript is accepted.

Ethical Approval
All animal experimental procedures were conducted according to a protocol approved by the Ethics Committee of the First Affiliated Hospital of Zhejiang University (No. 2015-130).  Figure S2: optimal cluster numbers determined by average silhouette analysis. Figure S3: identification of OTUs associated with the CCl 4 and negative control (NC) groups by linear discriminant analysis (LDA) effect size (LEfSe). Figure