A More Robust Gut Microbiota in Calorie-Restricted Mice Is Associated with Attenuated Intestinal Injury Caused by the Chemotherapy Drug Cyclophosphamide.

Improving the gut microbiota via calorie restriction is beneficial for human health. Our findings showed differential responses between calorie-restricted mice and ad libitum-fed mice. Compared with the ad libitum-fed mice, the calorie-restricted mice were less susceptible to cyclophosphamide side effects otherwise observed on the gut integrity and its microbiota. These results show the potential benefits of manipulating the gut microbiota with CR prior to cancer chemotherapy.

induced overgrowth of sulfate-reducing bacteria Desulfovibrio spp. that damaged the gut barrier (15,16). In contrast, increasing the abundance of protective bacteria, such as Bifidobacterium spp. and Lactobacilli spp., was associated with improvement of the intestinal barrier function by inhibiting the adhesion of pathogenic bacteria to the intestinal wall or modulating epidermal growth factor receptor-mediated intracellular signaling (17)(18)(19)(20)(21). Previous studies showed that chemotherapy can shift the structure of gut microbiota, such as significant enrichment of the Clostridium leptum in mice and reduction of the numbers of anaerobic bacteria in the human gut (22,23). Moreover, diminishment of the intestinal microbiota by antibiotic treatment aggravated the toxicity of chemotherapy to the intestinal epithelium (22). This result suggests that the manipulation of the gut microbiota may provide a promising strategy to counteract the side effect of CTX.
Diet is the most important factor to modulate the gut microbiota (24,25). Not only the composition of the diet but also the quantities of food consumed can modify the gut microbiota (26). Previous studies showed that calorie restriction (CR) can significantly shift the gut microbiota in both humans and mice (27,28). One major change in CR mice was the enrichment of protective bacteria such as Lactobacillus spp. (26). Our recent study identified a strain of Lactobacillus murinus that thrived in CR mice and contributed to the protection of the gut barrier and attenuation of ageing-associated inflammation (29). CR or short-term starvation has been shown to mitigate the CTXcaused oxidative stress in the mouse model (30)(31)(32). However, it was still unclear whether the CR-modulated gut microbiota contributes to alleviation of side effects from chemotherapy.
To investigate this question, we administered CTX to C57BL/6J mice after 4 weeks of CR, and the mice that were fed ad libitum were used as control group. We found that the CR-modulated gut microbiota was associated with protection of the intestinal barrier and epithelium from damage caused by CTX. These results suggest that manipulating the structure of the gut microbiota may be used as a strategy to alleviate the side effects of CTX.

RESULTS
Calorie restriction protected mice from the toxicity of cyclophosphamide observed in the intestinal epithelial cells. We randomized 8-week-old male C57BL/6J mice into two groups (n ϭ 35 per group): fed ad libitum with normal chow diet (ad libitum group) or fed with 70% normal chow of the ad libitum (CR group) (see Fig. S1 in the supplemental material). The body weight of mice in the CR group was significantly decreased and stayed stable after 14 days (Fig. S2B). Hematoxylin-eosin staining of the distal colon showed that the structure of the intestinal epithelium had no significant differences between the CR and ad libitum group at 28 days. However, the serum level of lipopolysaccharide (LPS)-binding protein (LBP) (33), which reflects the bacterial antigen load in host blood, was significantly lower in the CR group ( Fig. S2C to E).
After 4 weeks, half of the mice in each group were intraperitoneally injected with CTX (100 mg/kg body weight), and the other half were injected with normal saline (NS) as the control (Fig. S1). CTX induced significant decrease of food consumption in the ad libitum-fed mice, but not in CR mice (Fig. S3A). CTX showed no significant effects on body weight in both CR mice and ad libitum-fed mice (Fig. S3B).
We evaluated the effect of CR on protecting intestinal epithelium from the side effects of CTX treatment in mice. From the 2nd day of CTX injection, CTX induced inflammatory infiltration and epithelium injury in both CR and ad libitum-fed mice ( Fig. 1A and B), but the histological score was significantly higher in the ad libitumϩCTX group than in the CRϩCTX group. Moreover, although the intestinal stem cells (ISC) in the colon were damaged by CTX in both CR and ad libitum-fed mice (Fig. 1C), the number of ISC was significantly higher in the CRϩCTX mice than in the ad libitumϩCTX mice (Fig. 1D).
To detect the effect of CTX on intestinal permeability, we orally administered 4,000 Da fluorescein isothiocyanate (FITC)-labeled dextran to the mice and then measured translocation of fluorescence into plasma. The concentration of FITC in serum was significantly higher in the ad libitumϩCTX group than in the CRϩCTX group after 2 days of CTX injection (Fig. 1E). Seven days after CTX injection, we were able to detect marked bacterial growth in the mesenteric lymph nodes (MLN) and spleen of the ad libitumϩCTX group compared to the ad libitumϩNS mice, while the bacterial load was much lower in both the CRϩCTX and CRϩNS groups (Fig. 1F). Moreover, we found that LBP was significantly increased after CTX injection in both CR and ad libitum-fed mice. However, the concentration of LBP in serum was significantly lower in the CRϩCTX group than in the ad libitumϩCTX group (Fig. 1G). The decreased translocation of gut bacteria and bacterial antigen reflect the protection of gut barrier function by CR (22).
Taken together, these results suggest that the detrimental effects of CTX on the intestinal epithelium and gut barrier function were attenuated by CR.
The structure of microbiota in calorie-restricted mice was more robust against cyclophosphamide treatment-induced changes. Our previous studies showed that the CR mice had a unique gut microbiota dominated by potentially beneficial bacteria such as Lactobacillus spp. (27). In the current study, the overall structure of the gut microbiota in CR mice was significantly shifted after 4 weeks of CR, as shown by principal-coordinate analysis (PCoA) based on Bray-Curtis distance (Fig. S4A, P Ͻ 0.001 In ad libitum-fed mice, the gut microbiota of the CTX group was clearly separated from the NS group after CTX injection ( Fig. 2A and C). There was no significant difference between the CRϩNS and CRϩCTX groups ( Fig. 2B and C). The structure of gut microbiota showed significant differences between the CTX and NS group from the third to seventh day after CTX injection in the ad libitum-fed group ( Fig. 2C; see also Table S1 in the supplemental material). In CR mice, the CTX treatment showed smaller impact on their gut microbiota (Fig. 2D).
These results suggest that CTX could induce significant change of the microbiota structure in the ad libitum-fed group, but the gut microbiota in the CR group were more stable than the gut microbiota in the ad libitum group in response to CTX effects.
Specific phylotypes of the gut microbiota were modulated by calorie restriction and cyclophosphamide treatment. On the basis of the results of our redundancy analysis (RDA), we identified 79 operational taxonomic units (OTUs) that were significantly changed in CR mice compared to ad libitum-fed mice after 4 weeks of CR (Fig. S4B). Compared to ad libitum-fed mice, OTU1 that belongs to the genus Lactobacillus became the predominant bacteria in CR mice, and 26 other OTUs were also significantly enriched in the CR group. In contrast, 53 OTUs were decreased in the CR group, most of which were phylotypes from the family Porphyromonadaceae (Fig. 3).
After injecting CTX into the mice, we identified 83 OTUs that were shifted in responding to the CTX stress in ad libitum-fed mice ( Fig. 3 and Fig. S5). Twenty-one OTUs were increased and 62 OTUs were decreased in response to CTX. None of these OTUs were significantly affected by CTX in CR mice (Fig. 3). Among the 83 OTUs, 29 OTUs were also significantly different between CR and ad libitum-fed groups at 28 days. Six OTUs were enriched in CR mice and reduced by CTX in the ad libitum-fed group. These OTUs belonged to the Lachnospiraceae (2 OTUs), Porphyromonadaceae (2 OTUs), and Lactobacillus (1 OTU). However, five OTUs that belong to Porphyromonadaceae were reduced in CR mice and increased by CTX in the ad libitum-fed group OTUs are arrayed by the Spearman correlation cluster. The key comparisons and direction of changes are summarized in three columns to the right of the heat map. A two-sided Mann-Whitney test was used for analysis. The FDR of Ͻ0.05 are shown with blue and pink box. The "ad libitum Vs. CR" column compares CRϩNS mice to ad libitumϩNS mice at 28d. A blue box indicates that the OTU was significantly lower in CRϩNS mice, and a pink box indicates that it was significantly higher. The "NS Vs. CTX in ad libitum" column compares ad libitumϩCTX mice to the ad libitumϩNS mice. A blue box indicates that the OTU was significantly decreased by CTX, and a pink box indicates that the OTU was significantly increased by CTX. The "NS Vs. CTX in CR" column compares CRϩCTX mice to the CRϩNS mice.
Gut Microbiota in Calorie-Restricted Mice and CTX ® (Fig. 3). On the other hand, 54 OTUs were affected only by CTX. For example, 11 OTUs were enriched by CTX, and all these OTUs belong to Erysipelotrichaceae, Porphyromonadaceae, and Sutterellaceae. However, 43 OTUs were reduced by CTX, most of which were species from the family Lachnospiraceae and Ruminococcaceae, whose members contained butyrate-producing bacteria (34).
Key phylotypes showed guild-like response to calorie restriction and cyclophosphamide treatment. Bacteria may function together as a functional group called guilds (35). Coabundance analysis may capture such structures in gut microbiota (36). We then constructed a coabundance network to illustrate the potential interaction among the 146 nonredundant OTUs that were significantly affected by CTX or CR. These OTUs were clustered into 22 coabundance groups (CAGs) based on Spearman correlation analysis (Fig. 4A).
Six CAGs were significantly different between the ad libitumϩCTX group and ad libitumϩNS group (Fig. 4B). For example, CAG3, which was mainly composed of the OTUs from Lachnospiraceae, was significantly decreased by CTX. However, CAG8, CAG15, CAG16, and CAG21 were increased by CTX (Fig. S6). Twenty CAGs were significantly shifted by CR (Fig. 4B). In contrast to ad libitum-fed mice, all of the CAGs The size of the node indicates the mean abundance of each OTU. Lines between nodes represent correlations between the nodes they connect, with the line width indicating correlation magnitude and color for correlation (red for a negative correlation, gray for a positive correlation). Only lines corresponding to correlations with a magnitude greater than 0.5 are drawn. OTUs are grouped into 22 CAGs by permutation multivariate analysis of variance (PERMANOVA) when P Ͻ 0.05. (B) Heat map of the FDR value for CAGs, indicating the changes induced by diet or CTX stress. The tests of significance were performed using a two-sided Mann-Whitney test, and the calculation of FDR was performed using the procedure originally introduced by Benjamini and Hochberg.
were not significantly changed by CTX in CR mice. These results suggested that CR can significantly reduce the modulating effects of CTX on the gut microbiota.
CAG8, which was composed of OTUs belonging to Porphyromonadaceae, was significantly enriched by CTX but reduced by CR. In contrast, four CAGs were significantly negatively correlated with CAG8, including CAG19, which mainly contained OTU1 in Lactobacillus spp.

DISCUSSION
Our study shows that CR attenuated the mucositis caused by CTX, including intestinal permeability, bacterial translocation, and epithelium damage, while at the same time, CR changed the gut microbiota structure which became more robust under the later CTX treatment.
CTX exacerbated the dysbiotic state of the gut microbiota, which not only reduces the effects of tumor suppression but also decreases the quality of life of the patients (37,38). Clinically, more and more pieces of evidence focused on the role of gut microbiota in chemotherapy. For instance, recently probiotics were used in combination with chemotherapy to alleviate the side effects, such as diarrhea (39). CR or starvation could reduce the toxicity of CTX (32,40,41). Starvation can be efficient in protecting the host from oxidative stress induced by CTX in both rats and mice (32,41). However, the mechanism of such protective effects are still unclear. In our current work, we focus on CR modeled gut microbiota to mitigate the side effects of the chemotherapy drug.
Our findings suggest that the gut microbiota structure in CR mice was linked to reduced CTX-induced damage. A significant change in the overall structure of the gut microbiota was observed in CTX-treated mice in our work. Interestingly, CR mice maintained their gut microbiota structure despite the treatment effects of CTX, showing a significant robustness against perturbations. Accumulating evidence suggests that robust gut microbiota is tightly linked with the host health (42,43). A previous study of mice showed that greater perturbation on the gut microbiota by antibiotic treatment increased higher susceptibility to intestinal colonization, more disruption in the microbiota, and more severe intestinal pathology than mice whose gut microbiota were not perturbed by Salmonella enterica serovar Typhimurium (44). These results suggest that more stable gut microbiota structure may bring more benefits to the host against disorders induced by various perturbations such as chemotherapy.
In addition to the more robust overall structure of the gut microbiota, specific bacterial phylotypes or their "guilds" may play a more significant role in mediating the CR-induced protective effects. In the current study, we also detected similar modulation of gut microbiota by CR, especially significant increase of Lactobacillus spp. (27). The gut microbiota predominated by Lactobacillus spp. has been shown as the unique characteristic of mice on both long-term and short-term CR (27,29). The Lactobacillus murinus CR147 strain, isolated from healthy CR mice, can enhance intestinal barrier function and reduce systemic inflammatory marker in old microbiota-colonized mice (29). Other strains, such as Lactobacillus plantarum DSM 2648, also can tightly adhere to the intestinal epithelium and abrogate bacterial translocation to mesenteric lymph nodes (45). The L. plantarum MB452 strain can also enhance the intestinal barrier function via adjusting the tight junction-related genes (46). Moreover, the Lactobacillus rhamnosus Gorbache Goldin strain can reduce intestinal and liver oxidative stress (47). Thus, these results support our finding that the increased abundance of Lactobacillus phylotypes in our CR mice may contribute to reducing the intestinal permeability and oxidative stress caused by CTX.
We also found that specific phylotypes of Lachnospiraceae, members of which are known as butyrate-producing bacteria (34,48), were increased by CR and decreased by CTX. Butyrate acts as an energy substrate for the colonocytes and has a trophic effect on mucosa (49,50). The enrichment of these Lachnospiraceae phylotypes might play an important role in the protection of intestinal epithelium by CR. On the other hand, the majority of Barnesiella phylotypes decreased upon CR. The members of Barnesiella were Gut Microbiota in Calorie-Restricted Mice and CTX ® reported to be associated with systemic inflammation and oxidative DNA damage after radiotherapy (51). Reduction of this group of bacteria might be one of the reasons for the lower amount of damage caused by CTX in CR mice. The results from the current work, supported by previous findings, show that the gut microbiota improved by CR may play an important role in protecting the host from the damage caused by chemotherapy.
In conclusion, our work suggests that CR-remodeled gut microbiota mitigate the side effects of chemotherapy. Gut microbiota should have a critical role in the development of precision treatment strategies for cancer, and it will be increasingly seen as a component for next-generation cancer therapies.

MATERIALS AND METHODS
Animal trial and samples. Specific-pathogen-free, 6-week-old male C57BL/6 mice (n ϭ 70) were purchased from SLAC Inc. China (Shanghai, China). All mice were housed individually and randomly separated into two groups, the calorie-restricted (CR) group (n ϭ 35) (fed with 70% normal chow diet of the ad libitum-fed mice) and the ad libitum-fed group of mice (n ϭ 35) (see Fig. S1 in the supplemental material). The daily consumption of food in the ad libitum-fed group was recorded over 1 week and averaged to determine the amount of food given every day for the following week for CR. We weighed and changed food once a day for all the mice. After 28 days of calorie restriction, the CR group mice were assigned into two groups, the CRϩNS group and CRϩCTX group. At day A0, cyclophosphamide (CTX) (100 mg/kg of body weight) was intraperitoneally injected into CRϩCTX mice, and normal saline (NS) was intraperitoneally injected into CRϩNS mice (once) for a control. The same grouping was also applying to the ad libitum-fed mice: ad libitumϩCTX intraperitoneally injected with CTX and ad libitumϩNS intraperitoneally injected with normal saline (Fig. S1). To determine toxicity and efficacy, the mice were monitored routinely for weight loss and general behavior (52).
Fresh feces were collected daily. All fecal samples were stored at Ϫ80°C until analysis. Five mice from each group were sacrificed on day 28 (28d), A2d (day 2 ad libitum), A7d, and A10d. The blood, colon contents, colon tissues, and cecum content were collected. Mice were humanely euthanized prior to serum and tissue sample collection.
Histological analysis of colon tissues. For histological analysis, tissues were fixed in 4% paraformaldehyde, then ethanol dehydrated, embedded in paraffin, and sectioned as described previously (54). Hematoxylin-eosin (HE) staining was performed by standard methods. Histological scoring was performed by measuring inflammation and damage as previously described (54).
All the images were taken by using a Leica CTR6000 microscope. Brightness and contrast were adjusted linearly across the entire image for any particular image.
Intestinal permeability assay and bacterial translocation detection. The gut barrier integrity was assessed by permeability to fluorescein isothiocyanate-dextran (FITC-dextran; Sigma). After injecting the mice for 48 h with NS or CTX at 100 mg/kg body weight, mice were forced to fast for 4 h and then orally fed with FITC-dextran at 0. 6 mg/g body weight. After 4 h, the mice were euthanized and exsanguinated by cardiac puncture. Plasma FITC levels were subsequently determined using a fluorescence spectrophotometer (485/545 nm) (22).
Mesenteric lymph nodes and spleens were aseptically removed, smashed in PBS (200 l), and plated onto LB agar plates. After 48 h of aerobic culture, the numbers of CFU were calculated and analyzed statistically.
Serum LBP measurement. Blood samples were collected from the eyes and centrifuged at 12,000 ϫ g for 30 min to pellet blood cells, and the serum samples were stored at Ϫ80°C until further analyses. Serum LBP was determined after a dilution of 1:1,600 using the Mouse Lipopolysaccharide Binding Protein ELISA kit (Cell Sciences, Canton, MA, USA) according to the instructions of the manufacturer.
Microbiota DNA extraction and Illumina V3-V4 regions in 16S rRNA gene sequencing. DNA was extracted from fecal samples at the 28d, A2d, A3d, A4d, A7d, and A10d and analyzed as previously described (55). The extracted DNA purified with the Omega Gel Extraction kit (catalog no. D2501-01; OMEGA Bio-Tek, Taiwan, China) using both physical and chemical lysis. DNA concentration and integrity were determined both visually by electrophoresis on a 1% agarose gel and spectrophotometrically by using a biodrop instrument.
The microbiota composition was assessed by Illumina targeting the V3-V4 region of the bacterial 16S rRNA gene with the primer. The primers were used to build the library by PCRs as previously described (56). PCR was performed using the following program: predenaturation at 94°C for 3 min; 22 cycles, with 1 cycle consisting of denaturation at 94°C for 30 s, annealing at 55°C for 30 s, and elongation at 72°C for 30 s; final extension at 72°C for 8 min.
For the Index PCR (attachment of dual indices and Illumina sequencing adapter using the Nextera XT Index kit), PCR program of Index PCR was the same as for Amplicon PCR except the cycle number decreased to 8. The purified products from different samples were mixed at equal ratios for sequencing using the Illumina MiSeq platform (Illumina Inc., USA).
Bioinformatics and statistical analysis. Both the forward and the reverse ends were cut off from the same reads at the first base for which the Q value was less than 2. All of the reads were kept when the length was more than 399 bp and the expected error was less than 0.5 (57).
High-quality sequences were clustered into OTUs (operational taxonomic units) at 97% identity by Usearch and the representative nonchimeric OTU sequences were picked by Uparse's default (58). The number of high-quality reads was more than 10,000 for each sample. The representative sequences of each OTU were classified by the RDP classifier online, and the RDP-classified sequences were used for taxonomical assignments at 80% confidence level (59).
The tree, together with sequence abundance data, was then used for beta-diversity analysis based on weighted metric by QIIME 1.6 (60). The relative abundances of OTUs were used for principal-component analysis, multivariate analysis of variance, and redundancy analysis via Matlab R2015a (The MathWorks, Natick, MA, USA) and Canoco for Windows 4.5 (Microcomputer Power, NY, USA).
Two-way ANOVA test and Mann-Whitney test were used to test the statistical significance of the physiological and biochemical data via software SPSS 19.0 (SPSS Inc, Chicago, IL, USA). P values were adjusted by the method of Benjamini and Hochberg (61).
Statistical analyses. Redundancy analysis was introduced to identify specific bacterial phylotypes that contributed to the segregation of gut microbiota by calorie restriction and CTX. The samples from all the mice at 28 days were used to establish classification models of diet. Samples from mice on the ad libitumϩCTX and ad libitumϩNS on A3 day, A4 day, and A7 day were used to establish classification models of CTX.
The correlation among OTUs was calculated using the Spearman algorithm. PERMANOVA (9,999 permutations, P Ͻ 0.05) based on Spearman correlation coefficients was used to cluster the OTUs into coabundance groups (CAGs) using the R program.
Accession number(s). The 16S rRNA gene sequence information in this study has been submitted to the GenBank Sequence Read Archive database under accession number SRP166816 (https://www .ncbi.nlm.nih.gov/).

ACKNOWLEDGMENTS
This work was supported by grants from the National Natural Science Foundation of China (31330005, 81401141, 81871091) and the Science and Technology Commission of Shanghai Municipality (14YF1402200).