A total of 60 patients with CRC who were due to undergo radical colorectomy at Shanghai Jiao Tong University Affiliated Shanghai Sixth People's Hospital (Shanghai, China), were selected between September 2011 and December 2013. The inclusion and exclusion criteria are listed in (Table I). A total of 60 patients were assessed for eligibility; however, only 38 were finally enrolled due to the fact that 22 failed to meet the inclusion criteria or refused to participate in the study. A total of 22 patients completed the trial (Fig. 1). All eligible patients (n=22) were randomized and divided into two treatment groups: The CGT group (n=11) received perioperative placebos and the PGT group (n=11) received probiotics. There were no significant differences in gender, age, body mass index (BMI), cancer stage and time between onset of symptoms and hospital admission between the two groups. No significant differences in the levels of albumin, hemoglobin or creatinine were observed between the two groups (Table II).

In order to analyze further differences in the microbiota between patients with CRC and the healthy population, an additional 11 healthy volunteers (HGT group) were enrolled, who were part of a larger population-based case-control study at the Shanghai Jiao Tong University (Table II). The inclusion and exclusion criteria for the controls are listed in Table I.

A single-center prospective randomized control study design was used. Equal randomization was accomplished using a computer-generated random allocation schedule. The probiotic and placebo were sealed with aluminum foil and appeared identical in the two groups, with an identical smell and taste. Only one nurse not directly involved in the trial was able to solve the treatment codes in the event of an emergency. An acid-resistant coating was used to prepare the capsules containing the probiotics and placebo. Patients in the PGT group received an encapsulated probiotics preparation (Shanghai Xinyi Pharmaceutical Co., Ltd., Shanghai, China) containing live combined Bifidobacterium longum, Lactobacillus acidophilus and Enterococcus faecalis (1:1:1) with no less than 1.0×10⁷ CFU/g viable cells, three times/day, with a total daily dose of 6.0×10⁷ CFU for five days. Patients in the CGT group received encapsulated maltodextrin (Haiyan Liuhe Pharmaceutical Industry Co., Ltd., Jiangsu, China) three times a day. During the study period, no parenteral or enteral nutritional supplementation was administered. All patients received a regular diet preoperatively and conventional bowel preparation was performed on pre-operative day 1, including the administration of a low-residue diet and mechanical bowel preparation. All patients were given Soffodex (Jiangxi Hygecon Pharmaceutical Co., Shangrao, China), containing 2.4 g monobasic sodium phosphate and 0.9 g dibasic sodium phosphate. Parenteral hydration was administered on the morning of the surgery supplied via a central venous catheter. A 12F catheter was placed through a jejunal limb during surgery for gastric aspiration to reduce colon anastomotic fluxion. For prophylaxis, 500 mg metronidazole (Sine Wanxiang Pharmaceutical Co., Shanghai, China) and 1 g ceftriaxone (Shanghai Roche Pharmaceutical Co., Shanghai, China) were administered 1 h prior to induction and continued for 48 h following the operation. All subjects were interviewed by the study nurse, and reactions to the product, medications taken and adverse events that occurred during the five-day period were recorded.

Healthy subjects underwent a standard bowel cleansing preparation with oral polyethylene glycol electrolyte (139.12 g/2,000 ml; Jiangxi Hygecon Pharmaceutical Co.) the evening prior to colonoscopy. Endoscopically, the colonic mucosa appeared normal in all subjects. Colonic mucosal tissue biopsies (~1×2 mm each) were collected from the ascending colon (n=3), descending colon (n=1), sigmoid colon (n=3) and rectum (n=4) during colonoscopy.

The present study was approved by the Ethics Committee of the Sixth People's Hospital Affiliated to Shanghai Jiaotong University on Human Subjects in Medical Research [protocol number 2011 (L)-11] and all individuals provided written informed consent prior to participating in the study.

Colorectal mucosa tissue samples were collected from the tumor sites during surgery. Normal colorectal mucosa tissue samples were collected from the cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum during colonoscopy. Fisher's exact test was used for comparison of different anatomic sites. All samples from the study participants, including healthy individuals, were placed in liquid nitrogen immediately and transported to the laboratory within 30 min of collection. DNA was extracted from all intestinal tissue samples using MoBio Powersoil DNA extraction kits (MoBio Laboratories, Inc., Carlsbad, CA, USA), according to the manufacturer's instructions. Intestinal tissue samples were completely lysed by overnight incubation at 56°C in buffer ATL (Qiagen, Düsseldorf, Germany) and proteinase K (Shanghai SAIBAISHENG Gene Technology Co., Ltd., Shanghai, China). The standard tissue protocol was modified to extend the 70°C incubation step from 10 to 30 min.

DNA was eluted from a column in a final volume of 200 µl elution buffer and stored at −20°C prior to the amplification steps.

The V3 region of the bacterial 16S rRNA gene was amplified from extracted DNA using primers (Sangon Biotech Co., Ltd., Shanghai, China) 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 533R (5′-TTACCGCGGCTGCTGGCAC-3′) (33), with a sample barcode sequence and the FLX Titanium adapters (Roche Diagnostics, Basel, Switzerland) were used to amplify the V3 region of each tissue sample by polymerase chain reaction (PCR). The PCR mixture was composed as follows: 10 ng template, 0.4 ml FastPfu Polymerase (Beijing TransGen Biotech Co., Ltd., Beijing, China), 4 ml 56 FastPfu buffer (Transgene, Illkirch-Graffenstaden, France), 2 ml deoxynucleotide diphosphate (2.5 mM each; Takara Bio, Inc., Otsu, Japan), 0.4 ml forward primers (5 mM) and 0.4 ml reverse primers (5 mM). The reaction was performed in an ABI GeneAmp H 9700 Cycler (Applied Biosystems Life Technologies, Foster City, CA, USA) and the cycling parameters were as follows: 5 min denaturation at 95°C followed by 25 cycles of 30 sec at 95°C (denaturation), 30 sec for annealing at 55°C and 30 sec at 72°C (elongation), with a final extension at 72°C for 5 min. Triplicate PCR reactions were performed on each sample. Amplified products from stool samples were verified by gel electrophoresis using 5 ml PCR reaction mixture in a 2.0% agarose gel. The PCR products were purified using the AxyPrepDNA Gel extraction kit (Corning Life Sciences, Corning, NY, USA) and quantified on a QuantiFluor™-ST Fluorometer (Promega Corporation, Madison, WI, USA). The products from different samples were mixed at equal ratios for pyro-sequencing using the Roche GS FLX 454 Sequencer (Roche Diagnostics) according to the manufacturer's instructions. All pyrosequencing reads were then removed of their primers, barcodes and adaptor sequences, and were further screened and filtered according to the standards for quality control as follows: The elimination of sequences that did not perfectly match the proximal PCR primer (over two mismatches to the primers), those with short sequencing length (<200 nucleotides) sequences that contained mononucleotide repeats of six nucleotides, sequences with ambiguous characters, or sequences with a read quality score <25. Finally, a total of 416,599 high-quality sequences from 22 samples were produced, which accounted for 72.9% of valid sequences according to barcode- and primer-sequence filtering.

In order to gain high-quality and more precise bioinformation, effective sequences that contained certain point mutations and macromolecular homopolymers were narrowed down using QIIME, version 1.17 (http://qiime.org/) based on the following methods and parameters: Number of mismatching of primers, <2; maximum number of 3′ primers, 3; no ambiguous bases or >6 homologous mono-bases; and sequence length, <200 base pairs (34). The optimized sequences were then clustered into operational taxonomic units (OTUs) using Usearch, version 7.1 (http://drive5.com/uparse/) with a criterion of minimum similarity of 97%. Chimera sequences arising from the PCR amplification were detected and excluded from the OTUs using UCHIME version 4.2.40 (http://drive5.com/usearch/manual/uchime_algo.html) (35). Using the usearch_global command, representative OTUs were aligned to the optimized sequences and the abundance of OTUs per sample was obtained for performing the following further analysis.

Applying Bayesian algorithms of ribosomal database project (RDP) classifier (http://rdp.cme.msu.edu/), representative OTUs were identifed in the following databases: 16S bacteria and archaeal ribosomes [SILVA (36); release 115; http://www.arb-silva.de]; RDP (21) (release 11.1; http://rdp.cme.msu.edu/); Greengene (37) (release 13.5; http://greengenes.secondgenome.com/); ITS fungus [Unite (33); release 5.0; http://unite.ut.ee/index.php]; and functional genes (FGR; release 7.3; http://fungene.cme.msu.edu/). Subsequently, analysis with Good's coverage estimator, diversity estimators (Shannon and Simpson), richness estimators [Chao and abundance-based coverage estimators (ACE)] were performed and the rarefaction curve was generated using the Mothur software package (http://www.mothur.org/wiki/Main_Page) at an 80% confidence level. In addition, the Bray-Curtis dissimilarities were used to construct a cluster dendrogram. A metagenomic biomarker discovery approach was employed with linear discriminant analysis (LDA) coupled with effect size measurement (LEfSe), which was used to perform a non-parametric Wilcoxon sum-rank test followed by LDA analysis using online software (http://huttenhower.sph.harvard.edu/galaxy/) to assess the effect size of each differentially abundant taxon.

Results were analysed using SPSS, version 18.0 for Windows (SPSS, Inc., Chicago, IL, USA). Quantitative values are expressed as the mean ± standard deviation. Comparison of categorical data between groups was performed using the Pearson χ² test or, where indicated, Fisher's exact test. Analysis of variance or the Kruskal-Wallis test was used for continuous variables, as appropriate. P<0.05 was considered to indicate a statistically significant difference.

The 16S sequence data generated in the present study were submitted to the GenBank Sequence Read Archive with the accession number SRP043381.

A total of 416,599 high-quality, classifiable reads were obtained, with an average of 12,624±2,423 (n=33) reads/sample. UPARSE, version 7.1 (http://drive5.com/uparse/) was used to cluster the above reads. At a 3% dissimilarity level, 45,453 OTUs in all samples and an average of 1,377 OTUs (n=33) were identified per sample.

The PGT group exhibited significantly higher Chao (P=0.013) and ACE (P=0.01) indices in colorectal mucosa compared with those in the HGT group. There were no significant differences between the HGT and CGT groups or CGT and PGT groups (Fig. 2).

Compared with the HGT group, the Shannon (P=0.028) and Simpson (P=0.046) indices were significantly reduced in the CGT group. This demonstrated that the mucosal microflora diversity was reduced in patients with CRC, while the diversity of the intestinal flora increased following oral administration of probiotics.

The gut micro-flora structure was compared at various bacterial levels. At the phylum level, there were no significant differences in classification among the three groups (27 in the CGT group, 29 in the PGT group and 25 in the HGT group). A micro-flora whose abundance was >0.1% contributed to 99.8% of the total category, including Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria and Fusobacteria. At the genus level, 188 were present in the CGT group, 198 in the PGT group and 201 in the HGT group. Bacteria whose abundance was >0.01% contributed to 96.86% of the total category in the HGT group, compared with 99.76% in the CGT group and 99.13% in the PGT group. A cladogram representation of the structure of the mucosal microbiota (Fig. 3) and a bar graph showing the predominant bacteria (Fig. 4) were generated by LEfSe. The greatest differences in taxa between the three communities are displayed. Pyrosequencing data demonstrated that probiotics-treated patients exhibited a significant reduction in Peptostreptococcus, Comamonas, Fusobacterium and expansion of Enterococcus and Proteobacteria in the mucosa-adherent microbiota.

To determine whether microbial alterations in the mucosa-associated microbiota mediate the effects of probiotics in CRC, a taxonomy-based comparison was performed to determine the differences between the microbiota of CRC and healthy individuals (Table III). Fusobacterium, which constitutes less than 0.1% of total bacteria in healthy mucosal tissue, was the most prevalent in the mucosa of patients with CRC (10.08 vs. 0.01%; P=0.03). In addition, probiotics treatment was observed to be associated with a significant (~6-fold) reduction in the relative abundance of Fusobacterium (1.91%; P=0.03). Accordingly, it was concluded that a reduction of the mucosa-associated Fusobacterium group may be a potential mediator of the effects of probiotics on CRC.