The research protocols for the present investigation were approved by the Ethics Committee at Sun Yat-sen University (Guangzhou, China). Written informed consent was provided by all participants prior to the initiation of the experiment.

A total of 15 patients with colorectal cancer (nine males and six females) and 12 healthy control individuals from the Physical Examination Center at the Department of Gastroenterology, the Third Affiliated Hospital of Nanchang University (Nanchang, China) participated in the present study between June 2013 and October 2014 at the Third Hospital Affiliated of Nanchang University. All patients with colorectal cancer were diagnosed for the first time according to the diagnostic criteria proposed by the International Union Against Cancer and the American Joint Committee on Cancer in 2003 (15). Patient exclusion criteria included those who experienced colorectal cancer recurrence post-surgery, underwent chemotherapy, had colorectal cancer complicated with metabolic diseases (such as diabetes mellitus), received antibiotics within one month, administered nonsteroidal anti-inflammatory drugs (NSAIDS), statins or probiotics within two months prior to the initiation of the experiment, suffered chronic intestinal diseases and had a history of food allergies. The average age of the patients was 52.5 years (range, 40–60 years). Among the 15 patients, three cases were of ascending colon cancer, two were transverse colon cancer, four were descending colon cancer, one was sigmoid colon cancer and five were rectal colon cancer. The clinical stages of these patients were stage II in four cases, stage III in six cases and stage IV in five cases.

The general characteristics of healthy control individuals were recorded, including age, gender and medical history. The exclusion criteria for healthy controls included those who had a medical history of cancer, diabetes, heart disease and other metabolic syndrome-related diseases, had recently received antibiotics, NSAIDS, statins or probiotics, had suffered from chronic intestinal diseases and had a history of food allergy.

Stool samples were collected prior to surgery or bowel preparation. All participants consumed a bland diet and did not smoke or consume alcohol one day prior to sample collection. A stool sample (500 mg) was collected from the center of the stool using a sterilized cotton swab and stored at −20°C. Prior to gut flora detection, a stool sample (100 mg) was emulsified with phosphate-buffered saline followed by vibration for 1 min. Samples were subsequently placed at 0°C for 5 min and the top water-soluble layer of extraction was collected and centrifuged at 3,000 x g for 10 min at 4°C. Following this, the sample was filtered and stored at −80°C. Pretreatment of stool samples prior to metabolic profiling analysis was conducted as follows: A total of 100 mg stool sample was mixed with 1 ml of isopropanol:acetonitrile:water (3:2:2), homogenized and centrifuged at 6,500 × g for 5 min at 4°C. Following this, the samples were dried in a quick vacuum gauge and re-suspended in 50 µl pyridine-methoxy amino acid salt solution (15 mg/ml). Subsequently, the solution was incubated at 60°C for 45 min, sonicated for 10 min, and centrifuged at 3,000 × g for 5 min at room temperature. Following cooling to room temperature, GC/MS analysis was performed using the sample. For the analysis of short-chain fatty acids, 0.5 g stool was mixed with 3 ml double distilled H₂O, vibrated for 2 min and subsequently centrifuged at 10,000 × g for 5 min at room temperature. The supernatant was filtered using a 0.45-µm filter. A total of 1 ml filtrate was mixed with 0.1 ml sulfuric acid solution (50%) and 1 ml diethyl ether, followed by vibration for 30 min. Subsequently, the sample was centrifuged at 9,500 × g for 5 min and placed at −20°C for 30 min. Finally, the upper layer of diethyl ether was collected for GC/MS analysis.

Genomic DNA of flora was extracted using PowerSoil-htp 96-well DNA isolation kit (Mo Bio Laboratories, Inc.; Qiagen, Inc., Valencia, CA, USA), according to the manufacturer's instructions. According to the Jacob method, pyrosequencing was employed to specifically detect the V4 region of bacterial 16S ribosomal RNA on the isolated genomic DNA (3). The primer sequences used were: 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 909R (5′-GACTACHVGGGTATCTAATCC-3′). The reaction cycle parameters were as follows: Preheating at 95°C for 2 min; followed by 25 cycles of denaturation at 95°C for 30 sec, annealing at 55°C for 30 sec, extension at 72°C for 30 sec; and terminal extension at 72°C for 7 min. GenElute gel extraction reagent (Sigma-Aldrich; Merck KgaA, Darmstadt, Germany) was used for initial purification of the amplified product and the secondary purification was performed using AMpure beads (Beckman Coulter, Inc., Brea, CA, USA). DNA concentration was measured using a PicoGreen DNA detection kit (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA). A total of 1 µg DNA product was used for standardized 454 pyrophosphate sequencing analysis using a high throughput sequencing machine (Illumina MiSeq; Illumina, Inc., San Diego, CA, USA).

Samples were isolated using a 30-cm TG-5MS column (Thermo Fisher Scientific, Inc.), with parameters set as follows: Split ratio, 1:10; starting temperature, 80°C maintained for 30 sec. The temperature was subsequently raised to 330°C at a speed of 15°C/min and maintained for 8 min using ultra-pure helium as a carrier gas at a flow rate of 1.2 ml/min. The remaining parameters were: Injection port temperature, 260°C; splitless sampling, 1 µl; interface temperature, 280°C, solvent delay, 3 min; and scanning range, 50–650 M/z 5 times/sec. The parameters were adjusted for short-chain fatty acid analysis. The starting temperature of 22°C was maintained for 30 sec, followed by a temperature increase to 180°C at a speed of 8°C/min, which was maintained at 180°C for 1 min and subsequently raised to 200°C at a speed of 20°C/min. This temperature was maintained for 5 min with ultra-pure helium as a carrier gas at a flow rate of 1.2 ml/min. The remaining parameters were: Injection port temperature, 230°C; splitless sampling, 1 µl; interface temperature, 280°C; solvent delay, 3 min; and scanning range, 50–300 M/z 5 times/sec.

Gene sequences were read, edited and analyzed using Mothur 1.25 software (the University of Michigan, Ann Arbor, MI, USA). The obtained sequence reads were pre-clustered, allowing two mismatches. Operational taxonomic units (OTUs) were calculated via clustering by average neighbor principle at 97% genetic similarity. The species composition of samples was obtained through species annotation and species analysis. Subsequently, the numbers of sequence reads in each classification unit, and thus the relative abundance of each species, in each sample were calculated. Shared files and R language were utilized to create a profiling bar chart, and principal component analysis and species heatmap analysis were performed. For stool metabolic profile analysis, the spectral peak was distinguished and extracted according to Matlab script 7.0 (MathWorks, Natick, MA, USA), and the three-dimensional matrix table, consisting of retention time, mass-to-charge ratio and peak intensity, was obtained. The table was imported into SIMCA 14.0 software, (Umetrics, Umea, Sweden) and multi-dimensional statistical analysis was conducted.

Statistical analysis was performed using SPSS 6.0. (SPSS, Inc., Chicago, IL, USA). Analysis of molecular variance and Student's t-test were used to analyze the difference between bacterial species and metabolic profiling. Relationships between metabolites and bacteria were represented by the Pearson coefficient (r≥0.50 and 0.70 indicated moderate and intense correlation, respectively). P<0.05 was considered to indicate a statistically significant difference.

A total of 261,144 high quality reads were obtained from 27 samples using high-throughput sequencing of gut flora, with an average of 9,672 reads per sample. Based on the principle of 97% genetic similarity, 1,409 OTUs, with an average of 103 OTUs, were obtained per sample. No significant differences in Chaol abundance indices and Shannon diversity indices were observed between patients with colorectal cancer and healthy controls (P>0.05).

A total of 13 bacterial phyla were found in the stool samples of the colorectal cancer and healthy control groups, including Verrucomicrobia, Tenericutes, Synergistetes, Proteobacteria, Fusobacteria, Firmicutes, Cyanobacterium, Chlorobacteria, Bacteroidetes, Actinobacteria, Acidobacteria, Crenarchaeota and Euryarchaeota (Fig. 1). The composition of gut flora exhibited evident individual differences between the two groups; however, the relative abundance of bacterial phyla between the two groups was similar. The dominant bacterial phyla in samples from both groups were Firmicutes (44.0% in the colorectal cancer group vs. 40.9% in the healthy control group; P>0.05), Bacteroidetes (35.6% in the colorectal cancer group vs. 47.1% in the healthy control group; P>0.05) and Verrucomicrobia (7.9% in the colorectal cancer group vs. 3.7% in the healthy control group; P>0.05). Several stool samples from the colorectal cancer group contained Synergistetes; however, there was no significant difference observed between the two groups (P>0.05).

At the bacterial strain/genus level, there was a significant difference observed in 18 bacteria between the colorectal cancer group and the healthy controls. As presented in Table I, the proportions of Adlercreutzia, Anaerosporobacter, Megamonas, Hypersegal, Dialister, Faecalibacterium, Bacteroides, Prevotella, Pseudobutyrivibrio, Enterobacter, Roseburia, Ruminococcus and Dorea forsicigenerans significantly decreased (P<0.05) in the colorectal cancer group compared with the healthy control group, whereas the proportions of Citrobacter farmer, Fusobacterium, Akkermansia muciniphila, Peptostreptococcus and Streptococcus significantly increased (P<0.05).

As shown in Fig. 2, the metabolic profiling of the colorectal cancer group was distinct from that of the healthy control group (R2Y=0.990; QY2=0.914). The levels of short-chain fatty acid metabolites, including acetic acid, valeric acid, butyric acid and isovaleric acid, in the patients with colorectal cancer were significantly higher than those of the healthy control patients (P<0.0001, P=0.024, P<0.0001 and P=0.002, respectively), whereas the level of isobutyric acid was significantly lower in patients with colorectal cancer than in the healthy control individuals (P<0.0001; Fig. 3). No significant difference was observed in the levels of propionic acid between the two groups (P>0.05; Fig. 3). As shown in Table II, compared with the healthy control group, levels of nine amino acid metabolites in stool samples from patients with colorectal cancer increased by 44–82%, whereas three unsaturated fatty acids (oleic acid, elaidic acid and linoleic acid), two polyhydric alcohols (glycerin and monoacyl glycerol) and one saturated fatty acid (myristic acid) significantly decreased (P<0.05). Furthermore, ursodesoxycholic acid, a metabolic product of bile acid, and pantothenic acid, a metabolic product of vitamin B, significantly decreased (P<0.05) in colorectal cancer patients compared with the healthy controls.

Pearson rank correlation analysis demonstrated that there was a close correlation between gut flora and certain metabolic products (Fig. 4). The concentration of free fatty acids in stool samples was highly positively correlated with the levels of Bacteroides, Dialister and Pseudobutyrivibrio (r=0.87) and moderately correlated with the levels of Fusobacterium and Ruminococcus. The concentration of ursodesoxycholic acid in stool samples was highly positively correlated with the level of Ruminococcus (r=0.75). The concentrations of phenylalanine and glutamic acid were highly positively correlated with the levels of Phascolarctobacterium and Acidiminobacter (r=0.74). The concentrations of serine and threonine were moderately positively correlated with the levels of Phascolarctobacterium and Acidiminobacter (r=0.63).