Colorectal cancer (CRC) is the third most commonly diagnosed cancer and the second leading cause of cancer mortality worldwide.1Bray F. et al.CA Cancer J Clin. 2018; 68: 394-424Crossref PubMed Scopus (50539) Google Scholar Epidemiologic studies2Esposito K. et al.Diabetes Care. 2012; 35: 2402-2411Crossref PubMed Scopus (711) Google Scholar, 3Rampal S. et al.Gastroenterology. 2014; 147: 78-87Abstract Full Text Full Text PDF PubMed Scopus (32) Google Scholar have suggested that metabolic syndrome (MS) increases the risk of CRC. Early screening of CRC effectively increased the 5-year survival rate4Allemani C. et al.Lancet. 2015; 385: 977-1010Abstract Full Text Full Text PDF PubMed Scopus (421) Google Scholar and was of importance for the MS population at high risk of developing CRC.5Choi Y.J. et al.Eur J Epidemiol. 2018; 33: 1077-1085Crossref PubMed Scopus (34) Google Scholar Low sensitivity and specificity limit various biomarkers in the early screening of CRC.6Lauby-Secretan B. et al.N Engl J Med. 2018; 378: 1734-1740Crossref PubMed Scopus (140) Google Scholar Metabolomics is an emerging tool in systems biology for the discovery of small-molecule metabolic markers. This study aimed to define reliable serum biomarkers for diagnosis of CRC. The study participants were enrolled based on strict inclusion and exclusion criteria (see Supplementary Methods). In the screening set, 360 serum samples were collected from Taizhou Hospital (Zhejiang, China) and community-based study in the Xiaoshan district (Zhejiang, China). The validation set included 1594 participants enrolled from Zhejiang Cancer Hospital, a community-based study in the Xiaoshan district, and Taizhou Hospital. A third independent cohort comprised 1528 participants enrolled as a prediction set from Zhejiang Cancer Hospital and a community-based study in the Putuo district (Zhejiang, China). Untargeted analyses were performed on an Agilent 1290 liquid chromatographic (Agilent Technologies, Santa Clara, CA) and an Agilent 6545 Quadrupole time-of-flight system . Targeted analyses were carried out on a triple quadrupole coupled with a liquid chromatographic (Nexera x2)–mass spectrometry (TQ8050) system (Shimadzu, Kyoto, Japan). All statistical analyses were conducted using R software, version 3.4.1 (R Core Team, Vienna, Austria). Details are available online as Supplementary Methods. A total of 3482 participants, consisting of healthy volunteers, patients with MS, and patients with CRC from 4 independent centers were recruited (Figure 1A). The clinical characteristics of the participants are presented in Supplementary Table 1. The screening set included 120 healthy volunteers, 120 patients with MS, and 120 age- and sex-matched patients with CRC. The principal component analysis of this phase showed obvious differences among samples from healthy, MS, and CRC participants (Figure 1B). Subsequently, the orthogonal partial least squares discriminant analysis shows apparent separations of healthy from MS participants and healthy from CRC participants, and a significant distinction between MS and CRC participants (Supplementary Figure 1A). The representative total ion chromatograms for healthy and CRC participants are shown in Supplementary Figure 1B. A total of 1793 features were detected in both both positive and negative electrospray ionizaiton (ESI+ and ESI–) modes. The differential metabolites that satisfied the criterion of variable importance in the projection of >1.0 and false discovery rate of <0.05 were considered as biomarker candidates. Hence, 61, 37, and 59 differential metabolites were identified from the comparisons of healthy vs CRC, healthy vs MS, and MS vs CRC participants, respectively. Importantly, 30 differential metabolites were significantly altered in the 3 comparisons (Figure 1C). The relative levels of these 30 differential metabolites are presented as a heatmap (Figure 1D). Perturbed metabolic pathways were mainly related to phenylalanine metabolism, glycerophospholipid metabolism, phenylalanine, tyrosine, and tryptophan biosynthesis (Supplementary Figure 1C). The validation set enrolled 1594 participants comprising 580 healthy volunteers, 577 MS patients, and 437 CRC patients. The outcome of principal component analysis and orthogonal partial least squares discriminant analysis are presented in Supplementary Figure 1D and E. Ultimately, 7 biomarker candidates—glutamine-leucine [Glu-Leu], tryptophan, PE(18:2/P-18:0), coumaric acid, tyrosine, isoleucine, and phenylalanine—were confirmed in the validation set and selected as potential biomarkers. The heatmap shows the concentration changes of 7 potential biomarkers in each group (Figure 1E). A correlation analysis between clinical characteristics and 7 potential biomarkers is shown in Supplementary Figure 1F. Finally, a biomarker panel of tyrosine and Glu-Leu, capable of discriminating CRC participants from those without CRC, was selected by binary logistic regression. Subsequently, a convenient and rapid isotope-labeled ultra-performance liquid chromatography coupled with a triple quadrupole mass spectrometry method was established to accurately quantify the levels of tyrosine and Glu-Leu in serum. In total, 900 serum samples (from 300 healthy volunteers, 300 MS patients, and 300 CRC patients) were selected randomly for analysis from the validation cohort. The area under the curve and specificity and sensitivity in receiver operating characteristic curves were used to evaluate the performance of tyrosine and Glu-Leu in the comparisons of healthy vs CRC, healthy vs MS, and MS vs CRC participants (Figure 1F). An independent cohort of 1528 serum samples was further used to evaluate the diagnostic performance of the biomarker panel (tyrosine and Glu-Leu) in the prediction phase. As shown in Figure 1G, the average concentrations of tyrosine and Glu-Leu in the model application were reproducible compared with those in the model establishment set. Predictive values were 95.24% and 92.38%, respectively, for the healthy vs CRC groups, 68.65% and 69.34%, respectively, for the healthy vs MS groups, and 97.53% and 92.94%, respectively, for the MS vs CRC groups (Figure 1H). In this study, a large multicenter cohort of 3482 participants was enrolled that included healthy volunteers (n = 1204) and MS (n = 1183) and CRC patients (n = 1095) from 4 independent centers. Thirty metabolites were altered as differential metabolites through untargeted metabolomics, 7 of which were validated in multiple centers. Based on logistic regression analysis, a simplified panel of Glu-Leu and tyrosine was selected for targeted metabolomics. Tyrosine is associated with cancer-related alterations of the trichloroacetic acid cycle.7Leichtle A.B. et al.Metabolomics. 2012; 8: 643-653Crossref PubMed Scopus (102) Google Scholar The decreased level of tyrosine observed in this study may be due to the metabolic disturbance resulting from colorectal tumor development. Glu-Leu, a dipeptide, is a byproduct of reduced glutathione synthesis catalyzed by γ-glutamylcysteine synthetase. The γ-glutamyl dipeptides play roles in the development of liver-related diseases.8Soga T. et al.J Hepatol. 2011; 55: 896-905Abstract Full Text Full Text PDF PubMed Scopus (189) Google Scholar The increased serum concentration of Glu-Leu may be reflective of the oxidative stress and inflammatory conditions in MS patients. This biomarker panel showed a satisfactory diagnostic performance with regard to the receiver operating characteristic curves. The predictive values were higher than 92.38% in discriminating CRC participants from those without CRC in the prediction set. These results suggest that the combination of Glu-Leu and tyrosine in serum is potentially a novel biomarker panel for early diagnosis of CRC. The limitation of this study lies in the possible occurrence of diverse phenotypes, particularly in the control cohorts—a situation that could result in a lot of mixed signals. However, the participants of the control and case groups were selected based on matched age and sex in the screening set, thereby decreasing these effects. Also, the false positive rate was controlled by multiple testing adjustment and validated by independent samples. The authors thank Dr Xiaodong Wen from the Cellular and Molecular Biology Center of China Pharmaceutical University for laboratory support, Lei Zhang and Siyi Chen from China Pharmaceutical University, and Xiaohui Sun from Zhejiang University for preparing serum samples. Special thanks go to Majie Wang for assisting with statistical analysis and Dr. Raphael N. Alolga for editing the manuscript. The authors also appreciate the scholarly comments of Dr. Fengguo Xu, China Pharmaceutical University. Author contributions: Maode Lai, Yimin Zhu, and Lian-Wen Qi conceived the original idea and designed the study. Jiankang Li and Huan Wang contributed to the experiments. Jiankang Li and Jing Li conducted data analysis and data interpretation. Jiankang Li prepared the draft of the study. Maode Lai, Yimin Zhu, and Lian-Wen Qi revised the manuscript. All authors agreed with the conclusion and approved the final version of manuscript. The physical examination and laboratory test results of healthy volunteers were normal; MS patients were enrolled according to the criteria of the Chinese Diabetes Society. Metabolic abnormalities included: (1) overweight or obesity (body mass index ≥ 25 kg/m2), (2) fasting plasma glucose ≥ 6.1 mmol/L and/or 2-hour fasting plasma glucose ≥ 7.8 mmol/L and/or diagnosis of type 2 diabetes, (3) systolic blood pressure/diastolic blood pressure ≥ 140/90 mm Hg and/or diagnosis of hypertension, (4) triglyceride levels (≥1.7 mmol/L) and/or high-density lipoprotein cholesterol (men, <0.9 mmol/L; women, <1.0 mmol/L). A diagnosis of MS was made in participants with 3 or more metabolic abnormalities; CRC patients were undergoing CRC surgery, and all of the cases had been confirmed and evaluated after colonoscopy. Pathologic diagnoses were determined by pathologists via biopsy reports. The informed consent statements and clinical protocol were reviewed and approved by the institutional review board at Zhejiang University (Zhejiang, China) and the committee on human rights related to human experimentation of each cooperating hospital. The serum samples of participants were collected without anticoagulant from each overnight-fasted individual. The serum samples were immediately stored at –80°C until metabolomic analysis. In this study, the ultra-performance liquid chromatography–quadrupole time-of-flight mass spectrometry–based analysis was used to find the differential metabolites in the discovery set and ascertain the stable metabolites as potential biomarkers in the validation set. For untargeted extraction, a 150-μL aliquot of methanol was added to 50-μL serum samples in a 1.5-mL tube, and the mixture was vortexed for 30 seconds and then centrifuged at 13,000g for 10 minutes at 4°C. A 150-μL aliquot of the supernatant was then transferred into another 1.5-mL tube. Two 75-μL aliquots of the supernatant were dried under a gentle stream of nitrogen at room temperature and reconstituted with 100 μL methanol solution containing 100 ng/mL l-2-chlorophenylalanine or 1 μg/mL ketoprofen as the internal standard for analysis in the positive and negative electrospray ionization (ESI+ and ESI−) modes, respectively. The resultant solution was then vortexed for 30 seconds and centrifuged at 13,000g for 10 minutes at 4°C, and 80 μL of the supernatant was transferred into the sample vial. A 1-μL aliquot of this was injected into the ultra-performance liquid chromatography–quadrupole time-of-flight mass spectrometry setup for untargeted analysis. The analytical column used was a Waters BEH C8 (100 × 2.1–mm inner diameter, 1.7-μm particle size) (Waters Corporation, Wexford, Ireland) column. All data were acquired with Agilent 6545 MassHunter Workstation software, version B.06.00 (Agilent Technologies). After the acquisition by MassHunter Workstation software, data were exported to the .mzdata format and then subjected to data preprocessing with the XCMS package in R, version 3.4.1. The differential metabolites were tentatively identified by database matching, that is, MassHunter METLIN Metabolite PCDL (Agilent Technologies) and Human Metabolome Database. Subsequently, some of the differential metabolites were confirmed by using available reference compounds. The mobile phase consisted of acetonitrile and water (both containing 0.1% formic acid) for the positive ion mode and methanol and water (both containing 5 mmol/L ammonium acetate) for the negative ion mode. For the 2 ESI modes, solvent A was the aqueous phase and solvent B was the organic phase, under the gradient elution program as follows: 0–1 minutes, 5% B; 1–4 minutes, 5%–30% B; 4–9 minutes, 30%–90% B; 9–10 minutes, 90%–100% B; 10–12 minutes, 100% B. The running time was 12 minutes, and the posttime was 3 minutes. The flow rate was 0.4 mL/minute, the injected volume was 1 μL, and the column oven temperature was set at 50°C. The fragmental voltage was set at 100 V, nebulizer gas at 35 lb/in2 gauge, capillary voltage at 3500 V, drying gas flow rate at 10 L/min, and temperature at 300°C. Reference masses at m/z 112.9855 and 980.0163 were introduced for accurate mass calibration. The essence of the quality control samples was to correct the analytical blocks and calculate technical precision. The serum samples were assigned a random number by the study administrator. During analyses of the sample sequence, 5 blank samples were injected first to ensure a stable baseline, and 1 quality control and 1 blank sample were run after every 10 injections of the prepared sera. After complete run, the data were acquired by MassHunter Workstation software, exported to the .mzdata format, and then preprocessed (nonlinear retention time alignment, peak discrimination, filtering, alignment, matching, and identification) using the XCMS package (Scripps Center for Metabolomics and Mass Spectrometry, La Jolla, CA). The ion features that were present in fewer than 80% of the samples were screened out. The intensities of each peak detected were generated by virtue of the retention times and the m/z data pairs for each ion. A targeted analysis was used to quantitatively determine the content of tyrosine and Glu-Leu in the test set. Analyses were performed on a triple quadrupole liquid chromatographic (Nexera x2)-mass spectrometry (TQ8050) system (Shimadzu, Kyoto, Japan). The analytical column used was a Waters BEH Amide (50 × 2.1–mm inner diameter, 1.7-μm particle size) column (Waters Corporation). All data were acquired with LabSolutions, version 5.86 SP1 (Shimadzu, Kyoto, Japan). The gradient elution program was 95%–80% B at 0.0–2.0 minutes, 80%–50% at 2.0–2.5 minutes, 50% at 2.5–3.5 minutes, 50%–90% at 3.5–4.0 minutes, 95% at 4.0–6.0 minutes, and back to initial conditions. A 1-μL sample was injected into the system with the autosampler conditioned at 4°C and the column temperature maintained at 40°C. The ion source parameters were set as follows: nebulizing gas flow at 3 L/minute; desolvation line temperature at 250°C; heat block temperature at 400°C; interface temperature at 300°C; drying flow at 10 L/minute, and heating gas flow at 10 L/minute. For targeted extraction, methanol/acetonitrile (3:1) was chosen as the optimal extraction solvent and best solvent for redissolution compared with methanol, acetonitrile, 50% acetonitrile, methanol/acetonitrile (1:1), and methanol/acetonitrile (1:3). Other extraction conditions were the same as the untargeted extraction. The quantitation of tyrosine and Glu-Leu in the test cohort was performed with reference to their corresponding isotope-labeled internal standards (105.0 ng/mL tyrosine-C13 and 20.4 ng/mL Glu-Leu -6C13, N15).Supplementary Table 1Clinical Characteristics of the Participants in Each SetCharacteristicsDiscovery set (n = 360)Validation set (n = 1594)P valueTest set (n = 1528)P valueHealthyMSCRCP valueHealthyMSCRCHealthyMSCRCn120120120580577437504486538Age, y61.7 ± 12.263.1 ± 13.063.1 ± 13.1>.05aThe differences were measured by continuous variables among 3 groups using the Kruskal-Wallis test.58.1 ± 11.459.5 ± 9.659.1 ± 12.5>.05aThe differences were measured by continuous variables among 3 groups using the Kruskal-Wallis test.68.9 ± 11.668.4 ± 11.162.9 ± 11.6>.05aThe differences were measured by continuous variables among 3 groups using the Kruskal-Wallis test.Sex, male/female73/4773/4773/47>.05bThe differences were measured by discrete variables among 3 groups using the chi-square test.298/282286/289252/185<.05bThe differences were measured by discrete variables among 3 groups using the chi-square test.250/254205/281281/257<.01bThe differences were measured by discrete variables among 3 groups using the chi-square test.BMI, kg/m221.3 ± 1.926.5 ± 2.3<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.21.6 ± 1.927.1 ± 2.7<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.22.6 ± 3.126.1 ± 3.4<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.Waist/hip ratio (male)0.87 ± 0.050.94 ± 0.05<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.0.84 ± 0.050.92 ± 0.05<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.0.86 ± 0.050.93 ± 0.05<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.Waist/hip ratio (female)0.83 ± 0.050.89 ± 0.06<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.0.81 ± 0.040.87 ± 0.05<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.0.86 ± 0.050.88 ± 0.06<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.SBP, mm Hg121.9 ± 11.9161.5 ± 19.1<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.121.0 ± 11.3158.2 ± 17.0<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.128.3 ± 26.7140.7 ± 19.7<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.DBP, mm Hg73.6 ± 8.3492.3 ± 11.2<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.73.5 ± 8.1092.6 ± 9.7<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.77.3 ± 9.9484.6 ± 11.1<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.TC4.25 ± 0.775.14 ± 1.05<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.4.41 ± 0.795.02 ± 0.90<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.4.83 ± 0.855.09 ± 2.68<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.TG1.18 ± 0.313.08 ± 2.40<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.1.17 ± 0.303.06 ± 1.90<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.0.96 ± 0.451.82 ± 1.04<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.Ua302.4 ± 64.7353.0 ± 88.2<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.299.5 ± 74.5347.5 ± 89.5<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.303.4 ± 62.7363.0 ± 85.2<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.Glu4.90 ± 0.466.39 ± 1.97<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.4.87 ± 0.466.07 ± 1.61<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.5.76 ± 1.456.41 ± 2.39<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.lnhoma–0.59 ± 0.540.34 ± 0.58<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.–0.54 ± 0.520.35 ± 0.55<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.–0.55 ± 0.510.36 ± 0.52<.01cThe differences were measured by continuous variables between 2 groups using the Mann-Whitney U test.Stage I117692 II35126144 III62153214 IV128288NOTE. Data are presented as mean as mean ± standard deviation. All P values were adjusted by the Benjamini-Hochberg false discovery rate correction, and an adjusted P value of <.05 was considered statistically significant.BMI, body mass index; DBP, diastolic blood pressure; Glu, glucose; SBP, systolic blood pressure; TC, total cholesterol; TG, triglyceride; Ua, uric acid.a The differences were measured by continuous variables among 3 groups using the Kruskal-Wallis test.b The differences were measured by discrete variables among 3 groups using the chi-square test.c The differences were measured by continuous variables between 2 groups using the Mann-Whitney U test. Open table in a new tab NOTE. Data are presented as mean as mean ± standard deviation. All P values were adjusted by the Benjamini-Hochberg false discovery rate correction, and an adjusted P value of <.05 was considered statistically significant. BMI, body mass index; DBP, diastolic blood pressure; Glu, glucose; SBP, systolic blood pressure; TC, total cholesterol; TG, triglyceride; Ua, uric acid.