Metabolomic Study On The Screening Of Esophageal Mucosal Lesions

Background:Esophageal cancer is defined as malignant lesions formed by the esophageal squamous epithelial dysplasia or glandular epithelial dysplasia.Esophageal cancer is the sixth most lethal diseases in the world.The high incidence and high mortality have aroused extensive attention.Esophageal squamous epithelial dysplasia is regarded as an important esophageal precancerous lesion,and the development of esophageal squamous epithelial dysplasia to cancer typically may need years or even more than ten years.This phase is the most precious time to the early detection and treatment of esophageal cancer.The early detection and intervention of mild and moderate dysplasia may promote the reverse of esophageal squamous epithelial dysplasia,thus the risk of the development of esophageal squamous epithelial dysplasia to cancer will be significantly reduced.Therefore,the early recognition and intervention of esophageal squamous epithelial dysplasia is of great significance to reduce its canceration rate and promote the prevention of esophageal cancer.Since 2008,China has screened for the precancerous condition or cases of early esophageal cancer in 40?69 year-old people in Shandong province Feicheng city,with the help of iodine staining indicative technique under gastroscope biopsy.But even in area with high incidence of esophageal cancer,the positive rate of esophageal mucosal lesions was low.As much as 90%of people had negative screening results,suffering from unnecessary distress in gastroscopy.Therefore,there is an urgent need to develop a new method to explore the changes in the internal environment in the development of esophageal cancer,to provide the scientific evidenceand medicine support for early screening before using the gold standard in esophageal cancer screening methods(iodine staining indicativetechnique under gastroscope biopsy),to distinguish between people with esophageal mucosal lesions and people without esophageal mucosal lesions,to narrow the scope of the gastroscope screening,to improve the positive rate of gastroscope screening,so as to avoid esophageal perfectly healthy people to bear the pain of invasive gastroscope,to make the early detection of esophageal cancer treated early more targeted.Although the past esophageal cancer metabolomics study had found the disturbance and exception in amino acid metabolism,energy metabolism,lipid metabolism and other multiple metabolic pathways.But most studies were traditional hospital-based case-control design.They only found the differences of metabolic profile between patients and healthy controls,which couldn't be used for early screening.Basing on "national demonstration base of early detection and early treatment for esophageal cancer",this project collected serum specimens of newly discovered subjects with esophageal mucosal lesions(including esophagitis,esophageal precancerous lesions and patients with early esophageal cancer)and subjects without esophageal mucosal lesions(health subjects without gastrointestinal lesions).Our study set up a metabolic profile study for early screening of esophageal cancer under the design of population-based study,so as to overcome the drawback of lacking the esophageal precancerous lesions and early cases in hospital-based case-control study.This study aims to establish a population-based metabonomics case-control study,with the help of Ultra-performance Liquid Chromatography detection platform and statistical pattern recognition and other statistical methods,finding metabonomics indicators for esophageal cancer screening in high-risk area,so as to narrow the scope of the gastroscope screening which only includes high-risk individuals.It is important to the implementation and promotion of the esophageal cancer screening.Methods:In this study,a nontargeted metabolomics approach based on the ultra performance liquid chromatography-mass spectrometry(UPLC-MS)was employed to determine global alterations in the metabolic profiles of the serum samples collected from participants.To test the validity of the model against overfitting,response permutation testing was computed in partial least squares discriminant analysis(PLS?DA)by permutation testing using 100 random permutations.Discriminating variables were selected according to variable importance in projection values and nonparametric test of single variable.Furthermore,the AUC value was used to determine whether different potential biomarkers were statistically significant between subjects with esophageal mucosal lesions and subjects without esophageal mucosal lesions.The original chromatography mass spectrometry data was checked to remove the obvious noise.R-package CAMERA was used to identify the isotope ions,adduct ions and fragment of the same metabolites.Thus,potential metabolic markers could be effectively found out.Fragment ions cracking rules were described according to the retention time,mass spectrometry(MS)and MS/MS experiments of metabolites.The structure of the underlying metabolic markers was identified by searching the free databases or confirmed by standard compounds.Samples can be divided into two parts:one part was used to build esophageal cancer screening model to distinguish subjects with esophageal mucosal lesions and subjects without esophageal mucosal lesions,the other part was used to verify the validity of the model.Results:To obtain a high quality data,we strictly abided by the standard operation procedure and totally collected serum samples from 641 participants under gastroscope,including 187 cases of subjects without esophageal mucosal lesions and 454 cases of subjects with esophageal mucosal lesions.The subjects with esophageal mucosal lesions consisted of 95 cases of esophageal inflammation(20.93%),184cases of mild atypical hyperplasia(40.53%),133 cases of moderate and severe dysplasia(29.30%),42 cases of carcinoma in situ and early screening squamous carcinoma(9.29%).A nontargeted metabolomics approach based on UPLC/QTOF/MS was employed in the metabolic profiles of the serum samples that had been disrupted the group order.Metabolomic data was acquired after pretreatment.In the positive ion detection mode,UPLC-MS spectral data generated 522 variables in the matrix,and 212 variables were found in the negative ion detection mode.Thus,we finally obtained 2 dimensions standard data format which can be used for statistical analysis.R-package CAMERA was used to obtain the isotope ions,adduct ions,fragment and other identity information of the metabolites.Metabolomic data of QC samples was analyzed by relative standard deviation and principal component analysis to monitor the quality control of the system.The distribution plot of relative standard deviation in each metabolite and principal component score plot of the QC samples demonstrated the good quality control in the process of experiment.The experimental data was reliable and stable.The result of principal component analysis found a preliminary classification trend between subjects with esophageal mucosal lesions and subjects without esophageal mucosal lesions,but there was a certain overlap between different groups.To maximize the separation of sample classes,a further supervised analysis using PLS-DA was performed,R2X=0.231,R2Y=0.749,Q2cum=0.638.The result showed the metabolic mode difference between subjects with esophageal mucosal lesions and subjects without esophageal mucosal lesions.There was obvious trend of classification between the two groups.The validation plot of response permutation testing proved that PLS-DA model had no risk of over fitting.According to variable importance in projection values,nonparametric test of single variable and the AUC value that was used to determine if different biomarker candidates were statistically significant between subjects with esophageal mucosal lesions and subjects without esophageal mucosal lesions,we finally found 22 potential metabolic markers.Combined with the retention time and fragment ions cracking rules from mass spectrum or secondary mass spectrometry experiments,the structure identification of the underlying metabolic markers was conducted the information from free databases and standard methods of identification.We determined the structure of three major metabolic markers,L-tryptophan,aloperine and Sphinganine-1-phosphate.The original data was randomly in proportion assigned two thirds as training set,one third as the testing set.Training set was used to construct the PLS-DA and random forest model to distinguish subjects with esophageal mucosal lesions and subjects without esophageal mucosal lesions.Constructed random forest model was used to predict the classification of testing set.The area under the ROC curve was 0.994.According to the PLS-DA 3 d chart of the testing set,there was an obvious trend of classification between subjects with esophageal mucosal lesions and subjects without esophageal mucosal lesions.The results indicated that the models builded by serum metabolic markers had good ability to distinguish subjects with esophageal mucosal lesions and subjects without esophageal mucosal lesions.The logistic regression model was established by using identified metabolic markers L-tryptophan,aloperine and Sphinganine-1-phosphate.According to the testing set,the ROC curve was draw by using real and estimated values acquired in the logistic regression model.The value of AUC was 0.835,the 95%confidence interval(0.776,0.776),showed that the model had good predictive ability.According to the identification results of potential metabolic marker and the roles they played in the metabolic pathway,we found amino acids and membrane phospholipids metabolic pathway disorders in subjects with esophageal mucosal lesions.Conclusion:The logistic regression model was established by using identified metabolic markers L-tryptophan,aloperine and Sphinganine-1-phosphate.The testing set showed that the model had good predictive ability.Metabolic markers in serum and corresponding metabolic pathways changed in subjects with esophageal mucosal lesions.There were metabolic disorders in amino acid and membrane phospholipids metabolic pathways.