Combining Targeted And Untargeted Metabolomics To Research Metabolic Characteristics Of Papillary Thyroid Carcinoma

BackgroundPapillary thyroid carcinoma (PTC) is the most common pathological type of thyroid cancer. The incidence of thyroid carcinoma is increasing globally. Papillary thyroid carcinoma (PTC) is the most common subtype of malignant thyroid nodule, and a significant increase in the number of thyroid cancer cases has been attributed to the increase in the incidence of PTC. Research has demonstrated that surgical therapy of thyroid cancer at an early stage yields a better prognosis (lower recurrence and mortality) in both children and adult patients. Ultrasound (US) and US-guided fine needle aspiration biopsy (FNAB) are commonly used methods for diagnosing thyroid cancer, but the negative predictive value of these methods for microcancers is only 72.2%. The disadvantage of using US to diagnose a malignant node is its simple reliance on individual US features. Therefore, biomarkers should be assessed in combination with FNAB samples before surgery to decrease the rate of unnecessary thyroidectomies.Metabolomics involves the simultaneous quantitative analysis of a comprehensive profile of metabolites in living systems. Nowadays, metabolomics is a well-recognized metabolic system-wide approach that characterizes small molecules found in urine, plasma or tissues. Therefor, it could serve as a new platform to extensively screen for tumor biomarkers and to further study the association of cancer pathogenesis with metabolic disorders. Targeted and untargeted metabolomics are two main mass spectrometry techniques for metabolite research and each of them has its advantages and disadvantages. The untargeted metabolomics is good at comprehensively screening for biomarkers but has disadvantages in precision. Fortunately, targeted metabolomics specializes in measuring a set of known metabolites with high sensitivity and good reproducibility, which could fill the gap of the untargeted experiment. Research is advancing through a great diversity of technologies, including gas chromatography-mass spectrometry (GC-MS), liquid chromatography-mass spectrometry (LC-MS), and nuclear magnetic resonance (NMR).ObjectivesThe aim of our study was to construct a tissue-targeted metabolomic analysis method based on untargeted and targeted metabolic multiplatforms to identify a comprehensive PTC metabolic network in clinical samples. Firstly, we tried to find out the specific methods for analyzing PTC and normal tissues based on GC-TOF-MS. Secondly, based on untargeted metabolomics, we aimed to preliminarily analyze the samples, to evaluate the grouping effect of this platform and to find differentially abundant metabolites. Thirdly, in order to accurately qualify and absolutely quantify target metabolites, we explored measuring methods for detecting galactinol, melibiose and other biomarkers based on UHPLC-QqQ-MS and GC-TOF-MS platforms and verified them. Thus, our study focused on using targeted and untargeted mass spectrometric techniques, which are highly efficient methods that combine multiparametric analysis methods, to identify candidate molecular biomarkers of malignant disease among the components of human metabolic pathways to improve the accuracy of FNAB-based diagnosis. So it will make the PTC patient to accept early diagnosis and early treatment, meanwhile it will avoid unnecessary invasive treatment and excessive patient anxiety. In addition, we attempted to explain the potential metabolism related molecular mechanisms responsible for PTC occurrence. This comprehensive analysis provides new insights into oncogenesis and thyroid cancer treatment.Materials and methods1. Clinical population and sample collection. Surgically resected thyroid tissue was collected from patients undergoing a thyroid operation. Finally 50 tissue samples, including 25 cancer and 25 healthy thyroid samples, were collected from these patients.2. Untargeted GC-TOF-MS sample processing and analyses. Totally 30 samples (15 PTC tissue and 15 normal thyroid tissue) were collected for GC-TOF-MS analysis. After being homogenized, centrifuged and extracted, almost all metabolites were separated and were added internal standard mixture for GC analysis. Various metabolites were isolated in capillary column of GC and then were introduced into the ion source of TOF-MS for detecting mass and RT parameters. The Chroma TOF4.3X software (LECO) combined with the LECO-Fiehn Rtx5 database was utilized to analyze the original data.3. Untargeted GC-TOF-MS statistical analysis. Further, to simulate the missing raw data values, a method of half-minimum numerical simulation was applied. Next, an internal standard normalization method was used to standardize the data, and an interquartile range for filtering the data was used to remove noise. Thereafter, the normalized data were imported into the SIMCA-P+13.0 software package for multivariate analysis of pattern recognition, including principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA) and orthogonal partial least squares discriminant analysis (OPLS-DA). To estimate the robustness and the predictive ability of our model, a 7-fold cross-validation method was used based on the PLS-DA model. Based on OPLS-DA, differentially abundant metabolites were selected. The criteria for selecting differentially abundant compounds were a variable importance in the projection (VIP) value greater than 1, and a P value of student’s T test less than 0.05. In addition, the differentially abundant metabolites were cross-referenced to the pathways by further searching commercial databases, KEGG (http://www.genome.jp/kegg/).4. Targeted UHPLC-QqQ-MS processing and statistical analysis. Totally 20 samples (10 PTC tissue and 10 normal thyroid tissue) were prepared for UHPLC analysis. For individual mother solutions (1 mg/ml), each analytical standard was dissolved in methanol or water in accordance with its solubility and was then diluted in methanol into a series of 10 ng/μl to 100 ng/μl. Next, to select the MRM mode (positive or negative), the declustering potential (DP), collision energy (CE) and collision chamber export pressor (CXP) parameters were optimized through filling in precursor and product ion pairs. After pretreatment, samples were introduced in UHPLC system and then in QqQ-MS for metabolites analysis. Standard curves were drawn based on testing of a series of standard solutions and then the quantification of biomarkers was conducted. Student’s t-test (with p< 0.05) was applied to discern the differences in tissue metabolite concentrations. Besides, the ROC curve, gotten from Daim Package of R Programming Language, was also used to assess the potential diagnostic value of each metabolite and combinatorial biomarker by calculating their specificity and sensitivity in classification.5. Targeted GC-TOF-MS processing and statistical analysis. All analytical standards were dissolved in methanol or water in accordance with its solubility and were then diluted in methanol into a series of 10 ng/μl to 100 ng/μl. Totally 20 samples (10 PTC tissue and 10 normal thyroid tissue) as same as which were analyzed based on UHPLC-QqQ-MS. After pretreatment (as same as process of untargeted GC-TOF-MS), all samples were introduced in GC and combined TOF-MS. The quantification results of potential biomarkers were calculated based on standard curves. And statistical analysis is same with UHPLC-QqQ-MS analysis.Results1. Total ion chromatograms (TICs) for all samples were gotten from GC-TOF-MS, and no drift was observed in any of the peaks displaying a stable retention time. In total,761 ion peaks were identified and 686 metabolites remained after removing noise based on the interquartile ranges.2. Statistic pattern recognition model constructed for differentiation of cancer tissues and normal tissues based on GC-TOF-MS data. With Initially, PCA was performed using two principal components:R2X=0.427 and Q2= 0.159. The PCA score plot demonstrates as a whole that each sample can be clearly divided and no abnormal sample should be rejected. Certain groups that could not be clearly distinguished were further examined through subsequent supervised discriminating analyses. Subsequently, the parameters of the PLS-DA score scatter plot were R2X =0.225 and R2Y=0.944; these results reflect the data stability and the good fitness of model parameters. A 7-fold cross-validation permutation test was applied and this test results excluded the random effects in the constructed model. A loading plot was constructed based on OPLS-DA using one predictive and one orthogonal component (R2X= 0.225, R2Y= 0.944, Q2Y= 0.687). Using this model, the clearest separation was produced between the normal tissue (NT) and thyroid cancer (TC). These values indicate that a majority of the variation in the statistical data are attributable to the separation between the healthy and cancer groups.3. Totally 45 differentially abundant metabolites with a variable importance in the projection (VIP) value greater than 1 and a P value less than 0.05 were identified based on OPLS-DA model. Metabolites exhibited higher concentrations in the cancer tissue than in the healthy tissue including alpha-aminoadipic acid (FC=20.05), a-linolenic acid (FC=4.176) and so on. The pathways that matched based on the KEGG database contained galactose metabolism which included the largest number of significantly altered metabolites, including glucose, sorbitol, galactinol, and melibiose. Furthermore, lysine degradation, neuroactive ligand-receptor interactions, pyrimidine metabolism, biosynthesis of unsaturated fatty acids and a two-component system were analyzed further. The metabolites involved in these pathways include OA, noradrenaline, arachidic acid, a-AA, glutaric acid, a-linolenic acid, uridine 1, 5,6-dihydrouracil 1, and melatonin 2. Combined with the 4 metabolites in the galactose metabolism pathway,13 different metabolites were selected for subsequent validation experiments.4. Based on targeted metabolomics,11 metabolites were verified and 3 biomarkers were proposed with statistical significance. A linear standard curve was generated for the mixed standard solutions and except for glucose, glutaric acid and noradrenaline, the metabolites displayed R2 values of linearity greater than 0.98.11 metabolites out of the 13 selected biomarkers were detected with excellent sensitivity and reproducibility, except for noradrenaline and 5,6-dihydrouracil. T-test was performed and only 3 of the 11 verified metabolites, galactinol, melibiose and melatonin, were validated as significantly altered biomarkers (p<0.05) for further study in PTC auxiliary diagnoses and in supporting clues for finding therapeutic targets.5. Galactinol, melibiose and melatonin as combinatorial biomarkers for PTC diagnosis with the higher sensitivity and specificity. Screening for potential biomarkers for clinical diagnoses is a significant task of metabolomics, the application of ROC curves with the area under the ROC curve (AUC) value for evaluation of diagnostic accuracy of biomarkers were conducted. Our results show that the AUC of these analytes, except galactinol, was less than 0.9 based on the untargeted and targeted metabolomics data. As the combinatorial biomarker has been applied in previous studies, the sensitivity and specificity of the combinatorial biomarkers for PTC diagnosis could be evaluated with ROC curve. Based on data from untargeted metabolomics, the combination of biomarkers comprising galactinol, melibiose and melatonin displayed an AUC value of 0.95, while this diagnosis panel achieved 93.3% sensitivity and 100% specificity for the prediction of PTC. Next, to validate the auxiliary diagnostic application using the combination of metabolites, the concentrations of galactinol, melibiose and melatonin determined based on targeted GC-TOF-MS were analyzed using the ROC curve. The validation results show that the AUC of the three combined biomarkers was 0.96, confirming the 90.0% sensitivity and 100% specificity were satisfactory. Therefore, the potential combinatorial biomarker provided a new orientation for exploring auxiliary ways of PTC diagnosis.6. Galactose metabolism participates in PTC development. To identify the pathway most relevant to PTC development, the metabolomics data were analyzed using MetaboAnalyst 2.0 (http://www.metaboanalyst.ca/MetaboAnalyst/). The analysis result show that the impact value of the galactose metabolic pathway is 0.1, which indicates that this altered pathway obviously affects PTC carcinogenesis. Moreover, based on the galactose metabolism map in the KEGG database, a critical enzyme, the alpha-galactosidase (GLA) plays significant role in this pathway. Based on our data, galactose is consistently converted into mannose, sorbitol, galactinol, melibiose and glucose via GLA enzymolysis. Coincidentally, the concentrations of these five metabolites were lower in malignant thyroid tissue than in normal tissue. Thus, we hypothesize that the level of GLA activity is lower in PTC patients or that thyroid cancer tissues express deficient GLA, which leads to a decrease in GLA-related metabolites. And GLA may will serve as a target for therapy in future study.Conclusions1. First of all, the integrated application of targeted and untargeted mass spectrometry techniques for metabolic analysis of tissue enhanced the accuracy and reproducibility of the results. To the best of our knowledge, validation experiments using targeted metabolomics techniques to search for diagnostic biomarkers of PTC have not been conducted in previous related studies.2. And then, we creatively detected analysis methods of PTC tissues based on untargeted GC-TOF-MS and targeted GC-TOF-MS, UHPLC-QqQ-MS platforms. And we proved that GC-TOF-MS was an effective method for discrimination of PTC and normal tissues.3. Further, as metabolic profiling is susceptible to various factors, focusing on a single pathological cancer type, PTC, had good effects on the results, because disturbance from the metabolic changes caused by different thyroid cancer subtypes were avoided. Tissue-targeted metabolomics focuses on tissue-specific profiles of the full complement of metabolites in tissue homogenates.4. Thus, it’s difficult to provide sufficiently high diagnostic value relying on one metabolic biomarker and a combination of more analytes could comprehensively reflect the pathologic status of PTC. We verified galactinol, melibiose and melatonin with statistical significance in our targeted metabolomics, so we proposed them as combinatorial biomarkers for PTC diagnosis.5. That identifying the pathways involved in PTC reveals new factors associated with PTC pathogenesis. Based on the metabolite enrichment analysis results, the galactose metabolism pathway was regarded as an important factor influencing PTC development by affecting energy. metabolism. Alpha-Galactosidase (GLA) was considered to be a potential target for PTC therapy.