Establishment Of Genomics-radiomics Model And Molecular Mechanisms Of Colorectal Cancer With Liver Metastasis

Objective:Colorectal cancer is one of the most common malignancies worldwide.In recent years,the mortality rate of colorectal cancer has been declining in many countries due to the rapid development of early disease screening and treatment.However,the prognosis of patients with advanced colorectal cancer is still not ideal.Liver is the most common site of colorectal cancer metastasis,and liver metastasis is one of the leading causes of death in patients with advanced colorectal cancer.Although the development of chemotherapy combined with targeted therapy and a multidisciplinary approach has significantly improved survival in patients with liver metastases,survival in patients with advanced colorectal cancer is still less than 3 years.Genes are key factors in the development of colorectal cancer.Studies have confirmed that colorectal cancer patients with different RAS(KRAS and NRAS)and BRAF gene status have different clinical and prognosis characteristics.But whether the radiomics features are different is still unclear,and the key molecular mechanism of liver metastasis of colorectal cancer has not been comprehensively concluded.Therefore,we aimed at the clinical problem of liver metastasis from colorectal cancer in this study.First,the artificial neural network(ANN)method was used to construct a genomics prediction model based on CT radiomics features,so as to achieve non-invasive prediction of RAS and BRAF gene mutation status in patients with advanced colorectal cancer liver metastasis.At the same time,the potential molecular mechanism of key genes in the development of liver metastasis from colorectal cancer was also discussed.Methods : 1.Patients with pathologically confirmed liver metastases from colorectal cancer who received systemic treatment between January 2014 and October 2019 were enrolled in a multi-center study.2.The liver metastasis lesions of patients were manually delineated and segmented layer by layer by 3D-Slicer software.3.Py Radiomics method was used to extract the radiomics features of liver metastasis lesions.4.Intraclass correlation coefficient(ICC)and concordance correlation coefficient(CCC)were used to screen the radiomics features with strong stability and reproducibility.5.CT images of patients were evaluated with two semantic features: 1)whether there were other metastatic lesions beyond the liver or regional lymph nodes;2)whether there were "microsatellites"(a single metastatic lesion with a large cross-sectional area was surrounded by multiple smaller lesions).6.Seven machine learning methods were used to construct prediction models for RAS and BRAF gene mutation status based on semantic features,radiomics features and combination.7.The prediction performance of different prediction models was evaluated according to the accuracy,specificity,sensitivity,AUCs and other indicators.The prediction model with the best performance was selected.Evaluate the predictive performance of the best model in multiple subgroups.8.Compare the predictive performance of the best model in patient cohorts from different hospitals.9.Single cell transcriptome sequencing data was used to screen the key genes of liver metastasis from colorectal cancer.10.Online public database was used to verify the differential expression of Polymeric Immunoglobulin Receptor(PIGR)11.The expression of PIGR in patients with colorectal cancer was detected by immunohistochemistry and Enzyme-linked Immunosorbent Assay method.12.Western blotting was used to detect the expression of PIGR in six colon cancer cell lines.13.Transwell and wound healing assesses the ability of migration in colon cancer cell lines.14.PIGR was overexpressed by c DNA plasmid or temporarily silenced by si RNA technology.15.Gene Set Enrichment Analysis(GSEA)was used to investigate PIGR related signaling pathways.16.Statistical analysis: Graph Pad Prism,Pyhton and R were used for statistical analysis.The data of molecular biology experiment was calculated by the mean ± standard deviation of three independent repeated experiments.The Student's T test was used to evaluate the differences between the two sets of data.P <0.05 was considered statistically significant.Results:1.One hundred and fifty-nine patients with colorectal cancer liver metastasis were enrolled in multiple centers.2.851 reproducible radionics features of liver metastases were extracted and screened using 3D-Slicer software and Py Radiomics method.3.The Radiomics and Semantic Features Model Based on Artificial Neural Network(RS-ANN)has the best prediction performance.According to the accuracy,specificity,sensitivity,AUCs and other indicators of each prediction model,the results indicate that the RS-ANN model has an AUC of 0.95 and an accuracy of 81.10% in the training set,and an AUC of 0.79 and an accuracy of 71.43% in the validation set.The prediction performance of RS-ANN model is better than that of all other models.4.In the subgroup analysis,the positive predictive value of RS-ANN model in patients with RAS or BRAF gene mutation reached 100%.5.The RS-ANN model has good consistency and extrapolation in the prediction performance of patients in different centers.6.Polymeric Immunoglobulin Receptor(PIGR)was identified as the key gene of liver metastasis in right colon cancer using single cell transcriptome sequencing data.7.In GEPIA,TCGA-COAD and GSE39582 datasets,the expression level of PIGR in colon cancer tissues was higher than that in normal colon tissues,and the expression level of PIGR in right colon cancer tissues was higher than that in left colon cancer tissues(P values<0.05).8.Immunohistochemical results of tissue samples from 44 patients after surgery indicated that the expression level of PIGR in colon cancer tissue was higher than that in normal colon tissue,and the expression level of PIGR in right colon cancer tissue was higher than that in left colon cancer tissue.ELISA results of PIGR in plasma of 78 patients indicated that the expression level of PIGR in plasma of right colorectal cancer patients was higher than that of left colorectal cancer patients.9.Western blotting was used to detect the expression level of PIGR in six colon cancer cell lines,and the results showed that the expression level in RKO cell lines was the highest,while the expression level in HCT15 cell lines was the lowest.Therefore,si RNA was subsequently used to conduct gene silencing in RKO cells,and overexpressed plasmid was used to conduct gene overexpression in HCT15 cells.10.PIGR affects the epithelial mesenchymal transition(EMT)in colon cancer cell lines.Western blotting was used to detect EMT-related proteins after PIGR overexpression induced by c DNA plasmid or transient silencing induced by si RNA technology.It was found that the expression of ZEB1 and N-Ca were up-regulated in HCT15 cell lines after PIGR overexpression.The expression of ZEB2 and N-Ca was down-regulated after transient PIGR silencing in RKO cell lines.11.PIGR affects the migration ability of colon cancer cells.Transwell results showed that compared with NC control group,the migration ability of HCT15 cells was significantly enhanced after overexpression of PIGR.After the transient silencing of PIGR in RKO cell lines,the migration ability of RKO cells was significantly reduced.12.The results of Gene Set Enrichment Analysis(GSEA)suggested that Wnt/?-catenin pathway activity was significantly down-regulated in patients with lower level of PIGR expression in right colon cancer.13.The expression level of PIGR affects the activity of Wnt/?-catenin pathway.Western blotting was used to detect proteins related to the Wnt/?-catenin pathway.Compared with NC control group,the activity of Wnt-?-catenin pathway was increased after overexpression of PIGR in HCT15 cell lines.The activity of Wnt-?-catenin pathway decreased after transient silencing of PIGR in RKO cell lines.Conclusions:1.The RS-ANN model constructed with artificial neural network method based on the radiomics and semantic features of CECT can accurately predict the mutation status of RAS and BRAF genes in patients with colorectal cancer liver metastasis.2.PIGR is a key gene in the occurrence and development of liver metastasis of right colon cancer.3.PIGR could promote EMT of colon cancer cells by regulating Wnt/?-catenin pathway,as well as promoting the migration ability of cancer cells.These might promote liver metastasis of colon cancer.