Establishment Of Comprehensive Prognostic Models For Breast Cancer Based On LncRNA Expression Profiles

BackgroundBreast cancer is the most common malignancies threatening to the health of female worldwide.Its highly heterogeneity brings wide variability in biological phenotypic,treatment response and even prognosis.Human epidermal growth factor receptor 2(HER2)-positive breast cancer is a specific subtype of breast cancer with higher recurrence rate and mortality.Currently,long non-coding RNA(lncRNA)has been a novel class of RNA which played an important role in the development of breast cancer.In this study,on the basis of The Cancer Genome Atlas(TCGA)database,we aimed at mining the prognostic information from genomes and constructing a IncRNA risk-score model for breast cancer patients.We further combined the IncRNA-based model with clinicopathological factors and developed a prognostic nomogram which could accurately predict the individual survival of breast cancer patients.In addition,for HER2-positive breast cancer patients,we tried to construct a IncRNAs risk-score model that could predict the prognostic risk of patients.This study will provide a comprehensive prognostic evaluation system for the clinical management of breast cancer.Part One:Construction of a prognostic nomogram for breastcancer and function study of LHX1-DT in breast cancerObjective:To establish a prognostic nomogram for predicting the 3-year and 5-year survival probability of breast cancer patients and to explore the function role of the key prognostic lncRNA in the progression of breast cancer.Methods:1.The RNA expression profiles and clinicopathological data of 1109 breast cancer patients and 113 normal control were downloaded from TCGA database."DESeq2" package of R was used to identify differently expressed lncRNAs(DELs)between breast cancer tissues and normal tissues.Based on univariate Cox regression analysis and Kaplan-Meier curves,overall survival(OS)-related DELs were screened as candidate prognostic lncRNAs.2.In the training dataset,multivariate Cox regression analysis(forward stepwise)was then used to select the best-fit prognostic lncRNAs and construct a lncRNA risk-score model.The prognostic value of the model was evaluated by Kaplan-Meier curves,time-dependent receiver operating characteristic curves(ROC)in the training dataset,validation dataset and total dataset.Furthermore,Kaplan-Meier stratified analysis was used to test the independent prediction ability of lncRNA model and the univariate and multivariate Cox regression analysis was performed to screen the independent prognostic factors for breast cancer patients.3.By using the "rms" package in R software,the lncRNAs risk-score model and the above independent prognostic factors were combined to construct a nomogram for predicting the 3-year and 5-year survival probability of breast cancer patients.Concordance index(C-index),time-dependent ROC curves and calibration curves were used to evaluate the prediction performance of the nomogram in the training dataset,validation dataset and total dataset.In addition,we also performed time-dependent ROC curves to compare the prognostic performance of the nomogram with other clinicopathological factors.4.Moreover,we explored the potential biological functional of these prognostic lncRNAs by functional enrichment analysis.Based on the fold change and the amount of co-expressed different expressed mRNAs of the above prognostic lncRNAs,we choose the key prognostic lncRNA for further study.In situ hybridization(ISH)were performed to detect the expression level of the key prognostic lncRNA between breast cancer tissues.Kaplan-Meier analysis was used to detect whether the expression of the key prognostic lncRNA was associated with the prognosis of breast cancer patients.5.Transfection of the specific siRNA and shRNA were performed to knockdown the expression of the key prognostic lncRNA in breast cancer cell lines.The EdU incorporation assay,RTCA assay and colony formation assay were used to detect the cell proliferation after knockdown the key prognostic lncRNA.The nude mouse xenograft models were further used to verify the effect of the key prognostic lncRNAs on primary tumor growth in vivo.Results:1.1103 DELs were screened by the "DESeq2" package,including 794 up-regulated DELs and 309 down-regulated DELs.Then,we identified 71 prognostic lncRNAs from 1103 DELs.2.In the training dataset,a 10-lncRNAs risk-score model was successfully established by multivariate Cox regression analysis:Rick Score=(0.634×ExpressionAL138789.1)+(0.488×ExpressionAL513123.1)+(0.254 ×ExpressionLIN00536)+(0.262×ExpressionBCAR4)+(0.825×ExpressionAC079414.1)+(0.253 × ExpressionLHX1-DT)+(1.115 × ExpressionAC006262.3)+(-0.779×ExpressionMIR3150BHG)+(-1.414 × ExpressionAC 105398.1)+(-0.739×ExpressionAL133467.1).Kaplan-Meier curves showed that the model could effectively identify high-risk and low-risk breast cancer patients(P<0.0001).The area under ROC curve(AUC)at 3-year and 5-year were 0.886(95%confidence interval(CI):0.792-0.979)and 0.911(95%CI:0.835-0.987),respectively.3.The results of the validation dataset and the total dataset were consistent with those in the training dataset.The 3-year AUC of the model was 0.574(95%CI:0.400-0.748)and 5-year AUC was 0.650(95%CI:0.482-0.817)in the validation dataset,while in the total dataset,the AUC at 3-year and 5-year reached 0.734(95%CI:0.618-0.849)and 0.781(95%CI:0.688-0.0.875),respectively.Moreover,Kaplan-Meier stratified analysis showed that there existed the statistically differences in OS between high-risk and low-risk patients in multiple subgroups,indicating that our 10-lncRNAs risk-score model had a good independent prognostic performance regardless of age,TNM stage,estrogen receptor(ER)status,and progesterone receptor(PR)status and triple-negative breast cancer.4.Univariate and multivariate Cox regression analysis found that the 10-lncRNAs model,age,TNM stage and HER2 status were the independent prognostic factors of breast cancer.A nomogram combing these independent prognostic factors was then constructed by using the "rms" package of R software.The C-index of the nomogram were 0.891,0.813 and 0.851 in the training dataset,validation dataset and total dataset,respectively.The calibration curves indicated that the nomogram predicted survival probability closely corresponded to the actual survival probability.With respect to the total set,the time-dependent ROC curves revealed that the AUC of this nomogram was 0.804,which was the highest compared with the 10-lncRNAs model(AUC=0.781),age(AUC=0.614),HER2 status(AUC=0.600)and TNM stage(AUC=0.582).5.Functional enrichment analysis revealed that these prognostic lncRNAs were mainly involved in DNA replication,chromosome segregation biological function and p53,oocyte meiosis pathways.According to the fold change and the amount of co-expressed different expressed mRNAs of the above prognostic IncRNAs,we choose LHX1-DT as the key prognostic lncRNA for further study.The results of ISH showed that the expression level of LHX1-DT was significantly up-regulated in breast cancer tissues(P<0.05).Kaplan-Meier analysis found the overexpression of the LHX1-DT was associated with the poor prognosis of breast cancer patients(P<0.05).6.After transfected the specific siRNA and shRNA,LHX1-DT was knockdown in the breast cancer cell lines.All of the EdU incorporation assay,RTCA assay and colony formation assay detected a significant decrease in the cell proliferation after knockdown of LHX1-DT(P<0.05).The nude mouse xenograft models were further verified that LHX1-DT promoted the tumor growth in vivo(P<0.05).Conclusions:1.Our study established a 10-lncRNAs risk-score model which could independently distinguish breast cancer patients with significantly different prognosis even in the intrinsic clinicopathological subgroup.2.Integrating 10-lncRNAs risk-score model and clinical independent prognostic factors to establish nomogram,which could visually predict the 3-year and 5-year survival probability of patients,and will provide new insight into the individual clinical management of breast cancer patients.3.Ten prognostic lncRNAs might involve:in the progression of breast cancer by influencing DNA replication.LHX1-DT promoted the proliferation of breast cancer cells in vitro and in vivo,which might be the novel therapeutic target for breast cancer.Part Two:Establishment of a prognostic risk-score model for HER2-positive breast cancer patientsObjective:To construct a lncRNA risk-score model for predicting the prognosis of HER2-positive breast cancer patients based on TCGA database.Methods:1.A total of 161 HER2-positive breast cancer patients and 113 normal control were extracted from the TCGA database.According to the "DESeq2" package,univariate Cox regression analysis and Kaplan-Meier curves,lncRNAs that were dysregulated in HER2-positive breast cancer patients and correlated with their OS were identified as the candidate prognostic lncRNAs.2.A multivariate Cox regression analysis was then used to mine the independent prognostic lncRNAs and combined them into a multigene risk-score model in the training set.Kaplan-Meier analysis and time-dependent ROC curves were performed to evaluate the prognostic performance of the model in the training set,validation set and total set.Meanwhile,patients in the total dataset was stratified according to TNM stage,ER status,and PR status.Kaplan-Meier analysis was further used to assess the prediction ability of the risk-score model in different subgroups.Results:1.There were 25 DELs related to the prognosis of breast cancer patients,of which 3 lncRNAs(LINC01833,LINC00536,and LINC02725)were the independent prognostic risk factors for patients.2.Based on the coefficients of multivariate Cox regression and the expression of three lncRNAs,those three prognostic lncRNAs were then integrated into a 3-lncRNAs risk-score model:Risk score=0.710×ExpressionLINC01833+1.869×ExpressionLINC00536+ 2.992×ExpressionLINCr2725.In all the three datasets,this model could effectively divided patients into the high-risk group and low-risk group,meanwhile,the low-risk group had a longer OS than the high-risk group(all P<0.05).Time-dependent ROC curves showed that the 3-lncRNAs model had the highest AUC at 5-year(AUC=0.825)compared to single lncRNA(AUCLINC01833=0.787;AUCLINC00536=0.735;AUCLINC02725=0.702),TNM stage(AUC=0.605),ER status(AUC=0.546)and HER2 status(AUC=0.554).What's more,the 3-lncRNAs model also showed better prognostic performance in TNM ?-? subgroup,ER-positive subgroup,PR-negative subgroup,and PR-positive subgroup.Conclusions:The proposed 3-lncRNAs risk-score model is a reliable tool for predicting the prognosis in patients with HER2-positive breast cancer,and it is expected to provide a new reference for clinical individualized management of such patients.