| [Background] According to the2012Annual Report of Cancer Registration in China,colorectal carcinoma (CRC) and gastric cancer (GC) accounted for23.0%of total newmalignant cases and22.2%of malignant disease-related deaths in Mainland China. Localrecurrences and metachronous metastases after surgery are the leading causes ofcancer-related death. Except the tumor-node-metastasis (TNM) staging system, which wasbased on the anatomical extent of the disease after surgery, there are not prognostic orpredictive biomarkers currently available for clinical decision makings. Therefore, it is ofgreat significance to develop prognostic and/or predictive biomarkers useful in accurateprediction of prognosis and therapeutic response, paving the way for personalized treatment(e.g. targeted therapy). Gene expression profiling provides an opportunity to develop robustprognostic and predictive predictors of CRC and GC. However, published gene signatureshave been not yet applied clinically due to a lack of overlapping genes, adequate validation,and easy implementation. Thus, it is an indispensible step to translate current findings byintegrating the related published signatures into a new immunohistochemistry signature andvalidate its prognostic and predictive values in large clinical cohorts. In our previous study,we had found a key molecule involving in GC metastasis, NR4A2, based on gene expressionprofiling and protein-protein interaction (PPI) network assay. The role of NR4A2inpredicting postoperative prognosis and therapeutic outcome of gastrointestinal cancers havenot been reported.[Objective] To develop an immunohistochemistry signature with stable prognostic andpredictive values for CRC in distinct cohorts by integrating current microarray data, andclarify the role of nuclear orphan receptor NR4A2on chemo-resistance and in predictingtherapeutic outcome in GC and CRC, paving the way for personalized treatment of cancers.[Method] With the use of bioinformatics, we initially searched for and assessed valid genesignatures by checking original databases and excluding those with small sample size andduplications. We recruited24published signatures and constructed an in-house CRCprognosis-related gene signature. A total of9available microarray datasets with prognosisinformation were combined into4according to their platforms. After gene identity mapping,prognostic values of the25gene signatures were reassessed separately in each of the4combined microarray datasets using a modified nearest template prediction (NTP) method Signatures with high concordance and high prognostic values were integrated into asub-network to select the network-centric and the top overlapping molecules as candidatebiomarkers. Tissue micoarrarys were prepared using formalin-fixed paraffin-embeddedspecimens of the CRC patients who received radical surgery. Expression patterns of theselected molecules in surgical tissues were examined using immunohistochemistry. The lassopenalized multivariate Cox proportional hazards model was employed to perform the variableselection and shrinkage by using leave-one out cross-validation (LOOCV). We assessedprognostic values of the candidates’ expression in the specimens of682CRC patientsrecruited from Changhai Hospital, Second Military Medical University, Shanghai, China,between2001and2009(training cohort) using immunohistochemistry and constructed asignature. The signature was validated in343CRC patients from The Sixth AffiliatedHospital, Sun Yet-sen University, Guangzhou, China, between2000and2006(validationcohort). Human NR4A2gene was transfected into human gastric cancer cells to investigatethe effect of NR4A2expression on cell growth/invasiveness, tumorigenicity in nude mice,and chemo-resistance to5-fluorouracil and oxaliplatin. We also investigated PGE2-inducedNR4A2expression and its effect on the chemo-resistance and anti-apoptosis. Postoperativesurvival analyses were carried out in245GC patients and682CRC patients at I-III stages.The expression of NR4A2was also examined using immunohistochemistry.[Results] Five signatures with high concordance and prognostic value were integrated into anetwork. Network-centric molecules (GRB2, PTPN11, and ITGB1) and the top overlappingmolecule (POSTN), each significantly associated with disease-free survival (DFS), wereoptimized to construct our signature. As continuous variables, signature score and TNM stageindependently predicted a decreased DFS in both cohorts, furthermore, the signature hadbetter prognostic values for5-years DFS than did TNM staging in the training (AUC,0.860vs0.669; P<0.0001) and the validation cohorts (AUC,0.788vs0.589; P<0.0001). The patientswere dichotomized into high-risk and low-risk subgroups with the median risk score.Compared to patients with low-risk scores, those with high-risk scores had shorter DFS in thetraining (HR,8.32;95%CI,4.54-15.25) and the validation cohorts (HR,2.94;95%CI,1.82-4.73) in multivariate Cox regression analysis. In all patients with postoperativechemotherapy, those with high-risk scores had shorter DFS (HR,7.15;95%CI,4.35-11.76)and Disease-specific survival (DSS)(HR,5.86;95%CI,3.16-10.86) than those with low-riskscores. Ectopic expression of NR4A2increased the tumorigenicity in vivo and thechemo-resistance to F-FU and attenuated the chemotherapy-induced apoptosis in bothintestinal-type moderately differentiated GC and diffuse-type undifferentiated metastatic GCcells. Transient treatment of PGE2significantly up-regulated NR4A2expression via proteinkinase A pathway, increased the chemo-resistance, and attenuated the apoptosis. Theexpression of NR4A2was high in the nuclei of cancer cells of stage I-III GCs. High NR4A2expression in the surgical specimens was significantly associated with a reduced DSS of GCpatients (log-rank test, P=0.011). This effect was only observed in those who receivedchemotherapy (DSS in GC, P=0.016). In multivariate Cox regression analyses, high NR4A2expression independently predicted poor prognosis of GC patients (HR=1.51, P=0.044).Ectopic expression of NR4A2up-regulated stem-like markers CD44and EpCAM on CRCcells. High NR4A2expression, either ectopic expression or transient treatment with PGE2,increased the chemo-resistance and attenuated the5-fluorouracil and oxaplatin-inducedapoptosis in both DNA mismatch repair (MMR)-proficient and the MMR-deficient CRC cells.NR4A2was high expressed in the nuclei of cancer cells of stage I-III tumors. High NR4A2expression in specimens was significantly associated with a reduced survival of CRC patients(DSS，HR，1.94; log-rank P=0.015). Further stratification analysis indicated that this effectwas only observed in the patients who received chemotherapy with colon cancer (HR=1.86,log-rank P=0.020). In multivariate Cox regression analyses, high NR4A2expressionindependently predicted poor prognosis of CRC patients (DSS, HR=1.94, P=0.017).[Conclusion] In this study, we developed an immunohistochemistry signature with a higherprognostic power than the TNM staging for postoperative prognosis and therapeutic outcomeof CRC. Inflammatory milieu promotes the expression of NR4A2and high NR4A2expression confers chemo-resistance, attenuates chemotherapy-induced apoptosis, andsignificantly predicts poor postoperative prognoses of GC and colon cancer patients,especially for those who received chemotherapy. The findings of this study are of greatsignificance to improve postoperative prognosis of CRC and GC using personalizedtreatment. |
PDF Full Text Request