Development of a seven-gene tumor immune microenvironment prognostic signature for high-risk grade III endometrial cancer
2021; Elsevier BV; Volume: 22; Linguagem: Inglês
10.1016/j.omto.2021.07.002
ISSN2372-7705
AutoresMingjun Zheng, Yuexin Hu, Rui Gou, Siting Li, Xin Nie, Xiao Li, Bei Lin,
Tópico(s)Ovarian cancer diagnosis and treatment
ResumoUterine corpus endometrial carcinoma locally infiltrates numerous immune cells and other tumor immune microenvironment components. These cells are involved in malignant tumor growth and proliferation and the process of resistance toward immunotherapies. Here, we aimed to develop a tumor immune microenvironment-related prognostic signature for high-risk grade III endometrial carcinoma based on The Cancer Genome Atlas. The signature was systematically correlated with immune infiltration characteristics of the tumor microenvironment. The seven-gene Riskscore signature was robust and performed well in training, testing, and Gene Expression Omnibus-independent cohorts. A nomogram comprising the gene signature accurately predicted patient prognosis, with our model performing better than other endometrial cancer-related signatures. Analysis of the IMvigor210 immunotherapy cohort revealed that subgroups with a low Riskscore had a better prognosis than subgroups with a high Riskscore. Subgroups with a low Riskscore exhibited immune cell infiltration and inflammatory profiles, whereas subgroups with a high Riskscore experienced progressive disease. The receiver operating characteristic curve indicated that risk score, neoantigen, and tumor mutation burden models together accurately predicted treatment response. Taken together, we developed a tumor microenvironment-based seven-gene prognostic stratification system to predict the prognosis of patients with high-risk endometrial cancer and guide more effective immunotherapy strategies. Uterine corpus endometrial carcinoma locally infiltrates numerous immune cells and other tumor immune microenvironment components. These cells are involved in malignant tumor growth and proliferation and the process of resistance toward immunotherapies. Here, we aimed to develop a tumor immune microenvironment-related prognostic signature for high-risk grade III endometrial carcinoma based on The Cancer Genome Atlas. The signature was systematically correlated with immune infiltration characteristics of the tumor microenvironment. The seven-gene Riskscore signature was robust and performed well in training, testing, and Gene Expression Omnibus-independent cohorts. A nomogram comprising the gene signature accurately predicted patient prognosis, with our model performing better than other endometrial cancer-related signatures. Analysis of the IMvigor210 immunotherapy cohort revealed that subgroups with a low Riskscore had a better prognosis than subgroups with a high Riskscore. Subgroups with a low Riskscore exhibited immune cell infiltration and inflammatory profiles, whereas subgroups with a high Riskscore experienced progressive disease. The receiver operating characteristic curve indicated that risk score, neoantigen, and tumor mutation burden models together accurately predicted treatment response. Taken together, we developed a tumor microenvironment-based seven-gene prognostic stratification system to predict the prognosis of patients with high-risk endometrial cancer and guide more effective immunotherapy strategies. IntroductionUterine corpus endometrial carcinoma (UCEC) is one of the three most common gynecological malignancies and the fourth most common cancer type affecting women in developing countries.1Ferlay J. Soerjomataram I. Dikshit R. Eser S. Mathers C. Rebelo M. Parkin D.M. Forman D. Bray F. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012.Int. J. Cancer. 2015; 136: E359-E386Crossref PubMed Scopus (20798) Google Scholar In 2015, the number of new endometrial cancer cases in China was expected to be 63,400, with an estimated death toll of 21,800.2Chen W. Zheng R. Baade P.D. Zhang S. Zeng H. Bray F. Jemal A. Yu X.Q. He J. Cancer statistics in China, 2015.CA Cancer J. Clin. 2016; 66: 115-132Crossref PubMed Scopus (13681) Google Scholar Approximately three- quarters of the patients with UCEC are diagnosed early, with a 5-year survival rate of more than 70%.3Ferlay J. Steliarova-Foucher E. Lortet-Tieulent J. Rosso S. Coebergh J.W. Comber H. Forman D. Bray F. Cancer incidence and mortality patterns in Europe: estimates for 40 countries in 2012.Eur. J. Cancer. 2013; 49: 1374-1403Abstract Full Text Full Text PDF PubMed Scopus (4051) Google Scholar, 4Bendifallah S. Koskas M. Ballester M. Genin A.-S. Darai E. Rouzier R. The survival impact of systematic lymphadenectomy in endometrial cancer with the use of propensity score matching analysis.Am. J. Obstet. Gynocol. 2012; 206: 500.e1-500.e11Abstract Full Text Full Text PDF Scopus (24) Google Scholar, 5Creasman W.T. Odicino F. Maisonneuve P. Quinn M.A. Beller U. Benedet J.L. Heintz A.P. Ngan H.Y. Pecorelli S. Carcinoma of the corpus uteri. FIGO 26th Annual Report on the Results of Treatment in Gynecological Cancer.Int. J. Gynaecol. Obstet. 2006; 95: S105-S143Crossref PubMed Scopus (755) Google Scholar However, the prognosis of patients with high-risk UCEC, including FIGO (International Federation of Gynecology and Obstetrics) grade III endometrioid cancer, clear cell carcinoma, serous carcinoma, and mixed adenocarcinoma, is still poor in addition to the risk of metastasis and recurrence being very high. The 5-year survival rate is only 17%.6Nugent E.K. Bishop E.A. Mathews C.A. Moxley K.M. Tenney M. Mannel R.S. Walker J.L. Moore K.N. Landrum L.M. McMeekin D.S. Do uterine risk factors or lymph node metastasis more significantly affect recurrence in patients with endometrioid adenocarcinoma?.Gynecol. Oncol. 2012; 125: 94-98Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar, 7Bendifallah S. Canlorbe G. Raimond E. Hudry D. Coutant C. Graesslin O. Touboul C. Huguet F. Cortez A. Daraï E. Ballester M. A clue towards improving the European Society of Medical Oncology risk group classification in apparent early stage endometrial cancer? Impact of lymphovascular space invasion.Br. J. Cancer. 2014; 110: 2640-2646Crossref PubMed Scopus (64) Google Scholar, 8Bendifallah S. Canlorbe G. Collinet P. Arsène E. Huguet F. Coutant C. Hudry D. Graesslin O. Raimond E. Touboul C. et al.Just how accurate are the major risk stratification systems for early-stage endometrial cancer?.Br. J. Cancer. 2015; 112: 793-801Crossref PubMed Scopus (94) Google Scholar, 9Moore K.N. Fader A.N. Uterine papillary serous carcinoma.Clin. Obstet. Gynecol. 2011; 54: 278-291Crossref PubMed Scopus (84) Google ScholarThe tumor microenvironment (TME) is comprised of tumor cells, stromal cells, endothelial cells, immune cells, and extracellular matrix secreted by tumor-associated cells. TME components interact with tumor cells to regulate their growth and development.10Xiao Y. Yu D. Tumor microenvironment as a therapeutic target in cancer.Pharmacol. Ther. 2021; 221: 107753Crossref PubMed Scopus (73) Google ScholarIn recent years, increasing evidence has shown that the occurrence, development, and metastasis of malignant tumor cells are all related to the TME.11Ren B. Cui M. Yang G. Wang H. Feng M. You L. Zhao Y. Tumor microenvironment participates in metastasis of pancreatic cancer.Mol. Cancer. 2018; 17: 108Crossref PubMed Scopus (199) Google Scholar, 12Luo Z. Wang Q. Lau W.B. Lau B. Xu L. Zhao L. Yang H. Feng M. Xuan Y. Yang Y. et al.Tumor microenvironment: The culprit for ovarian cancer metastasis?.Cancer Lett. 2016; 377: 174-182Crossref PubMed Scopus (102) Google Scholar, 13Jiang Y. Wang C. Zhou S. Targeting tumor microenvironment in ovarian cancer: Premise and promise.Biochim. Biophys. Acta Rev. Cancer. 2020; 1873: 188361Crossref PubMed Scopus (53) Google Scholar The tumor immune microenvironment (TIME), as part of the TME, plays an important role in tumor progression.14Yang L. Lin P.C. Mechanisms that drive inflammatory tumor microenvironment, tumor heterogeneity, and metastatic progression.Semin. Cancer Biol. 2017; 47: 185-195Crossref PubMed Scopus (81) Google Scholar In addition, it suppresses immune cells, allowing for immune evasion and tolerance of tumor cells, which in turn affects tumor occurrence and progression. Given the abundance of immune cells and cytokines found in UCEC,15Degos C. Heinemann M. Barrou J. Boucherit N. Lambaudie E. Savina A. Gorvel L. Olive D. Endometrial Tumor Microenvironment Alters Human NK Cell Recruitment, and Resident NK Cell Phenotype and Function.Front. Immunol. 2019; 10: 877Crossref PubMed Scopus (56) Google Scholar it can be inferred that the TME plays a major role in UCEC development and immunotherapy response.In this genomic era, a large number of genome-sequencing technologies and data have emerged,16Wang Z. Gerstein M. Snyder M. RNA-Seq: a revolutionary tool for transcriptomics.Nat. Rev. Genet. 2009; 10: 57-63Crossref PubMed Scopus (8248) Google Scholar and researchers began to focus on the prognosis prediction of UCEC using The Cancer Genome Atlas (TCGA)-UCEC cohort.17Wang X. Dai C. Ye M. Wang J. Lin W. Li R. Prognostic value of an autophagy-related long-noncoding-RNA signature for endometrial cancer.Aging (Albany NY). 2021; 13: 5104-5119Crossref PubMed Scopus (11) Google Scholar, 18Jiang Y. Chen J. Ling J. Zhu X. Jiang P. Tang X. Zhou H. Li R. Construction of a Glycolysis-related long noncoding RNA signature for predicting survival in endometrial cancer.J. Cancer. 2021; 12: 1431-1444Crossref PubMed Google Scholar, 19Liu J. Li S. Feng G. Meng H. Nie S. Sun R. Yang J. Cheng W. Nine glycolysis-related gene signature predicting the survival of patients with endometrial adenocarcinoma.Cancer Cell Int. 2020; 20: 183Crossref PubMed Scopus (18) Google Scholar, 20Zhou M. Zhang Z. Zhao H. Bao S. Sun J. A novel lncRNA-focus expression signature for survival prediction in endometrial carcinoma.BMC Cancer. 2018; 18: 39Crossref PubMed Scopus (49) Google Scholar Wang et al.17Wang X. Dai C. Ye M. Wang J. Lin W. Li R. Prognostic value of an autophagy-related long-noncoding-RNA signature for endometrial cancer.Aging (Albany NY). 2021; 13: 5104-5119Crossref PubMed Scopus (11) Google Scholar constructed an autophagy-related long-noncoding RNA signature to predict the prognosis of UCEC. Jiang et al.18Jiang Y. Chen J. Ling J. Zhu X. Jiang P. Tang X. Zhou H. Li R. Construction of a Glycolysis-related long noncoding RNA signature for predicting survival in endometrial cancer.J. Cancer. 2021; 12: 1431-1444Crossref PubMed Google Scholar and Liu et al.19Liu J. Li S. Feng G. Meng H. Nie S. Sun R. Yang J. Cheng W. Nine glycolysis-related gene signature predicting the survival of patients with endometrial adenocarcinoma.Cancer Cell Int. 2020; 20: 183Crossref PubMed Scopus (18) Google Scholar constructed a glycolysis-related gene signature to predict the prognosis of UCEC. However, these studies are limited to the analysis of specific genes and prognosis in all TCGA-UCEC samples and lack external verification. Therefore, the prediction of the prognosis of high-risk UCEC accurately remains a challenge.In this study, we included all high-risk grade III tumor samples in TCGA-UCEC, comprising mixed-type, serous, and endometrioid endometrial adenocarcinoma (SEA and EEA, respectively). Then, an immune prognostic signature of grade III UCEC was constructed based on TIME genes and validated in different cohorts. In addition, we evaluated the prognostic merit of the seven-gene signature for immunotherapy response. The current findings revealed that the gene signature could be used to evaluate the prognosis of patients with grade III UCEC and guide clinical decision-making, including immunotherapy-related choices.ResultsIdentification of two molecular subtypes based on TIME genesAfter deduplication and filtering, the expression profile of 1,356 tumor immune-related genes in TCGA was extracted and analyzed by univariate Cox analysis (Table S1), and 195 genes related to the prognosis of UCEC were obtained (Table S2; p < 0.05). Non-negative matrix factorization (NMF) helps extract the biological correlation coefficients of the data in the gene expression matrix and obtain the internal characteristic structure of the data and finally groups the samples. This approach is widely used in the molecular classification of cancers at present. With the use of the NMF algorithm, the optimal number of subtypes was two (Figures 1A, S1A, and S1B). The C2 subtype has a significantly worse prognosis than C1 in terms of overall survival (OS) and progression-free survival (PFS) (Figures 1B and 1C; log-rank p < 0.01). The R software package ESTIMATE was used to evaluate StromalScore, ImmuneScore, and ESTIMATEScore between C1 and C2 subtypes. Further, MCPcounter was used to evaluate 10 immune cell types, and CIBERSORT was employed to evaluate 22 immune cell types. ESTIMATE and MCPcounter results revealed that the tumor immune infiltration degree of the C1 subtype was mostly higher than that of the C2 subtype (Figures 1E and 1F). CIBERSORT results indicated that M1 macrophages, CD8+ T cells, and follicular helper T cell abundance in the C1 subtype were significantly higher than in the C2 subtype (Figure 1D). Furthermore, the distribution of tumor-infiltrating immune cells (TIICs) between the two subtypes, as evaluated by the three methods, was nearly the same. We, therefore, speculated that the C1 subtype may have a better response to immunotherapy. The heatmap of TIIC distribution for the two subtypes is shown in Figure 1G.Further, we compared the distribution of histopathological subtypes' age, FIGO stage, and survival between the two molecular subtypes. The results showed that the C1 subtype is dominated by EEA samples, whereas in the C2 subtype, most samples are SEA (Figure S1C). The survival rates for the two subtypes were significantly different, with the mortality rate of the C2 subtype being higher (Figure S1D). The age ratios of the two subtypes were significantly different, as the proportion of middle-aged patients in subtype C2 was higher (Figure S1E). There was also a significant difference in the proportion of FIGO stages between the two subtypes, as the proportion of patients with FIGO stage II, III, and IV was higher in the C2 subtype than in the C1 subtype (Figure S1F).Identification of differentially expressed genes (DEGs) between subtypesAs previously described, we obtained 883 DEGs (Table S3), and the volcano map of upregulated and downregulated DEGs between the C1 and C2 subtypes is shown in Figure S1G. Among them, 443 genes were upregulated, whereas 440 genes were downregulated. We selected the top 100 genes with the most significant upregulation or downregulation to construct a heatmap, which is shown in Figure S1H. DEGs between C1 and C2 have evident distribution characteristics.Construction of a prognostic multi-gene signature based on TIME subtypeTCGA cohort was divided into training and testing cohorts, with 156 samples each (Table 1). Based on the training cohort, the univariate Cox proportional hazard regression model was employed to identify prognostic DEGs between the subtypes with the threshold value of p < 0.05. Finally, 42 prognostic hub genes were obtained (Table S4). These genes may serve as potential TIME-related characteristic genes.Table 1TCGA training set and validation set sample informationClinical featuresTCGA-UCEC trainTCGA-UCEC testpOS01191240.585313732StageI81840.6354II1915III4642IV1015Age≤6575780.9436>658077Unknown11 Open table in a new tab As an excessively large number of genes are not conducive to clinical detection, we further narrowed the range of immune-related genes. We employed the least absolute shrinkage and selection operator (Lasso) regression analysis, and the resulting change trajectory of each independent variable is shown in Figure 2A. With the gradual increase of lambda, the number of independent variable coefficients gradually increased to zero. Five-fold cross-validation was used to build the model, and the confidence interval under each lambda is shown in Figure 2B, indicating that when log(lambda) = −3.65, the model was optimal. Thus, we selected 18 genes at lambda = 0. 0262 as the candidate genes. As described in Materials and methods, in order to obtain the best fit of the model, the Akaike information criterion (AIC) method was then employed from which we obtained seven genes, namely DRAM1, TNFRSF14, SCGB2A1, EMX2, DNER, DAPL1, and interferon-induced protein with tetratricopeptide repeats sequence 1 (IFIT1). The seven-gene signature estimated Riskscore (RS) is as follows: RS = −0.621 ⋅ DRAM1 − 0.343 ⋅ tumor necrosis factor receptor (TNFR)SF14 − 0.106 ⋅ SCGB2A1 − 0.433 ⋅ EMX2 + 0.259 ⋅ DNER + 0.248 ⋅ DAPL1 + 0.344 ⋅ IFIT1.Figure 2Analysis of Lasso regressionShow full caption(A) The changing trajectory of each independent variable (the abscissa represents the corrected lambda, and the ordinate represents the coefficient of the independent variable). (B) The log value of the independent variable lambda (the abscissa represents the confidence interval of each lambda, and the ordinate represents errors in cross validation). (C) The KM curve of the seven-gene signature-based stratification in TCGA training cohort. (D) The 1-, 3-, and 5-year ROC curve based on seven-gene signature stratification.View Large Image Figure ViewerDownload Hi-res image Download (PPT)To obtain a fixed grouping threshold between different cohorts, we converted RS to standard score; the samples with RS greater than zero are divided into high-risk groups (HRGs), whereas those with less than zero are divided into low-risk groups (LRGs). The Kaplan-Meier (KM) curve is shown in Figure 2C. Seventy-one samples were added into the HRG and 85 samples into the LRG. There was a significant difference in survival between the HRG and LRG (p < 0.0001). R software package timeROC was used to analyze the prognostic efficiency of the RS. The area under the curves (AUCs) of gene signature for 1-year, 3-year, and 5-year survival were 0.79, 0.82, and 0.89, respectively (Figure 2D).Internal and external validation of the robustness of the signatureTo determine the robustness of the signature, internal cohorts (TCGA testing and TCGA-UCEC), along with an external cohort, named Gene Expression Omnibus (GEO): GSE119041, were used for validation. The RS was calculated using the same formula as the training cohort. Subsequently, the samples are divided into HRG and LRG according to the previous step.Significant prognostic differences were observed between the HRG and LRG from TCGA testing cohort (Figure 3A; p < 0.01). The AUCs of 1-year, 3-year, and 5-year survival in TCGA testing cohort were 0.62, 0.71, and 0.69, respectively (Figure 3B).Figure 3Validation of the seven-gene prognostic signatureShow full caption(A) The OS curve of seven-gene signature classification in TCGA testing cohort. (B) The 1-, 3-, and 5-year ROC curve based on seven-gene signature stratification. (C and D) The OS curve of seven-gene signature-based stratification and 1-, 3-, and 5-year ROC curve of TCGA-UCEC cohort. (E and F) The OS curve of the seven-gene signature-based stratification and the 1-, 3-, and 5-year ROC curve in the GEO: GSE119041 cohort.View Large Image Figure ViewerDownload Hi-res image Download (PPT)In the entire TCGA-UCEC cohort, significant prognostic differences were also observed between the HRG and LRG (Figure 3C; p < 0.01). The AUCs of 1-year, 3-year, and 5-year survival in TCGA-UCEC cohort were 0.72, 0.77, and 0.80, respectively (Figure 3D).In the GEO: GSE119041 cohort, the prognosis of the HRG was significantly worse than that of the LRG (Figure 3E; p < 0.05). Lastly, the AUCs of 1-year, 3-year, and 5-year survival in the GEO: GSE119041 cohort were 0.76, 0.79, and 0.78, respectively (Figure 3F). These results indicated that our model performed robustly in different cohorts.Performance of RS with regard to clinical featuresThe RS constructed by the seven-gene signature could effectively distinguish between the HRG and LRG with regard to age and FIGO stage (Figures 4A−4D ; p < 0.05). The RS of the group over 65 years old was higher than that of the group ≤65 years old (Figure 4E). In addition, the advanced FIGO stage had a higher RS than the early stage (Figure 4F). Between the molecular subtypes, the RS of the poor prognosis C2 subtype was higher than the C1 subtype (Figure 4G). These findings indicated that our signature had a good ability to predict prognosis based on different clinical characteristics.Figure 4KM survival curves in different clinical subgroups stratified based on the seven-gene risk modelShow full caption(A) Age >65 years. (B) Age ≤65 years. (C) FIGO stages I + II. (D) FIGO stages III + IV. (E) Correlation diagram between risk score and age. (F) Correlation diagram between risk score and FIGO stage. (G) Correlation diagram between risk score and subtype cluster. (H) The top 22 pathways with RS correlation greater than 0.3. (I) Heatmap of the relationship between pathways and RS; the horizontal axis represents the sample, and the RS increases from left to right.View Large Image Figure ViewerDownload Hi-res image Download (PPT)We further explored the relevant pathways that characterize the different clinical features of RS. R software package gene set variation analysis (GSVA) was used to perform gene set enrichment analysis (GSEA) for each sample, then calculated the correlation between biological pathways and the RS, and selected the top 22 pathways with a correlation greater than 0.3 for visual display (Figure 4H). It is evident that INTESTINAL_IMMUNE_NETWORK_FOR_IGA_PRODUCTION, T_CELL_RECEPTOR_SIGNALING_PATHWAY, and NATURAL_KILLER_CELL_MEDIATED_CYTOTOXICITY tumor immune-related pathways are negatively correlated with RS (Figure 4I).Construction and evaluation of nomograms comprising the signatureThe independence of the seven-gene signature in clinical application was evaluated by univariate and multivariate Cox regression analyses. The results revealed that RS was significantly correlated with prognosis in both the univariate model (hazard ratio [HR] = 3.80, p < 0.001) and multivariate model (HR = 3.60, p < 0.001) (Figures 5A and 5B ), indicating that the seven-gene signature had a good clinical predictive value.Figure 5Clinical value of the predictive modelShow full caption(A) Forest plot of univariate Cox analysis. (B) Forest plot of multivariate Cox analysis. (C) Nomogram predicting the 1-, 3-, and 5-year OS of patients. The nomogram is applied by adding up the points from the point scale for each variable to a total score. Based the total score, the probability of 1-, 3-, or 5-year survival is projected on the bottom scales. (D) Calibration curves for nomogram-predicted 1-, 3-, and 5-year OS in relation to actual survival. (E−G) ROC curves of nomograms compared with those of other clinical variables with regard to 1-, 3-, and 5-year survival. The DCA curves can evaluate the clinical potential of nomograms. Black indicates that all samples are negative, and none are treated. Therefore, the net benefit is 0. Gray indicates that all samples are positive, and all are treated. The x axis represents threshold probabilities of patients experiencing: (H) 1-year survival; (I) 3-year survival; and (J) 5-year survival.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Based on the results of multivariate Cox analysis, significant clinical features such as FIGO stage and RS were combined to construct a nomogram (Figure 5C). The nomogram comprising the RS and FIGO stage proved useful to predict survival.Calibration curves were used to visualize the performance of 1-, 3-, and 5-year nomograms. In each case, the 45° line represents the best predictive ability. The calibration results indicated that the nomogram performed well (Figure 5D). The AUC of the 1-, 3-, and 5-year nomograms was larger than that of the other clinical variables (Figures 5E−5G). Decision curve analysis (DCA) was used to assess the validity of the signature, and the nomogram showed the greatest net benefit (Figures 5H−5J). These results suggested that the nomogram is better for predicting the survival of patients with grade III UCEC than using a single clinical factor. Thus, it may be useful during the clinical decision-making process and for choosing individualized treatments.The seven-gene signature performed better than others in prognostic predictionTo determine whether our seven-gene signature had a superior predictive ability for TCGA-UCEC cohort, we compared it with four published prognostic signatures, namely, a nine-gene signature (Jiang et al.21Jiang P. Sun W. Shen N. Huang X. Fu S. Identification of a metabolism-related gene expression prognostic model in endometrial carcinoma patients.BMC Cancer. 2020; 20: 864Crossref PubMed Scopus (14) Google Scholar), a seven-gene signature (Liu et al.22Liu L. Lin J. He H. Identification of Potential Crucial Genes Associated With the Pathogenesis and Prognosis of Endometrial Cancer.Front. Genet. 2019; 10: 373Crossref PubMed Scopus (36) Google Scholar), a six-gene signature (Wang et al.23Wang Y. Ren F. Chen P. Liu S. Song Z. Ma X. Identification of a six-gene signature with prognostic value for patients with endometrial carcinoma.Cancer Med. 2018; 7: 5632-5642Crossref PubMed Scopus (30) Google Scholar), and another nine-gene signature (O'Mara et al.24O'Mara T.A. Zhao M. Spurdle A.B. Meta-analysis of gene expression studies in endometrial cancer identifies gene expression profiles associated with aggressive disease and patient outcome.Sci. Rep. 2016; 6: 36677Crossref PubMed Scopus (26) Google Scholar). To make signatures comparable, we calculated the RS of each UCEC sample in all TCGA cohorts by the same method and converted the RS according to the previous methods in the four signatures. All four signatures could effectively divide patients into two subgroups with significantly different prognoses (Figures 6A−6C and 6G ). However, receiver operating characteristic (ROC) analysis revealed that the AUC values of the four signatures for 1-, 3-, and 5-year survival were lower than those of our model (Figures 6D−6F, and 6H). The restricted mean survival (RMS) package was used to calculate the C-index of all prognostic signatures. Our model had the highest C-index at 0.72 (Figure 6I). These findings highlighted the superior predictive performance of our TIME gene signature.Figure 6Comparison of the seven-gene risk model with other modelsShow full caption(A and B) The ROC and KM curves of a nine-gene signature (Jiang et al.21Jiang P. Sun W. Shen N. Huang X. Fu S. Identification of a metabolism-related gene expression prognostic model in endometrial carcinoma patients.BMC Cancer. 2020; 20: 864Crossref PubMed Scopus (14) Google Scholar). (C and D) The ROC and KM curves of another seven-gene signature (Liu et al.22Liu L. Lin J. He H. Identification of Potential Crucial Genes Associated With the Pathogenesis and Prognosis of Endometrial Cancer.Front. Genet. 2019; 10: 373Crossref PubMed Scopus (36) Google Scholar). (E and F) The ROC and KM curves of a six-gene signature (Wang et al.23Wang Y. Ren F. Chen P. Liu S. Song Z. Ma X. Identification of a six-gene signature with prognostic value for patients with endometrial carcinoma.Cancer Med. 2018; 7: 5632-5642Crossref PubMed Scopus (30) Google Scholar). (G and H) The ROC and KM curves of a nine-gene signature (O'Mara et al.24O'Mara T.A. Zhao M. Spurdle A.B. Meta-analysis of gene expression studies in endometrial cancer identifies gene expression profiles associated with aggressive disease and patient outcome.Sci. Rep. 2016; 6: 36677Crossref PubMed Scopus (26) Google Scholar). (I) C-indexes of the five risk models.View Large Image Figure ViewerDownload Hi-res image Download (PPT)The seven-gene signature effectively predicted the efficacy of immunotherapyThe identification of novel predictive markers is essential for effective immunotherapy. We obtained an immunotherapy cohort (IMvigor210) to explore whether the seven-gene signature could predict the benefit of immunotherapy.25Rosenberg J.E. Hoffman-Censits J. Powles T. van der Heijden M.S. Balar A.V. Necchi A. Dawson N. O'Donnell P.H. Balmanoukian A. Loriot Y. et al.Atezolizumab in patients with locally advanced and metastatic urothelial carcinoma who have progressed following treatment with platinum-based chemotherapy: a single-arm, multicentre, phase 2 trial.Lancet. 2016; 387: 1909-1920Abstract Full Text Full Text PDF PubMed Scopus (2499) Google Scholar The IMvigor210 cohort contains a gene expression profile for patients with or without beneficial responses to anti-programmed death ligand 1 (PD-L1) immunotherapy for metastatic urothelial carcinoma (MUC). The KM curve indicated that the prognosis of HRG was worse than that of patients with LRG (Figure 7A). The ROC curve indicated that the combination of the RS, neoantigen (NEO), and tumor mutational burden (TMB) models with logistic regression could predict treatment response with 83% accuracy, which was higher than that of NEO (AUC = 0.62) or TMB (AUC = 0.56) alone (Figure 7B). MCPcounter was used to calculate the immune cell scores of IMvigor210 samples. The results revealed
Referência(s)