An early relapse-associated gene signature optimizes prognostic prediction for gastric cancer patients

doi:10.21203/rs.3.rs-1494225/v1

Download PDF

Research Article

An early relapse-associated gene signature optimizes prognostic prediction for gastric cancer patients

https://doi.org/10.21203/rs.3.rs-1494225/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: Gastric cancer (GC) is one of the most common cancers in the world. Patients with GC who experienced early relapse have poor prognosis. We aim to develop an early relapse-associated gene signature to optimize prognosis prediction in patients with replapsing GC.

Methods: The GC cohorts including GSE62254 set (N=300) and GSE15459 set (N=192) were extracted from Gene Expression Omnibus (GEO) database. Propensity score matching (1:1) based on pathological stage was executed between patients with early relapse and long-term survival from GSE62254 set. Global transcriptome analysis was performed between the two groups to identify early relapse-associated gene. Based on the differentially expressed genes, we developed a classifier incorporating 5 genes that using LASSO Cox regression model. The signature’s prognostic value was internally validated in 210 GC patients and validated in GSE15459 set externally. The patients from GSE62254 set could be divided into high-risk or low-risk group.

Results: In the train set, patients in high-risk group had poor prognosis as compared to those in low-risk group [hazard ratio (HR): 3.002, 95% confidence interval (CI): 2.132-4.226, P<0.001)]. Good reproducibility for the prognostic value of early relapse-associated gene signature was verified in the internal validation set (HR: 2.772, 95% CI: 1.836–4.184, P<0.001) and another external validation set (HR: 1.733, 95% CI: 1.149–2.614, P=0.009). Also, we developed a nomogram which integrated the five mRNA classifier, pathological stage and lymph node ratio (LNR) to evaluate prognosis based on GSE62254 set. Time-dependent receiver-operating characteristic at 1 year demonstrated that integrated signature had better prognostic accuracy [area under curve (AUC=0.849)] than the American Joint Commission on Cancer TNM staging system (AUC=0.773) and LNR (AUC=0.811) in GSE62254 set.

Conclusions: This study suggest that an early relapse-associated gene signature that can robustly divide gastric cancer patients into two groups with distinctive prognosis. This classifier may contribute to select gastric cancer patients with poor prognosis who require more frequent follow-up and more aggressive therapeutic intervention.

gastric cancer

GEO database

early relapse

prognosis

gene signature

Gastric cancer (GC) is the fifth most common malignancy and remains the third leading cause of cancer death in the world in 2018 [1]. Although the systemic treatments of GC including surgical resection, chemotherapy or chemo-radiotherapy have significantly improved the long-term survival of GC patients, the prognosis of advanced GC patients remains poor [2]. Previous studies reported that approximately 30% of patients developed recurrence or distant metastasis even after postadjuvant chemotherapies with a 14-month median recurrence-free survival time[3, 4]. Multiple factors resulted in the early relapse of GC, such as age, gender, tumor location, differentiation type, tumor invasion, regional lymph nodes metastasis, worse histopathological type, distant metastasis and poor response to adjuvant chemotherapy[5–7]. The early relapse of GC following surgery is a main cause of cancer-related deaths which lead to poorer long-term survival rates.

Many kinds of biomarkers or strategies have been applied to evaluate the risk of early relapse of GC, and the American Joint Commission on Cancer/International Union against Cancer (AJCC/UICC) tumor–node–metastasis (TNM) staging system is the most commonly used parameter for prediction. However, GC patients with the same TNM stage may have different prognosis due to the tumor heterogeneity[8], indicating the classical staging methods may have its shortcomings for precise prediction. Therefore, researchers are extensively looking for novel, reliable biomarkers or indicators to identify GC patients with high risk of early relapse. Recent studies on many malignant tumors have shown that multiple gene expression patterns or gene signatures can be a good predictor of cancer prognosis[9–11]. Zhu et al. provided an efficient classification method for clinical prognosis prediction of GC with a long non-coding RNA signature[12]; Zhang et al. reported a circular RNAs signature method can predict the early recurrence of stage III GC[13]. However, few precious gene maps have been used to detect genes associated with early relapse of GC. Looking for a genetic signature may help to predict early relapse in GC patients.

In the present study, GC cohorts (GSE62254[14] set and GSE15459[15] set) were extracted from Gene Expression Ominus (GEO) database. We identified a 5 gene-related signature from the test set in GSE62254 by using Cox regression model. The signature was subsequently validated in the internal validation set. This early relapse-associated gene signature can robustly separate GC patients into two groups with distinctive prognosis, and may contribute to select GC patients with poor prognosis who require more frequent follow-up and more aggressive therapeutic intervention.

Preprocessing of transcriptome microarray

Raw microarray GC datasets were acquired from the GEO database (http://www.ncbi.nlm.nih. gov/geo/) and were normalized by using Robust Multichip Average[16]. All data are under the Affymetrix Human Genome U133 + 2.0 chip platform. All probes are mapped to the newest version of NetAffx annotation files. If multiple probes are mapped to the identical EntrezGeneID, the average value is used to represent its average expression level. Combat methods are used to eliminate potential internal and external batch effects.

Dataset selection

GC datasets of GSE62254 set (N=300) and GSE15459 set (N=192) were identified in this research to construct and verify the predictive value of gene signature. The inclusion criteria for the datasets include: (a) all sets were created by Affymetrix HG-U133 plus 2.0 platform; (b) basic clinical information and follow-up data of overall survival (OS) and status should be available. The datasets that lack of necessary clinicopathological or follow-up information were excluded.

Identification of early relapse-associated genes

In our present study, early relapse refers to the locoregional recurrence or distant metastasis within 1 year after radical surgery. Patients from GSE62254 set were classified into early relapse group and long-term survival group (no relapse was observed with a minimum of 5 years follow-up). Propensity score matching (PSM) analysis was conducted between the two groups to adjust for pathological stage, which was the most significant factor in predicting relapse. 67 matched patients were later included in the study to determine the gene expression profile changes between the early relapse group and the long-term survival group. Analysis of differentially expressed genes (DEGs) between two groups was conducted using the linear models for microarray data (LIMMA) method. The threshold for screening of DEGs was set as P < 0.05 and fold change ≥ 1.5. Furthermore, LASSO Cox regression model was utilized to choose the most significantly relapse-related mRNA of all the DEGs.

Development of risk score and Statistical analysis

LASSO Cox regression model was used to identify a list of mRNAs constructed a multi-gene-based classifier for predicting the early relapse of GC patients in the training set. According to the standard formula of risk score, the patients in different sets were divided into high-risk and low-risk groups with the median of training set as the cutoff point.

Kaplan–Meier estimate with log-rank test was used to evaluate the survival differences between high-risk and low-risk of early relapse of GC patients. Multivariate Cox regression analysis was used to test the independent role of risk score in OS prediction. Receiver operating characteristic (ROC) curves were used to study the prediction or prediction accuracy of our signature. R (version 3.6.1, https://www.r-project.org/) was used to test all the statistical analyses and P < 0.05 were considered as statistically significant.

Preparation of GEO datasets

The GC datasets and corresponding clinical information were downloaded from the publicly available GEO database. 492 GC patients were identified and studied in total, including 300 patients from GSE62254 set and 192 patients from GSE15459 set. Patients from GSE62254 set were used as train series to develop the signature. Two-thirds of patients were selected randomly for internal validation, while patients from GSE15459 were used as external validation.

Development of early relapse signature from the GSE62254 set

Samples from the GSE62254 set were divided into early relapse group and long-term survival group. PSM analysis with 1:1 match was conducted between the two groups to adjust for pathological stage. Global changes of mRNA expression profiles were analyzed between two groups. 180 genes were differentially expressed between early relapse group and long-term survival group (P < 0.05, fold change ≥ 1.5; Fig. 1). LASSO Cox regression model were established. Of these, compared to the long-term survival group, 3 mRNA were downregulated and 2 mRNA were upregulated in the early relapse group (Fig. 2). We subsequently developed a 5-mRNA signature, weighted by the level of 5-mRNA expression and their regression coefficients, to calculate the risk score for each patient. Risk score = (0.568259164224088 × expression level of MATN3) + (-0.516812065573742 × expression level of CDC42EP5) + (-0.865365153482474 × expression level of TNFRSF11A) + (0.416093717199612 × expression level of TAC1) + (0.501124977450989 × expression level of LOC400043).

The prognostic value of 5-gene signature in train, internal validation, external validation series

Patients in train set were classified into high-risk group (N = 150) and low-risk group (N = 150) according to the median risk score as cutoff point. In the train set, patients in high-risk group had poor prognosis as compared to those in low-risk group [hazard ratio (HR): 3.002, 95% confidence interval (CI): 2.132–4.226, P < 0.001)]. As expected, good reproducibility for the prognostic value of early relapse-associated gene signature was verified in the internal validation set (HR: 2.772, 95% CI: 1.836–4.184, P < 0.001) and another external validation set (HR: 1.733, 95% CI: 1.149–2.614, P = 0.009) (Fig. 3). Multivariate analysis showed that our five-gene signature was an independent prognostic factor in train set and external validation set (Table 1). ROC analysis was used to evaluate the prognostic accuracy of the signature (Fig. 4). Furthermore, the distribution of the gene risk score, the survival status of patients and gene expression differences between the two groups were also demonstrated (Fig. 5–7).

Development of a nomogram for predicting prognosis in gastric cancer patients

Lastly, we developed a nomogram which integrated the five-mRNA classifier, pathological stage and lymph node ratio (LNR) to evaluate prognosis based on GSE62254 set. Time-dependent receiver-operating characteristic at 1 year demonstrated that integrated signature had better prognostic accuracy [area under curve (AUC = 0.849)] than the AJCC/TNM staging system (AUC = 0.773) and LNR (AUC = 0.811) in GSE62254 set (Fig. 8).

While gastrectomy with D2 lymph node dissection and adjuvant chemotherapy for GC has greatly reduced locoregional recurrence, 30% of GC patients suffered from early relapse, which caused poor prognosis. Therefore, new predictive factors are needed to identify early recurrence. Previous studies have identified some ideal clinical or molecular markers to predict the early relapse of GC. Hiroki Imaoka et al. reported that serum miR-203 may be a noninvasive biomarker for prognosis and metastasis prediction in patients with GC[17]. Kohei Wakatsuki et al. found that pN ≥ 14, CA19-9 ≥ 37 IU/ml, and BL ≥ 445 ml were independent risk factors for early recurrence of pStage III GC after radical surgery[18]. Min Ma et al. developed a nomogram model including tumors location, positive lymph node ratio, pTNM stage III, lymphocyte count, postoperative infection complications and adjuvant chemotherapy, which can be used to predict the possibility of early recurrence after radical resection of stage II/III GC[19]. However, the high-throughput expression profile datasets is rarely used to predict the early relapse of GC.

In the present research, we developed a 5 gene-related signature to predict the early relapse of GC. In the train set, patients in high-risk group had poorer prognosis as compared to those in low-risk group. Early relapse-associated gene signature showed good producibility which was later verified in both the internal and external validation sets. Also, we developed a nomogram which integrated the five mRNA classifier, pathological stage and LNR to evaluate prognosis based on overall survival. Time-dependent receiver-operating characteristic at 1 year demonstrated that integrated signature had better prediction accuracy than the AJCC TNM staging system and LNR in GSE62254 set. Therefore, our study identified a robust integrated 5 gene-related signature that could help identify high-risk patients with early relapse and could be a new option in addition to the AJCC staging system.

Most genes included in the signature have been experimentally studied to be related to cancer. Among them, four mRNAs including MATN3, CDC42EP5, TNFRSF11A and TAC1 have been previously reported. GEO and TCGA data predicted that MATN3 mRNA levels were significantly higher in GC tissue than normal tissue. Further survival analyses revealed that GC patients with high mRNA expression of MATN3 had significantly lower DFS and OS than those with low mRNA expression[20]. In prostate cancer, CDC42EP5 was closely related to tumor heterogeneity in association with PTEN deletion[21]. Previous studies revealed that TNFRSF11A system worked as a possible marker for disease progression and aggressiveness in breast cancer[22]. Similarly, the TNFRSF11A system is a potential target for BRCA1 mutation carriers to prevent breast cancer [23]. Yunxian Yu et al. suggested that genes in the TAC1 pathway participate in the development of colorectal cancer [24]. Hypermethylation of TAC1 promoter is not only a risk factors for the occurrence and development of esophageal adenocarcinoma, but also is a tissue biomarker of esophageal adenocarcinoma with poor prognosis[25]. TAC1 and TACR1 played an important role in the development of squamous cell carcinoma of the head and neck through CpG hypermethylation as well, which may be an important biomarker of the carcinoma[26].

There are several limitations in the current study. First, our data was based on the open datasets of GEO and was not validated in prospective clinical trials. Second, some other important clinicopathological characteristics which can influence prognosis, such as tumor differentiation and positive lymph node ratio was not included due to the limitation in the study method.

In conclusion, the study first developed an early relapse-associated gene signature that can robustly divide GC patients into two groups with low and high risks of distinctive prognosis. This classifier may help to select GC patients with poor prognosis who require more frequent follow-up and more aggressive therapeutic intervention.

GC: Gastric cancer; GEO: Gene Expression Omnibus; TCGA: The Cancer Genome Atlas; HR: Hazard ratio; CI: Confidence interval; LNR: Lymph node ratio; AUC: Area under curve; PSM: Propensity score matching; DEGs: Differentially expressed genes; LIMMA: Linear models for microarray data; ROC: Receiver operating characteristic;

Acknowledgements

Not applicable.

Authors’ contributions

Pengchao Wang, Qingyu Song and Qingcheng Xia designed this study. Pengchao Wang and Qingyu Song were responsible for the interpretation of the results. Shipei Qiu , Xin Tang , Pengchao Wang and Ming Lu processed the data processing and performed the statistical analysis. Pengchao Wang, Qingyu Song, and Qingcheng Xia wrote the manuscript. Qinghong Zhao and Xiang Ma revised and improved the manuscript. All authors approved the submission.

Funding

Not applicable.

Availability of data and materials

The datasets included in the current study are openly available in GSE62254 set and GSE15459 set at http://www.ncbi.nlm.nih. gov/geo/.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that there are no competing interests.

Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018; 68:394–424.
Yamaguchi K, Yoshida K, Tanahashi T, Takahashi T, Matsuhashi N, Tanaka Y, et al. The long-term survival of stage IV gastric cancer patients with conversion therapy. Gastric Cancer. 2018; 21:315–323.
Spolverato G, Ejaz A, Kim Y, Squires MH, Poultsides GA, Fields RC, et al. Rates and patterns of recurrence after curative intent resection for gastric cancer: a United States multi-institutional analysis. J Am Coll Surg. 2014; 219:664–675.
Kanaji S, Suzuki S, Matsuda Y, Hasegawa H, Yamamoto M, Yamashita K, et al. Recent updates in perioperative chemotherapy and recurrence pattern of gastric cancer. Ann Gastroenterol Surg. 2018; 2:400–405.
Pattison S, Mitchell C, Lade S, Leong T, Busuttil RA, Boussioutas A. Early relapses after adjuvant chemotherapy suggests primary chemoresistance in diffuse gastric cancer. PLoS One. 2017; 12:e0183891.
Suh DD, Oh ST, Yook JH, Kim BS, Kim BS. Differences in the prognosis of early gastric cancer according to sex and age. Therap Adv Gastroenterol. 2017; 10:219–229.
Kim YJ, Chung WC, Cho IH, Kim J, Kim S. Prognostic effect of different etiologies in patients with gastric cardia cancer. Medicine (Baltimore). 2019; 98:e18397.
Li Z, Wang Y, Ying X, Shan F, Wu Z, Zhang L, et al. Different prognostic implication of ypTNM stage and pTNM stage for gastric cancer: a propensity score-matched analysis. BMC Cancer. 2019; 19:80.
Hou JY, Wang YG, Ma SJ, Yang BY, Li QP. Identification of a prognostic 5-Gene expression signature for gastric cancer. J Cancer Res Clin Oncol. 2017; 143:619–629.
Song P, Jiang B, Liu Z, Ding J, Liu S, Guan W. A three-lncRNA expression signature associated with the prognosis of gastric cancer patients. Cancer Med. 2017; 6:1154–1164.
Dai W, Li Y, Mo S, Feng Y, Zhang L, Xu Y, et al. A robust gene signature for the prediction of early relapse in stage I-III colon cancer. Mol Oncol. 2018; 12:463–475.
Zhu X, Tian X, Yu C, Shen C, Yan T, Hong J, et al. A long non-coding RNA signature to improve prognosis prediction of gastric cancer. Mol Cancer. 2016; 15:60.
Zhang Y, Li J, Yu J, Liu H, Shen Z, Ye G, et al. Circular RNAs signature predicts the early recurrence of stage III gastric cancer after radical surgery. Oncotarget. 2017; 8:22936–22943.
Cristescu R, Lee J, Nebozhyn M, Kim KM, Ting JC, Wong SS, et al. Molecular analysis of gastric cancer identifies subtypes associated with distinct clinical outcomes. Nat Med. 2015; 21:449–456.
Ooi CH, Ivanova T, Wu J, Lee M, Tan IB, Tao J, et al. Oncogenic pathway combinations predict clinical prognosis in gastric cancer. PLoS Genet. 2009; 5:e1000676.
Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, et al. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics. 2003; 4:249–264.
Imaoka H, Toiyama Y, Okigami M, Yasuda H, Saigusa S, Ohi M, et al. Circulating microRNA-203 predicts metastases, early recurrence, and poor prognosis in human gastric cancer. Gastric Cancer. 2016; 19:744–753.
Wakatsuki K, Matsumoto S, Migita K, Kunishige T, Nakade H, Miyao S, et al. Risk Factors and Risk Scores for Predicting Early Recurrence After Curative Gastrectomy in Patients with Stage III Gastric Cancer. J Gastrointest Surg. 2019.
Ma M, Xiao H, Li L, Yin X, Zhou H, Quan H, et al. Development and validation of a prognostic nomogram for predicting early recurrence after curative resection of stage II/III gastric cancer. World J Surg Oncol. 2019; 17:223.
Wu PL, He YF, Yao HH, Hu B. Martrilin-3 (MATN3) Overexpression in Gastric Adenocarcinoma and its Prognostic Significance. Med Sci Monit. 2018; 24:348–355.
Yun JW, Lee S, Ryu D, Park S, Park WY, Joung JG, et al. Biomarkers Associated with Tumor Heterogeneity in Prostate Cancer. Transl Oncol. 2019; 12:43–48.
Sirinian C, Papanastasiou AD, Schizas M, Spella M, Stathopoulos GT, Repanti M, et al. RANK-c attenuates aggressive properties of ER-negative breast cancer by inhibiting NF-kappaB activation and EGFR signaling. Oncogene. 2018; 37:5101–5114.
Nolan E, Vaillant F, Branstetter D, Pal B, Giner G, Whitehead L, et al. RANK ligand as a potential target for breast cancer prevention in BRCA1-mutation carriers. Nat Med. 2016; 22:933–939.
Yu Y, Pan Y, Jin M, Zhang M, Zhang S, Li Q, et al. Association of genetic variants in tachykinins pathway genes with colorectal cancer risk. Int J Colorectal Dis. 2012; 27:1429–1436.
Jin Z, Olaru A, Yang J, Sato F, Cheng Y, Kan T, et al. Hypermethylation of tachykinin-1 is a potential biomarker in human esophageal cancer. Clin Cancer Res. 2007; 13:6293–6300.
Misawa K, Kanazawa T, Misawa Y, Imai A, Uehara T, Mochizuki D, et al. Frequent promoter hypermethylation of tachykinin-1 and tachykinin receptor type 1 is a potential biomarker for head and neck cancer. J Cancer Res Clin Oncol. 2013; 139:879–889.

Table 1 Univariable and multivariable Cox regression analysis in gastric cancer

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

An early relapse-associated gene signature optimizes prognostic prediction for gastric cancer patients

Status:

Version 1

Abstract

Figures

Introduction

Materials And Methods

Result

Discussion

Conclusion

Abbreviations

Declarations

References

Table

Additional Declarations

Status:

Version 1