Transcription Elongation Factor A-like 7 is a Stemness-Related Prognostic Factor in Gastric Cancer


 Background: Stemness is described as the potential for self-renewal and differentiation from the cell-of-origin. A previous study calculated the mRNA expression-based stemness index (mRNAsi) based on a one-class logistic regression machine learning algorithm for describing stemness features of cancer. We aim to identify stemness-related prognostic genes in gastric cancer (GC) based on mRNAsi using bioinformatics analysis.Methods: The WGCNA analysis was performed to find the relevant gene modules to mRNAsi. Gene ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathways annotation analysis were performed on genes in blue module. The overall survival analysis, univariate Cox regression analysis and the least absolute shrinkage and selection operator (LASSO) regression model were used to identify prognostic genes highly associated with survival. The multivariate Cox regression analysis was performed to analysis prognostic factors. The nomogram was constructed according to the result of multivariate analysis. qPCR, Western Blot and IHC staining were appliedResults: The mRNAsi of tumors is higher than normal tissues, and there was a significant difference in overall survival (OS) between the high and low mRNAsi GC groups. TCEAL7 was selected to be the key gene associated with mRNAsi and prognosis according to the result of least absolute shrinkage and selection operator (LASSO) Cox regression. The expression level of TCEAL7 was lower in tumors than in normal tissues, but high TCEAL7 level group showed a worse OS than low TCEAL7 level group in GC. Based on the result of multivariable Cox regression analysis which including TCEAL7 and clinical characteristics, a nomogram for predicting GC 1-, 3-, and 5-year survival was established. The C-index and the AUC (Area Under Curve) of the model indicated that the model has a good discrimination ability. Additionally, the calibration curves of 3- and 5-year OS rates showed the model fits well. The experimental validation of the expression of TCEAL7 in GC and normal tissues were consistent with the above.Conclusions: In summary, we verified mRNAsi was associated with the prognosis of GC patients. And TCEAL7 was finally identified as the key gene correlated with stemness features and prognosis in GC.


Introduction
As an important disease worldwide, gastric cancer (GC) is the 5th most common cancer and 3rd most common cause of cancer-related death, with about 784,000 deaths globally in 2018 [1]. Though surgery and chemotherapy improve survival and quality of life for patients with advanced gastric cancer, overall survival (OS) in GC patients is still poor due to recurrence, metastasis, and chemoresistance [2]. Stemness is described as the potential for self-renewal and differentiation from the cell-of-origin. The progression of cancer involves gradual loss of differentiated phenotype and the acquisition of stemness features [3].
Previous studies have shown that tumors with more stemness features are more apt to result in metastasis and chemoresistance, leading to disease progression and poor prognosis [3][4][5]. Therefore, it would be helpful to identify key genes associated with cancer stemness features to predict and improve prognosis.
In a previous study, Malta et al. applied a one-class logistic regression machine learning algorithm (OCLR) to extract transcriptomic and epigenetic feature sets derived from non-transformed pluripotent stem cells and their differentiated progeny [6]. Then, they calculated the mRNA expression-based stemness index (mRNAsi), DNA methylation-based stemness index (mDNAsi), and mRNAsi and the epigenetic regulation based-index (EREG-mRNAsi) based on the OCLR-derived feature sets. This provided a new way for describing stemness features of cancer. Many scholars have identi ed stemness featuresassociated genes and possible signal pathways in bladder cancer (BLCA), breast cancer (BRCA), and lung adenocarcinoma (LUAD) based on the OCLR-based stemness index [7][8][9]. Preceding studies also suggested the importance of mRNAsi-related genes in risk strati cation and survival prediction in lowergrade glioma (LGG) and gastric cancer [10,11]. However, the predictive power of the nine-mRNAsi-relatedgene risk model in the study of gastric cancer was not very satisfactory compared to that of LGG, and this gene model also lacked external validation.
The transcription elongation factor A-like 7 (TCEAL7) is a member of the transcription elongation factor A (SII)-like gene family, which was rst cloned as a proapoptotic nuclear protein and shown to function as a tumor suppressor gene contributing to the activation of oncogenic pathways in ovarian cancer [12][13][14][15].
Besides, decreased expression of TCEAL7 has been reported in different types of cancer cell lines including endometrial cancer, glioblastoma, non-small cell lung cancer and gastric cancer [16][17][18][19][20], but the role of TCEAL7 in these cancers is still not very clear.
In this study, we explored the mRNAsi level in different clinical features and veri ed its relationship with prognosis. By applying weighted gene co-expression network analysis (WGCNA) and least absolute shrinkage and selection operator (LASSO) Cox regression analysis based on The Cancer Genome Atlas (TCGA) database, TCEAL7 was identi ed as the most relevant gene to mRNAsi and prognosis, which showed a good prognostic value in GC. The microarray GSE66229 (composed of GSE62254 and GSE66222) dataset from the Gene Expression Omnibus (GEO) database was utilized to further com rm the expression and prognostic value of TCEAL7.

Data Processing
The RNA-seq data of 32 normal and 375 stomach adenocarcinoma (STAD) samples from The Cancer Genome Atlas (TCGA) database and 174 normal gastric tissue samples from the Genotype-Tissue Expression (GTEx) project were downloaded from the University of California Santa Cruz (UCSC) Xena website. The clinical data of tumor samples were also obtained, which included: gender, age, gender, TNM stages, survival time, and vital status. The RNA-seq expression pro les were quanti ed by transcripts per kilobase of exon model per million mapped reads (TPM) normalized estimation and log2-based transformation. Next, R package "sva" was used to remove the batch effects by the methods of ComBat. Then, differentially expressed genes (DEGs) were identi ed by "limma" package of R software under the selection criteria of absolute value of the log2-transformed fold change (FC) > 1, the adjusted P-value < 0.05, and false discovery rate (FDR) < 0.05.

The Expression Level and Survival Analysis of mRNAsi
In a previous study [6], the gene expression-based stemness index of the STAD samples from TCGA database was calculated by a one class logistic regression machine learning algorithm (OCLR), which ranges from zero (no gene expression) to one (complete gene expression). The mRNAsi was obtained from this research. Boxplots were drawn to observe mRNAsi expression in different clinical features. The "surv_cutpoint" function of "survminer" package was used to nd the optimal cutoff value for the dichotomy of mRNAsi related to overall survival. The Kaplan Meier analysis was performed for samples with high and low mRNAsi.
Weighted Gene Co-expression Network Analysis (WGCNA) The WGCNA analysis was performed by the "WGCNA" package [21] to build a co-expression network of DEGs. First, the co-expression similarity matrix and Pearson correlation matrix were constructed. Next, the power function a mn = |c mn |β (c mn = Pearson correlation between gene m and gene n; a mn = adjacency between gene m and gene n) was used to construct a weighted adjacency matrix. The parameter β was set to achieve a scale-free co-expression network. Then, a topological overlap matrix (TOM) was built to measure the network connectivity of genes de ned as the sum of adjacent genes generated by all other networks. Finally, a hierarchical clustering tree and gene modules were performed based on TOM-based dissimilarity measurements. The minimum module size was 30 and cut height was 0.2.
The mRNAsi and epigenetically regulated mRNAsi (EREG-mRNAsi) were selected as the external traits to identify the mRNAsi-related modules and genes. The relationship between the modules and mRNAsi was calculated, and the most relevant module was selected for subsequent analysis to explore key genes.
Gene signi cance (GS, the correlation between genes and sample traits) and module membership (MM, correlation between the module's genes and gene expression pro les) for each gene in the module were calculated. The thresholds for screening key genes in the module were as follows: cor. gene MM > 0.8 and cor. gene GS > 0.7.

Functional Enrichment Annotation and Pathway Analysis of Key Genes
Gene ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathways annotation analysis were performed by clusterPro ler package [22] to explore the biological functions of the key genes. The statistical signi cance values were set as follows: P-value < 0.05, and false discovery rate (FDR) < 0.05.

LASSO Cox Regression Analysis and Construction of TCEAL7 Expression-based Model
The univariate Cox regression analysis was used to identify prognostic genes highly associated with survival. Next, the least absolute shrinkage and selection operator (LASSO) regression model was performed to further optimize the prognostic genes, and the TCEAL7 was nally selected as the prognostic gene. As mentioned above in the mRNAsi analysis, boxplots were drawn to observe the expression of TCEAL7 in different clinical features, and the Kaplan Meier analysis was performed for samples with high and low expression level. Multivariate Cox regression analysis was performed to analysis prognostic factors in the model, including age, gender, TNM stages, TCEAL7. The nomogram was constructed according to the result of multivariate analysis using "rms" package.
The Cox model validation was performed by AUC (Area Under Curve) and C-index for discrimination ability, calibration curves for calibration ability. Bootstraps with 1000 resamples were adopted to decrease over t bias.
To compare the accuracy and discrimination of different models, the net reclassi cation improvement (NRI), the integrated discrimination improvement (IDI), and the median improvement in risk score (MIRS) were applied by using "survIDINRI" package [23].

Data Validation
To further verify the expression and prognostic value of TCEAL-7, the microarray GSE66229 (composed of GSE62254 and GSE66222) dataset was downloaded from the Gene Expression Omnibus (GEO) database, which had 100 normal and 300 tumor samples. Boxplots for expression of TCEAL-7 in different clinical features and the Kaplan Meier analysis of high and low expression group were performed. We also built Cox regression models to manifest the prognostic value of TCEAL-7.
Clinical specimens 5 pairs of GC tissues and paired adjacent normal tissues were obtained from patients who underwent surgery at Nanjing Medical University's First A liated Hospital in China. All samples were con rmed by histopathology. The samples were frozen in liquid nitrogen and then stored at -80 ℃ until use. Written consent was obtained from all patients. The study were approved by First A liated Hospital of Nanjing Medical University Ethics Committee.

Cell culture
Human GC cell lines (SGC-7901, HGC-27, and AGS) and the normal human gastric epithelial cell line (GES-1) were purchased from the American Type Culture Collection (ATCC, USA). All these cells were cultured at 37 °C in a humidi ed atmosphere with 5% CO2, and cultured in RPMI 1640 medium supplemented with 1% penicillin/streptomycin, 10% fetal bovine serum (FBS). All culture medium reagents were obtained from Gibco, Brazil.
Chemiluminescence western blotting detection system (proteintech) was used for protein detection.

Statistical Analysis
The statistical analysis was carried out using the R software (version 3.6.2). Wilcox test was used to assess the difference in mRNAsi and TCEAL7 between tumor samples and normal samples. Kruskal-Wallis test was applied to explore the correlation in mRNAsi and TCEAL7 versus clinical characteristics.
Pearson regression was applied to illustrate the correlation between TCEAL7 and mRNAsi. Survival was analyzed using the Kaplan-Meier method and log-rank test. Statistical signi cance was set at P < 0.05.

Data Processing for DEGs
The work ow of our study was brie y summarized in Figure 1. The RNA-seq data of 174 normal gastric tissue samples from GTEx project, 32 normal and 375 stomach adenocarcinoma (STAD) tissue samples from the TCGA database were included in the current study. We con rmed the necessity of normalization and batch effect removal before identi cation of DEGs ( Supplementary Fig. S1a-b). And it was not di cult to nd out the TCGA and GTEx samples performed better after normalization and batch effect removal ( Supplementary Fig. S1c-d). Then 5416 DEGs were identi ed between normal and tumor tissue samples ( Supplementary Fig. S1e), of which 3744 of them were upregulated and 1672 were downregulated, using the criteria of the absolute value of the log2-transformed fold change (FC) > 1 and the adjusted P-value < 0.05 (Fig. 2a).

The Expression Level and Survival Analysis of mRNAsi
The mRNAsi of normal tissues and tumors were respectively illustrated in Figure 2b, with the expression level of tumors obviously higher than that of normal tissues. According to the calculated mRNAsi, all the GC samples were further assigned to either a high mRNAsi score group or a low mRNAsi score group by a cutoff value of 0.42. As presented in Figure 2c, the OS showed notable difference between the two groups (P < 0.001) by applying Kaplan-Meier survival analysis. Furthermore, the expression level of mRNAsi was compared among T stages, lymphatic metastasis, M stages, and AJCC stages. Howesver, as shown in Figures 2d-g, there was no statistically signi cant difference among these subgroups.

WGCNA of DEGs
After all the data process was completed, a correlation analysis of 5416 DEGs was conducted. To assure a scale-free topology model, the soft threshold power of β was set at 5 ( Supplementary Fig. S2b). Next, eleven various modules (magenta, red, turquoise, brown, yellow, green, blue, purple, black, pink, and gray) were validated by a clustering analysis (module size>30 and cut height>0.2), with each color module presenting the same gene expression pattern (Fig. 3a). And the correlation relationships between every two modules were displayed in the gures (Supplementary Fig. S2a, c). Under the condition of fold enrichment > 1 and P-value < 0.05, a total of eight modules among all the eleven modules were validated signi cantly associated with mRNAsi. The green, blue, purple, black and pink modules were correlated negatively with mRNAsi (ME green : r =-0.19, P = 2E -4 , ME blue : r = -0.79, P = 6E -79 , ME purple : r = 0.52, P = 1E -26 , ME black : r = -0.39, P = 2E -14 , ME pink : r = -0.64, P = 4E -43 ). The turquoise, brown and yellow modules were correlated positively to mRNAsi (ME turquoise : r =0.15, P =0.004, ME brown : r = 0.7, P =2E -55 , ME yellow : r = 0.38, P =6E -14 ) (Fig. 3b). According to the module-trait relationships, the blue module was identi ed as the most signi cantly relevant module to mRNAsi, with the highest correlation value (r = -0.79, P = 6E -79 ). The blue module was therefore selected for subsequent analyses to explore key genes. And based on threshold limits (cor. gene GS > 0.7 and cor. Gene MM > 0.8), 55 out of 504 hub genes were identi ed after selection in the blue module (Fig. 3c).

Function Annotation of the Key Genes in the Blue Module
To elaborate the de nite functions of key genes in blue module, the KEGG signaling pathway annotation analysis was applied. As a result, the key genes in blue module were con rmed closely related to 18 pathways which were associated with "Signal transduction", "Endocrine system", "Circulatory system", "Digestive system", "Cardiovascular disease", "Cell growth and death" and "Excretory system" ( Supplementary Fig. S3a). Then 9 biological processes (BP) terms, 9 cellular components (CC) terms and 6 molecular functions (MF) terms were enriched as a total 24 GO processes based on the Gene ontology (GO) enrichment analysis, with GO terms consisting of "regulation of cardiac conduction," "potassium ion transport," "Z disc" and "n ATPase binding" ranked at the top of the list (Supplementary Fig. S3b). These results could preliminarily explain the negative correlation between mRNAsi and the key genes in blue module.

LASSO Cox Regression Analysis
By applying univariable Cox regression analysis, 23 out of 55 genes were identi ed as the prognostic genes highly associated with survival, which were subjected to LASSO Cox regression. Then, TCEAL7 was selected to be the key gene associated with mRNAsi and survival according to the result of LASSO Cox regression ( Supplementary Fig. S4a, b).

The Expression Level and Survival Analysis of TCEAL7
As shown in the Fig. 4a, the expression level of TCEAL7 was obviously lower in tumors than in normal tissues (P < 0.001). Moreover, TCEAL7 was validated at the lowest expression level in T1 among all T stages of GC (P < 0.001, P-value for T1 vs T2, T1 vs T3, and T1 vs T4 were all < 0.001) and so was in AJCC classi cation (P = 0.003, P-value for stage I vs stage II, stage I vs stage III, and stage I vs stage IV were all < 0.05) (Fig. 4d, g). But in lymphatic metastasis (P = 0.203) and M stage groups (P = 0.301), the expression level showed no statistically signi cant difference (Fig. 4e, f). Simultaneously, Pearson regression was applied to illustrated TCEAL7 had a negative correlation with mRNAsi (Pearson r = -0.8404, P < 0.001) (Fig. 4c). Kaplan-Meier survival analysis found a signi cant difference in OS between the high TCEAL7 level group and the low TCEAL7 level group (Fig. 4b, cutoff value = 1.82, P = 003).
To further investigate this gene, we also analyzed the expression level of TCEAL7 in other tumors from TCGA database. It was shown that the expression level of TCEAL7 was lower in tumors than in normal tissues in most types of tumors. (Supplementary Fig. S5a), which was consistent with the current results.
All these results con rmed the prognostic value of TCEAL7. Nevertheless, it was truly worth thinking that the expression level of TCEAL7 was lower in tumors while high expression level of TCEAL7 was associated with the worse survival outcome in GC.
Next, other two Cox models were constructed to further assess the accuracy and discrimination of Cox model 2, with age, gender, TNM stages enrolled in Cox model 1 and age, gender, TCEAL7 enrolled in Cox model 3. As shown in Figure 5d, the AUC for 1-to 6-year OS of Cox model 2 is higher than Cox 1 and Cox model 3.

Validation of the Prognostic Value of TCEAL7
To further verify the expression level and the prognostic value of TCEAL-7, the microarray GSE66229 dataset was obtained, which had 100 normal and 300 tumor samples. Consistent with the results above, the expression level of TCEAL7 was signi cantly lower in tumors than in normal tissues (P < 0.001, Fig.  6a), and the OS of high TCEAL7 level group was obviously worse than that of low TCEAL7 level group (Fig. 6b, cutoff value = 1.54, P < 0.001). It seemed that lower TCEAL7 expression level was associated with lower TNM stages and AJCC stages ( Supplementary Fig. S6a-d). Correspondingly, the AUC of the three models were shown in Figure 6e to explain the accuracy and discrimination of models. And the Cindex was 0.646 (95% CI: 0.6-0.692) for model 1, 0.663 (95% CI: 0.619-0.708) for model 2 and 0. 0.596 (95% CI: 0.55-0.643) for model 3. These results veri ed the prognostic value of TCEAL7. Moreover, the calibration plots of model 2 were also established (Fig. 6d, e).

Validation of the TCEAL7 expression in GC tissues and cells
The expression of TCEAL7 was veri ed in GES-1 cells and in another three GC cell lines, including SGC-7901, HGC27 and AGS. The gastric cancer cell lines had a relatively low expression level of TCEAL7 compared with the GES-1 cells, both at the mRNA and protein levels ( Fig. 7a-b). To further con rmed the low expression of TCEAL7 in GC tissues, IHC staining was employed and the results obtained were consistent (Fig. 7c-d).

Discussion
Since the pan-cancer cohorts study of Malta et al. provided a new way for describing stemness features of cancer, many studies have focuesed on stemness features-associated key genes and possible signal pathways in cancers based on the OCLR-based stemness index [7][8][9][10][11]. These studies veri ed that tumor samples had a higher stemness index when compared with normal samples in BLCA, BRCA, LUAD, LGG, and STAD. However, the high level of mRNAsi was not always related to a poor prognosis. In our study, we analyzed the mRNAsi expression in different clinical features and its relationship with survival in GC patients. It showed that the high level of mRNAsi might represent a good prognosis. In the pan-cancer cohorts study[6], a trend toward higher stemness index with better outcome was also found in COAD (Colon adenocarcinoma) and LGG. Zheng et al. used a relative expression orderings (REOs)-based way to describe the stemness features of cancers, and in this study, there was no signi cant relationship was found between REO-based stemness index and prognosis of STAD with Hazard Ratio (HR) value < 1 but P-vlue > 0.05 [24]. Possible explanations for these results might be that whether OCLR-based or REO-based stemness index can not accurately describe the stemness features in these cancers or the relationship between cancer stemness features and progression or prognosis of cancers is more complicated than we know, which may be associated with intratumoral heterogeneity. Further studies on stemness features of these cancers will need to be undertaken. In short, we preliminarily demonstrated the prognostic value of OCLR-based stemness index despite the negative correlation between mRNAsi level and prognosis.
WGCNA is a tool to analyze the gene expression pattern in multiple samples, and it can classify those genes with similar expression patterns into clusters and further analyse the correlations between different gene clusters and certain characteristics. By applying WGCNA, we identi ed more than one gene module related to mRNAsi. Different from the study of Chen et al. who selected the highest positive correlation module with mRNAsi [11], the blue module which had the highest correlation value (r = -0.79, P = 6E − 79 ) with mRNAsi was selected and 55 out of 504 hub genes in this module were further analyzed according to the threshold limits. The KEGG signaling pathway annotation analysis and GO enrichment analysis showed that these gene were more associated with some organismal system-speci c functions such as secretion of some digestive juices, which preliminarily explained the negative correlation between the key genes in blue modul and stemness features.
By virtue of Cox regression analysis and LASSO regression, TCEAL7 was nally identi ed as the key gene correlated with mRNAsi and prognosis in GC. Consistent with previous study [19,20], the expression level of TCEAL7 was signi cantly lower in tumors than in normal tissues in two GC cohorts including samples from TCGA database and GSE66229 dataset. However, contrary to the results of miRNAsi mentioned above, OS of high TCEAL7 exprssion level group was worse than low TCEAL7 expression level group in both two GC cohorts. Moreover, a trend that lower TCEAL7 expression level was associated with lower TNM stage and AJCC stage was discovered in the cohorts of GSE66229 dataset. These ndings are unexpected and are contrary to the study of Huang et al. who found low expression of TCEAL7 was associated with worse prognosis of gastric adenocarcinoma [20]. The inconsistency may be explained by differences in races and areas between the cohorts in the two studies. Several reports have shown that TCEAL7 functions as a tumor suppressor gene in glioblastoma, ovarian and gastric cancer [13,16,19]. As a cell death-regulatory protein, TCEAL7 has been reported to be associated with a reduced risk of invasive serous ovarian cancers [12]. Further studies proved that TCEAL7 negatively modulated Myc activity by binding to E-boxes of gene promoters in ovarian cancer [15]. Moreover, TCEAL7 could negatively regulate nuclear factor κB (NF-κB) signaling by modulating the transcriptional activity of NF-κB on its target gene promoters in ovarian cancer cells [13]. In glioblastoma, exo-miR-301a was manifested to activates wnt/βcatenin signaling and promote radiation resistance by targeting TCEAL7 [16]. What's more, Heart and Neural Crest Derivatives Expressed 2 antisense RNA 1 (HAND2-AS1) was found to regulate the migration, invasion, and apoptosis of GC cells by suppressing TCEAL7 expression via targeting miR-769-5p [19]. In this study, we con rmed that the expression level of TCEAL7 was lower in tumors than in normal tissues in most types of tumors. Nevertheless, univariable Cox regression analysis showed that TCEAL7 was considered as a risk factor in BLCA, KIRP, STAD, KICH, and ACC. This result may be explained by the fact that TCEAL7 might plays different roles in different tumors and the prognosis of tumors is affected by various factors including TCEAL7. Therefore, further research should be undertaken to investigate the function of TCEAL7 and its correlation with stemness features in GC.
To explore the prognostic value of TCEAL7, three models were constructed to predict the prognosis of GC, with age, gender, TNM stages enrolled in Cox model 1, age, gender, TNM stages, TCEAL7 enrolled in Cox model 2 and age, gender, TCEAL7 expression level enrolled in Cox model 3. Cox model 2 showed better accuracy and discrimination than Cox model 1 and Cox model 3 in both two GC cohorts. Moreover, the accuracy and discrimination of this one gene-based model are comparable with a nine-mRNAsi-relatedgene risk model of a previous study [11]. These results veri ed the prognostic value of TCEAL7, and the TCEAL7 expression based model could be expected to serve as an e cient tool to predict the prognosis of GC patients for clinicians.
There are several limitations in the current study. First of all, as mentioned above, the OCLR-based stemness index was used to describe the stemness features of tumors, which might not accurately describe the stemness features in GC. Next, though this study was based on two cohorts of GC from TCGA database and GSE66229 dataset, validation in cellular experiments, and animal and tissue models are needed. Finally, considering inadequate treatment information in both TCGA database and GSE66229 dataset, the role of TCEAL7 in chemoresistance of GC patients needs to be further explored.

Conclusion
In conclusion, we veri ed mRNAsi was associated with the prognosis of GC patients. And TCEAL7 was nally identi ed as the key gene correlated with stemness features and prognosis in GC, which was demonstrated as a risk factor. Nomogram including TCEAL7 and clinical features showed good accuracy and discrimination for predicting the prognosis of GC patients. Further study should be conducted to con rm our conclusion and investigate the role of TCEAL7 in GC and its correlation with stemness features. G.Z. and X.L.: designing the project; H.Q., X.G., R.M. and W. L.: data analysis and writing the article; Y.Z.: revising the syntax.

Index
Cox model 2  Figure 1 The complete process of stemness-related prognostic genes identi cation and veri cation.       The expression of TCEAL7 was veri ed in GES-1 cells and in another three GC cell lines, including SGC-7901, HGC27 and AGS. The gastric cancer cell lines had a relatively low expression level of TCEAL7 compared with the GES-1 cells, both at the mRNA and protein levels ( Fig. 7a-b). To further con rmed the low expression of TCEAL7 in GC tissues, IHC staining was employed and the results obtained were consistent

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.