Identification of Potential Novel Prognosis-Related Genes Through Transcriptome Sequencing, Bioinformatics Analysis, and Clinical Validation in Acute Myeloid Leukemia

Abstract

Background: Acute Myeloid Leukemia (AML) is a complex and heterogeneous hematologic malignancy. However, the function of prognosis-related signature genes in AML remains unclear. Methods: In the current study, transcriptome sequencing was performed on 15 clinical samples, differentially expressed RNAs were identified using R software. The potential interactions network was constructed by using the common genes between target genes of differentially expressed miRNAs with transcriptome sequencing results. Functional and pathway enrichment analysis was performed to identify candidate gene-mediated aberrant signaling pathways. Hub genes were identified by the cytohubba plugin in Cytoscape software, which then expanded the potential interactions regulatory module for hub genes. TCGA-LAML clinical data were used for the prognostic analysis of the hub genes in the regulatory network, and GVSA analysis was used to identify the immune signature of prognosis-related hub genes. qRT-PCR was used to verify the expression of hub genes in independent clinical samples. Results: We obtained 1,610 differentially expressed lncRNAs, 233 differentially expressed miRNAs, and 2,217 differentially expressed mRNAs from transcriptome sequencing. The potential interactions network is constructed by 12 lncRNAs, 25 miRNAs, and 692 mRNAs. Subsequently, a sub-network including 15 miRNAs as well as 12 lncRNAs was created based on the expanded regulatory modules of 25 key genes. The prognostic analysis results show that CCL5 and lncRNA UCA1 was a significant impact on the prognosis of AML. Besides, we found three potential interactions networks such as lncRNA UCA1/hsa-miR-16-5p/COL4A5, lncRNA UCA1/hsa-miR-16-5p/SPARC, and lncRNA SNORA27/hsa-miR-17-5p/CCL5 may play an important role in AML. Furthermore, the evaluation of the immune infiltration shows that CCL5 is positively correlated with various immune signatures, and lncRNA UCA1 is negatively correlated with the immune signatures. Finally, the result of qRT-PCR showed that CCL5 is down-regulated and lncRNA UCA1 is up-regulated in AML samples separately. Conclusions: In conclusion, we propose that CCL5 and lncRNA UCA1 could be recognized biomarkers for predicting survival prognosis based on constructing competing endogenous RNAs in AML, which will provide us novel insight into developing novel prognostic, diagnostic, and therapeutic for AML.

Keywords: Acute Myeloid Leukemia; Competing endogenous RNA; bioinformatics; prognosis-related genes; transcriptome sequencing (RNA-seq).

FIGURE 1

The heatmap and volcano plot of DEmRNAs, DEmiRNAs, and DElncRNAs (A) The heatmap and volcano plot of DEmRNAs (B) the heatmap and volcano plot of DEmiRNAs (C) the heatmap and volcano plot of DElncRNAs. The heatmap shows the clustering results of mRNA, miRNA, and lncRNA expression between AML and normal samples. The volcano plot shows the differential analysis results of transcriptome sequencing. The green and red dots indicate the down-regulated and up-regulated differentially expressed genes that reached |log2 FC|> 2 and P-value <0.05, respectively, and the blue dots indicate that the differentially expressed genes are not meet the screening criteria.

FIGURE 2

The Venn diagram of interacted mRNAs and lncRNAs (A) The Venn diagram of interacted mRNA between DEmRNA and DEmiRNA predicted targets (B) The Venn diagram of interacted lncRNA between DElncRNA and DEmiRNA predicted targets.

FIGURE 3

GO and KEGG analysis of hub mRNAs in the potential interactions network (A)Top 10 biological processes (B) Top 10 cellular components (C) Top 10 molecular functions; The circles in the outer circle of the circle plot show the logFC values of genes for each term using scatter. Red circles indicate upregulation, and blue indicates downregulation (D) Results of the top 10 enriched KEGG pathways.

FIGURE 4

Construction of protein-protein interaction (PPI) network in the potential interactions network (A) The hub genes PPI network (B) The top 25 hub genes interactions. The darker color (red) of the mRNAs represents the gene with a higher centrality in the interaction network.

FIGURE 5

The hub genes interacted with miRNAs and lncRNAs in the potential interactions network (A) The potential interactions network of lncRNA-miRNA-mRNA regulation relationships. The green dot represents hub mRNAs in the potential interactions network; the yellow triangle represents the interacted miRNAs in the potential interactions network; the red rhombus represents the interacted lncRNAs in the potential interactions network (B) The prognosis-related genes sub-network of lncRNA-miRNA-mRNA regulatory relationships.

FIGURE 6

The univariate and multivariate Cox regression analysis on hub mRNAs (A) The result of mRNAs Cox regression analysis showed that the clinical information such as cytogenetics risk, age, and leukocyte could be identified as the risk factors on AML, the genetic features including CCL5, CXCL5, and THBS1 also could be recognized as the risk factors on AML, and GNG2 is the protective factor on AML (B) The result of multivariate Cox regression analysis also showed that the clinical information such as cytogenetics risk, age, and leukocyte could be identified as the risk factors on AML, and the genetic features including ADCY1, CCL5 could be recognized as the risk factors on AML. However, CXCR6 and COL6A2 are the protective factors on AML.

FIGURE 7

The univariate and multivariate Cox regression analysis on lncRNAs (A) The result of univariate Cox regression analysis showed that the clinical information such as cytogenetics risk, age, and leukocyte could be identified as the risk factors on AML, and the genetic features including lncRNA SNORA31 and lncRNA UCA1 are recognized as the protective factors in AML (B) The result of multivariate Cox regression analysis also showed that the clinical information such as cytogenetics risk, age, and leukocyte could be identified as the risk factors on AML, the genetic features including lncRNA FOXO3B, lncRNA XIST could be recognized as the risk factors on AML. However, lncRNA UCA1 is the protective factor on AML.

FIGURE 8

K-M survival analysis of hub mRNAs and lncRNAs in the potential interactions network. The result showed that high expression of THBS1, CCL5, and CXCL5 were significantly associated with poor prognosis in AML, while high expression of CXCL12, HGF, GNG2, lncRNA UCA1, and lncRNA SNPRA31 were significantly associated with good prognosis (p <0.05).

FIGURE 9

The Nomogram of risk factors influencing the prognostic outcome (A) The 1, 2, and 3 years overall survival of AML was predicting in Nomogram (B) The calibration curve for the overall survival nomogram. A grey diagonal line shows the ideal Nomogram, the red, green, and blue lines represent the nomograms observed at 1, 2, and 3 years, respectively.

FIGURE 10

The ROC curve for analyzing the diagnostic value of hub mRNAs and lncRNAs. The result showed that both CCL5 (AUC = 0.836, 95% CI: 0.773–0.899) and lncRNA UCA1 (AUC = 0.696, 95% CI: 0.581–0.811) have good diagnostic value in AML. Furthermore, the diagnostic value’s comparison results between CCL5 and lncRNA UCA1 represent better diagnostic sensitivity and specificity in CCL5 (P <0.05).

FIGURE 11

The correlations of hub gene expression levels with the tumor microenvironment (TME) (A) A positive correlation of CCL5 expression (Normalized expression levels) and immune and stromal scores (B) Negative correlation between lncRNA UCA1 expression (Normalized expression levels) and immune score.

FIGURE 12

Association between hub genes and immune signatures in AML (A) The expression of CCL5 (Normalized expression levels) exhibits a significant positive correlation with ten immune cells (Macrophages, NK cells, CD8⁺ T cells, CAFs, Treg, CD4⁺ regulatory T cells, MDSC, TAM, B cells, and Th17). The results of Spearman’s correlation test are shown as P <0.05 (B) The expression of UCA1 (Normalized expression levels) exhibits a significant positive correlation with two immune cells (CAFs and CD8 cells); besides, UCA1 also indicates a significant negative correlation with three immune cells (TAM, MDSC, and M2 macrophages). The results of Spearman’s correlation test are shown as P <0.05.

FIGURE 13

Differential expression of pivotal genes in clinical samples between AML and healthy individuals. Clinical validation results showed that the expression of CCL5 and lncRNA UCA1 was significantly lower and higher in primary AML, respectively. ***, p <0.001.