The role of ARHGAP9: clinical implication and potential function in acute myeloid leukemia

Background Rho GTPase activating protein 9 (ARHGAP9) is expressed in various types of cancers and can inactivate Rho GTPases that mainly regulate cytoskeletal dynamics. However, the exact role of ARHGAP9 in acute myeloid leukemia (AML) has yet to be clarified. Methods We compared the transcriptional expression, prognosis, differentially expressed genes, functional enrichment, and hub genes in AML patients on the basis of the data published in the following databases: UALCAN, GEPIA, Gene Expression Omnibus, the Human Protein Atlas, Cancer Cell Line Encyclopedia, LinkedOmics, Metascape, and String. Data from the Cancer Genome Atlas database was used to evaluate the correlations between ARHGAP9 expression and various clinicopathological parameters, as well as the significantly different genes associated with ARHGAP9 expression. Results We found that ARHGAP9 expression was higher in the tissues and cell lines extracted from patients with AML than corresponding control tissues and other cancer types. ARHGAP9 overexpression was associated with decreased overall survival (OS) in AML. Compared with the ARHGAP9low group, the ARHGAP9high group, which received only chemotherapy, showed significantly worse OS and event-free survival (EFS); however, no significant difference was observed after treatment with autologous or allogeneic hematopoietic stem cell transplantation (auto/allo-HSCT). The ARHGAP9high patients undergoing auto/allo-HSCT also had a significantly better prognosis with respect to OS and EFS than those receiving only chemotherapy. Most overlapping genes of the significantly different genes and co-expression genes exhibited enriched immune functions, suggesting the immune regulation potential of ARHGAP9 in AML. A total of 32 hub genes were identified from the differentially expressed genes, within which the KIF20A had a significant prognostic value for AML. Conclusions ARHGAP9 overexpression was associated with poor OS in AML patients and can be used as a prognostic biomarker. AML patients with ARHGAP9 overexpression can benefit from auto/allo-HSCT rather than chemotherapy.

chemotherapy intensification, and autologous or allogeneic hematopoietic stem cell transplantation (auto/ allo-HSCT), the outcome for AML patients remains discouraging because of recurrence and refractory [1,2]. Approximately 10% to 40% of younger patients and a larger percentage of patients aged 60 and above have refractory after standard chemotherapy (40%-60%) [3]. Therefore, identifying robust prognostic markers is crucial to provide optimal care for AML patients.
The Rho family of GTPases is a family of small (~ 21 kDa) signaling G proteins, which act as molecular switches and are tightly controlled by guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs) by generating active GTP-binding and inactive GDP-binding proteins, respectively [4,5]. Activated Rho GTPases interact with their downstream effectors to regulate the cytoskeleton of the cell membrane or other cellular compartments [6,7]. Considerable research has supported the importance of Rho GTPases in hematopoiesis and confirmed that Rho GTPases are related to cytoskeleton rearrangement including adhesion, cytokinesis, differentiation, migration, engraftment, aging, and self-renewal in the cellular process [8][9][10][11][12][13][14]. The reorganization of the actin and microtubuleb cytoskeleton is essential for platelet adhesion and thrombus formation to avoid excessive bleeding.
A total of 80 Rho family GAPs have thus far been identified, but fewer than half of them in cancer have been clearly investigated. Rho GTPase activating protein 9 (ARHGAP9) containing RhoGAP, SH3, WW, and PH domains, is a member of the Rho GAPs family. Research indicates that ARHGAP9 suppresses the adhesion of KG-1 (a human leukemia cell line) to fibronectin and collagen via the activation of Cdc42 and Rac1 rather than RhoA [15]. ARHGAP9 is also considered as a MAP kinase docking protein, and the WW domain of ARHGAP9 interacts with the CD domains of Erk2 and p38alpha, leading to the inactivation of MAP kinases [16]. ARHGAP9 inhibits the migration and invasion of hepatocellular carcinoma cell by increasing FOXJ2/Ecadherin expression [17]. By contrast, silencing ARH-GAP9 reduces the proliferation, migration, and invasion of breast and gastric cancer cells in vitro [18,19]. These occurrences suggest that ARHGAP9 plays distinct roles in various physiological conditions or various tissues and cells.
No studies have thus far been conducted to access the expression profile and functions of ARHGAP9 in AML, although ARHGAP9 expression in peripheral blood leukocytes has been demonstrated [15]. In the present study, we investigated ARHGAP9 expression in human AML samples and cell lines, as well as explored its associations with clinicopathological factors. We then we evaluated the prognostic significance of ARHGAP9. We also investigated the differentially expressed genes associated with ARHGAP9 expression and discussed their potential functions in AML.

Analysis of The Human Protein Atlas (HPA) database
The HPA (https ://www.prote inatl as.org/) is user-friendly online server, which contains the human transcriptomic and proteomic data in cells, tissues, and organs from human normal or pathological tissues via RNA sequencing (RNA-Seq) analysis and immunohistochemistry (IHC) [20]. Thus, ARHGAP9 expression was analyzed in multiple cell lines, such as various leukemia cell lines, lung cell lines, breast cell lines, and brain cell lines in HPA by using the search term "ARHGAP9".

The Cancer Cell Line Encyclopedia (CCLE) database analysis
The CCLE project (https ://porta ls.broad insti tute.org/ ccle) is an effort to conduct comprehensive genetic characterization of a large panel of human cancer cell lines from individuals of various lineages and ethnicities [21]. It provides public access to analysis and visualization of mRNA expression, mutation data, DNA methylation, DNA copy number, and histone H3 modification for > 1100 cancer cell lines, such as breast cell lines, gastric cancer cell lines, and AML cell lines. RNA-Seq of RNA expression was used to verify ARHGAP9 expression in various cell lines.

UALCAN database analysis
The UALCAN database (http://ualca n.path.uab.edu/) contains level-3 RNA-seq and clinical data from 31 cancer types selected by the Cancer Genome Atlas (TCGA) [22]. It is an interactive web resource for the in-depth analysis of RNA-Seq expression. The function module of TCGA analysis in the UALCAN database was used in the pan-cancer analysis of ARHGAP9 expression.

GEPIA database analysis
The GEPIA database (http://gepia .cance r-pku.cn) is a comprehensive resource for the systematic analysis of gene expression [23]. This database includes 9736 tumor and 8587 normal tissue samples from the TCGA and the Genotype-Tissue Expression (GTEx) projects. In the function module, single-gene analysis was used to create plots for ARHGAP9 expression in various cancers and normal tissues. The threshold was selected as the default value.

The Gene Expression Omnibus (GEO) database analysis
ARHGAP9 expression profiles in various chromosome abnormalities were acquired from the GEO (https :// www.ncbi.nlm.nih.gov/geo) database. Data on ARH-GAP9 expression of GSE14468 and GSE13159 were downloaded from the profile graph of the GEO2R online program in the GEO according to the gene ID. A total of 200 and 512 AML patients were from GSE14468 and GSE13159, respectively. The data were the raw counts provided by the submitter.

The Cancer Genome Atlas (TCGA) database analysis
We studied 151 AML patients with RNA-seq and clinical data from the TCGA (https ://porta l.gdc.cance r.gov) to analyze the significantly different genes, prognosis, and association between ARHGAP9 expression and different clinicopathologic features [24]. The patient IDs used in the present study are listed in the Additional file 1: Table S1. Specifically, 79 patients received only chemotherapy, whereas 67 patients underwent chemotherapy and auto/allo-HSCT. The main clinical and genetic characteristics of the AML patients are presented in Table 1 and Additional file 1: Table S2. RNA-seq and clinical data are available on the TCGA website. These patients were divided into two groups (ARHGAP9 low and ARHGAP9 high ) based on the median values of the ARHGAP9 transcript by using the RNAseq data from TCGA. The Limma package in R 3.3.3 was used to screen significantly different genes between the ARHGAP9 high and ARHGAP9 low groups in AML. Adjusted P < 0.05 and |log1.2 FC|≥ 1 were used as cut-off values for identifying significantly different genes.

LinkedOmics database analysis
The co-expression genes in correlated with ARHGAP9 expression were analyzed using the LinkedOmics database (http://www.linke domic s.orglo gin.php) [25]. The LinkedOmics database, including mRNA sequencing data from 173 AML patients from the TCGA database, was used to determine ARHGAP9 co-expression in AML. In the LinkFinder module of LinkedOmics, Pearson's correlation coefficient was calculated to analyze the data. Volcano plots displaying the results were generated.
Overlapping genes between significantly different genes and co-expression genes were determined using the website Draw Venn Diagram (http://bioin forma tics. psb.ugent .be/webto ols/Venn/). The overlapping genes were used for the subsequent analysis, including ARH-GAP9 functional enrichment and protein-protein interaction (PPI) analysis.

Metascape
(http://metas cape.org/gp/index .html#/ main/step1 ) is a free gene-list analysis tool for gene functional enrichment analysis [26]. The identified overlapping genes were inputted into the Metascape database for Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, tissues, and disease enrichment analysis.

ARHGAP9 overexpression in AML cell lines
To elucidate the significance of ARHGAP9 expression in AML cells, we first analyzed ARHGAP9 expression based on the RNA-Seq data obtained from cell lines recorded in the CCLE and HPA databases. In the HPA database, the ARHGAP9 mRNA expression levels were higher in AML cell lines, such as HEL, HL60, NB4, and U937, than that in lymphoid cell lines; meanwhile, ARHGAP9 mRNA was almost not expressed in other cell lines representing the brain, breast, lung and so on (Fig. 1a). Moreover, ARHGAP9 showed the highest expression in AML cell lines in the CELL database (Fig. 1b).

ARHGAP9 overexpression in AML
Using the UALCAN and GEPIA databases, we subsequently tested the mRNA expression of ARHGAP9 in different human tumor samples. The mRNA expression levels of ARHGAP9 was the highest among all types of human cancers (Fig. 2a, b). The transcriptional levels of ARHGAP9 in cancers were then compared with those in normal samples by using the GEPIA database. The results indicated that the expression level of ARHGAP9 was significantly upregulated in patients with AML (Fig. 2c). To further study whether ARHGAP9 expression was influenced by different chromosomal abnormalities, we retrieved two microarray data (GSE14468 and GSE13159) from the GEO database and evaluated ARHGAP9 expression among AML patients with major recurrent chromosomal translocations, including inv(16), t(8;21), t(15;17), 11q23, and complex, as well as the normal karyotype. Analysis results of both data sets showed that t(15;17) AML patients exhibited the lowest ARH-GAP9 expression among the patients with cytogenetic abnormalities (Fig. 2d, e). These results for the AML samples corresponded with those in the cell lines.

Relationships between ARHGAP9 and clinicopathological characteristics of patients with AML
In this study, the clinical and molecular characteristics of all patients from the TCGA database are summarized in Table 1. No significant differences in age, sex, BM blasts, WBC, and PB blasts (p > 0.05) were found between ARHGAP9 low and ARHGAP9 high groups (Table 1). Significant differences were determined in the distribution of FAB classification, cytogenetics, risk stratification, and gene mutations. High expression of ARHGAP9 was significantly correlated with FAB-M1 (p = 0.019), whereas low expression of ARHGAP9 was significantly correlated with FAB-M3 (p < 0.001), FAB-M4 (p = 0.011), and FAB-M5 (p = 0.005) in the distribution of FAB classifications. For cytogenetics, ARHGAP9 overexpression was particularly associated with cytogenetically normal AML (CN-AML) (p = 0.026), and low expression of ARHGAP9 was associated with the t(15;17) (p < 0.01) and complex (p = 0.041) subtypes. Moreover, ARHGAP9 low cases tended to be associated with a good prognosis, whereas ARHGAP9 high cases were obviously correlated with intermediate risk. Among the mutated genes, high expression of ARHGAP9 was only correlated with WT1 mutation (Additional file 1: Table S2).

Prognostic values of ARHGAP9 in AML
We investigated whether ARHGAP9 expression was associated with the prognosis of AML patients. The OS against ARHGAP9 expression was evaluated using the GEPIA and LinkedOmics databases. As shown on the GEPIA and LinkedOmics databases, high expression of ARHGAP9 was correlated with poor OS in AML (Fig. 3a,  3b). We further analyzed the survival data from the TCGA database and found ARHGAP9 overexpression was related to shorter event-free survival (EFS) (Fig. 3c), albeit statistical significance was not achieved. No significant difference in OS (Log-rank p = 0.8792) and EFS (p = 0.7039) was found between the ARHGAP9 low and ARHGAP9 high groups in CN-AML (Fig. 4a, b). To determine whether AML patients with high expression of ARHGAP9 could benefit from chemotherapy or auto/allo-HSCT, we divided the tested AML patients into two groups by their treatment regimens. Significantly shorter OS (Log-rank p = 0.0046) and EFS (Log-rank p = 0.0057) were found in AML patients who received with only chemotherapy (Fig. 4c, d). Among the patients who received auto/allo-HSCT, no significant differences in OS (Log-rank p = 0.6830) and EFS (Log-rank p = 0.7368) were found between the ARHGAP9 low and ARHGAP9 high groups (Fig. 4e, f ). Patients undergoing auto/allo-HSCT had higher OS (p < 0.0001) and EFS (p = 0.0015) than those who received only chemotherapy (Fig. 4g, h). Acute promyelocytic leukemia (APL) (FAB M3) is a unique subtype of AML associated with peculiar clinical features and treatment strategies and the prognosis of APL is good. The prognostic values of ARHGAP9 expression in CN-AML, chemotherapy, and auto/allo-HSCT after the exclusion of FAB M3 from the AML cases were the same as those without the exclusion (Additional file 2: Figure S1), which indicated that the aforementioned prognosis results were not subject to the effect of FAB M3.
Overall, ARHGAP9 expression is a poor factor for AML rather than CN-AML. Moreover, chemotherapy alone showed no prognostic influence on ARHGAP9 high cases. And AML patients with high expression of ARH-GAP9 could benefit from auto/allo-HSCT.

ARHGAP9-associated gene analysis between ARHGAP9 high and ARHGAP9 low in AML patients
To further explore the role of ARHGAP9 in AML, we firstly compared the transcriptomes of the ARHGAP9 low and ARHGAP9 high groups base on the TCGA database. A total of 2,948 genes were identified as significantly different between the ARHGAP9 low and ARHGAP9 high groups (p ≤ 0.05, |log1.2 FC|≥ 1, Additional file 1: Table S3); 1173 genes denoted by red circles and 1,775 genes denoted by green circles, were significantly upregulated and downregulated in the ARHGAP9 high group, respectively (Fig. 5a). Subsequently, we analyzed the co-expressed genes in conjunction with the ARHGAP9 genes by using the LinkedOmics database. As shown in Fig. 5b, a total of 3,310 genes represented by dark red dots were positively correlated with ARHGAP9 expression, and 4,268 genes represented by dark green dots were negatively correlated with ARHGAP9 in AML. Among the genes correlated with ARHGAP9 expression, 3,121 co-expression genes were significant correlations with ARHGAP9 in AML (False discovery rate, FDR ≤ 0.05, p ≤ 0.05, and |cor.|≥ 0.3; Additional file 1: Table S4).
Comparison of the significantly different genes and co-expressed genes led to the determination of 963 overlapping genes. These 963 genes contained 537 positively upregulated genes and 426 negatively downregulated genes (Fig. 5c, d and Additional file 1: Table S5). The overlapping genes were used for subsequent studies.

Functional analysis of the overlapping genes
We then sought to investigate the possible biological function of ARHGAP9 in patients with AML. The 963 genes were analyzed by using tools in Metascape to study KEGG and GO annotation. The top 20 clusters of enriched sets are shown in Fig. 6a. ARHGAP9 mediated the function of Rho GTPases, including the regulation of small GTPases mediated signal transduction, regulation of cell adhesion, actin cytoskeleton organization, cell cycle, microtubule cytoskeleton organization,  organelle localization, phagocytosis and regulation of G2/M transition of mitotic cell cycle. Notably, seven clusters belonged to the immune system including leukocyte activation involved in the immune response, signaling by interleukins, the adaptive immune system, regulation of leukocyte-mediated immunity, regulation of myeloid leukocyte-mediated immunity, interferon signaling, and interferon signaling. ARHGAP9 expression was also associated with protein autophosphorylation, regulation of phosphatidylinositol 3-kinase signaling (PI3K), positive regulation of protein kinase activity, CXCR4 pathway, and hemostasis. In addition, the overlapping genes were enriched in blood, spleen, and bone marrow (Fig. 6b), further suggesting immunological function of ARHGAP9 in leukemogenesis. Moreover, among different diseases, the overlapping genes participated primarily in APL (Fig. 6c).

Validation of hub genes
The PPI network was constructed using the String website and the figures were generated by Cytoscape (MCODE plug-in). A total of 963 common genes were imported into the PPI network. We obtained 881 nodes and 5103 edges (Fig. 7a). The most significant module (MCODE score = 29.613) contained 32 genes. These genes are identified in Fig. 7b. Subsequently, OS analysis of the hub genes was performed using the GEPIA database. Of the 32 hub genes, 31 genes exerted no effect on OS in AML patients. Only KIF20A was significantly related to OS in AML (p = 0.02, Fig. 7c). KIF20A in the AML samples was significantly lower than that in normal tissues (Fig. 7d).

Discussion
Rho GAPs inactivate Rho GTPases by the conservative GAP domain that promotes GTP hydrolysis and accelerates intrinsic GTPase activity, limiting the duration of the regulated reaction [29,30]. Numerous reports suggest that the abnormal expression of Rho GAPs is observed in various tissues from patients with cancer and immunological diseases [31][32][33][34]. The expression of Rho GAPs is found to be heterogeneous among different tumors. For instance, ARHGAP30 was downregulated in lung cancer and colorectal cancer but overexpressed in pancreatic Genome-wide genes associated with ARHGAP9 expression. a Volcano plot of different gene-expression profiles between the ARHGAP9 low and ARHGAP9 high groups. b Volcano plots for the analysis of the co-expression genes associated with ARHGAP9 expression (LinkedOmics). c Overlapping genes between positively correlated genes and significantly increased genes. d Overlapping genes between negatively correlated genes and significantly reduced genes cancer [35][36][37]. Low expression of ARHGAP9 was found in hepatocellular carcinoma and bladder cancer, whereas high expression of ARHGAP9 was observed in breast cancer [17,19,38]. In the current study, ARHGAP9 is overexpressed in both AML samples and cell lines compared, relative to that in their normal counterparts [19].
Nonetheless, the low expression of ARHGAP9 was found in t(15; 17) AML. Several studies investigating the role of Rho GAPs in cancer have found that most Rho GAPs were associated with good outcomes in various kinds of solid tumors [33,36,37,39]. ARHGAP9 also showed a good prognosis in bladder cancer, hepatocellular carcinoma, and gastric cancer [17,18,38]. However, we found that, ARHGAP9 overexpression was associated with a poor prognosis in AML. These results suggest that ARHGAP9 can play different roles in various cancers. Meanwhile, DOCK2 belonging to the Rho GEFs family activates GTPase, which has a function contrary to that of Rho GAPs. DOCK2 was also an independent favorable prognostic factor for both EFS and OS in AML [40]. Thus, some regulators of Rho GTPase may have other functions in cancer, in addition to being associated with GTPase. Despite the association of the high expression of ARH-GAP9 with CN-AML, no relationship between ARH-GAP9 expression and prognosis of CN-AML was found in the current study. In addition, several studies have shown that abnormal expression of some genes such as NCALD, IL2RA, and BCL2 are associated with prognosis in AML patients with auto/allo-HSCT and/or chemotherapy [41][42][43]. Our findings showed that ARHGAP9 high groups had poor prognosis in post-chemotherapy AML patients, whereas no significant differences in OS and EFS were found between the ARHGAP9 high group and ARHGAP9 low group in patients who underwent after auto/allo-HSCT. These results suggest that the effects of ARHGAP9 over-expression can be eliminated by auto/ allo-HSCT, instead of chemotherapy.