APOBEC3B and SPAG5, the Target Genes of miR539, Involved into Imatinib-Resistant of Gastrointestinal Stromal Tumors

Background: Although imatinib can effectively treat gastrointestinal stromal tumor (GIST), some patients are still resistant to it or cannot tolerate the adverse reactions of the drug. This study aimed to investigate potential miRNAs, mRNA and tumor-inltrating lymphocytes (TILs) associated with the prognosis of imatinib-resistant GIST. Methods: mRNA, miRNA sequencing data and patient clinic traits of primary (imatinib-naive) and imatinib-resistant GIST were obtained from GEO (Gene Expression Omnibus) database. A systems biology approach combining Weighted co-expression network analysis (WGCNA) and differential expression analysis were utilized to detect the imatinib-resistant-related miRNA and gene modules and construct a miRNA-gene network. Tumor-inltrating immune cells were analyzed by Estimating the Proportion of Immune and Cancer cells (EPIC) and Tumor-Immune System Interactions (TISIDB). miR-539 was measured by qRT-PCR. SPAG5 and APOBEC3B was measured by qRT-PCR and WB. Transforming growth factor (TGF)-β and interleukin (IL)-10 were assessed with enzyme-linked immunosorbent assay (ELISA). The proportions of CD4+ T cells, CD8+ T cells , NK cells and B cells in tumor-inltrating lymphocyteswere analyzed via ow cytometry (FCM). Results: Two gene modules (brown and yellow) and one miRNA module were associated with the imatinib-resistant. Two hub genes (APOBEC3B and SPAG5) were associated with the imatinib-resistant. Three hub miRNAs were identied to be related to imatinib-resistant (miR-539, miR-376b and miR-18b). G1/S transition of mitotic cell cycle, G2/M transition of mitotic cell cycle, and cell proliferation were common pathways of the gene modules and miRNA module. apolipoprotein B mRNA editing catalytic polypeptide-like 3B (APOBEC3B) and Sperm-associated antigen 5 (SPAG5), which were both target genes of miR-539 was located at the core miRNA–gene APOBEC3B p 2.2e-16) p < 2.2e-16) positive correlated the inltration activated CD4 + T cells. T APOBEC3B imatinib-resistant tumor samples signicantly compared with samples in Imatinib-resistant tumor samples exhibited signicantly downregulation of miR-539 and TGF-β1. Over-expression of miR-539 sensitized imatinib resistant GIST48 cells and increased the secretion of TGF-β1 by inhibiting APOBEC3B and SPAG5. at 0, 0.05, 0.1, 0.2 μM by (CCK8) The miR-539 expression was measured by TaqMan Small RNA primer and probe sets (Applied Biosystems, USA) according to the manufacturers’ instruction. Total RNA was isolated from GIST48 cells or tumor of patient clinic traits of primary (imatinib-naive) and imatinib-resistant GIST using TRIzol. U6-snRNA was used as an endogenous control. APOBEC3B and SPAG5 mRNA expression level was investigated by SYBR green-based quantitative real-time polymerase chain reaction (qPCR) assay. β-actin was used as the internal control. The formula 2 (-ΔΔCt) was used for calculating relative mRNA or miRNA expression. The primers were: APOBEC3B (forward: 5’-ACCCATCCTCTATGGTCGGA-3’; reverse: 5’-GCTTGAAATACACCTGGCCTC -3’); SPAG5 (forward: 5’-CATCTCACAGTGGGATAACTAATAAAC-3’; reverse: 5’- CAGGGA TAGGTGAAGCAAGGATA-3’); β-actin (forward: 5’- AGTCATTCCAAATATGAGATGCGTT -3’; reverse: 5’-TGCTATCACCTCCCCTGTGT-3’).

G1/S transition of mitotic cell cycle, G2/M transition of mitotic cell cycle, and cell proliferation were common pathways of the gene modules and miRNA module. apolipoprotein B mRNA editing catalytic polypeptide-like 3B (APOBEC3B) and Sperm-associated antigen 5 (SPAG5), which were both target genes of miR-539 was located at the core of miRNA-gene network. APOBEC3B (rho=0.509, p < 2.2e-16) and SPAG5 (rho=0.468, p < 2.2e-16) was positive correlated with the in ltration levels of activated CD4 + T cells. he proportions of CD4 + T cells, and the mRNA and protein relative expression of APOBEC3B and SPAG5 in imatinib-resistant tumor samples signi cantly increased as compared with tumor samples in imatinib-naive group. Imatinib-resistant tumor samples exhibited signi cantly downregulation of miR-539 and TGF-β1. Over-expression of miR-539 sensitized imatinib resistant GIST48 cells and increased the secretion of TGF-β1 by inhibiting APOBEC3B and SPAG5.
Conclusion: APOBEC3B and SPAG5, the target genes of miR-539 may play as key factor for imatinibresistant GIST by increasing the proportion of tumor-in ltrated activated CD4+ T cells via TGF-β1. These ndings help to advance the understanding of imatinib-resistant GIST and provide potential therapeutic targets.

Background
Gastrointestinal stromal tumour (GIST) is mainly initialised by mutations of the receptor tyrosine kinase genes KIT or PDGFRA [1]. The advent of a small-molecule inhibitor imatinib mesylate, which targets several receptor tyrosine kinases, including both KIT and PDGFRA, dramatically improved patient outcome [2]. However, Still, 10% of GISTs have primary resistance to imatinib and 40-50% of GISTs have secondary resistance within 2 years [3]. MicroRNAs have been found to play a role in imatinib resistance of GIST [3].
MicroRNAs (miRNAs) are short, noncoding regulatory RNAs. By binding to the target mRNAs, miRNAs lead to mRNA degradation and/or translational repression. Accumulated evidence has shown that miRNAs are associated with imatinib response in GISTs [7]. Although distinct mRNA and miRNA expression signatures have been found in GISTs imatinib-resistant, miRNA-target relationships are not well considered.
Weighted gene co-expression network analysis (WGCNA) is a novel gene co-expression network-based approach focusing on gene or miRNA sets other than individual genes or miRNA from different group.
WGCNA was used to explore molecular interaction mechanism analysis and drug resistant correlation networks resolving based on RNA sequencing or microarray data [8].
In this study, weighted co-expression network analysis was performed on the mRNA and miRNA to identify the potential key miRNAs and genes associated with the GIST imatinib-resistant. We further investigated the expression and functional role of miR-539 and its potential targets APOBEC3B and SPAG5 in imatinib resistance of GIST and GIST48, a imatinib-resistant GIST cell line. The results may provide novel information for the study of GIST imatinib-resistant, and provide novel potential biomarkers for the clinical therapy of GIST.

Data collection and preprocessing
The genes or miRNAs estimated with verbose less than 3 were removed using Weighted Gene Coexpression Network Analysis (WGCNA) package of R (Version 3.6.1). The outlier samples with verbose more than 40% were removed detecting by cluster diagram. abnormally low expressed genes and miRNAs with reads per kilobase of exon per million reads mapped (RPKM) less than 10 were removed.
The microRNA and mRNA expression pro les datasets GSE and clinical information of primary (imatinibnaive) and imatinib-resistant with GIST (n=126) were obtained from from Gene Expression Omnibus database (GEO; http://www.ncbi.nlm.nih.gov/geo/). Datasets GSE132542, GSE69465, GSE51697, and GSE45901 were chosen for further study. Data processing was performed using the R programming language (v3.6.1).
Weighted Gene co-expression network analysis (WGCNA) Brie y, correlation coe cient all pairs of genes or miRNA was calculated based on their expression values. A soft-thresholding power β was selected and the co-expression similarity was transformed into a weighted undirected network and topological overlap matrix (TOM).Then, genes or miRNA were clustered into modules by using dynamic tree cut with the following parameters: maxBlockSize of 6000, TOMType of unsigned, and minModuleSize of 30. Next, The module eigengene (ME) were identi ed based on the rst principal component (PC) of gene or miRNAs expression in each module.
Identi cation of differentially expressed genes (DEGs) and miRNAs (DEMs) The differentially expressed genes (DEGs) and differentially expressed miRNAs (DEMs) between imatinibresistant and imatinib-naïve group were screened by the R package, edgeR. The thresholds for DEGs were set as the values of p<0.05 and false discovery rate (FDR)<0.1. The p<0.05 and false discovery rate (FDR) <0.2 were set as the cut-off criteria for DEMs.

Relating modules to clinical parameters and identifying characteristic genes
Hub genes were identi ed based on gene signi cance (GS) and module membership (MM). The gene signi cance (GS) de ned as the correlation between the gene and four clinical factors of GIST (c-KIT mutation, location, malignancy risk, and imatinib-resistant).The module membership (MM) de ned as the correlation between the module eigengene (ME) and the gene expression pro le. The module signi cance (MS) de ned as the average miRNA signi cance or gene signi cance (GS). Hub Genes with high GS and MM were common in the module signi cance (MS) and differentially expressed genes (DEGs) pro le de ned as characteristic genes. characteristic miRNAs were also identi ed by this method.

Functional enrichment analysis of genes and miRNAs
The characteristic genes were submitted to Database for Annotation, Visualization and Integrated Construction of the miRNA-gene interaction network The miRNA-gene interaction network was constructed based on the Imatinib-resistant characteristic miRNAs and Imatinib-resistant characteristic genes. Target genes of characteristic miRNAs were predicted based on the information in microT-CDS (v5.0) and TargetScan (http://www.targetscan.org/).
Only the interactions that were common in the two databases with false discovery rate (FDR) < 0.05 were used to construct the miRNA the interaction network. Cytoscape 3.6.1 was used to visualize the network.

Evaluation of Tumor Microenvironment
Differentially expressed genes (DEGs) were uploaded to Estimating the Proportion of Immune and Cancer cells (EPIC), a method for characterizing tumor in ltrating lymphocytes (TILs) of solid tumor tissues from their gene expression pro les (http://epic.gfellerlab.org) [10]. We cross-check the role of APOBEC3B and SPAG5 in tumor-immune interactions in sarcoma using the Tumor-Immune System Interactions (TISIDB) (http://cis.hku.hk/TISIDB) [11].

miRNAs and mRNAs detection by quantitative Real time PCR
The miR-539 expression was measured by TaqMan Small RNA primer and probe sets (Applied Biosystems, USA) according to the manufacturers' instruction. Total RNA was isolated from GIST48 cells or tumor of patient clinic traits of primary (imatinib-naive) and imatinib-resistant GIST using TRIzol. U6-snRNA was used as an endogenous control. APOBEC3B and SPAG5 mRNA expression level was

Statistical analysis
The comparison between the two groups were evaluated by Student's t-test using Prism 5.0 software. P<0.05 was considered to be statistically different. Cell viability assay were conducted by 2-way ANOVA and Tukey's post-hoc test.

Results
mRNA expression module analysis of WGCNA After data preprocessing, for the mRNA sequencing data, the samples (n=29) were well clustered and 5880 of 23520 mRNAs were remained for further analysis (Fig. 1a). Using the discovery set (GSE132542 dataset), 17 mRNA network modules were identi ed using WGCNA. The modules were assigned a color label (black, brown, yellow, midnightblue, pink, salmon, red, turquoise, purple, tan, cyan, lightcyan, magenta, greenyellow, blue, green, and grey) (Fig. 1b). Correlations between these gene network modules and clinical data were analyzed. Module brown was found to be markedly positive correlated with imatinib-resistant (r = 0.54, p = 0.002); Module yellow was found to be markedly positive correlated with imatinib-resistant (r = 0.43, p = 0.02) (Fig. 1c). To reveal the functions of the mRNAs of the modules correlated with imatinib-resistant of GIST, GO enrichment analysis was performed. According to the results, the mRNAs in the brown module (n=786) and yellow module (n=582), they were involved with cell division, mitotic nuclear division, chromosome segregation, G1/S transition of mitotic cell cycle, sister chromatid cohesion, DNA replication, DNA replication initiation, spindle organization, G2/M transition of mitotic cell cycle, cell proliferation, mitotic sister chromatid segregation, mitotic cytokinesis, and mitotic spindle organization. (Fig. 1d and Table1).
Weighted Gene co-expression network analysis (WGCNA) of miRNA After data preprocessing, for the miRNA sequencing data, the samples (n=53) were well clustered and 722 miRNAs were remained for further analysis (Fig. 2a). According to the module analysis of miRNA coexpression networks, 7 miRNA network modules were identi ed, and assigned the color labels red, blue, green yellow, brown, turquoise, and grey. The correlation analysis of modules and clinical prognostic data revealed that module red was signi cantly negative associated with imatinib-resistant of GIST (r = -0.29, p = 0.04) (Fig. 2b). Cytoscape 3.6.1 was used to visualize the miRNA network in the red module (n=23) (Fig. 2c). To reveal the functions of the miRNAs of the modules correlated with imatinib-resistant of GIST, GO enrichment analysis was performed using mirPath v.3. According to the results, 210 GO terms were totally enriched in biological process (BP). G1/S transition of mitotic cell cycle, G2/M transition of mitotic cell cycle, and cell proliferation were common pathways of the gene modules and miRNA module. (Fig.  2d).

Associations of miRNA target genes and gene network modules
The thresholds for the differentially expressed genes (DEGs) were set as the values of p<0.05 and false discovery rate (FDR)<0.1 between Imatinib-Naive and Imatinib-Resistant GIST Samples. 436 signi cantly differentially expressed genes were detected between the two groups. 102 genes were found in module brown and 16 genes were found in module yellow of WGCNA such as CDK1, AURKB, AURKA, CD44, HMMR, BIRC5 and CCNB1 ( Fig. 3a and Table 2). The Connectivity Map (Cmap) was used to predict smallmolecule drugs against GIST. We found 28 small-molecule drugs could target CDK1 such as JNJ-7706621, Terameprocol, 1-Azakenpaullone, and Kenpaullone. We found 24 small-molecule drugs could target AURKB such as JNJ-7706621, tozasertib and orantinib. We found 25 small-molecule drugs could target AURKA such as JNJ-7706621, tozasertib and orantinib. We found 1 small-molecule drugs could target CD44 such as hyaluronic-acid. We found 2 small-molecule drugs could target BIRC5 such as Terameprocol and YM-15. We found 2 small-molecule drugs could target CCNB1 such as Kenpaullone and 1-Azakenpaullone. These small-molecule drugs might have a therapeutic effect against imatinibresistant GIST. (Table 3). A two-sample t-test (p<0.05) Benjamini-Hochberg false discovery rate (FDR) (FDR<20%) was used to determine differentially expressed microRNAs between Imatinib-Naive and Imatinib-Resistant GIST Samples. 35 signi cantly differentially expressed miRNAs (DEMs) were detected between the two groups (Table 4). Three miRNAs (miR-539, miR-376b and miR-18b) were found in both in module red of WGCNA and DEMs. The results are shown in the Venn diagram (Fig. 3b).
Three miRNAs (miR-539, miR-376b and miR-18b) contained in the modulel red and DEMs were chosen for target genes. A total of 265 target genes were screened out based on the information in TargetScan, miRanda, and miRwalk. Three critical miRNAs target gene regulatory network (miR-539, miR-376b and miR-18b) and the regulatory network of 120 mRNA gene in module brown and yellow were then constructed. Interestingly, Two target gene APOBEC3B and SPAG5 of miR-539 are also in network of mRNA gene in module brown (Fig. 3c).

Relationship between the expression of APOBEC3B/SPAG5 and immune in ltration
To investigate the association between differentially expressed genes (DEGs) and the tumor microenvironment, the 8 immune cell faction was calculated by EPIC. We found that imatinib-naive group has a lower density of CD4 + T cells and endothelial cells compared to the imatinib-resistant group (Fig.  4a). The correlation between APOBEC3B/SPAG5 expression and immune invasion in sarcoma using the Tumor-Immune System Interactions (TISIDB) (Fig. 4b). We found that both the expression of APOBEC3B (rho=0.509, p<2.2e-16) and SPAG5 (rho=0.468, p<2.2e-16) was positive correlated with the in ltration levels of activated CD4 + T cells (Fig.4c).
We compared the expressions of miR-539, APOBEC3B, SPAG5 and TGF-β in imatinib-naive (n=14) and imatinib-resistant tumor samples (n=15). Consistent with bioinformatics results, imatinib-resistant tumor samples exhibited signi cantly downregulation of miR-539 (p=0.0005) (Fig.5a). We found that the mRNA relative expression of APOBEC3B and SPAG5 in imatinib-resistant tumor samples signi cantly increased as compared with tumor samples in imatinib-naive group (p=0.0051 and 0.0035, respectively) (Fig.5b). As expected, the protein relative expression of APOBEC3B and SPAG5 in imatinib-resistant tumor samples were signi cantly increased (p=0.0051 and 0.0085, respectively) (Fig.5c). We detected Transforming growth factor (TGF)-β1 and interleukin (IL)-10, known as important immunosuppressive cytokines, by ELISA. We found that TGF-β1 secretion of imatinib-resistant tumor samples were signi cantly decreased compared with tumor samples in imatinib-naive group (p=0.0051) (Fig.5d). The proportions of CD4 + T cells, CD8 + T cells , NK cells and B cells in tumor-in ltrating lymphocytes from imatinib-naive and imatinib-resistant patients were analyzed via ow cytometry (FCM). We found that the proportions of CD4 + T cells of imatinib-resistant tumor samples were signi cantly increased compared with tumor samples in imatinib-naive group (p=0.003) (Fig.5e). We analyzed the correlation of miR-539 expressions and CD4 + T cells proportions in 29 GIST patients. We observed a signi cantly negative correlation between miR-539 expressions and CD4 + T cells proportionsin GIST patients (P < 0.001, R 2 = 0.5353) (Fig.5f).
Overexpression of miR-539 sensitized imatinib resistant GIST48 cells and increased the secretion of TGF-b1 by inhibiting APOBEC3B and SPAG5.
We investigated whether miR-539 can change the effects of imatinib on GIST48, a imatinib-resistant GIST cell line. We found that miR-539 can signi cantly sensitized the inhibitory effects of GIST48 cells proliferation induced by imatinib (Fig.6a). ELISA was employed to measure the concentration of TGF-β1 and IL-10 secreted by in the supernatants of GIST48 cells. We found that both miR-539 and imatinib can signi cantly rise the TGF-β1 secretion of GIST48 cells, compared with GIST48 cells treated with imatinib only (p=0.0231) (Fig.6b). miR-539 targeted a site at 3'-UTR of APOBEC3B and SPAG5 mRNA analyzed by bioinformatics. (Fig.6c). We identi ed that APOBEC3B and SPAG5 protein expression was signi cantly downregulated in miR-539 and imatinib as compared with cells GIST48 cells treated with imatinib only (p=0.001) (Fig.6d).

Discussion
In the present study, 17 mRNA network modules were identi ed based on the WGCNA of sequencing data of GIST samples, of which four modules were signi cantly related to imatinib-resistant. Meanwhile, seven miRNA network modules were identi ed, of which one module was associated with tumor stage of imatinib-resistant. We found that G1/S transition of mitotic cell cycle, G2/M transition of mitotic cell cycle, and cell proliferation were common pathways of the gene modules and miRNA module. 118 genes such as CDK1, AURKB, AURKA, CD44, HMMR, BIRC5, CCNB1, APOBEC3B, and SPAG5 were both involved in the differentially expressed genes (DEGs) and the markedly positive correlated with imatinib-resistant module.
Previous studies have shown that the speci c activities of Cyclin-dependent kinase 1 (CDK1) signi cantly correlated with recurrence of GIST [12]. Altered expression of AURKB has been observed previously GIST [13]. Aurora kinase A (AURKA) could be a poor risk marker and potential treatment target for metastatic GIST [14,15]. CD44 cleavage activity was signi cantly associated with disease progression and poor survival in GIST [16,17]. Survivin (BIRC5) was associated with high risk and a poor prognosis in GIST [18,19]. Increased expression cyclin B1 (CCNB1) could predict the proliferative activity of GIST [20,21]. The Connectivity Map (Cmap) was used to predict small-molecule drugs targeting these gene. JNJ-7706621 had three target genes of CDK1, AURKB and AURKA. Hyaluronic-acid had two target genes of CD44 and HMMR. Terameprocol had two target genes of CDK1 and BIRC5. 1-Azakenpaullone had two target genes of CDK1 and CCNB1. Kenpaullone had two target genes of CDK1 and CCNB. So, these ve smallmolecule drugs JNJ-7706621, Hyaluronic-acid, Terameprocol, 1-Azakenpaullone, kenpaullone were predicted to have a therapeutic effect against imatinib-resistant GIST.
Apolipoprotein B mRNA editing enzyme catalytic polypeptide-like 3B (APOBEC3B), an antiviral cytidine deaminase that induces DNA mutations, was considered as a medium for cancer development and therapeutic resistance [36]. Recent study have shown that the overexpression of APOBEC3B in tumor increased the resistance to chemotherapy and triggered anti-tumor T cells [37]. Sperm-associated antigen 5 (SPAG5), a new ampli cation gene of Ch17q11.2, could have clinical utility as a prognostic biomarker and chemosensitivity predictor for breast cancer [38]. In the present study, we found that two hub genes (APOBEC3B and SPAG5) were associated with the imatinib-resistant. Both APOBEC3B [39] and SPAG5 [40] expression were related to drug resistance. We found that mRNA and protein relative expression of APOBEC3B and SPAG5 in imatinib-resistant tumor samples signi cantly increased as compared with tumor samples in imatinib-naive group. Interestingly, We also found that miR-539 can directly target APOBEC3B and SPAG5, and the previous studies supports our results [41] [42]. However, up to now, the roles of miRNA-539, APOBEC3B and SPAG5 in GIST are still unclear. We found that overexpression of miR-539 sensitized imatinib resistant GIST48 cells by inhibiting APOBEC3B and SPAG5.
Recent studies have shown that in ltrating CD4 + T cells attenuate chemotherapy sensitivity in prostate cance [43]. APOBEC3 family expression correlated with T-Cell in ltrationin in high-grade serous ovarian carcinoma [44]. SPAG5 expression level was positively correlated with the in ltration levels of CD8 + T cells [45]. In the present study, we found that both the expression of APOBEC3B and SPAG5 was positive correlated with the in ltration levels of activated CD4 + T cells, and miRNA-539 was negative correlated with the in ltration levels of activated CD4 + T cells. Tumor-derived TGF-β suppresses antitumor function of CD4 + T cells [46]. We found that transforming growth factor (TGF)-β1, known as important immunosuppressive cytokines, were signi cantly decreased in imatinib-resistant tumor samples.
Overexpression of miR-539 sensitized imatinib resistant GIST48 cells and increased the secretion of TGF-β1 by inhibiting APOBEC3B and SPAG5.

Conclusions
In conclusion, APOBEC3B and SPAG5, the target genes of miR-539, might play as key factor for imatinibresistant GIST based on the WGCNA of sequencing data. We infered that low expression of miR-539 in imatinib-resistant GIST might increase the expression of APOBEC3B and SPAG5. And then, APOBEC3B and SPAG5 might increase imatinib resistance in GIST patients by increasing the proportion of tumorin ltrated activated CD4 + T cells.

Consent for publication
Written informed consent for publication was obtained from all participants.
Availability of data and materials