Long noncoding RNA AFAP1-AS1 acts as a competing endogenous RNA of miR-423-5p to facilitate nasopharyngeal carcinoma metastasis through regulating the Rho/Rac pathway

1Nasopharyngeal carcinoma (NPC) is a head and neck epithelial malignancy that occurs frequently in Southern China. Epstein-Barr virus exposure, diet, and genetic factors are implicated in its etiology. Although radiotherapy is an effective treatment for NPC patients in the early stages of the disease, the majority (75–90%) of NPC cases are predisposed to metastasis at initial diagnosis, which hampers efficacious treatment and poses a high risk of disease recurrence. Enhancing our understanding of the molecular mechanisms underlying NPC may promote the development of effective metastasis-targeted therapy and improve the overall prognosis of patients with this disease.

2Noncoding RNAs, including microRNAs (miRNAs), long noncoding RNAs (lncRNAs) and circular RNAs (circRNAs), have important roles in the regulation of gene expression and are involved in the development of a variety of human diseases, including cancer. The lncRNA actin filament associated protein 1 antisense RNA 1 (AFAP1-AS1) maps to the antisense DNA strand of AFAP1 and was first identified as highly expressed in esophageal cancer. AFAP1-AS1 is also highly expressed in lung cancer and pancreas cancer, and hepatocellular carcinoma. We previously discovered that AFAP1-AS1 was significantly highly expressed in NPC and associated with metastasis and poor prognosis. Numerous cytoskeleton proteins, including proteins in the small GTPase Rho/Rac signaling pathway, were identified and significantly regulated using the liquid chromatography-tandem mass spectrometry (LC-MS/MS) by siAFAP1-AS, suggesting that AFAP1-AS1 may regulate NPC migration and invasion through this pathway; however, as a newly identified lncRNA, the mechanism by which AFAP1-AS1 regulates the Rho/Rac signaling pathway remains unknown.

3A new model of lncRNA involvement in gene regulation has recently been proposed, called competing endogenous RNA (ceRNA). This model proposes that lncRNAs and mRNA transcripts affect one another by competitively combining with miRNA response elements (MREs) to influence posttranscriptional regulation. In this way, lncRNAs bind with miRNAs to modulate regulation of protein-coding gene expression and participate in the regulation of cell behavior. These interactions between mRNAs, miRNAs, and lncRNAs generate complex regulatory networks. In this study, we found that AFAP1-AS1 acts as a ceRNA, competitively binding with miR-423-5p and directly regulating RAB11B and LASP1 genes in the Rho/Rac pathway. AFAP1-AS1 can also regulate the expression of FOSL2 and its downstream genes by competing with FOSL2 to bind miR-423-5p, resulting in increased migration and invasion of NPC through the Rho/Rac signaling pathway.

4Thirty-two nasopharyngeal carcinoma samples and 13 nontumor nasopharyngeal epithelium tissues were obtained from patients. All samples were collected with the consent of patients and the experiments were approved by the ethics committee of Hunan Cancer Hospital, Changsha, China. All fresh tissues were immersed in RNALater (Ambion, Austin, TX, USA). The tissue samples were stored in a − 80 °C laboratory freezer. RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol. The NPC cell lines 5-8F and HNE2 were maintained in our laboratory in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS; Invitrogen), penicillin (100 U/ml, Sigma-Aldrich, St. Louis, MO, USA), and streptomycin (100 mg/ml, Sigma-Aldrich). All cell lines had been recently tested for mycoplasma contamination and identified no cross-contaminated cell lines or mycoplasma contamination.

5The synthetic miR-423-5p mimics and inhibitors were supplied by Ruibo Co. (Guangzhou, China). siRNAs and nontargeting scrambled control siRNAs were provided by Invitrogen. The sequences of the AFAP1-AS1- and FOSL2-targeting siRNAs are listed in Additional file 1: Table S1. To construct the AFAP1-AS1 expression vector, the entire AFAP1-AS1 sequence was amplified by reverse transcription PCR (RT-PCR) and cloned into the pcDNA3.1 vector. The open reading frame of FOSL2 (NM_005253) was cloned into the pENTR vector and tagged with the C-terminal Flag and His peptide sequences was purchased from the Vigene Co. (Rockville, MD, USA). Transfection of cells with plasmids, siRNAs, and miRNAs was performed using Lipofectamine 3000 (Life Technologies, Grand Island, NY, USA) or HiPerFect (Qiagen, Hilden, Germany) transfection reagents, according to the manufacturer’s instructions.

6NPC data sets and the corresponding clinical data were downloaded from publicly available GEO databases. These included 62 patients with NPC and 6 non-NPC nasopharyngeal epithelial tissues from GSE73460, 6 patients with NPC and 12 non-NPC epithelial tissues from GSE32906 and 122 patients with NPC with survival status from GSE70970.

7Five software tools, RNAhybrid, RNA22, PITA, TargetScan, and RegRNA, were used for miRNA target prediction in this study. For miRNA target validation assays, 58 bp fragments of the AFAP1-AS1 sequence, or the RAB11B, LASP1, RAC1, and FOSL2 3′-UTRs containing the wild-type (WT) and mutant (MT) binding sites for miR-423-5p, were cloned into the pMIR-REPORT vector. The sequences of these synthetic oligonucleotides are listed in Additional file 1: Table S1. Cells (5-8F and HNE2) were cotransfected with pMIR-REPORT vectors containing the WT sequences of the RAB11B, RAC1, LASP1, or FOSL2 3′-UTRs or their corresponding mutants, along with the pRL-TK control vector and miR-423-5p mimics, inhibitors, or negative control.

8For the LASP1 promoter activity analysis, the Promoter 2.0 Prediction Server (http://www.cbs.dtu.dk/services/Promoter/) was used to predict potential AP1-binding sites in the LASP1 promoter sequence. Three tandem the wild type or mutant AP1-binding sites were inserted into the TA-luciferase vector as synthetic oligonucleotides encoding the wild type or mutant AP1-binding sites (Additional file 1: Table S1).

9The AP-1 cis-element reporter plasmid (the AP-1 reporter), which contains multiple conserved AP1-binding sites for sensitive detection of AP-1 transcriptional activity, was purchased from Promega (pAP1-TA-luc; Promega, Fitchburg, WI, USA) to measure the AP-1 transcriptional activity. Plasmid transfection and luciferase assays were performed as described previously. Cells were harvested 48 h posttransfection and assayed using a Dual Luciferase Assay (Promega) according to the manufacturer’s instructions. All transfection assays were carried out in triplicate.

10For wound healing assays, 5-8F and HNE2 cells were seeded and grown to 90% confluence in 6-well culture plates, and a 200 μl pipet tip was used to create a scratch in the cell monolayer. Images were captured and measured with an ocular ruler to ensure that all wounds were the same width at the beginning of each experiment.

11For transwell assays, 5-8F and HNE2 cells in 200 μl serum-free media were placed into upper transwell chambers (8.0 μl pore size, BD Biosciences, Franklin Lakes, NJ, USA) for invasion assays after transfection, according to the manufacturer’s protocol. Invaded cells were fixed and stained with methanol and 0.1% crystal violet, counted under a microscope, and imaged. Independent experiments were carried out in triplicate.

12Proteins were extracted and separated according to previously described methods. Membranes were incubated overnight at 4 °C with primary anti-RHOA, anti-RHOC, anti-RAC2, anti-RAB10, anti-RAB11B, anti-RAB11A, anti-RHOGDI, anti-PFN1, anti-LASP1, or anti-FOSL2 antibodies (Proteintech, Wuhan, China). Blots were visualized by exposure to X-ray films using an enhanced chemiluminescence detection system (EMD Millipore, Burlington, MA, USA). Blots were probed with anti-GAPDH antibody (Cell Signaling Technology, Danvers, MA, USA) as an endogenous control for equal loading.

13Total RNA was isolated with TRIzol reagent (Invitrogen) according to the manufacturer’s protocol. For miRNA real-time PCR, cDNA was synthesized using the miScript system (Ruibo) according to the manufacturer’s instructions. miR-423-5p expression levels were measured using a Ruibo miRNA primer assay (Ruibo) and a miDETECT Track miRNA qRT-PCR kit (Ruibo) according to the manufacturer’s instructions. Data were normalized to the small nuclear RNA RNU6B (U6 snRNA) expression level. For mRNAs and lncRNAs, real-time PCR expression analysis was carried out using a SYBR green real-time PCR kit (Applied Biological Materials, Inc., Richmond, BC, Canada). Data were normalized to the GAPDH expression level and further normalized to the negative control, unless otherwise indicated. The primers used for PCR are listed in Additional file 1: Table S1. Fold changes in expression were calculated using a relative quantification (2-ΔΔCt) method. All reactions were performed in triplicate and repeated in three independent experiments.

14Male BALB/C nude mice (5–6 weeks of age, 16–18 g) were maintained in the experimental animal center of Central South University in a pathogen-free facility. The Institutional Animal Ethics Committee of Central South University approved this animal study. Mice were divided into four groups, including the over-AFAP1-AS1, negative control (NC), miR-423-5p, and over-FOSL2 groups. For metastasis assay, 2 × 106 5-8F cells in a volume of 200 ml with corresponding treatment were tail-vein injected into five-week-old male BALB/c nude mice which were randomly divided into four groups (n = 10 for each group). 60 days later, the mice were killed, and all the lungs are surgically removed and the number of macroscopically visible pulmonary metastases nodules per mouse was counted by two experienced pathologists. Then, the lung tissues were fixed in 10% neutral phosphate-buffered formalin, followed by HE staining.

15All continuous variables are presented as the means ± standard deviation (SD), and categorical data are expressed as percentages. GraphPad Prism 5.0 (GraphPad Software, Inc., La Jolla, CA, USA) was used for statistical analysis. Student’s t-test was performed to assess significant differences between two groups using p < 0.05 as the significance threshold. Survival analysis was performed by Kaplan-Meier curves and log-rank test for significance. A two-sided p-value of < 0.05 was considered statistically significant.

16We first confirmed the function of AFAP1-AS1 in NPC cells migration and invasion. The overexpressing AFAP1-AS1 vector was constructed and transfected into the NPC cell lines 5-8F and HNE2, and the expression of AFAP1-AS1 was confirmed by real-time PCR (Fig. 1a). Wound healing and transwell assays demonstrated that overexpression of AFAP1-AS1 accelerated 5-8F and HNE2 cell migration (Fig. 1b) and invasion (Fig. 1c).

17To identify the mechanism of AFAP1-AS1 in NPC, we used four programs (PITA, RNAhybrid, RNA22, and RNAreg2.0) to predict potential lncRNA-miRNA interactions involving AFAP1-AS1. Among the results, miR-423-5p was selected for further investigation, since it was the only interacting miRNA proposed by all these four tools. Then, whether endogenous AFAP1-AS1 levels were affected by miR-423-5p was examined through transfection of miR-423-5p mimics or inhibitors into 5-8F and HNE2. Real-time PCR indicated that AFAP1-AS1 expression was reduced after transfection with miR-423-5p mimics (Fig. 2a) and induced after transfection with miR-423-5p inhibitors in 5-8F and HNE2 cells (Fig. 2b). These results indicated that AFAP1-AS1 was negatively regulated by miR-423-5p in NPC cells. A total of 2 binding sites for miR-423-5p were predicted in the AFAP1-AS1 sequence. One binding site between 5021 and 5043 bp of AFAP1-AS1 (NR_026892, Fig. 2c) with the lowest minimum free energy was selected, and its wild-type (WT) and mutant (MT) sequences were cloned into a pMIR-REPORT luciferase vector for luciferase activity assay. The result showed that transfection of miR-423-5p mimics into 5-8F and HNE2 cells reduced the luciferase activity generated by the wild-type AFAP1-AS1 reporter (WT), but not that of the empty vector or the reporter vector containing the mutant miR-423-5p binding site (Fig. 2d). Moreover, transfection with miR-423-5p inhibitors increased the luciferase activity of the wild type AFAP1-AS1 reporter (WT, Fig. 2e), suggesting that miR-423-5p binds with AFAP1-AS1 through this site.

18Based on the above findings, we screened online published miRNA datasets and found that miR-423-5p expression was low-expressed in NPC clinical samples compared with nontumor nasopharyngeal epithelium tissues in a gene expression profiling (GEP) dataset, GSE73460 (Additional file 2: Figure S1a), and another NPC GEP dataset, GSE32906 showed that the expression of miR-423-5p was tightly associated with the TNM stages of NPC patients (Additional file 2: Figure S1b). In the GSE70970 dataset, decreased miR-423-5p expression was also associated with poor overall survival and poor relapse-free survival (Additional file 2: Figure S1c) of NPC patients. Then, the expression of AFAP1-AS1 and miR-423-5p were examined in 32 NPC clinical samples and 13 nontumor nasopharyngeal epithelium biopsies by qRT-PCR. The results showed that miR-423-5p was lowly expressed in NPC tissues compared with that in nontumor nasopharyngeal epithelium tissues (Fig. 3a). Conversely, AFAP1-AS1 was highly expressed in the NPC tissues and tightly associated with the clinical TNM stages, neck lymph node metastasis, and the T stages of the patients (Fig. 3b). There was an inverse correlation between AFAP1-AS1 and miR-423-5p in the NPC tissues (Fig. 3c).

19Next, we measured the cell migration and invasion capacity of miR-423-5p in NPC cells. Wound healing and transwell assays demonstrated that miR-423-5p overexpression using mimics inhibited cell migration and invasion compared with those of negative control (Fig. 3d and e), consistently with the functional effects of AFAP1-AS1 knockdown. Moreover, transfection with miR-423-5p inhibitors had a consequence similar to the function of AFAP1-AS1 overexpression in 5-8F and HNE2 cells (Fig. 3f and g). Taken together, these results indicated that AFAP1-AS1 might promote NPC cell migration and invasion via mutually regulated with miR-423-5p.

20We previously used a proteomics strategy to determine that AFAP1-AS1 regulates the expression of several small GTPase family members, and modulates cancer cell metastasis via regulation of actin filament integrity. However, the mechanism by which AFAP1-AS1 regulates the GTPase family members expression remains unknown. Since AFAP1-AS1 interacts with miR-423-5p, we hypothesized that AFAP1-AS1 might act as a ceRNA to regulate the GTPase family members through miR-423-5p.

21We first confirmed that AFAP1-AS1 regulates the small GTPase family members at the mRNA level. After AFAP1-AS1 knockdown (siRNA) or overexpression in 5-8F and HNE2 cells, multiple members of the Rho/Rac pathway were affected at the mRNA level, as determined by real-time PCR (Fig. 4a), and the expression of these molecules was also altered at the protein level, as revealed by western blotting (Additional file 2: Figure S2); these results were consistent with those of our previous proteomics analysis. Similar results were obtained in 5-8F and HNE2 cells after transfection with miR-423-5p mimics and inhibitors (Fig. 4b and c). miR-423-5p regulated the expression of numerous molecules, including RAB10, RAB11A, RAC2, PFN1, RHOA, RAC1, RHOC, LASP1, and RAB11B, in the Rho/Rac signaling pathway at both the mRNA (Fig. 4b) and protein levels (Fig. 4c), suggesting that AFAP1-AS1 may regulate the Rho/Rac signaling via miR-423-5p.

22As shown in Fig. 4, we found that the expression of RAC1, RHOC, LASP1, and RAB11B was reduced by miR-423-5p mimics and induced by miR-423-5p inhibitors at both the mRNA and protein levels. We next investigated whether miR-423-5p directly targeted these molecules to regulate the Rho/Rac signaling pathway. First, the 3′-UTR sequences of these genes were analyzed using the TargetScan software. Interestingly, the 3′-UTRs of RAC1, LASP1, and RAB11B contained numerous elements exactly complementary to the miR-423-5p seed region. Next, the vectors containing the wild type or mutant (mutations targeting the seed miR-423-5p sequence) RAB11B, LASP1, and RAC1 3′-UTR sequences were constructed, and luciferase analysis showed that miR-423-5p mimics reduced luciferase activity in 5-8F and HNE2 when transfected with vectors containing the wild type RAB11B (Fig. 5a and b) and LASP1 (Fig. 5c and d) 3′-UTR sequences, but not in 5-8F and HNE2 transfected with the wild type RAC1 3′-UTR (pMIR-WT) (Additional file 2: Figure S3). The effect of miR-423-5p mimics on the luciferase activity of the vectors containing the mutant RAB11B and LASP1 sequence was abolished in 5-8F and HNE2 cells (Fig. 5). Transfection of miR-423-5p inhibitors in these cells had the opposite effect on the luciferase activity of the RAB11B and LASP1 3′-UTR reporter constructs (Fig. 5). These results suggested that miR-423-5p regulated the Rho/Rac signaling through direct binding the 3′-UTRs of RAB11B and LASP1.

23The ceRNA hypothesis posits that specific lncRNA molecules function as sinks for pools of active miRNAs, functionally liberating mRNA transcripts targeted by certain miRNAs. Therefore, the expression of lncRNAs that act as ceRNAs is usually positively correlated with the expression levels of their affected mRNAs. To determine which mRNAs are regulated by AFAP1-AS1 through miR-423-5p via ceRNA, we first performed bioinformatics analysis to predict targets of miR-423-5p. The NPC gene expression omnibus (GEO) datasets were screened to identify mRNAs whose expression levels were positively correlated with those of AFAP1-AS1. The expression of FOSL2 was positively correlated with that of AFAP1-AS1 (Additional file 2: Figure S4). Real-time PCR showed that FOSL2 was also overexpressed in those 32 NPC clinical samples compared with nontumor nasopharyngeal epithelial samples and tightly positively associated with the TNM stages, the T stages, and lymphoma metastasis in patients with NPC (Fig. 6a), positively correlated with the AFAP1-AS1 expression (Fig. 6b) while negatively correlated with miR-423-5p (Fig. 6c). The 3′-UTRs of FOSL2 were also predicted with several potential binding sites complementary to the miR-423-5p seed region (Fig. 6d). Transfection of miR-423-5p mimics or inhibitors into 5-8F and HNE2 cells reduced or induced the mRNA and protein levels of FOSL2, respectively (Fig. 6e). Luciferase assays also showed that miR-423-5p mimics decreased the luciferase activity of the wild type FOSL2 3′-UTR sequence (pMIR-WT), while no effect was observed on the luciferase activity from vectors containing the mutant miR-423-5p binding sites in 5-8F and HNE2 cells. Transfection with miR-423-5p inhibitors had the opposite effect on the luciferase activity generated from the FOSL2 3′-UTR (Fig. 6f).

24We hypothesized that, as a ceRNA, AFAP1-AS1 competed with FOSL2 to bind miR-423-5p; therefore, we determined whether there was specific interplay among AFAP1-AS1, miR-423-5p, and FOSL2. We monitored both the mRNA and protein levels of FOSL2 after AFAP1-AS1 overexpression and knockdown in 5-8F and HNE2 cells. As expected, AFAP1-AS1 overexpression increased both the mRNA and protein levels of FOSL2, and siAFAP1-AS1 transfection reduced FOSL2 expression (Fig. 6g). Overall, these results suggested that the role of AFAP1-AS1 in modulating FOSL2 expression is mediated by its competitive binding with miR-423-5p in 5-8F and HNE2 cells. Together, these data indicated that AFAP1-AS1 functions as a ceRNA by competitively binding miR-423-5p, thereby relieving the suppression of FOSL2 expression by miR-423-5p in NPC.

25To investigate the function of FOSL2 in NPC, a FOSL2 overexpression vector, including three Flag tags, was constructed and its expression confirmed in 5-8F and HNE2 cells (Fig. 7a). Wound healing and transwell assays demonstrated that overexpression of FOSL2 accelerated cell migration and invasion compared to those in negative control (Fig. 7b). Moreover, FOSL2 knockdown by siRNA (Fig. 7c) had the opposite effect (Fig. 7d). Taken together, these results indicated that FOSL2 induces NPC cell migration and invasion.

26FOSL2 is a member of the Fos family of transcription factors that heterodimerize with the Jun family members to form the AP-1 complex, which is able to activate or suppress the expression of many genes involved in tumor growth and progression. As a transcription factor, FOSL2 could modulate the expression of genes in the Rho/Rac pathway. Therefore, we determined whether FOSL2 could transcriptionally activate downstream genes in the Rho/Rac signaling pathway. Bioinformatics analysis indicated a potential FOSL2 binding site in the LASP1 promoter positioned at the − 322 to − 312 bp upstream of the transcription start site (Fig. 8a). Next, luciferase reporter constructs containing the wild-type and mutant LASP1 promoter sequences, including the potential FOSL2 binding site, were constructed (Fig. 8a). An AP-1 luciferase vector, containing multiple conserved AP1-binding sites for sensitive detection of the AP1 transcriptional activity, was used as a positive control. As shown in Fig. 8b, FOSL2 activated transcription from only the wild-type, but not the mutated LASP1 promoter in these two cell lines. AFAP1-AS1 also participated in regulation of the transcription activity of the AP-1 complex containing FOSL2. Overexpression of AFAP1-AS1 dramatically increased and knockdown decreased the AP-1 transcriptional activity from a synthetic promoter including the AP-1 consensus-binding site (pAP1-TA-luc) (Fig. 8c). FOSL2 knockdown also reduced LASP1 and AFAP1-AS1 expression (Fig. 8d). These results suggest that AFAP1-AS1 may regulate the Rho/Rac signaling pathway via the AP-1 transcriptional complex through modulating FOSL2 activity by competition with miR-423-5p.

27The function of AFAP1-AS1, miR-423-5p, and FOSL2 in metastasis was confirmed in nude mice. AFAP1-AS1-, miR-423-5p-, and FOSL2-overexpressing 5-8F cell lines were used in this study. The cells were injected into the tail veins of nude mice; 5-8F-mock served as the control. After 60 days, the mice were sacrificed, and the lungs, livers, and lymph nodes were observed. Compared with the control, metastasis to the lungs of 5-8F cells transfected with AFAP1-AS1 and FOSL2 significantly increased, while metastasis of 5-8F transfected cells with miR-423-5p mimics were reduced (Fig. 9). These results indicated that AFAP1-AS1 and FOSL2 enhanced NPC metastasis and miR-423-5p reduced its metastasis.

28A ceRNA hypothesis was proposed by Salmena et al. in 2011. The theory explains why mRNA, pseudogenes, lncRNA, and other molecules share miRNA response elements (MREs) and bind to identical miRNAs competitively to affect cell status. miRNAs and MREs are two key elements in this hypothesis; miRNAs bind not only to MREs in mRNAs but also to pseudogenes and lncRNAs, influencing mRNA and lncRNA expression levels posttranscriptionally. For example, the tumor suppressor gene phosphatase and tensin homolog (PTEN), a negative regulator of the oncogenic phosphoinositide 3-kinase/Akt signaling pathway, provides the best evidence for disease-associated ceRNAs. PTEN pseudogene 1 (PTENP1) can increase PTEN expression by binding to several miRNAs to suppress proliferation of cancer cells. Conversely, when PTENP1 is downregulated, more miRNAs are available to inhibit PTEN expression, facilitating tumor growth. Both PTEN and PTENP1 are controlled by ceRNA circuitry in many cancers. LncRNAs interact with miRNAs and mRNAs as components of the cellular RNA language, which comprises a complex, finely controlled network.

29AFAP1-AS1 is highly expressed and tightly associated with tumor progression in many cancers including NPC. We previously showed that AFAP1-AS1 was associated with poor prognosis in NPC and it promoted cell migration and invasion via regulation of actin filament integrity and the Rho/Rac signaling. In this study, we explored the detailed molecular mechanism of AFAP1-AS1 as a ceRNA in regulation of NPC cell migration and invasion. We hypothesized that AFAP1-AS1 may act as a ceRNA to regulate the Rho/Rac pathway; therefore, we employed bioinformatics methods to predict potential miRNAs regulated by AFAP1-AS1. miR-423-5p was the only miRNA predicted by all four tools used for these analyses and had the lowest minimum free energy, suggesting that AFAP1-AS1 is likely a direct target of miR-423-5p. Many reports have identified miR-423-5p as an oncogene in various cancers, including gastric cancer and hepatocellular carcinoma. In this study, we found that miR-423-5p was low-expressed in NPC and its expression was negatively correlated with the clinical TNM stages in NPC tumors. Overexpression of miR-423-5p inhibited NPC cells migration and invasion, and inhibition of miR-423-5p produced the opposite phenotype. The function of miR423-5p had an opposite effect to those of AFAP1-AS1 on migration and invasion. Luciferase assays confirmed that miR-423-5p regulated the 3’-UTR of AFAP1-AS1. These results suggested that AFAP1-AS1 may regulate downstream genes through binding of miR-423-5p, and that miR-423-5p may have a role as a tumor suppressor gene through its effects on different downstream genes in NPC.

30We previously determined by the proteomic analysis that expression levels of at least 209 proteins were significantly altered after AFAP1-AS1 knockdown in NPC cells. Among these, the expression levels of many proteins involved in regulation of cytoskeletal dynamics were confirmed as significantly altered using liquid chromatography-tandem mass spectrometry (LC-MS/MS) after AFAP1-AS1 knockdown, including those in the small GTPase Rho/Rac signaling pathway: RAB10, RAB11A, RAC2, PFN1, RHOA, RAC1, RHOC, LASP1, and RAB11B. The Rho/Rac signaling pathway is a well-established regulator of the actin cytoskeleton, as highlighted by the RAC1 dependence of membrane ruffling and RHOA-induced stress-fiber formation. In this study, we focused on the role of AFAP1-AS1 as a noncoding RNA in regulation of the expression of these proteins. We investigated whether AFAP1-AS1 acted as a ceRNA through miR-423-5p to regulate the Rho/Rac pathway. miR-423-5p inhibited the expression of RAC1, RHOC, LASP1, and RAB11B and increased that of other molecules, including RAB10, RAB11A, RAC2, PFN1, and RHOA, consistently with the results after AFAP1-AS1 inhibition or overexpression. These results implied that AFAP1-AS1 regulated the Rho/Rac signaling pathway through miR-423-5p. As a microRNA, miR-423-5p may directly regulate expression or translation of its target genes; therefore, we first examined whether some molecules including RAC1, RAB11B, RHOC, and LASP1 in the Rho/Rac pathway were direct target genes of miR-423-5p, since their expression levels were reduced by the specific miR-423-5p mimics. Bioinformatics analysis predicted the RAB11B and LASP1 genes as targets of miR-423-5p. Further experiments confirmed that miR-423-5p directly targeted these two genes and that AFAP1-AS1 acted as a ceRNA of RAB11B and LASP1, regulating their expression by competitively binding miR-423-5p, thus affecting the migration and invasion of NPC. As we know, LASP1 is an actin-binding phosphoprotein that is overexpressed in several cancers and it is a cAMP- and cGMP-dependent signaling protein which binds to the actin cytoskeleton at extensions of the cell membrane. RAB11B belongs to the Rab family and regulates exocytotic and endocytotic pathways. Our experiments confirmed the hypothesis that AFAP1-AS1 acts as a ceRNA to regulate the expression of key genes, such as LASP1 and RAB11B, by competitively binding miR-423-5p, resulting in dysregulation of the Rho/Rac1 signaling pathway and migration and invasion of human NPC cells.

31The expression of RAB10, RAB11A, RAC2, PFN1, and RHOA were upregulated in 5-8F and HNE2 cells transfected with miR-423-5p mimics, which is consistent with the results of experiments using siAFAP1-AS1 and suggests that AFAP1-AS1 and miR-423-5p are tightly associated with the Rho/Rac signaling pathway. miR-423-5p may also indirectly regulate the expression of these proteins through its effects on other molecules. To identify other direct targets of miR-423-5p, we screened genes whose expression levels were highly correlated with those of AFAP1-AS1 in NPC tissues based on the results of previous microarray analyses and the ceRNA theory. FOSL2, a member of the Fos family, was analyzed using various prediction tools and found to contain several potential miR-423-5p binding sites in its 3′-UTR, while the expression of AFAP1-AS1 and FOSL2 was significantly correlated in NPC GEO datasets. FOSL2 is a member of AP-1 transcription factor complexes. FOS and Jun protein family members can form homo- or heterodimers and bind to the promoters of specific genes to promote or inhibit their transcription in combination with other transcription and regulatory factors. In this study, we found that FOSL2 could regulate the LASP1 transcription by binding its promoter, suggesting that AFAP1-AS1 acts as a ceRNA in regulation of FOSL2 and the AP-1 transcription factor activation, thereby affecting the expression of small GTPase molecules. Whether AFAP1-AS1 also regulates the expression of other downstream genes through FOSL2 is worthy of further investigation.

32In this study, we confirmed that AFAP1-AS1 functions as a ceRNA by adsorption of miR-423-5p and verified that LASP1, RAB11B, and FOSL2 are downstream target genes of miR-423-5p (Additional file 2: Figure S5). However, miRNAs can target multiple genes, and there may be other protein-encoding genes regulated by miR-423-5p among the 209 proteins regulated by AFAP1-AS1 in NPC cells using our previous proteomic approach; this is also worthy of further investigation.

33AFAP1-AS1 may regulate a complex and sophisticated gene network through miR-423-5p as a ceRNA. Identification of the ceRNA network involving AFAP1-AS1 may contribute to a better understanding of NPC carcinogenesis and tumor progression. Our data also suggest that targeting AFAP1-AS1 and/or miR-423-5p have potential as a new therapeutic strategy and that these molecules are potential biomarkers in NPC.