miR-3188 regulates nasopharyngeal carcinoma proliferation and chemosensitivity through a FOXO1-modulated positive feedback loop with mTOR–p-PI3K/AKT-c-JUN

The biological role of miR-3188 has not yet been reported in the context of cancer. In this study, we observe that miR-3188 not only reduces cell-cycle transition and proliferation, but also significantly prolongs the survival time of tumour-bearing mice as well as sensitizes cells to 5-FU. Mechanistic analyses indicate that miR-3188 directly targets mTOR to inactivate p-PI3K/p-AKT/c-JUN and induces its own expression. This feedback loop further suppresses cell-cycle signalling through the p-PI3K/p-AKT/p-mTOR pathway. Interestingly, we also observe that miR-3188 direct targeting of mTOR is mediated by FOXO1 suppression of p-PI3K/p-AKT/c-JUN signalling. In clinical samples, reduced miR-3188 is an unfavourable factor and negatively correlates with mTOR and c-JUN levels but positively correlates with FOXO1 expression. Our studies demonstrate that as a tumour suppressor, miR-3188 directly targets mTOR to stimulate its own expression and participates in FOXO1-mediated repression of cell growth, tumorigenesis and NPC chemotherapy resistance.

M icroRNAs (miRNAs or miRs) play important roles in development, cellular differentiation, proliferation, cellcycle control and cell death 1 , and have been implicated in a variety of human diseases, including cancer 2,3 . A growing body of evidence has demonstrated the importance of miRNAs in managing chemotherapy efficacy in multiple human cancers 4,5 . Despite being one of the original miRNAs discovered, the biological role of miR-3188 and its molecular mechanisms underlying cancer initiation and progression have not been reported.
Nasopharyngeal carcinoma (NPC) is a tumour type arising from the epithelial cells that line the nasopharynx 6 . It is common in certain regions of East Asia and Africa, with Epstein-Barr virus (EBV) exposure, diet and genetic factors implicated in its aetiology 7,8 . Although relatively rare in the USA, NPC accounts for one-third of childhood nasopharyngeal neoplasms 9 . In recent studies, abnormal expression of miRNAs was been broadly implicated in the pathogenesis of NPC. For example, EBVencoded miRNA BART1 induces tumour metastasis by regulating the PTEN-dependent pathways 10 . In addition, tumour suppressor PDCD4 modulates miR-184-mediated direct suppression of c-MYC and BCL2-blocking cell growth and survival 11 .
FOXO1 is a transcription factor and a member of the FOXO subfamily of the Forkhead/winged helix family. The phosphorylation of FOXO1 by AKT leads to its inactivation after nuclear to cytoplasmic translocation 12,13 . Previous evidence has supported that FOXO1 functions as tumour suppressor on the basis of its role in regulating cell-cycle progression, differentiation, metabolism and survival 14,15 . Furthermore, decreased FOXO1 expression has been demonstrated in many tumour types, such as Hodgkin lymphoma 16 , breast cancer 17 and alveolar rhabdomyosarcoma 18 . Recent evidence suggested that LMP1 silencing slows cell growth and enhances chemosensitivity through inhibition of the AKT signalling pathway and its downstream factor phospho-FOXO1 in EBV-positive NPC cell line 19 . Elevated levels of phosphorylated AKT also correlated with phospho-FOXO1 in NPC samples 20 . However, the detailed role of FOXO1 in the suppression of NPC cell growth remains unclear.
Here, we examined the relationship between miR-3188, mammalian target of rapamycin (mTOR) and FOXO1 in NPC, and found an atypical miR-3188-mTOR-p-PI3K/AKT-c-JUN feedback loop modulated by FOXO1. This pathway suppresses proliferation and sensitizes NPC cells to 5-fluorouracil (5-FU). Altogether these results provide a mechanism by which miR-3188 modulates NPC cell growth.

miR-3188 suppresses cell growth and 5-FU chemoresistance.
To identify the role of miR-3188 in NPC development, we first examined its expression levels in normal epithelium (NP) and NPC cell lines. miR-3188 expression was elevated in NP69 and SXSW-1489 cells but weakly expressed in NPC cells (Fig. 1a). To further explore its biological role in NPC, miR-3188 mimics or inhibitors were respectively introduced into NPC or NP69 cell lines. More than threefold increase in miR-3188 expression was observed in HONE1-EBV and SUNE1 cells treated with miR-3188 mimics compared with the control group by qRT-PCR (Student's t-test, with Po0.05 for both) (Supplementary Fig. 1A). Due to higher endogenous miR-3188 expression in NP69 and 5-8F cells, miR-3188 inhibitors were transiently transfected into these lines. Expression of miR-3188 was significantly lower in the inhibitor-treated NP69 and 5-8F cells than in the control cells ( Supplementary Fig. 1B, Student's t-test, Po0.05 for both).
Next, we conducted an in vivo tumour formation experiment by subcutaneously injecting HONE1-EBV-miR-3188 and SUNE1-miR-3188 or control cells ( Supplementary Fig. 1D) into nude mice. After 18 days of implantation, the mice injected with HONE1-EBV-miR-3188 and SUNE1-miR-3188 cells had smaller tumour burdens (Fig. 1g) and displayed lower expression of Ki67 and proliferating cell nuclear antigen (PCNA) in tumour tissues relative to controls (Fig. 1h). These results suggested miR-3188 significantly inhibits tumorigenesis in vivo.
NPC cell lines stably overexpressing miR-3188 exhibited significantly increased sensitivity to 5-FU. Inhibition rates 48 h after treatment at different concentrations of 5-FU were calculated for cells with or without miR-3188 transfection (Fig. 1i). The IC50 for 5-FU in SUNE1 cells was reduced from 42 to 14 mM after miR-3188 transfection. A similar IC50 reduction from 23 to 8 mM occurred in 5-8F cells. Interestingly, obvious changes in the IC50 for diamminedichloroplatinum (DDP) treatment was not observed in miR-3188-treated NPC cells ( Supplementary Fig. 1E).
We then evaluated the in vivo anti-tumour efficacy of 5-FU in mice bearing tumours originating from miR-3188-overexpressing cells or their control lines. The weight of each group were measured every 3 days, and the results showed that tumour burden in mock þ 5-FU and miR-3188 þ 5-FU groups was slightly reduced compared to those in mock þ NS and miR-3188 þ NS groups, but there were no significant difference among the four groups ( Supplementary Fig. 1F). This suggested that 5-FU was well tolerated by the mice. Kaplan-Meier analysis showed the survival times mice in the mock þ 5-FU and miR-3188 þ NS groups were much longer than the mock þ NS group. However, the survival time of miR-3188 þ 5-FU treatment group was significantly longer than the other three groups (Fig. 1j) (log-rank test, Po0.001). There was no significant difference between the mock þ 5-FU group and mock þ miR-3188 groups. Average survival times of the mock þ NS, miR-3188 þ NS, mock þ 5-FU and miR-3188 þ 5-FU groups were 28.7, 36.4, 37.0 and 49.0 days, respectively.
To explore the mechanisms by which miR-3188 suppresses NPC cell proliferation, we found that miR-3188 overexpression downregulated c-JUN and CCND1 but enhanced p27 and p21. miR-3188 inhibitors rescued these decreased levels. Interestingly, miR-3188 knock-down in 5-8F cells exhibited opposite results and miR-3188 mimics could restore levels of these cell-cycle regulators (Fig. 1k). Furthermore, we found levels of p-PI3K and p-AKT were decreased in miR-3188-overexpressing SUNE1 and HONE1-EBV cells, yet increased in miR-3188-inhibited 5-8F cells (Fig. 1l). These results suggest that miR-3188 decreases cell growth by inactivating phosphoinositide-3-kinase (PI3K)/AKT as well as downstream c-JUN and G1/S cell-cycle transition signalling.
miR-3188 directly targets mTOR. Through TargetScan and RNAhybrid algorithms, mTOR was predicted to be a direct target of miR-3188 (Fig. 2a). Overexpression of miR-3188 downregulated mTOR mRNA and protein levels as well as p-mTOR levels in HONE1-EBV and SUNE1 cells. Conversely, miR-3188 downregulation elevated mTOR and p-mTOR levels in 5-8F and NP69 cells (Fig. 2b,c). Consistent with in vitro results, immunohistochemistry of xenografts generated from HONE1-EBV-3188 and SUNE1-3188 cells revealed a marked reduction in    mTOR expression (Fig. 2d). Simlarly, cotransfection miR-3188 mimics significantly decreased mTOR luciferase reporter activity (Fig. 2e Fig. 2A) enhanced cell proliferation by MTT (Fig. 3a) and EdU incorporation assays (Fig. 3b) as well as promoted G1 to S cell-cycle transition (Fig. 3c). mTOR overexpression significantly reversed the 5-FU sensitizing effects of miR-3188 in SUNE1 and 5-8F cells (Fig. 3d). Furthermore, we found that mTOR overexpression induced expression of c-JUN and CCND1 but reduced p27 and p21 (Fig. 3e). These results indicate that mTOR overexpression can overcome NPC cell growth suppression induced by miR-3188. Subsequently, we found that levels of mTOR, p-mTOR, p-PI3K, p-AKT, CCND1 and c-JUN were significantly decreased, while p27 and p21 were elevated after mTOR siRNA treatment (Fig. 3f). These results were consistent with miR-3188 overexpression, suggesting that mTOR is a direct target of miR-3188 responsible for suppressing cell growth and inducing NPC sensitization to 5-FU.
c-JUN inhibites miR-3188 by binding to its promoter region. To test the transcriptional regulatory mechanisms of miR-3188 expression, UCSC, PROMO and TFSEARCH bioinformatics software was utilized to analyze a 3-kb region upstream of the transcription start site of miR-3188. Three c-JUN-binding motifs at À 492 to À 498, À 1,628 to À 1,634 and À 2,356 to À 2,362 were identified inside the putative miR-3188 promoter region. These three transcription factor-binding sites (TFBSs) were named A, B and C (Fig. 4a). To examine the role of c-JUN in regulating miR-3188, we first used small-interfering RNAs (siRNAs) to suppress c-JUN expression in HONE1-EBV, SUNE1 and 5-8F cells ( Supplementary Fig. 2B). Next, quantitative PCR (qPCR) analysis indicated that miR-3188 expression was markedly increased in all lines after c-JUN knock-down ( Fig. 4b), suggesting that c-JUN is an upstream regulator of miR-3188.
To identify whether c-JUN-A, c-JUN-B or c-JUN-C was functional, we first performed electrophoresis mobility shift assay (EMSA) experiment to check whether nuclear extracts of SUNE1 and HONE1-EBV cells could bind to predicted sites A, B or C. As shown in Fig. 4c, a shift band was formed when the probe of digoxygenin (DIG)-ddUTP-labelled c-JUN was incubated with the nuclear protein extracted from SUNE1 and HONE1-EBV cells (lanes 2 and 8), whereas the band was nearly gone when unlabelled oligonucleo tides of c-JUN were added as binding competition (lanes 6 and 12). Bands were not affected when mutated A, B or C was added to compete with DIG-ddUTPlabelled A, B or C in SUNE1 cell and HONE1-EBV cell (lanes 3-5, lanes 9-11). The EMSA results demonstrate that the three predicted c-JUN-binding sites in the promoter region of miR-3188 were functional. Chromatin immunoprecipitation (ChIP) assays further confirmed that c-JUN protein was recruited to all the three binding sites in the putative miR-3188 promoter in SUNE1 and HONE1-EBV (Fig. 4d). Furthermore, a reduction of the wild-type miR-3188 promoter luciferase activity was observed on upregulation of c-JUN in the HEK293T, SUNE1 and HONE1-EBV cell lines (One-way ANOVA and Dunnett's multiple comparison test, Po0.05). A similar effect was observed when sites A and B, sites A and C, sites B and C were mutated respectively in 293T, SUNE1 and HONE1-EBV cells (One-way ANOVA and Dunnett's multiple comparison test, Po0.05) (Fig. 4e). These data indicate that c-JUN binds to specific promoter TFBS of miR-3188 and inhibits transcription. Taken together, our data suggest c-JUN is involved in miR-3188 transcription, and all the three c-JUN-binding sites of the miR-3188 promoter are functional in SUNE1 and HONE1-EBV cells.
FOXO1 acts as a tumour suppressor reducing cell growth. The latent membrane protein 1 (LMP1) of EBV is closely associated with NPC pathogenesis. To determine whether the expression of FOXO1 was influenced by EBV, HONE1 and HONE1-EBV lysates were examined by western blot. No obvious difference in FOXO1 was observed for the EBV infected variant of the line ( Supplementary Fig. 3A). In addition, no significant changes in FOXO1 protein were noted after siLMP1 or pSG5-LMP1 transfection in HONE1-EBV cells ( Supplementary Fig. 3B). On the basis of these data, we infer that total FOXO1 levels are not modulated by LMP1 in NPC.
To further confirm the growth-suppressive effect of FOXO1, we performed in vivo tumorigenesis experiment in nude mice. Tumour volumes and growth rates were significantly decreased in tumours derived from FOXO1-overexpressing HONE1-EBV and SUNE1 cells (Fig. 5e,f and Supplementary Fig. 3F). These tumours also exhibited a reduction in Ki67 and PCNA expression by immunohistochemistry (Fig. 5g). These results suggest that FOXO1 exerts a significant inhibitory effect on tumorigenesis in vivo. FOXO1 has been reported to induce the PI3K/AKT signalling in gastric cancer 21 , thus we sought to examine this effect in NPC. Overexpression of FOXO1 significantly reduced the levels of p-PI3K, p-AKT, mTOR and p-mTOR. Furthermore, we found that ectopic FOXO1-reduced expression of c-JUN and CCND1 but upregulated p21 and p27. Interestingly, the opposite results were observed after siRNA-mediated suppression of ectopic FOXO1 (Fig. 5h). Further, the specific PI3K inhibitor Ly294002 significantly reversed the expression of p-PI3K, p-AKT, mTOR, p-mTOR, c-JUN, CCND1, p21 and p27 (Fig. 5h).
As a downstream regulator of the PI3K/AKT pathway, c-JUN has been observed to directly suppress miR-3188 expression in NPC. We used immunofluorescence to confirm reduced expression of c-JUN in FOXO1-overexpressing NPC cells (Fig. 5i). This was also confirmed by immunohistochemistry of FOXO1-overexpressing tumour tissues derived from NPC mouse models (Fig. 5j). Finally, ChIP assay revealed less c-JUN binding to the miR-3188 promoter in FOXO1-overexpressing NPC cells compared to control cells (Fig. 5k).
All the results suggest that FOXO1 regulates NPC cell proliferation and cell-cycle progression through the PI3K/AKT/ c-JUN pathway.    Fig. 4). miR-3188 was confirmed as a positive modulator of FOXO1 via qRT-PCR in NPC cells treated with Mock, FOXO1 or both FOXO1 and siFOXO1 (Fig. 6a). Reduction of miR-3188 by its specific inhibitor could reverse the growth-suppressive effect after ectopic FOXO1 expression in MTT (Fig. 6b) and Edu assays (Fig. 6c,d). Western blot analysis showed that treatment with a miR-3188 inhibitor increased expression of p-PI3K, p-AKT, c-JUN and CCND1, but reduced p27 and p21 levels in FOXO1overexpressing NPC cells (Fig. 6e). These results indicate that miR-3188 is induced by FOXO1 and suppresses NPC cell growth. Specific PI3K inhibitor Ly294002 reversed the changes in miR-3188 expression in NPC cells with both FOXO1 overexpression or silencing (Fig. 6a). This suggests that FOXO1 positively regulated the expression of miR-3188 through the PI3K/AKT pathway.
Taken together, these results support that miR-3188 expression is induced by FOXO1 through PI3K/AKT/c-JUN signalling.
Pathoclinical features of miR-3188 expression. Levels of miR-3188 were significantly decreased in 8 NPC cell lines and NPCs compared to NP tissues by qPCR analysis (Student's t-test, P ¼ 0.00037, P ¼ 0.00033, respectively) (Fig. 7a). Further, in situ hybridization assay confirmed reduced expression of miR-3188 in NPC tissues compared to NP tissues ( Fig. 7b; Table 1). Clinical characteristics of the NPC patients are summarized in Table 2.
We did not find a significant association between miR-3188 expression level and patient age, sex, clinical stage (I-II versus III-IV), lymph node metastasis (N classification; N0-N1 versus N2-N3) or distant metastasis stage (M classification) in the 142 NPC cases. However, we observed that reduced-miR-3188 expression was negatively correlated with tumour size (T classification; w 2 test, P ¼ 0.011) ( Table 2). Subsequently, we found that NPC patients with high miR-3188 expression had longer survival times than those of patients with low miR-3188 levels (Log-rank test, P ¼ 0.009, Fig. 7c).
Correlation of miR-3188 with other key genes. To further confirm the relationship between miR-3188, mTOR, FOXO1 and c-JUN, we analyzed their mRNA expression in NPC and NP samples. As shown in Fig. 7d, mTOR and c-JUN expression were significantly higher in NPC than in NP samples (Student's t-test, P ¼ 0.0282, Po0.0001, respectively), while FOXO1 expression was significantly lower in NPC samples (Student's t-test, Po0.0001). miR-3188 expression was positively correlated with FOXO1 expression (Fig. 7e; Spearman's correlation coefficient, P ¼ 0.0326), but negatively associated with mTOR ( Fig. 7e; Spearman's correlation coefficient, P ¼ 0.0288) and c-JUN expression ( Fig. 7e; Spearman's correlation coefficient, P ¼ 0.0006) in the same NPC specimens. but also suppressed tumourigenicity in vivo. Furthermore, we also found that miR-3188 overexpression sensitized NPC cells to 5-FU, but not DDP. These results suggest that miR-3188 functions as a potential tumour suppressor in NPC. It is well established that cell-cycle progression is a predominant factor promoting tumour cell proliferation and inducing chemotherapeutic resistance to 5-FU and DDP 24 . The biological functions of miR-3188 identified in this study provide a mechanism for its role in carcinogenesis. miR-3188 forms a negative feedback loop via key oncogenic genes and signal including mTOR, PI3K/AKT and c-JUN, which suppress cellcycle signal transition, thus inhibiting cell growth and sensitizing cells to 5-FU. However, miR-3188 overexpression did not affect NPC cells response to DDP, which might attributed to the fact that miR-3188 induces ZEB2 expression and thus promotes an epithelial-mesenchymal transition (EMT)-like process in NPC ( Supplementary Fig. 5). While EMT has been widely studied for its role in early development and cancer metastasis, it can also affect resistance to platinum-based therapies 25 . The development of DDP resistance in NPC cells is accompanied by inducible EMT-like changes with an increased metastatic potential in vitro 26 .
mTOR functions as an oncogene in many tumour types [27][28][29] , including NPC 30 . The PI3K/Akt signalling pathway controls fundamental cellular processes, such as cell survival, growth, proliferation, cell repair, cell migration and angiogenesis, and is constitutively activated in nearly all cancer types [31][32][33] . Activation of PI3K/Akt/mTOR signalling through mutation of pathway components as well as through activation of upstream signalling molecules occurs in a majority of cancers 34,35 . The existence of a negative feedback loop between mTOR and PI3K/AKT has been demonstrated in many reports [36][37][38] . In our investigation, we found that miR-3188 directly targets mTOR and suppresses PI3K/AKT signalling. This finding contrasted previous studies in  breast cancer that mTOR suppresses PI3K/AKT signalling 36,37 . Expression of PI3K/AKT downstream components including cell-cycle factors, c-JUN and p-mTOR were dysregulated in mTOR-silenced NPC cells, a pattern consistent with miR-3188 overexpression. Furthermore, overexpression of mTOR reversed the inhibitory growth effect mediated by miR-3188 and promoted NPC cell proliferation. These results support that miR-3188 directly targets mTOR to suppress PI3K/AKT signalling, especially downstream cell-cycle factors c-JUN and p-mTOR. An analysis of a region upstream to the miR-3188 locus revealed multiple putative binding sites for c-JUN, an essential regulator of cell proliferation 39 , invasiveness, metastasis 40 and PI3K/AKT signalling 11,41 . Subsequent experiments confirmed that c-JUN could negatively regulate miR-3188 expression by directly binding its promoter region. These results thus indicate that miR-3188 can induce its own expression though a complex miR-3188-mTOR-pPI3K/AKT-c-JUN loop in NPC pathogenesis.
Although the tumour suppressive activity of FOXO1 has been well characterized for some cancers, little is known of its function and molecular mechanisms in NPC. Similar to previous reports 17,18 , FOXO1 significantly inhibited NPC cell G1/S cellcycle transition and proliferation in vitro and in vivo. Upstream cell-cycle signalling from the PI3K/AKT pathway was decreased in NPC cells after FOXO1 overexpression. These results are in contrast to a previous report in gastric cancer 21 . Two downstream regulators of PI3K/AKT signalling, c-JUN and p-mTOR, as well as total mTOR levels were reduced in NPC. Taken together, FOXO1 suppresses NPC cell growth by blocking PI3K/AKTmediated cell-cycle transition, a role consistent with that of miR-3188. To our knowledge, the detailed mechanisms for FOXO1-induced inhibition of mTOR protein have not been previously documented.
To investigate the effect of FOXO1 on miRNAs, we used a miRNA chip after FOXO1 overexpression in SUNE1 NPC cells. Among miRNAs differentialy express were miR-141 (ref. 42) and miR-29c (ref. 43), which have been reported to be involved in NPC carcinogenesis. We observed a marked upregulation of miR-3188 in FOXO1-overexpressing NPC cells and further confirmed these results by qPCR. We observed that miR-3188 could not induce the expression of FOXO1, which suggested FOXO1 could be an upstream regulator of miR-3188. Suppression of miR-3188 partially rescued the inhibitory growth effects of FOXO1 and promoted cell proliferation. This was achieved by suppressing mTOR-mediated activation of p-PI3K/AKT/p-mTOR, c-JUN and cell-cycle signalling. These data demonstrated that miR-3188 is a downstream effector of FOXO1 signalling and participates in FOXO1-induced growth suppression in NPC.
We observed that miR-3188 was negatively modulated by c-JUN, a downstream regulator of the PI3K/AKT pathway. Interestingly, this signalling cascade is suppressed by FOXO1. We suspected that miR-3188 is induced by FOXO1 by inhibiting PI3K/AKT/c-JUN signalling. Indeed, we found that miR-3188 expression was significantly reduced after PI3K inhibition in FOXO1-overexpressing NPC cells.
Consistent with their roles in vitro and in vivo, we observed that miR-3188 levels were significantly decreased in NPC tissues compared to NP tissues. Furthermore, reduced miR-3188 expression was negatively correlated with T classification and positively associated with the survival time of NPC patients. Patients that exhibited low miR-3188 expression had an overall shorter survival time compared to patients with high miR-3188 expression. Further, we also observed reduced FOXO1 but elevated mTOR and c-JUN levels in clinical NPC tissues compared to NP samples. miR-3188 was positively correlated with FOXO1 expression but negatively correlated with mTOR and c-JUN mRNA levels. Altogether these results suggest that miR-3188 exerts an important role in NPC tumorigenesis between FOXO1, c-JUN and mTOR dysfunction.
As summarized in our working model in Fig. 8, miR-3188 does not suppress cell growth and enhance chemotherapeutic sensitization to 5-FU alone. Instead, miR-3188 forms a negative feedback loop through mTOR/PI3K/AKT/c-JUN signalling that is modulated by FOXO1. There is an increasing appreciation that miRNAs form regulatory motifs with protein regulators, which underlie pathogenesis when dysregulated 44 . Therefore, context is important for understanding the role of specific miRNAs in regulatory networks. We hypothesize that once induced by FOXO1, the miR-3188-mediated feedback loop allows NPC cells to become less autonomous reducing cell proliferation. Tissue specimens. Forty nine (49) primary fresh NPC samples with tumour node metastasis (TNM) staging and 20 non-cancerous fresh nasopharyngeal samples were collected from the People's Hospital of Zhongshan City, China, at the time of diagnosis before any therapy. All fresh samples were immediately preserved in liquid nitrogen. One hundred and forty two (142) paraffin-embedded NPC specimens and 36 paraffin-embedded were obtained from the People's Hospital of Zhongshan City, Guangdong Province, China. In the 142 NPC cases, there were 99 male and 43 female with age ranging from 20 to 84 years (median, 58.9 years). For the use of these clinical materials for research purposes, prior consent from the patients and approval from the Ethics Committee of the People's Hospital of Zhongshan City were obtained. All specimens had confirmed pathological diagnosis and were staged according to the 1997 NPC staging system of the UICC.

Methods
In situ hybridization. The expression levels of miR-3188 in 142 paraffin-embedded NPC specimens and 36 paraffin-embedded NP tissues were detected by in situ hybridization. Tissue sections were dewaxed in xylene, rehydrated through an ethanol gradient and then treated with 3% H 2 O 2 for 10 min. Subsequently, sections were treated with pepsin dilution in 3% fresh citrate buffer at 37°C for 30 min and then washed. Further, hybridization with DIG-labelled miRCURY LNA probes (probe sense: Bis-P22401; Biosense Bioscience Co. Ltd, Guangzhou, China) was performed overnight at 37°C after pre-hybridization was carried out using 20 ml of pre-hybridization solution for 2 h at 37°C. Sections were subjected to high stringency washes with 2 Â SSC, 0.5 Â SSC and 0.2 Â SSC for 5, 15 and 15 min at 37°C. Afterwards, the sections were incubated in blocking solution for 30 min at 37°C and then incubated with alkaline phosphatase-conjugated sheep anti-DIG Fab fragments for 60 min at room temperature (RT). Positive staining of miR-3188 was observed by adding BM purple AP substrate (Roche, Basel, Switzerland) according to the manufacturer's instructions.
Lentivirus production and infection. Lentiviral particles carrying hsa-miR-3188 precursor and pGC-FU-FOXO1-RFP vector and their flanking control sequence (Mock for short) were constructed by GeneChem (Shanghai, China). SUNE1, HONE1, HONE1-EBV and 5-8F cells were infected with lentiviral vector, and polyclonal cells with green or red fluorescent protein signals were selected for further experiments using fluorescence-activated cell sorting. Hsa-miR-3188 expression was confirmed by qPCR and the levels of FOXO1 (Cat. No. 2880, 1:1,000, CST) protein were measured using western blotting.
Cell transfection. siRNA for FOXO1, c-JUN and mTOR or miR-3188 mimics and its inhibitor were designed and synthesized by RiboBio Inc. (Guangzhou, China) (Supplementary Table 1 qRT-PCR. RNA was extracted from the NPC cell lines, tissues and normal nasopharynx tissues by Trizol (Takara Bio, Inc., Shiga, Japan). U6 and ARF5 genes were used as miRNA and gene internal controls, respectively. Cycling conditions were 95°C for 10 min to activate DNA polymerase, followed by 45 cycles of 95°C for 15 s, 60°C (for miR-3188, c-JUN, LMP1 and FOXO1 ) for 15 s and 72°C for 10 s. Specificity of amplification products was confirmed by melting curve analysis. Independent experiments were done in triplicate. Specific sense primers for miR-3188, c-JUN, LMP1, FOXO1, U6 and ARF5 are shown in Supplementary Table 2.
Immunofluorescent staining. NPC cells grown on coverslips were rinsed with phosphate-buffered saline (PBS) and fixed with cold 4% paraformaldehyde for 5 min at RT. Subsequently, the cells were blocked with Triton X-100 at a concentration of 0.3% for 30 min and incubated with primary monoclonal antibodies c-JUN (Cat. No. 9165, 1:50, CST) in PBS for 2 hrs at RT. After three washes in PBS, the coverslips were incubated for 1 h in the dark room with Alexa Fluor 488 goat anti-rabbit IgG (1:500, Bioworld Technology, Inc.). Further, the coverslips were washed three times and then stained with 4-6-diamidino-2-phenylindole (DAPI) for 5 min at 4°C. Finally, ECLIP SE 80i fluorescent microscope (Nikon, Japan) was used to observe the expression of c-JUN in NPC cells.
Cell proliferation and colony formation assays. The MTT assay was used to examine cell viability. NPC cells (1,000/well) were seeded in 96-well plates. For lentivirus-mediated FOXO1 overexpression, the cells were incubated for 1, 2, 3, 4, 5, 6 or 7 days. For transient transfection with siFOXO1, miR-3188 mimics, miR-3188 inhibitor or mTOR plasmid et al, the cells were cultured for 1, 2, 3 or 4 days. Subsequently, 20 ml of MTT (5 mg ml -1 in PBS) (Sigma, St Louis, MO) solution was added to each well and incubated for 4 h. The formazan crystals formed by viable cells were solubilized in 150 ml dimethyl sulfoxide (Sigma, St Louis, MO) and then the absorbance value (OD) was measured at 490 nm. For colony formation assay, NPC cells were seeded in 6-well culture plates at a density of 100 cells/well and each group had 2 wells. After incubation for 14 days at 37°C, colonies were washed twice with PBS and stained with hematoxylin solution. The colonies composed of more than 50 cells in a well were counted under a microscope. All the experiments were repeated for at least three times. Cell-cycle analysis and EdU incorporation assay. Cell-cycle analyses and EdU incorporation assays were performed according to a previous description 45 . For cell-cycle analysis, a total number of 5 Â 10 6 NPC cells were harvested after 48 h incubation and then washed with cold PBS. The cells were further fixed with 70% ice-cold ethanol at 4°C overnight. After incubation with PBS containing 10 mg ml À 1 propidium iodide and 0.5 mg ml À 1 RNase A for 15 min at 37°C, fixed cells were washed with cold PBS three times. FACS caliber flow cytometry (BD Biosciences) was used to gain the DNA content of labelled cells. For EdU incorporation assay, proliferating NPC cells were examined using the Cell-Light EdU Apollo 488 or 567 In Vitro Imaging Kit (RiboBio) according to the manufacturer's protocol. Briefly, after incubation with 10 mM EdU for 2 h, NPC cells were fixed with 4% paraformaldehyde, permeabilized in 0.3% Triton X-100 and stained with Apollo fluorescent dyes. A total of 5 mg ml À 1 of DAPI were used to stain cell nuclei for 10 min. The number of EdU-positive cells was counted under a fluorescent microscope in five random fields. All assays were independently performed for three times.
In vivo tumorigenesis in nude mice. A total of 5 Â 10 6 logarithmically growing NPC cells transfected with miR-3188 and FOXO1 or the control (N ¼ 5 per group) in 0.1 ml 1,640 medium without FBS were subcutaneously injected into the leftright symmetric flank of the mice (BALB/C, nu/nu, 4-5-weeks-old, female). The mice were maintained in a barrier facility on HEPA-filtered racks. The animals were fed an autoclaved laboratory rodent diet. All animal studies were conducted in accordance with the principles and procedures outlined in the Southern Medical University Guide for the Care and Use of Animals under assurance number SCXK (Guangdong) 2008-0002. After 19 days, the mice were killed and tumour tissues were excised and weighed.