MiR-34c downregulation leads to SOX4 overexpression and cisplatin resistance in nasopharyngeal carcinoma

Background A major cause of disease-related death in nasopharyngeal carcinoma (NPC) is the development of distant metastasis (DM) despite combination chemoradiotherapy treatment. We previously identified and validated a four microRNA (miRNA) signature that is prognostic for DM. In this study, characterization of a key component of this signature, miR-34c, revealed its role in chemotherapy resistance. Methods Two hundred forty-six NPC patient biopsy samples were subject to comprehensive miRNA profiling and immunohistochemistry (IHC). Two human normal nasopharyngeal cell lines (immortalized; NP69 and NP460), as well as the NPC cell line C666–1, were used for miR-34c gain-of-function and loss-of-function experiments. Signaling pathways were assessed using quantitative real-time PCR (qRT-PCR) and Western blot. Cell viability was measured using the ATPlite assay. Results MiR-34c was downregulated in NPC patient samples, and confirmed in vitro to directly target SOX4, a master regulator of epithelial-to-mesenchymal transition (EMT). MiR-34c downregulation triggered EMT-representative changes in NP69 and NP460 whereby Snail, ZEB1, CDH2, and SOX2 were upregulated, while Claudin-1 and CDH1 were downregulated. Phenotypically, inhibition of miR-34c led to cisplatin resistance, whereas miR-34c over-expression sensitized NPC cells to cisplatin. TGFβ1 decreased miR-34c and increased SOX4 expression in vitro. The TGFβ receptor 1 inhibitor SB431542 reduced SOX4 expression and increased cisplatin sensitivity. Finally, IHC revealed that lower SOX4 expression was associated with improved overall survival in chemotherapy-treated NPC patients. Conclusion miR-34c is downregulated in NPC. Repression of miR-34c was shown to increase SOX4 expression, which leads to cisplatin resistance, while TGFβ1 was found to repress miR-34c expression. Taken together, our study demonstrates that inhibition of the TGFβ1 pathway could be a strategy to restore cisplatin sensitivity in NPC.

Our group previously completed a global miRNA NPC patient sample profiling, discovering and validating a four-microRNA (miRNA) prognostic signature associated with risk for DM (low miR-34c, low miR-140, high miR-154, and high miR-449b) [11]. A subsequent study demonstrated that elevated levels of miR-449b were significantly associated with poor OS in patients receiving concurrent chemoradiotherapy [12]. MiR-449b overexpression in NPC was found to decrease transforming growth factor beta-induced (TGFBI), leading to an increase in transforming growth factor beta 1 (TGFβ1), TGFβ pathway activation, and cisplatin resistance [12].
TGFβ1 is a secreted protein involved in the regulation of many cellular mechanisms, such as metastasis formation, chemoresistance, epithelial-to-mesenchymal transition (EMT) [13,14], and more recently, miRNA expression [15,16]. This latter process occurs via TGFβ1-mediated Smad activation whereby Smads bind to miRNA promoter regions that contain Smad-binding elements, as well as the Drosha complex [17]. Conversely, numerous miRNAs have been shown to negatively regulate TGFβ pathways [18].
In this current study, the biological mechanisms and effects of miR-34c downregulation were investigated. The data suggest that this downregulation is caused by TGFβ1, which leads to SOX4 disinhibition, which in turn promotes EMT and cisplatin resistance in NPCtwo features that contribute to the formation of DM.

Patient samples
In compliance with the Institutional Research Ethics Board at the University Health Network (UHN), all patients provided written consent for the use of their tissues in this study. Diagnostic formalin-fixed paraffinembedded (FFPE) blocks were obtained from NPC patients (n = 246) treated at the Princess Margaret Cancer Center (PMCC) between 1993 to 2009, as previously described [11]. FFPE tissues from patients who underwent quadroscopy and were not diagnosed with NPC (n = 17) were used as normal nasopharyngeal epithelial tissues.

NanoString analysis
RNA was isolated using the Recover All Total Nucleic Acid Isolation Kit for FFPE (Ambion, Austin, TX, USA). Total RNA (200 ng) was assayed using the nCounter Human miRNA Assay v1.0 (Nanostring; 734 unique human and viral miRNAs). Please note that this experiment was also used for a previous study. Full analyses and protocols can be found in Bruce et al. [11].

Cell culture
The EBV-positive NPC cell line C666-1, the nontumorigenic human nasopharyngeal cell lines NP69 (SV40-immortalized) and NP460 (hTert-immortalized), and HEK 293 T cells were cultured as previously described [12]. NP69 and NP460 cell lines were generated by SW Tsao's group [55,56] and served as "normal" cells throughout this study. Every new batch of cells underwent mycoplasma testing and STR analyses [12]. C666-1, NP69 and NP460 cells were used for gain-and loss-of-function assays; HEK 293 T (ATCC CRL-32 L) cells were used for lentiviral generation and luciferase assays.

Lentiviral transduction
Lentiviral transduction was used to generate stable cell lines as previously described [12]. pLV-miRNA-34c (Biosettia, San Diego, CA, USA), pLV-miR-34c-lockers (Biosettia, San Diego, CA, USA), and their respective control vectors were used. All stable cell lines were generated for the purpose of this work.

Quantitative real-time PCR (qRT-PCR)
The Total RNA Purification Kit (Norgen Biotek, Thorold, ON, Canada) was used for both mRNA and miRNA isolation. Reverse-transcription of total RNA (1 μg) was performed using the iScript cDNA Synthesis Kit (BioRad, Hercules, CA, USA). qRT-PCR was performed using SYBR Green (Roche, Basel, Switzerland) and the primers are listed in Table 1. mRNA expression was normalized to the average expression of two housekeeping genes (β-actin and GAPDH, as in [12]) and melting curves were generated for each experiment. MiRNA levels were assessed using the TaqMan MicroRNA Assay, and processed according to manufacturer's instructions (Applied Biosystem, Foster City, CA, USA). RNU44 and RNU48 were used to normalize miR-34c expression [57,58]. Relative expression was calculated using the 2 -ΔΔCt method [59].

Cell viability assays
Stable cell lines generated from C666-1, NP69 and NP460 cells were seeded in 96-well plates (2000 cells/ well). After 1 day, they were exposed to decreasing concentrations of cisplatin (CDDP) for 72 h as indicated in the figures. Dose-response curves for cisplatin were determined through treatment using two-fold serial dilutions starting from 12.5 μg/mL (which induced~90% cell death in NP69/NP460 cells after 72 h of treatment). Cell viability was assessed using the ATPlite 1 Step Luminescence Assay System (PerkinElmer, Waltham, MA, USA).

Immunohistochemistry (IHC)
Sections from FFPE blocks were subject to IHC using microwave antigen retrieval. Citric acid (0.01 M, pH 6.0) and the LSAB+ System-HRP (Dako, Les Ulis, France) were used. Rabbit polyclonal anti-SOX4 (PA5-41442, lot#SB2344261A, Invitrogen: 1/40) antibody was used, but omitted for negative control staining. Positive nuclear SOX4 localization was detected by light microscopy. The percentage of positive tumour cells was quantified by evaluating a total of at least 300 tumour cells from the three most densely staining fields (magnification 400×). A final score was calculated as the product of the percentage of positive tumour cells and staining intensity (0 = negative; 1 = weak; 2 = moderate; 3 = strong) as previously described [62]. No samples had an intensity score of 3. All scoring was performed blinded to any knowledge of clinical or pathological parameters. Each section was scored at least twice.

Statistical analyses
All experiments were performed at least three times. In order to maintain independence between replicates, new frozen batches of cells were used each time. Data are presented as the mean ± SEM. GraphPad Prism (Graph-Pad Software, San Diego, CA, USA) was used for statistical analyses. Intergroup statistical significance was determined using the ANOVA test, with the Bonferroni post-test (if applicable), or the Mann-Whitney U test (socscistatistics.com).

MiR-34c is downregulated by TGFβ1
In order to investigate the role of miR-34c downregulation in the validated prognostic signature for NPC DM [11], we first confirmed that miR-34c expression was significantly reduced in NPC diagnostic FFPE samples compared to normal nasopharyngeal tissues using previously generated NanoString data [11] (Fig. 1a). Cell line models were then assessed for miR-34c expression. EBV-positive NPC cell line C666-1 exhibited significantly lower levels of miR-34c compared to the two normal (immortalized) nasopharyngeal cell lines NP69 and NP460 (Fig. 1b), consistent with clinical observations.
In order to confirm the association between increased miR-449b, increased TGFβ1, and decreased miR34c, NP69 cells stably expressing pre-miR-449b were compared to NP69 cells stably expressing miR-control or anti-miR-34c. NP69-pre-miR-449b cells expressed higher levels of active TGFβ1 protein compared to NP69-miR-control or NP69-anti-miR-34c cells (Fig. 1f, top); associated with a correspondingly lower expression of miR-34c compared to NP69-miR-control (Fig. 1f,   bottom). Taken together, these data support the hypothesis that TGFβ1 decreases miR-34c expression, although the mechanism of regulation remains unknown.
The expression of the potential miR-34c targets was then determined through qRT-PCR on NP69 and NP460 cells transiently transfected with pre-miR-34c (Additional file 2: Figure S2C and D), as well as on NP69, NP460, and C666-1 cells stably expressing pre-miR-34c and anti-miR-34c (Additional file 2: Figure S2E, F, G, H, I and J). Together, these data show that only SOX4 was both significantly and inversely related to miR-34c in all tested cell line models. SOX4 is potentially important in the tumorigenesis of a number of different cancers (reviewed in [70]), including NPC [35,71]. It is also known to be regulated by TGFβ1 [19], although its relationship with miR-34c remains to be investigated. Thus, we proceeded to interrogate the relationship between the TGFβ pathway, miR-34c, and SOX4.
First, miR-34c-mediated direct inhibition of SOX4 expression was confirmed using a luciferase reporter assay (Fig. 2d). The data were corroborated in NP69 cells, wherein TGFβ1 treatment significantly increased SOX4 expression, which was abrogated with miR-34c overexpression (Fig. 2e). Furthermore, RNA-seq performed on 53 diagnostic NPC biopsy samples revealed that patients with higher than median SOX4 transcript levels experienced a lower 10-year distant relapse-free survival (DRFS) compared to those with lower levels (p = 0.063) (Fig. 2f). Taken together, these data suggest that elevated TGFβ1 (via miR-449b upregulation (Fig. 1f) and consequent TGFBI degradation [12]) may lead to the downregulation of miR-34c, which directly upregulates SOX4 overexpression, possibly leading to an inferior 10-year DRFS, as seen in this dataset.

TGFBR1 inhibition sensitizes C666-1 cells to cisplatin
Our group previously demonstrated that miR-449b overexpression was associated with EMT and cisplatin sensitivity in NPC [12], with EMT being a well-described mediator of chemoresistance [75]. In this current study, miR-34c was found to be downregulated by TGFβ1 (Fig.  1), leading to EMT. On this basis, the potential involvement of miR-34c in cisplatin resistance was examined. Downregulation of miR-34c using anti-miR-34c significantly increased resistance to cisplatin in NP69, NP460, and C666-1 stable cell lines ( Fig. 4a and b, Additional file 3: Figure S3C). Conversely, overexpression of miR-34c using pre-miR-34c increased cisplatin sensitivity in NP69, NP460, and C666-1 stable cell lines (Additional file 3: Figure S3D and E, Fig. 4c). Additionally, SB431542 treatment had a cytotoxic effect on C666-1 cells in a dose-dependent manner in vitro (Additional file 3: Figure S3F). The combination of SB431542 and cisplatin had an additive effect on the cell death of C666-1 cells (Fig. 4d). Finally, IHC performed on NPC biopsy samples from patients treated with chemoradiation (n = 25) demonstrated that lower SOX4 nuclear immunostaining was associated with a superior 10-year OS compared to patients with high SOX4 immunostaining (p = 0.031; Fig.  4e, and Additional file 4: Figure S4). These data all support a role for the TGFβ1-miR34c-SOX4-SOX2 pathway in mediating cisplatin sensitivity in NPC.
In summary, miR-34c acts as a switch that controls EMT and chemoresistance in NPC. With TGFβ1 stimulation, miR-34c is repressed, directly leading to an increase in SOX4, which consequently upregulates SOX2, leading to EMT and cisplatin resistance in NPC (Fig. 4f).

Discussion
This study revealed a novel role of miR-34c in EMT and chemoresistance in NPC. Downregulation of miR-34c in our cellular model, caused at least partially by miR-449b overexpression and consequent TGFβ1 activity, resulted in SOX4 and SOX2 overexpression, which triggered EMT and cisplatin resistance (Fig. 4f). Concordantly, miR-34c overexpression sensitized NPC cells to cisplatin-a phenotype corroborated in other cancer types [76][77][78][79].
Interestingly, miR-34c and miR-449b belong to the same miRNA family, as their seed sequences are highly similar (reviewed in [80]). Despite having potentially overlapping predicted targets however, as illustrated in this study, they do not function in the same manner in every context. Our data do demonstrate a similar effect wherein both miR-449b and miR-34c lead to the same cellular outcome: EMT and cisplatin resistance. Further experiments would be required to unravel the roles of the other members of the miR-34/449 family in NPC. d and e Western blots (WBs) were performed using anti-Snail (αSnail), anti-Claudin-1 (αClaudin-1), anti-ZO-1 (αZO-1), anti-ZEB1 (αZEB1), anti-CDH1 (αCDH1), anti-CDH2 (αCDH2), with anti-β-actin (αβ-actin) as loading control. d WBs using NP69-anti-miR-34c stable cells and their control. Fulllength blots are presented in Additional file 6: Figure S6. (E) WBs using NP460-anti-miR-34c stable cells and their control. ZO-1 was undetectable in NP460 cell lines. Full-length blots are presented in Additional file 7: Figure S7. f Relative SOX2 expression assessed by qRT-PCR in C666-1 cells treated with SB431542 (10 or 20 μM). The data are expressed as the mean ± SEM of at least three independent experiments. * P < 0.05; *** P < 0.001 In NPC, miR-34c downregulation has been previously reported by several groups [11,38,39], but its mechanism of action has never been determined. This study elucidated a clear signaling pathway and provides data suggesting a myriad of other miR-34c effects. For example, our data demonstrated that miR-34c overexpression increased the expression of well-known pro-apoptotic genes, such as BAX [81] and PML [82]. Interestingly, the inhibition of PML nuclear bodies by the EBV protein EBNA1 has been described to contribute to tumorigenesis in NPC cells [83,84]. MiR-34c has also been reported to suppress tumorigenesis through MET inhibition [38]. These and other miR-34c relationships remain to be further investigated in NPC.
Other miR-34 family members have been shown to be pro-apoptotic [44], with a liposome containing a miR-34a mimic (MRX34) being developed and evaluated Fig. 4 MiR-34c and the TGFβ pathway are involved in cisplatin sensitivity in NPC cells. a to d Cell viability was assessed 72 h after cisplatin treatment using the ATPlite assay. a Stable NP69-anti-miR-34c (or control) cells. b Stable NP460-anti-miR-34c (or control) cells. c Stable C666-1pre-miR-34c (or control) cells. d C666-1 cells were treated simultaneously with combinations of cisplatin and SB431542 with varying doses. The data are represented as the mean ± SEM of at least three independent experiments. * P < 0.05; ** P < 0.01; *** P < 0.001. e Kaplan-Meier curve of OS based on low (<median) vs. high (>median) SOX4 expression (nuclear staining of tumour cells) using an anti-SOX4 polyclonal antibody in 25 NPC patients treated with chemoradiation (median follow-up time = 5 years). f Proposed model for the miR-449-TGFβ1-miR-34c-SOX4 pathway [12]. The red dotted line indicates that the mechanism remains unknown clinically as a therapeutic agent [85]. Additionally, while miR-34a regulates SOX2 expression through PAI-1 [86], its overexpression reverts EMT, which suppresses invasion in NPC [53] and enhances docetaxel sensitivity in prostate cancer [87].
There has been increasing evidence supporting a primary role for TGFβ pathway activation in NPC [12,53,63,[65][66][67]. This current study demonstrated that miR-34c can be downregulated by TGFβ1, and that miR-449b overexpression can cause similar effects. Correspondingly, miR-449b upregulation and miR-34c downregulation were components of the four-miRNA prognostic signature for DM in NPC [11]. Cellular models mimicking these miRNA dysregulations display mesenchymal features and resistance to cisplatin, which are known contributors to disease recurrence and metastasis [12,88,89]. Furthermore, in C666-1 cells, TGFβ pathway inhibition produced a similar gene expression profile to transient miR-34c overexpression (i.e. NOTCH1, TGIF2, BAX, and PML), suggesting a close relationship between TGFβ1 and miR-34c pathways. The relationship between these pathways and chemoresistance should be a potential avenue of investigation for future translational studies.

Conclusion
This study elucidates the novel role of miR-34c in EMT and cisplatin resistance. TGFβ1 negatively regulates miR-34c, which in turn increases the expression of SOX4 and SOX2, mediators of EMT triggering leading to cisplatin resistance (Fig. 4f). Correspondingly, miR-34c overexpression and TGFβ pathway inhibition leads to cisplatin sensitivity in NPC, highlighting a potential therapeutic strategy for this complex disease.
Additional file 1: Figure S1. Genes with no significant differences in expression between C666-1 cells and NP69/NP460. Note that MARCKS and PML expression were significant only between C666-1 and NP460 cells. The data are represented as the mean ± SEM of at least three independent experiments. * P < 0.05; ** P < 0.01; *** P < 0.001. The funding bodies did not influence the design, collection, analysis, or interpretation of the study. The funding bodies did not influence the writing of the manuscript.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
In compliance with the Institutional Research Ethics Board at the University Health Network (UHN), all patients provided written consent for the use of their tissues in this study.

Consent for publication
Not applicable.