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

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 paraffin-embedded (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.

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].

The EBV-positive NPC cell line C666–1, the non-tumorigenic 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.

SB431542 (#S1067, SelleckChem, Houston, TX, USA), a TGFβ receptor I (TGFβR1, also known as ALK5) inhibitor, was used as indicated. Human TGFβ1 (#8915; Cell Signaling, Danvers, MA, USA) was used where indicated after overnight starvation of cells in Minimum Essential Media (MEM) supplemented with 0.5% FBS.

Polyplus-transfection JetPRIME (Graffenstaden, France) was used for transfection of C666–1, NP69, NP460, and HEK 293 T cells, according to manufacturer’s specifications. C666–1, NP69, and NP460 cells were transfected with pre-miR-34c or pre-miR negative control (20 nM and 50 nM, Ambion, Austin, TX, USA).

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.

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].

Immunoprecipitation buffer (150 mM NaCl, 5 mM EDTA, 50 mM Hepes pH 7.6, 1–2% Nonidet P-40; with protease inhibitor cocktail, Roche), was used for protein extraction. Electrophoresis was performed with Bolt 4–20% Gels (Life Technologies, Carlsbad, CA, USA).

The Epithelial-Mesenchymal Transition Antibody Sampler Kit (Cell Signaling; #9782; 1/1000 each), anti-TGFβ1 (Cell Signaling; #3711; 1/1000), and anti-β-actin (Sigma: 1/5000) antibodies were used. The SuperSignal West Femto ECL (Pierce, #34095, Thermo Scientific, Waltham, MA, USA) was used for ZEB1, CDH1 and ZO-1 detection. Pierce ECL (#32209) was used to detect all other proteins.

RNA from our cohort of FFPE samples was isolated (200 ng/sample), processed (Ribo-Zero Gold rRNA Removal Kit (Illumina, San Diego, CA, USA)), and sequenced as previously described (as in the NanoString section of [11]). A subset of these samples (n = 53) was processed for RNA-seq. Library preparation was performed using the TruSeq Stranded Total RNA Sample Prep Kit (Illumina, San Diego, CA, USA). Sequencing was conducted on the Illumina HiSeq 2000 to > 100 million paired-end 100 bp reads. STAR (v2.4.2a) was used to align the reads [60], and RSEM (v1.2.21) was used to summarize expression values [61].

MiR-34c was predicted to target the wild-type (WT) 3′-untranslated region (3’UTR) of SOX4 in silico. This region was inserted into the pMIR-REPORT vector (Ambion). JetPRIME was used to reverse transfect HEK 293 T cells with pre-miR-control or pre-miR-34c. Twenty-four hours later, JetPRIME was used to co-transfect pRL-SV Renilla vector (Promega, Madison, WI, USA) with either pMIR-SOX4 3’UTR WT (CTAGTGCTCAGCTCAAGTTCACTGCCTGTCAGAT) or pMIR-SOX4 3’UTR Mutant (CTAGTGCTCAGCTCAAGTTTCTGTAAAGTCAGAT). The Dual-Luciferase Reporter Assay (Promega) was used to measure luciferase activity 24 h post-transfection.

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).

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.

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 (GraphPad 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).

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.

We had previously demonstrated that miR-449b overexpression, another component of the validated prognostic DM signature [11], led to TGFBI mRNA degradation with subsequent TGFβ1 accumulation [12]. Given that TGFβ1 plays an important role in NPC progression [53, 63‐68] and in the regulation of miRNAs, particularly miR-34a [52], we sought to measure TGFβ1 in these cell lines. Indeed, C666–1 cells (which have high miR-449b expression [12]) expressed higher levels of active TGFβ1 compared to either NP69 or NP460 cells (both of which have lower miR-449b expression [12]) (Fig. 1c). We therefore hypothesized that TGFβ1 could be regulating miR-34c in these cells. Treatment with recombinant TGFβ1 significantly reduced miR-34c expression in both NP69 and NP460 cells (Fig. 1d and e). Conversely, a TGFβ receptor 1 (TGFBR1) inhibitor (SB431542) increased miR-34c expression in C666–1 cells (Additional file 1: Figure S1A).

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.

In order to identify miR-34c target candidates, 17 genes at the intersection between computationally predicted targets and genes upregulated in patient NPC samples [69] were examined (Fig. 2a). Using qRT-PCR, 6 of the 17 genes were observed to be upregulated in C666–1 (low miR-34c) compared to NP69 and NP460 cells (high miR-34c) (Additional file 1: Figure S1B and C). These genes were then assessed for response to transient miR-34c overexpression (pre-miR-34c transfection) (Fig. 2b for the 6 genes; Additional file 2: Figure S2A for the other 11 genes), and TGFβ pathway inhibition using SB431542 (a TGFBR1 inhibitor, which also upregulates miR-34c) (Fig. 2c for the 6 genes; Additional file 2: Figure S2B for the remaining 11 genes) in C666–1 cells. As can be seen in Fig. 2b and c, elevated miR-34c conditions consistently and significantly downregulated ARID5A, BIK, and SOX4. Interestingly, BAX and PML were consistently and significantly upregulated (Additional file 2: Figure S2A and B), suggesting that they are not direct targets of miR-34c, but possibly further downstream or altered via a more complex mechanism.

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.

SOX4 has been characterized as a master regulator of EMT [25, 27], notably by upregulating SOX2 [19‐22], a well-known mediator of tumour initiation and cancer stem cell maintenance [72‐74]. We therefore hypothesized that miR-34c could affect EMT via SOX4 and SOX2. First, SOX2 was confirmed to be highly expressed in C666–1 cells (low miR-34c; high SOX4) compared to NP69 and NP460 cells (high miR-34c; low SOX4) (Fig. 3a). NP69 cells stably expressing SOX4 had a significant increase in SOX2 expression (Additional file 3: Figure S3A), corroborating previous reports [19‐22]. Moreover, downregulation of miR-34c in both NP69 and NP460 anti-miR-34c stable cell lines led to the significant upregulation of SOX2 (Fig. 3b and c). The overexpression of miR-34c in C666–1 correspondingly decreased SOX2 transcript levels (Additional file 3: Figure S3B).

The expression of well-known EMT markers were then investigated. NP69 anti-miR-34c stable cells overexpressed SNAI1 (Snail), ZEB1, and CDH2, while under-expressing CLDN1 (Claudin-1), ZO-1, and CDH1 (Fig. 3d). Similar results were observed in NP460 anti-miR-34c stable cells (Fig. 3e), supporting the role of miR-34c downregulation in the promotion of EMT in normal nasopharyngeal cell lines. C666–1 cells were not amenable to this gene expression analysis (ZEB1, CDH2, and CLDN1 are not expressed). However, TGFBR1 inhibition using SB431542 decreased SOX2 transcript expression in C666–1 cells (Fig. 3f). Taken together, the data show that high levels of TGFβ1 downregulate miR-34c, which directly leads to SOX4 overexpression and consequent SOX2 upregulation, promoting EMT in nasopharyngeal cells.

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).