MicroRNA 744-3p promotes MMP-9-mediated metastasis by simultaneously suppressing PDCD4 and PTEN in laryngeal squamous cell carcinoma

MicroRNA controls cancer invasion by governing the expression of gene regulating migration and invasion. Here, we reported a novel regulatory pathway controlled by miR-744-3p, which enhanced expression of matrix metallopeptidase 9 (MMP-9) in laryngeal squamous cell carcinoma (LSCC). We profiled the differential micoRNA expression pattern in LSCC cell lines and normal epithelial cultures derived from the head and neck mucosa using microRNA microarray. MiR-7-1-3p, miR-196a/b and miR-744-3p were expressed differentially in the LSCC cell lines. Subsequent validation using real-time PCR revealed that high miR-744-3p level was positively correlated with regional lymph node metastasis of LSCC. Real-time cellular kinetic analysis showed that suppressing miR-744-3p could inhibit migration and invasion of LSCC cell lines and reduce the number of lung metastatic nodules in nude mice modules. In silico analysis revealed that miR-744-3p targeted 2 distinct signaling cascades which eventually upregulated MMP-9 expression in LSCC. First, miR-744-3p could suppress programmed cell death 4 (PDCD4), a direct suppressor of NF-κB (p65). PDCD4 could also prevent AKT activation and suppress MMP-9 expression. Further, suppressing miR-744-3p expression could restore phosphatase and tensin homolog (PTEN) expression. PTEN could inhibit AKT activation and inhibit MMP-9 expression in LSCC cells. The results revealed that suppressing miR-744-3p was effective to inhibit LSCC metastasis by inactivating AKT/mTOR and NF-κB (p65) signaling cascade. Targeting miR-744-3p could be a valuable therapeutic intervention to suppress the aggressiveness of LSCC.


INTRODUCTION
Cancer of the larynx originating from the head and neck region is the second most common upperaerodigestive cancer [1,2]. Histologically, laryngeal squamous cell carcinoma (LSCC) is the major form. Despite the advances of cancer treatment in the past few decades, the prognosis of LSCC patients remains poor without significant improvement [3]. Recent data reveals that the morbidity and mortality associates with LSCC are increasing in recent year [3,4]. Local invasion and metastatic spread of cancer cells is one of the main causes of death in LSCC patients [5]. Laryngeal cancer patients present with cervical lymph node involvement have poorer disease free interval and overall survival [6]. Given that the key cellular and molecular mechanisms resulting in LSCC migration and invasion are less understood, investigation on the underlying mechanism will be clinically beneficial for developing molecular targeting therapy.
MicroRNAs are highly conserved non-coding RNAs (~18-25 nucleotides) involved in gene regulation at posttranscription level. The single-stranded mature microRNA can bind to the target mRNA at the 3′ untranslated region (3′UTR) forming a stable duplex by Watson-Crick MicroRNA 744-3p promotes MMP-9-mediated metastasis by simultaneously suppressing PDCD4 and PTEN in laryngeal squamous cell carcinoma complementarities at a partial complementary manner. By recruiting and incorporating with the RNA-induced silencing complex (RISC), translation of the target mRNA transcript is hindered and/or the mRNA is subjected to degradation [7,8]. MicroRNA is recognized as a master gene regulator as individual microRNA could target a plethora of mRNA transcript at multiple sites the same time [9]. This multiple target feature makes microRNA an efficient effector in target-based cancer therapy which offers unique advantages as compared with the conventional treatment modalities [10].
In order to identify the key functional microRNA in LSCC, we performed microRNA profiling in LSCC cell lines using microRNA microarray and validated the results in LSCC tissues. We found that deregulation of miR-744-3p was a common event in LSCC. At present, however, its functions in LSCC remains to be defined. In silico analysis showed that miR-744-3p could directly target the mRNA transcript of both programmed cell death 4 (PDCD4) and phosphatase and tensin homolog (PTEN), both of which had been reported to be correlated with LSCC metastasis [11][12][13]. Reduced PDCD4 was usually found in aggressive head and neck cancers [14]. PDCD4 knock-out mice showed high systematic dissemination rate implying the functional implication in the metastatic process [15,16]; PTEN, on the other hand, was a well-known anti-neoplastic factor [17], which antagonized the action of PI3K by converting PIP3 to PIP2 via dephosphorylation [18]. Both PTEN and PDCD4 were identified as the upstream suppressors of matrix metallopeptidase 9 (MMP-9) which facilitated cancer cell migration through degrading the collagenous substrates in the surrounding extracellular matrix [19]. Our results revealed a novel pathway employed by LSCC in promoting LSCC migration and metastasis by overexpressing miR-744-3p. Table 1 showed the deregulated microRNA expression profile in LSCC cell lines. MiR-7-1-3p, miR-196a, miR-196b and miR-744-3p were detected in LSCC but not normal epithelial cultures. In comparison, let-7a-3p, miR-34a-3p, miR-338-5p and miR-365a-5p could only be detected in normal epithelial culture. Twenty-three microRNAs showed significant difference in expression level between the LSCC and normal cell lines (1.5-fold, P < 0.05) (Figure 1). The microarray data are publicly available at GEO (Accession No. GSE73171).
Three microRNA (miR-196a, miR-196b and miR-744-3p) were significantly upregulated in the LSCC tissue (P < 0.05). Let-7a-3p was not detected in all the LSCC tissues and the paired normal epithelia. MiR-7-1-3p, miR-338-5p and miR-365a-5p were detected in the LSCC tissues and the paired normal epithelia. However, there were no significant difference in the expression level between cancer and the normal tissues (P > 0.05). In the microarray results, miR-34a-3p expression was found in normal epithelial cell lines and was undetectable in the LSCC cell lines. In the validation set using laryngeal tissues, however, miR-34a-3p was significantly upregulated in tumor (P = 0.013). Thus, we shortlisted miR-196a, miR-196b and miR-744-3p as candidate microRNA and explored their clinical significance by evaluating the statistical association with the clinicopathological parameters of LSCC patients.
All the LSCC cases were grouped into high expression and low expression group using median expression level of each microRNA in LSCC as cut-off points. As shown in Table 2, expression levels of miR-196a and miR-196b were not statistically associated with the age, smoking habit, drinking habit, T-stage, regional lymph node status, or clinical stage of our LSCC patients. In contrast, high miR-744-3p expression was significantly associated with the cervical lymph node metastasis in LSCC (P = 0.007). Thus, we proposed that miR-744-3p overexpression could promote migration and invasion of the LSCC cells.

Suppressing miR-744-3p reduced the metastatic ability of LSCC
To explore the functional impact of miR-744-3p overexpression on LSCC migration and invasion, we monitored the dynamic changes of migration and invasion capacities of LSCC cell lines (SNU899 and SNU1076) expressing the miR-744-3p shRNA. Significant reduction in migration and invasion ability was observed in both SNU899 and SNU1076 ( Figure 2F). To confirm the in vitro data, the miR-744-3p suppressed LSCC cell line was injected into nude mice and the formation of lung metastatic foci were evaluated on day 60 after injection. Mice injected with miR-744-3p suppressed LSCC cells group had lower number of metastatic cancer nodules in comparison with the control group ( Figure 2G-2J).

In silico analysis showed that miR-744-3p targeted PDCD4 and PTEN simultaneously
To evaluate the functional target of miR-744-3p, the target transcript was predicted using miRWalk. In total, www.impactjournals.com/oncotarget 2004 genes were simultaneously predicted by 2 or more databases. Of which, 13 genes had been reported to be downregulated in LSCC tissues [20][21][22][23][24][25][26][27][28]. Next, we used the mirSVR score to narrow down the gene list. Six transcripts (PDCD4, PTEN, SCN2B, IGFBP5, GPR64, BOC) had mirSVR score < -0.01 suggesting that they could potentially form thermodynamically stable complexes with the mature miR-744-3p (Table 3). Among the 6 genes, only PDCD4 and PTEN were reported to consistently suppress head and neck cancer cell metastasis in previous studies. PTEN restoration could significantly suppress migration and invasion in LSCC and oral carcinoma. On the other hand, low PDCD4 expression was significantly related with lymph node metastasis in head and neck malignancy, and PDCD4 upregulation could inhibit invasiveness of oral cancer and nasopharyngeal carcinoma cells [29][30][31][32][33][34].  Yes Yes No ----------Twenty-three microRNAs showed significant difference in expression level between the LSCC and normal cell lines (1.5-fold, P < 0.05). MiR-7-1-3p, miR-196a, miR-196b and miR-744-3p were detected in LSCC but not normal epithelial cultures. In comparison, let-7a-3p, miR-34a-3p, miR-338-5p and miR-365a-5p could only be detected in normal epithelial culture. www.impactjournals.com/oncotarget  Using the median of individual microRNA expression level as cut-off point, the LSCC cases were allocated into high level and low level groups regarding the corresponding microRNA expression level. Pearson Chi square test was adopted to examine the correlation between the clinical feature and each microRNA expression level. The high miR-744-3p expression level could significantly increase the risk of lymph node metastasis in LSCC.

Reduced PDCD4 and PTEN expression were correlated with the regional nodal status of LSCC patients
In the LSCC tissues, PDCD4 and PTEN expression were significantly reduced ( Figure 3A). Low PDCD4 (P = 0.037) and PTEN (P < 0.001) were significantly associated with the regional lymph node metastasis of LSCC patients. Reduced PTEN expression was also found to be significantly associated with the advanced clinical stage (P = 0.036). Furthermore, MiR-744-3p was negatively correlated with PTEN (R = -0.319, P = 0.029) and PDCD4 (R = -0.574, P < 0.001) in LSCC tissues ( Table 4).

MiR-744-3p targeted PDCD4 and PTEN transcript in LSCC
As shown in Figure 3B, suppressing miR-744-3p in the LSCC cell lines (SNU899 and SNU1076) resulted in significant increase in PTEN and PDCD4 transcript expression (P < 0.05). Target gene prediction analysis revealed that PDCD4 and PTEN 3′-UTR harbored the binding sites for miR-744-3p seed sequence. In the Ago2-IP/IgG-IP, PDCD4 and PTEN mRNA levels were reduced when the endogenous expression of miR-744-3p were suppressed by shRNA ( Figure 3C). The association between miR-744-3p and PDCD4 or PTEN was further validated using luciferase activity reporter containing PDCD4 or PTEN 3′-UTR. As shown in Figure 3D and 3E, there was a significant increase in the luciferase activity when miR-744-3p was suppressed endogenously. In the mutant construct of which miR-744-3p binding sequence on the PDCD4 or PTEN 3′-UTR were disrupted, the luciferase activity remained stable with no significant changes.

PDCD4 restoration reduced the metastatic ability of LSCC
As a famous tumor suppressor gene, PDCD4 was described as inhibitor of cancer cell migration and metastasis in different cancer forms including LSCC. Low PDCD4 expression was observed to be significantly related with poor differentiation and lymph node metastasis in LSCC patients [12,13]. However, the mechanisms underlying its suppression remained to be determined. We continuously monitored the dynamic changes of migration and invasion capacities of LSCC cell lines with PDCD4 restoration. LSCC cell lines (SNU899 and SNU1076) expressing PDCD4 ( Figure 5A and 5B) In total 13 potential target genes of miR-744-3p were retrieved, which were also downregulated in LSCC according to literature review. www.impactjournals.com/oncotarget showed significant reduction in the migration and invasion ability in comparison with the mock controls ( Figure 5C).

PDCD4 restoration reduced AKT and NF-κB p65 activation as well as MMP-9 expression in LSCC
In the PDCD4 overexpressing clones, phosphorylated AKT and phosphorylated NF-κB (p65) levels were significantly reduced, while PDCD4 protein level was significantly increased ( Figure 5E). The phosphorylated IKKα/β and IκB-α level remained unchanged in the LSCC cell lines with high PDCD4 expression ( Figure 5E). Moreover, MMP-9 mRNA and protein level were remarkably reduced in SNU899 and SNU1076 with PDCD4 restoration (Figure 5D and 5E).

DISCUSSION
As there was currently no data on the pathological functions of miR-744-3p, we first examined the tissue expression level and evaluated the statistical association with the clinicopathological characteristics of our LSCC patients. The association between high miR-744-3p expression and regional lymph node positive suggested that miR-744-3p could possibly play a part in controlling LSCC migration and invasion. To establish the causal relationship, we established the miR-744-3p suppressed LSCC cell lines and measured the changes in migration and invasion property of LSCC cells using real-time cellular kinetic assays and animal models. Our results confirmed that the expression level of miR-744-3p could have a functional impact on the migration and invasion propensity of LSCC cells.
MMP-9 expression is regulated by the upstream regulator PDCD4 and PTEN. MMP-9 could shape the development of epithelial-mesenchymal transition (EMT) phenotype in LSCC [35,36]. Further, LSCC patients with low MMP-9 expression had a better 5-year survival rate [37]. In our LSCC samples, both PDCD4 and PTEN were significantly reduced. As miR-744-3p expression level in LSCC tissue was negatively correlated to the PDCD4 and PTEN transcript levels in the same tissue cohort, Using the median expression level of PTEN, PDCD4 and MMP-9 mRNA in LSCC tissues as cut-off value, the patients were divided into high expression level group and low expression level group. Low PTEN expression in LSCC was significantly correlated with lymph node metastasis and late clinical stage. Low PDCD4 expression in LSCC was significantly correlated with lymph node metastasis. High MMP-9 expression in LSCC was significantly correlated with lymph node metastasis. Furthermore, the expression level of PTEN and PDCD4 were negatively correlated with miR-744-3p expression level.  we proposed that miR-744-3p could possibly function as MMP-9 enhancer by inhibiting PDCD4 and PTEN expression in LSCC.
Our data showed that decreased PDCD4 was associated with the nodal status of LSCC patients indicating the role in LSCC migration and invasion. PDCD4 could suppress MMP-9 expression by inactivating AKT/mTOR signaling or reducing NF-кB (p65) activation [38,39]. In the PDCD4 overexpressing LSCC cell lines, we noticed that the signal intensity of both pathways was reduced significantly. NF-кB (p65) was a critical oncogenic transcription factor. Activated NF-кB (p65) was involved in promoting cancer cell survival, inhibiting the apoptotic machinery, promoting angiogenesis and inducing cancer invasion [40]. PDCD4 could inactivate NF-кB (p65) and prevent NF-кB (p65)-mediated transcription activation of MMP-9 [41]. Hence, suppressing PDCD4 could promote cancer migration and invasion. PDCD4 could inactivate nuclear NF-κB (p65) without interfering the activation status of upstream IKKα/β and IκBα in brain tumor [41]. This was because PDCD4 could interact with NF-κB (p65) physically. Our data illustrated that the NF-κB (p65) inhibitory property of PDCD4 was effectively abolished by the high expression of miR-744-3p in LSCC.
The phosphatidylinositol 3 kinase (PI3K) / protein kinase B (PKB, also known as AKT) and mammalian target of rapamycin (mTOR) signaling cascade is the most commonly altered signaling cascade in head and neck carcinoma. PI3K (class I, II, and III) is a family of intracellular lipid and serine/threonine kinases. It functions by phosphorylating the downstream phosphatidylinositols and phosphoinositides at the 3′-hydroxyl group which triggers a cascade of signaling event. Upon receiving external stimulus, PI3K is recruited to the plasma membrane by the activated receptor and catalyzes the formation of phosphatidylinositol 3, 4, 5-bisphosphate (PIP3) from phosphatidylinositol 4, 5-bisphosphate (PIP2). PIP3 will activate the serine/threonine protein kinase Akt (Akt1, Akt2, and Akt3) and activate its activity by phosphorylating threonine 308 (by 3-phosphoinositide dependent kinase) or serine 473 (by rapamycin-insensitive mTOR complex/mTORC2) after binding AKT with its PH domain [18,42,43]. AKT and mTOR control key oncogenic signaling molecule including TP53, mitogenactivated protein kinase (MAPK)/ extracellular signalregulated kinase, and eukaryotic initiation factor 4E (eIF4E) as well as NF-κB (p65) [44]. Phosphorylated AKT activates mTOR complex 1 (mTORC1), which later releases eIF4E by phosphorylation of eIF4E-binding protein (4EBP1). It has been shown that free eIF4E could activate MMP-9 expression [45]. In this signaling cascade, the tyrosine phosphatase PTEN functions as a central negative regulator which antagonize the action of PI3K by converting PIP3 to PIP2 via dephosphorylation [18] and hence reducing the quantity of activated AKT in the cells [46,47]. It had been reported that reduced PTEN expression was associated with the deep invasion and lymph nodes metastasis of LSCC [11]. Low PTEN was observed in precancerous lesions suggesting that loss of PTEN was an early event in the development process of LSCC [27]. Targeting the PI3K/AKT/mTOR signaling pathway in head and neck cancer was suggested to be promising in improving the treatment efficacy and clinical outcome [48,49]. Here, our results suggested that miR-744-3p could possibly control the mTOR signaling cascade by inhibiting PTEN expression in LSCC.
Everolimus is a selective inhibitor of mTOR approved by FDA for use in cancer treatment. As miR-744-3p could simultaneously target dual signaling channels responsible for MMP-9 activation, we hypothesized that suppressing miR-744-3p could have a therapeutic impact on inhibiting MMP-9 expression and should have a better efficacy than everolimus. In comparison with everolimus, MMP-9 expression was significantly reduced in miR-744-3p suppressed LSCC cell lines indicating that targeting miR-744-3p had a therapeutic impact in preventing MMP-9 upregulation in LSCC.

Cell cultures and drug
LSCC cell lines (SNU46, SNU899, and SNU1076) were obtained from Korean Cell Line Bank [50]. Normal epithelial cultures derived from oral cavity epithelia (HOK1 and HOK2) and tonsil epithelia (HTEpiC) were obtained from ScienCell Research Laboratories. The LSCC cell lines were maintained in RPMI 1640 medium supplemented with 10% FBS, 200 units/ml penicillin G sodium, 200 μg/ml streptomycin sulfate and 0.5 μg/ ml amphotericin B. HOK1 and HOK2 were maintained in oral keratinocyte medium (ScienCell Research Laboratories) together with oral keratinocyte growth supplement (ScienCell Research Laboratories). HTEpiC was maintained in tonsil epithelial cell medium (ScienCell Research Laboratories) supplemented with tonsil epithelial growth supplement (ScienCell Research Laboratories). Everolimus were obtained from Sigma-Aldrich.

Tissue samples
Forty-one paired LSCC and the corresponding normal epithelia were obtained from the First Affiliated Hospital, Sun Yat-sen University, Guangdong, China. Six paired LSCC tissues were obtained from Department of Surgery, Queen Mary Hospital, The University of Hong Kong. Consent of tissue donation for research purpose was obtained before tissue collection. Tissues were collected from patients with no prior treatment including radiotherapy and/ or chemotherapy. All the tissues were snap-frozen in liquid nitrogen and were stored at -80℃ until use. The protocol was approved by the human research ethics committee of Sun Yet-sen University (reference number: 2013_23) and University of Hong Kong (reference number: UW 10-142).

Figure 7: Illustration of potential pathways for miR-744-3p mediating metastasis in LSCC.
MiR-744-3p promotes LSCC metastasis by simultaneously inhibiting PTEN and PDCD4. Both of PTEN suppression and PDCD4 suppression can activate AKT and mTOR, and then further promote transcription activation of MMP-9, resulting in enhancement of LSCC metastasis. Moreover, loss of PDCD4 leads to NF-κB (p65) activation, which also promotes MMP-9 expression. Hence, suppression of miR-744-3p can powerfully prevent LSCC metastasis by inhibiting MMP-9 through inactivation of those 2 pathways above. www.impactjournals.com/oncotarget

MicroRNA microarray
RNA quality was evaluated using Agilent 2100 bioanalyser (Agilent Technologies) prior to microarray experiment. All the samples had rRNA Ratio (28s/18s) over 1.8 and RNA Integrity Number (RIN) over 8.0 (Figure 8). The experiments were performed in the Centre for Genomic Sciences, The University of Hong Kong. Data were analyzed by miRNA QC Tool software (Affymetrix). MicroRNA with P < 0.05 and 1.5-fold difference were considered as deregulated microRNA in LSCC.

Quantitative reverse transcription polymerase chain reaction (QPCR)
Total RNA from the cell lines and tissues were extracted with TRIZOL reagent (Invitrogen). For microRNA cDNA preparation, microRNA was polyadenylated by using the Poly (A) Polymerase Tailing Kit (Illumina). The cDNA product was then generated using High-Capacity cDNA reverse transcription kit (Invitrogen) following the methods described by Reichenstein et al. [51]. For conventional cDNA conversion, mRNA was reversely transcribed into cDNA using High Capacity cDNA Reverse Transcription Kit (Invitrogen). Real-time PCR were performed using FastStart Universal Probe Master (Roche Applied Science) on LightCycler ® 480 (Roche Applied Science). Reaction was performed at 95°C for 10 minutes followed by 45 cycles of 95°C for 15 seconds and 60°C for 1 minute. Table 5 listed all the primers and the probes used in this study.

Real-time migration and invasion assays
xCELLigence Real-Time Cell Analyzer (Roche Applied Science) was used to measure the changes in migration and invasion ability of LSCC cells in response to the changes in miR-744-3p or PDCD4 levels. CIM-Plate 16 was adopted for the migration and invasion assay. The dynamic changes of cell number (expressed as cell migration index) from upper chamber (coated with fibronectin) to lower chamber due to the migratory capacitance of LSCC cells were measured continuously over 24 hours. For the measurement of invasion propensity, the upper chamber of CIM-Plate 16 was coated with Matrigel (1:30 dilution; BD Biosciences) for 4 hour at 37°C before cell seeding. Subsequently, the dynamic invasion of LSCC cells through the Matrigel layer was recorded and expressed as cell invasion index.
MiRWalk retrieves predicted target gene lists from 12 databases (DIANA-microTv4.0, DIANA-microT-CDS, miRanda-rel2010, mirBridge, miRDB4.0, miRmap, miRNAMap, PicTar2, PITA, RNA22v2, RNAhybrid2.1 and Targetscan6.2). Mature miR-744-3p sequence, MIMAT0004946 was used in the computational analysis. Genes simultaneously reported by 2 or above databases were selected. In addition, gene list was generated from literature. Genes reported to be downregulated in LSCC were selected and generate the second gene list. The common genes in the 2 gene lists were further filtered using the machine learning regression method mirSVR scoring method, which measures the likelihood of target mRNA down-regulation by the microRNA from sequence and structure features in microRNA/mRNA predicted target sites [53]. A lower mirSVR score implies a greater likelihood of the downregulation of potential target genes due to specific microRNA. Gene with a mirSVR score less than -0.01 was selected for subsequent validation.

Animal models
LSCC cells (5 × 10 5 ) were injected intravenously into the tail veins of athymic nu/nu mice (5-weeks old, weight range: 18-22 g). The animals were euthanized on day 60 after injection and the lung were removed for histological examination. Hematoxylin and Eosin (H&E) staining was performed on the lung sections to evaluate the number of metastatic foci. The protocol for animal study was approved by the Institutional Committee on the Use of Live Animals in Teaching and Research (Protocol number 3675-15).

Statistical analysis
Statistical analyses were performed using the SPSS 18.0 software package (SPSS Inc.). All results were presented as mean ± SE from three or more independent experiments. The difference between primary tumors and their normal counterparts was calculated by Wilcoxon Signed Rank test. Chi-square test and Pearson correlation test was used for evaluation on the clinical significance of microRNAs or genes in LSCC. The difference between control groups and engineered cell lines were analyzed used student-t-test. All the tests were 2-sided. Unless otherwise noted, P value < 0.05 was considered as statistically significant.