MicroRNA expression profile in head and neck cancer: HOX-cluster embedded microRNA-196a and microRNA-10b dysregulation implicated in cell proliferation

Fifteen patients with OSCC (tongue and floor of the mouth) were selected for the microarray experiments. In order to validate the microarray results, 35 additional patients with HNSCC (oral cavity, oropharynx and larynx) were selected. The clinical and pathological profile of patients is shown in Table 1. The average age of patients was 55.5 years (SD 9.8, range 38–82 years), and the male/female ratio was 24:1. Most patients were smokers or former smokers and had a history of chronic alcohol abuse. Tumor and corresponding cancer free surgical margins containing the corresponding epithelium were collected from patients submitted to surgical resection of primary tumor at Hospital das Clinicas, Hospital Heliopolis and Arnaldo Vieira de Carvalho Cancer Institute, in Sao Paulo, Brazil. All patients provided written informed consent, and the research protocol was approved by review boards of all institutions involved and by the National Committee of Ethics in Research (CONEP 1763/05). Samples corresponding to the oral cavity, base of the tongue and larynx were snap-frozen in liquid nitrogen immediately after surgery and stored in liquid nitrogen until RNA preparation. Frozen samples were sectioned using a cryostat, and tissue sections were stained with RNAse-free reagents. Analysis of hematoxylin and eosin-stained sections by the study pathologists confirmed >75% tumor cells in all HNSCC samples and that surgical margins were tumor-free. The diagnosis of HNSCC was confirmed, and tumors were histologically examined for perineural invasion (tumor cells in the perineural space or epineurium), tumor differentiation (well, moderated or poorly differentiated, according to the WHO guidelines), lymphatic-vascular invasion, surgical margins, and lymph node metastasis. Tumors were staged according to the TNM clinical staging system, as proposed by the International Union Against Cancer.

Total RNA was prepared from tissue using mirVana miRNA Isolation Kit (Ambion, Austin, TX) in compliance with the manufacturer’s protocol. RNA integrity and concentration were assessed using the RNA 6000 Nano Assay kit with Agilent 2100 Bioanalyzer according to the manufacturer’s instructions (Agilent Technologies, Palo Alto, CA).

MiRNA expression profiling was performed using the Illumina miRNA arrays version 1.0. Sample preparation and hybridization followed the manufacturer’s instructions. Briefly, 200 ng of total RNA was first polyadenylated and converted to biotinylated cDNA, which was attached to a solid phase and hybridised with a pool of miRNA-specific oligonucleotides (MSO). Each single MSO is used to assay one miRNA on the panel. Universal PCR amplification was then performed, creating fluorescently labeled products identifiable by their unique MSO sequence. These products were hybridized on the Illumina miRNA array, with the address sequence from each MSO enabling the hybridization of specific miRNA products to specific locations on the BeadArray substrate. Hybridization signals were detected and quantified using Illumina scanner and BeadStudio version 3.2.7. Average signals were quantile normalized and then filtered according their detection p-value: only miRNAs for which the detection p-value was consistently equal or lower than 0.05 across at least one group of samples (marginal and/or tumor samples) were considered expressed and further analyzed. All data is MIAME compliant and the raw data has been deposited in a MIAME compliant database (Gene Expression Omnibus database) under accession Number GSE31277. Differentially expressed miRNAs were determined by using the Rank Product, non-parametric statistical method based on ranks of fold-changes [19]. We used the RankProd package on R for this analysis; the percentage of false positives was calculated and p-values were accordingly corrected for multiple comparisons.

To validate the microarray expression data, miRNAs were subjected to quantitative Real Time-PCR using the TaqMan miRNA assay system (Applied Biosystems, Foster City, CA). Briefly, about 100 ng of total RNA was subjected to a reverse transcription reaction using miRNA-specific looped primers according to the manufacturer’s protocol to obtain the cDNA. Subsequent PCRs used miRNA specific forward and reverse primers along with appropriate cDNA product and TaqMan universal mix. PCR was carried out in AB7500 (Applied Biosystems, Foster City, CA) in a 20 ul volume reaction following thermal cycling parameters suggested by the manufacturer: 50°C for 2 min, 95°C for 10 min and 45 cycles of 95°C for 15 s and 60°C for 1 min.

The expression data was normalized to the RNU48 expression. RNU48 was chosen as a suitable endogenous control gene following analysis of gene expression stability of three candidate genes across our samples. For a detailed description of this step refer to the next Methods’ section. Expression levels were determined using the comparative ∆Ct method [20].

For miRNAs individually studied in independent sets of samples by quantitative real-time PCR, the nonparametric test Wilcoxon Signed Rank Test was used to detect the statistically significant differences between paired normal tissue (N) and tumor (T) samples obtained from the same individual. This test was performed using SPSS for Windows® Software. The same software was used to calculate the mean and standard deviation of all variables.

The expression of three snoRNAs (RNU6B, RNU44 and RNU48) was measured by quantitative real-time PCR with TaqMan miRNA assays (Applied Biosystems, Foster City, CA), as previously described for all samples assayed by miRNA microarrays. This data was analyzed using the SLqPCR package in R [21] to determine the expression stability of these snoRNAs across samples. The stability factor M was calculated for each snoRNA (M (for RNU48) = 0.69; M (for RNU44) = 0.78; M (for RNU6B) = 0.75). Since high expression stability is associated to low M values, RNU48 appeared to be the snoRNA with most stable expression across the set of samples analyzed, hence was chosen as control for normalisation.

Potential miRNA targets were identified using Ingenuity Pathway Analysis (IPA Ingenuity Systems, http://​www.​ingenuity.​com). Only experimentally validated targets were selected, using miRecords (http://​mirecords.​biolead.​org/​), Tarbase (http://​microrna.​gr/​tarbase) or TargetScan (http://​www.​targetscan.​org/​). For fuctional annotation of potential targets we used KEGG pathways term enrichment analysis using the computational tool Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.7 (http://​david.​abcc.​ncifcrf.​gov/​home.​jsp).

The HNSCC cell lines SCC25 and SCC9, derived from a SCC of the tongue, and FaDu, derived from a SCC of the hypopharynx were used in this study. They were obtained from American Type Culture Collection (SCC25 catalog number CRL-1628, SCC9 catalog number CRL-1629, and FaDu catalog number HTB-43). The cell lines were grown in a Dulbecco’s Modified Eagle’s medium/Nutrient Mixture F-12 Ham (DMEM/F12) supplemented with 10% fetal bovine serum in a humidified atmosphere of 5% CO² and 95% air at 37°C. Oral keratinocytes were obtained from primary cultures of the buccal mucosa, from voluntary donor patients undergoing surgery performed in out-patient clinics in the Dentistry School of USP. The patients were informed and signed the required Informed Consent. This study was approved by the Research Ethics Committee of the Instituto de Pesquisas Energéticas e Nucleares (IPEN/CNEN-SP) [Institute of Energy and Nuclear Research] (approval number 087/CEP-IPEN/SP). Keratinocytes were plated on a support layer, called feeder-layer, composed of murine fibroblasts of the type 3T3-Swiss albino (ATCC, catalog number CCL-92), which were irradiated (60 Gy), and maintained in an incubator at 37°C, in a humidified atmosphere containing 5% CO₂ and grown as previously described [22].

Normal keratinocytes transfected with the miRNA precursor and the negative control were cultured in Lab-Tek Chamber Slides (Nalge Nunc International, Rochester, NY, USA) for the immunofluorescence assay. Cells were fixed with methanol, blocked with 3% bovine serum in PBS, and incubated for 1 h with antihuman Ki67 (Monoclonal Mouse, clone MIB-1, DAKO, Denmark A/5), diluted 1:400. Cells were washed with PBS and incubated at room temperature for 45 minutes with secondary antibody conjugated with fluorescein (1:50) (Novocastra Laboratories, UK), in a dark chamber. Following washing, chambers containing the cells were mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories, Ind. Buelingame, CA 94010). Results were analyzed by fluorescence microscopy (Zeiss Axiophot II, Carl Zeiss, Oberköchen, Germany). The percentage of cells displaying Ki67 labeling was determined by counting the number of positive Ki67 stained cells as a proportion of the total number of cells counted. Cells were counted manually in the whole chamber area.

Cell lines SCC9, SCC25 and FaDu were stained with Cell Trace Violet (Molecular Probes®), according to the manufacturer protocol. Briefly, the cells were incubated with 5 μM Cell Trace Violet for 20 minutes at 37°C, washed twice with fresh and warmed medium and cultured under regular conditions. The cells were run on BD LSR Fortessa flow cytometer with 405 nm laser at day zero and after 72 hours of cell culture for cell proliferation rate assessment. Proliferation rate was determined by fluorescence decay. Analysis was performed using Flow Jo software (Tree Star™). For cell proliferation rates after transfection, cell lines SCC25 and FaDu were stained 24 h after transfection (at the time of medium exchange). Proliferation rates were compared between scramble (negative control) and cells overexpressing miR-10b.

Following the transfection assays, the global gene expression analysis was conducted using the Agilent Human Whole Genome Oligonucleotide Microarray (44K; Agilent Technologies) following the manufacturer’s protocols. Oligonucleotide microarrays were scanned using the GenePix 4000B Microarray Scanner (Molecular Devices) and features were automatically extracted and analyzed for quality control using Agilent Feature Extraction Software. Raw data was deposited in a MIAME compliant database (Gene Expression Omnibus database) under the accession Number GSE31277. Partek Genomics Suite 6.6 (Partek Incorporated) was used for normalization of gene expression levels and for fold-change in gene expression calculation. To gain insights into the potential mechanisms affected by the overexpression of the miR-10b and miR-196a in cells, deregulated genes were mapped to regulatory networks using Ingenuity Pathway Analysis (IPA Ingenuity Systems, http://​www.​ingenuity.​com).

HNSCC can involve multiple anatomical sites, each with individual molecular characteristics, and highly affected by the drinking and smoking habits of patients [13, 23, 24]. In an attempt to limit data variability due to HNSCC subsites and environmental factors, we assessed miRNA expression levels in 15 OSCC samples limited to tongue and floor of the mouth, from patients possessing similar demographic and clinico-pathological characteristics (Table 1, detailed in Methods). Samples were paired with tumor-free surgical margins. The expression profiles of tumor samples revealed significant differential expression for 72 miRNAs compared to their corresponding tumor-free margins (Table 2). Several studies have analysed the miRNA expression profile of OSCC cell lines and tumor samples, with little overlap among results [25, 26]. This inconsistency in results justifies additional studies.

In order to access biological processes possibly targeted by deregulated miRNAs we performed a functional analysis of validated targets through KEGG term enrichment analysis using the computational tool DAVID. Thirty-eight of the 72 deregulated miRNAs possessed mRNA targets that have been experimentally observed; in total 609 genes are potentially regulated (Additional file 1). These genes were mapped to KEGG pathways and were shown to be broadly involved in cancer development (Additional file 2).

Specifically, members of the miR-17-92 cluster were deregulated in our dataset: miR-19a and miR-19b were strongly up-regulated, in addition to moderate up-regulation of miR-17-3p/miR-17-5p and miR-92b. These results are in line with the observation that the miR-17-92 cluster is up-regulated in many cancer types, including lung cancer and lymphoma [27, 28]. Accordingly, miR-17-92 cluster members have been shown to take part in feedback loops determining the role of c-MYC as tumor suppressor and/or oncogene [29, 30]. Specifically, c-MYC apparently possesses a tumorigenic role in HNSCC, constituting a current candidate for anticancer strategies [31]. Recently, the miR-17-92 cluster has been also shown to regulate multiple components of the TGF-β pathway in neuroblastoma [32]. Other cancer-related miRNAs up-regulated in our OSCC samples are members of the miR-34 family: miR-34b and miR-34c. To our knowledge this is the first report of their altered expression profile in HNSCC, although the deregulation of miR-34a has been recently addressed in HNSCC [33]. These results are interesting in light of the finding that miR-34 is a direct target of p53, functioning downstream of the p53 pathway as a tumor suppressor [34, 35]. Similar to other types of cancer, inactivation of p53 is an extremely common event in head and neck cancers, with mutant p53 status found in nearly 50% of the cases and commonly associated with poor prognosis [36]. However, the role of miR-34b/c in the context of p53 regulation has not been addressed in HNSCC.

In agreement with most miRNA profiles in HNSCC samples and tumor cell lines, miR-133a was also down-regulated in our cancer set as compared to tumor-free samples. Its tumor suppressor activity, for instance by controlling the target genes actin-related protein 2/3 complex subunit 5 (ARPC5) and moesin (MSN), has been already demonstrated in squamous cell carcinoma of the tongue [37‐39]. Since this seems to be a robust characteristic in HNSCC, its function should be further investigated as well as its possible use as a biomarker for early cancer detection.

MiR-196a/b was over-expressed and miR-10b was down-regulated in the OSCC samples compared with tumor-free surgical margins (Table 2). Both miRNAs are dysregulated in a variety of cancers [40, 41], but have not been previously associated with OSCC. We validated our microarray results in an additional subset of OSCC samples as well as in samples belonging to other HNSCC subsites (Figure 1, and Table 1 for sample characteristics). Both miRNAs clearly presented differential expression between tumor and tumor-free samples, suggesting a role in HNSCC.

MiR-196a/b and miR-10b are embedded within homeobox (HOX) clusters of developmental regulators [42]. Schimanski and collaborators [43] demonstrated that HOX genes are targeted by miR-196, and HOX transcripts were also experimentally validated as miR-10 targets [44, 45]. Given that molecular events involved in carcinogenesis interfere in the regulation of cell identity, it has been proposed that HOX proteins could be oncogenic regulators [46]. HOX genes have not been implicated in the development of HNSCC, as judged from reviewing the available literature, including HNSCC gene expression profiles. This suggests that the homeobox-cluster embedded miRNA could have a different role in HNSCC. Thus, we performed gain-of-function experiments aiming to outline a possible role for these molecules.

Precursor molecules of miR-10b were transfected into SCC25 and SCC9 (tongue squamous cell carcinoma-derived cell lines) and FaDu (a cell line derived from hypopharyngeal squamous cell carcinoma), while miR-196a precursor molecules were transfected into human keratinocytes derived from normal oral epithelium. We chose SCC cell lines and oral keratinocytes as models for the investigation of miRNA function in a HNSCC genetic background, emulating cancer and tumor-free cellular systems, respectively.

Two SCC cell lines were initially chosen for the gain-of-function experiments due to differences in their proliferation rates, as reported in Figure 2 and Table 3, a characteristic that could affect results.

As expected, in untreated SCC cell lines, miR-196 was up-regulated and miR-10b was downregulated when expression levels were compared to untreated keratinocytes (data not shown).

After transfection, we assessed the over-expresssion of the respective mature miRNAs in each cell line (Figure 3). Despite the successful overexpression of miR-10b in SCC9, these cells were very sensitive to the transfection, with intense effects in cell proliferation seen also for the negative control. Thus this cell line was not used in subsequent experiments.

Since tumor cells evade programmed cell death and sustain proliferative status [47], we tested whether miR-10b and/or miR-196a could play a role in this scenario. Assessment of Ki-67 antigen expression, a cell proliferation marker, revealed that keratinocytes over-expressing miR-196a were mostly quiescent, as defined by a lack of Ki-67 antigen expression. When compared to transfection controls, cell proliferation was reduced approximately 5 fold in keratinocytes over-expressing miR-196a (Figure 4).

Despite inhibiting cell proliferation, specific occurrences at the cell surface level, such as surface blebbing, considered as a marker for apoptosis, were absent at 72 h after transfection (data not shown). Changes in nuclear morphology indicating late apoptosis were also absent.

Inhibition of cell proliferation upon over-expression of miR-10b in SCC25 and FaDu was assessed by flow cytometry (Figure 5). This is consistent with the miRNA expression data: miR-10b was detected in low levels in HNSCC in this study, a downregulation that would thwart the inhibitory effect of miR-10b on cell proliferation. The effect of the overexpression of miR-10b was clearly more intense in FaDu as compared to SCC25.

Given the fact that miR-196a was found to be overexpressed in HNSCC samples, the inhibitory effect of miR-196a overexpression on proliferation of normal keratinocyes cannot easily put into context. One hypothesis could be that miR-196a overexpression in HNSCC could also be the consequence of uncontrolled proliferation, as a means to counteract it, rather than the cause of its perturbation.

Efforts to understand the global effects of miR-196a, which might be cell-type dependent, are essential considering that it has been recently addressed as a potential therapeutic target [41].

The identification of miRNA target genes is critical in order to understand their roles. However, this task is challenging. MiRNAs are usually imperfectly complementary to the 3′UTR region of their mRNA targets in mammalian cells, with target effect hardly detected at the gene expression level. Additionally, the cellular environment is key in determining miRNA functions, which will vary depending on the cell type.

While keeping these drawbacks in mind, global gene expression profiling using DNA microarrays was performed in this study to identify deregulated cellular processes upon the overexpression of miR-10b and miR-196a in each cell model. Broad consequences of the overexpression should be detected by global gene expression profiling, even when the deregulation of specific target genes might not be detected by this kind of experiment.

Experimentally validated mRNA targets were searched in Tarbase and miRecord databases. None of the miR-10b targets HOXA1, HOXD10 and KLF4[40, 48, 49] were affected at the mRNA level by the overexpression of miR-10b in SCC25 or FaDu (Additional files 3 and 4, respectively). The same was true for the miR-196a gene targets ANXA1, HOXA7, HOXB8, HOXC8, HOXD8, KRT5 and S100A9[42, 50, 51] (Additional file 5). These results suggest that, at least at the mRNA levels these genes are not targeted by miR-196a in the cells used here.

Among the above mentioned gene targets, only ANXA1 down-regulation has been previously reported in HNSCC [52, 53]. For this reason, we checked for alterations of this target also at the protein level. Our results demostrate that ANXA1 is not targeted by miR-196a under the conditions studied here (Figure 6).

We performed a functional analysis of deregulated genes aiming to pinpoint alterations that could explain impaired proliferation. A total of 353 annotated genes were downregulated (at 1.5 fold-change) following miR-196a over-expression in keratinocytes (Additional file 5). The relationships among these genes were assessed using Ingenuity Pathway Analysis (IPA), while considering only experimentally proven connections between genes or proteins. The most significant interaction network consisted of genes associated with DNA replication, recombination and repair, cell cycle and, consequently, cancer. Figure 7 depicts this network and genes involved in cell cycle arrest are highlighted. This network contains 8 deregulated genes from our dataset: CDK2, SYNM (DMN), TP73, AKT1, NFATC4, HOXA9, HSPB3 and CD40LG. Of particular interest is the downregulation of CDK2 and AKT1 and the upregulation of TP73. CDK2 is a subunit of the cyclin-dependent protein kinase complex, expressed in G1-S phase, and essential for cell cycle G1/S phase transition. TP73, up-regulated in cells overexpressing miR-196a, transcriptionally activates target genes leading to apoptosis and growth arrest [54]. The activation of PI3K/AKT pathway in HNSCC is well known; the pathway regulates cell proliferation and has been addressed as a therapeutical target. Thus, the expression patterns of these three genes, following over-expression of miR-196a, would be in agreement with the observed arrest of the cell cycle. However, none of them are direct targets of this miRNA and further studies are needed in order to comprehend the observed effect.

Overexpression of miR-10b in SCC25 and in FaDu provided relatively similar results. Two hundred and ten annotated genes were downregulated and 169 were upregulated when SCC25 cells overexpressing miR-10b were compared to controls (Additional file 3) while 161 genes were downregulated and 169 upregulated in FaDu overexpressing miR-10b (Additional file 4), when at least a 2-fold difference was considered. Sixteen common genes were downregulated in both cell lines, but none of these genes were miR-10b predicted targets.

Regulatory networks provided by IPA did not contain a significant number of genes directly implicated in cell proliferation or cell cycle arrest for SCC25 cell line. This analysis, however, highlighted enrichment of terms belonging to the G-protein-coupled-receptor signaling pathway - with 9 molecules regulated in our dataset (ADRB3, AVPR2 and GRM2, upregulated 2-fold in SCC25 overexpressing miR-10b; and CNR1 (16-fold), DRD3 (4-fold), HCAR2 (2-fold) and OPRK1 (14-fold), downregulated in these cells). A recent review addresses mechanisms by which G-protein-coupled-receptors participate in the regulation of cell cycle [55] and, in the context of HNSCC, G-protein-coupled-receptors have been associated with EGFR signaling and cell survival [56].

A significant regulatory network built with deregulated genes upon overexpression of miR-10b in FaDu includes genes involved in the regulation of cell cycle progression and arrest (Figure 8). Although none of these genes have been implicated in HNSCC or heavily studied in the context of cancer, it is noteworthy the fact that they relate to cell cycle regulation through key players in HNSCC: TP53, NOTCH1, MYC and HRAS [57].

From this analysis it became clear that the effect of the overexpression of miR-10b in SCC25 and FaDu, and miR-196a in keratinocytes do not act upon a large number of cellular processes but may rather target a small set of genes, some of which directly or indirectly involved in the progression of cell cycle.