Multitargeting activity of miR-24 inhibits long-term melatonin anticancer effects

We have previously shown that melatonin exerts tumor suppressor activities by inducing the p38-p53 axis. This occurred within a few hours while no data are available on how melatonin pathway can be sustained on the long term. Here we show that miR-24, which has been demonstrated to target genes involved in the DNA repair process, targets p38, p53, PML and H2AX simultaneously. We show that long-term treatment with melatonin can decrease miR-24 levels post-transcriptionally, which pairs with a long-wave regulation of genes involved in cell proliferation, DNA damage, RNA metabolism and cell shape and transformation. Moreover, we show that melatonin can inhibit cell proliferation and migration, at least in part, by downregulating miR-24. Furthermore, we propose the involvement of hnRNP A1, which is downregulated by melatonin and involved in miRNA processing, in the regulation of miR-24 levels by melatonin. We conclude showing that miR-24 is upregulated in colon, breast and head and neck datasets and its levels negatively correlate with overall survival.


INTRODUCTION
Melatonin exerts a plethora of activities, among which radical scavenging and activation of tumor suppressor pathways [1][2][3][4][5][6][7][8][9][10][11][12]. These activities explain in part the negative correlation observed between melatonin levels and the risk of developing cancer. In particular, it is worth to mention three epidemiological prospective studies conducted on healthy subjects, two of them on the ORDET cohort [13,14] and the third one on the Nurses' Health cohort [15], which showed a negative correlation between the urinary 6-sulphatoxymelatonin (aMT6s) levels and breast cancer risk.
A few studies have shown that melatonin can activate phosphorylation cascades, mediated by MEK1/2, ERK1/2, JNK and p38 MAPK [10,11,16,17], through binding to its G protein-coupled membrane receptors MT1 and MT2. This leads to the activation of tumor suppressor pathways, such as p53. In particular, melatonin treatment induces a p38-mediated phosphorylation of p53 in Ser15, which caused a transient cell cycle arrest and the accumulation of DNA repair proteins (e.g. H2AX and PML) and the induction of DNA repair mechanisms themselves [6,10]. These phenomena occur only in the presence of intact MT1 and MT2 signalling [11].
These studies have been conducted on short intervals and showed fast phosphorylation kinetics, at least concerning p53. Although it is evident that melatonin activities are sustained over long periods of time (e.g. increase in p53 levels following 72 hours treatment with melatonin), no data are available on their longterm regulation. Being deregulated in many cancers and involved in the carcinogenic process [18][19][20][21], micro-RNAs www.impactjournals.com/oncotarget (also known as miRNAs and miRs) are good candidates for long-term regulation of tumor suppressive pathways, such as p53. They are small non-coding RNAs [22], which regulate gene expression by inducing degradation or inhibiting translation of their numerous target genes [23]. They are transcribed as a long primary RNA (pri-miRNA or pri-miR), which usually contains multiple miRNAs. Before leaving the nucleus, pri-miRs are cleaved by Drosha and DGCR8 [24][25][26][27] into a smaller precursor RNA (pre-miRNA or pre-miR). pre-miR is then exported into the cytoplasm [28][29][30], where Dicer cleaves it into 22 nucleotides long double-stranded miRNAs [31,32]. These are bound by the RNA-induced Silencing Complex (RISC), which catalyzes the separation of the two complementary strands, indicated as mature miR-5p and miR-3p (reviewed in [33]). Both strands are able to regulate gene expression and show different target genes [34]. They can also elicit sometimes completely different outcomes, as in the case of miR-10b and miR-10b*. Indeed, overexpression of miR-10b positively regulates cell migration and invasion in metastatic-and its overexpression initiates robust invasion and metastasis in non-metastatic breast tumors [35], while overexpression of miR-10b* associates with longer relapseand metastasis-free survival [18,36].
A variety of miRNAs regulate the DNA repair process by targeting DNA repair proteins [44], among which miR-24, which regulates histone H2AX expression, thereby causing genomic instability and reducing DNA repair [45]. miR-24 has been found in both pleural effusions and sera from patients with lung cancer, therefore it is now regarded to as a potential diagnostic tumor marker [46,47]. In addition, miR-24 is upregulated in highly differentiated CD8+ T cells upon etoposide treatment and associated with a decreased DNA damage response [48] and in oral squamous cell carcinomas (OSCC) as compared to their matched controls [49], as well as in hepatocellular carcinomas (HCC), its expression being higher in higher grade HCC [50]. Moreover, miR-24 is overexpressed in glioblastomas and promotes cell growth and migration in glioma cells both in vitro and in vivo [51].
In the present manuscript, we show that melatonin downregulated miR-24, which targets four components of the melatonin activated pathway: p38, p53, PML and H2AX. By targeting these genes, miR-24 enhance cell proliferation and induce migration of cancer cells, while melatonin can revert these phenotypes in the presence of physiological levels of miR-24 but not when miR-24 is overexpressed by mimic transfection. We also show that miR-24 downregulation occur through intact MT1/MT2 receptor signalling, as it is abolished in the presence of the melatonin antagonist Luzindole. In addition, we propose the involvement of hnRNP A1 in the processing of miR-24, since miR-24 downregulation by melatonin seems to occur at the post-transcriptional level and melatonin downregulates hnRNP A1 expression through receptor signalling.

Melatonin downregulates miR-24 leading to upregulated expression of its targets
We have previously shown that melatonin activates the p53 pathway with a fast kinetics through stimulation of its membrane receptors, MT1 and MT2 [10,11]. However, it remained unclear how melatonin could sustain the p53 pathway activation on the long term, which would eventually lead to inhibition of cell proliferation and protection from DNA damage accumulation. To clarify this, we performed a search in silico to identify miRNAs predicted to regulate the four components of the melatonin pathway: p53, p38, PML and H2AX. We used the microRNA.org database and identified 60 miRNAs putatively targeting all four genes simultaneously. Among those, only five had validated targets in Tarbase: miR-24, miR-217, miR-296-5p, miR-324-3p and miR-876-3p (Table 1). miR-24 (alias miR-24-3p) raised our interest as it showed the highest number of validated targets, it had been previously reported to target H2AX and inhibit DNA repair [45]. It is also upregulated in a variety of cancers as compared to non-tumoral tissues (Supplementary Figure S1). We also interrogated the expression of predicted miR-24 putative target genes in a RNA-Seq experiment previously performed in our laboratory in vehicle and melatonin treated cells for 48 hours, to look at global changes in mRNA expression following melatonin treatment. We identified 14 upregulated genes following melatonin treatment which were also miR-24-3p putative targets (Table 2). Among the 14 genes we identified H2AX and p38-γ ( Figure 1A). To assess whether miR-24 could be regulated by melatonin, we measured its expression levels after treatment with melatonin of both HCT 116 and MCF-7 cells. We found that miR-24 levels were strongly reduced following 72 hours of melatonin treatment in both cell lines and this appear already evident at 48 hours in MCF-7 cells ( Figure 1B). This prompted us to validate our four genes of interest as targets of miR-24. We analyzed their expression upon miR-24 transfection by Dual Luciferase assays, Real-Time qPCR and Western Blot. Dual Luciferase assays confirmed that all four genes were direct targets of miR-24 ( Figure 1C) and deletion of the seed sequence of miR-24 abolished their regulation by miR-24 ( Figure 1C). Both Real-Time qPCR and Western Blot analyses showed that p53, p38 and PML expression was decreased upon transfection of miR-24 ( Figure 1D, 1F). H2AX was downregulated by miR-24 at the mRNA but not protein level, probably due to the targeting of a specific 3ʹUTR variant of H2AX by miR-24. Conversely, inhibition of miR-24 expression by LNA transfection induced the expression of p38, p53, PML and H2AX both at the mRNA ( Figure 1E) and the protein ( Figure 1G) levels. miR-24 overexpressed and downregulated levels are monitored by qPCR (Supplementary Figure S2A-S2B). Strikingly, p38, p53, PML and H2AX transcripts are up-regulated in response to melatonin thus it is opposite to that of miR-24 upon the identical experimental conditions in both HCT 116 and MCF-7 cell lines ( Figure 2A).

miR-24 causes genomic instability
We have previously reported that melatonin can reduce DNA fragmentation caused by chemotherapeutic administration through activation of the p38-p53 axis [10,11]. Lal et al., have shown that miR-24 causes DNA damage through downregulation of histone H2AX [45]. We reasoned that, p53-and p38-dependent DNA repair induced by melatonin could be mediated by its ability to reduce miR-24 expression levels. To this purpose, we assessed DNA damage accumulation in HCT 116 cell lines stably overexpressing miR-24 (miR-Vec-24) or its negative control (miR-Vec). Our results confirmed that miR-24 could increase DNA damage induced by chemotherapy as assessed by comet assays (Figure 2B-2C). Interestingly, melatonin was able to reduce DNA fragmentation only when miR-24 was expressed at physiological levels ( Figure 2B-2C). This suggests that melatonin's ability to protect cells from DNA damage is exerted, at least in part, through downregulation of miR-24 levels. miR-24 stably overexpressed levels are verified in qPCR analysis (Supplementary Figure S2C).

Melatonin inhibits miR-24-mediated proliferation and migration
Melatonin has been shown to inhibit cell proliferation through activation of p53 [6,8,10,11]. We found that inhibition of miR-24 expression by LNA transfection reduced HCT 116 cells proliferation ( Figure 3A). As miR-24 has been shown to regulate migration and invasion [52], we performed transwell migration assays in the presence or in the absence of melatonin in HCT 116 and HCC1143 cells transfected with either control or mimic-24. Melatonin inhibited cell migration at physiological levels of miR-24, which per se enhanced the percentage of migrating cells, while this ability was lost when miR-24 was overexpressed ( Figure 3B-3C). Similar findings were found by LNAmediated inhibition of miR-24 expression ( Figure 3D-3E). miR-24 downregulated levels are verified in qPCR analysis (Supplementary Figure S2D). To understand whether melatonin and miR-24 regulated the invasive processes through the p38-p53 axis, we performed transwell migration assays in the presence and in the absence of the p38 inhibitor SB202190, which we previously showed to impair melatonin-induced p38-mediated phosphorylation of p53 [10]. Melatonin reduced partially the migration of HCT 116 cells in the presence of SB202190, but the extent of reduction was far lower ( Figure 3F).
These results were corroborated by wound healing assays in HCC1143 cells: melatonin inhibited the wound healing process, while blockage of p38 impaired melatonin's ability to reduce wound closure ( Figure 3G). Both sets of data suggest that, at least, part of melatonin inhibitory activity towards migration is exerted through the p38 signalling.

Melatonin downregulates miR-24 through inhibition of hnRNP A1
We reported that melatonin anticancer activities are exerted mainly through activation of MT1 and MT2 receptors [11]. To ascertain whether downregulation  of miR-24 by melatonin could be exerted through MT1 and MT2, we investigated miR-24 expression levels following melatonin treatment in the absence and in the presence of the melatonin receptors inhibitor Luzindole. Quantitative Real-Time PCR revealed that blockage of melatonin receptors impaired melatonin's ability to reduce miR-24 levels ( Figure 4A). Likewise, inhibition of p38 activity by treatment with SB202190 impaired miR-24 downregulation by melatonin ( Figure 4B), suggesting that miR-24 expression was regulated through the p38-mediated melatonin receptor signalling. miR-24 is transcribed from two different clusters: miR-23b~27b~24-1 and miR-23a~27a~24-2 [53]. To understand whether miR-24 was regulated at either the transcriptional or the post-transcriptional level by melatonin, we examined the expression levels of miR-23a, miR-23b, miR-27a and miR-27b. Interestingly their expression did not seem to vary upon melatonin treatment ( Figure 4C), suggesting that the two clusters were not regulated transcriptionally. To corroborate these results, we amplified both pri-miR-24 and pre-miR-24 by Real-Time qPCR, both in the absence and in the presence of melatonin. Interestingly, while we could not appreciate a variation in the ratio between pre-miR-24-1 and pri-miR-24-1, the ratio between pre-miR-24-2 and pri-miR-24-2 increased upon melatonin treatment ( Figure 4D), this may suggest that in the presence of melatonin pre-miR-24-2 would not be converted to mature miR-24.
Given that melatonin treatment induced the expression of miR-24 target genes (Table 2), which are involved in cell proliferation (WDR18, PRELID1, CRLF1, TMEM115), cell shape and transformation (TAGLN, REEP6), RNA metabolism (RPS23, RPL10) and DNA damage response (RFC2, MAPK12, H2AFX), we reasoned that miR-24 downregulation could be a major event in melatonin anticancer activities. Therefore, we searched for a possible regulator of miR-24 maturation in a proteomic screening previously performed in our laboratory, to look at global changes in protein expression following melatonin treatment (data not shown). We found that hnRNP A1, a protein involved in both mRNA splicing and miRNA maturation [42,43,[54][55][56], was downregulated upon melatonin treatment (data not shown). We confirmed this finding by Western Blot ( Figure 5A). Of notice, hnRNP A1 mRNA levels were not affected by melatonin treatment (Supplementary Figure S3), suggesting a post-transcriptional regulation by melatonin.
To confirm hnRNP A1 involvement in miR-24 maturation, we overexpressed hnRNP A1 and treated the cells with melatonin. While melatonin led to a decrease in miR-24 at physiological levels of hnRNP A1, we could not observe a decrease in miR-24 when hnRNP A1 was overexpressed ( Figure 5B), suggesting that melatonininduced miR-24 downregulation was exerted through inhibition of hnRNP A1 expression. We then sought to analyze whether the decrease in hnRNP A1 levels occurred through activation of melatonin membrane receptors. To this purpose, we knocked down the expression of MT1 and MT2 by siRNA and investigated hnRNP A1 protein levels. Results showed that removal of either receptors led to an increase of hnRNP A1 protein levels ( Figure 5C).

miR-24 and hnRNP A1 are deregulated in human cancers
We investigated miR-24 expression levels in both colon and breast cancer datasets. For colon cancer we used the Pan-Cancer colorectal adenocarcinoma dataset [57,58], comprising 314 cases and 11 controls, while for breast cancer we interrogated both the METABRIC [59] and the TCGA [60] databases, comprising 1284 cases (208 basal-like) and 116 controls, and 694 cases (90 basal-like) and 83 controls, respectively. We found that miR-24 levels were far higher in tumor vs normal samples in both datasets for colorectal carcinoma ( Figure 6A). We observed a statistically significant upregulation of miR-24 in the basal-like breast cancer subtype as compared to normal samples ( Figure 6B) for breast cancer. Kaplan-Meier analyses in the METABRIC dataset showed that high miR-24 levels correlated with a lower disease specific survival rate ( Figure 6C). To verify that miR-24 acted as an oncogene in diverse cancer types, we analyzed its expression from our head and neck squamous cell carcinomas dataset [61]. We found that, also in this dataset,   Histograms show the ratio between pre-miR-24-1 and pri-miR-24-1 and the ratio between pre-miR-24-2 and pri-miR-24-2 levels. *p < 0,01. miR-24 was upregulated in tumors as compared to nontumoral matched samples ( Figure 6D) and that high levels of miR-24 correlated with a lower survival ( Figure 6E). These data, together with the proliferation and migration data shown in Figure 3, corroborate our and others finding that miR-24 acts as an oncogenic microRNA [52,62].

DISCUSSION
Epidemiological studies have shown that melatonin levels negatively correlate with the risk of developing various kinds of cancers [13][14][15]. We have previously shown that melatonin can induce the p38-p53 axis through membrane receptor activation [10,11]. However, while our previous data focused on a fast kinetics of induction of the p38-p53 axis occurring within 3 hours of treatment (Figure 7, green arrows), others have observed an upregulation of p53 at later time points [3,6,8] but there is no evidence of how melatonin can exert its longterm oncostatic effects. A possible mechanism could be miRNA regulation. However, at present there is only one report about miRNAs regulation by melatonin, in which the authors identified a set of miRNAs differentially modulated in untreated and treated MCF-7 cells [70] but neither the mechanisms by which melatonin can influence miRNA expression nor the mechanisms by which miRNAs can regulate melatonin's anticancer activities are known. Our work identifies the oncogenic miR-24 as a miRNA strongly repressed by melatonin 72 hours post treatment (Figure 7, red arrow), which has p38, p53, PML and H2AX among its target genes (Figure 1). In addition, we identified 14 putative targets of miR-24 after melatonin treatment (Table 2), among which H2AX and p38-γ. Of notice, H2AX had the highest mirSVR score (Supplementary Table S1), which could account  for the fact that it was upregulated already 48 hours post treatment, when we observed a mild downregulation of miR-24 ( Figure 1B). The decrease in miR-24 expression following melatonin treatment translated into activation of DNA repair and inhibition of cell proliferation and migration, due in part to the upregulation of the four key components of the melatonin receptor activated pathway (Figures 2 and 3). Thus, upon melatonin treatment, p38 induced a mechanism of self-sustainment through downregulation of miR-24 ( Figure 4B). Our data on the melatonin-induced p38-dependent reduction of hnRNP A1 protein are supported by previous observations reporting that p38 can phosphorylate hnRNP A1 [71,72], thereby changing its localization from the nucleus to stress granules in the cytoplasm, as well as reducing its protein levels [73]. Previous reports have also shown that hnRNP A1 can bind to miRNA precursors and regulate their cleavage acting as a miRNA chaperone [42,43]. We suggest that hnRNP A1 mediates miR-24 processing in the absence of melatonin, as its ectopic expression abrogates miR-24 downregulation. We reported previously that melatonin membrane receptors MT1 and MT2 could be involved in the carcinogenic process, since their expression is lower in tumors as compared to normal samples [11]. Conversely, here we show that hnRNP A1 is upregulated in tumor vs normal samples in the same colon cancer casuistries where we observed repression of MT1 and MT2 and in an additional casuistry as well (Supplementary Figure S4A). This occurs also in four breast cancer casuistries deposited in Oncomine (Supplementary Figure S4B). Despite we show that miR-24 targets four crucial proteins involved in melatonin anticancer effects, this does not rule out the possibility that other target proteins as well as other miRs will play a major role in such activities.
miR-24 is regarded to as an oncogenic miRNA in various types of cancer because not only it targets DNA repair genes, such as H2AX [44,45] but it has also been found upregulated in OSCC [49] (Figure 6D), hepatocellular carcinoma [50], glioblastoma [51] and breast cancer [74]. In addition, it is involved in the carcinogenic process by inducing transformation and migration [50,51]. Our data provide evidence for both miR-24 oncogenic role in breast and colon cancer and melatonin inhibition of an oncogenic pathway. First, we provide evidence that miR-24 contributes to transformation of breast and colon cancer cell lines, as it conferred a higher proliferative and migratory potential to HCT 116 and HCC1143 cells and that its ectopic expression could overcome melatonin repression of these oncogenic features (Figure 3). Then, we show that miR-24 is upregulated in colon cancer, as well as in breast cancer. Long-term activation of Melatonin receptors sustains p38 activation and leads to inhibition of hnRNP A1 (red arrow) thereby causing a decrease in miR-24, which results in p53, PML, H2AX and p38 activity reduction. Through downregulation of hnRNP A1 and miR-24, melatonin impairs the migrating capability of cancer cells.
In particular, its expression levels are higher in colorectal adenocarcinoma as compared to normal samples in the Pan-Cancer casuistry, and in the basal-like subtype of breast cancer in the METABRIC and TCGA datasets ( Figure 6A-6B). Moreover, we evaluated the association between expression levels of miR-24 and recurrence-free survival by Kaplan-Meier analysis and found a strong negative correlation for breast cancer ( Figure 6C). As for breast cancer, we found that miR-24 was overexpressed in head and neck tumors and higher levels correlated with a worse prognosis (e.g. worse disease-free survival) ( Figure 6D-6E). This suggests that miR-24 deregulation could be a carcinogenic event common to various types of cancer. Recently, miR-24 has been identified as a circulating miRNA in lung cancer patients. It has been found, together with miR-30d, as a cell-free miRNA in both malignant and benign pleural effusions, with higher levels in the malignant ones [47], as well as in preoperation sera from lung cancer patients [46]. In the light of these observations, miR-24 could be used as a potential cancer biomarker.
These data strengthen the role of melatonin as a chemopreventive agent. In fact, we can speculate that not only melatonin could induce DNA repair ( Figure 2B-2C) [75], thereby restraining the side effects of radio-and chemo-therapy on normal cells, but it could per se reduce migration and invasion, through downregulation of miR-24, thus ameliorating the prognosis of invasive cancers. Obviously, this speculation needs to be validated by assessing the impact of melatonin treatment on the DNA repair of untransformed cells. The use of melatonin as an adjuvant for both chemo-and radio-therapy could therefore be envisioned in patients with metastatic cancers and could be used as a therapeutic approach in tumors harboring p53 wild-type. In aggregate, it emerges that melatonin treatment instigates the activation of a tumor suppressor network, which finely tunes transcriptional events and non-coding mRNAs activity. The outputs of such network might depend from both the type and the stage of a given tumor.

Western blots
Protein extracts were prepared as previously described [11]. All protein extracts were quantified by Bradford assay and equal amounts were loaded onto SDS-PAGE, transferred to polyvinylidene fluoride membranes (PVDF, Immobilon-P, Merck-Millipore, Life Science, Darmstadt, Germany) and subjected to immunoblot with the indicated antibodies. Antibodies

Comet assays
Following transfection with the indicated constructs, cells were treated with melatonin for 2 hours and then subjected to DNA damaging treatments with CDDP (850 nM, TEVA, Italia, Italy), Doxorubicin (10 nM, Ebewe, Pharma, Austria) and 5-Fluorouracile (1,5 µM, Hospira, Illinois, US), or Ultra Violet B (UVB) irradiated at a dose of 0,05 J/cm2 using a Bio-Sun irradiation apparatus (Vilbert Lourmat, Marne-la-Vallée, France) and allowed to repair DNA for 4 hours. After treatment, cells were detached with trypsin and subjected to comet assay as described previously [10,11]. DNA was stained with propidium iodide (Sigma-Aldrich, Saint Louis, Missouri, US) and pictures were taken using 60x magnification on an Axiovert 200 M microscope and Axiovision acquisition program (Zeiss, Oberkochen, Germania). At least 300 cells were scored for each slide.

Transwell migration assays
Transfected cells were detached and counted. 5 × 10 4 cells, in a volume of 100 µl DMEM containing 0,1% FBS, were seeded in the upper chamber while the bottom chamber of the transwell was filled with 600 µl of DMEM with 10% FBS. Either melatonin or ethanol was added to the upper chamber. Cells were allowed to migrate for 20 hours. Then cells remaining in the upper chamber were scrubbed away with a cotton pad and cells remaining on the bottom layer of the upper chamber were subjected to DAPI staining as follows: the upper chamber was washed twice with PBS and then cells were fixed with 4% formaldehyde for 20 minutes, permeabilized with 0,2% Triton for 25 minutes and, after being washed with PBS, they were stained with 5 µM DAPI for 5 minutes. Membranes were cut and mounted on microscopy slides. All the cells on each membrane were counted. Each sample was assayed in triplicate. Graphs show percentage of cells relative to control. Student t-test has been used to calculate p-values.
miRNA were reverse transcribed from 10 ng RNA using the TaqMan ® microRNA Reverse transcription kit (4366596, Life Technologies, Carlsbad, California, US). 1 µl of the reaction was used for Real-Time PCR. Amplification was carried out using specific miRNA probes and the TaqMan

RNA-Seq
Total RNA from melatonin or vehicle treated HCT 116 cells was isolated using miRNeasy including the optional DNase treatment (Qiagen, Valencia, CA, US). RNA quality was verified using an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA, US; RNA 6000 Nano kit). All RNAs used for subsequent library preparation had an RNA integrity number greater than 9.0. RNA libraries for sequencing were generated according to the Epicenter ScriptSeq v2 RNA-Seq library preparation kit with an initial ribosomal depletion step using Ribo-Zero Magnetic Gold. Starting material was 5 µg of total RNA. Based on qPCR quantification, libraries were normalized to 1 nM and denatured by using 0.1 N NaOH. Cluster amplification of denatured templates was carried out according to manufacturer protocol (Illumina, Inc., San Diego, CA, US). Sequencing was performed on a Genome Analyzer IIx (Illumina, Inc., San Diego, CA, US) in paired-end mode, sequencing 2 × 75 bp. For each sample generated by the Illumina platform, a pre-process step for quality control has been performed to assess sequence data quality and to discard low quality reads.
For the analysis we exploited the RNA-Seq analysis workflow RAP [76] , that comprises of read mapping, transcript assembly and abundancy estimation followed by transcript-based differential expression via the Tuxedo suite [77]. Paired-end reads were mapped to the human genome assembly hg19 with TopHat and further analyzed by the Cufflinks-Cuffdiff pipeline to identify differentially expressed transcripts. We run the pipeline without novel transcript discovery, correcting for multiply mapping reads with the "-u" option of Cuffdiff and a mask file for rRNAs and tRNAs. The RNA-Seq was conducted with three biological replicates.
Putative miR-24-3p targets were identified using the predictive target search of miRWalk [78] with the programs DIANAmT, miRanda, miRDB, miRWalk, RNA22 and Targetscan, taking only those genes in consideration that were predicted to be miR-24-3p targets by at least 3 different programs. Differentially expressed genes and transcripts that are putative miR-24-3p targets are reported in Table 2, along with p-value and q-value as reported by Cuffdiff.

Statistical analyses
No statistical method was used to predetermine sample size. Experiments were not randomized. For biochemical experiments we performed the experiments at least three independent times. Experiments for which we showed representative images were performed successfully at least 3 independent times. No samples were excluded from the analysis. The investigators were not blinded to allocation during experiments and outcome assessment. All p-values were determined using two-tailed t-tests and statistical significance was set at p = 0.05. The variance was similar between groups that we compared. For head and neck patients' samples, the ethical committee of the "Regina Elena" National Cancer Institute (protocol CE/379/08) approved the study and informed consent was obtained.