A total of 20 tumor tissue samples were obtained from OSCC patients who underwent radical tumor resection at the Department of Oral and Maxicallifacial Surgery, the Affiliated Stomatology Hospital of Sun Yat-sen University (Guangzhou, China) and were used in the experiments. The study was approved by the ethics committee of the Institute of Stomatological Research at the university. Of the 20 tumor tissue samples, 8 matched samples of tumor and adjacent nonmalignant tissue were obtained from 4 OSCC patients, as well as 12 samples with/without metastasis. Saliva samples (1 ml) were obtained from each patient prior to surgery and mixed with 3 ml RNAlater (Ambion, Foster City, CA, USA). Fresh tissue samples were obtained from the patients, with informed consent, and cut into fragments <0.5 cm in any single dimension. The tissues were then immersed into 2 ml RNAlater. All the saliva and tissue samples were frozen 30 min following surgery and stored in liquid nitrogen until use. Tissue sections from each OSCC sample were reviewed and classified by a pathologist. Tumor tissue fragments subsequently processed for RNA isolation should contain ≥80% neoplastic cells, as determined by hematoxylin and eosin (H&E) staining of tissue sections and the adjacent nonmalignant tissues were also validated as no neoplastic cells were present. The clinical status of the disease was heterogeneous across the patients: seven patients showed metastasis, three showed locally advanced disease and six showed localized tumors. The anatomopathological data of the OSCC patients are listed in Table I.

The OSCC tissue and saliva samples were thawed from the liquid nitrogen before use. The tissues were retrieved from RNAlater with sterile forceps. Excess RNAlater solution was blotted away with an absorbent lab wipe. Saliva samples were centrifuged at 20,000 × g for 5 min at 4°C and the supernatants were discarded. Total RNA from tissue and saliva samples were extracted using TRIzol (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions.

Reverse transcription reactions were performed using 1 μg total RNA with the PrimeScript^® RT reagent kit (Takara Bio, Shiga, Japan) using random hexamer primers. Real-time PCR was then performed using the SYBR^® Premix Dimer Eraser kit (Takara Bio). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was evaluated as a housekeeper gene for the qPCR reaction using the commercial PCR primers (Takara Bio). Each test was performed in triplicate and the 2^−ΔCt method was used to calculate the expression of each lncRNA in all the tissue and saliva samples. All the reactions were performed on a Bio-Rad CFX-96 real-time PCR system (Bio-Rad, Hercules, CA, USA) according to the manufacturer's instructions. The experiments were in strict compliance with the Minimum Information about Quantitative Real-Time PCR Experiments (MIQE) guidelines. The nucleotide sequence of the qPCR primers are listed in Table II.

The difference in tissue lncRNA expression among the patients was analyzed using the Mann-Whitney U test concerning clinical parameters such as median age, gender and lymph node metastasis. For paracarcinoma-carcinoma paired tissues, the difference in lncRNA expression was evaluated with a Wilcoxon matched-pairs signed-ranks test. All the statistical analyses were performed with STATA 10.0 (StataCorp LP, College Station, TX, USA). P<0.05 was considered to indicate a statistically significant difference. The lncRNA expression heat-maps shown in Figs. 1A and 2A were generated by matrix2png (http://chibi.ubc.ca/matrix2png/bin//matrix2png.cgi). Rows of each gene expression matrix were normalized to have mean zero, variance one.

We firstly compared the expression profiles from four OSCC samples with those from their matched adjacent non-tumor tissue samples. We found that most of the selected transcripts (4/6) were upregulated in tumors relative to matched adjacent nonmalignant tissue (Fig. 1A and B). One gene, MEG-3, was downregulated in cancer compared with its adjacent nonmalignant tissue (Fig. 1A and B). However, we analyzed a gene (MALAT-1) identified by Ji et al(24), which did not have a significantly different expression between primary tumors and matched adjacent nonmalignant tissues (Fig. 1B). An increased expression of the lncRNA MALAT-1 was firstly observed in metastatic non-small cell lung cancer (24), followed by endometrial stromal sarcoma of the uterus (29) and in six additional types of cancer, including hepatocellular carcinoma, breast, pancreas, lung, colon and prostate cancer (30).

Next, we focused on the expression of these genes in metastatic tumors compared with nonmetastatic tumors. Higher expression levels in the metastatic compared with the nonmetastatic group were confirmed in 3 (HOTAIR, NEAT-1, UCA) of the 6 transcripts analyzed by qPCR (Fig. 2A and B) and the expression of MEG-3 was found to be decreased in the nonmetastatic tissues, which was consistent with data from primary cancer compared with nonmalignant tissue, suggesting that the expression levels of these lncRNAs increased following tumor development and progression. These analyses demonstrated the reliability and the validity of the procedure, since most of these transcripts have been described as cancer- or metastasis-associated genes. For instance, increased expression of lncRNA HOTAIR has previously been shown to be associated with metastasis in breast cancer patients, having a unique association with patient prognosis (8). Subsequently, HOTAIR expression level has been found to correlate with metastasis in colorectal carcinoma (9). Consistent with previous studies, MEG-3 expression was also downregulated in the metastatic compared with the nonmetastatic group (Fig. 2B).

We compared the expression of lncRNA in different gender and age groups. No statistically significant difference in gene expression was found between males and females. Similarly, no statistically significant difference was found between the younger and elder patients, after the classification of patients in different age groups based on the median age (50 years; data not shown).

It has been previously shown that a subset of the salivary mRNAs can be used as biomarkers for oral cancer detection (31,32). To discover whether lncRNAs are present in salivary fluid, whole saliva was used for lncRNA identification.

In a pilot experiment, 1 ml salivary samples obtained from patients with OSCC (n=9), in which there were 4 patients with lymph node metastasis, were examined using qPCR for the presence of lncRNAs. Samples with a triplicate lncRNA qPCR result of Ct<40 were considered to be positive. Of the lncRNAs investigated, MALAT-1 was present in all the participants, while the mRNA expression was not significantly different between metastatic and nonmetastatic samples. However, HOTAIR was detected in 5/9 patients, 3 were positive of the 4 patients with lymph node metastasis, which is higher compared with those without metastasis. There was no detectable expression for the other lncRNAs investigated (Table III). Taken together, these data indicate that whole saliva contains detectable amounts of certain lncRNAs which may be potential markers for OSCC diagnosis.