Bioinformatics analysis of the expression and role of microRNA-221-3p in head and neck squamous cell carcinoma

Background Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide, associated with a high rate of morbidity and mortality. However, the target genes of miR-221-3p and the underlying mechanism involved in HNSCC are still not clear. Therefore, in the current study, we studied the role of miR-221-3p in the HNSCC. Methods Tissues collected from 48 control and 21 HNSCC patients were processed to check the differential expression of miR-221-3p by RT-qPCR. Overexpression of microRNA-221-3p (miR-221-3p) is significantly correlated to the onset and progression of HNSCC. We also conducted the meta-analysis of the cancer literature from the cancer genome atlas (TCGA) and the Gene Expression Omnibus (GEO) database to estimate the expression of miR-221-3p in HNSCC. The miR-221-3p target genes in the HNSCC were predicted with the miRWalk and TCGA databases, and functionally annotated via the Gene Ontology. Finally, Spearman’s analysis was used to determine the role of the related target genes in important pathways involved in the development of HNSCC. Results We observed a significantly higher expression of miR-221-3p in HNSCC compared to the normal with a summary receiver operating characteristic (sROC) of 0.86(95% Cl: 0.83,0.89). The KEGG and GO comprehensive analysis predicted that miR-221-3p might be involved in the development of HNSCC through the following metabolic pathways, viz. Drug metabolism - cytochrome P450 UGT1A7 and MAOB may be important genes for the role of miR-221-3p. Conclusion Based on bioinformatics analysis, our results indicate that miR-221-3p may be used as a non-invasive and hypersensitive biomarker in the diagnosis. Thus, it can be concluded that miR-221-3p may be an extremely important gene locus involved in the process of the deterioration and eventual tumorigenesis of HNSCC. Hopefully, additional work will validate its usefulness as a target for future clinical research.


Background
Head and neck squamous cell carcinoma (HNSCC), encompassing oral squamous cell carcinoma, oropharyngeal squamous cell carcinoma, hypopharyngeal squamous cell carcinoma and laryngeal squamous cell carcinoma, is the sixth most common cancer worldwide, accounting for approximately 1-2% of all cancer deaths, with approximately 600,000 new cases detected each year globally [1][2][3]. The treatment of patients at the early stage is relatively successful, but the overall survival rate of recurrent or metastatic HNSCC remains low and has barely seen any improvements for decades [4,5]. Although the past three decades have seen several improvements in diagnostic tools and treatment regimens, the overall survival rate for advanced (stage III-IV) HNSCC is only 65% [6,7]. The current standard of care for such patients is surgery along with radiation and chemotherapy, but these have not significantly improved the 5-year survival of HNSCC patients. The main reason for the decline in patient survival is the lack of effective therapeutic targets for the development of HNSCC [8,9].
Recent years have seen an increasing concentration of research on microRNAs (miRNAs) [10,11], which are a type of highly conserved single-stranded noncoding RNA containing 17~22 nucleotides and are involved in the process of tumorigenesis, cell survival, and chemosensitivity [12][13][14]. They bind to the 3′ untranslated region (3'UTR) of different target mRNA (messenger RNA) genes to degrade or inhibit the mRNA translation of target genes associated with a tumor suppressor function [13,15,16]. In normal healthy individuals, miR-221-3p is observed to play a role in the process of vascular proliferation [17], while the tumor promoter microRNA-221 is involved in the process of regulating apoptosis of tumor cells [18][19][20] and is associated with a variety of cancer types, including hepatocellular cancer [21], cutaneous melanoma [19], prostate cancer [20], and non-small cell lung cancer [22].
Studies have shown that specific miRNA profiles can be identified between tumor tissue and adjacent healthy tissue in HNSCC patients [2,23]. For instance, studies have reported a link between miR-221 and vascular invasion in HNSCC [14,15,[24][25][26]. However, the definite target gene of miR-221-3p and its biological mechanism of action are still not clear. Thus, in the current study, we investigated the expression of miR-221-3p in HNSC C and attempted to explore the correlation between the two.

RT -qPCR
Tissue specimens were collected from 48 control and 21 HNSCC patients from the Department of Pathology, First Affiliated Hospital of Guangxi Medical University (Nanning, Guangxi Medical University). Total RNA was extracted from formalin-fixed, paraffin-embedded (FFPE) tissues using an FFPE RNA kit (Omega Bio-Tek, Norcross, GA, USA). RNA was reverse-transcribed into cDNA using the Mir-X miRNA qRT-PCR SYBR Kit (Takara, California, USA). The cDNA samples were processed for qPCR with SYBR-Green Master Mix (Takara, Tokyo, Japan) on an ABI 7500 cycler (Applied Biosystems) under the following conditions: initial denaturation at 95°C for 30 s, followed by 40 cycles of denaturation at 95°C for 5 s and annealing at 60°C for 34 s. The expression of miR-221-3p in HNSCC tissues relative to negative control (NC) tissues was calculated using the 2-ΔΔCT method, with U6 as the internal control. The primers for miR-221-3p and U6 were synthesized by TaKaRa (Dalian, Liaoning, China), and the sequences were as follows: miR-221-3p: forward 5′-AGCUAC AUUGUCUGCUGGGUUUC − 3′ and reverse 5′ mRQ 3′; U6: forward 5′-GGAACGATACAGAGAAGATTAG C-3′ and reverse 5′-TGGAACGCTTCACGAATT TGCG-3′. All the experiments were repeated three times.  MiR-221-3p levels in HNSCC and NC tissue specimens. RT-qPCR was used to detect miR-221-3p was significantly up-regulated in the tumor tissues of HNSCC (n = 21) relative to control tissues (n = 48)(p<0.01) SCC OR "squamous cell cancer" OR "squamous cell carcinoma") AND (oropharynx OR oropharyngeal OR "head and neck" OR nose OR nasopharynx OR "nasal sinus" OR "nasal cavity" OR "oral cavity" OR hypopharynx OR oral OR laryngopharynx OR larynx OR laryngopharyngeal OR laryngeal OR pharyngeal OR tongue OR tonsil OR tonsillar OR cheek OR palatal OR "paranasal sinuses" OR buccal OR lip) AND (microRNA-221-3p OR miRNA-221-3p OR "miR 221-3p" OR "miRNA 221-3p" OR miRNA221-3p OR miR221-3p).

Selection criteria and data extraction
The databases were searched independently by two researchers who selected the studies based on the following inclusion criteria: (1) comparison of HNSCC and noncancerous tissues; (2) validation of miR-221-3p expression levels via reverse transcription quantitative PCR (RT-qPCR); (3) evaluation of the association between miR-221-3p expression and clinical outcomes; and (4) availability of sufficient data to calculate the mean, standard deviation (SD) and 95% confidence intervals   (2) inclusion of unqualified data; (3) publication language other than English or Chinese; (4) overlapping or duplicate publications; and (5) letters, reviews, comments, editorials, conference articles, laboratory studies or case reports. The reviewers independently appraised the quality of data in each eligible study and extracted the first author name, year of publication, country of origin, sample type, sample size and analysis method. For articles with incomplete information, the authors were contacted to obtain relevant information.

Microarray data collection from GEO
The microarray data of HNSCC samples uploaded until October 1, 2019, were obtained from the GEO database (https://www.ncbi.nlm.nih.gov/gds/) using the following keywords: (HNSCC OR SCC OR "squamous cell cancer" OR "squamous cell carcinoma") AND (oropharynx OR oropharyngeal OR "head and neck" OR nose OR nasopharynx OR "nasal sinus" OR "nasal cavity" OR "oral cavity" OR hypopharynx OR oral OR laryngopharynx OR larynx OR laryngopharyngeal OR laryngeal OR pharyngeal OR tonsil OR tonsillar OR tongue OR cheek OR palatal OR "paranasal sinuses" OR buccal OR lip) AND (microRNA OR miRNA OR "miR" OR "miRNA"). The inclusion criteria for the microarray datasets were as follows: (1) comprising data from HNSCC and noncancerous tissues; (2) evaluation of the association between miR-221-3p expression and clinical outcomes; and (3) availability of sufficient data to calculate the mean, SD and 95% CI. The exclusion criteria were as follows: (1) unrelated to this study; (2) unqualified data; and (3) publication language other than English.

RNA sequencing data selection from TCGA
The miR-221-3p expression data of HNSCC and normal tissues and the clinicopathological parameters of patients were downloaded and extracted from the OncoLnc website (http://www.oncolnc.org/).

Predicting the target genes of miR-221-3p in HNSCC
We use the MiRWalk 2.0 [27] (http://zmf.umm. uniheidelberg.de/apps/zmf/mirwalk2/) and GEPIA (http://gepia.cancer-pku.cn/index.html) databases to retrieve differentially expressed genes of miR-221-3p in HNSCC. Gene Ontology (GO) [28] analysis, which was used to define the biological processes (BPs), molecular functions (MFs), and cellular components (CCs) of the target genes, and Kyoto Encyclopedia of Genes and Genomes (KEGG) [29] pathway enrichment analyses were conducted using R (version 3.6.1), and the ClusterProfiler package of R was used to visualize the results. The Search Tool for the Retrieval of Interacting Genes (STRI NG) (https://string-db.org/) was then used to establish a protein-protein interaction (PPI) network of the miR-221-3p target genes associated with the significantly enriched pathways. The expression levels of the miR-221-3p target genes in HNSCC and nontumor tissues were determined with UALCOULD [30] (http://ualcan. path.uab.edu/index.html). Finally, the LinkedOmics [31] (http://www.linkedomics.org/) Spearman's analysis tool was applied to determine the correlation between the expression levels of miR-221-3p and the potential target genes involved in notable signaling pathways.

Statistical analysis
Independent-Samples T Test for the RT-qPCR analysis of two groups. SPSS 24.0 software (IBM, Somers, NY) was used for the statistical analysis. The mean ± SD (X ± s) were calculated, and t test was used for comparisons between groups for all the measurement data. Data that were used for our meta-analysis as well as TCGA data (Nanning, Guangxi Zhuang Autonomous Region, China) were analyzed using STATA 15.0 software. The standard mean difference (SMD) and 95% confidence interval were used to estimate the expression value of miR-221-3p. Scatter plots of the data of both the normal and HNSCC tissues were prepared with GraphPad Prism 6 (Nanning, Guangxi Zhuang Autonomous Region, China) and used to assess the expression level of miR-221-3p. Evidence of bias was assessed by visual funnel plots and Egger's regression asymmetry test. In addition, the I 2 index was used to evaluate the potential heterogeneity of the selected data. When I 2 > 50% or P < 0.05, we used the fixed effect model; otherwise, the random effect model was used.. The count rate was expressed as a percentage (%), and we used the X 2 test for comparisons.   false negatives (FNs) and true negatives (TNs) were used to determine the diagnostic value of miR-221-3p. We used Meta-disc software to calculate the specificity, sensitivity, negative and positive likelihood ratios, and summary receiver operating characteristic (sROC) curve. We also performed sensitivity analysis to assess the differences between the sample sizes. Moreover, all the miR-221-3p expression data, which included the TCGA sequencing data, were normalized to log2 to improve the normality of the measurements. P < 0.05 was taken as a statistically significant difference for all analyses.
Result miR-221-3p is highly expressed in the tumor tissues The RT-qPCR data clearly showed that miR-221-3p is highly expressed in the tissues of nasopharyngeal carcinoma patients compared to the controls (p < 0.05) (Fig.1).

Selection of relevant literature and microarray data extraction
A total of 17 articles were retrieved from the initial search, and the following paper was selected after the full-text review based on the inclusion criteria: Zhou Cheng. et al.
Risk of bias assessment and meta-analysis of miR-221-3p in HNSCThe results of the visual funnel plots show a symmetrical shape, which indicates the absence of any significant publication bias as a whole. Sensitivity analysis performed to assess the heterogeneity of the samples showed that there were no significant differences between the studies (Fig.3). There were 911 HNSCC samples and 289 normal samples from the PubMed, GEO, and TCGA datasets. The meta-analysis performed on all data to explore the expression level of miR-221-3p in HNSCC indicated a high degree of heterogeneity between these studies. Therefore, the random-effects model was selected, and the combined standard mean difference (SMD) was observed to be 0.72 (95% CI: 0.28, 1.15) (Fig.4). The meta-analysis data indicate that miR-221-3p expression is significantly upregulated in HNSC C, which is consistent with the results obtained by RT-qPCR. In addition, miR-221-3p expression was significantly increased in HNSCC in the GSE11163, GSE103931, GSE113956, GSE31277, GSE45238, GSE82064, TCGA and PMID30928631 datasets, and the expression levels of miR-221-3p in HNSCC and normal tissues from each included dataset are shown in Fig.5.

Clinical features of HNSCC based on the TCGA
Based on the HNSCC sample data extracted from the TCGA database, we analyzed 527 samples, and the results indicated the absence of any statistically significant difference between miR-221-3p expression and age, lymphatic invasion, neoplasm histologic grade and any other clinicopathological features. However, the results indicated statistically significant differences in the expression of miR-221-3p in different tissues, sexes, and even alcohol consumption statuses ( Table 2).
The prospective target genes of miR-221-3p in HNSCC By using the MiRWalk 2.0 and GEPIA databases, we retrieved 5311 and 467 differentially expressed genes, respectively. Bioinformatics analysis showed a total of 117 overlapping genes that were involved (Fig.9, Table 3). These overlapped 117 genes could be considered as the target genes that miR-221-3p might play a role in HNSC C. The enriched GO and KEGG pathway categories of   the overlapped 117 genes with p < 0.05 are shown in Fig.10, Table 4 and Table 5. For the cellular components, the identified target genes were mostly enriched in the actin cytoskeleton, sarcomeres, and contractile fiber part; for the molecular functions, the target genes were mainly enriched in aromatase activity, oxidoreductase activity, and cofactor binding. KEGG enrichment analysis showed that miR-221-3p plays a significant role in HNSCC through a variety of pathways, including the drug metabolism pathways of the cytochrome P450 signaling pathway. The drug metabolism-cytochrome P450related genes include GSTA1, UGT1A7, CYP3A5, FMO2, MAOB, ADH1C and ADH1B, and the protein-protein interaction (PPI) network of these genes is shown in Fig.11.

Discussion
The drug metabolism-cytochrome P450 signaling pathway is an essential signal transduction pathway in cells.
It is an important biological function of cell survival, proliferation and apoptosis. For example, Liping Wang and his colleagues suggest [33] that the inhibition of CYP3A5 in cytochrome P450 drug metabolism can drive the migration, proliferation, and invasion of HNSCC cells. Dongfang Wang et al. showed that cytochrome P450 is correlated with the overall survival and vascular invasion of hepatocellular carcinoma (HCC) patients.   [38][39][40][41][42] control the progression of tumors. For example, miR-221-3p has been regarded as a tumor biomarker that can be used to assess the clinical prognosis of breast cancer [15]. In healthy humans, it has been shown that miR-221-3p plays a role in the process of vascular proliferation, while the tumor promoter miR-221-3p regulates the apoptosis of tumor cells. The current study focused on the expression pattern of miR-221-3p in HNSCC, identifying the exact target genes of this miRNA and its biological mechanism of action through bioinformatics analysis. The biological pathways relevant to the actions of miR-221-3p will guide further understanding of the mechanisms of HNSCC.
Of course, the limitations of our study should also be mentioned. On the one hand, this study is restricted in that the analysis of differential miRNAs was only based on HNSCC and noncancerous tissues, and other samples (e.g., blood) were not assessed. On the other hand, the target genes of miR-221-3p have not been experimentally confirmed, and our conclusions will need to be confirmed by clinical or molecular biological methods in the future. In addition, only English studies were used for the basis of our meta-analysis, which did not include

Conclusion
In our study, we found that miR-221-3p levels were apparently upregulated in HNSCC compared to normal tissues. MAOB and UGT1A7 are potentially important targets of miR-221-3p. MiR-221-3p may be used as a noninvasive and hypersensitive biomarker for the diagnosis of HNSCC and is an extremely important gene locus involved in the process of the deterioration and eventual tumorigenesis of HNSCC. Finally, the biological pathways relevant to the actions of miR-221-3p will provide insights into its potential molecular mechanisms in HNSCC. Larger-scale studies will be needed to validate its diagnostic promise. Hopefully, additional work will validate its usefulness as a target for future clinical research.