WASJ World Academy of Sciences Journal 2632-2900 2632-2919 D.A. Spandidos WASJ-0-0-00077 10.3892/wasj.2020.77 Articles Analysis of TCGA data of differentially expressed EMT-related genes and miRNAs across various malignancies to identify potential biomarkers KyritsisKonstantinos A. 1 AkrivouMelpomeni G. 1 GiassafakiLefki-Pavlina N. 1 GrigoriadisNikolaos G. 1 2 VizirianakisIoannis S. 1 3 4 Laboratory of Pharmacology, School of Pharmacy, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece Biogenea Pharmaceuticals Ltd., 54627 Thessaloniki, Greece Functional Proteomics and Systems Biology Research Group at AUTH (FunPATH) Research Group, KEDEK-Aristotle University of Thessaloniki, Balkan Center, 57001 Thessaloniki, Greece Department of Life and Health Sciences, University of Nicosia, CY-1700 Nicosia, Cyprus Correspondence to: Professor Ioannis S. Vizirianakis, Laboratory of Pharmacology, School of Pharmacy, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece ivizir@pharm.auth.gr 01 2021 02 12 2020 3 1 6 08 09 2020 30 11 2020 Copyright: © Kyritsis et al. 2020 This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

Tumor heterogeneity presents a hindering factor that leads to therapeutic failures and limits the improvement of clinical outcomes within the concept of precision medicine. This heterogenous characteristic provides the epithelial mesenchymal plasticity that is considered an advantage for cancer cell metabolism and genome function to be adjusted within the microenvironment, and also plays a role in the development of drug resistance and metastasis. To this respect, identifying druggable molecular targets that modulate signaling networks, which contribute to cancer cell heterogeneity, could provide innovative therapeutics with improved safety and efficacy profiles. The present study attempted to identify potentially druggable molecular targets that have been connected to the process of epithelial-to-mesenchymal transition (EMT). Towards this goal, gene and miRNA differential expression analyses were performed for cancer patients with 4 and 3 different tumor types, respectively, using data that were retrieved from The Cancer Genome Atlas (TCGA) program. Furthermore, the dbEMT 1.0 database was used to limit the results to differentially expressed molecular targets that have already been associated with EMT. The analysis resulted in the identification of multiple EMT-associated genes and miRNAs for all types of cancer, which, through pairwise comparisons, were separated into groups of common potential targets for different malignancies. Differential gene expression profiling by RT-qPCR analysis was also carried out for a number of selected genes and miR-21 in human cancer cell lines. Notably, EMT-associated homeobox B9 (HOXB9) and miR-137 were found to have a deregulated expression in all malignancies examined, thus increasing their potential as druggable targets for cancer therapy. Overall, the present study presents an approach that, through systematic in silico analysis, could lead to the selection of potential druggable biomarkers of broader utility for several tumor types, irrespective of their tissue of origin.

 epithelial-to-mesenchymal transition the cancer genome atlas RNA-Seq miRNA-Seq meta-analysis druggable targets anticancer therapeutics Introduction

Cellular and genomic heterogeneity in tumors involves the microenvironment, cancer stem cells and epithelial-to-mesenchymal transition (EMT) (1,2). EMT constitutes a complex and dynamic biological process during which epithelial cells transdifferentiate towards a mesenchymal phenotype. Unlike epithelial cells, which are characterized by polarity and maintain firmly cell-to-cell adhesion contacts through cellular adhesion molecules (CAMs), mesenchymal cells display an increased mobility and loose organization within the extracellular matrix. EMT plays an important role in physiological processes, including organ formation during embryogenesis and tissue regeneration; however, its involvement has also been confirmed in tumor initiation, progression and metastasis. The transition can be triggered through various stimuli, including different factors of the tumor microenvironment (cytokines, growth factors, etc.), as well as immune responses, hypoxia and antitumor drug treatment. Notably, EMT is reversible, exhibiting plasticity, with mesenchymal cells being capable of converting back to an epithelial phenotype through a process known as mesenchymal-to-epithelial transition (MET). The combination of EMT and MET can lead to a mixed, dynamic population of cancer cells exhibiting both epithelial and mesenchymal characteristics that also promote circulating tumor cell (CTC) formation. This can result in the disruption of cellular adhesion and increased migratory and invasive capabilities, which can lead to metastasis (3,4).

The EMT phenotypic plasticity of tumor cells contributes to molecular and cellular heterogeneity, that leads to acquired drug resistance to cytotoxic or molecularly targeted therapy in clinical practice. Moreover, the differential pharmacological response limits the productivity and clinical outcomes of innovative therapeutic approaches (5-7). Moreover, the interplay of transcription (including the Snail, Twist and Zeb families) and epigenetic factors (e.g., the miR-200 family, miR-205, miR-203, miR-34 and miR-29b) drive the regulatory network program of EMT plasticity in cancer (6). It would be of interest if common molecular drivers in these EMT and MET processes that are deregulated in various types of tumors could be characterized; following clinical validation, biomarkers could be developed which may be used in cancer therapy.

Previously the authors characterized the expression levels of several epithelial markers, namely desmoglein 3 (DSG3), E-cadherin and β-/γ-catenins (β-/γ-catenins) in monolayer (ML) and multicellular aggregates (MCAs) of the HSC-3 cell line (oral squamous carcinoma) in vitro, as well as in clinical samples of oral leukoplakia (OL) and oral squamous cell carcinoma (OSCC) in vivo (8). Of note, the downregulation of DSG3, E-cadherin and β-/γ-catenins was observed to be significantly associated with the grade of OL-dysplasia and OSCC samples (8). Furthermore, the switch of expression and potent perinuclear aggregation of DSG3 and γ-catenin were observed in both HSC-3 cells and OL/OSCC samples. These observations support the involvement of DSG3 and γ-catenin in the progression of oral epithelial cell malignancy. It was also suggested that these genes may serve as potential predictive biomarkers, along with E-cadherin and β-catenin, of the malignant transformation risk of oral dysplasia and the biological behavior (aggressiveness) of oral cancer, respectively (8).

In the present study, an in silico analysis was performed using RNA-Seq data from The Cancer Genome Atlas (TCGA) database (9), in order to identify genes that are involved in the EMT and MET processes, and that may serve as possible biomarkers and/or therapeutic targets in clinical practice. For this purpose, differential gene and microRNA (miRNA/miR) expression analyses were carried out between solid tissue normal (STN) and primary solid tumor (PST) samples from 4 different types of cancer (head and neck, prostate and breast cancer, and glioblastoma). The accurate separation of STN and PST samples into distinct clusters was confirmed using principal component analysis (PCA) based solely on the identified differentially expressed (DE) genes/miRNAs (Fig. 1). Additionally, dbEMT 1.0, a database containing EMT-related genes collected through extensive literature search (10), was used to further filter DE genes and miRNAs, and keep those that have been found to be involved in EMT and/or MET. On the whole, the present study identified and reported several DE and EMT/MET-related genes and miRNAs in various types of malignancies, whose potential in clinical utilization will be further evaluated by characterizing their expression levels in cell line EMT models and in clinical samples in the future.

 Materials and methods <sec> <title>Cancer data

RNA-Seq and miRNA-Seq data from TCGA (9) were retrieved using TCGAbiolinks (11). Specifically, from the available pre-processed data types, gene and miRNA count data, derived from HTSeq software (12), were selected for 4 different types of cancer: i) Head-and-neck squamous cell carcinoma (TCGA-HNSC); ii) breast cancer (TCGA-BRCA); iii) prostate adenocarcinoma (TCGA-PRAD); and iv) glioblastoma multiforme (TCGA-GBM).

 Differential expression analysis

To ensure the power of statistical testing, only STN and PST samples were selected to perform differential expression analysis for genes and miRNAs using DESeq2(13).

Due to the lack of sufficient miRNA-Seq samples in TCGA-GBM, miRNA differential expression analysis was performed only for TCGA-BRCA, TCGA-HNSC and TCGA-PRAD.

A minimum threshold of 100 and 10 total number of counts was set to filter out genes and miRNAs with very low counts, respectively. Genes and miRNAs with an absolute log2 fold change (LFC) >1 and a false discovery rate (FDR) (14) adjusted P-value <0.001 were reported as statistically significant, DE targets. Variance-stabilizing transformation (VST) was applied to DE gene or miRNA expression values of all samples, followed by PCA. Both VST and PCA were performed using DESeq2(13).

 EMT-associated genes and miRNAs

Genes and miRNAs that have been associated with EMT were retrieved from dbEMT 1.0, a database containing EMT-related genes and miRNAs that were collected through extensive literature search (10). Comparisons of EMT targets with DE genes and miRNAs was performed with R statistical programming language. Venn diagrams were created using limma (15).

 Survival analysis of DE genes and miRNAs

Survival analysis was performed on DE genes and miRNAs using clinical metadata from TCGA. Specifically, for each DE gene or miRNA, PST samples were assigned into 2 separate groups, depending on whether the target expression of each sample was higher (high expression) or lower (low expression) than the median. Kaplan-Meier analysis on the 2 groups was performed for each DE gene or miRNA and selected, statistically significant targets are reported in Kaplan-Meier survival curves. For the statistical analysis of DE genes (P-value <0.001) and miRNAs (P-value <0.1) the non-parametric log-rank test was used. An exception to this was miR-16-1 in HNSC cancer samples, with a P-value of ~0.12. Survival analysis was performed using a survival analysis package (16,17) and survminer (18).

 Cell cultures, RNA isolation and RT-qPCR analysis

The established cell lines of human breast epithelial carcinoma MCF-7 (Cellosaurus, CVCL-0031) and MDA-MB-231 (Cellosaurus, CVCL-0062), as well as the human tongue squamous carcinoma HSC-3 (Cellosaurus, CVCL-1288) and keratinocyte HaCaT (Cellosaurus, CVCL-0038) cancer cells that are routinely used in the authors' laboratory were cultured as previously described (19,20). Moreover, the RNA isolation, RT-cDNA synthesis, as well as the RT-qPCR analysis were carried out as previously described (19,20). In brief, total RNA was extracted from cells using TRItidY G (Panreac, Applichem), quantified using a Nanodrop ND-100 Spectrometer and reverse transcribed into cDNA by applying the QuantiTect Reverse Transcription kit (Qiagen, Inc.). qPCR was performed on a 7500 Real-Time PCR System (Applied Biosystems; Thermo Fisher Scientific, Inc.) using KAPA SYBR® FAST qPCR Master Mix (KAPA Biosystems) under optimized conditions: 95˚C for 3 min followed by 40 cycles at 95˚C for 3 sec and 60˚C for 20 sec. Primers designed and used during the present study were as follows: (PPARG forward, 5'-TCG-AGG-ACA-CCG-GAG-AGG-3' and reverse, 5'-CAC-GGA-GCT-GAT-CCC-AAA-GT-3'; HMGA2 forward, 5'-GAA-AAA-CGG-CCA-AGA-GGC-AG-3' and reverse, 5'-AGA-GCT-ATC-CTG-GAC-TCC-TCC-3'; FOXM1 forward, 5'-ACC-GCT-ACT-TGA-CAT-TGG-AC-3' and reverse, 5'-GGG-AGT-TCG-GTT-TTG-ATG-GTC-3'; CAV-1 forward, 5'-CCC-AGG-GAA-ACC-TCC-TCA-CAG-3' and reverse, 5'-GGC-AGA-TAG-CAG-AAG-CGG-AC-3'; TGFB1-F forward, 5'-ACT-GCG-GAT-CTC-TGT-GTC-ATT-G-3' and reverse, 5'-ACA-GTA-GTG-TTC-CCC-ACT-GGT-C-3'; Vimentin forward, 5'-GGC-TCG-TCA-CCT-TCG-TGA-AT-3' and reverse, 5'-GAG-AAA-TCC-TGC-TCT-CCT-CGC-3'; β-actin forward, 5'-TTG-CTG-ACA-GGA-TGC-AGA-AG-3' and reverse, 5'-TGA-TCC-ACA-TCT-GCT-GGA-AG-3'). β-actin was used as an endogenous control to normalize the gene expression levels.

The expression of miRNAs was also carried out by RT-qPCR using the miScript SYBR®-Green PCR kit (Qiagen, Inc.). Total cellular RNA extraction and quantification was performed as indicated above, whereas cDNA synthesis was executed with the miScript II RT kit (Qiagen, Inc.). Hsa-miR-21-5p (miR-21, 5'-UAGCUUAUCAGACUGAUGUUGA-3') was designed and used during this experiment and SNORD6 small nucleolar RNA, C/D box 6 (also known as mgh28S-2412) (Qiagen, Inc.) was used as reference RNA gene. The reaction conditions consisted of polymerase activation/denaturation at 95˚C for 15 min, followed by 40 cycles at 94˚C for 15 sec and 55˚C for 30 sec.

In both cases, the relative mRNA/miRNA concentrations were calculated using the 2-ΔΔCq method (21) and the results obtained were represented as fold changes in the diagrams.

 Statistical analysis

The results from 2 independent biological experiments (triplicate measurements) are shown and the data are expressed as the means ± standard deviation (SE). Comparisons were carried out using a Student's t test, whereas the statistical analysis was performed using GraphPad Prism 6.0 (GraphPad Software, Inc.). P<0.05 was considered to indicate a statistically significant difference.

 Results <sec> <title>DE genes and miRNAs

Accessing the GDC data portal through TCGAbiolinks enabled the retrieval of RNA-Seq for 4 different types of malignancies. For each project, STN and PST were identified as the predominant categories containing the majority of samples. Specifically, for TCGA-BRCA, TCGA-HNSC, TCGA-PRAD and TCGA-GBM, STN + PST samples numbering 113 + 1,102, 44 + 523, 52 + 498 and 5 + 156, respectively, were obtained, filtered for low count genes and compared to identify DE genes. Based on strict criteria (please see the ‘Materials and methods' section) a subset of all genes analyzed was determined to be DE genes in each tumor (percentage of DE genes in each TCGA-project: TCGA-BRCA, 8.95%; TCGA-HNSC, 6.1%; TCGA-PRAD, 2.56%; TCGA-GBM, 6.52 %) (Table I). Similarly, miRNA-Seq data were retrieved for 3 TCGA projects and major sample categories STN and PST were used for DE miRNA identification (TCGA-BRCA, TCGA-HNSC and TCGA-PRAD with STN + PST samples numbering 104 + 1,096, 44 + 523, 52 + 498, respectively). Following analysis, the percentage of DE miRNAs were determined to be as follows: TCGA-BRCA, 19.1%; TCGA-HNSC, 18.33%; TCGA-PRAD, 9.04% (Table I).

To further confirm these findings, PCA analysis was performed using DE gene and miRNA expression values for all samples of each TCGA-project. Following dimension reduction, loadings of principal components (PC) 1 and 2 were plotted against each other and displayed the formation of distinct groups between STN and PST, as well as other sample types, such as metastatic, in each type of cancer (Figs. 1 and 2).

Moreover, survival analysis was performed on DE genes and miRNAs (please see the ‘Materials and methods’ section). In total, 6 genes (Fig. 3) and 3 miRNAs (Fig. 4) were reported, for which patients with high or low expression levels presented considerable differences in survival. The results concerning the association of the genes PGK1, PCMT1, FDG3, HOXB9, NSUN5 and ZNF330 to patient survival probability, are in agreement with those previously reported by the Human Protein Atlas project (22), with the exception of the unfavorable prognosis of HOXB9 overexpression in glioblastoma, which was identified during the current analysis (Fig. 3). As regards the miRNAs, miR-16, known for its tumor suppressive functions (23-25), miR-92a-1(26) and miR-484 (27-29) exhibited an associated with a favorable and poor prognosis, respectively, for patients with breast and head and neck cancer (Fig. 4). These results support the approach for the identification of DE targets.

 EMT-associated genes and miRNAs

In total, 344 genes and 20 miRNAs that, following an exhaustive literature search, constitute a collection of well-characterized EMT-associated targets, were retrieved from dbEMT 1.0(10). Direct comparisons between the 2 gene collections revealed that only a small fraction of DE genes has been identified as directly related to the EMT process (percentage of DE genes that were EMT-related: TCGA-BRCA, 1.7%; TCGA-HNSC, 1.5%; TCGA-PRAD, 1%; TCGA-GBM, 1.9%) (Table I). A small number of miRNAs reported in dbEMT 1.0 was also found in the present collection of DE miRNAs (percentage of DE miRNAs that were EMT-related: TCGA-BRCA, 3.7%; TCGA-HNSC, 2.4%; TCGA-PRAD, 3.6%) (Table I).

Furthermore, DE genes and miRNAs related to EMT were found both up- and downregulated in each type of cancer (Figs. 5 and 6). Careful inspection of these results is required to decipher the role of each gene and miRNA in EMT, at the context of each malignancy.

 EMT targets amongst different malignancies

With an aim of identifying targets that are commonly deregulated and EMT-associated between different types of cancer, pairwise comparisons of DE genes and miRNAs associated with EMT were performed. It was found that different types of malignancies shared several deregulated molecules (Tables II and III), with HOXB9 and miR-137 being common for all types of malignancies that were examined (Fig. 7).

 Assessment of DE genes and miRNAs related to EMT in breast, and head and neck cell carcinoma lines

Based on the data obtained, the present study wished to assess the expression level in a number of DE genes in well-characterized human breast, and head and neck cancer cell lines. To this end, the selection was made for caveolin-1, FOXM1 and Vimentin for the human breast epithelial carcinoma cell lines, MCF-7 and MDA-MB-231. As for the head and neck cancer cell lines, the human oral HSC-3 and keratinocyte HaCaT cancer cells were used to assess the gene expression levels of HMGA2, TGFB1, FOXM1 and PPARG. Moreover, from the DE miRNAs, the expression of miR-21 was selected and was assessed in all these 4 cell lines.

As shown in Fig. 8, for the breast epithelial carcinoma cell lines, the expression of caveolin-1 (Fig. 8A) was markedly higher in the MDA-MB-231 cells than in the MCF-7 cells (P<0.001). As regards the expression of miR-21 (Fig. 8B), it was higher again in the MDA-MB-231 cells compared to the MCF-7 cells, although the difference was not statistically significant. Such gene and miR-21 differential expression profiles may contribute to the observed metabolic and phenotypic behavior of these 2 cell lines. Indeed, although they are both invasive ductal breast carcinoma cells, they have a number of phenotypic and genotypic differences. MCF-7 are estrogen receptor-positive cells, whereas MDA-MB-231 cells are estrogen and progesterone receptor-negative; in addition, MCF-7 cells express the epithelial phenotype in contrast to the MDA-MB-231 cells that are more mesenchymal (30-32).

Similarly, as shown in Fig. 9, the expression profiles of the FOXM1 and HMGA2 genes exhibited higher levels in the HSC-3 as compared to the HaCaT cells (Fig. 9A), although this difference was not statistically significant. On the contrary, miR-21 was found to have an increased expression in HaCaT compared to HSC-3 cells, a result that was again, not significant. It is interesting that the pattern of expression of these EMT-related genes and that of miR-21 varies in these 2 cell lines; however, it is still not known to what extent such behavior contributes to any metabolic and phenotypic property seen in these 2 cell lines.

 Discussion

Through preliminary in silico analysis, the present study identified a list of genes and miRNAs that were DE and associated with EMT and/or MET in different types of cancer, such as head and neck, breast and prostate cancer, and glioblastoma. Moreover, RT-qPCR analysis revealed differential expression profiles of selected EMT-related genes and miR-21 in a number of human breast, and head and neck carcinoma cell lines. TCGA-BRCA and -HNSC cancer samples shared the most DE and dbEMT 1.0 reported genes and miRNAs, in accordance with their common epithelial tissue origin. The present study, by analyzing the expression profiles of EMT-related genes and miRNAs in patient cancer samples, suggests that HOXB9 and miR-137 present the same deregulated patterns, independent of tumor type. HOXB9 belongs to HOX gene family that in human plays crucial role in physiology and pathophysiology, by modulating cell development, differentiation and growth. It is noteworthy that the aberrant expression of HOX genes has been shown to contribute to cancer progression and development (33,34). Indeed, it has been observed that HOX genes exhibit a dysregulated expression in leukemia, ovarian and lung cancer (35-38). The function of HOXB9 novel tumor suppressor in the regulation of colon adenocarcinoma progression has also been identified (39). Moreover, it has been shown that HOXB9 is associated with the emergence of radioresistance, as well as the development of resistance in anti-VEGF therapy in colorectal cancer (40,41). Importantly, a recent study identified HOXB9 as one key gene in a 5-gene molecular prognostic signature in patients with laryngeal cancer (42). In addition, the implication of HOX9 in prostate cancer cell progression has been recently proposed (43). Furthermore, the present study reports HOXB9, along with PGK1 (44,45), PCMT1 (46,47), NSUN5 (48,49) and ZNF330(50), as unfavorable markers for the survival of patients with the types of cancer examined herein. Of note, the present study demonstrated showed that HOXB9 overexpression was associated with a poor prognosis for patients with both head and neck cancer, previously reported in Human Protein Atlas (22), and those with glioblastoma. An exception was FGD3 (51,52), with its overexpression being predictive of a favorable outcome for patients with head and neck cancer (Fig. 3).

miRNAs represent important players in the post-transcriptionally regulation of gene expression. In this manner, they affect signaling pathways and cellular processes with implication in cancer progression and development (53-56). miR-137 has shown to control tumorigenesis, invasion and metastasis in pancreatic neuroendocrine tumors (56). Moreover, miR-137 exhibits crucial developmental roles in neuronal differentiation (57). In addition, miR-137, by modulating SLC1A5-dependent glutamine uptake, is involved in the progression of head and neck squamous cell carcinoma (58,59). The significant role of miR-137 in the progression, diagnosis and prognosis of hepatocellular carcinoma has also been documented (60). The therapeutic potential of targeting miR-137 in non-small cell lung cancer (NSCLC) has been recently proposed (61). By retrospectively analyzing tumor patient data, the task of identifying potential druggable genes and miRNAs related to the process of epithelial mesenchymal plasticity and exhibiting deregulated expression levels in different malignancies has been set forth. Moreover, the data presented in the present study justify the affordability of identifying common druggable cancer biomarkers applicable to various tumors, in order to proceed thereafter, through the pharmacological assessment, to the development of successful anticancer therapeutics. Of note, HOXB9 and miR-137 were found to be deregulated in all types of malignancies that were analyzed. However, both were expressed in very low levels compared to other genes and/or miRNAs. Therefore, careful examination is required upon attempting to further clinically validate their usefulness and therapeutic applicability, as several DE genes and miRNAs related to EMT are also shared by different types of cancer that were analyzed herein. To this end, the regulatory network of miRNAs in EMT plasticity has been previously evaluated in breast cancer (62). Moreover, the EMT regulatory network includes a number of EMT-related transcription factors (e.g., the Snail and Zeb family) and epigenetic collaborative regulators (e.g., miR-34 for Snail and miR-200s for Zeb). Such interplay drives the well-orchestrated epithelial-mesenchymal transcriptional program, thus mediating the downstream biological effects (6). In the present study, the bioinformatics analysis focused on EMT-related miRNAs and genes dysregulated in various origin tumor patient samples and thus the findings obtained highlight only such conclusions. Whether or not there exists any functional involvement between the miRNAs and the genes identified, needs to be experimentally validated. It is interesting, however, to note that recent data highlight the yet unexplored role of miRNAs as regulators of Hox genes in hematopoiesis, through the elucidation of the role of miR-708 as a novel regulator of the Hoxa9 program in leukemia myeloid cells (63).

Overall, the analysis approach, is further strengthening the previously published data regarding the modulation of EMT in tumors and proposes that targeted research efforts focused on identifying common biomarkers could provide effective anticancer drugs. The results obtained support the notion that as such, druggable biomarkers could be considered the HOXB9 gene and/or miR-137, irrespective of the cured tumor type, although further clinical and experimental studies are also needed. Importantly, however, such a direction is expected to provide valuable therapeutic interventions in malignancies by contributing toward overcoming the existed cellular and genomic heterogeneity (inter- and intra-tumoral) and the differential pharmacological response seen among patients. In particular, the data provided herein support the notion of identifying druggable biomarkers that impinge on the fundamental cancer cell traits that provide the needed advantageous capacity of tumor cell metabolism to abrogate the molecular balance, as well as the existing physiological restriction signals between differentiation, apoptosis and proliferation. This notion of the pan-cancer clinical intervention and the implementation of informed clinical decisions are based on the profiling of genomic signatures and molecular biomarkers in cancer patients; the origin and type of histology of the tumor have already begun to be left. Indeed, the development of therapeutics showing pan-cancer capabilities present a new revolutionary therapeutic era, an approach mentioned as ‘tumor-agnostic therapies’. Complementary to this, the ability to identify and clinically validate cancer biomarkers working irrespectively of the tumor type, permits the implementation of personalized cancer therapy in the clinical setting. The already marketed anticancer drugs, pembrolizumab, larotrectinib and entrectinib, belong to the class of tumor-agnostic therapies, by successfully receiving approval and being clinical used in patients with various types of tumor bearing common molecular features (64). Furthermore, the ability to implement cancer therapy with pharmacogenomics-guided therapeutic decisions offers the needed precision in clinical practice for the practical utilization of molecular profiling and biomarkers, as well as the outcome improvement in patients (65).

 Acknowledgements

The statistical methods used in the present study were reviewed by Angelis Eleftherios, Professor of Statistics and Information Systems, Head of the School of Informatics, Faculty of Sciences, Aristotle University of Thessaloniki, Greece.

 Funding

The present study was funded in the context of the project ‘Molecular signatures analysis of three-dimensional cell cultures and circulating tumor cells in the treatment of cancer’ (MIS 5004622) under the call for proposals ‘Supporting researchers with emphasis on new researchers’ (EDULLL 34). The project was co-financed by Greece and the European Union (European Social Fund-ESF) by the Operational Programme Human Resources Development, Education and Lifelong Learning 2014-2020.

 Availability of data and materials

All data generated or analyzed during this study are included in this published article or are available from the corresponding author on reasonable request.

 Authors' contributions

KAK was involved in the acquisition of data, analysis and interpretation of data, drafting the article and providing final approval. MGA was involved in the interpretation of the data, cell culture experiments and providing final approval. LPNG was involved in the interpretation of the data, drafting the article and providing final approval. NGG was involved in the interpretation of the data, revising the article and providing final approval. ISV conceived and designed the study, drafting the article, critical revision and provided final approval.

 Ethics approval and consent to participate

Not applicable.

 Patient consent for publication

Not applicable.

 Competing interests

NGG declares ownership of Biogenea Pharmaceuticals Ltd. The company had no role in the design or outcomes of the study. The remaining authors declare that they have no competing interests.

 References BocciFGearhart-SernaLBoaretoMRibeiroMBen-JacobEDeviGRLevineHOnuchicJNJollyMKToward understanding cancer stem cell heterogeneity in the tumor microenvironmentProc Natl Acad Sci USA11614815720193058758910.1073/pnas.1815345116 VizirianakisISMiliotouANMystridisGAAndriotisEGAndreadisIIPapadopoulouLCFatourosDGTackling pharmacological response heterogeneity by PBPK modeling to advance precision medicine productivity of nanotechnology and genomics therapeuticsExpert Rev Precis Med Drug Dev41391512019 BrabletzTKalluriRNietoMAWeinbergRAEMT in cancerNat Rev Cancer1812813420182932643010.1038/nrc.2017.118 RocheJThe epithelial-to-mesenchymal transition in cancerCancers (Basel)105220182946290610.3390/cancers10020052 CuligZEpithelial mesenchymal transition and resistance in endocrine-related cancersBiochim Biophys Acta Mol Cell Res18661368137520193110811710.1016/j.bbamcr.2019.05.003 LuWKangYEpithelial-mesenchymal plasticity in cancer progression and metastasisDev Cell4936137420193106375510.1016/j.devcel.2019.04.010 WilliamsEDGaoDRedfernAThompsonEWControversies around epithelial-mesenchymal plasticity in cancer metastasisNat Rev Cancer1971673220193166671610.1038/s41568-019-0213-x KyrodimouMAndreadisDDrougouAAmanatiadouEPAngelisLBarbatisCEpivatianosAVizirianakisISDesmoglein-3/γ-catenin and E-cadherin/ß-catenin differential expression in oral leukoplakia and squamous cell carcinomaClin Oral Investig1819921020142343033910.1007/s00784-013-0937-z Cancer Genome Atlas Research NetworkWeinsteinJNCollissonEAMillsGBShawKROzenbergerBAEllrottKShmulevichISanderCStuartJMThe cancer genome atlas pan-cancer analysis projectNat Genet451113112020132407184910.1038/ng.2764 ZhaoMKongLLiuYQuHdbEMT: An epithelial-mesenchymal transition associated gene resourceSci Rep51145920152609946810.1038/srep11459 ColapricoASilvaTCOlsenCGarofanoLCavaCGaroliniDSabedotTSMaltaTMPagnottaSMCastiglioniITCGAbiolinks: An R/Bioconductor package for integrative analysis of TCGA dataNucleic Acids Res44e7120162670497310.1093/nar/gkv1507 AndersSPylPTHuberWHTSeq-a Python framework to work with high-throughput sequencing dataBioinformatics3116616920152526070010.1093/bioinformatics/btu638 LoveMIHuberWAndersSModerated estimation of fold change and dispersion for RNA-seq data with DESeq2Genome Biol1555020142551628110.1186/s13059-014-0550-8 ShafferJPMultiple hypothesis testingAnnu Rev Psychol465615841995 RitchieMEPhipsonBWuDHuYLawCWShiWSmythGKlimma powers differential expression analyses for RNA-sequencing and microarray studiesNucleic Acids Res43e4720152560579210.1093/nar/gkv007 TherneauTMGrambschPMModeling survival data: Extending the Cox model. Springer-Verlag New York, 2000. TherneauTA package for survival analysis in RR package version3272020 KassambaraAKosinskiMBiecekPsurvminer: Drawing survival curves using‘ggplot2’, 2017. TseligkaEDRovaAAmanatiadouEPCalabreseGTsibouklisJFatourosDGVizirianakisISPharmacological development of target-specific delocalized lipophilic cation-functionalized carboranes for cancer therapyPharm Res331945195820162711587810.1007/s11095-016-1930-4 AkrivouMGDemertzidouVPTheodoroulaNFChatzopoulouFMKyritsisKAGrigoriadisNZografosALVizirianakisISUncovering the pharmacological response of novel sesquiterpene derivatives that differentially alter gene expression and modulate the cell cycle in cancer cellsInt J Oncol532167217920183022658610.3892/ijo.2018.4550 LivakKJSchmittgenTDAnalysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) methodMethods2540240820011184660910.1006/meth.2001.1262 UhlenMZhangCLeeSSjöstedtEFagerbergLBidkhoriGBenfeitasRArifMLiuZEdforsFA pathology atlas of the human cancer transcriptomeScience357eaan250720172881891610.1126/science.aan2507 LiangHFuZJiangXWangNWangFWangXZhangSWangYYanXGuanWXmiR-16 promotes the apoptosis of human cancer cells by targeting FEATBMC Cancer1544820152603177510.1186/s12885-015-1458-8 QuYLiuHLvXLiuYWangXZhangMZhangXLiYLouQLiSLiHMicroRNA-16-5p overexpression suppresses proliferation and invasion as well as triggers apoptosis by targeting VEGFA expression in breast carcinomaOncotarget8724007241020172906979710.18632/oncotarget.20398 RuanLQianXMiR-16-5p inhibits breast cancer by reducing AKT3 to restrain NF-κB pathwayBiosci Rep39BSR2019161120193138378310.1042/BSR20191611 NilssonSMöllerCJirströmKLeeABuschSLambRLandbergGDownregulation of miR-92a is associated with aggressive breast cancer features and increased tumour macrophage infiltrationPLoS One7e3605120122256343810.1371/journal.pone.0036051 ZearoSKimEZhuYZhaoJTSidhuSBRobinsonBGSoonPSMicroRNA-484 is more highly expressed in serum of early breast cancer patients compared to healthy volunteersBMC Cancer1420020142464180110.1186/1471-2407-14-200 VoliniaSCroceCMPrognostic microRNA/mRNA signature from the integrated analysis of patients with invasive breast cancerProc Natl Acad Sci USA1107413741720132358984910.1073/pnas.1304977110 YeFGSongCGCaoZGXiaCChenDNChenLLiSQiaoFLingHYaoLCytidine deaminase axis modulated by miR-484 differentially regulates cell proliferation and chemoresistance in breast cancerCancer Res751504151520152564369610.1158/0008-5472.CAN-14-2341 ThompsonEWReichRShimaTBAlbiniAGrafJMartinGRDicksonRBLippmanMEDifferential regulation of growth and invasiveness of MCF-7 breast cancer cells by antiestrogensCancer Res486764676819882846159 GjerdrumCTironCHøibyTStefanssonIHaugenHSandalTCollettKLiSMcCormackEGjertsenBTAxl is an essential epithelial-to-mesenchymal transition-induced regulator of breast cancer metastasis and patient survivalProc Natl Acad Sci USA1071124112920102008064510.1073/pnas.0909333107 TheodossiouTAAliMGrigalaviciusMGrallertBDillardPSchinkKOOlsenCEWälchliSInderbergEMKubinASimultaneous defeat of MCF7 and MDA-MB-231 resistances by a hypericin PDT-tamoxifen hybrid therapyNPJ Breast Cancer51320193099319410.1038/s41523-019-0108-8 ShaLDongLLvLBaiLJiXHOXB9 promotes epithelial-to-mesenchymal transition via transforming growth factor-β1 pathway in hepatocellular carcinoma cellsClin Exp Med15556420152508102210.1007/s10238-014-0276-7 BhatlekarSFieldsJZBomanBMRole of HOX genes in stem cell differentiation and cancerStem Cells Int2018356949320183015486310.1155/2018/3569493 HayashidaTTakahashiFChibaNBrachtelETakahashiMGodin-HeymannNGrossKWVivancoMdWijendranVShiodaTHOXB9, a gene overexpressed in breast cancer, promotes tumorigenicity and lung metastasisProc Natl Acad Sci USA1071100110520102008056710.1073/pnas.0912710107 De BraekeleerEDouet-GuilbertNBasinkoALe BrisMJMorelFDe BraekeleerMHox gene dysregulation in acute myeloid leukemiaFutur Oncol1047549520142455945210.2217/fon.13.195 EohKJKimHJLeeJYNamEJKimSKimSWKimYTDysregulated expression of homeobox family genes may influence survival outcomes of patients with epithelial ovarian cancer: Analysis of data from the cancer genome atlasOncotarget8705797058520172905030310.18632/oncotarget.19771 SongJWangTXuWWangPWanJWangYZhanJZhangHHOXB9 acetylation at K27 is responsible for its suppression of colon cancer progressionCancer Lett426637220182965488910.1016/j.canlet.2018.04.002 ZhanJNiuMWangPZhuXLiSSongJHeHWangYXueLFangWZhangHElevated HOXB9 expression promotes differentiation and predicts a favourable outcome in colon adenocarcinoma patientsBr J Cancer11188389320142502596110.1038/bjc.2014.387 ChibaNComaillsVShiotaniBTakahashiFShimadaTTajimaKWinokurDHayashidaTWillersHBrachtelEHomeobox B9 induces epithelial-to-mesenchymal transition-associated radioresistance by accelerating DNA damage responsesProc Natl Acad Sci USA1092760276520122193094010.1073/pnas.1018867108 CarboneCPiroGSimionatoFLigorioFCremoliniCLoupakisFAlìGRossiniDMerzVSantoroRHomeobox B9 mediates resistance to anti-VEGF therapy in colorectal cancer patientsClin Cancer Res234312432220172829854510.1158/1078-0432.CCR-16-3153 ZhangGFanEYueGZhongQShuaiYWuMFengGChenQGouXFive genes as a novel signature for predicting the prognosis of patients with laryngeal cancer. J Cell Biochem, Oct 31, 2019 (Online ahead of print). XuHWuSShenXWuDQinZWangHChenXSunXSilencing of HOXB9 suppresses cellular proliferation, angiogenesis, migration and invasion of prostate cancer cellsJ Biosci4540202032098919 FuDHeCWeiJZhangZLuoYTanHRenCPGK1 is a potential survival biomarker and invasion promoter by regulating the HIF-1α-mediated epithelial-mesenchymal transition process in breast cancerCell Physiol Biochem512434244420183053774410.1159/000495900 HeYLuoYZhangDWangXZhangPLiHEjazSLiangSPGK1-mediated cancer progression and drug resistanceAm J Cancer Res922802302201931815035 SambriICapassoRPucciPPernaAFIngrossoDThe microRNA 15a/16-1 cluster down-regulates protein repair isoaspartyl methyltransferase in hepatoma cells: Implications for apoptosis regulationJ Biol Chem286436904370020112203392110.1074/jbc.M111.290437 DongLLiYXueDLiuYPCMT1 is an unfavorable predictor and functions as an oncogene in bladder cancerIUBMB Life7029129920182951783910.1002/iub.1717 SchossererMMinoisNAngererTBAmringMDellagoHHarreitherECalle-PerezAPircherAGerstlMPPfeifenbergerSMethylation of ribosomal RNA by NSUN5 is a conserved mechanism modulating organismal lifespanNat Commun6615820152563575310.1038/ncomms7158 HeissenbergerCLiendlLNagelreiterFGonskikhYYangGStelzerEMKrammerTLMicutkovaLVogtSKreilDPLoss of the ribosomal RNA methyltransferase NSUN5 impairs global protein synthesis and normal growthNucleic Acids Res47118071182520193172242710.1093/nar/gkz1043 de MeloISIglesiasCBenítez-RondánAMedinaFMartínez-BarberáJPBolívarJNOA36/ZNF330 is a conserved cystein-rich protein with proapoptotic activity in human cellsBiochim Biophys Acta17931876188520091989585310.1016/j.bbamcr.2009.10.011 WillisSSunYAbramovitzMFeiTYoungBLinXNiMAchuaJReganMMGrayKPHigh expression of FGD3, a putative regulator of cell morphology and motility, is prognostic of favorable outcome in multiple cancers. JCO Precis Oncol 1: PO.17.00009, 2017. RendaIBianchiSVezzosiVNoriJVanziETavellaKSusiniTExpression of FGD3 gene as prognostic factor in young breast cancer patientsSci Rep91520420193164562410.1038/s41598-019-51766-w VidigalJAVenturaAThe biological functions of miRNAs: Lessons from in vivo studiesTrends Cell Biol2513714720152548434710.1016/j.tcb.2014.11.004 PengYCroceCMThe role of microRNAs in human cancerSignal Transduct Target Ther11500420162926389110.1038/sigtrans.2015.4 BartelDPMetazoan MicroRNAsCell173205120182957099410.1016/j.cell.2018.03.006 MichaelIPSaghafiniaSHanahanDA set of microRNAs coordinately controls tumorigenesis, invasion, and metastasisProc Natl Acad Sci USA116241842419520193170476710.1073/pnas.1913307116 CheroneJMJorgjiVBurgeCBCotargeting among microRNAs in the brainGenome Res291791180420193164905610.1101/gr.249201.119 DongJXiaoDZhaoZRenPLiCHuYShiJSuHWangLLiuHEpigenetic silencing of microRNA-137 enhances ASCT2 expression and tumor glutamine metabolismOncogenesis6e35620172869203210.1038/oncsis.2017.59 ZhangZLiuRShuaiYHuangYJinRWangXLuoJASCT2 (SLC1A5)-dependent glutamine uptake is involved in the progression of head and neck squamous cell carcinomaBr J Cancer122829320203181917810.1038/s41416-019-0637-9 WuPXiaoYGuoTWangYLiaoSChenLLiuZIdentifying miRNA-mRNA pairs and novel miRNAs from hepatocelluar carcinoma mirnomes and TCGA databaseJ Cancer102552255920193125876110.7150/jca.28167 NuzzoSCatuognoSCapuozzoMFiorelliASwiderskiPBoccellaSde NigrisFEspositoCLAxl-targeted delivery of the oncosuppressor miR-137 in non-small-cell lung cancerMol Ther Nucleic Acids1725626320193127695610.1016/j.omtn.2019.06.002 Drago-GarcíaDEspinal-EnríquezJHernández-LemusENetwork analysis of EMT and MET micro-RNA regulation in breast cancerSci Rep71353420172905156410.1038/s41598-017-13903-1 SchneiderEPochertNRuessCMacPheeLEscanoLMillerCKrowiorzKDelsing MalmbergEHeravi-MoussaviALorzadehAMicroRNA-708 is a novel regulator of the Hoxa9 program in myeloid cellsLeukemia341253126520203176801810.1038/s41375-019-0651-1 LooneyAMNawazKWebsterRMTumor-agnostic therapiesNat Rev Drug Discov1938338420203249404710.1038/d41573-020-00015-1 AstrasGPapagiannopoulosCIKyritsisKAMarkitaniCVizirianakisISPharmacogenomic testing to guide personalized cancer medicine decisions in private oncology practice: A case study. FrontOncol1052120203241159210.3389/fonc.2020.00521 GuZEilsRSchlesnerMComplex heatmaps reveal patterns and correlations in multidimensional genomic dataBioinformatics322847284920162720794310.1093/bioinformatics/btw313

PCA of all samples from (A) breast cancer, (B) head and neck cancer, (C) prostate cancer and (D) glioblastoma using the expression values of differentially expressed genes. PCA was performed and scatterplots were created using DESeq2(13). PCA, principal components analysis; TCGA, The Cancer Genome Atlas; BRCA, breast cancer; HNSC, head and neck squamous cell carcinoma; PRAD, prostate adenocarcinoma; GBM, glioblastoma multiforme.

PCA of patient samples from (A) breast cancer, (B) head and neck cancer and (C) prostate cancer using the expression values of differentially expressed miRNAs (please see the ‘Materials and methods’ section). PCA was performed and scatterplots were created using DESeq2(13). differentially expressed genes. PCA was performed and scatterplots were created using DESeq2(13). PCA, principal components analysis; TCGA, The Cancer Genome Atlas; BRCA, breast cancer; HNSC, head and neck squamous cell carcinoma; PRAD, prostate adenocarcinoma.

Kaplan-Meier survival curves of the differentially expressed genes, PGK1, PCMT1, FGD3, HOXB9, NSUN5 and ZNF330, whose expression levels were significantly associated (P-value <0.001; non-parametric log-rank test) with a favorable or poor survival probability of patients with breast (BRCA), head and neck (HNSC), glioblastoma (GBM) or prostate (PRAD) cancer. PST samples were assigned into two separate groups depending on whether target expression of each sample is higher (high expression) or lower (low expression) than the median. Survival analysis was performed and plots were created using survival methods (16,17) and survminer (18). PST, primary solid tumor; (Time: Is shown in days).

Kaplan-Meier survival curves of differentially expressed miRNAs miR-16-1/-2, miR-92-1 and miR-484, whose expression levels are significantly associated (P-value ~0.1; non-parametric log-rank test) with favorable or poor survival probability of patients with breast (BRCA) or head and neck (HNSC) cancer. PST samples were assigned into two separate groups depending on whether target expression of each sample is higher (high expression) or lower (low expression) than the median. Survival analysis was performed and plots were created using survival methods (16,17) and survminer (18). PST, primary solid tumor; (Time: Is shown in days).

Heatmaps of expression for genes that are differentially expressed and EMT-associated in at least one of the following malignancies: (A) breast cancer, (B) head and neck cancer, (C) prostate cancer and/or (D) glioblastoma. Gene counts were transformed using variance-stabilizing transformation and scaled to z-score for each gene. Heatmaps were created using ComplexHeatmap (66). The number of genes reported for each malignancy heatmap, is available in Table I (differentially expressed genes reported by dbEMT 1.0). Gene names for each malignancy heatmap are also given in Table II (marked as ‘TRUE’ for each malignancy). EMT, epithelial-to-mesenchymal transition.

Heatmaps of expression for miRNAs that are differentially expressed and EMT-associated in at least one of the following malignancies: (A) breast cancer, (B) head and neck cancer and/or (C) prostate cancer. miRNA counts were transformed using variance-stabilizing transformation and scaled to z-score for each miRNA. Heatmaps were created using ComplexHeatmap (66). The number of miRNAs reported for each malignancy heatmap, is available in Table I (differentially expressed miRNAs reported by dbEMT 1.0). miRNA names for each malignancy heatmap are also given in Table III (marked as ‘TRUE’ for each malignancy). EMT, epithelial-to-mesenchymal transition.

Venn diagrams showing common and unique differentially expressed and EMT reported (A) genes and (B) miRNAs for all types of malignancies that were analyzed. Plots were created using limma (15). EMT, epithelial-to-mesenchymal transition; The Cancer Genome Atlas; BRCA, breast cancer; HNSC, head and neck squamous cell carcinoma; PRAD, prostate adenocarcinoma; GBM, glioblastoma multiforme.

Assessment of expression profiles of (A) EMT-related genes and (B) miR-21 by RT-qPCR analysis of the MCF-7 and MDA-MB-231 cell lines. EMT, epithelial-to-mesenchymal transition. The results represent the means ± SD (A) ***P<0.001; (B) no statistically significant difference was observed; Student's t-test).

Assessment of expression profiles of (A) EMT-related genes and (B) miR-21 by RT-qPCR analysis of the HSC-3 and HaCaT cell lines. EMT, epithelial-to-mesenchymal transition. The results shown represent the mean ± SD (no statistically significant difference is observed; Student's t-test).

Number of genes and miRNAs identified as differentially expressed and EMT-associated in different malignancies.

TCGA project ID Type of malignancy Number of DE genes (LFC >1 and FDR adjusted P-value <0.001) Number of DE Genes reported by dbEMT 1.0 Number of DE miRNAs (LFC >1 and FDR adjusted P-value <0.001) Number of DE miRNAs reported by dbEMT 1.0
TCGA-HNSC Head and neck squamous carcinoma 2,509 37 255 6
TCGA-BRCA Breast cancer 4,227 73 271 10
TCGA-PRAD Prostate cancer 1,049 11 110 4
TCGA-GBM Glioblastoma multiforme 2,350 44    

DE, differentially expressed; EMT, epithelial-to-mesenchymal transition.

Differentially expressed and EMT-associated genes in different malignancies.

Gene symbol Ensemble gene ID TCGA_BRCA TCGA_HNSC TCGA_PRAD TCGA_GBM
ZEB2 ENSG00000169554 TRUE FALSE FALSE FALSE
EGFR ENSG00000146648 TRUE FALSE FALSE TRUE
EPAS1 ENSG00000116016 TRUE FALSE FALSE FALSE
ERBB2 ENSG00000141736 TRUE FALSE FALSE FALSE
MET ENSG00000105976 TRUE FALSE FALSE FALSE
CDH2 ENSG00000170558 TRUE FALSE FALSE FALSE
KLF4 ENSG00000136826 TRUE FALSE FALSE FALSE
KLF6 ENSG00000067082 TRUE FALSE FALSE FALSE
WT1 ENSG00000184937 TRUE TRUE FALSE FALSE
LEF1 ENSG00000138795 TRUE FALSE FALSE FALSE
SIM2 ENSG00000159263 TRUE FALSE TRUE FALSE
FN1 ENSG00000115414 TRUE TRUE FALSE TRUE
EGR1 ENSG00000120738 TRUE FALSE FALSE FALSE
KIT ENSG00000157404 TRUE TRUE TRUE FALSE
BMP2 ENSG00000125845 TRUE FALSE FALSE FALSE
CAV1 ENSG00000105974 TRUE FALSE TRUE FALSE
PPARG ENSG00000132170 TRUE TRUE FALSE FALSE
TGFBR3 ENSG00000069702 TRUE TRUE TRUE FALSE
FOXA1 ENSG00000129514 TRUE FALSE FALSE FALSE
STAT5A ENSG00000126561 TRUE FALSE FALSE FALSE
GATA3 ENSG00000107485 TRUE FALSE TRUE FALSE
ANXA1 ENSG00000135046 TRUE TRUE FALSE TRUE
DDR2 ENSG00000162733 TRUE FALSE FALSE FALSE
FOXM1 ENSG00000111206 TRUE TRUE FALSE TRUE
HSPB1 ENSG00000106211 TRUE FALSE FALSE FALSE
VIM ENSG00000026025 TRUE FALSE FALSE TRUE
SMAD9 ENSG00000120693 TRUE FALSE FALSE FALSE
GSN ENSG00000148180 TRUE FALSE FALSE FALSE
CYR61 ENSG00000142871 TRUE FALSE FALSE FALSE
MST1R ENSG00000164078 TRUE FALSE FALSE FALSE
VCAN ENSG00000038427 TRUE FALSE FALSE FALSE
MYCN ENSG00000134323 TRUE FALSE FALSE FALSE
TCF21 ENSG00000118526 TRUE FALSE FALSE FALSE
HMGA2 ENSG00000149948 TRUE TRUE FALSE FALSE
CDKN2A ENSG00000147889 TRUE TRUE FALSE FALSE
MUC1 ENSG00000185499 TRUE FALSE FALSE FALSE
AURKA ENSG00000087586 TRUE TRUE FALSE TRUE
FGF2 ENSG00000138685 TRUE FALSE FALSE FALSE
MCAM ENSG00000076706 TRUE FALSE FALSE FALSE
PAX2 ENSG00000075891 TRUE FALSE TRUE FALSE
PTPN14 ENSG00000152104 TRUE FALSE FALSE FALSE
SHH ENSG00000164690 TRUE FALSE FALSE FALSE
SPP1 ENSG00000118785 TRUE TRUE FALSE FALSE
CDH13 ENSG00000140945 TRUE FALSE FALSE FALSE
VTN ENSG00000109072 TRUE FALSE FALSE FALSE
FBLN5 ENSG00000140092 TRUE FALSE FALSE FALSE
KRT19 ENSG00000171345 TRUE FALSE FALSE FALSE
EZH2 ENSG00000106462 TRUE FALSE FALSE TRUE
ECT2 ENSG00000114346 TRUE TRUE FALSE FALSE
PLAUR ENSG00000011422 TRUE FALSE FALSE FALSE
MMP9 ENSG00000100985 TRUE TRUE FALSE TRUE
PTN ENSG00000105894 TRUE FALSE FALSE FALSE
POSTN ENSG00000133110 TRUE TRUE FALSE TRUE
CXCL12 ENSG00000107562 TRUE FALSE FALSE FALSE
EPO ENSG00000130427 TRUE FALSE FALSE FALSE
FGF1 ENSG00000113578 TRUE FALSE FALSE FALSE
DLX4 ENSG00000108813 TRUE FALSE FALSE FALSE
MMP3 ENSG00000149968 TRUE TRUE FALSE FALSE
LEP ENSG00000174697 TRUE FALSE FALSE FALSE
PRSS8 ENSG00000052344 TRUE FALSE FALSE FALSE
MMP13 ENSG00000137745 TRUE TRUE FALSE FALSE
KL ENSG00000133116 TRUE FALSE FALSE FALSE
HOXB9 ENSG00000170689 TRUE TRUE TRUE TRUE
SLC39A6 ENSG00000141424 TRUE FALSE FALSE FALSE
SDC1 ENSG00000115884 TRUE FALSE FALSE FALSE
GIPC2 ENSG00000137960 TRUE FALSE FALSE FALSE
HMGB3 ENSG00000029993 TRUE FALSE FALSE FALSE
TMPRSS4 ENSG00000137648 TRUE FALSE FALSE FALSE
RGCC ENSG00000102760 TRUE FALSE FALSE FALSE
VWCE ENSG00000167992 TRUE FALSE FALSE FALSE
MUC2 ENSG00000198788 TRUE FALSE FALSE FALSE
KCNH1 ENSG00000143473 TRUE FALSE FALSE TRUE
HSPB2 ENSG00000170276 TRUE TRUE FALSE FALSE
TGFB1 ENSG00000105329 FALSE TRUE FALSE FALSE
SNAI2 ENSG00000019549 FALSE TRUE FALSE FALSE
EGF ENSG00000138798 FALSE TRUE FALSE FALSE
TNC ENSG00000041982 FALSE TRUE FALSE TRUE
ITGA6 ENSG00000091409 FALSE TRUE FALSE FALSE
PTHLH ENSG00000087494 FALSE TRUE FALSE FALSE
ITGA5 ENSG00000161638 FALSE TRUE FALSE TRUE
ROR2 ENSG00000169071 FALSE TRUE FALSE FALSE
CLDN4 ENSG00000189143 FALSE TRUE FALSE FALSE
HPGD ENSG00000164120 FALSE TRUE FALSE FALSE
GREM1 ENSG00000166923 FALSE TRUE FALSE FALSE
CLU ENSG00000120885 FALSE TRUE TRUE FALSE
LAMA1 ENSG00000101680 FALSE TRUE FALSE FALSE
PROM1 ENSG00000007062 FALSE TRUE FALSE FALSE
LOXL2 ENSG00000134013 FALSE TRUE FALSE TRUE
FSCN1 ENSG00000075618 FALSE TRUE FALSE FALSE
COL8A1 ENSG00000144810 FALSE TRUE FALSE TRUE
EIF5A2 ENSG00000163577 FALSE TRUE FALSE FALSE
PDPN ENSG00000162493 FALSE TRUE FALSE TRUE
TWIST1 ENSG00000122691 FALSE FALSE TRUE FALSE
PTGS2 ENSG00000073756 FALSE FALSE TRUE FALSE
FOXQ1 ENSG00000164379 FALSE FALSE TRUE FALSE
TP53 ENSG00000141510 FALSE FALSE FALSE TRUE
MAP2K1 ENSG00000169032 FALSE FALSE FALSE TRUE
PRKCE ENSG00000171132 FALSE FALSE FALSE TRUE
CD44 ENSG00000026508 FALSE FALSE FALSE TRUE
MMP2 ENSG00000087245 FALSE FALSE FALSE TRUE
YBX1 ENSG00000065978 FALSE FALSE FALSE TRUE
TGFB1I1 ENSG00000140682 FALSE FALSE FALSE TRUE
CXCR4 ENSG00000121966 FALSE FALSE FALSE TRUE
MDK ENSG00000110492 FALSE FALSE FALSE TRUE
MSN ENSG00000147065 FALSE FALSE FALSE TRUE
SIX1 ENSG00000126778 FALSE FALSE FALSE TRUE
S100A4 ENSG00000196154 FALSE FALSE FALSE TRUE
PAK1 ENSG00000149269 FALSE FALSE FALSE TRUE
IGFBP3 ENSG00000146674 FALSE FALSE FALSE TRUE
MMP14 ENSG00000157227 FALSE FALSE FALSE TRUE
ST14 ENSG00000149418 FALSE FALSE FALSE TRUE
MKL2 ENSG00000186260 FALSE FALSE FALSE TRUE
ETV4 ENSG00000175832 FALSE FALSE FALSE TRUE
WNT1 ENSG00000125084 FALSE FALSE FALSE TRUE
LOX ENSG00000113083 FALSE FALSE FALSE TRUE
LIMA1 ENSG00000050405 FALSE FALSE FALSE TRUE
GRIN1 ENSG00000176884 FALSE FALSE FALSE TRUE
LOXL3 ENSG00000115318 FALSE FALSE FALSE TRUE
VSNL1 ENSG00000163032 FALSE FALSE FALSE TRUE
IDH1 ENSG00000138413 FALSE FALSE FALSE TRUE
CAMK1D ENSG00000183049 FALSE FALSE FALSE TRUE
MGAT3 ENSG00000128268 FALSE FALSE FALSE TRUE
HAS2 ENSG00000170961 FALSE FALSE FALSE TRUE

EMT, epithelial-to-mesenchymal transition; The Cancer Genome Atlas; BRCA, breast cancer; HNSC, head and neck squamous cell carcinoma; PRAD, prostate adenocarcinoma; GBM, glioblastoma multiforme.

Differentially expressed and EMT-associated miRNAs in different malignancies.

miRNA TCGA_BRCA TCGA_HNSC TCGA_PRAD
hsa-mir-137 TRUE TRUE TRUE
hsa-mir-193a TRUE FALSE FALSE
hsa-mir-200a TRUE FALSE FALSE
hsa-mir-200b TRUE FALSE FALSE
hsa-mir-200c TRUE FALSE TRUE
hsa-mir-205 TRUE TRUE FALSE
hsa-mir-21 TRUE TRUE FALSE
hsa-mir-33a TRUE FALSE FALSE
hsa-mir-9-1 TRUE TRUE FALSE
hsa-mir-429 TRUE FALSE FALSE
hsa-mir-30a FALSE TRUE FALSE
hsa-mir-34c FALSE TRUE FALSE
hsa-mir-221 FALSE FALSE TRUE
hsa-mir-222 FALSE FALSE TRUE

EMT, epithelial-to-mesenchymal transition; The Cancer Genome Atlas; BRCA, breast cancer; HNSC, head and neck squamous cell carcinoma; PRAD, prostate adenocarcinoma; GBM, glioblastoma multiforme.