Journal of Clinical Medicine Article A Novel Saliva-Based miRNA Signature for Colorectal Cancer Diagnosis Óscar Rapado-González 1,2,3 , Blanca Majem 4,5 , Ana Álvarez-Castro 6 , Roberto Díaz-Peña 2,7 , Alicia Abalo 2 , Leticia Suárez-Cabrera 4 , Antonio Gil-Moreno 3,8 , Anna Santamaría 4 , Rafael López-López 3,9 , Laura Muinelo-Romay 2,3, * and María Mercedes Suarez-Cunqueiro 1,10, * 1 Department of Surgery and Medical-Surgical Specialties, Medicine and Dentistry School, Universidade de Santiago de Compostela (USC), 15782 Santiago de Compostela, Spain; oscar.rapado@rai.usc.es 2 Liquid Biopsy Analysis Unit, Translational Medical Oncology (Oncomet), Health Research Institute of Santiago (IDIS), 15706 Santiago de Compostela, Spain; roberdp78@gmail.com (R.D.-P.); Alicia.Abalo.Pineiro@sergas.es (A.A.) 3 Centro de Investigación Biomédica en Red de Cáncer (CIBERONC), Instituto de Salud Carlos III, 28029 Madrid, Spain; antonioimma@yahoo.es (A.G.-M.); rafa.lopez.lopez@gmail.com (R.L.-L.) 4 Cell Cycle and Cancer Laboratory, Group of Biomedical Research in Urology, Vall Hebron Research Institute (VHIR), Universitat Autònoma de Barcelona (UAB), 08035 Barcelona, Spain; b.majem@gmail.com (B.M.); leticia.suarez@vhir.org (L.S.-C.); anna.santamaria@vhir.org (A.S.) 5 Cell Death Regulation Group, Oncobell Program, Bellvitge Biomedical Research Institute (IDIBELL), L’Hospitalet de Llobregat, 08908 Barcelona, Spain 6 Medical Digestive Service, Complexo Hospitalario Universitario de Santiago de Compostela (SERGAS), 15706 Santiago de Compostela, Spain; anacastro189@gmail.com 7 Faculty of Health Sciences, Universidad Autónoma de Chile, Talca 3460000, Chile 8 Unit of Gynecologic Oncology, Department of Obstetrics and Gynecology, Hospital Universitario Vall d’Hebron, 08035 Barcelona, Spain 9 Translational Medical Oncology (Oncomet), Health Research Institute of Santiago (IDIS), Complexo Hospitalario Universitario de Santiago de Compostela (SERGAS), 15706 Santiago de Compostela, Spain 10 Oral Sciences, Health Research Institute of Santiago de Compostela (IDIS), 15706 Santiago de Compostela, Spain * Correspondence: lmuirom@gmail.com (L.M.-R.); mariamercedes.suarez@usc.es (M.M.S.-C.); Tel.: +34-981-955-073 (L.M.-R.); +34-881-812-437 (M.M.S.-C.)  Received: 24 October 2019; Accepted: 18 November 2019; Published: 20 November 2019  Abstract: Salivary microRNAs (miRNAs) are of high interest as diagnostic biomarkers for non-oral cancer. However, little is known about their value for colorectal cancer (CRC) detection. Our study aims to characterize salivary miRNAs in order to identify non-invasive markers for CRC diagnosis. The screening of 754 miRNAs was performed in saliva samples from 14 CRC and 10 healthy controls. The differential expressed miRNAs were validated by RT-qPCR in 51 CRC, 19 adenomas and 37 healthy controls. Receiver operating characteristic (ROC) curves and logistic regression models were performed to analyze the clinical value of these miRNAs. Twenty-two salivary miRNAs were significantly deregulated in CRC patients vs. healthy individuals (p < 0.05) in the discovery phase. From those, five upregulated miRNAs (miR-186-5p, miR-29a-3p, miR-29c-3p, miR-766-3p, and miR-491-5p) were confirmed to be significantly higher in the CRC vs. healthy group (p < 0.05). This five-miRNA signature showed diagnostic value (72% sensitivity, 66.67% specificity, AUC = 0.754) to detect CRC, which was even higher in combination with carcinoembryonic antigen (CEA) levels. Overall, after the first global characterization of salivary miRNAs in CRC, a five-miRNA panel was identified as a promising tool to diagnose this malignancy, representing a novel approach to detect cancer-associated epigenetic alterations using a non-invasive strategy. J. Clin. Med. 2019, 8, 2029; doi:10.3390/jcm8122029 www.mdpi.com/journal/jcm J. Clin. Med. 2019, 8, 2029 2 of 13 Keywords: liquid biopsy; saliva; colorectal cancer; microRNA; circulating biomarkers; epigenomics 1. Introduction Colorectal cancer (CRC) is the third most commonly diagnosed cancer in men and the second in women worldwide. Regarding mortality data, it represents the fourth and the third cause of cancer death in men and women, respectively [1]. In 2030, the global burden CRC is expected to increase by 60% to over 2.2 million new cases and 1.1 million deaths [2]. Approximately 50% of CRC patients present metastasis at diagnosis or develop it during the course of the disease [3]. Currently, CRC diagnostic procedures are invasive or show low sensitivity and specificity, and thus research efforts are directed towards the search for less invasive biomarkers with higher accuracy to detect patients. Recently, a growing number of researchers have reported the potential of microRNAs (miRNAs) as diagnostic biomarkers in CRC [4,5]. MiRNAs are small, noncoding RNA molecules of 18–22 nucleotides that act as regulators of gene expression, interacting mainly with messenger RNA. MiRNAs expression can be deregulated in the pathogenesis and development of human cancer, and they can act as oncogenes or tumor suppressors [6]. Several miRNAs are released by cells [7] in physiological and pathological conditions into the extracellular space and can then be incorporated, free or encapsulated, into the bloodstream and other body fluids, such as saliva, urine, and cerebrospinal fluid [8], as “circulating miRNAs”—they have high stability due to their small size and their association with extracellular vesicles and RNA-binding proteins, such as argonaute-2 or high-density lipoproteins [9]. Due to their unique characteristics, miRNAs detected in liquid biopsies are promising biomarkers for cancer diagnosis, with a clear potential application for patients management in daily clinical practice. To date, several authors have described circulating miRNAs in blood with high diagnostic value and prognostic impact in CRC [10]. Ongoing scientific evidence in the salivary research field reinforces the interest of this biofluid as a source of potential biomarkers, identified in different salivary ‘-omics’ including the genome, the microbiome, the epigenome, the transcriptome, the proteome and the metabolome [11]. In this sense, saliva presents many advantages as a diagnostic tool over blood or tissue. The sample collection is even less invasive than blood and easy to store and transport [12]. In addition, saliva can be collected in shorter intervals than other diagnostic samples in order to provide updated tracking and patient follow-up without involving any discomfort for the patient. Aberrant expression profiles of salivary miRNAs have been detected in different types of cancer, showing their power as a discriminatory clinical method [13]. Unlike other tumors, only salivary miR-21 was analyzed in CRC by Sazanov et al. [14], finding miR-21 expression significantly higher in the saliva of CRC patients compared to healthy controls. Based on these findings, we have analyzed miRNAs profiles in a large cohort of saliva samples from CRC patients, adenomas and healthy individuals. With this approach, we were able to identify and validate salivary miRNAs with discriminatory value in detecting CRC. 2. Experimental Section 2.1. Study Subjects and Sample Collection This study was approved by the Galician Ethics Committee of Clinical Research (Ref. No. 2013/371) and written informed consent was obtained from all participants involved in this study at the time of recruitment. The methods were performed in accordance with relevant guidelines and regulations, following the recommendations of the PRoBE design [15]. This study complied the principles outlined in the Declaration of Helsinki [16]. One hundred and seven saliva samples from 51 CRC patients, 19 adenomas and 37 healthy individuals were collected from 2014 to 2018 at Oncology and Digestive Services from Complexo J. Clin. Med. 2019, 8, 2029 3 of 13 Hospitalario Universitario of Santiago de Compostela (SERGAS) in Galicia, Spain. All CRC and adenoma cases were confirmed by histopathological analysis of the corresponding tissue biopsy. Tumor stage in CRC patients was determined in accordance with the Tumor-Node-Metastasis (TNM)-based staging system promulgated by the American Joint Commission on Cancer (AJCC). None of the patients under study received chemotherapy or radiotherapy before sample collection. Eight patients underwent resective surgery before sample collection. Patients with a history of familial adenomatous polyposis, hereditary non-polyposis CRC, inflammatory bowel disease or diagnosis of other cancer in the last 5 years were excluded from this study. This study was designed as two phases: a discovery phase including 24 saliva samples (10 healthy controls and 14 CRC, including 7 non-metastatic and 7 metastatic patients) and a validation phase comprising 107 saliva samples (37 healthy controls, 19 adenomas and 51 CRC), including those used for the discovery phase (Figure S1). Clinicopathologic characteristics and lifestyle habits are summarized in Table S1. The detailed saliva collection and processing can be found in Supplementary Methods. 2.2. Saliva RNA Extraction Total RNA containing miRNAs was extracted from 500 µL of saliva using miRNeasy extraction Micro Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol with some modifications (Supplementary Methods). The concentration and quality of total RNA were analyzed using the Agilent 2100 Bioanalyzer and Quantus™ Fluorometer (Promega). 2.3. Salivary MiRNA Discovery Profiling with TaqMan Low-Density Arrays MiRNA expression profiling was performed using TaqMan® Low-Density Human MicroRNA Arrays (TLDA, Applied Biosystems, Foster City, California, USA). Each TLDA card detects 384 probes including 377 human miRNAs, 3 endogenous control assays and 1 negative control assay. Together, the 2 array cards (A and B Cards v3.0) can detect 754 mature miRNAs present in miRBase v14. This miRNA microarray assay was performed using 3 µL (1–350 ng) of total RNA from each sample, according to the manufacture’s protocol (Supplementary Methods). 2.4. MiRNA Expression Analysis Using RT-qPCR Expression levels of candidate miRNAs were validated by miRNA TaqMan RT-qPCR assays (Applied Biosystems) according to the manufacturer´s protocol (Supplementary Methods). Relative gene expression was determined by the 2-∆∆Ct method [17] and normalized with miR-193b-3p, which showed high stability among our samples. The selection of this internal control for quantification of salivary miRNA was described in Supplementary Methods (Table S2 and Figure S2). 2.5. Functional Enrichment Analysis Pathway enrichment analysis was performed using the DNA Intelligent Analysis (DIANA)- miRPath v3.0 software [18] to determine the miRNA regulatory roles and to identify the significant Kyoto Encyclopedia of Genes and Genomes (KEGG) molecular pathways. The algorithm miRTarBase v7.0 was used to identify the validated targets miR-186-5p, miR-29a-3p, miR-29c-3p, miR-766-3p and miR-491-5p. We selected the “pathways union” option of the miRPath software. The threshold (p < 0.05) and the false discovery rate (FDR) (p < 0.05) were calculated by Fisher´s exact test. 2.6. Statistical Analysis All statistical analyses with IBM SPSS Statistics 20 and graphs were generated using GraphPad Prism 5.0 (GraphPad Software, Inc., San Diego, California, USA). Hierarchical clustering was performed using the R environment (http://www.r-project.org/) and a heatmap was generated using a function of heatmap.2 in gplots package. Two-tailed Mann–Whitney U test or the Kruskal–Wallis test was used to evaluate the differential expression of salivary miRNAs. Multivariate logistic regression analyses J. Clin. Med. 2019, 8, 2029 4 of 13 were performed to establish the best miRNA panel to discriminate CRC from healthy individuals and those stage IV CRC with bad evolution of the disease (defined as progression or death within the 10 months after sample collection). The following formulas were generated after the logistic regression to provide a score for classifying patients vs. healthy controls and good/bad prognosis stage IV CRC patients: score for diagnosis = −9 − (1.4 * expressionmiR-29a-3p ) + (0.6 * expressionmiR1-86-5p ) + (2.1 * expressionmiR-29c-3p ) − (0.38 * expressionmiR-491-5p ) + (0.01 * expressionmiR-766-3p ); risk-score for prognosis = 1.09 + (5.4 * expressionmiR-29a-3p ) − (1.4*expressionmiR-186-5p ) − (5.7 * expressionmiR-29c-3p ) + (1.03 * expressionmiR-491-5p ) + (0.39 * expressionmiR-766-3p ). Receiver operating characteristic (ROC) curves were constructed, and area under the ROC curve (AUC) with 95% of confidence intervals (CIs) was obtained to evaluate the diagnostic accuracy of individual salivary miRNAs and the miRNA panel. Internal validation with the bootstrap method was used to adjust the overfitting. A total of 2000 random samples with replacement were generated. The AUC and the 95% CIs for the sensitivity and specificity were estimated using the pROC package in R software. Cut-off selection was made based on the value that provided the highest sensitivity and specificity to discriminate healthy controls vs. CRC patients and CRC patients with good vs. bad outcomes. For survival analyses Univariate/Multivariate Cox Regression was applied together with the Kaplan–Meier curves and the Log-Rank test. Spearman correlations were performed to determine the relationship among the different miRNAs and between each miRNA and the carcinoembryonic antigen (CEA) blood levels. p-value < 0.05 was set as the level of statistical significance. 3. Results 3.1. Salivary miRNA Expression Profiling Identified a Specific Expression Pattern in CRC During the discovery phase, 754 miRNAs were profiled in saliva samples from 14 CRC patients and 10 healthy controls. From all miRNAs included in the panel, 31 miRNAs (4.11%) were detected in all samples, 194 miRNAs (25.73%) were detected in >50% and 294 miRNAs (38.99%) were not present in any salivary sample. To detect those miRNAs characterizing CRC patients, we selected miRNAs present in at least 70% of patients, being significantly differentially expressed in patients (metastatic and/or non-metastatic) vs. healthy controls with a minimum fold change of 1.5 (up or down). Following these criteria, 22 miRNAs were found significantly deregulated in the saliva of patients compared to healthy controls. Expression levels of 20 miRNAs were found upregulated while only two were downregulated. The Venn diagram (Figure S3) shows differentially expressed miRNAs in cancer and controls taking into account the tumor stage. Thus, 18 miRNAs were found differentially expressed in non-metastatic patients, two in metastatic patients and six in both non-metastatic and metastatic patients. 3.2. A Panel of Salivary miRNAs Was Validated for CRC Ten miRNAs (miR-186-5p, miR-29a-3p, miR-766-3p, miR-16-5p, miR-195-5p, miR-140-3p, miR-142-5p, miR-29c-3p, miR-150-5p and miR-491-5p) were chosen from a total of the 22 differential miRNAs based on their fold change and their previous description as miRNAs expressed in CRC tumors (Figure 1). To verify whether the selected miRNAs were upregulated in saliva, we first examined their levels in a set of 30 CRC and 15 healthy individuals. Significant differences were found in seven miRNAs (miR-186-5p, miR-29a-3p, miR-766-3p, miR-16-5p, miR-29c-3p, miR-150-5p and miR-491-5p). These miRNAs were further validated in the full large-scale set of saliva samples from a total of 107 CRC, 19 adenomas and 37 healthy individuals. As shown in Figure 2, after the complete analysis, we found that salivary levels of miR-186-5p (p = 0.0136), miR-29a-3p (p = 0.0376), miR-29c-3p (p = 0.0112), miR-766-5p (p = 0.0381) and miR-491-5p (p = 0.0366) were statistically significantly different between the CRC and healthy groups. Salivary expression levels of miR-16-5p and miR-150-5p were higher in CRC than in healthy individuals, but no significant differences were obtained. In addition, the expression levels of validated miRNAs were also increased in patients with adenomas compared to healthy controls, however no significant differences were found. J. Clin. Med. 2019, 8, 2029 5 of 13 J. Clin. Med. 2019, 8, 2029 5 of 14 Figure 1. Hierarchical clustering Figure 1. Hierarchical heatmap clustering heatmapofof1010 miRNAs selected miRNAs selected forvalidation for the the validation phase with a > with a > phase 1.5-fold change1.5-folddifference (p < 0.05) change difference between (p < 0.05) betweenCRCCRC patients and patients and healthy healthy controls. controls. Each Each row row represents represents one sampleone andsample eachand each column column represents represents the expression the expression profile profile of of a singlemiRNA. a single miRNA. The The relative relative miRNA miRNA expression changes are expressed by three colors from red to yellow, as shown at the top, expression changes are expressed by three colors from red to yellow, as shown at the top, where red where red represents the high expression, orange medium expression, and yellow low expression. representsThe themiRNA high expression, clustering treeorange is shownmedium on the topexpression, and the sampleand yellowdendrogram clustering low expression. The miRNA is presented clustering on tree theisleft. shown S1–S12on the top to correspond and the sample salivary clustering samples of dendrogram CRC patients and SC1–SC10 is correspond presentedtoon the left. salivary samples S1–S12 correspond of healthy to salivary controls.ofNote: samples CRC miRNAs, patientsmicroRNAs; CRC, colorectal and SC1–SC10 cancer.to salivary samples correspond J. Clin. Med. 2019, 8, 2029 6 of 14 of healthy controls. Note: miRNAs, microRNAs; CRC, colorectal cancer. miR-186-5p miR-29a-3p miR-29c-3p Relative Expression (2-Δ Δ CT) Relative Expression (2-Δ Δ CT) Relative Expression (2-Δ Δ CT) 15 * 20 * 20 * 15 15 10 10 10 5 5 5 0 0 0 HC A CRC HC A CRC HC A CRC miR-766-3p miR-491-5p miR-150-5p Relative Expression (2-Δ Δ CT) Relative Expression (2-Δ Δ CT) Relative Expression (2-Δ Δ CT) 15 * 20 * 20 15 15 10 10 10 5 5 5 0 0 0 HC A CRC HC A CRC HC A CRC miR-16-5p Relative Expression (2-Δ Δ CT) 20 15 10 5 0 HC A CRC Figure 2. Salivary miRNA expression levels in healthy controls, adenomas and CRC patients in Figure 2. Salivary miRNA expression levels in healthy controls, adenomas and CRC patients in the the validation phase of the study. The relative expression levels of selected salivary miRNAs were validation phase of the study. The relative expression levels of selected salivary miRNAs were normalized to endogenous normalized control (miR-193b-3p). to endogenous The two-tailed control (miR-193b-3p). Mann–Whitney The two-tailed Mann–Whitney U test U was test performed was to examine the difference performed to examinebetween groups the difference of seven between miRNAs groups of seven (* p < 0.05). miRNAs (* p