ORIGINAL RESEARCH published: 03 November 2020 doi: 10.3389/fonc.2020.586268 Five microRNAs in Serum Are Able to Differentiate Breast Cancer Patients From Healthy Individuals Andrea Feliciano 1 , Lucila González 1 , Yoelsis Garcia-Mayea 1 , Cristina Mir 1 , Mireia Artola 2 , Nieves Barragán 2 , Remedios Martín 2 , Anna Altés 2 , Josep Castellvi 1 , Sergi Benavente 1 , Santiago Ramón y Cajal 1 , Martín Espinosa-Bravo 3 , Javier Cortés 4,5 , Isabel T. Rubio 6 and Matilde E. LLeonart 1,7* 1 Biomedical Research in Cancer Stem Cells Group, Vall d’Hebron Research Institute (VHIR), Universitat Autònoma de Barcelona, Barcelona, Spain, 2 Primary Care Center CAP-Vallcarca-Sant Gervasi, Barcelona, Spain, 3 Breast Pathology Unit, Vall d’Hebron University Hospital, Barcelona, Spain, 4 Institute of Breast Cancer, Quiron Group, Barcelona, Spain, 5 Vall d’Hebron Institute of Oncology (VHIO), Barcelona, Spain, 6 Breast Surgical Oncology, University of Navarra Clinic, Madrid, Spain, 7 Spanish Biomedical Research Network Center in Oncology, Madrid, Spain Breast cancer is the cancer with the most incidence and mortality in women. microRNAs are emerging as novel prognosis/diagnostic tools. Our aim was to identify a serum microRNA signature useful to predict cancer development. We focused on studying Edited by: the expression levels of 30 microRNAs in the serum of 96 breast cancer patients vs. Chris Albanese, 92 control individuals. Bioinformatic studies provide a microRNA signature, designated Georgetown University, United States as a predictor, based on the expression levels of five microRNAs. Then, we tested the Reviewed by: Hadi Parsian, predictor in a group of 60 randomly chosen women. Lastly, a proteomic study unveiled Babol University of Medical the overexpression and downregulation of proteins differently expressed in the serum of Sciences, Iran Krishna Beer Singh, breast cancer patients vs. that of control individuals. Twenty-six microRNAs differentiate University of Pittsburgh, United States cancer tissue from healthy tissue, and 16 microRNAs differentiate the serum of cancer *Correspondence: patients from that of the control group. The tissue expression of miR-99a, miR-497, Matilde E. LLeonart miR-362, and miR-1274, and the serum levels of miR-141 correlated with patient survival. matilde.lleonart@vhir.org Moreover, the predictor consisting of miR-125b, miR-29c, miR-16, miR-1260, and Specialty section: miR-451 was able to differentiate breast cancer patients from controls. The predictor was This article was submitted to validated in 20 new cases of breast cancer patients and tested in 60 volunteer women, Cancer Metabolism, a section of the journal assigning 11 out of 60 women to the cancer group. An association of low levels of miR-16 Frontiers in Oncology with a high content of CD44 protein in serum was found. Circulating microRNAs in serum Received: 22 July 2020 can represent biomarkers for cancer prediction. Their clinical relevance and the potential Accepted: 17 September 2020 Published: 03 November 2020 use of the predictor here described are discussed. Citation: Keywords: breast cancer, serum, microRNAs, prognosis, diagnosis Feliciano A, González L, Garcia-Mayea Y, Mir C, Artola M, Barragán N, Martín R, Altés A, INTRODUCTION Castellvi J, Benavente S, Ramón y Cajal S, Espinosa-Bravo M, Cortés J, Breast cancer is one of the most frequent carcinomas and the second leading cause of death in Rubio IT and LLeonart ME (2020) Five women (1). Specifically, in the United States and Europe, about 1 in 8 women (12.5%) will develop microRNAs in Serum Are Able to Differentiate Breast Cancer Patients invasive breast cancer over the course of their life. Therefore, comprehensive research should be From Healthy Individuals. devoted to cancer prevention in order to scale down these numbers and reach higher life expectancy Front. Oncol. 10:586268. in affected patients, lower mortality rates, and decline socio-economical burdens due to the high doi: 10.3389/fonc.2020.586268 cost of chemotherapeutical treatments. Frontiers in Oncology | www.frontiersin.org 1 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum Currently there is no precise model to estimate breast cancer that have had no cancer episodes in the past. For the validation risk. Most of the predictor models consider clinical factors, study, we had additional serum from 20 breast cancer patients. including the density of breast tissue, biopsy history, and several Finally, for the test study, we had serum from 60 volunteer clinical parameters. However, such models are not informative women where no selection criteria were applied. The pathological at an individual level. Predictive tests (i.e., Oncotype DX, and clinical characteristics of the patients include the presence Prosigna, MammaPrint) based on the status of genetic and non- of estrogen receptor (ER), progesterone receptor (PR), Ki-67 genetic factors in cancer tissue have proven their prognostic and expression, p53, tumor grade determined by tumor heterogeneity predictive ability in a personalized way (2). Currently, the liquid (low, medium, and high), tumor stage determined by the size biopsy is being used to establish the biomarkers that are able of the tumor and its infiltrating capacity to neighboring local to predict or envisage a potential future cancer development areas (T1b, T1c, or T2), subtype of breast cancer (molecular risk (3). classification), presence of metastasis, disease-free survival, and microRNAs are key factors in oncogenesis because they overall survival. All patients included in the study were recruited contribute to the modulation of key oncogenic and tumor from the Vall d’Hebron Hospital and selected for primary breast suppressor proteins. In particular, microRNA expression cancer. Patients were not treated with radio- or chemotherapy profiling can be used to classify human cancer (4). On the before sample collection. Control individuals were recruited from other hand, recent evidence suggests that microRNAs are very the Castilla-La Mancha Blood Bank and the Government of stable molecules in serum and that they have been established Catalonia Blood and Tissue Bank. Volunteer women came from as biomarkers for some cancer types (5). Interestingly, the the Primary Care Center (CAP-Vallcarca Sant Gervasi). The level of certain microRNAs in combination with known tumor study was conducted in accordance with the instructions and markers (e.g., CEA or CA15-3) improves sensitivity to breast requirements stated in the Declaration of Helsinki international cancer detection (6). Thereby routine monitoring of circulating standards for studies and approved by the Ethics Committee of microRNAs can result in significant benefits for the prognosis, Vall d’Hebron Hospital (CEIC). Informed consent was obtained diagnosis, and breast cancer treatment (7). from the patients to participate, analyze, and publish their data. Previously, we identified a molecular signature based on 35 microRNAs that vary significantly in normal tissue vs. Sample Collection cancer tissue in breast cancer patients (8). According to our Serum was collected from each patient prior surgery. Hemolytic previous work and literature search, we selected 30 cancer- sera (representing 5%) were discarded from the study. Summing related microRNAs that could be potentially detected in serum up, blood sample was obtained and centrifuged at 1,300 rpm for (miR-96, miR-451, miR-155, miR-195, miR-200c, miR-106b, 10 min and the supernatant fraction (serum) was collected and miR-141, miR-21, miR-486, miR-16, miR-125b, miR-99a, miR- stored at −80◦ C. The collection and pre-processing of the cancer 497, miR-191, miR-145, miR-100, miR-144, miR-382, miR-29c, samples vs. the healthy ones were treated with the same technical miR-10b, miR-133a, miR-1260, miR-1274a, miR-1274b, miR- conditions. Normal and tumor tissue were collected from the 133b, miR-92, miR-376c, miR-411, miR-299, and miR-215) (8– surgery room and stored at −80◦ C before RNA extraction. 11). In the present study, we compared the expression of 30 Hematoxylin and eosin staining of the slides from frozen microRNAs in tumor vs. normal tissue and serum from 96 breast biopsies was validated histologically to ensure that the tissue cancer patients (in comparison with control serum). Through area had an adequate tumor density (>80%). RNA was isolated statistical and bioinformatic studies, we determined a predictor, with a MirVana kit (Ambion R Life Technologies) according comprised by five microRNAs, that categorize an individual in to the manufacturer’s instructions. The RNA concentration the control group or breast cancer group. The potential benefit from tissue was quantified using the Nanodrop-2000 UV-Vis of this classifier and its validation for breast cancer prediction Spectometer (Fisher Scientific) and its quality was determined by is discussed. the Bioanalyser (RIN ratio> 8). On the other hand, to verify that in RNA extractions from sera, there was enough RNA to analyze the 30 microRNAs considered MATERIALS AND METHODS in this study, each sample was amplified using RNU and cel-miR- Patients and Controls 39-3p probes individually using quantitative real-time qRT-PCR This study comprises 96 breast cancer patients. For each patient, (data not shown). we had samples of cancer tissue (CANtum), normal tissue (CANnorm), and serum (CANse). For comparison purposes, we qRT-PCR had serum from 92 control individuals (CTLse). The method to The reverse transcription was performed on 10 ng of RNA select the control group established the following criteria: 20- using specific primers for the 30 selected microRNAs, including to 80-year-old women, non-smokers, non-drinkers, no evidence endogenous control RNU6 (ID 00973) and the exogenous control of breast cancer in their family history, and healthy women cel-miR-39-3p (ID 000200) with the TaqMan commercial kit microRNA Reverse Transcription Kit (Applied Biosystems, Life Technologies, CA, USA) as described (8). The pre-amplification Abbreviations: CANnorm, normal tissue; CANtum, cancer tissue; CANse, cancer serum; Ct, cycle threshold; CTLse, control individual serum; dCt, Ct reference reaction was carried out on 5 µl of cDNA product using gene – Ct gene of interest; Down, downregulated; FC, Fold change; qRT-PCR, a pool containing the specific preamplification primers for Quantitative Real-time PCR; RIN, RNA integrity number; Up, upregulated. each microRNA with the TaqMan R PreAmp Master Mix 2× Frontiers in Oncology | www.frontiersin.org 2 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum solution (Applied Biosystems, Life Technologies, CA, USA). Mann–Whitney U-test was used to identify microRNAs Reactions were performed in the VeritiTM Thermal Cycler Assays differently expressed between patients and controls. Benjamini– thermocycler (Applied Biosystems, Life Technologies, CA, USA). Hochberg’s false discovery rate (FDR) method was used to The study of each microRNA levels was conducted by triplicate. correct for multiple testing. The analysis to select the differently The references used for each microRNA are the following: expressed microRNAs has been based on the fitting of a mir-125b-5p (ID 000449), mir-99a-5p (ID 000435), mir-100-5p linear model. (ID 000437), mir-497-5p (ID 001043), mir-1274b (ID 002884), For the predictor, we considered that the best classification mir-106b-5p (ID 000442), mir-1260a (ID 002896), mir-141-3p method was CART (Classification and Regression Trees) (ID 000463), mir-96-5p (ID 000186), mir-21-5p (ID 000397), (16). The statistical analyses have been performed using miR-1274a (ID 002883), mir-145-5p (ID 002278), mir-299-5p ExpressionSuite (Life Technologies, CA, USA) (R version 3.5.1, (000600), mir-376c-3p (ID 002122), mir-451a (ID 001141), mir- copyright© 2018, Foundation for Statistical Computing, Vienna, 486-5p (ID 001278), cel-miR-39-3p (ID 000200), U6, miR- Austria) and the libraries developed for microRNA-target 16-5p (ID 000391), mir-195-5p (ID 000494), mir-191-5p (ID analysis by the Bioconductor Project (www.bioconductor.org). 002299), mir-215-5p (ID 000518), mir-382-5p (ID 000572), mir- Regarding the validation of the microRNA expression with the 411-5p (ID 001610), mir-10b-5p (ID 002218), mir-155-5p (ID pathological characteristics of the patients, ANOVA and t-test 002623), mir-200c-3p (ID 002300), mir-144-5p (ID 002148), methods were used (SPSS v9.3). A statistical analysis to determine mir-92a-3p (ID 000431), mir-133a-3p (ID 002246), mir-133b differential proteins and peptides was performed using DanteR (ID 002247), mir29c-3p (ID 000587), and miR-362 (ID478058). software (http://omics.pnl.gov/software/danter). p < 0.05 were Supplementary Tables 1, 2 show the raw qRT-PCR data for the considered significant. indicated microRNAs in tissue and serum samples, respectively. Supplementary Tables 3, 4 show the qRT-PCR results for the indicated microRNAs in tissue and serum samples respectively RESULTS upon normalization. The probes cel-miR-39-3p and RNU6 were used as internal controls, both to monitor the efficiency of RNA Tumor-Associated microRNAs in Breast isolation and subsequent retrotranscription and to normalize Cancer possible variations between samples during RNA isolation. For the selection of the microRNAs studied here, they were Although RNU6 is used as one of the most frequent endogenous selected: (a) the 17 most significantly deregulated microRNAs controls to study profiling microRNA in cell and tissue samples, in breast cancer based on our previous work (miR-21, miR-96, it is not a suitable endogenous control to study the expression miR-141, miR-1274a, miR-1260, miR-1274b, miR-106b, miR-299, of serum microRNAs (12). Therefore, in order to compare the miR-486, miR-376c, miR-497, miR-195, miR-100, miR-145, miR- data, results were normalized using the quantile method (using 99a, miR-451, and miR-125b) and (b) the potentially relevant the normalized CtData function of the R package HTqRTPCR) microRNAs in the serum of breast cancer patients (miR-155, including all patients per sample type. miR-200, miR-16, miR-191, miR-144, miR-382, miR-29c, miR- 10b, miR-133a, miR-133b, miR-92, miR-411, and miR-215) (8– Proteomic Study 11, 17). The following microRNAs were studied in serum and Serum from 70 breast cancer patients and 70 controls was cancer tissue in comparison with control individuals: miR-21, studied at protein level. Each sample was depleted individually miR-96, miR-141, miR-1274a, miR-1260, miR-106b, miR-1274b, using the PierceTM Abundant Protein Depletion Spin Columns miR-299, miR-376c, miR-497, miR-195, miR-100, miR-145, miR- kit (ref. 13434319, Thermo ScientificTM ) according to the 99a, miR-451, miR-125b miR-486, miR-16 (only serum), miR- manufacturer’s instructions. This kit eliminates ∼95% of 12 191, miR-215, miR-382, miR-411, miR-106, miR-155, miR-200c, abundant proteins in serum (α1-Acid Glycoprotein, Fibrinogen, miR-144, miR-92a, miR-133a, miR-133b, miR-29c, and miR-362 α1-Antitrypsin, Haptoglobin, α2-Macroglubulin, IgA, Albumin, (only tissue) (8). IgG, Apolipoprotein A-I, IgM, Apolipoprotein A-II, and Supplementary Table 5 shows the 26 microRNAs differently Transferrin), allowing the identification of other proteins in expressed when comparing tumor tissue with normal tissue the samples. The quantitative study of proteins was performed in 96 breast cancer patients and 92 control individuals (p < through Tandem Mass Tag marking as previously described 0.05). The volcano plot shows the most relevant microRNAs (13). Then, samples were grouped by pools (nine cancer pools (Figure 1A) (p < 0.01). Supplementary Table 6 shows that and nine control pools) for sequencing. Each pool (80 µg of 16 microRNAs (out of 30 initially selected) are significantly protein) was composed of equivalent amounts of seven samples deregulated when comparing the serum from cancer patients of each type (cancer or control). Sequencing was performed by vs. the serum from control individuals. The volcano plot shows quantitative liquid chromatography tandem mass spectrometry the top significant microRNAs (Figure 1B). The miR-125b and using an LTQ-Orbitrap XL instrument as described above (14). RNU6 levels were validated by another approach based on the manual performance of the Assays-on-Demand Taqman Statistical Analysis Gene Expression Assays according to the procedure previously The study has been conducted using Leave-One-Out Cross described (data not shown) (18). In order to check if the Validation (LOOCV) as cross-validation technique, thus microRNAs expressed in the tumor reflect the same trend in ensuring greater robustness in the results obtained (15). the serum of breast cancer patients, we compared significant Frontiers in Oncology | www.frontiersin.org 3 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum FIGURE 1 | MicroRNAs expression in breast cancer. (A) Volcano plots indicating the top deregulated microRNAs when cancer tissue vs. normal tissue was compared (green, downregulated microRNAs; red, upregulated microRNAs). (B) Volcano plots indicating the top deregulated microRNAs when the serum of cancer patients was compared with a group of control serum. (C) Left side: Eleven commonly deregulated microRNAs when tissue and serum samples were compared. Right side: Table showing the potential association between the expression of 11 microRNAs in tissue and serum. It can be observed that three microRNAs (miR-497, miR-133b, and miR-96) have a statistically significant correlation coefficient (R) for a 95% confidence interval (CI) (p < 0.05). As indicated in the table, the microRNA values of the cancer tissue are relativized to normal tissue and the microRNA values of the cancer sera are relativized to the control sera. Up, upregulated; Down, downregulated. microRNAs in the tumor tissue and serum in all patients. miR-62, and miR-1274a correlated with overall survival Eleven out of 16 significant microRNAs were deregulated in both (Figure 2A). In addition, miR-362 and miR-133b expression samples: tumor tissue of cancer biopsies and serum (Figure 1C, correlated with disease-free survival (Figure 2A). In addition, Supplementary Tables 5, 6). Three microRNAs, miR-191, miR- we found that high miR-141 expression in the serum 141, and miR-96, followed the same trend when the tumor and of breast cancer patients correlated with better survival serum of cancer patients were compared (Figure 1C). (Figure 2B). There is a lack of correlation regarding the studied microRNAs with the molecular classification of tumors (19). Pathological and Clinic Characteristics of the Tumors The pathological characteristics of the patients are shown Construction of a Predictor (Supplementary Table 7). Supplementary Figure 1 shows The experimental design of the study is summarized in Figure 3. the serum microRNAs that correlate with tumor stage. In order to establish a microRNA signature designated here as Supplementary Figure 2 shows the tumor microRNAs predictor, statistical and bioinformatic studies were performed that correlate with tumor grade. Supplementary Figure 3 in the serum from 92 control women and 96 breast cancer shows the tumor microRNAs that correlate with tumor patients. Accordingly, the minimal number of microRNAs able stage. We found that the expression of miR-99a, miR-497, to predict whether a serum sample should be categorized as Frontiers in Oncology | www.frontiersin.org 4 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum FIGURE 2 | Prognosis-related microRNAs. (A) Kaplan–Meier curves showing that miR-99a, miR-497, miR-362, and miR-1274a levels in the tissue sample correlate with survival and/or disease-free survival. Up, upregulated; Down, downregulated. (B) Kaplan–Meier curve showing that miR-141 levels in the serum correlate with survival. Low, low expression; High, high expression. control or cancer was reduced to five: miR-125b, miR-29c, miR- 86.25, 100, and 81.67%, respectively (Figure 3). The internal 16, miR-1260, and miR-451 (Figure 3). The proposed microRNA classification error was 5.45%. signature that derives exclusively from serum samples has the following percentages of accuracy, sensitivity, and specificity: Proteomic Study 90.43, 90.62, and 90.22%, respectively (Figure 3). The internal A total of 110 significantly deregulated proteins were found when classification error was 9.26%. comparing the serum of cancer patients vs. the serum of healthy Later on, in an external validation phase, the predictor was individuals (Supplementary Table 10). Thirty-five proteins were used to verify the status of the serum from 20 additional selected as the top differently expressed ones between cancer cancer patients plus 60 serum samples from a group of vs. normal serum using a fold change (FC) ratio above 1.2 or volunteer women taken randomly to be tested by the predictor. below 0.8 (Figure 4A). By using the multiMiR Bioconductor’s Supplementary Table 8 shows the raw qRT-PCR data for the package, microRNA–gene target interactions were explored (20). indicated microRNAs in serum samples. Supplementary Table 9 The search for validated targets was performed across miRecords, shows the qRT-PCR results for the indicated microRNAs in miRBase, and TarBase databases. A total of 3,947 validated serum samples upon normalization. All serum samples were unique target genes were found to the 16 microRNAs deregulated confirmed as cancer patients (Figure 3). Eleven out of 60 samples in serum (data not shown). CD44 protein (upregulated in the were classified as cancer patients (Figure 3). The percentages serum pools from breast cancer patients vs. the pools from the of accuracy, sensitivity, and specificity of this later study are control group patients) was found in the list of the 3,947 validated Frontiers in Oncology | www.frontiersin.org 5 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum FIGURE 3 | Flow chart of the analysis design in the present study. The expression change-based method pipeline is described (left). The interval validation phase provides 5 microRNAs revealed by the predictor (middle). The validation and test phase comprises 20 patients and apparently 60 healthy women (right). Indicated values of accuracy, sensitivity, and specificity are shown for each phase. targets. CD44 inversely correlates with miR-16 expression, which and downregulation of miR-145, miR-99a, and miR-125b in appears downregulated in the serum from cancer patients in breast cancer tissue (8, 21–24). Secondly, we found that 16 comparison with controls (Supplementary Table 10, Figure 4B). out of 30 microRNAs were significantly deregulated in the The 35 proteins were classified accordingly to their involvement serum of cancer patients vs. the serum of the control group. in different regulatory pathways (Figure 4C). Among them, Interestingly, in serum of breast cancer patients, downregulation CST3 (Cystatin C) seems to be involved in the modulation of of miR-411, miR-376c, miR-16, and miR-155 (9, 17) and different pathways (Figure 4D). upregulation of miR-125b, miR-1260, and miR-96 had been previously described, confirming the validation of our results (9, 17, 21, 25). Some of these 16 microRNAs have been DISCUSSION associated with breast cancer diagnosis including miR-125b, miR- 191, miR-411, miR-155, and miR-215 (26, 27). In particular, The final purpose of this research is to establish a microRNA 11 deregulated microRNAs were found in the serum and signature associated with breast cancer to determine molecular tissue of breast cancer patients (Figure 1C). Most of them evidence of cancer that will lead to future cancer development are contrarily overexpressed among both types of samples, in serum samples. Firstly, we found 26 microRNAs significantly that is, although we found 11 deregulated microRNAs that deregulated in the cancer vs. the healthy tissue from 96 are common to serum and tissue, their expression (either breast cancer patients. Our results corroborate previous studies upregulated or downregulated) was inversely correlated when showing upregulation of miR-96, miR-200c, and miR-141, comparing serum and tissue. The fact that the expression of a Frontiers in Oncology | www.frontiersin.org 6 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum FIGURE 4 | Proteins deregulated in the serum of breast cancer patients. (A) Proteins significantly deregulated in the serum pools of breast cancer patients vs. controls (upregulated proteins are indicated in red and those that are downregulated are indicated in green). (B) microRNAs and validated target proteins found in serum revealed miR-16 and CD44. FC, fold change. (C) Different molecular pathways involving significantly deregulated proteins. (D) Interactions of CST3 with other proteins which might have a relevant role in cancer. specific microRNA in different sample types can have inverse novel microRNAs associated with survival in breast cancer: miR- implications in prognosis/diagnosis, has already been described 497 and miR-362 (36). Apart from miR-362, miR-133b correlates as well as microRNA deregulation in the opposite direction with disease-free survival, the latter already been described when comparing their expression in serum vs. tissue (21, 28–30). as a diagnostic marker in breast cancer (28). Interestingly, For example, miR-125b, known to be downregulated in breast serum levels of miR-125b and miR-29c (the top 2 in order cancer tissue (8, 31), is upregulated in the bloodstream of breast of significance; Figure 1C) were associated to tumor stage. cancer patients (21, 25, 32). Possible explanations include (i) Moreover, high levels of miR-141 in serum were correlated with extracellular and cellular microRNAs profiles differ, and freely better survival. Contrary to our results, Debel et al. found that circulating microRNA might not reflect their abundance in miR-141 expression in serum was associated with shorter brain cancer cells (33); (ii) the total level of free microRNAs in the metastases (37). bloodstream might be masked by certain microRNAs present Lastly, despite the growing interest in assessing predictive into exosomes (34). cancer models based on microRNA signatures, most of the In relation with the use of microRNAs as biomarkers, it has reported studies need to be further evaluated in larger cohorts been suggested that the association of miR-99a in breast cancer of breast cancer patients (21, 24, 38). In this study, we identified tissue with survival differs depending on the molecular subtype a predictor (based on the following microRNAs: miR-125b, miR- (35). Our study corroborates the fact that high levels of miR- 29c, miR-16, miR-1260, and miR-451), capable of differentiating 1274a are associated with worse prognosis and proposes two the serum of breast cancer patients from that of control Frontiers in Oncology | www.frontiersin.org 7 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum individuals with ∼90% of accuracy, sensitivity, and specificity. identified. This suggests the prognosis value of CD44 protein in The fact that the predictor model includes microRNAs less serum as a potential marker of breast cancer. Collectively, our statistically significant such as miR-16 and miR-1260 than other results support the fact that microRNA detection in serum can more deregulated microRNAs is because the predictors work by represent a viable predictive method applicable to breast cancer. combining different variables in a unique model to maximize discrimination between groups. The advantage of using a combination of variables is that predictive ability is obtained from DATA AVAILABILITY STATEMENT the combination of this precise set of variables. That is, although The information has been made public and accessible in the some variable may show a small difference between groups, it repository https://figshare.com/ with the updated information may be the case that its contribution is different from other in the new document “Supplementary materials” provided. variables, so that including this variable in the model results in an increase of its global predictive capability. In a second phase, the predictor was validated and tested in 20 additional breast ETHICS STATEMENT cancer cases plus 60 volunteer women, respectively. While the 20 patients were correctly categorized, the predictor included 11 The studies involving human participants were reviewed and out of 60 women into the cancer group. Although the theoretical approved by Clinical Research Ethics Committee of Vall breast cancer risk in the overall women population of Europe and d’Hebron Hospital (CEIC). The patients/participants provided United States is 12.5%, according to our predictor, we found a their written informed consent to participate in this study. percentage of 18.3% women that will develop cancer in the future. This percentage (18.3%) represents an increase of ∼1.5 over the AUTHOR CONTRIBUTIONS expected values. A possible explanation of this high incidence could be the fact that, unlike the control group, this group of 60 AF, LG, and ML conceived and designed the experiments. AF, women were not selected by any criteria; therefore, they could LG, YG-M, and CM performed the experiments and designed the have a higher risk of developing breast cancer than the control graphs. JCa and SB performed the statistical study. JCo, IR, NB, group. It would be interesting to determine the health condition RM, and AA collected and organized the patients’ information. of those 60 women in the following 5–10 years with the purpose AF, ME-B, SR, IR, JCa, and ML analyzed and interpreted the data. of establishing the validation of our predictor in the future. MA contributed to technical support. AF, YG-M, CM, NB, RM, On the other hand, differently expressed proteins in the serum AA, IR, SR, ME-B, JCo, and JCa contributed to scientific support. of breast cancer patients vs. controls have been described (39). ML wrote the paper and conducted the study supervision. All The deregulated proteins found in the pools of cancer vs. control authors read and approved the final manuscript. serum samples—PEDF, IGKC, CD44, and CST3—have been previously reported (39–41). High levels of CD44 in serum are an independent prognosis indicator in primary breast cancer, since FUNDING it correlates with overall survival and disease-free survival (42). This work was supported by ISCIII (Instituto de Salud Carlos Interestingly, we found that lower expression of miR-16 in the III) Ref. FIS PI15/01262, co-financed by the European Regional serum of cancer patients correlated with high expression of its Development Fund (ERDF) and AECC Project GEC Ref. CD44 target protein. Our results reinforce the potential relevance GC16173720CARR (ML). AF, YG-M, and CM were granted with of CD44 as a potential marker of breast cancer as well as propose P-FIS, VHIR, and iP-FIS fellowships, respectively. other proteins that might play key roles as biomarkers such as CST3, which needs to be extensively and individually studied in the serum of large series of patients (40). ACKNOWLEDGMENTS Liquid biopsy (i.e., serum) is gaining importance in the clinical practice as novel biomarkers (i.e., microRNAs and proteins) are We are grateful to the Statistics and Bioinformatics Unit (UEB), being considered to monitor healthy individuals. We hope that Vall d’Hebron Research Institute (VHIR), since they carried out the results here reported open new avenues for future cancer the statistical analysis (Miriam Mota and Alex Sanchez-Pla) and prevention and diagnosis. the Proteomics Laboratory CSIC/UAB (Proteored) (Montserrat Overall, while much effort is being devoted to cancer Carrascal and Joaquin Abian). We are grateful to the members of predictive methods, it is not yet possible to detect cancer the CAP-Vallcarca-Sant Gervasi, the Castilla-Leon Blood Bank, before the appearance of the first clinical symptoms. A and the Government of Catalonia Blood Bank. We are grateful to molecular signature based on the detection in serum of five the laboratory technicians Teresa Moliné and Rosa Somoza from microRNAs capable of differentiating breast cancer patients from the Pathology Department (HUVH). healthy individuals was found. The clinical application of the molecular signature herein described will be determined in large SUPPLEMENTARY MATERIAL women’s cohorts. New microRNAs detected in serum and biopsy from breast The Supplementary Material for this article can be found cancer patients have been discovered. An association of low levels online at: https://www.frontiersin.org/articles/10.3389/fonc. of miR-16 with a higher content of CD44 protein in serum was 2020.586268/full#supplementary-material Frontiers in Oncology | www.frontiersin.org 8 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum REFERENCES involvement, recurrence patterns and disease free survival. Breast. (2016) 29:181–5. doi: 10.1016/j.breast.2016.07.016 1. DeSantis CE, Ma J, Goding Sauer A, Newman LA, Jemal A. Breast cancer 20. Ru Y, Kechris KJ, Tabakoff B, Hoffman P, Radcliffe RA, Bowler R, et al. statistics. 2017 racial disparity in mortality by state. CA Cancer J Clin. (2017) The multiMiR R package and database: integration of microRNA-target 67:439–48. doi: 10.3322/caac.21412 interactions along with their disease and drug associations. Nucleic Acids Res. 2. Pease AM, Riba LA, Gruner RA, Tung NM, James TA. Oncotype DX((R)) (2014) 42:e133. doi: 10.1093/nar/gku631 recurrence score as a predictor of response to neoadjuvant chemotherapy. Ann 21. Matamala N, Vargas MT, Gonzalez-Campora R, Minambres Surg Oncol. (2019) 26:366–71. doi: 10.1245/s10434-018-07107-8 R, Arias JI, Menendez P, et al. Tumor microRNA expression 3. Kleivi Sahlberg K, Bottai G, Naume B, Burwinkel B, Calin GA, Borresen- profiling identifies circulating microRNAs for early breast cancer Dale AL, et al. A serum microRNA signature predicts tumor relapse and detection. Clin Chem. (2015) 61:1098–106. doi: 10.1373/clinchem.201 survival in triple-negative breast cancer patients. Clin Cancer Res. (2015) 5.238691 21:1207–14. doi: 10.1158/1078-0432.CCR-14-2011 22. Xiong DD, Lv J, Wei KL, Feng ZB, Chen JT, Liu KC, et al. A 4. Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D, et al. nine-miRNA signature as a potential diagnostic marker for breast MicroRNA expression profiles classify human cancers. Nature. (2005) carcinoma. An integrated study of 1,110 cases. Oncol Rep. (2017) 435:834–8. doi: 10.1038/nature03702 37:3297–304. doi: 10.3892/or.2017.5600 5. Bahmanpour Z, Sheervalilou R, Choupani J, Shekari Khaniani M, Montazeri 23. Chen L, Li Y, Fu Y, Peng J, Mo MH, Stamatakos M, et al. Role of deregulated V, Derakhshan SM. A new insight on serum microRNA expression as microRNAs in breast cancer progression using FFPE tissue. PLoS ONE. (2013) novel biomarkers in breast cancer patients. J Cell Physiol. (2019) 234:19199– 8:e54213. doi: 10.1371/journal.pone.0054213 211. doi: 10.1002/jcp.28656 24. Yu X, Liang J, Xu J, Li X, Xing S, Li H, et al. Identification and validation 6. Zaleski M, Kobilay M, Schroeder L, Debald M, Semaan A, Hettwer K, of circulating MicroRNA signatures for breast cancer early detection et al. Improved sensitivity for detection of breast cancer by combination based on large scale tissue-derived data. J Breast Cancer. (2018) 21:363– of miR-34a tumor markers CA 15-3 or CEA. Oncotarget. (2018) 9:22523– 370. doi: 10.4048/jbc.2018.21.e56 36. doi: 10.18632/oncotarget.25077 25. Liu B, Su F, Chen M, Li Y, Qi X, Xiao J, et al. Serum miR-21 7. Gong C, Tan W, Chen K, You N, Zhu S, Liang G, et al. Prognostic and miR-125b as markers predicting neoadjuvant chemotherapy response value of a BCSC-associated MicroRNA signature in hormone and prognosis in stage II/III breast cancer. Hum Pathol. (2017) 64:44– receptor-positive HER2-negative breast cancer. EBioMedicine. (2016) 52. doi: 10.1016/j.humpath.2017.03.016 11:199–209. doi: 10.1016/j.ebiom.2016.08.016 26. Madhavan D, Cuk K, Burwinkel B, Yang R. Cancer diagnosis and prognosis 8. Feliciano A, Castellvi J, Artero-Castro A, Leal JA, Romagosa C, Hernandez- decoded by blood-based circulating microRNA signatures. Front Genet. Losa J, et al. miR-125b acts as a tumor suppressor in breast tumorigenesis via (2013) 4:116. doi: 10.3389/fgene.2013.00116 its novel direct targets ENPEP, CK2-alpha, CCNJ, and MEGF9. PLoS ONE. 27. Eichelser C, Flesch-Janys D, Chang-Claude J, Pantel K, Schwarzenbach (2013) 8:e76247. doi: 10.1371/journal.pone.0076247 H. Deregulated serum concentrations of circulating cell-free 9. Fortis SP, Vaxevanis CK, Mahaira LG, Sofopoulos M, Sotiriadou NN, microRNAs miR-17, miR-34a, miR-155, and miR-373 in human Dinou A, et al. Serum miRNA-based distinct clusters define three breast cancer development and progression. Clin Chem. (2013) groups of breast cancer patients with different clinicopathological and 59:1489–96. doi: 10.1373/clinchem.2013.205161 immune characteristics. Cancer Immunol Immunother. (2019) 68:57– 28. Chan M, Liaw CS, Ji SM, Tan HH, Wong CY, Thike AA, et al. 70. doi: 10.1007/s00262-018-2252-7 Identification of circulating microRNA signatures for breast cancer 10. Sochor M, Basova P, Pesta M, Dusilkova N, Bartos J, Burda P, et al. detection. Clin Cancer Res. (2013) 19:4477–87. doi: 10.1158/1078-0432.CCR- Oncogenic microRNAs: miR-155, miR-19a, miR-181b, and miR-24 enable 12-3401 monitoring of early breast cancer in serum. BMC Cancer. (2014) 29. Grimaldi AM, Incoronato M. Clinical translatability of “Identified” circulating 14:448. doi: 10.1186/1471-2407-14-448 miRNAs for diagnosing breast cancer. Overview and update. Cancers. (2019) 11. Shin VY, Siu JM, Cheuk I, Ng EK, Kwong A. Circulating cell-free miRNAs 11:E901. doi: 10.3390/cancers11070901 as biomarker for triple-negative breast cancer. Br J Cancer. (2015) 112:1751– 30. Huang GL, Sun J, Lu Y, Liu Y, Cao H, Zhang H, et al. MiR-200 family 9. doi: 10.1038/bjc.2015.143 and cancer. from a meta-analysis view. Mol Aspects Med. (2019) 70:57– 12. Xiang M, Zeng Y, Yang R, Xu H, Chen Z, Zhong J, et al. U6 is not a suitable 71. doi: 10.1016/j.mam.2019.09.005 endogenous control for the quantification of circulating microRNAs. Biochem 31. Zhang Y, Yan LX, Wu QN, Du ZM, Chen J, Liao DZ, et al. miR-125b is Biophys Res Commun. (2014) 454:210–4. doi: 10.1016/j.bbrc.2014.10.064 methylated and functions as a tumor suppressor by regulating the ETS1 13. Bonjoch L, Casas V, Carrascal M, Closa D. Involvement of exosomes in lung proto-oncogene in human invasive breast cancer. Cancer Res. (2011) 71:3552– inflammation associated with experimental acute pancreatitis. J Pathol. (2016) 62. doi: 10.1158/0008-5472.CAN-10-2435 240:235–45. doi: 10.1002/path.4771 32. Wang H, Tan G, Dong L, Cheng L, Li K, Wang Z, et al. Circulating MiR-125b 14. Guillen-Gomez E, Bardaji-de-Quixano B, Ferrer S, Brotons C, Knepper MA, as a marker predicting chemoresistance in breast cancer. PLoS ONE. (2012) Carrascal M, et al. Urinary proteome analysis identified neprilysin and 7:e34210. doi: 10.1371/journal.pone.0034210 VCAM as proteins involved in diabetic nephropathy. J Diabetes Res. (2018) 33. Pigati L, Yaddanapudi SC, Iyengar R, Kim DJ, Hearn SA, Danforth 2018:6165303. doi: 10.1155/2018/6165303 D, et al. Selective release of microRNA species from normal 15. Kearns M, Ron D. Algorithmic stability and sanity-check and malignant mammary epithelial cells. PLoS ONE. (2010) bounds for leave-one-out cross-validation. Neural Comput. (1999) 5:e13515. doi: 10.1371/journal.pone.0013515 11:1427–53. doi: 10.1162/089976699300016304 34. Rodriguez-Martinez, de Miguel-Perez D, Ortega FG, Garcia-Puche JL, Robles- 16. Breiman FJL, Stone CJ, Olshen RA. Classification and Regression Trees. Fernandez I, Exposito J, et al. Exosomal miRNA profile as complementary Belmont, CA: Wadsworth Inc (1984). tool in the diagnostic and prediction of treatment response in localized 17. Huo D, Clayton WM, Yoshimatsu TF, Chen J, Olopade OI. Identification of breast cancer under neoadjuvant chemotherapy. Breast Cancer Res. (2019) a circulating microRNA signature to distinguish recurrence in breast cancer 21:21. doi: 10.1186/s13058-019-1109-0 patients. Oncotarget. (2016) 7:55231–48. doi: 10.18632/oncotarget.10485 35. Sokilde R, Persson H, Ehinger A, Pirona AC, Ferno M, Hegardt C, 18. Feliciano A, Garcia-Mayea Y, Jubierre L, Mir C, Hummel M, Castellvi et al. Refinement of breast cancer molecular classification by miRNA J, et al. miR-99a reveals two novel oncogenic proteins E2F2 and expression profiles. BMC Genomics. (2019) 20:503. doi: 10.1186/s12864-019- EMR2 and represses stemness in lung cancer. Cell Death Dis. (2017) 5887-7 8:e3141. doi: 10.1038/cddis.2017.544 36. Madhavan D, Peng C, Wallwiener M, Zucknick M, Nees J, Schott S, 19. Vasconcelos I, Hussainzada A, Berger S, Fietze E, Linke J, Siedentopf et al. Circulating miRNAs with prognostic value in metastatic breast F, et al. The St. Gallen surrogate classification for breast cancer cancer and for early detection of metastasis. Carcinogenesis. (2016) 37:461– subtypes successfully predicts tumor presenting features, nodal 70. doi: 10.1093/carcin/bgw008 Frontiers in Oncology | www.frontiersin.org 9 November 2020 | Volume 10 | Article 586268 Feliciano et al. microRNAs in Breast Cancer Serum 37. Debeb BG, Lacerda L, Anfossi S, Diagaradjane P, Chu K, Bambhroliya 42. Kong Y, Lyu N, Wu J, Tang H, Xie X, Yang L, et al. Breast cancer stem A, et al. miR-141-mediated regulation of brain metastasis from breast cell markers CD44 and ALDH1A1 in serum: distribution and prognostic cancer. J Natl Cancer Inst. (2016) 108:djw026. doi: 10.1093/jnci/ value in patients with primary breast cancer. J Cancer. (2018) 9:3728– djw026 35. doi: 10.7150/jca.28032 38. Taslim C, Weng DY, Brasky TM, Dumitrescu RG, Huang K, Kallakury BV, et al. Discovery and replication of microRNAs for Conflict of Interest: JC holds Stock, patents and intellectual property of the breast cancer risk using genome-wide profiling. Oncotarget. (2016) company Medica Scientia Innovation Research (MedSIR). 7:86457–68. doi: 10.18632/oncotarget.13241 39. Pendharkar N, Dhali S, Abhang S. A novel strategy to investigate The remaining authors declare that the research was conducted in the absence of tissue-secreted tumor microenvironmental proteins in serum toward any commercial or financial relationships that could be construed as a potential development of breast cancer early diagnosis biomarker signature. conflict of interest. Proteomics Clin Appl. (2019) 13:e1700119. doi: 10.1002/prca.2017 00119 Copyright © 2020 Feliciano, González, Garcia-Mayea, Mir, Artola, Barragán, 40. Tumminello FM, Flandina C, Crescimanno M, Leto G. Circulating Martín, Altés, Castellvi, Benavente, Ramón y Cajal, Espinosa-Bravo, Cortés, Rubio cathepsin K and cystatin C in patients with cancer related bone disease: and LLeonart. This is an open-access article distributed under the terms of clinical and therapeutic implications. Biomed Pharmacother. (2008) 62:130– the Creative Commons Attribution License (CC BY). The use, distribution or 5. doi: 10.1016/j.biopha.2007.07.001 reproduction in other forums is permitted, provided the original author(s) and the 41. Chen C, Zhao S, Karnad A, Freeman JW. The biology and role of CD44 copyright owner(s) are credited and that the original publication in this journal in cancer progression: therapeutic implications. J Hematol Oncol. (2018) is cited, in accordance with accepted academic practice. No use, distribution or 11:64. doi: 10.1186/s13045-018-0605-5 reproduction is permitted which does not comply with these terms. Frontiers in Oncology | www.frontiersin.org 10 November 2020 | Volume 10 | Article 586268