Plasma microrna expression profile for reduced ejection fraction in dilated cardiomyopathy

The left ventricular (LV) ejection fraction (EF) is key to prognosis in dilated cardiomyopathy (DCM). Circulating microRNAs have emerged as reliable biomarkers for heart diseases, included DCM. Clinicians need improved tools for greater clarification of DCM EF categorization, to identify high-risk patients. Thus, we investigated whether microRNA profiles can categorize DCM patients based on their EF. 179-differentially expressed circulating microRNAs were screened in two groups: (1) non-idiopathic DCM; (2) idiopathic DCM. Then, 26 microRNAs were identified and validated in the plasma of ischemic-DCM (n = 60), idiopathic-DCM (n = 55) and healthy individuals (n = 44). We identified fourteen microRNAs associated with echocardiographic variables that differentiated idiopathic DCM according to the EF degree. A predictive model of a three-microRNA (miR-130b-3p, miR-150-5p and miR-210-3p) combined with clinical variables (left bundle branch block, left ventricle end-systolic dimension, lower systolic blood pressure and smoking habit) was obtained for idiopathic DCM with a severely reduced-EF. The receiver operating characteristic curve analysis supported the discriminative potential of the diagnosis. Bioinformatics analysis revealed that miR-150-5p and miR-210-3p target genes might interact with each other with a high connectivity degree. In conclusion, our results revealed a three-microRNA signature combined with clinical variables that highly discriminate idiopathic DCM categorization. This is a potential novel prognostic biomarker with high clinical value.

Validation study and association with a population that has a reduced ejection fraction and DCM. The ischemic and idiopathic DCM populations are presented as DCM MOD and DCM SEV groups based on their LVEF, ≤ 30% and > 31-49%, respectively. We investigated the potential value of the miRNA candidates to discriminate between the DCM SEV vs. the DCM MOD population and their association with echocardiographic parameters (Table 3 and Table S2). Only the idiopathic DCM SEV group showed an association between LVEF and fourteen miRNA candidates (Table 4). These miRNAs were let-7a-5p, let 7 g-5p, miR-19b-3p, miR-25-3p, miR-29a-3p, miR-30b-5p, miR-30e-3p, miR-142-3p, miR-145-5p, miR-150-5p, miR-199a-3p, miR-210-3p, , miR-324-3p and miR-660-5p. No association was found in the idiopathic DCM MOD cohort with LVEF (Table 4). Moreover, the correlation between other echocardiographic variables and individual miRNAs was also identified. All these miRNAs significantly correlated with LVEF in DCM SEV patients. In addition, these miRNAs in the DCM MOD group were significantly correlated with LV end-systolic dimension (LVESD) and the left atrium dimension (LA).
A graphical illustration of the potential miRNA expression levels to discriminate between DCM SEV and DCM MOD in the idiopathic group is represented in Figs. 2 and 3. In all cases, the average miRNA levels were significantly higher in the idiopathic DCM group than in healthy subjects. Moreover, we found a significant increase in these circulating miRNA levels in plasma from DCM SEV compared with DCM MOD idiopathic patients. www.nature.com/scientificreports/ Combining miRNAs for DCM detection and subtype categorization. The ability of these miRNAs to discriminate between DCM with moderate and severe EF impairment was assessed using AUC-ROC. All individual miRNAs show an AUC ≥ 0.7, except for miR-30b-5p, miR-210-3p and miR-324-3p that were 0.69, 0.68 and 0.69, respectively. The highest AUC values achieved by single miRNAs was obtained for miR-145-5p with an AUC of 0.78 (95% CI: 0.65-0.91; p = 0.0029). The AUC values obtained for let-7a-5p, let 7 g-5p, miR-19b-3p, miR-25-3p, miR-29a-3p, miR-30e-3p, miR-142-3p, miR-150-5p, miR-199a-3p, and miR-660-5p when DCM SEV were compared to DCM MOD in the idiopathic group, as well as sensitivity for a whole range of specificities, are shown in Table 5.
A predictive model to determinate the association of microRNAs with a very reduced ejection fraction in idiopathic DCM. Statistical (Table 6). In addition, the Akaike Information Criterion (AIC) of our model showed a value of 35.406. The AIC for a model just with a combination of miR-150-5p and miR-210-3p levels showed a higher value (AIC: 61.393). This lower value indicates that our model has a better fit. The potential usefulness of our model for identifying the idiopathic DCM SEV population was obtained with a ROC curve analysis. As shown in Fig. 4, the AUC was 0.96 (95% CI: 0.884-1.00; p < 0.001). miRNA-gene network analysis. The hypothetical functions of three miRNAs were assessed with the miRDB database to reach the target mRNAs. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways database identified the biological and molecular processes. The miR-130b-3p, miR-150-5p and miR-210-3p were associated with 917, 902 and 83 mRNAs in the miRDB database, respectively (Fig. 5A). The analysis revealed that most of these genes are involved in signalling and regulation of the cellular and metabolic process and genetic information processing (Figs. 5B-C). For instance, the genes are deeply involved in the transcription pathway since these miRNAs regulate the expression of the RNA Polymerase II, transcription factors or alternative splicing. In addition, other target genes are implicated in the nitrogen compound metabolic process or in response to oxygen levels, among others (extended in Fig. S3).
On the other hand, the protein-protein interaction networks (PPI) analysis networks related to miR-210-3p were analysed using the search tool for the retrieval of interacting genes (STRING) (Fig. 5D). Our in-silico model showed that ISCU and other mitochondrial activity-related proteins, as well as SDHD and NDUFA4, are targets of miR-210-3p. Moreover, the analysis revealed that genes targeted by miR-150 and miR-210 showed a significant PPI enrichment (p = 0.00459) ( Figure S4 and Table S3). These results demonstrated that target genes of miR-210-3p and miR-150-5p may interact with each other.

Discussion
We analysed the circulating miRNA signature of the ischemic and idiopathic DCM population stratified in severely and moderately reduced EF, ≤ 30% and 31-49% respectively. Our study reports, for the first time, a signature of three circulating miRNAs (miR-130b-3p, miR-150-5p and miR-210-3p) that allow us to discriminate between idiopathic DCM MOD and DCM SEV subjects. Furthermore, this circulating miRNA panel in plasma defines idiopathic DCM SEV from ischemic etiology.
LVEF is the cornerstone of DCM outcome. Optimal pharmacological and non-pharmacological decisions, mostly based on LVEF, have improved DCM prognosis drastically 16 . Major arrhythmogenic episodes, sudden death events and mortality are related to DCM with lower EF ≤ 35% 7 . The impact of stratifying the severity of systolic impairment is crucial in DCM as shown 6,7,17 , but access to the EF has several limitations. The need for expert professionals, intra-and inter-observational differences or time-consuming imaging techniques in transthoracic www.nature.com/scientificreports/ echocardiography or cardiac magnetic resonance, as well as inaccessible or excessive radiation imaging techniques, make it difficult to monitor the patient. The lack of a specific circulating biomarker to stratify LVEF and the limitations of cardiac imaging tests lead to the need to investigate a novel tool for this at-risk population. Some studies have struggled to relate preserved and impaired LVEF to HF categorization [18][19][20] . The dysregulation of miRNAs related to DCM diagnosis and its etiology has scarcely been characterized nor categorization of the EF has not been reported in this population 14,15,21 . Some groups have worked on the reverse remodeling of LV in DCM or the signalling network 22 . To the best of our knowledge, no study has focused on LVEF in the DCM population and circulating miRNA.
In the present study, fourteen plasma miRNAs showed discriminative power to define DCM SEV vs. DCM MOD idiopathic patients based on their correlation to some echocardiographic variables. The idiopathic DCM SEV group showed an association between LVEF with these miRNAs. In the case of DCM MOD patients, the miRNAs were associated with LVESD, and LA. Remarkably, it has been suggested that pressure overload in earlier stages of DCM and elevated LV filling pressures lead to a hemodynamic imbalance that justifies the increased LA and LVESD in this cohort 23,24 .. Interestingly, some of these fourteen miRNAs have been related to HF and DCM patients with an EF < 30%, compared with healthy controls. Circulating levels of let-7a-5p were significantly upregulated in HF patients 25 . Myocardial and circulating miR-30 levels were significantly increased in DCM patients 22 . We present a novel, reliable fingerprint of circulating miRNAs to discriminate between idiopathic DCM MOD and DCM SEV subjects.
A multiparametric model was performed as a potential tool for the diagnosis of idiopathic DCM SEV . A panel comprised of a combination of three circulating miRNAs (miR-130b-3p, miR-150-5p and miR-210-3p) with four clinical variables (LBBB, LVESD, SBP and smoking habit) showed a high-yield diagnostic accuracy with an AUC of 0.96. Significantly, miR-150-5p and miR-210-3p had a positive value while miR-130b-3p had a negative value for a diagnosis of idiopathic DCM. SEV .
A total of 1902 genes were predicted to be targeted by these three miRNAs. Although they have been described as key hypoxia-related miRNA, our in-silico analysis revealed that they are more than silent players in hypoxia. Hence, these miRNAs may influence DCM SEV devolvement, promoting heart injury. MiR-210-3p and miR-150-5p share targeted genes linked within the cellular proliferation, differentiation oxidative stress and apoptosis [26][27][28][29] . The miR-210 expression has been extensively studied in several cardiovascular-related diseases with controversial results [30][31][32][33] . Its overexpression reduces mitochondrial reactive oxygen species production, and promotes cardiomyocyte proliferation, cell survival and angiogenesis post-myocardial infarction 30,31 . In disagreement,  34 . Our in-silico model showed that several mitochondrial activity-related proteins, are targets of miR-210-3p. Downregulation of these proteins may influence mitochondrial activity. Hence, our data are in agreement with previous studies that demonstrate impaired mitochondrial function in idiopathic DCM 35,36 . On the other hand, miR-150 levels have been described as a responsive protector in ischemic heart disease 37-39 and DCM 40 . MiR-150 protects the heart from ischemic injury by inhibiting the inflammatory response and repressing pro-apoptotic gene expression 39 .
Otherwise, a different role has been defined for miR-150. Its levels were upregulated in the infarct area compared with other cardiac zones, altering the LV remodeling 37 . Furthermore, the downregulation of miR-150 targeted genes intensifies cardiac injury by deregulation of cell cycle-dependent intracellular Ca 2+ concentration, and oxidative stress increase 29 ; accompanied by an increase in reactive nitrogen disease, DNA damage and cardiomyocyte death 41 . MiR-130b-3p is negatively associated in our model. Instead of having a cytoprotective role, miR-130b-3p promotes cardiomyocytes injury 42 . Although these three miRNAs have been studied in several cardiovascular-related diseases, their role is still unknown regarding idiopathic DCM.
Using predictive tools, we report an additive value to our model, including some clinical variables such as LBBB, LVESD, SBP and smoking habit. LBBB, SBP and LVESD are a risk factor marker of HF 5,43 . Interestingly, LBBB and lower EF have widely shown a better response to non-pharmacological treatment, such as an implantable defibrillator [44][45][46] . Both clinical markers are related to neurohormonal activation in HF leading to severe myocardial damage 43 . Our circulating levels of miRNAs may reflect a dysregulation of the cardiomyocyte protecting processes arising from severe impairment of the LVEF in the DCM SEV population.
We have described a circulating miRNA signature as a useful and non-invasive tool for LVEF classification in idiopathic DCM patients. In addition, we have defined a parametric model, a miRNA panel and four clinical parameters that categorize DCM based on LV systolic impairment with the highest diagnostic values reported to date. The bioinformatics analysis may elucidate the underlying molecular and cellular mechanisms of idiopathic DCM. Our results present a step-forward personalized therapeutic strategy and management of this entity. www.nature.com/scientificreports/ There are several limitations to the current study. The study sample was small, but the patients who were included were strictly DCM subjects. Recently, it has been reported that sex differences might play a role in the prognosis of DCM 47,48 . In our case, larger sample size is needed to validate these data. Therefore, before these novel biomarkers can be routinely used in clinical practice, the data should be replicated and extended to larger populations. Secondly, data on natriuretic peptides, troponin, or reverse remodeling were not available in all patients to compare to our results. Our sample was of patients in a chronic situation, who were not hospitalized but were prospectively followed in the outpatient clinic. Future studies should include these data. The dysregulation of miRNAs related to DCM diagnosis and its etiology has scarcely been characterized and characterization of the EF has not been reported in this population. In the bioinformatic analysis, we only identified the genes targeted by miR-210-3p and miR-150-5p, since both are positively related to our logistic regression model. MiR-130b-3p was not included in the PPI analysis since it is negatively related to our logistic model. In addition, the analysis of 1902 genes did not show clear results due to a large number of interactions and data overload. Finally, we have no evidence that these circulating miRNAs are directly secreted into the extracellular space from the heart. Thus, dedicated experiments on human heart samples are necessary to verify these results and bioinformatic predictions.

Conclusion
In conclusion, we identified a miRNAs fingerprint that is differentially expressed in the idiopathic DCM SEV population. This signature arises as a potential clinical biomarker to discriminate DCM etiology and stratify its risk, based on the LVEF. The clinical usefulness of this miRNA panel as a diagnostic tool could lead to tailored treatment improving the DCM population outcome. In addition, the bioinformatics analysis results may be useful in elucidating the underlying mechanisms of idiopathic DCM.

Methods
Study population and design. This was a prospective cross-sectional study. Inclusion criteria were patients with clinical features of DCM 1 , a LV end-diastolic diameter larger than 56 mm and a LVEF below 50%. Exclusion criteria were genetic DCM or any cardiovascular, life-limiting systemic condition or an infectious or tumoral condition that could influence the DCM definition or miRNA results. Only patients older than 18 years old were enrolled. A total of 159 consecutive subjects were included in the study: (1) 55 idiopathic DCM patients, (2) 60 ischemic DCM and (3) 44 healthy controls who had been referred to the Cardiology Department of the University Hospital Puerta del Mar, Cádiz, Spain. The study design is shown in Fig. 1. LVEF was classified into two categories based on prior studies: ≤ 30%, severely reduced EF (DCM SEV ) and 31-49% moderately reduced EF (DCM MOD ) 49,50 . A transthoracic echocardiography protocol was performed as described previously 13,15 . Echocardiographic methods for the measurement of EF were the apical biplane method of disks (modified Simpson's rule) and when this was not possible, Teicholz 51 . Detailed anthropometric, clinical and pharmacological information was obtained from each subject including family history, symptoms of HF, an electrocardiogram, a 24-h Holter electrocardiogram and, when appropriate, cardiac magnetic resonance. Our institution's ethics committee (Comité de Ética de la Investigación de Cádiz) approved the study protocol. The study was performed in full compliance with the Helsinki II Declaration. All participants provided written informed consent.
Blood collection. Ten millilitres of peripheral blood were collected in K2-ethylenediaminetetraacetic acid tubes (BD) after 10 h overnight fasting and were immediately centrifuged (1500 xg, 15 min, 4 °C). The blood was processed within 4 h after isolation. The upper layer containing plasma was divided into aliquots and stored at -80ºC until further analysis. MicroRNA real-time reverse transcriptase-polymerase chain reaction. RNA was reverse transcribed using the miRCURY LNA RT Kit (Qiagen) as previously described 15 . The reverse transcription reaction was performed under the following conditions: 60 min at 42 ºC, heat-inactivated for 5 min at 95 ºC and imme- www.nature.com/scientificreports/ To account for the variability between the plates, the interplate calibrator UniSp3 was analysed. Cqs above 35 cycles were censored at the minimum level observed for each miRNA. cel-miR-39-3p levels were stable across all samples. Relative quantification was performed using the 2 − ΔCq method, where ΔCq = CqmiRNA − Cqcel-miR-39-3p miRNA levels were log-transformed before being used in the statistical analyses. We have previously used this approach in the field of circulating non-coding RNAs 13,15 . miRNA-gene network analysis. The miRNAs obtained were tested using the miRDB database (http:// mirdb. org/) to predict the targeted genes 52 . A database analysis to identify the biological function was performed using GO enrichment analysis (http:// geneo ntolo gy. org/) 53 and the KEGG (https:// www. kegg. jp/ kegg/) software 54 . R software (www.r-proje ct. org) was assessed to build up miRNA-mRNA 55 . The STRING database (http:// www. string-db. org/) was used to analyse the PPI networks 56 .

RNA isolation.
Statistical analysis. Continuous variables are shown as mean ± standard deviation. Categorical variables are expressed as frequency and percentage of patients (%). Intergroup comparisons of miRNAs levels were performed using non-parametric Mann-Whitney and Kruskal-Wallis rank tests for continuous variables. The Pearson correlation coefficient was used for correlations between echocardiographic and clinical parameters vs. log 2 miRNAs. An analysis of differences between groups was performed using analysis of variance. Receiver operating characteristic (ROC) curves that characterize the diagnostic performance of candidate miRNAs and logistic regression models were plotted to determine the area under the curve (AUC) and the specificity and sensitivity of the optimal cut-offs. ROC curves were generated by plotting sensitivity against 100-specificity. Data were presented as the AUC and 95% confidence intervals. The relationships between miRNAs and LVEF status were assessed using logistic regression. Multiple logistic regression modelling was applied to perform the miRNA panel. The regression coefficients of each miRNA that was significantly associated with the outcome were applied to estimate the miRNA panel value. The Wilcoxon test and iterating combinations between our miRNA candidates, as well as echocardiographic and clinical covariates, were used to construct several models. The changes in p-values of their variables were evaluated by the Wald test and a likelihood ratio. The statistical software package R (www.r-proje ct. org) was used for all analyses 55 .  www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.