Introduction

The amount of patients hospitalized in Western countries for chest pain accounts for several millions each year. In approximately half of the cases, chest pain is of cardiac origin [1]. Among these patients, approximately 50% exhibit an underlying coronary artery disease (CAD), causing either stable (SA) or unstable (UA) angina pectoris or myocardial infarction. In the absence of myocardial necrosis leading to increased plasmatic levels of the cardiac-specific protein Troponin, there are currently no established circulating biomarkers that may support the diagnosis of SA or UA. Notably, missed diagnosis of cardiac ischemia has been demonstrated to cause an increase in early and mid-term mortality [2]. Thus, the identification of novel noninvasive biomarkers of CAD lacking myocardial necrosis is a compelling need still under investigation.

MicroRNAs (miRNAs) are ~22 nucleotides long non-coding RNAs, known to regulate complex biological process by 'fine-tuning' the translation of specific messenger RNA targets [3]. MiRNAs are pivotal modulators of mammalian cardiovascular development and disease and can be steadily found in the systemic circulation of both animals and humans, where they show a remarkable stability probably due to internalization in vesicles and binding to circulating proteins and other molecules [4]. Since their levels may significantly change upon stress, circulating miRNAs have been proposed as diagnostic biomarkers in different pathologic conditions (e.g. cancer, cardiac diseases, liver injury, and hepatitis) [4-6]. Interestingly, recent reports have suggested the diagnostic potential of miRNAs in heart diseases, such as heart failure [7] and myocardial infarction (MI) [8]. In particular, we previously reported the presence of a distinctive “signature” of 6 circulating miRNAs in patients with ST-Elevation Myocardial Infarction (STEMI) [8]. In a more recent work, we have shown that circulating miR-499-5p may be a useful marker for early discrimination between congestive heart failure and Non ST-Elevation Myocardial Infarction (NSTEMI) in elderly patients presenting to hospital with unclear symptoms [9].

In the search for diagnostic biomarkers of chest pain of cardiac origin, circulating miRNAs have been previously investigated in blood, serum, plasma, platelets, and peripheral blood mononuclear cells (PBMC) in patients with both stable and unstable angina [10]. However, the high heterogeneity in study design, patient population, and miRNA source and detection methods is likely to be responsible for the high variability and poor overlap of the aberrant miRNA signatures identified, making extrapolation of conclusive information difficult.

The purpose of the present study was to search for distinctive miRNA profiles in plasma of patients with angiographically-documented troponin-negative CAD in comparison to controls matched for cardiovascular risk factors. The potential diagnostic significance of circulating miRNAs has been challenged by ROC curve and hierarchical cluster analysis with the aim of identifying putative specific CAD-associated miRNA signatures.

Materials and Methods

Results

Confirmation step and analysis of expression

The confirmation step of the screening phase was conducted on restricted cohorts of SA and UA patients and control subjects (n=10 for each group). Three out of the 10 putatively modulated miRNAs showed a statistically significant (p<0.05) up-regulation: miR-337-5p, -433, and -485-3p. Conversely, the other 7 miRNAs selected from the screening were not confirmed as significant at this step. The discrepancies between screening and validation results could be attributable to the differences in detection and normalization methods and in the sample sizes. The expression analysis was then extended to all remaining patients and controls and this step further uphold the significant positive regulation of miR-337-5p, -433, and -485-3p. In particular, when compared to controls, miR-337-5p (Figure 1A) showed a significant increase in the SA group (9.4±4.4 folds, p=0.001), but not in UA subjects. Conversely, miR-433 (Figure 1B) and miR-485-3p (Figure 1C) showed a very strong positive modulation both in SA (9.6±2.1 and 42.4±19.5 folds, respectively, p<0.001 for both) and in UA patients (4.6±1.3 and 29.7±17.8 folds, p=0.002 and p<0.001, respectively), when compared to controls. Interestingly, none of the 3 miRNAs showed statistically significant expression differences between SA and UA patients.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Plasmatic miRNAs regulation by CAD in Stable and Unstable Angina patients.

Panels A to L. All investigated miRNAs showed increased expression in CAD subjects, both in SA (yellow bars, 5.7–42.4 fold increase with the exception of miR-145, not regulated) and UA (blue bars, 2.9–29.7 fold increase with the exception of miR-337-5p, not regulated) patients. Values indicate fold changes (expressed as mean ± SEM) of each miRNA vs. its level in control healthy subjects (CTRLS, red bars), arbitrarily set to 1. (*) indicates p≤0.05 vs. control; NS=not significant. Differences among groups were compared using either One-way ANOVA or non-parametric Kruskal-Wallis with Dunn’s post-hoc test, when appropriate.

https://doi.org/10.1371/journal.pone.0080345.g001

Diagnostic potential of circulating miRNAs

To determine the predictive capacity of miRNAs identified above, we performed ROC curve analyses. The most predictive AUC values (>0.85) were obtained in SA patients, when compared to Controls, for miR-1, -126, and -485-3p (0.918, 0.929, and 0.851, respectively; Figure 2), although good AUC values were obtained for almost all considered miRNAs (Figure S1). Similarly, AUC values >0.85 were observed in UA subjects for miR-1, -126, and -133a (0.920, 0.867, and 0.906, respectively; Figure 3), while all remaining miRNAs showed AUC values <0.85 (Figure S2). Interestingly, none of the considered miRNAs reached an acceptable AUC value to differentiate SA from UA patients (Figure 4), ranging from 0.404 (miR-145) to 0.678 (miR-337-5p).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. ROC curve analysis of CAD-miRNAs in Stable Angina patients and control subjects.

The figure depicts calculated ROC curve and respective AUC values for miR-1, miR-126, and miR-485-3p, which exhibited good accuracy (AUC>0.85) in differentiating Stable Angina (SA) patients from matched controls (C).

https://doi.org/10.1371/journal.pone.0080345.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. ROC curve analysis of CAD-miRNAs in Unstable Angina patients and control subjects.

The figure depicts calculated ROC curve and respective AUC values for miR-1, miR-126, and miR-133a, which exhibited good accuracy (AUC>0.85) in differentiating Unstable Angina (UA) patients from matched controls (C).

https://doi.org/10.1371/journal.pone.0080345.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. ROC curve analysis of CAD-miRNAs in Stable and Unstable Angina patients.

None of the investigated miRNAs exhibited adequate accuracy in differentiating Stable (SA) from Unstable Angina (UA) patients: AUC values ranged between 0.404 (miR-145) and 0.678 (miR-337-5p).

https://doi.org/10.1371/journal.pone.0080345.g004

In an attempt to identify a signature with diagnostic potential resulting from a combination of miRNAs, we performed a hierarchical cluster analysis on all miRNAs found regulated by CAD. Figure 5 (and Figure S3) shows that the combined expression of miR-1, -126, and -485-3p correctly classifies SA patients vs. Controls with an efficiency of 90.2%. Similarly, the combination of miR-1, -133a, and -126 presented 87.2% efficiency in discriminating UA patients from control subjects (Figure 5 and Figure S4). In contrast, the efficiency of CAD-regulated circulating miRNAs to correctly differentiate SA from UA subjects was not higher than 66% (see Figure 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. miRNA clusters efficiently differentiate SA and UA patients from Controls.

Hierarchical clustering demonstrated that different miRNA “signatures” efficiently classify between matched controls (CTRLS) and CAD patients. The cluster composed by miR-1 and miR-126 and miR-485-3p can be used to correctly classify SA patients from controls with 90.2% (yellow bar) efficiency. Similarly the miR-1, miR-126 and miR-133a cluster can be used to correctly classify UA patients from controls with 87.2% (red bar) efficiency. No signatures of miRNAs could be found to efficiently discriminate SA from UA patients with an accuracy > 66% (miR-126 and miR-337-5p cluster, blue bar).

https://doi.org/10.1371/journal.pone.0080345.g005

Discussion

The main findings of this work are the evidences that: i) specific plasmatic miRNA signatures may have the potential to unveil with high efficiency (>85%) the presence of a significant angiographically documented Stable (miR-1, miR-126, miR-485-3p) or Unstable (miR-1, miR-126, miR133a) CAD vs. matched Controls; ii) conversely, the ability of plasmatic miRNAs to discriminate Stable from Unstable CAD is low.

MiRNAs have recently gained interest for their role in the pathogenesis of cardiovascular disease, including atherosclerosis and acute coronary syndromes [15]. In the search for diagnostic and predictive biomarkers of CAD, circulating miRNAs have been previously investigated in blood, serum, plasma, platelets, and PBMC. A large number of miRNAs have been found to be either up- or down-regulated in patients with angiographically-documented significant Stable and Unstable CAD compared with controls. However, differences in study design, patient population, and, particularly, miRNA sources, are likely to be responsible for the high variability and poor overlap of the aberrant miRNA signatures identified, thus making difficult to extrapolate conclusive information. In particular, since miRNA profiles are likely to be specific for the investigated blood components [10], the starting material is crucial for the evaluation of the potential of miRNAs as circulating biomarkers of CAD. In view of a possible clinical transferability, we have isolated circulating miRNAs from EDTA-plasma, given the sensitivity of the TaqMan qPCR technique in these preparations [14,16]. Further, plasma allows the exclusion of the strong interference of PMBC and platelets as sources of RNA, given the high heterogeneity of miRNAs signatures previously identified in these components [10]. To our knowledge, a previous study by Fichtlscherer and coworkers [14] is the only one that investigated the plasmatic levels of miRNAs from patients with CAD. Thirty-six patients with stable angiographically-documented CAD and 7 healthy controls served as derivation cohort; results obtained were then prospectively tested in 14 negative controls and 31 patients. Circulating levels of endothelial-enriched miR-126, miR-17, and miR-92a, inflammation-associated miR-155 and muscle-enriched miR-145 have been found significantly reduced in CAD compared with controls. In contrast, cardiac-muscle-enriched miR-133a and miR-208a showed a tendency for an increased concentration in plasma, albeit not significant. Notably, there are some important differences between Fichtlscherer’s study and ours, which should be considered when comparing the results. First, along with stable CAD patients, we have prospectively enrolled a cohort of troponin-negative angiographically-documented unstable patients. Secondly, we have used a different normalization methodology that relies on an internal normalizer instead of a spiked-in synthetic exogenous miRNA (e.g. cel-miR-39). In our opinion, our technique allows a more reliable identification of a suitable and stable internal normalizer which compensates for the variability of RNA amount in different samples. The alternative spike-in method, more suited for normalization of differences in RNA recovery, has intrinsic flaws, because it postulates that equal volumes of plasma always contain the same amount of RNA. Thus, the addition of a fixed amount of an exogenous synthetic miRNA to variable amounts of total RNA may affect the final miRNA quantification [17]. Finally and most importantly, we put attention during the enrollment phase to match stable and unstable patients for known cardiovascular risk factors and pharmacological therapy, given the influence of these variables on circulating miRNAs [18-21].

The discriminating power of selected and validated CAD-related miRNAs was firstly assessed by means of ROC curve analysis. We considered as highly discriminating AUC values those above 0.85 [22]. MiR-1, miR-126, and miR-133a were found to be highly predictive as potential single biomarker of both stable and unstable angina. Interestingly, miR-1 and miR-133a are co-transcribed and have been previously extensively reported as associated to cardiovascular disease [23]. Intriguingly, the plasmatic up-regulation of the “endothelial-specific” miR-126 in presence of stable and unstable angina may mirror a condition of endothelial dysfunction occurring in the ischemic microcirculation during chronic angina or acute coronary syndromes [24]. There is no known relation between miR-485-3p expression and cardiovascular disease.

It is noteworthy mentioning that, using our own normalization technique, Sun and co-workers [25] recently reported the decrease in plasmatic expression of miR-126 in CAD patients with high low-density lipoprotein (LDL) cholesterol levels. In contrast, the level of miR-126 was significantly increased when LDL cholesterol was high in patients who had risk factors for CAD but did not have angiographically significant CAD. Different from our study and in analogy with Fichtlscherer’s study, however, patient enrolment was performed in a population unmatched for age and major cardiovascular risk factors, and patients without significant coronary stenosis served as internal control group. These dissimilarities in patient’s selection may at least partly explain the discrepancy in results with our study.

MiRNA signatures in tissues and plasma have been shown in oncologic patients to be able to efficiently predict disease development and aggressiveness [26]. We have tested by means of a hierarchical cluster analysis the hypothesis that plasma miRNA signatures have the potential to unveil CAD. The combination of cardiac-enriched miR-1, endothelial-enriched miR-126, and miR-485-3p allowed the best discrimination between SA patients and control donors with good efficiency. The addition to this cluster of the cardiac-enriched miR-133a with exclusion of miR-485-3p identified successfully the presence of UA. These results may have a relevant clinical read-out, given the unmet need of circulating diagnostic biomarkers particularly in patients with stable angina [27].

We failed to identify any circulating miRNA (either by ROC curve or cluster analysis) with a satisfying potential in discriminating SA from UA. The two signatures overlapped for miR-1 and miR-126, and only differed for the presence of miR-485-3p and miR-133a in SA and UA patients, respectively. We and others have previously described an aberrant circulating miRNAs regulation in presence of acute coronary syndromes with Troponin dispersion [28]. Although it has been suggested that unstable angina may entail a higher expression of miRNA levels [29], SA and UA patients did not show a markedly diverse circulating miRNA profile in our clinical setting. It is worth to note that our UA population was carefully documented for its troponin-negative profile.

A relevant issue is whether plasmatic signatures elicited in angina are specific of troponin-negative CAD and not features in common with MI. Interestingly, a data lookup from our previous results in STEMI [8] and NSTEMI [9] showed that: i) miR-1 and miR-133a were found increased both in angina and MI patients; ii) miR-126 and miR-485-3p were up-regulated in patients with angina only; iii) miR-499-5p was observed specifically raised in MI patients only. A potential drawback of the aforementioned comparison lies in the use of different internal standards (miR-17-5p for MI and miR-16 for CAD groups). Nonetheless, a retrospective analysis of published MI plasma signatures showed that a restricted cohort of miRNAs have been repeatedly found dysregulated (namely miR-1, miR-133a, and miR-499-5p), irrespectively of the methodology of normalization adopted (see Table S1), and this is in full agreement with our observations. Notably, the aberrant plasmatic expression of miR-499-5p emerges as the most consistent marker of occurrence of heart damage [8,9,30,31]. Although the exact mechanism by which miRNAs are released into the bloodstream upon myocardial ischemia is a matter of investigation [32], this feature may be explained with a passive release of this specific miRNA triggered by cardiomyocyte death. Thus, mir-499-5p represents a potential discriminating factor within plasma miRNA signatures in CAD between Troponin-positive vs. negative coronary syndromes [23]. Confirmatory studies are warranted in the future to verify this issue under standardized experimental settings.

The main limitation of this work is its single-center set-up involving a limited number of patients. Multicenter large-scale studies will be required to definitively assess the potential of circulating miRNAs as diagnostic biomarkers of angina. Further, our screening method allowed the detection of only 367 human miRNAs: we cannot exclude that other miRNAs not present on our arrays may be modulated under the same experimental conditions. This work should, thus, be regarded of a proof-of-principle that specific circulating miRNAs may discriminate CAD patients from matched controls and, as such, have the potential to be used as reliable diagnostic biomarkers.

In conclusion, this work shows that plasmatic profiles of miRNAs are capable to accurately identify the presence of troponin-negative angiographically-documented CAD, causing either SA or UA. Due to the high overlap of the signatures pinpointed, the miRNAs discriminating power between SA and UA was found to be low. Further studies should address with larger population cohorts the potential utility of plasmatic miRNAs as biomarkers of CAD, along with the mechanisms underpinning their release in the bloodstream.

Supporting Information