MiR-320a as a Potential Novel Circulating Biomarker of Arrhythmogenic CardioMyopathy

Diagnosis of Arrhythmogenic CardioMyopathy (ACM) is challenging and often late after disease onset. No circulating biomarkers are available to date. Given their involvement in several cardiovascular diseases, plasma microRNAs warranted investigation as potential non-invasive diagnostic tools in ACM. We sought to identify circulating microRNAs differentially expressed in ACM with respect to Healthy Controls (HC) and Idiopathic Ventricular Tachycardia patients (IVT), often in differential diagnosis. ACM and HC subjects were screened for plasmatic expression of 377 microRNAs and validation was performed in 36 ACM, 53 HC, 21 IVT. Variable importance in data partition was estimated through Random Forest analysis and accuracy by Receiver Operating Curves. Plasmatic miR-320a showed 0.53 ± 0.04 fold expression difference in ACM vs. HC (p < 0.01). A similar trend was observed when comparing ACM (n = 13) and HC (n = 17) with athletic lifestyle, a ACM precipitating factor. Importantly, ACM patients miR-320a showed 0.78 ± 0.05 fold expression change vs. IVT (p = 0.03). When compared to non-invasive ACM diagnostic parameters, miR-320a ranked highly in discriminating ACM vs. IVT and it increased their accuracy. Finally, miR-320a expression did not correlate with ACM severity. Our data suggest that miR-320a may be considered a novel potential biomarker of ACM, specifically useful in ACM vs. IVT differentiation.

with structural and intracellular signalling functions, mediated by the WNT/β-catenin pathway 4 , which is a key regulator of cardiac development, response to injury 5 , and cell fate 6 .
The diagnosis of ACM is challenging and often late after disease onset because of the broad spectrum of phenotypic presentations. Currently, no single gold standard diagnostic method is available. Even the identification of an associated genetic defect, when possible, may be not conclusive because of the high variability in penetrance and expressivity. According to the International Task Force recommendations, the decision-making process relies on a scoring system of major and minor criteria, including clinical, instrumental and genetic tests. ACM is ascertained in the presence of either two major, one major plus two minor, or four minor criteria 7 . Importantly, using non-invasive tools only (e.g. Fig. 1), a conclusive diagnosis is not reached in a substantial number of patients. Invasive procedures such as endomyocardial biopsy and/or electroanatomic mapping may be needed to confirm the presence of the disease 8, 9 . Particularly challenging is also the differential diagnosis for ACM, which includes Idiopathic Ventricular Tachycardia (IVT), a primary electrical cardiac condition with a similar arrhythmic phenotype. The two entities share several phenotypic similarities: left bundle branch block with inferior axis ventricular tachycardia, symptoms aggravation after exercise, predominant occurrence in young, otherwise healthy individuals. Nonetheless, while ACM is a life-threatening disease, IVT has a more favourable outcome and different therapeutic options 10 . Thus, early and definite distinction between ACM and IVT is of the upmost clinical importance. As a consequence, novel ACM early diagnostic biomarkers are currently under investigation. The plakoglobin mislocalization assay is a promising candidate as a new diagnostic tool 11 , however it is exploitable only in bioptic or autoptic samples only. Plasma BIN1 has recently been proposed as predictor of arrhythmias in ACM, although restricted to subjects presenting symptomatic HF 12 .
In the last few years microRNAs (miRNAs), short (22-24 nucleotides) non-coding RNAs, have been found to be stably expressed in the systemic circulation of both animals and humans 13 . Notably, several reports from others and our laboratory described the diagnostic potential of circulating miRNAs in different heart conditions, including myocardial infarction 14 , HF 15 and atrial fibrillation 16 .
No previous study has evaluated circulating miRNAs as potential biomarkers of ACM. In the present work, we assessed miRNAs expression in plasma samples obtained from ACM and NON-ACM subjects. Our results indicate that miR-320a is significantly less expressed in ACM patients' plasma and that it may be helpful in differentiating ACM from IVT.

Study population.
A total of 110 male subjects were enrolled in three independent recruiting centres: 53 healthy controls (HC), 36 patients affected by ACM and 21 patients affected by IVT. Baseline characteristics of the different cohorts of patients are reported in Tables 1 and 2 and Supplementary Tables S1, S2 and S3. HC without previous history of heart disease, age-and sex-matched with ACM patients have been selected. Healthy long-distance runners, classified in group D of cardiovascular risk according to the COCIS score 17 , were classified as HC athletes. Blood samples were withdrawn in a resting period.
ACM diagnosis was reached according to the International Task Force criteria 7 . Among ACM patients, 13 subjects were classified in group D of cardiovascular risk 17 and defined as ACM-athletes for comparison with HC athletes.
IVT patients were characterized by right or left ventricular outflow tract tachycardia in the absence of any sign of structural heart disease 18 as documented by 2D-echocardiography, magnetic resonance and/or electroanatomic mapping 9 . BNP assay. Brain Natriuretic Peptide plasma levels were analysed using the BNP ARCHITECT chemiluminescent microparticle immunoassay on the ARCHITECT i2000 System (Abbott Diagnostics), following the manufacturer's instructions.  Table 2. Established Arrhythmogenic CardioMyopathy (ACM) major or minor diagnostic criteria 7  Webster). Bipolar and unipolar electrograms were analysed. Reference values for identifying normal endocardial bipolar and unipolar signals were defined as >1.5 and >5.5 mV, respectively.

2D-Echocardiography. 2D-echocardiography studies were performed with an iE33 system (Philips Medical
Systems) using the S5-1 sector array probe. Complete standard 2D-echocardiography was performed according to clinical practice. In particular, biplane left ventricular ejection fraction (LVEF) was measured from apical 4-chamber and 2-chamber views by the area-length method. From the 4-chamber apical view, the M-mode cursor was positioned at the junction of the tricuspid valve plane with the RV free wall to evaluate tricuspid annular plane systolic excursion (TAPSE), defined as the difference in the displacement of the RV base from end-diastole to end-systole and used as an index of RV function 19 .
Magnetic Resonance. Patients were evaluated in a 1.5-T scanner (Discovery MR450; GE Healthcare). CMR data sets were analysed with a dedicated software (Report Card 4.0; GE Healthcare) as previously described 20 .
Plasma and RNA preparation. Blood samples (5 ml) were collected in EDTA coated tubes and centrifuged at 1500 × g for 15 min. Supernatants were collected, centrifuged again at 16000 × g for 15 min to obtain cells-and platelet-free plasma, and stored at −80 °C as 400 µl aliquots until usage. Total RNA was extracted from plasma as previously described 21 . Briefly, 1 ml of TRIzol reagent (Life Technologies) was used for each 400 µl of plasma followed by phase separation and RNA precipitation with 1 ml of isopropanol after addition of 30 μg of glycogen (Roche). RNA pellets were obtained by centrifugation, washed with 1 ml of 70% ethanol, and centrifuged again for 10 min at 12.000 × g. Supernatants were discarded and pellets were resuspended in RNase-free water. miRNA Screening. TaqMan Human microRNA Card A Arrays version 2.0 (Life Technologies) were used for expression screening of 377 miRNAs, using 3 µl of total RNA from 3 ACM and 3 HC subjects. Reverse transcription (RT) and pre-amplification steps were performed according to the manufacturer's protocol. Data were analysed with the ExpressionSuite software v1.0.3 (Life Technologies), normalized by the Global Normalization method. All miRNAs presenting a Ct ≥ 35 were considered as not expressed. miRNAs whose expression in ACM patients consistently (t-test p < 0.05) differed from controls more than 2-folds were considered for the validation phase.
miRNA expression analysis. The expression of each single miRNA was evaluated using TaqMan microRNA assays (Life Technologies) following manufacturer's instructions. No RNA quantification was possible neither by spectrophotometric methods nor by using 2100 Bioanalyzer (Agilent Technologies). Therefore, the same volume of RNA (2 µl) was used in the RT step for technical consistency. As previously described 14,[22][23][24] we used an internal reference, so that all subsequent calculations were independent from RNA input. We selected 3 miRNAs, basing on screening data analysis (miR-210, miR-375) and on previous reports 24 (miR-16) for validation in all samples as potential references. Our results showed high intra-group data variability in all groups when using miR-16. Similarly, data obtained using miR-375 showed a less than optimal expression stability, as it was undetectable in a few samples. Conversely, miR-210 showed a strong expression in all samples with very limited intra-group variability when used for normalization. In addition, we confirmed, using NormFinder (http:// moma.dk/normfinder-software) that miR-210 has the lowest intra group variability and the best stability value (data not shown). Thus, miR-210 was selected as reference miRNA. The 2 −ΔΔCt method was used to calculate miRNA expression relative to the mean expression of a calibrator group, indicated for each comparison.
Statistical Analysis. Continuous variables are presented as mean ± SEM, when not differently indicated.
Comparisons between classes were performed by either parametric (One-way ANOVA, t-test) or non-parametric (Kruskal-Wallis, Mann-Whitney) tests, after assessment of the normality of the data by Kolmogorov-Smirnov test. Categorical variables are reported as frequencies and compared by Fisher's exact test. The Random Forests method 25 was used to rank the importance of selected variables in classifying ACM vs. IVT, by constructing multiple decision trees, bootstrapping 70% of samples 100 times. The mean decrease in Gini index was used as measure of the variable importance: Gini index measures the divergences between the probability distributions of the variable values. Therefore, variables with higher mean decrease in Gini index are more important for classification of the samples. The area under the curve (AUC) of Receiver Operating Characteristic (ROC) plots was used to assess diagnostic accuracy of selected variables: each plot shows the mean curve of 100 reiterated analyses through different bootstrap sampling of training and test sets. Correlation of miR-320a expression with clinical variables was calculated and reported as Spearman rho (ρ).
Data Availability. The datasets generated and analysed during the current study are available from the corresponding authors on request. In particular, the raw data generated by the screening can be analysed using the dedicated free online tool (http://www.thermofisher.com/it/en/home/cloud.html).

miR-320a expression in Arrhythmogenic CardioMyopathy patients and Healthy Controls.
We conducted a screening of expression of 377 miRNAs in plasma samples from 3 symptomatic ACM patients and 3 age-and sex-matched healthy donors. 121 miRNAs were detectable in all screened plasma samples and 5 miR-NAs showed a potential differential expression (see Supplementary Table S4). We then tested their expression by RT-qPCR in the whole cohort of ACM patients and HC (n = 36 and n = 53, respectively; baseline characteristics of the study groups are reported in Table 1). Only miR-320a showed a statistically significant lower expression in ACM patients with respect to HC (0.53 ± 0.04 and 1.00 ± 0.11, respectively, p < 0.01; Fig. 2).

Assessment of athletic status influence on miR-320a expression.
Since 13 out of 36 ACM patients in our cohort were athletes (36.11%), and strenuous physical activity is considered a precipitating factor for the development of the disease 26 , we compared their miR-320a plasmatic expression with the sub-populations of HC athletes (n = 17; 32.07%), and these two subgroups with the non-athletic counterparts (the baseline characteristics of the four subgroups are reported in Tables S1 and S2). As depicted in Fig. 3, no difference was observed between non-athlete and athlete subgroups both in HC (1.00 ± 0.22 vs. 1.16 ± 0.13, respectively; p = 0.10) and ACM patients (0.57 ± 0.04 vs. 0.55 ± 0.07, respectively; p = 0.60). On the contrary, the lower plasmatic expression of miR-320a in ACM vs. HC was confirmed both in non-athlete (p = 0.05) and athlete (p < 0.01) subgroups.

miR-320a expression in Arrhythmogenic CardioMyopathy and Idiopathic Ventricular Tachycardia patients.
To ascertain its potential benefit in differential diagnosis, plasma miR-320a expression was compared in ACM (n = 36) vs. IVT (n = 21) patients. As presented in Fig. 4, ACM patients showed significantly lower plasma levels of miR-320a than IVT patients (0.79 ± 0.05 fold; p = 0.03). Baseline characteristics of patient groups are described in Table 1, clinical and pathological features are reported in Table 2 and Table S3.

Ranking of non-invasive variables discriminating between Arrhythmogenic CardioMyopathy and Idiopathic Ventricular Tachycardia.
To evaluate the performance of miR-320a in partitioning ACM vs. IVT patients, we used Random Forest learning method, considering substantial non-invasive variables that can be used to distinguish the two populations ( Table 2). As shown in Fig. 5A, compared to each non-invasive parameter currently used to identify ACM patients (i.e. diagnostic criteria 7 ), miR-320a showed a mean decrease in Gini Index value of 2.72[2.43-3.12] (median[Q1-Q3]), ranking second only to global/regional dysfunction and structural alteration 5.76 [5.40-6.39], suggesting a higher classification potential than other well-known clinical criteria, in our cohort. ROC curve analyses (Fig. 5B) showed a moderate discriminating capacity of miR-320a alone (AUC = 0.69) and a slight increase in ACM prediction when miR-320a was added to global/ Similarly, as shown in Fig. S1, miR-320a presented a mean decrease in Gini Index of 3.57[3.09-4.01], higher than other non-invasive parameters that may help clinical distinction between the two diseases, namely TAPSE and LVEF 27 , Major Arrhythmic Events (MAE, defined as prior sustained ventricular tachycardia or ventricular fibrillation or aborted sudden cardiac death) 18 , and elevated BNP 28 . Indeed, miR-320a slightly increased the ACM vs. IVT discrimination potentials of the four parameters (Fig. S2).
Assessment of miR-320a association with Arrhythmogenic CardioMyopathy severity. We then evaluated whether miR-320a expression was associated with ACM severity. No correlation was found with LVEF and TAPSE (Table 3). Similarly, no correlation was found with levels of circulating BNP, a marker of HF previously associated with RV dysfunction in ACM 29 . RV low voltage areas at electroanatomic mapping identifies fibro-fatty substitution in ACM 30 ; in particular, unipolar mapping allows detection of epicardial pathological involvement 31 . miR-320a plasmatic expression was not associated with low potentials areas of the RV measured neither by bipolar nor unipolar mapping (Table 3). In addition, no difference in miR-320a plasmatic expression was observed when comparing patients with or without history of MAE, as reported in Table 4. Finally, ACM patients were arbitrarily categorized as affected by either severe or mild disease based on the presence or absence (at imaging) of ventricular dilation and bulging: again, miR-320a did not show any difference in expression between the two groups (Table 4).

Discussion
The diagnostic flowchart for the assessment of ACM could decisively benefit from the discovery and exploitation of non-invasive biomarkers. Since circulating miRNAs are cheap, safe, and readily accessible, they represent good candidates to be investigated. To date, no study sought an association between circulating miRNAs and ACM. In the present work, we report, for the first time, that a plasma miRNA signature may help in ACM diagnosis. Specifically, we observed that miR-320a plasma levels are significantly lower in ACM patients compared to both HC and IVT patients. Moreover, plasma levels of miR-320a were not influenced by intense physical activity, which is known to influence ACM disease penetrance 2 , as they did not change in athletes, neither in the HC nor in the ACM groups. Thus, we may speculate that miR-320a differential expression is independent from increased mechanical stretch and training-induced adaptive heart remodelling. . Results are shown with respect to the mean expression of miR-320a in non-athlete HC, arbitrarily set to one. Kruskal-Wallis followed by Dunns test was used to determine that no significant difference is observed between non-athlete and athlete HC subgroups (1.00 ± 0.22 vs. 1.16 ± 0.13; p = 0.10) and between non-athlete and athlete ACM subgroups (0.57 ± 0.04 vs. 0.55 ± 0.07; p = 0.60). HC vs. ACM differences are confirmed both between the non-athlete (1.00 ± 0.22 vs. 0.57 ± 0.04; p = 0.05) and athlete (1.16 ± 0.13 vs. 0.55 ± 0.07; p < 0.01) subgroups.    The correct early distinction of a life-threatening condition such as ACM from a benign condition such as IVT remains a challenging task. The miR-320a significant differential expression and its high ranking among known ACM diagnostic criteria in distinguishing ACM from IVT suggest that this microRNA has the potential to be clinically exploited for differential diagnosis. Indeed, the classification accuracy of selected established ACM diagnostic criteria is improved by the contribution of miR-320a. In addition, miR-320a performed better in discriminating ACM vs. IVT than additional non-invasive clinical parameters such as TAPSE, MAE, LVEF, and BNP levels.
Considering that clinical ACM diagnosis is challenging particularly in the initial, concealed phase, new tools, independent from disease progression, are currently needed. Since we could not detect any correlation between miR-320a expression and indexes of ACM severity, we expect that miR-320a could be used as a marker of disease presence also at its onset. Indeed it would be interesting to test its diagnostic ability to discriminate patients with early-stage ACM from those with ventricular arrhythmias not bound to develop structural defects.
It will be of interest to address, in future studies, any mechanistic role of miR-320a in ACM pathogenesis. Intriguingly, miR-320a was found increased during adipogenic differentiation of human mesenchymal bone marrow cells 32 . Moreover, miR-320a expression regulates cardiomyocyte survival 33 and the WNT pathway by directly targeting β-catenin 34 , which are both involved in ACM pathogenesis 4 . In a previous study conducted by screening cardiac autopsy samples, 21 miRNAs were selected from all those putatively dysregulated, and were validated as being differentially expressed between ACM and HC 23 . Although miR-320a expression was not investigated, its regulation cannot be excluded. Thus, we speculate that our findings in plasma could mirror deregulation in cellular miR-320a expression, resulting in cardiac cell apoptosis and adipogenesis.
Some study limitations have to be acknowledged. First, the low samples size used for the screening may have influenced the results in terms of number of miRNAs identified as putatively regulated. Nevertheless, the subsequent validation in a consistent number of patients (given the low prevalence of the disease) confirmed the soundness of the identified dysregulation. In addition, due to the low ACM occurrence in females, we enrolled male patients only. Thus, transferability of our results to the female ACM population should be evaluated in further studies. Moreover, as a genetic characterization was not available for most of the recruited patients, a link between miR-320a expression and a specific genetic variant could not be addressed in this study. Finally, we did not investigate the ACM-specificity of miR-320a expression versus other cardiomyopathies or channelopathies. However, these conditions are generally not in the front line for differential diagnosis.
In conclusion, this is the first study addressing the involvement of circulating miRNAs in ACM. The present results pave the way to future prospective studies aimed at defining the predictive value of miR-320a in larger ACM patients' cohorts and represent the proof of concept that plasmatic miRNAs could be exploited as potential non-invasive biomarkers of ACM diagnosis.