Atrial fibrillation (AF) is associated with poor prognosis in patients with heart failure (HF). Although platelets play an important role in rendering a prothrombotic state in AF, the exact mechanism by which the effect is mediated is still debated. MicroRNAs (miRNAs), which have been shown to be involved in a variety of cardiovascular conditions, are abundant in platelets and in a cell-free form in the circulation. In the present study, we performed a genome-wide screen for miRNA expression in platelets of patients with systolic HF and in controls without cardiac disease, in pursuit of specific miRNAs that are associated with the presence of AF. MiRNA expression was measured in platelets from 50 patients with systolic HF and 50 controls, of which, samples from 41 patients with HF and 35 controls were used in the final analysis because of a quality control process. MiR-150 expression was 3.2-fold lower (p = 0.0003) in platelets of patients with HF with AF relative to those without AF. A similar effect was seen in serum samples from the same patients, in which miR-150 levels were 1.5-fold lower (p = 0.004) in patients with HF with AF. Furthermore, the serum levels of miR-150 were correlated to platelet levels in patients with AF (r = 0.65, p = 0.0087). In conclusion, miR-150 expression levels in platelets of patients with systolic HF with AF are significantly reduced and correlated to the cell-free circulating levels of this miRNA.
Atrial fibrillation (AF) is notoriously associated with heart failure (HF) and poor prognosis. The pathogenesis of AF is complex and results in atrial electrical and mechanical remodeling for which the therapeutic approach is debatable. MicroRNAs (miRNAs) are endogenous, small, noncoding ∼21- to 23-nucleotide RNAs that regulate gene expression by binding to the 3′ untranslated region of target gene messenger RNAs (mRNAs) to repress translation or induce mRNA cleavage. MiRNAs play an essential role in a variety of cardiovascular pathologies including hypertrophy, fibrosis, arrhythmia, ischemia, atherosclerosis, and HF. Interestingly, expression levels of specific miRNAs were demonstrated to play a role in AF pathogenesis, mainly through atrial remodeling, and circulating levels of specific miRNAs were found to be associated with AF. However, the possible association of platelet miRNAs with AF has not been studied before. It is important to note that although anucleated, platelets contain megakaryocyte-derived mRNAs that can be translated to produce protein molecules. Platelets also contain miRNAs that regulate the translation of mRNAs to proteins inside the cell and can be delivered to endothelial cells and affect their function. The aim of the present study is to evaluate the potential association of miRNAs in platelets with AF in patients with chronic systolic HF.
Methods
We recruited 50 patients with chronic, stable, stage C, systolic HF from our outpatient HF clinic and 50 volunteers who were age and gender matched to the HF group and had no known coronary, valvular, paroxysmal or persistent cardiac arrhythmia, or myocardial disease. All of the patients with HF had left ventricular ejection fraction <40%, were treated for at least 3 months according to the American College of Cardiology/American Heart Association guidelines, and were clinically stable, as judged by the treating HF specialized cardiologist on day of recruitment. The diagnosis of AF (paroxysmal, persistent, or chronic) was based on documentation of the arrhythmia by a senior cardiologist who reviewed the electrocardiograms or the Holter recordings or both, and was blinded to the results of the patients’ miRNA analyses. The study was approved by the Institution Review Board of the Lady Davis Carmel Medical Center, and all participants gave written informed consent before inclusion in the study and the initiation of any study related procedures.
For the platelet samples, 8 ml of blood was collected from each of the 100 participants into CPT collection tubes (BD Vacutainer CPT tubes 362761, Becton Dickinson and Company, Franklin Lakes, New Jersey). The tubes were allowed to stand for 2 to 9.5 hours at room temperature before being centrifuged at 1,800 g for 30 minutes at room temperature. The upper phase was then centrifuged at 4,000 g for 15 minutes at room temperature for separation of plasma poor platelets and the platelets pellet. The platelets pellet was stored at −80°C. After adding 1 ml mirVana Lysis/Binding Buffer to the platelets pellet, total RNA, including miRNA, was extracted according to the mirVana miRNA Isolation Kit manufacturer recommendations (Life Technologies, AM 1560, Grand Island, New York). Custom-designed arrays from Agilent Technologies (Wilmington, Delaware) were used for measuring levels of miRNAs in the samples as described previously using 0.28 to 1.5 μg of total RNA.
For the serum samples, 8 ml of blood was collected from each of the 100 participants. Sample handling, RNA extraction, and real-time reverse transcription-polymerase chain reaction were performed as described previously. Serum collection tubes were allowed to stand for 2 to 9.5 hours, and real-time reverse transcription-polymerase chain reaction was performed for 89 miRNAs whose signal was detected in serum samples but not in negative controls. Serum brain natriuretic peptide (BNP) and troponin I levels were measured using Triage Cardio2 Panel Test Kit (Biosite, San Diego, California) according to the manufacturer’s instructions.
Continuous clinical variables were compared between the 2 groups with a 2-sided unpaired t test except for BNP and troponin levels, which were compared using the Mann-Whitney U test. Chi-square tests were used to compare categorical clinical variables.
For the platelet data, array signals were normalized and analyzed as described previously. In short, analysis was performed on log-transformed signals to which polynomial normalization was applied, and fold changes for individual miRNAs were calculated based on the median values of the normalized expression in the compared groups.
For the serum data, signals from polymerase chain reaction were normalized and analyzed as described previously. In short, signals were normalized by scaling with the mean Ct of the samples and inverted by subtracting the normalized Ct from 50, so that high signals represent high expression levels. Fold changes for individual miRNAs were calculated as 2 Δ , where Δ is the absolute difference in median values of the normalized Ct in the 2 groups.
For both polymerase chain reaction and array data, normalized signals were compared between groups to find miRNAs that can be used to differentiate between the groups. Significance of differences was assessed by a 2-sided unpaired t test. The Benjamini-Hochberg false discovery rate method was used to control for multiple hypothesis testing, using a false discovery rate of 0.1.
Partitioning of samples by clinical parameters was performed according to dichotomous variables or by the median value for continuous variables. For each such partition the significance of differences in miRNA expression were assessed as described previously using a false discovery rate of 0.1. The ability to discriminate between HF with and without AF and control groups was characterized by the receiver operating characteristic curve and the area under the receiver operating characteristic curve was calculated. Logistic regression modeling was used to examine the association of AF with miRNA expression levels and clinical features. The model used AF status as the dependent variable, and miR-150 expression, patient age, and serum levels of BNP and troponin were the independent variables. The model was reduced in a stepwise manner until only statistically significant (p <0.05) terms remained. Pearson correlation was used to evaluate the correlation between the expression levels of miR-150 in serum and in platelets in samples taken from the same individuals.
Fifteen of the 100 platelet samples were excluded from the analysis because of failed RNA extraction or hybridization. Several additional samples had low overall miRNA expression. To objectively quantify the miRNA expression in the samples, the mean expression of each miRNA was calculated and 237 miRNAs with mean expression >300 were used to calculate the mean expression of each sample. Nine samples for which the mean expression was <1,000 were excluded from the analysis. The remaining samples from 41 patients with HF and 35 controls were used in the analysis. RNA extraction and polymerase chain reaction for the serum samples from the patients whose platelet samples passed quality assurance were successful, and all of them were used for the serum analysis.
Results
The clinical characteristics of the HF and control groups are listed in Table 1 ; differences in these characteristics reflect the common treatment modalities and co-morbidities of patients with HF. A comparison of the clinical characteristics of the HF with AF and without AF groups is listed in Table 2 . The 2 groups were similar except for predicted differences in age and serum levels of troponin and natriuretic peptide levels.
| Variable | HF (n = 41) | Control (n = 35) | p Value |
|---|---|---|---|
| Men | 35 (85) | 32 (91) | 0.41 |
| Smoker | 9 (22) | 4 (11) | 0.208 |
| Diabetes mellitus | 17 (41) | 4 (11) | 0.004 |
| Chronic renal failure ∗ | 12 (29) | 1 (3) | 0.003 |
| ACE-I/ARB | 38 (93) | 12 (34) | <0.001 |
| β Blocker | 39 (95) | 4 (11) | <0.001 |
| Aldosterone antagonist | 19 (46) | 0 (0) | <0.001 |
| Statin | 32 (78) | 17 (49) | 0.004 |
| Ischemic etiology | 21 (51) | — | — |
| ICD/CRT | 17 (41) | 0 (0) | <0.001 |
| Body mass index (kg/m 2 ) | 28.0 ± 5.6 | 27.2 ± 4.6 | 0.48 |
| Age (yrs) | 61.0 ± 11.0 | 65.7 ± 10.2 | 0.06 |
| Troponin I (μg/L) | 0.03 ± 0.09 | 0.00 ± 0.00 | 0.04 |
| BNP (pg/ml) | 147 (47–416) | 13 (5–31) | <0.001 |
| Ejection fraction | 30 (25–35) | — | — |
| AF | 15 (37) | 0 (0) | <0.001 |
| NYHA grade | |||
| I | 2 (5) | ||
| II | 19 (46) | ||
| III | 18 (34) | ||
| IV | 2 (5) | — |
| Variable | AF | p Value | |
|---|---|---|---|
| Yes (n = 15) | No (n = 26) | ||
| Men | 14 (93) | 21 (81) | 0.27 |
| Smoker | 1 (7) | 8 (31) | 0.08 |
| Diabetes mellitus | 5 (33) | 12 (46) | 0.42 |
| Chronic renal failure ∗ | 6 (40) | 6 (23) | 0.25 |
| ACE-I/ARB | 14 (93) | 24 (92) | 0.9 |
| β Blocker | 13 (87) | 26 (100) | 0.06 |
| Aldosterone antagonist | 6 (40) | 13 (50) | 0.67 |
| Statin | 13 (87) | 19 (73) | 0.41 |
| Ischemic etiology | 9 (60) | 12 (46) | 0.39 |
| ICD/CRT | 6 (40) | 11 (42) | 0.39 |
| Body mass index (kg/m 2 ) | 28.9 ± 6.2 | 27.5 ± 5.2 | 0.45 |
| Age (yrs) | 67.4 ± 10.2 | 57.3 ± 9.8 | 0.003 |
| Troponin I (μg/L) | 0.08 ± 0.14 | 0.01 ± 0.02 | 0.01 |
| BNP (pg/ml) | 266 (129–536) | 122 (33–327) | 0.03 |
| Ejection fraction | 30 (20–35) | 25 (20–30) | 0.57 |
| AF | 15 (100) | 0 (0) | <0.001 |
| NYHA grade | |||
| I | 1 (7) | 1 (4) | |
| II | 6 (40) | 13 (50) | |
| III | 7 (47) | 11 (42) | |
| IV | 1 (7) | 1 (4) | 0.91 |
In a comparison between miRNA expression levels in platelets of patients with HF with and without AF, miR-150 was the only miRNA that had significantly different expression between the 2 groups (see Methods ). MiR-150 expression levels were 3.2-fold lower in platelets of patients with AF, with a p value of 0.0003 and area under the receiver operating characteristic curve of 0.83 ( Figure 1 ).
To further validate the relation between AF and miR-150 expression levels, beyond the association with other clinical parameters, miR-150 levels and the 3 clinical parameters that were significantly different between patients with HF with and without AF ( Table 2 ) were incorporated into a logistic regression model for predicting AF status of patients with HF (see Methods ). MiR-150 expression level was the only significant parameter in the resulting model (p = 0.003).
In a comparison of miRNA expression levels in platelets from patients with HF and controls, none of the miRNAs, including miR-150, showed significant differences between the groups ( Figure 2 ), showing that platelet miRNA expression levels are not significantly altered by HF. MiR-150 expression levels in the control group were more similar to those in the HF-without-AF group; however, differences between miR-150 levels in the controls and patients with HF with and without AF did not pass the significance threshold ( Figure 3 ).
