Comparative expression profiles of microRNA in left and right atrial appendages from patients with rheumatic mitral valve disease exhibiting sinus rhythm or atrial fibrillation

Background The atrial fibrillation (AF) associated microRNAs (miRNAs) were found in the right atrium (RA) and left atrium (LA) from patients with rheumatic mitral valve disease (RMVD). However, most studies only focus on the RA; and the potential differences of AF-associated miRNAs between the RA and LA are still unknown. The aim of this study was to perform miRNA expression profiles analysis to compare the potential differences of AF-associated miRNAs in the right atrial appendages (RAA) and left atrial appendages (LAA) from RMVD patients. Methods Samples tissues from the RAA and LAA were obtained from 18 RMVD patients (10 with AF) during mitral valve replacement surgery. From these tissues, miRNA expression profiles were created and analyzed using a human miRNA microarray. Then, the results were validated using qRT-PCR analysis for 12 selected miRNAs. Finally, potential targets of 10 validated miRNAs were predicted and their functions and potential pathways were analyzed using the miRFocus database. Results In RAA, 65 AF-associated miRNAs were found and significantly dysregulated (i.e. 28 miRNAs were up-regulated and 37 were down-regulated). In LAA, 42 AF-associated miRNAs were found and significantly dysregulated (i.e. 22 miRNAs were up-regulated and 20 were down-regulated). Among these AF-associated miRNAs, 23 of them were found in both RAA and LAA, 45 of them were found only in RAA, and 19 of them were found only in LAA. Finally, 10 AF-associated miRNAs validated by qRT-PCR were similarly distributed in RAA and LAA; 3 were found in both RAA and LAA, 5 were found only in RAA, and 2 were found only in LAA. Potential miRNA targets and molecular pathways were identified. Conclusions We have found the different distributions of AF-associated miRNAs in the RAA and LAA from RMVD patients. This may reflect different miRNA mechanisms in AF between the RA and LA. These findings may provide new insights into the underlying mechanisms of AF in RMVD patients.


Background
Atrial fibrillation (AF) is the most common sustained arrhythmia in clinical practice and it is associated with pronounced morbidity, mortality, and socio-economic burden [1]. Effective therapy and prevention is crucial for the control of AF-related morbidity and mortality. However, to date, medical interventions for AF are relatively limited, because the precise mechanisms of AF have not been completely elucidated. Therefore, new methods leading to further insights into the underlying mechanisms of AF and potential novel mechanismbased therapeutic strategies are required [2].
MicroRNAs (miRNAs) are emerging as new regulators of gene expression at the post-transcriptional level and are shedding new light on the regulatory mechanisms underlying AF [3]. miRNAs are small, endogenous, single-stranded, non-coding RNAs of ∼22 nucleotides that bind target sequences in the 3′-untranslated region of target genes to induce mRNA instability or to inhibit translation [4]. Studies have shown that miRNAs regulate key genetic functions in cardiovascular biology and play an important role in the pathogenesis of cardiac diseases including cardiac development [5], cardiac hypertrophy/heart failure [6], myocardial infarction [7], myocardial ischemia-reperfusion injury [8], and arrhythmogenesis [9]. Recently, there are many studies indicating that miRNAs are involved in AF through their target genes [10][11][12][13].
AF often occurs concomitantly with other cardiovascular diseases such as hypertension, coronary artery disease, congestive heart failure, and valvular heart disease [14]. AF is also prevalent in rheumatic mitral valve disease (RMVD) [15]. RMVD is a major cardiovascular disease in developing countries where rheumatic fever is less well controlled, and is also a major clinical risk factor for AF [16]. Remodeling of the right atrium (RA) and left atrium (LA) associated with AF in RMVD patients are well established [17]. Studies have shown that miRNAs may be involved directly or indirectly in AF by modulating atrial remodelling [18]. Recently, the AFassociated miRNAs were respectively found in RA [19] and LA [20] from RMVD patients.
Morphological [21] and electrophysiological [17] differences have been demonstrated between the RA and LA, which at least in part, may reflect different mechanisms involved in AF between the RA and LA [22]. Thus, it is not surprising that AF-associated miRNAs of the RA may differ from those of the LA. However, most studies of AF-associated miRNAs focus on the RA; and the potential difference of AF-associated miRNAs between RA and LA are still unknown.
Thus, the aim of this study was to analyze miRNA expression profiles to compare the potential differences of AF-associated miRNAs in the RA and LA from RMVD patients who were either in an sinus rhythm (SR) or AF group.

Methods
Approval was obtained from the human ethics committee of the first affiliated hospital of Sun Yat-sen University. The investigation complied with the principles that govern the use of human tissues outlined in the Declaration of Helsinki. All patients gave informed consent before participating in the study.

Human tissue preparation
Tissue samples from the right atrial appendage (RAA) and left atrial appendage (LAA) were obtained from18 RMVD patients. 8 patients were in SR group and they did not have a history of AF; 10 patients were in AF group and they had documented arrhythmia for more than six months before surgery. The tissue samples were obtained at the time of the mitral valve replacement surgery, immediately snap frozen in liquid nitrogen, and stored at −80°C until used. The diagnosis of AF was made based on medical records and 12-lead electrocardiogram (ECG) findings. Patients with SR had no history of using antiarrhythmic drugs and were screened to ensure that they had never experienced AF [23]. Pre-operative color Doppler echocardiography was performed routinely on the patients. Preoperative functional status was recorded according to the New York Heart Association (NYHA) classifications.

RNA isolation
Total RNA was extracted from human tissue samples (RAA and LAA) using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. The RNA quality of each sample was determined using an Agilent 2100 Bioanalyzer (Agilent Technologies; Santa Clara, CA, USA) and the sample was immediately stored at −80°C.

MiRNA microarray processing and analysis
The miRNA microarray was processed by LC Sciences (Houston, TX, USA) as described previously [20]. In brief, the assay utilized 2 to 5 μg total RNA sample. The total RNA was size fractionated using a YM-100 Microcon centrifugal filter (Millipore, Billerica, MA) and RNA sequences with < 300 nt were isolated. These small RNA were then extended at 3′-end with a poly(A) tail using poly(A) polymerase, followed by ligation of an oligonucleotide tag to the poly(A) tail for later fluorescent dye staining. Hybridization was performed overnight on a μParaflo™ microfluidic chip using a micro-circulation pump (Atactic Technologies, Inc, Houston, TX). Each microfluidic chip contained detection probes and control probes. The detection probes were made in situ by photogenerated reagent (PGR) chemistry. These probes consisted of a chemically modified nucleotide coding sequence complementary to the target microRNA (all 1,921 human miRNAs listed in the Sanger's miRNAmiRBase, Release 18.0, http://microrna.sanger.ac.uk/sequences/) and a spacer segment of polyethylene glycol to extend the coding sequences away from the substrate. The hybridization melting temperatures were balanced by chemical modifications of the detection probes. Hybridization was performed using 100 μL of 6× SSPE buffer (0.90 M NaCl, 60 mM Na2HPO4, 6 mM EDTA, pH 6.8) containing 25% formamide at 34°C. Fluorescence labeling with tag specific Cy5 dye was used for after-hybridization detection. An Axon GenePix 4000B Microarray Scanner (Molecular Device, Union City, CA) was used to collect the fluorescent images, which were then digitized using Array-Pro Image Analysis software (Media Cybernetics, Bethesda, MD). Each miRNA was analyzed two times and the controls were repeated 4-16 times.
The miRNA microarray analysis was also performed at LC Sciences. The microarray data were analyzed by subtracting the background and then the signals were normalized using a locally weighed regression (LOW-ESS) filter as reported previously [24]. Detectable miR-NAs were selected based on the following criteria: signal intensity higher than 3× the background standard deviation and spot CV < 0.5 (where CV = standard deviation/signal intensity). When repeating probes were present on an array, a transcript was listed as detectable only if the signals from at least 50% of the repeating probes were above detection level. To identify miRNAs whose expression differs among the groups, statistic analysis was performed and the P-values of the t-test were also calculated. The ratio of two samples was calculated and expressed in log 2 scale (balanced) for each miRNA. The miRNAs were then sorted according to their differential ratios.

Quantitative real-time PCR (qRT-PCR) of miRNA expression
To validate the microarray results in the study, a stemloop qRT-PCR based on SYBR Green I was performed on differentially expressed miRNAs. The primers used are listed in Additional file 1. Total RNA was isolated using the TRIzol Reagent (Invitrogen) as previously described. A single-stranded cDNA for a specific miRNA was generated by reverse transcription of 250 ng of total RNA using a miRNA-specific stem-looped RT primer. Briefly, a reverse transcription reaction mixture contained 250 ng of total RNA, 0.5 μl of 2 μM stem-loop RT primer, 1.0 μl of 5× RT buffer, 0.25 μl of 10 mM each dNTPs, 0.25 μl of 40 U/μlRNase inhibitor and 0.5 μl of 200 U/μl M-MLV. An Eppendorf Mastercycler® (Eppendorf, Hamburg, Germany) was used to conduct the reverse transcription reaction at the following temperature conditions: 42°C for 60 min, 70°C for 15 min and finally held at 4°C.
Following the reverse transcription reaction, qRT-PCR was performed using an ABI PRISM® 7900HT sequencedetection system (Applied Biosystems, Foster City, CA, USA) with the Platinum SYBR Green qPCR SuperMix-UDG (Invitrogen). According to the manufacturer's instructions, the PCR reaction (a total of 20 μl) contained 0.5 μl of RT product, 10 μl of 2× SYBR Green Mix, 0.4 μl of ROX, 0.8 μl of 10 μM primer mix, and 8.3 μl of nuclease-free water. The reaction protocol was as follows: 95°C for 2 min, followed by 40 amplification cycles of 95°C for 15 s, and 60°C for 30 s.
In qRT-PCR analysis, the relative expression level for each miRNA was calculated using a comparative cycle threshold 2 -ΔΔCt method for three independent experiments [25]. RNU6B was used as an internal control for normalizing the results.

Target prediction and function analysis
We used the human miRNA information database miR-Focus (http://mirfocus.org/ Version 2.1) to predict potential human miRNA target genes. The miRFocus provides a full gene description and functional analysis for each target gene by combining the predicted target genes from other databases (TargetScan, miRanda, Pic-Tar, MirTarget and microT). The greater the number of databases predicting that a gene would be a target, the more likely the miRNA-mRNA interaction would be relevant [26]. In this study, we have included genes that were predicted by two or more databases. The miRFocus program also identifies miRNA-enriched pathways, incorporating those from the Kyoto Encyclopedia of Genes and Genomes (KEGG), Biocarta, and Gene Ontology (GO) databases, using Fisher's exact test.

Statistical analyses
All data are presented as mean ± standard deviation and analyzed by paired t-test. For microarray results, miRNAs with P-values < 0.1 and |log 2 ratio | > 0.5 were considered to be significantly differentially expressed, while P < 0.05 was considered statistically significant for qRT-PCR analysis.

Clinical characteristics of the SR and AF patients
Tissues from both RAA and LAA were obtained from each patient.There were no significant differences in terms of age, gender or NYHA classifications between the SR and AF groups. Pre-operative color Doppler echocardiography showed that the LA size of the patients with AF was significantly greater than patients with SR as previously reported [27]. There were no differences in the left ventricular ejection fraction (LVEF) between the groups (Table 1).  Figure 1A). Among these, 132 miRNAs were detected in all groups. A number of miR-NAs were expressed only in one of the four groups-6 in the SR-RAA group, 2 in the SR-LAA group, 40 in the AF-RAA group, and 19 in the AF-LAA group.
Among the 258 miRNAs, 178 miRNAs were expressed in SR patients (SR-RAA or SR-LAA), while 250 miRNAs were expressed in AF patients (AF-RAA or AF-LAA). AF resulted in greater expression of miRNAs than SR ( Figure 1B). A total of 235 miRNAs were expressed in RAA tissues (SR-RAA or AF-RAA), while 213 miRNAs were expressed in LAA tissues (SR-LAA or AF-LAA). RAA tissues had a larger number of miRNAs expressed ( Figure 1C).
For most of the detected miRNAs, the expression levels were low, which was evident by their low signal intensities (less than 500 units) during microarray analysis ( Figure 2). Of the 164 miRNAs detected in SR-RAA tissues, 88 miRNAs emitted signals less than 500 units, while only 8 miRNAs produced signals above 10,000 units. Of the 155 miRNAs detected in SR-LAA tissues, 86 miRNAs emitted signals less than 500 units, while only 5 miRNAs produced signals above 10,000 units. Of the 216 miRNAs detected in AF-RAA tissues, 106 miRNAs emitted signals less than 500 units, while only 7 miRNAs produced signals above 10,000 units. Of the 208 miRNAs detected in AF-LAA tissues, 129 produced signals below 500 units, while only 3 produced signals above 10,000 units.
The signal intensities of the miRNAs expressed in only one of the four groups were not high enough to consider them differentially expressed among groups, and hence were not be considered for further analysis.   The SR-RAA group expressed 164 miRNAs, while the AF-RAA group expressed 216 miRNAs. Along with the number of detectable miRNAs, the expression levels of these miRNAs were also significantly different. Statistical analysis showed that 65 of these miRNAs were significantly dysregulated between AF-RAA and SR-RAA. Among these AF-associated miRNAs, 28 were up-regulated and 37 were down-regulated (Table 2 and Figure 3A).

miRNA expression profiles changes associated with AF in LAA tissue
The SR-LAA group expressed 155 miRNAs, while the AF-LAA group expressed 208 miRNAs. Along with the number of detectable miRNAs, significant differences also existed in the expression levels of these miRNAs. Statistical analysis showed that 42 miRNAs were significantly dysregulated in AF-LAA relative to SR-LAA. Among these AF-  associated miRNAs, 22 were up-regulated and 20 were down-regulated (Table 3 and Figure 3B) Comparison of AF-associated miRNAs between RAA and LAA tissues A total of 84 AF-associated miRNAs were found (either in RAA or LAA tissues): 65 AF-associated miRNAs were found in RAA tissues, while 42 AF-associated miRNAs were found in LAA tissues. Among these, 23 AFassociated miRNAs were found both in RAA and LAA, while 45 AF-associated miRNAs were found only in RAA, and 19 AF-associated miRNAs were found only in LAA ( Table 4).

Comparing expression of AF-associated miRNAs in RAA and LAA tissues from RMVD patients with SR
We identified 10 AF-associated miRNAs using the microarray and qRT-PCR techniques. These AF-associated miR-NAs had different tissue distributions between the RAA and LAA (Table 4, Figure 5). We wanted to know whether expressions of these miRNAs were different between the RAA and LAA tissues based on SR status. So we next compared expression levels of these miRNAs between the SR-RAA and SR-LAA groups. The expression levels of most of the AF-associated miRNAs were identical between RAA and LAA in SR patients, with the exception of miR-23b-3p and miR-125b-5p ( Figure 6A). miR-23b-3p was more highly expressed in the SR-LAA group than in SR-RAA, and miR-125b-5p was more highly expressed in the SR-RAA group. Moreover, we compared the expression level of miR-23b-3p and miR-125b-5p among four groups (i.e. SR-RAA, SR-LAA, AF-RAA, and AF-LAA). In LAA tissues, the expression level of miR-23b-3p was down-regulated in the AF group relative to the SR group. In RAA tissues, the expression level of miR-23b-3p did not change between the SR group and AF group ( Figure 6B). Meanwhile, a change in miR-125b-5p expression between SR and AF groups occurred only in RAA tissues, where miR-125b-5p was more highly expressed in SR group ( Figure 6C).
The biological function and potential functional pathways of each putative gene target were classified using the GO term and KEGG pathway. Since every gene is associated with many GO terms and KEGG pathways, the significant GO term (see Additional file 3) and KEGG pathway (see Additional file 4) for each miRNA were identified using Fisher's exact test. Table 6 gives a few representative KEGG pathways for the putative target genes of the validated miRNAs as predicted by the miR-Focus. The pathway analysis suggested that these miR-NAs may potentially contribute to AF.

Discussion
Two recent studies investigating expression profiles of miRNA in mitral stenosis patients found the AFassociated miRNAs in RA [19] and LA [20], respectively, many of which were also found in our study. However, the studies investigated the alterations of miRNA expression profiles in relation to AF only in RA or LA tissue and they could not compare the potential differences of AFassociated miRNAs between the RA and LA. A recent study investigated changes in miRNA expression profiles in patients with valvular heart disease in relation to AF both in RA and LA tissue [28] and found the AF-associated miR-NAs only in RA; the lack of detectable AF-associated miR-NAs in LA may be partially due to lack of tissue availability; and the study also could not compare the potential differences of AF-associated miRNAs between the RA and LA. Thus, our study is the first to compare the potential differences of AF-associated miRNAs in the RA and LA from RMVD patients. We found that the development of AF in RMVD patients was associated with significant changes in miRNA expression in both RAA and LAA tissues, and these AF-associated miRNAs had different distributions in RAA and LAA. A total of 23 AF-associated miRNAs were both in RAA and LAA, while 45 AF-associated miRNAs were only in RAA, and 19  AF-associated miRNAs were only in LAA. The differential distributions of these AF-associated miRNAs may reflect different miRNAs mechanisms in AF between the RA and LA.
The main limitation of this study was the small number of patients included. This was due, in part, to the difficulty of finding RMVD patients with SR. Second, We could not conduct experiments to modulate miRNA levels in native human tissues. Hence, the evidence presented here is not a direct cause and effect relationship. Furthermore, the exact targets and pathways Figure 5 Validation of the miRNA microarray data using qRT-PCR. RNU6B was used as an internal control for normalizing the results. Data are reported as mean ± standard deviation for three independent experiments. Statistically significant differences between the two groups are indicated by *P < 0.05, compared with SR-RAA (A) or SR-LAA (B).
of AF-associated miRNAs causing AF in RMVD patients remain elusive and deserve further investigation [19]. Finally, the patients included in this study were a specific cohort with preserved systolic left ventricular function and little comorbidity; they were undergoing mitral valve replacement surgery. Changes identified in this population may not be representative of other cohort of populations [32].

Conclusions
We have found the different distributions of AF-associated miRNAs in the RAA and LAA from RMVD patients. This may reflect different miRNA mechanisms involved in AF between the RA and LA. These findings may provide new insights into the underlying mechanisms of AF in RMVD patients and provide potential novel mechanism-based therapeutic strategies for AF. Figure 6 Comparison of validated AF-associated miRNAs expression level between RAA and LAA in SR status. Data are reported as mean ± standard deviation for three independent experiments, *P < 0.05, compared with SR-RAA (A,B,C); **P < 0.05, compared with SR-LAA (B) or SR-RAA (C).