miR-144-3p and miR-425-5p are Negatively Associated With Atrial Fibrosis and Promote Atrial Remodeling by Targeting CREB1 in Atrial Fibrillation

Background: Progression of atrial brosis is vital for atrial remodeling in atrial brillation (AF). The main objective of this study was to explore the association between miR-144-3p and miR-425-5p, and atrial brosis as well as the resultant impact on atrial remodeling in AF. Methods and Results: Through miRNAs sequencing and qRT-PCR, we demonstrated that miR-144-3p and miR-425-5p were downregulated in plasma and atrial tissue among the patients who suffered from AF. We conrmed that the plasma’s miRNAs level was negatively correlated with left atrial brosis, which was evaluated with the left low voltage area using left atrial voltage matrix mapping. Catheter ablation restored decreased plasma miR-144-3p and miR-425-5p. Besides, ROC curve analysis revealed that the miRNAs not only differentiated AF from healthy control of AUC 0.928 and 0.921, respectively, but also discriminated persistent AF from paroxysmal AF of AUC 0.906 and 0.888, respectively. Furthermore, the downregulated miR-144-3p and miR-425-5p promoted atrial broblast proliferation by CCK-8. CREB1 was realized to be a common direct target for miR-144-3p and miR-425-5p by western blot and luciferase assay. Conclusions: Our ndings suggested that miR-144-3p and miR-425-5p could serve as novel atrial brosis biomarkers and hence contribute to atrial remodeling in AF.


Introduction
Atrial brillation (AF) is a common atrial arrhythmia, which increases the risk of stroke, congestive heart failure and myocardial infarction [1,2] . Radiofrequency catheter ablation has emerged as an important treatment option with favorable outcomes in patient with symptomatic, drug-resistant paroxysmal AF.
Atrial structural remodeling is the main pathological mechanism for the development of AF. Atrial brosis, which is induced by the proliferation of atrial broblast, myocytes apoptosis, and excessive deposition of intercellular collagen, is the most prominent pathological manifestation of structural remodeling [5] .
Previous studies have shown that the severity of atrial brosis is related to the prognosis of AF.
Homogenized left atrial brosis matrix can reduce the recurrence rate [6] . Therefore, seeking early biomarkers of atrial brosis and exploring the mechanism of structural remodeling have become necessary in the eld of AF.
The microRNA(miRNA) is a short non-coding RNA molecule with about 22 nucleotides. It regulates gene silencing by degrading target mRNA or inhibiting protein translation through complementary combining the 3'-untranslated regions (UTRs) of mRNA [7] . Previous study demonstrated that miRNAs are widely involved in regulating cardiovascular functions and pathogenesis [8,9] . Multiple studies have suggested that circulating miRNAs could be used as diagnostic biomarkers in various diseases [10,11] . Circulating miRNAs were also closely related to the procedure outcome in patients undergoing radiofrequency catheter ablation [12] , and radiofrequency catheter ablation can restore plasma altered miRNAs [13] .
Recently, several studies have shown that miRNA also contributed to the regulation of atrial remodeling in AF [14,15] .
The present study was aimed to screen the brosis biomarkers for AF and investigate its role in atrial structure remodeling. We found that miR-144-3p and miR-425-5p were downregulated in the plasma and atrial tissue in AF unlike in the healthy control. They were also negatively correlated with the left atrial brosis in persistent AF. Besides, it was discovered that catheter ablation could restore decreased plasma miR-144-3p and miR-425-5p. Downregulation of miR-144-3p and miR-425-5p could promote atrial broblast proliferation and subsequently result into brosis by targeting CREB1. These results suggest these miRNAs could be potential as non-invasive biomarkers for atrial brosis in AF.

Study population
This study included three cohorts. The rst was the miRNAs sequencing cohort which included 3 patients suffering from AF and 3 healthy controls. The second was the validation cohort which consisted a total of 80 AF (40 persistent AF and 40 paroxysmal AF) who were hospitalized in our department from October 2018 to October 2019. 80 genderand age-matched individuals without AF were enrolled as the control. 60 of them received radiofrequency catheter ablation. The third cohort included 6 valvular AF who underwent cardiac surgical operation, and 6 non-AF who undergoing cardiac surgical operation as the controls. Left atrial diameter, left ventricular ejection fraction and left ventricular end-diastolic diameter were assessed by transthoracic echocardiographic.
The study exclusion criteria of the rst and second cohort were age < 18 years or > 80 years, severe underlying structural heart disease, pregnancy, contraindication for anticoagulation, serious liver or renal dysfunction, LA diameter ≥ 50mm, and LA or LAA thrombosis. This study was approved by the Ethics Committees of the First People's Hospital of Yunnan Province. All the participants signed a written informed consent document. All of the procedures were done following the declamiceion of Helsinki and relevant policies in China.

Plasma and atrial tissue collection
Blood samples were collected in the morning with anti-EDTA tubes and processed within 1h. They were subjected to centrifugation at 2000g for 10 min, and then stored at -80℃ until further analysis. Rightatrial tissue specimens were collected after the right atrium was opened and preserved in liquid nitrogen for subsequent experiments.

qRT-PCR
A neasy mini kit (Qiagen, Germany) was used to isolate the total RNAs according to the manufacturer's instruction. Briefly, 1µg total RNA was reverse transcribed with Anchored-oligo (dT) 18 Primer for cDNA synthesis using Transcriptor First Strand cDNA Synthesis Kit (Roche, Switzerland) based the manufacturer's instruction. The polymerase chain reaction was performed in a 10μl reaction mixture on an ABI 7900 real-time PCR instrument (Applied Biosystems, USA). The mixture was heated for 1 minute at 95 • C for predenatumiceion, and then processed to 35-40 PCR cycles, which consisted denatumiceion at 95 • C for 15 seconds, annealing at 60 • C for 60 seconds, and extension at 95 • C for 15 seconds. The miR-16 served as an internal reference of plasma miRNAs. U6 and GAPDH were used as an internal reference of miRNA and mRNA for tissue and cells. 2 -ΔΔCt method was used to calculate the relative expression level.
The changes in miRNAs expression were calculated using ΔCt, whereby a lower ΔCt value meant a higher expression of miRNA.

miRNA Solexa sequencing
Solexa sequencing was carried out to analyze plasma miRNA expression from the rst cohort subjects. Through illumina sequencing, miRNA digitalization analysis was performed based on Hiseq4000 highthroughput sequencing. All the procedures were performed with respect to the manufacturer's guidelines.
At least 30 copies in either group and at least a two-fold difference in expression between the two groups was used to screen for miRNA. Differentially expressed miRNAs were de ned as the fold changes of 2 or higher and P-values of 0.05 or lower relative to the controls.

Left atrial voltage matrix mapping
Atrial brillation with indication for radiofrequency catheter ablation was enrolled after preoperative preparation. Local anesthesia and conscious sedation with fentanyl were used on all the study subjects. Two long sheaths (Swartz, SL1, 8.5F, St. Jude Medical, Minnesota, USA) were inserted into the femoral vein. Puncture of the atrial septum was performed twice. All the patients were subjected to the circumferential pulmonary vein isolation (PVI). Left atrial voltage matrix mapping was started in case the atrial brillation converted to sinus rhythm post-PVI, or the electrical cardioversion was performed before left atrial voltage matrix mapping. Left atrial voltage matrix mapping was performed under sinus rhythm with a pentary mapping catheter (pentary catheter, Biosense Webster, Diamond Bar, CA). It was created with a 3-dimensional mapping system (Carto-3, Biosense Webster), with the acquisition of 300 to 1000 points. The range of mapping voltage amplitude was set at 0.1-0.4 mV. The local electrogram that was lower than 0.4 mV, was considered as a low voltage area (LVA), and it represented the extent of left atrial brosis [16] . The extent of left atrial LVA was calculated as a percentage of the LVA to the left atrial surface area.

Isolation of atrial broblast
Using the classic surgery procedures, the mouse's heart was removed. The excess tissue was cut off. The atrial tissue was minced with scissors and digested at 37℃ in phosphate-buffered saline (PBS) that contained 0.1% trypsin (Gibco, BRL) and 0.05% type I collagenase (Gibco, BRL). After 8-10 cycles of digestion, for 6min each, the digestive solution was centrifuged at low-speed centrifugation for 5 minutes. The supernatant was discarded and the cells were suspended in DMEM (Gibco, BRL) containing 15% fetal bovine serum (Gibco, BRL). Percoll uid (Sigma, USA) was used to purify the cardiomyocytes and broblasts. Then the broblasts were appropriately diluted and cultured adherently for further analysis.

CCK-8
Cell counting kit-8 (CCK-8) was used to detect broblasts proliferation. The broblasts were seeded in 96well plates at a density of 2×10 3 /well. After transfection with miRNA for 48 hours, 10µl CCK-8 solution was added to a 90µl culture medium, and then incubated at 37°C in 5% CO2 for 2 hours. The optical density (OD) value was recorded on the 450 nm wavelength enzyme-label analyzer.

Luciferase report assay
Luciferase reporter assay was described in detail in our previous publication [17]. Brie y, DNA segments containing 3'UTRs of human CREB1 mRNA were inserted into EcoRI and XhoI sites of pcDNA3.0-EGFP (TaKaRa, China). HL-1 cells were cotransfected with 1.0µg re y luciferase reporter vectors containing the target sites, 100nM miR-144-3p, or miR-425-5p mimics or miR-control and 0.06µg Renilla luciferase control vectors using Lipofectamine3000 (Invitrogen, USA). The cells were harvested after 48 hours. The results were expressed as relative luciferase activity ( re y luciferase/Renilla luciferase). The mutation of the miRNAs binding sites for human CREB1 3'UTRs was generated using the Quick Change XL Site-Directed Mutagenesis Kit.

Western blot
Total protein extracts were obtained via lysis buffer containing protease inhibitor cocktail (Sigma). The proteins were transferred to polyvinylidene di uoride membranes and blocked in PBS containing a 5% fatfree milk powder. Then the antibodies against CREB1 (Abcam, UK) were used to detect the corresponding proteins. Goat anti-rabbit immunoglobulin conjugated to horseradish peroxidase (CST, USA) was used as the secondary antibody. It was scanned and photographed in the BioRAD gel imager. The relative expression of CREB1 was determined using the speci c bands densities in western blots, normalized to the density of GAPDH.

Statistical analysis
Data were presented as mean±SD or median [interquartile range] if their values were not normally distributed. Unpaired Student's t-tests or non-parametric Mann-Whitney U-test was used to compare the continuous variables. Categorical variables were presented as absolute or relative frequencies, and compared using the χ2 test. The correlation between miRNA and the relative parameter were evaluated using the Pearson's correlation coe cients. The valve was log-transformed to normalize their distribution before statistical analysis. Moreover, receiver operating characteristic (ROC) curves analysis was performed and the area under the curves (AUC) was calculated to evaluate the candidate miRNAs' diagnosis power. P-value of < 0.05 was considered statistically signi cant. SPSS 17.0 and GraphPad Prism7.0 software was used for all the statistical analyses and graph presentation.

Baseline characteristics of the subjects
The baseline demographic characteristics of the three cohorts were shown in Online Table1 to Table3. The distributed characteristics were similar between the two groups. 30 paroxysmal AF and 30 persistent AF received radiofrequency catheter ablation in the second cohort.
3.4 Veri cation of miRNAs 7 candidate miRNAs were detected in an independent cohort of 80 AF patients and 80 healthy controls by qRT-PCR. The miR-144-3p and miR-425-5p were downregulated in the AF group whereas hsa-miR-302 was up-regulated. This was consistent with the results in the rst miRNA sequencing cohort. The expression of the other 4 miRNAs was inconsistent with the sequencing results (Figure 2A-C). We found that the expression of miR-144-3p and miR-425-5p was downregulated in left atrial blood than in the peripheral blood. On the contrary, expression of miR-302 was upregulated in the left atrial blood than in the peripheral blood (Online Figure 1).

miRNAs expression in atrial tissue in AF patients and mice AF model
We established that the expression of miR-144-3p and miR-425-5p was decreased in the AF than in the SR whereas miR-302 was increased ( Figure 2D-E). This was in agreement with the plasma results. We also realized that the expression of miR-144-3p and miR-425-5p was lower in the left atrial tissue compared to the right atrial tissue, and miR-302 expression was increased in the left atrium tissue in mouse AF model (Online Figure 2). Therefore, these ndings implied that the expression level of miR-144-3p and miR-425-5p was lowered, while that of miR-302 was increased in left atrial in comparison to the right atrial tissue in AF.

miRNAs expression in different type of AF
To detect whether the differentially expressed miRNA was associated with AF type, the miR-302, miR-144-3p, and miR-425-5p were detected in 40 paroxysmal AF and 40 persistent AF, and results showed the expression level of miR-144-3p and miR-425-5p was lower in persistent AF than in paroxysmal AF ( Figures 2F-G). However, miR-302 expression difference was not signi cantly between the two different types of AF ( Figure 2H).

Comparison of plasma miRNA expression level in the pre-RFCA and post-RFCA
The plasma miR-144-3p and miR-425-5p level were detected in 20 AF patients during pre-RFCA (radiofrequency catheter ablation) and post-RFCA, and also compared with the control group. The relative plasma level of miR-144-3p increased in post-RFCA compared with pre-RFCA (the median and interquartile ranges of pre-RFCA vs. post-RFCA were 1.10:0.34 and 0.762:0.2963, respectively). The level of miR-425-5p also increased (the median and inter-quartile ranges of pre-RFCA vs. post-RFCA were 1.03:0.60 and 0.49:0.28, respectively). Besides, the plasma miR-144-3p and miR-425-5p expression level were not signi cantly different between post-RFCA and the control group (p=0.108 and 0.628, respectively) ( Figure 4D-F).

Effect of miR-144-3p and miR-425-5p on atrial broblast proliferation and brosis
Atrial broblast proliferation and brosis were the hallmark of atrial remodeling in AF. We investigated whether the aberrant expression of the miRNAs in uence broblast proliferation and brosis. The miR-144-3p and miR-425-5p inhibitor was transfected into the atrial broblast cell to downregulate the expression of miR-144-3p and miR-425-5p ( Figure 5A). We found downregulation of miR-144-3p and miR-425-5p promoted the atrial broblast proliferation ( Figure 5B-C), and the proliferation marker expression of Ki67 and Cyclin D ( Figure 5D). Besides, it also promoted the brosis marker expression of collagen I and collagen III (Figures 5E-F).

CREB1 was the common direct target of miR-144-3p and miR-425-5p
Two widely available mammalian target prediction programs (miRDB and TargetScan7.2 databases) were used to predict the target gene of the miRNAs. 645 and 97 targets of miR-144-3p and miR-425-5p respectively were indicated by miRDB and TargetScan7.2. 20 of them were among the common targets for miR-144-3p and miR-425-5p (Online Figure 4), including DIP2C, MAPK6, CPEB2, ZNF148, ATP1B1, CREB1, LCOR, SYNCRIP, CPEB1, IFFO2, SCAMP1, TM9SF3, AFF4, FST, ZFHX3, CREBZF, LARP4B, FBN1, GOLGA4, SMAD5. We then after analyzed the expression level of 20 genes in the atrial tissue in SR and AF (Primer showed in the Online Table4). We found only the expression of CREB1 was increased. Besides, FBN1 and ZFHX3 expression was decreased in AF than in SR ( Figure 6A). The expression of miRNA was negatively correlated with the target gene expression, on this basis, we speculated that CREB1 may be the target gene for miR-144-3p and miR-425-5p. More importantly, our results also showed the expression of miR-144-3p and miR-425-5p was negatively associated with CREB1 expression in the atrial tissue in AF ( Figures 6B-C). Finally, we found that inhibition of CREB1 expression could reverse the phenotype of down-regulation of miR-144-3p and miR-425-5p( Figure 6D-F).

Discussion
This study demonstrated that miR-144-3p and miR-425-5p were downregulated in the plasma as well as in atrial tissue in AF. And downregulation of miR-144-3p and miR-425-5p were negatively correlated with left atrial brosis and cntribured to atrial remodeling by targeting CREB1. Besides, catheter ablation could restore the decreased miR-144-3p and miR-425-5p in the plasma. This ndings suggested that miR-144-3p and miR-425-5p could serve as noninvasive early diagnostic biomarkers for AF as well as a predictive biomarkers for atrial brosis in AF.
Diagnosis of AF is dependent on the clinical feature and ECG monitoring characterized by disappeared P wave and irregular R-R interval [19] . However, a considerable proportion of AF is asymptomatic [20]. Hence, a sensitive and speci c non-invasive diagnostic technique should be sought to identify the asymptomatic atrial brillation during the early stage. Accumulating pieces of evidence suggest that miRNAs are pivotal regulators for cardiovascular functions. The circulating miRNAs could serve as a non-invasive diagnostic biomarker since they carry disease-speci c information [17,21] . Meanwhile, the miRNAs can severe as prognosis biomarker of various cardiovascular diseases due to the alter dynamically in different stages of disease [22] . In this study, we found that miR-144-3p and miR-425-5p were downregulated in plasma in AF patients. The expression of miR-144-3p and miR-425-5p was also lower in persistent atrial brillation than in paroxysmal atrial brillation. These results suggest that the miRNAs experience changes during different stages of AF. Besides, ROC curve analysis showed a lower level of miR-144-3p and miR-425-5p and higher level of miR-302 could not only differentiate AF from a healthy individual but also differentiate persistent AF from paroxysmal AF. The implication is that miR-144-3p and miR-425-5p could serve as a potential diagnostic biomarkers for AF.
Atrial brosis is an important hallmark for atrial structural remodeling, and hence it forms an important substrate for AF [5] , The extent of left atrial brosis was more serious in the persistent AF than paroxysmal AF, which was positively associated with AF prognosis [23,24]. Pulmonary vein isolation and scar homogenization were also associated with considerably better long-term outcomes than the pulmonary vein isolation alone in AF with severe left atrial scars [25] . On this basis, an early recognition and intervention of atrial brosis was the effective method to improve the prognosis of AF. The cardiac MRI and intraoperative left atrial voltage mapping are the main methods for evaluating atrial brosis extent [26] . However, these methods are either invasive or complex. As such, seeking noninvasive sensitivity brosis biomarkers, which can not only evaluate atrial brosis but also provide evidence for the selection of the appropriate procedures becomes apparent, so as to improve the prognosis. Our study established that the plasma miR-144-3p and miR-425-5p were negatively correlated with the left atrial brosis in persistent AF. We also discovered that catheter ablation could restore the decreased miR-144-3p and miR-425-5p. These results suggested that miR-144-3p and miR-425-5p may serve as sensitive prediction biomarkers for atrial brosis, thereby providing clues for assessing the pathogenic contribution of the miRNAs in AF. Furthermore, we found the expression of miR-144-3p and miR-425-5p was lower in the left atrial blood than in the peripheral blood. This implied that, the aberrant expression of miR-144-3p and miR-425-5p in the left atrial blood play a critical role in atrial brillation.
Atrial remodeling is the main pathological mechanism for the occurrence and persistence of AF [27] . Atrial brosis induced by the proliferation of atrial broblasts, cardiomyocytes apoptosis, and over deposition of the intercellular matrix, was the prominent pathological manifestation in structural remodeling [15] .
Atrial electrical remodeling characterized by the shortening of the effective refractory period and action potential duration, is another fundamental trigger for atrial brillation [28] . However, the prevention and treatment of atrial brillation are mainly focused on the improvement of atrial electrical remodeling. For instance, the antiarrhythmic drugs can only recover and prolong the effective refractory period of the atrium [29] . Radiofrequency catheter ablation normally focuses on the removal of the abnormal local excitations and blocks the conduction between the pulmonary vein and the left atrium. As such, it becomes necessary to explore the molecular mechanism of structural remodeling. Our results implied that the downregulation of miR-144-3p and miR-425-5p could promote broblast cell proliferation and hence cause brosis by targeting CREB1. This suggested that the inappropriate down-regulation of miR-144-3p and miR-425-5p in the atrial tissue and plasma could facilitate the occurrence and development of atrial brillation via atrial remodeling.
Previous studies have reported the correlation between miRNAs and atrial brillation. For example, The expression of miR-30d was upregulated in patients with persistent AF, and upregulation of miR-30d promoted atrial electrical remodeling by targeting KCNJ3 [30] . Besides, miR-133 and miR-590 were downregulated in canine AF model and contributed to AF development via regulating structure remodeling [31] . However, the relationship between miRNA and AF need to be further clari ed. In this study, we found two novelty miRNAs, which were associated with atrial brosis and contributed atrial structure remodeling.

Conclusions
The present study demonstrated that the expression levels of miR-144-3p and miR-425-5p decreased in the plasma in AF. However, catheter ablation could restore the declined miR-144-3p and miR-425-5p levels. Besides, miR-144-3p and miR-425-5p could serve as non-invasive biomarkers for atrial brosis and consequently regulate atrial remodeling by targeting CREB1.

Limitation
The study had two main limitations as follows: Firstly, the relationship between the candidate miRNAs and the recurrence of AF after radiofrequency catheter ablation was not assessed. Secondly, only the left atrial voltage matrix mapping was used to evaluate atrial b rosis, other methods such as cardiac MRI were not utilized [32] .
Abbreviations AF: Atrial brillation ROC: receiver operating characteristic Declarations Ethics approval and consent to participate: This study was approved by the Ethics Committees of the First People's Hospital of Yunnan Province. All the participants signed a written informed consent document. All of the procedures were done following the declamiceion of Helsinki and relevant policies in China.
Consent for publication: Consent for publication of all the data.
Availability of data and materials: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests Authors' contributions: Feiyu Wei did the clinical sample experiment and analyzed data, and was a major contributor in writing the manuscript. Li LV was responsible for cell experiment and analyzed data. Xiaohui Kuang was responsible for Left atrial voltage matrix mapping. Wenjun Ren was responsible for sample collection. Jie Fan was responsible for the design and revising manuscripts.All authors read and approved the nal manuscript.