Plasma Circular RNAs, Hsa_circRNA_025016, Predict Postoperative Atrial Fibrillation After Isolated Off‐Pump Coronary Artery Bypass Grafting
Background Circular RNAs (circRNAs) are pervasively expressed in highly divergent eukaryotes and are stable in body fluids. However, the link between circRNAs and onset of atrial fibrillation (AF) has not previously been investigated. We aimed to identify plasma circRNAs that are able predict AF after cardiac surgery.
Methods and Results Plasma circRNA expression profiles were investigated in a total of 769 patients with or without AF after isolated off‐pump coronary artery bypass grafting. First, a circRNA microarray was used to screen 13 617 human circRNAs in plasma samples from patients in the discovery cohort (n=30). A quantitative polymerase chain reaction assay was then applied to evaluate the expression of 9 selected circRNAs in the training cohort (n=365). This approach revealed that hsa_circRNA_025016 was upregulated in patients with new‐onset AF with a high diagnostic accuracy as assessed by the area under the receiver operating characteristic curve (=0.802). Additionally, a satisfactory diagnostic performance of hsa_circRNA_025016 was found in the validation cohort (n=284). Furthermore, Kyoto Encyclopedia of Genes and Genomes biological pathway analysis indicated that hsa_circ_025016 participated in melanogenesis, insulin secretion, and the thyroid hormone signaling pathway. A positive correlation between hsa_circ_025016 and fasting blood glucose was also identified in both cohorts.
Conclusions Hsa_circ_025016 is a potential biomarker for predicting new‐onset AF after isolated off‐pump coronary artery bypass grafting.
What Is New?
For the first time, this study revealed the high expression of circular RNA (hsa_circRNA_025016) in plasma before surgery predicted new‐onset atrial fibrillation after isolated off‐pump coronary artery bypass grafting.
What Are the Clinical Implications?
The present study demonstrates that this plasma hsa_circ_025016 holds potential as a biomarker for the prediction of postoperative atrial fibrillation and could enable us to more accurately target patients who are at risk and thereby improve the effects of prophylactic treatments.
Pervasive expression of circular RNAs (circRNAs) in highly divergent eukaryotes was discovered in 2012.1 Recently, thousands of human circRNAs have been identified using molecular biology strategies coupled with new bioinformatic approaches.2 CircRNAs are expressed in tissue‐specific and developmental stage‐dependent manners.3 The dysregulations of circRNAs contribute to many diseases because they act as microRNA target decoys, RNA binding protein sponges, and transcriptional regulators.4 Additionally, circRNAs are highly stable because of a resistance to debranching enzymes and RNA exonucleases. Therefore, circRNAs possess distinct advantages and may be useful as new biomarkers for diagnosis, prognosis, and therapeutic response prediction.2
Postoperative atrial fibrillation (PoAF) is one of the most common complications following coronary artery bypass grafting (CABG).5 The reported incidences of atrial fibrillation (AF) after CABG range from 15% to 50%. PoAF is associated with significantly increased morbidity, mortality, and total treatment costs.6 Although potential biomarkers of PoAF, such as mtDNA copy numbers and microRNAs,7, 8 have been actively researched, the diagnostic use of plasma circRNAs for PoAF has never been investigated. Our study screened the plasma circRNAs expression profiles (13 617 circRNAs) of patients with and without PoAF. The selected circRNAs that were found to be associated with PoAF were then tested in 2 additional independent populations of patients who underwent CABG.
Because of privacy, the data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure.
Study Design and Patients
Seven hundred sixty‐nine blood samples were collected from the Department of Cardiovascular Surgery of the General Hospital of Shenyang Military Area Command from March 1, 2015 to June 31, 2016. All patients underwent off‐pump CABG. The inclusion criteria, surgical procedure, and monitoring of AF onset were performed as described previously.8 Patients’ blood samples were collected within 1 week before surgery. Blood samples from 15 healthy donors were collected from the Health Examination Center of the General Hospital of Shenyang Military Area Command. All samples in this study were collected after informed consent was obtained from the participants, and all procedures were conducted according to an established protocol that was approved by the Ethics Committee of the General Hospital of Shenyang Military Area Command. The samples were allocated to 3 phases in chronological order (Figure 1A).
Potential associations between detectable circRNAs in EDTA‐plasma and the onset of PoAF were analyzed in a retrospective case‐controlled clinical trial (ClinicalTrials.gov, NCT02807532). Overall, 120 patients who underwent off‐pump CABG were recruited to this study, and the incidence of PoAF was 16.7% (20/120). In this cohort, 15 patients who developed PoAF were selected as case subjects, and 15 non‐PoAF patients with matched characteristics (ie, age, sex, smoking, left atrial diameter, left ventricle end‐diastolic volume, and the use of statins before surgery) were selected as control subjects. A circRNA microarray was used to screen 13 617 human circRNAs in these 15 pairs of case–control samples. The microarray information was submitted to the Gene Expression Omnibus under the accession number GSE97455.
The 9 circRNAs discovered via the microarray analyses were first tested by quantitative real‐time polymerase chain reaction in an independent cohort of plasma samples from 365 consecutive participants who were recruited from June 2015 to December 2015. Of these circRNAs, 6 were consistently amplified in all individual samples. Two circRNAs were differentially expressed between the PoAF and non‐PoAF groups in this cohort. One of the 2 circRNAs exhibited the greatest association with PoAF.
The optimal mean cutoff generated from the training phase was applied to an independent cohort of 284 consecutive participants who were included from January 2016 to June 2016 to evaluate the diagnostic performance of the circRNA (hsa_circRNA_025016).
Plasma Preparation and RNA Isolation
For the plasma preparation, peripheral blood (3–4 mL) was drawn into EDTA tubes. Within 30 minutes, each tube was subjected to centrifugation at 1000 g for 15 minutes. Next, 1 mL plasma was transferred to a 1.5‐mL Eppendorf tube and centrifuged at 11 000 g for 10 minutes to pellet any remaining cellular debris. Each EDTA tube can isolate nearly 2 mL plasma. Subsequently, the supernatants were transferred to fresh tubes and stored at −80°C. For the plasma samples, the total RNA was extracted using TRIzol LS according to the instructions from the manufacturer (Thermo Scientific, Waltham, MA). The concentrations were quantified with a NanoDrop 1000 Spectrophotometer (NanoDrop Technologies, Waltham, MA). The integrity of the RNA was assessed by electrophoresis on a denaturing agarose gel.
The sample preparation and microarray hybridization were performed based on the Arraystar's standard protocols. The microarray contained probes for 13 617 human circRNAs. Briefly, the total RNAs were digested with Rnase R (Epicentre, Madison, WI) to remove linear RNAs and enrich the circRNAs. Then, the enriched circRNAs were amplified and transcribed into fluorescent cRNA utilizing a random priming method (Arraystar Super RNA Labeling Kit; Arraystar, Rockville, MD). The labeled cRNAs were hybridized onto the Arraystar Human circRNA Array (8×15 K). After washing the slides, the arrays were scanned with an Agilent Scanner G2505C (Arraystar, Rockville, MD).
Quantitative Real‐Time Polymerase Chain Reaction
To test the candidate circRNAs that were acquired on the microarrays, circRNA detection was next validated by real‐time polymerase chain reaction. For this purpose, the isolated RNAs were reverse transcribed with random primers. Specific circRNAs were amplified using the primers listed in Table 1. The assays for the following circRNAs were first performed on 365 samples from 9 candidates: hsa_circRNA_404686, hsa_circRNA_000367, hsa_circRNA_001729, hsa_circRNA_100790, hsa_circRNA_104270, hsa_circRNA_030162, hsa_circRNA_100789, hsa_circRNA_025016, and hsa_circRNA_102049. The expression level of 18S RNA was used as a stable endogenous control for normalization. All assays were performed in triplicate.
Annotation and Functional Prediction for hsa_circ_025016
To explore the targeted miRNA profile, the miRNA support vector regression algorithm was used to score and rank the efficiencies of the predicted miRNA targets. Hsa_circ_025016 was used as a seed to enrich a circRNA‐miRNA‐gene network according to analyses with TargetScan (http://www.targetscan.org/) combined with miRanda (http://www.microrna.org/). The biological pathways were defined with the Kyoto Encyclopedia of Genes and Genomes.
In the discovery phase, paired t tests were performed to identify the circRNAs that were differentially expressed in the PoAF and non‐PoAF groups. Multiple comparison corrections were performed by calculating the Benjamini‐Hochberg false discovery rate. Filtering of all the circRNAs with fold changes ≥4 and false discovery rate values <0.05 yielded 9 circRNAs that were differentially expressed between the PoAF and non‐PoAF groups. The training cohort (365) was divided into training data and validation data via 10‐fold cross‐validation method using the R package caret. The cutoff value, sensitivity, specificity, and area under the receiver operating characteristic curve of each training data were estimated by roc () function of R package Daim. The sensitivity and specificity of each validation data were calculated based on the corresponding cutoff of each training data, while the sensitivity and specificity of validation cohort (284) were calculated based on the mean cutoff of training data, both with a diagnostic test 4‐fold table. Based on the expression of hsa_circ_025016 and the incidence of PoAF (16.7%) in discovery phage, at an α level of 5% and a power of 90%, 15 patients of PoAF and 75 patients of no‐PoAF would be required. Therefore, sample size in training and validation cohorts was sufficient. All statistical calculations were performed using SPSS 18.0 software (IBM, Chicago, IL). In all statistical analyses, the circRNAs levels were log‐transformed using the base 2 logarithm to account for the skewnesses of their distributions. Continuous variables were compared using unpaired Student t tests. Categorical variables were compared using χ2 tests.8 A logistic regression was used to select the diagnostic circRNA markers based on the training data set. P<0.05 was considered significant in all comparisons.
Levels of Circulating circRNAs Were Altered in Patients Who Developed PoAF
The characteristics of the 120 patients included in this study are summarized in Table 2. Because of the relatively small size of the cohort and low incidence of AF after surgery, none of the characteristics were found to be associated with PoAF. In consideration of our previous study and other clinical studies, patients with and without PoAF were matched for age, sex, smoking, left atrial diameter, left ventricle end‐diastolic volume, and the use of statins before surgery. The characteristics of the 30 patients are presented in Table 3. In an attempt to study alterations in the circulating circRNAs in the PoAF patients, microarray analyses were performed on the RNA from the plasma of these 30 patients.
As illustrated in Figure 1B, 31 circRNAs were specifically dysregulated (24 circRNAs were upregulated, and 7 circRNAs were downregulated with the fold changes ≥2 and FDR (false discovery rate) <0.05, Table 4) in patients with PoAF compared with who did not develop AF. Filtering of all of the differentially expressed circRNAs for changes ≥4‐fold yielded 9 circRNA candidates (hsa_circRNA_404686, hsa_circRNA_000367, hsa_circRNA_001729, hsa_circRNA_100790, hsa_circRNA_104270, hsa_circRNA_030162, hsa_circRNA_100789, hsa_circRNA_025016, and hsa_circRNA_102049) for further testing via quantitative real‐time polymerase chain reaction.
Hsa_circRNA_025016 Was Elevated in Patients With PoAF
The characteristics of the training and validation cohorts’ participants are presented in Tables 5 and 6. The 9 candidate circRNAs discovered in the microarray analyses were first tested with quantitative real‐time polymerase chain reaction using an independent cohort (training cohort) of 365 plasma samples. Six of the 9 circRNAs were stably detected in the plasma samples. Only 2 circRNAs—hsa_circRNA_025016 and hsa_circRNA_404686—were significantly upregulated and predicted PoAF in patients after CABG (odds ratio, 1.355 95% CI [1.250–1.468], P<0.0001 and 1.098 95% CI [1.009–1.195], P=0.03, Figure 1C). Because the levels of hsa_circRNA_404686 were positively correlated with hsa_circRNA_025016 (Figure 1D), and hsa_circRNA_025016 exhibited the strongest association with PoAF, hsa_circRNA_025016 was selected for further analyses.
Next, we investigated the expression of hsa_circRNA_025016 in the training and validation cohorts. We found that hsa_circRNA_025016 was elevated in patients with coronary heart disease compared with the healthy controls. Additionally, an increase in the hsa_circRNA_025016 level was observed in patients with PoAF compared with those without PoAF (Figure 2A and 2B).
Hsa_circRNA_025016 as a Prognostic Indicator for PoAF
The means and 95% CIs of the cutoff, sensitivity, and specificity were calculated via a 10‐fold cross‐validation method using the R package caret. The mean cutoff for 10 training data set was −12.118, and we considered this to be the optimum cutoff. The area under the receiver operating characteristic curve was 0.802, and the 95% confidence interval was 0.798 to 0.806. For the training data, the sensitivity was 0.794% (0.777–0.811%), and the specificity was 0.776% (0.763–0.789%); these results indicated a good predictive effect (Table 7).
The mean cutoff of −12.118 was used to evaluate the predictive effect in an independent validation cohort set (284 plasma samples). In the validation cohort, patients were divided into a high hsa_circRNA_025016 expression group and a low hsa_circRNA_025016 expression group according to the cutoff value. The predicted probability was evaluated with the χ2 test. This cutoff value was able to predict PoAF in the independent cohort (Pearson χ2=56.6, P<0.0001, sensitivity=73.52% and specificity=77.83%, Table 8).
Prediction and Annotation of the hsa_circRNA_025016 Targeted miRNA‐mRNA Network
We assumed that hsa_circRNA_025016 exerts its effects by acting as a miRNA sponge. The miRNA support vector regression algorithm was used to score and rank the efficiency of the predicted miRNA targets. Five miRNAs were identified as conserved between the TargetScan and miRanda predictions, including hsa‐miR‐6882‐3p, hsa‐miR‐4632‐5p, hsa‐miR‐3972, hsa‐miR‐4651, and hsa‐miR‐7974 (Figure 3). To explore the possible mechanism of hsa_circRNA_025016, Kyoto Encyclopedia of Genes and Genomes pathway analysis was utilized based on the predicted results. The results suggested that melanogenesis, insulin secretion, and the thyroid hormone signaling pathway may be modulated by hsa_circRNA_025016 (Figure 4A).
To explore the effect of hsa_circRNA_025016 on the insulin secretion pathway, the fasting blood glucose levels were compared according to the hsa_circRNA_025016 expression levels. As presented in Tables 9 and 10, the fasting blood glucose levels were significantly higher in patients with high levels of hsa_circRNA_025016 expression in both cohorts. Moreover, hsa_circRNA_025016 exhibited significant linear correlations with fasting blood glucose in the training and validation cohorts (Figure 4B and 4C).
PoAF is related to increased morbidity and mortality after cardiac surgery and is an independent risk factor for poor long‐term outcomes.9 However, poor understandings of the underlying mechanism and predisposing factors have limited the possible strategies for preventive preoperative measures. In the present study, we performed a wide screen to identify circRNAs that are differentially present in the plasma of patients with and without PoAF. Fifteen of the circRNAs were distinguished through hierarchical clustering analysis. When tested in the training cohort, at least 3 of these circRNAs candidates were found to be significantly upregulated in PoAF patients, and the strongest association was found for hsa_circRNA_025016. Surprisingly, the cutoff value for hsa_circRNA_025016 obtained from the training data resulted in high accuracy in the diagnosis of PoAF in the independent validation cohort set.
The study by Harling et al7 demonstrated that the levels of circulating and atrial tissue microRNAs are potential markers of PoAF. Other studies have found dysregulations of microRNAs in AF.10, 11 However, one of the main mechanisms of circRNAs is acting as inhibitors of miRNA that function by binding (“sponging”) a specific miRNA. Furthermore, circRNAs are resistant to the exonuclease RNase R because of the lack of free ends; thus, circRNAs possess distinct advantages as new biomarkers for diagnoses, prognoses, and therapeutic response predictions. Moreover, circRNAs are a special class of noncoding RNAs that are abundant in body fluids. Therefore, we believe that circRNAs represent a class of highly stable biomarkers for the prediction of PoAF.
Hsa_circ_025016 was selected and further tested in 2 independent larger cohorts to demonstrate its high diagnostic value as a biomarker for PoAF. However, the mechanism of hsa_circ_025016 has not yet been demonstrated. To our surprise, the host gene symbol is calcium voltage‐gated channel subunit α1 C (CACNA1C), which encodes an α‐1 subunit of a voltage‐dependent calcium channel. Loss‐of‐function mutations in CACNA1C have been reported in short QT syndrome.12
Another mechanism of circRNAs is acting as miRNA sponges and regulating the circRNA‐miRNA‐mRNA network. Therefore, we predicted the top 5 candidate miRNA binding targets according to the miRNA support vector regression scores and gained further insight into the functions of hsa_circ_025016. Kyoto Encyclopedia of Genes and Genomes pathway analysis was utilized based on the predicted results from the TargetScan and miRanda analyses. Interestingly, melanogenesis, insulin secretion, and the thyroid hormone signaling pathways were found to be the potentially affected pathways. Tatsuishi et al reported that blood glucose is a very strong predictor of PoAF after CABG.13 More importantly, we found that fasting blood glucose was increased in patients with high hsa_circ_025016 expression levels. Moreover, hsa_circ_025016 expression levels were positively correlated with fasting blood glucose in patients before surgery. These findings suggest that modulating insulin secretion may be one of the functions of hsa_circ_025016. Kirchhof et al14 found that the dysregulation of PITX2 expression could be responsible for susceptibility to AF. Expression array analyses have also suggested that melanogenesis is affected by the reduced expression of Pitx2c. Intriguingly, melanocyte‐like cells are present in mouse and human hearts that are predisposed to AF.15 Additionally, melanocortin can exert electrophysiological effects and stimulate the expression of interleukins.16 A recent review demonstrated that subclinical hyperthyroidism is associated with increased coronary heart disease events, mortality, and the incidence of AF.17, 18 Therefore, thyroid hormone may be another potential target that is affected by hsa_circ_025016. However, the underlying mechanisms by which hsa_circ_025016 regulates AF via affecting melanogenesis, insulin secretion, and the thyroid hormone signaling pathway need to be explored in future studies.
In conclusion, we found a plasma circRNA in a large number of participants that is unregulated in patients who developed PoAF when compared with non‐PoAF patients with a high degree of accuracy. Our study demonstrates that this plasma hsa_circ_025016 holds potential as a biomarker for the prediction of PoAF and could enable us to more accurately target patients who are at risk and thereby improve the effects of prophylactic treatments.
Sources of Funding
This work was supported by grants from the National Natural Sciences Fund Project of China (81500315), the Postdoctoral Science Foundation Grant of China (2016M593029 and 2014M562573), the National Scientific Research Foundation of Liaoning province in China (2015020449), and the Doctoral Scientific Research Foundation of Liaoning province in China (201501026).
We thank Ning Li, Lihui Wang, and Tong Liu for collecting the blood samples. We thank the Kang Chen Company (Shanghai, China) for the microarray analyses and their excellent technical assistance. We thank Prof. Song Wan (Division of Cardiothoracic Surgery, Department of Surgery, Prince of Wales Hospital, The Chinese University of Hong Kong) and Dr Jingyi Sheng (Department of Medicine & Therapeutics, Li Ka Shing Institute of Health Sciences, The Chinese University of Hong Kong) for helpful discussion and language editing.
- ↵Chen Y, Li C, Tan C, Liu X. Circular RNAs: a new frontier in the study of human diseases. J Med Genet. 2016;53:359–365.
- ↵Guo Y, Luo F, Liu Q, Xu D. Regulatory non‐coding RNAs in acute myocardial infarction. J Cell Mol Med. 2017;21:1013–1023.
- ↵Ebbesen KK, Kjems J, Hansen TB. Circular RNAs: identification, biogenesis and function. Biochim Biophys Acta. 2016;1859:163–168.
- ↵Amar D, Shi W, Hogue CW Jr., Zhang H, Passman RS, Thomas B, Bach PB, Damiano R, Thaler HT. Clinical prediction rule for atrial fibrillation after coronary artery bypass grafting. J Am Coll Cardiol. 2004;44:1248–1253.
- ↵El‐Chami MF, Kilgo P, Thourani V, Lattouf OM, Delurgio DB, Guyton RA, Leon AR, Puskas JD. New‐onset atrial fibrillation predicts long‐term mortality after coronary artery bypass graft. J Am Coll Cardiol. 2010;55:1370–1376.
- ↵Zhang J, Xu S, Xu Y, Liu Y, Li Z, Zhang Y, Jin Y, Xue X, Wang H. Relation of mitochondrial DNA copy number in peripheral blood to postoperative atrial fibrillation after isolated off‐pump coronary artery bypass grafting. Am J Cardiol. 2017;119:473–477.
- ↵Goren Y, Meiri E, Hogan C, Mitchell H, Lebanony D, Salman N, Schliamser JE, Amir O. Relation of reduced expression of MiR‐150 in platelets to atrial fibrillation in patients with chronic systolic heart failure. Am J Cardiol. 2014;113:976–981.
- ↵Wei XJ, Han M, Yang FY, Wei GC, Liang ZG, Yao H, Ji CW, Xie RS, Gong CL, Tian Y. Biological significance of miR‐126 expression in atrial fibrillation and heart failure. Braz J Med Biol Res. 2015;48:983–989.
- ↵Tatsuishi W, Adachi H, Murata M, Tomono J, Okonogi S, Okada S, Hasegawa Y, Ezure M, Kaneko T, Ohshima S. Postoperative hyperglycemia and atrial fibrillation after coronary artery bypass graft surgery. Circ J. 2015;79:112–118.
- ↵Kirchhof P, Kahr PC, Kaese S, Piccini I, Vokshi I, Scheld HH, Rotering H, Fortmueller L, Laakmann S, Verheule S, Schotten U, Fabritz L, Brown NA. PITX2c is expressed in the adult left atrium, and reducing Pitx2c expression promotes atrial fibrillation inducibility and complex changes in gene expression. Circ Cardiovasc Genet. 2011;4:123–133.
- ↵Levin MD, Lu MM, Petrenko NB, Hawkins BJ, Gupta TH, Lang D, Buckley PT, Jochems J, Liu F, Spurney CF, Yuan LJ, Jacobson JT, Brown CB, Huang L, Beermann F, Margulies KB, Madesh M, Eberwine JH, Epstein JA, Patel VV. Melanocyte‐like cells in the heart and pulmonary veins contribute to atrial arrhythmia triggers. J Clin Invest. 2009;119:3420–3436.
- ↵Collet TH, Gussekloo J, Bauer DC, den Elzen WP, Cappola AR, Balmer P, Iervasi G, Asvold BO, Sgarbi JA, Volzke H, Gencer B, Maciel RM, Molinaro S, Bremner A, Luben RN, Maisonneuve P, Cornuz J, Newman AB, Khaw KT, Westendorp RG, Franklyn JA, Vittinghoff E, Walsh JP, Rodondi N. Subclinical hyperthyroidism and the risk of coronary heart disease and mortality. Arch Intern Med. 2012;172:799–809.
- ↵Floriani C, Gencer B, Collet TH, Rodondi N. Subclinical thyroid dysfunction and cardiovascular diseases: 2016 update. Eur Heart J. 2017;0:1–7. Feb 27. [Epub ahead of print].