Biomarkers for Predicting the Occurrence and Progression of Atrial Fibrillation: Soluble Suppression of Tumorigenicity 2 Protein and Tissue Inhibitor of Matrix Metalloproteinase-1

Background Soluble suppression of tumorigenicity 2 protein (sST2) and tissue inhibitor of matrix metalloproteinase (TIMP)-1 are involved in multiple pathogenic pathways, including cardiac remodeling, which is the main pathology of atrial fibrillation (AF). This study aims to investigate the previously unexplored relationship between the serum levels of sST2, TIMP-1, and AF. Methods This was a prospective cross-sectional study conducted at the Capital Medical University Affiliated Beijing Anzhen Hospital between June 2019 and July 2020, with a total of 359 participants. The clinical characteristics and laboratory results of the patients were compared, and multivariable ordinal logistic regression was used to evaluate the relationship between serum sST2, TIMP-1, and AF. Results The participants included 110 patients with sinus rhythm (SR), 113 with paroxysmal AF (the paroxysmal AF group), and 136 with persistent AF (the persistent AF group). It was found that the sST2 levels gradually increased in these three groups, from 9.1 (6.7–12.4 pg/ml) in the SR group to 14.0 (10.4–20.8 pg/ml) in the paroxysmal AF group and to 19.0 (13.1–27.8) pg/ml) in the persistent AF group (p  <  0.001). The multivariable ordinal logistic regression model for sST2 and TIMP-1 demonstrated that sST2 had an area under the receiver operating characteristic (ROC) curve (AUC) of 0.797 (95% confidence interval (CI) 0.749–0.846, p  <  0.001) and TIMP-1 had an AUC of 0.795 (95% CI 0.750–0.841, p=0.000). The multivariable ordinal logistic regression model for sST2 and TIMP-1 showed good discrimination between SR and AF, with an AUC of 0.846, and the addition of clinical factors, such as brain natriuretic peptide (BNP), left atrial diameter, age, and gender, to the biomarker model improved the detection of SR and AF (AUC 0.901). Conclusions In this cohort study, sST2 and TIMP-1 were associated with AF progression, independent of clinical characteristics and biomarkers. Soluble ST2 and TIMP-1 combined with age, elevated N-terminal-pro hormone BNP(NT-BNP), and an enlarged left atrium were able to demonstrate the progression of AF reliably.


Introduction
Atrial fbrillation (AF) is the most common cardiac arrhythmia in clinical practice, and it is a major risk factor for stroke, heart failure (HF), and other cardiovascular-related complications [1]. In the current study, the prevalence of AF in adults was between 2% and 4%, and the annual rates of paroxysmal AF progression to persistent AF ranged from <1% to 15%, up to 27%-36% in studies with a ≥ 10yearfollow-up [2,3]. In other words, AF begets AF and is irreversible. Te clinical outcome of patients with AF progression with regard to hospital admissions and major adverse cardiovascular events is worse compared with patients demonstrating no AF progression [4]. Catheter ablation is a widely accepted treatment for this type of arrhythmia, but the success rate is lower in patients with persistent AF than it is in those with paroxysmal AF [5].
Tere is increasing evidence that there is a link between oxidative processes and AF [6], and various infammatory markers and mediators have been independently linked to AF, suggesting a strong association between infammation and arrhythmia [7]. Infammatory mediators can refect changes in atrial electrophysiology and structural substrates, which lead to increased vulnerability to AF. Te recent guidelines for AF diagnosis and treatment have shown that serum N-terminal-pro hormone brain natriuretic peptide (NT-proBNP) and left atrial diameters (LADs) are more powerful biomarkers than other clinical variables [1]. However, there is also ongoing research into novel biomarkers in the diagnosis and prognosis of AF. Tere are data indicating that C-reactive protein (CRP) and other proinfammatory cytokine levels are higher in blood samples drawn from patients with AF than those from patients with sinus rhythm (SR) [1], and the levels decrease gradually after catheter ablation of persistent or long-lasting AF [8]. In addition, soluble suppression of tumorigenicity 2 protein (sST2) has not been restricted to infammation, but it is also expressed as a response to myocardial stress [9]. Tissue inhibitor of matrix metalloproteinase (TIMP)-1 is upregulated by infammatory factors, such as interleukin [10], and AF progression is associated with a gradual increase in matrix metallopeptidase 9 (MMP-9)/TIMP-1 [11]. However, the role of these novel biomarkers in the progression of AF is not yet clear [11,12], and neither is it clear whether serum sST2 and TIMP-1 can be detected and predict AF onset, especially the onset of paroxysmal AF. Clearly, an early diagnosis would be benefcial in the prevention of complications, such as HF and stroke.
Tis study aims to identify the clinical and laboratory variables that can predict the development of AF, ascertain whether the serum levels of sST2 and TIMP-1 are diferent between patients with SR and those with AF, observe whether sST2 and TIMP-1 are related to the progression of AF, and, if so, determine the cutof point of sST2 for predicting such progression.

Study Design.
In this cross-sectionalcase-control study, a review of the medical records of patients who were registered for catheter ablation to treat paroxysmal or persistent AF as well as the control group patients with SR was undertaken. Te data collected included clinical variables, laboratory test results, and blood sample results. Te study aimed to identify the clinical and laboratory variables differentiating the SR group, paroxysmal AF group, and persistent AF group and explore whether serum levels of sST2 and TIMP-1 could demonstrate AF progression.

Participants.
A total of 249 inpatients with AF who were admitted to Capital Medical University Afliated Beijing Anzhen Hospital between June 2019 and July 2020 were included in the study, along with 110 patients with SR, who constituted the control group. Te inclusion criteria were as follows: patients aged >18 years, patients clinically diagnosed with AF, and patients who signed an informed consent form. Te exclusion criteria were as follows: patients with a malignant tumor, infammation, or other end-stage disease. Te defnition of paroxysmal AF used in this study was that AF terminates spontaneously or with intervention within seven days of onset, whereas persistent AF is continuously sustained beyond seven days, including episodes terminated by cardioversion (drugs or electrical cardioversion) after seven or more days. Te diagnosis of paroxysmal or persistent AF was based on a 12-lead electrocardiogram and a 24-hour ambulatory electrocardiogram.

Data Collection.
Clinical and laboratory data were extracted from the medical records by two independent doctors. Te former included age, sex, and history of hypertension, coronary artery disease, diabetes, HF, and other similar indicators, and the laboratory data consisted of white and red blood cell counts, platelet counts, and hemoglobin (Hb) levels. In addition, the LAD, left ventricular endsystolic dimension, left ventricular end-diastolic dimension, and left ventricular ejection fraction were measured. Blood markers, namely, TIMP-1, sST2, highsensitivity (hs)-CRP, and NT-proBNP were measured using the same blood sample within six hours of blood sampling. Te blood sample was collected for paroxysmal and persistent patients during ongoing AF. Te CHA2DS2-VASc and HAS-BLED scores, for the assessment of stroke and bleeding, respectively, were also calculated for all the participants.

Statistical Analysis.
All statistical analyses were performed using SPSS 20.0 (IBM Corp., Armonk, NY, USA). Continuous data were expressed as mean ± standard deviation and analyzed using Student's t-test (for comparisons of two groups). Nonnormally distributed continuous data were described as median and interquartile range, and comparisons between the groups were made using the Wilcoxon rank sum test. Categorical variables were presented as a number (percentage) and analyzed using the chi-square test or Fisher's exact test and considered appropriate. Parameters with p < 0.05 in univariable ordinal logistic regression analysis were included in multivariate logistic regression analysis using the enter method. Multivariate ordinal logistic regression analysis was performed to assess the relationship between the serum levels of the biomarkers and the risk of AF, which was described using correlation coefcients and presented with 95% confdence intervals (CIs). To assess the discriminatory abilities of the biomarkers to predict AF, a receiver operating characteristic (ROC) curve was constructed, and the AUC was calculated. A value of p < 0.05 was considered statistically signifcant.

Baseline Characteristics.
A total of 359 participants were enrolled in the study between June 2019 and July 2020. Teir clinical characteristics and laboratory data are summarized in Tables 1 and 2, respectively, and a patient selection fowchart is shown in Figure 1.  Table 1). Te echocardiography results also found that left atrial anteroposterior diameter is associated with the progression of AF.
Te laboratory test data are shown in Table 2. Patients with AF had signifcantly higher levels of Hb than the SR group: the SR group, 144 (135-155) g/l; the persistent AF group, 152 (142-162.0) g/l, p < 0.001; and the paroxysmal AF group, 148 (139-160) g/l, p � 0.012. Te persistent AF group had the highest level of BNP among the three groups, and the diference was signifcant: the persistent AF group, 125 (87-215); the paroxysmal AF group, 67 (32-123); and the SR group, 32 (10-76.5), p < 0.001). Te serum level of hs-CRP was the highest in patients with persistent AF (1.2 [0.6-3.1] mg/l), and the diference between this group and the paroxysmal AF group (0.8 [0.5-1.9] mg/l) was signifcant (p � 0.022), as it was between the persistent AF group and the SR group (1 [0.45-1.9] mg/l, p � 0.0190. Te serum levels of mean cell volume and mean cell hemoglobin were higher in the persistent AF group than those in the paroxysmal AF and SR groups. However, the serum levels of platelets, total protein, and albumin in the AF group were lower than those in in the SR group. Te serum levels of sST2 and TIMP-1 were signifcantly higher in both the paroxysmal AF and persistent AF groups than those in the SR group. Te serum levels of sST2 increased signifcantly between the SR and the persistent AF group (SR group 9.1 [6.7-12.4], paroxysmal AF group 14 [10.4-20.8], p < 0.001, and persistent AF group 19 [13.1-27.8], p < 0.001). In addition, the serum level of TIMP-1 increased signifcantly from the SR group to the paroxysmal AF group and then to the persistent AF group as follows: SR group 67.4 (56.9-81.3) ng/ml; paroxysmal AF group 89.7 (70.6-115.5); and persistent AF group 127.3 (87.0-173.7), p < 0.001. Tese results indicated that the serum levels of TIMP-1 and sST2 were associated with the progression of AF.

Multivariate Ordinal Logistic Regression Analysis of Factors Associated with Atrial Fibrillation.
Te results of the univariate ordinal logistic regression analysis are shown in Table 3, which lists the variables that are independently associated with AF. All the independent risk factors associated with AF were included in the following multivariable ordinal logistic regression analysis. To identify the factors correlated with the progression of AF, a multivariable ordinal logistic regression analysis was performed. Te results of the multivariable ordinal logistic regression analysis are detailed in Table 4. TIMP-1, sST2, BNP, and LAD were associated with the progression of AF as their levels gradually increased starting from the SR group to the paroxysmal AF group and then to the persistent AF group ( Figure 2).
Te ROC curve for assessing the ability of these factors to predict AF progression is shown in Figures 3 and 4. Te ROC curve shows that sST2 and TIMP-1 were the best biomarkers for predicting the progression of AF ( Figure 3). Te AUC of sST2 for the progression of AF was 0.7973 (95% CI 0.749-0.8456; p < 0.001). According to the highest  Youden's index, the cutof value with the highest sensitivity and specifcity was 12.81 ng/ml of sST2 (78.4% and 75.4%, respectively). Te AUC for predicting AF (both paroxysmal and persistent AF) was 0.7954 (95% CI 0.813-0.922, p < 0.001). According to the highest Youden index, the cutof of TIMP-1 for predicting AF was 90.57 ng/ml, and sensitivity and specifcity were 63% and 94.7%, respectively. Both sST2 and TIMP-1 had a high predictive value for the progression of AF. However, although the serum level of sST2 had both high sensitivity and specifcity, the prediction value of TIMP-1 was even higher. Tus, a combination of sST2 and TIMP-1 could efectively predict the progression of AF. Te ROC curve for assessing the ability of sST2 and TIMP-1 alone and in combination with other biomarkers and clinical parameters to predict AF progression is shown in Figure 4. When these two biomarkers were combined, the AUC was 84.6% (95% CI 0.806-0.887, p � 0.000), sensitivity  was 63.4%, and the specifcity was 90.0%. When these two biomarkers were combined with BNP and LAD, the AUC was 89.7% (95% CI 0.855-0.939, p � 0.000) and the sensitivity and specifcity were 88.0% and 78.9%, respectively. Finally, the combination of the two biomarkers with BNP and LAD as well as age and gender resulted in an AUC of 0.901 (95% CI 0.858-0.944, p � 0.000). Te sensitivity and specifcity of these six variables for predicting AF progression were 82.2% and 86.0%, respectively. Te clinical models of TIMP-1 and sST2 for predicting AF progression showed a modest improvement with the addition of BNP, LAD, age, and gender.

Discussion
Tere are two main fndings in this present study. Te serum level of sST2 and TIMP-1 may be a biomarker for diferentiating paroxysmal AF from SR when AF cannot be diagnosed using electrocardiography in clinical practice. Te cutof values of TIMP and sST2 were 90.57 ng/ml and 12.81 ng/ml, respectively, and both of them had high sensitivity and specifcity. TIMP-1 and sST2 may be AF treatment targets. Te lifetime risk of AF mainly depends on age, but it is also infuenced by genetic and clinical factors (including gender, BMI, smoking history, HF, and hypertension) [1]. Tis study found that age and BMI are higher in the AF group, as is the proportion of males, which is consistent with one previous study [13], and the AF group also had a wider LAD, as found elsewhere [14]. Although the early intervention and control of modifable risk factors were previously found to reduce the incidence of AF [13], the predictive value of biomarkers has not been well-defned until now.
Atrium enlargement can result in physiological stretching in AF. Physiological stretching causes myofbroblasts to release IL-33, which binds the ST2 receptor (ST2L) to the cardiomyocyte membrane, promoting cell survival and integrity. In chronic conditions, however, local and neighboring cells can increase the release of the IL-33 decoy, sST2, which blocks IL-33/ST2L binding and promotes tissue fbrosis. Novel aspects of ST2/IL-33 signaling mediating cardiac fbrosis represent some new biomolecular targets for the prevention and treatment of maladaptive remodeling and disease progression [15]. Matilla et al. found that sST2 could afect myofbroblast activation, leading to an increase in collagen synthesis and profbrotic molecules in human cardiac fbroblasts. NRP-1, a molecule upregulated by Sst2, has emerged as an interesting new target in cardiac fbrosis. Te proinfammatory and profbrotic efects triggered by sST2 via nuclear factor kappa-light-chain-enhancer of activated B cells highlighted the key role of the latter on cardiac fbrosis [16]. In this cross-sectional study, the role of the biomarkers sST2 and TIMP-1 in predicting the International Journal of Clinical Practice 5 progression of AF was explored. Previous studies had led to a number of fndings. Te systemic biomarker sST2 was independently associated with increased primary outcomes in patients with HF and AF, and the prognostic performance of sST2 was stronger in AF for all-cause mortality [17]. Higher sST2 was associated with a higher prevalence of AF, possibly refecting remodeling phenomena in AF [18], and, in patients with persistent AF, increased sST2 served as a marker of recurrence after radiofrequency ablation, and patients with sST2 ≥ 39.25 ng/ml were more likely to have a recurrence within one year [19]. In the current study, the results also showed a higher serum level of sST2 in patients with persistent AF. In another study, patients with AF showed signifcantly higher sST2 than the control group did [20]. Tis study consistently found that sST2 was higher in patients with SR and increased gradually as a patient moved from SR to paroxysmal persistent AF. Atrial fbrillation is dependent on the electrical and structural remodeling of the atrium, and myocardial fbrosis plays a critical role in the maintenance of AF through the heterogeneity of atrial electrical conduction [21]. Atrial structural remodeling and maintenance depend on the synthesis and degradation of extracellular matrix (ECM) proteins (matrix metalloproteinases [MMPs] and TIMPs) [21]. An imbalance between MMPs and TIMPs can lead to an abnormal turnover of the ECM and result in atrial remodeling and fbrosis. In 15 rapid atrial pacing-induced AF pig models, the MMPs and TIMPs of in situ activity and the expression of gelatinases (MMP-2 and MMP-9) and their relationship with TIMP-1 in the atria were explored. Chen et al. found that in situ gelatinase activity was signifcantly higher in AF than in SR. Te signifcant increases in MMP-9 in its pro-form and the messenger ribonucleic acid level were shown to be responsible for the increased gelatinase activity in AF. Te inhibitory activities of glycosylated TIMP-1 and TIMP-3 in AF tissues were markedly elevated and localized in the atrial interstitium. In addition, although TIMP-1 was found to be mostly colocalized with gelatinase activity in the AF tissues, implying the coexistence of gelatinase activity and TIMP-1, TIMP-3 appeared only partially colocalized and halted the gelatinase activity surrounding the cardiomyocytes. Te increased activity of gelatinase, TIMP-1, and TIMP-3 as well as their interaction may have contributed to the atrial ECM remodeling of AF [22]. In other studies, in patients with persistent AF, TIMP-1 levels were increased when compared with patients with paroxysmal AF, and the levels of TIMP-1 were also higher in patients with paroxysmal AF than in the SR group [23,24]. Tese fndings were consistent with the present study. In another recent study, the expression of TIMP-1 was high in patients with persistent AF and chronic AF but not in those with paroxysmal AF [25], and elsewhere it was found that there was no signifcant diference in the plasma levels of TIMP-1 in patients with paroxysmal AF or those with persistent AF [24], which difered slightly from the fndings of the present study. Te reason may be due to inadequate power (i.e., a type II error). In the present study, TIMP-1 was higher in the paroxysmal AF group than the SR group and higher in the persistent AF group than the paroxysmal AF group. In one study, AF development and progression (from   International Journal of Clinical Practice paroxysmal to persistent) were associated with a gradual increase in the serum levels of TIMP-1 [11]. A recent metaanalysis revealed that increased MMP-1 and decreased TIMP-2 levels are signifcantly associated with an increased risk of AF [26]. Moreover, Wakula et al. demonstrated that the levels of TIMP-1 in patients with paroxysmal AF were higher than those in patients without AF [27].
Te clinical course of AF is marked by the development of atrial fbrosis [28][29][30][31], and sST2 is a marker of fbrosis, particularly cardiac fbrosis [15]. Elevated sST2 levels in patients with AF were associated with higher left atrium lowvoltage areas [32]. Moreover, sST2 levels were higher in paroxysmal AF with a low-voltage zone greater than 20% compared to those with a smaller low-voltage zone. Te dynamic balance of MMP-9 and TIMP-1 determines the fbrosis signal strength [21], and immunohistochemical and double-immunofuorescence staining for TIMP-1 in AF and SR tissues in the atrial interstitium showed that the increased levels of TIMP-1staining in the atrial interstitium reached statistical signifcance in the AF group when compared with the SR group [22]. Terefore, these two biomarkers have high specifcity for left atrial remodeling. Brain natriuretic peptide is an indicator of cardiac wall stress, and it is responsible for fuid and blood pressure homeostasis through its diuretic and vasodilatory efects and also afects cardiovascular remodeling. Tus, there is a strong causal relationship between natriuretic peptides and the incidence of AF [33,34]. Moreover, the most prominent risk factors for AF development, such as age, sex, increased BMI, hypertension, and HF, have all been related to elevated natriuretic peptide concentrations. Previous studies showed that patients with AF had signifcantly higher CRP levels than the SR group [20], and elevated hs-CRP levels were signifcantly associated with an increased risk of AF [35][36][37]. However, circulating BNP and CRP are known biomarkers for systemic circulation and can be infuenced by many factors, such as HF and acute or chronic infammation. In the present study, the serum levels of sST2 and TIMP-1 played an important role in AF progression. Tese two biomarkers had high sensitivity and specifcity for diferentiating SR from AF, especially when combined with BNP, LAD, age, and gender. Compared to the clinical variables, sST2 and TIMP-1 had high sensitivity and specifcity for diferentiating SR and AF, as they were involved in the development of AF, but BNP did not. Te increased sST2 levels suggested that the processes of infammation and fbrosis were overactivated. However, unlike BNP, the expression of sST2 was not infuenced by age or BMI.

Limitations
Tis study had some limitations. First, it was an observational study with a relatively small sample size in each group.
Certain known predictors of recurrence, such as hs-CRP, were not found to be signifcant in this study, which may have been because of a lack of statistical power. Terefore, research with a larger population is necessary to confrm the prognostic value of sST2 and TIMP-1 in identifying patients at high risk of paroxysmal AF. Second, serum markers are not heart-specifc, and the fndings here were not supported with atrial tissue biopsy data or coronary sinus sampling. Moreover, patients with conditions associated with fbrosis  were excluded from the study. In addition, the fact that this was a prospective, single-center, and cross-sectional study may have introduced unavoidable selection bias. Tus, to further confrm the efciency of sST2 and TIMP-1 in evaluating AF progression, a prospective and multicenter study is required. Finally, the relationship between sST2, TIMP-1, and atrial tissue fbrosis was not investigated in this study, and this issue needs to be addressed.

Conclusions
In patients with AF, serum TIMP-1, sST2, BNP, hs-CRP, and LAD are associated with the progression of AF. In this study, sST2 and TIMP-1, which both had high sensitivity and specifcity, were the best biomarkers for evaluating the progression of AF. Terefore, in clinical practice, sST2 and TIMP-1 may be able to serve as biomarkers to diferentiate paroxysmal AF from SR.

Data Availability
Te data that support the fndings of this study are available from the corresponding author upon reasonable request.

Ethical Approval
Te study was conducted in accordance with the Declaration of Helsinki (as was revised in 2013). Te study was approved by Ethics Committee of the Beijing Anzhen Hospital.

Consent
A written informed consent was obtained from all participants (2022042X).

Conflicts of Interest
Te authors declare that they have no conficts of interest.