Blood biomarkers and atrial remodeling in patients at risk of atrial fibrillation

Introduction

Left atrial (LA) enlargement is commonly seen in arterial hypertension, due to increased atrial pressure related to diastolic dysfunction, and in endurance training, due to volume overload (Trivedi et al., 2020). Pressure and volume loading result in stretching of the atrium and, over time, alterations in the cardiomyocytes, fibroblasts, and non-collagen infiltrative compartments of the atrium (Tsang et al., 2002; De Jong et al., 2011; Goette et al., 2016; Otaki et al., 2014), creating a potential substrate for atrial fibrillation (AF). Serum blood biomarkers related to myocardial damage, endothelial dysfunction and fibrosis have been increasingly explored in relation to cardiovascular risk and therefore evaluated to predict the particular risk of patients for cardiovascular events. (Fashanu et al., 2017; Ionin et al., 2020; Huxley et al., 2013).

Our aim was to assess the influence of patient characteristics and echocardiographic measures of left ventricular (LV) and LA size and function on a comprehensive set of serum blood biomarkers of myocardial injury, endothelial dysfunction and fibrosis in patients at risk for AF.

Methods

The study design was cross-sectional observational with the cohort (n = 511) consisting of two subgroups: endurance sports practitioners (athletes) training more than 3 h per week for at least 5 years (n = 286); and patients with established systemic arterial hypertension treated with at least one antihypertensive drug for at least 3 years (n = 225). The study workflow and inclusion/exclusion criteria are shown in Figure 1. All participants were assessed for cardiovascular risk factors, medical history of target organ damage, comorbidities, and pharmacological treatment; followed by an echocardiographic examination and the collection of a peripheral venous blood sample. The athlete cohort filled out a questionnaire regarding their average weekly training sessions (i.e., type(s) of sport, session duration and intensity) with the aim of estimating the cumulative MET minutes per week for each athlete (Ainsworth et al., 2011). The study was approved by the hospital ethics committees, and participants gave written informed consent.

The echocardiographic examination was performed on a Vivid E95 system (GE, Vingmed Ultrasound, Horten, Norway) equipped with a M5S and 4Vc transducer. Cardiac dimensions were measured in appropriate 2D views and derived parameters calculated, following relevant society recommendations (Lang et al., 2015). Real-time three-dimensional (3D) scans were obtained in the apical view in patients with an appropriate imaging window (in 84% of the cohort, see Figure 1), whereas 3D LA volumes were analyzed using the ‘4D LA quantification tool’ available in GE EchoPAC.

Blood was extracted from a peripheral vein at the same clinical visit as the imaging study. Samples were processed for complete blood count, metabolic panels and lipid profile. Additional blood biomarkers were assessed evaluating myocardial damage (Brain Natriuretic Peptide (BNP); troponin I, Oncostatin M (OSM) (Kubin et al., 2011), serum heart-type Fatty Acid-Binding Protein (FABP3) (Otaki et al., 2014; Setsuta et al., 2014), and Placenta Growth Factor (PIGF) (Draker et al., 2019)), vascular endothelial dysfunction (Endothelial cell-specific molecule 1 (Endocan-1) (Afsar et al., 2014; Çimen et al., 2016)); and fibrosis (Matrix Metalloproteinase 1, 2 and 9 (MMP 1,2,9) (Huxley et al., 2013; Boixel et al., 2003)), C-X-C motif chemokine Ligand 6 and 16 (CXCL 6, 16) (Ma et al., 2016; Xia et al., 2013), and Galectin-3 (Gal-3) (Ionin et al., 2020; Sonmez et al., 2014; Yalcin et al., 2015)). Additional information on echo measurements, phasic volume calculations, LV/LA strain analysis and laboratory methodology can be found in the Supplemental Methods.

The data was analyzed using IBM SPSS Statistics version 23.0. The quantitative variables were expressed as mean ± standard deviation or median and interquartile range based on the normality of their distribution evaluated by the Shapiro-Wilk test. Parameters that demonstrated a non-normal distribution were logarithmically transformed for the purpose of assessing correlations and fitting linear regression models. The qualitative variables were expressed as a total number and percentage. Differences between groups were analyzed for statistical significance with the t-test, when comparing variables with normal distribution, and the Mann Whitney test for non-normally distributed variables. When comparing categorical data, contingency tables and a Chi-square or the Fisher’s exact test were used for comparison. The strength and direction of the linear relationships between pairs of continuous variables was performed using the bivariate Pearson correlation, resulting in a correlation coefficient (R).

The association of patient/echo characteristics with biomarker levels in the athlete and hypertension cohorts, respectively, was assessed through linear regression models with individual biomarker levels as the dependent variable. Firstly, each parameter was examined in univariate analysis. Correlations were checked for age, gender, BMI, BSA, systolic blood pressure, diastolic blood pressure, diabetes, dyslipidaemia, smoking status, duration of arterial hypertension and the number of hypertensive drugs (hypertensive cohort only), MET/week (athlete cohort only), basal septal thickness, LV end-diastolic diameter, LV posterior wall thickness, LV mass index, RWT, LV end-diastolic volume, LV end-systolic volume, LA/LV volume ratio, LV ejection fraction, LV 4-chamber GLS, TAPSE, septal and lateral e’ and a’ velocities, average e’ velocity, E/e’ ratio, 2D and 3D LA maximal, minimal and preP volumes, 2D and 3D LA active, passive and total emptying fraction, LA reservoir, conduit and contractile strain. When selecting variables for the multiple regression model variables with a significant zero order R correlation with the biomarker level at hand (R > 0.100) were considered. Before running the analysis, the interactions between the independent variables were assessed for correlations stronger than 0.700. In case two independent variables were highly correlated, one of them was dropped, the decision made based on strength of correlation and clinical judgement. The models were fitted using a stepwise backward variable selection. Patients were excluded based on missing values using the pairwise approach. The proportion of the variance in the dependent variable that was predictable from the independent variables was assessed using the coefficient of determination (R squared). A value of p < 0.05 was considered statistically significant.

Results

Exploring the patient cohort

The comparison of patient characteristics and echocardiographic parameters can be found in the Supplementary Table S1. The comparison of biomarker levels and renal function between patient groups is shown in Table 1. The histograms in Figure 2 show the ranges of biomarker serum concentrations within the two subgroups. Markers of myocardial damage, BNP and troponin, were represented in a narrow range of values within our study, whereas OSM and FAB3 had a wider distribution of values, however without a statistical difference between subgroups. On the other hand, PIGF levels, although also narrow in range, were significantly lower in athletes as compared to hypertensive patients. Endocan-1, the biomarker related to endothelial dysfunction, was significantly higher in athletes. Finally, biomarkers of fibrosis (i.e., CXCL6, CXCL16, MMP-1, MMP-9 and Gal-3) were uniformly elevated in arterial hypertension. The estimated glomerular filtration rate was statistically lower in arterial hypertension; however, the median value was not clinically different between groups. Only FAB3 levels showed a weak correlation with impaired renal function (R −0.216, p = 0.005). Furthermore, when exploring the relationship between ACE inhibitor (n = 83, 37%) or angiotensine II receptor blocker (ARB) (n = 100, 44%) use and biomarker levels in hypertensive patients, patients on ACEi had slightly lower CXCL16 levels (p = 0.028), whereas biomarker levels did not differ based on ARB use. The correlations between biomarkers and patient characteristics, LV and LA echo measurements are shown in Figure 3 - revealing the notable influence of age and BMI on most biomarker levels. Correlations with 3D volumes is shown in Supplementary Figure S1.

TABLE 1. Blood biomarker levels in the patient cohorts.

Predicting blood biomarker levels

The univariable and multivariable regression models were fitted for the athlete and hypertensive cohorts, respectively (Tables 2–4). Considering biomarkers of myocardial damage (Table 2), the multivariable model showed that dyslipidemia predicted BNP and troponin levels in the hypertensive cohort, whereas LA active emptying fraction and average e’ mitral annulus velocity predicted troponin levels in athletes. OSM levels were related to smoking status, LV dimensions, and LA active emptying fraction in hypertension, and solely to age in athletes. Age and right ventricular longitudinal function predicted FABP3 in hypertension and athletes alike, while LA reservoir strain was significant only in the univariable model within the athlete cohort. Furthermore, BMI was associated with PIGF levels in hypertension, and LV mass and LA active emptying fraction in athletes, whereas LA contractile strain and LA maximal size were not found predictive outside univariable analysis, respectively.

TABLE 2. Biomarkers of myocardial damage–The association of patient and echo characteristics with biomarker levels in the athlete and hypertension cohorts, respectively, assessed through linear regression models with individual biomarker levels as the dependent variable (as explained in the Methods section). Abbreviations are explained in the manuscript.

TABLE 3. Biomarker of vascular endothelial dysfunction - The association of patient and echo characteristics with biomarker levels in the athlete and hypertension cohorts, respectively, assessed through linear regression models with individual biomarker levels as the dependent variable (as explained in the Methods section). Abbreviations are explained in the manuscript.

TABLE 4. Biomarkers of fibrosis - The association of patient and echo characteristics with biomarker levels in the athlete and hypertension cohorts, respectively, assessed through linear regression models with individual biomarker levels as the dependent variable (as explained in the Methods section). Abbreviations are explained in the manuscript.

Statistical Notes:* indicates statistical significance at p < 0.05. † indicates statistical significance at p < 0.01.

Abbreviations: BSA, body surface area; BMI, body mass index; LA, left atrium; LV, left ventricle; TAPSE, tricuspid annular plane systolic excursion; 4C-GLS, 4-Chamber Global Longitudinal Strain; MET, metabolic equivalent of task.

Endocan-1, the biomarker related to vascular endothelial dysfunction, was related only to systolic blood pressure in hypertensive patients, however, to BMI, LV septal wall thickness and LA maximal size in athletes (Table 3).

Regarding the biomarkers related to fibrosis (Table 4) BSA, LV longitudinal function and LA active emptying fraction predicted MMP-1 levels in hypertension, whereas, only LA minimal volume was found relevant in athletes after multivariable adjustment. No correlations were found between patient/echo characteristics and MMP-2 levels in hypertension, while the amount of METs/week predicted them within the athlete cohort. MMP-9 related to blood pressure in both cohorts, however, in hypertension it was also predicted by BSA and LA contractile function. CXCL 6 was associated with gender in both cohorts, but also LA maximal volume in hypertension, whereas CXCL 16 levels did not correlate with any characteristics besides age in athletes. Finally, Gal-3 was related to diastolic function and LA reservoir strain (R −0.286, β −0.254, p = 0.007) in hypertension, and to BMI in athletes. The main findings are summarized in the Central illustration.

Discussion

Blood biomarkers of myocardial damage, such as BNP, serve as a direct signal of atrial remodeling in conditions such as heart failure. In our cohort, most biomarkers of myocardial damage showed a narrow range of values reflecting low prevalence of heart failure and acute myocardial damage in this relatively healthy population. Therefore, predictive models aiming to relate biomarker levels associated with myocardial damage and LA function were of limited yield (Table 1). FABP3, a protein in the cardiomyocyte cytosol released into the circulation from damaged cardiac tissue, has been associated with cardiovascular risk in the general population (Otaki et al., 2014) and with cardiovascular events in patients with arterial hypertension (Setsuta et al., 2014). Although FABP3 is not specific to atrial remodeling, levels have been shown independently associated with LA reservoir strain in athletes, presumably reflecting overall cardiac strain in the at-risk athlete cohort (Lindsay and Dunn, 2007). This relationship was also noted in our athlete’s group, however, it did not remain significant after multivariable adjustment.

Atrial fibrosis, seen in arterial hypertension (Choisy et al., 2007) and endurance training (Peritz et al., 2020), is associated with an increasing risk for atrial arrhythmias (Kottkamp, 2013; Burstein and Nattel, 2008). Circulatory biomarkers (i.e., cytokines and markers of collagen turnover, extracellular matrix remodeling, inflammation, or myocardial stretch) have been suggested to be reflective of atrial injury and fibrosis, and associated with the risk of AF. (Turagam et al., 2016). While the circulating serum levels of MMP-9 correlate with LA size (Sonmez et al., 2014), the follow-up of the Atherosclerosis Risk in Communities (ARIC) cohort demonstrated MMP-1, MMP-2 and MMP-9 levels relate to AF after adjustment for age, sex and race–with only MMP-9 levels remaining significant in the fully-adjusted model. (Huxley et al., 2013). Moreover, in hypertension, serum MMP-9 related to cardiovascular risk (Tayebjee et al., 2004). In our study, MMP-1 and 9 were found elevated in the hypertensive cohort, correlating with LA active emptying fraction, however, both levels were associated with BSA, inferring interpretation of blood levels would need to take into account patient size (Figure 3). CXCL-6 is another circulatory biomarker and a proinflammatory response chemokine associated with fibrosis. In our study CXCL-6 levels were elevated in hypertension and related to LA maximal volume, however, the levels were also influenced by sex. Finally, Gal-3, a beta-galactoside binding lectin and mediator of profibrotic pathways, has been related with LA size and fibrosis (Ionin et al., 2020; Sonmez et al., 2014; Yalcin et al., 2015) with levels predictive of AF development (Gong et al., 2020). In the 15-year follow-up of the ARIC study, plasma Gal-3 remained associated with AF risk in a multivariate model (Fashanu et al., 2017), whereas this was not seen in the 10-year follow-up of the Framingham Offspring cohort (Ho et al., 2014). Our study showed hypertensive patients had higher values of Gal-3 as compared to athletes. Moreover, Gal-3 was found independently associated with LA reservoir strain in the hypertensive cohort in the multivariable regression model, without confounding influences of patient characteristic that would complicate interpretation (i.e., gender, BMI, BSA) (Table 4; Figure 3).

The findings of this study have to be interpreted in light of the studied patient population, and the lack of biomarker specificity to atrial/myocardial remodeling. The lack of a matched control group and the underlying age and sex differences limit direct comparison between the two patient groups. However, these changes are a reflection of a real-world population characteristics, and should not bias the exploration of imaging parameters and biomarker associations within each group respectively. However, when comparing biomarker levels between different groups (e.g., as shown in Table 1) caution should be applied as biomarker levels are influenced by renal clearance and hepatic metabolism, age, sex, inflammation, and other systemic factors unrelated to atrial pathology (Forman et al., 2020) (Figure 3 and the Central illustration). Indeed, fibrosis biomarkers are not cardiac-specific but reflect systemic collagen turnover. As the LA has a relatively low tissue mass even significant atrial fibrosis might marginally impact total biomarker concentrations. Fibrosis of larger organs, such as the liver, better correlate with blood biomarkers, due to larger organ mass (Loomba and Adams, 2020). These circulating biomarkers reflect active fibrogenesis and fibrolysis whereas imaging biomarkers capture both long-term structural remodelling (such as atrial fibrosis) as well as acute conditions (such as acute changes in preload or afterload). Finally, corrections for multiple comparisons were not performed in the analysis, which increases the risk of type I error and false positive findings.

Overall, the weak correlation between circulating and imaging biomarkers reflects the complementary rather than interchangeable nature of these biomarkers. The modest relationship between gal-3 and LA reservoir strain highlights the prospective limitations of using galectin-3 as a strong predictor of LA fibrosis and remodelling, at least if used as a single parameter. Potentially, a prospective follow-up of this cohort, identifying individuals developing AF, could confirm the role of biomarker levels as another complementary screening marker of atrial cardiopathy and the risk for atrial arrhythmia development.

Conclusion

Serum levels of fibrosis biomarkers differ between populations at risk for developing AF, although caution is needed in interpreting these levels in the context of age, sex and renal function. Gal-3 levels were found independently associated with LA reservoir function in patients with arterial hypertension, suggesting its potential role as a a biomarker related to LA fibrosis. Clinical follow-up and additional studies are needed to discern the diagnostic performance and relationship of biomarker levels and clinical outcomes.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmmed.2026.1801523/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Xingbo Xu, University Medical Center Göttingen, Germany

Reviewed by: Abel Martin Garrido, Heidelberg University, Germany

Ana Maria Vergel Leon, University Medical Center Göttingen, Germany