The Biomarkers NT-proBNP and CA-125 are Elevated in Patients with Idiopathic Atrial Fibrillation

Abstract

Background

Blood biomarkers related to AF could be useful to detect silent AF and to develop stratified strategies for AF prevention. Previous studies identified markers that predict incident AF. However, it is difficult to differentiate whether biomarkers relate to underlying cardiovascular diseases, are generated by the atria in response to an AF episode, or both. We therefore measured a panel of blood biomarkers in patients without overt CVD with and without AF to investigate the association between biomarkers and atrial fibrillation (AF) in patients without overt cardiovascular disease (CVD).

Methods

Blood samples – drawn remote from an AF episode – of 60 patients with AF but without overt forms of CVD (idiopathic AF; iAF) were compared to 120 matched patients with sinus rhythm only. A novel antibody-based method for quantification of blood biomarkers (OlinkProseek Multiplex Cardiovascular) was used to compare 92 biomarkers between the two groups.

Results

N-terminal pro-B-type natriuretic peptide (NT-proBNP), Cathepsin L1, Endothelial cell-specific molecule 1, Cancer Antigen-125 (CA-125), Heat shock 27kDa protein, Galanin peptides, Proteinase-activated receptor 1, Stem cell factor, and CD40-ligand were all higher in iAF patients than in SR controls. Both NT-proBNP (OR1.55(1.07–2.25);p=0.022) and CA-125 (OR1.68(1.07–2.64);p=0.026) were independently associated with iAF.

Conclusions

This exploratory study, investigating over 90 cardiovascular blood biomarkers in patients without known CVD, identified one established biomarker for paroxysmal AF, NT-proBNP, and a novel marker, CA-125. CA-125 - previously unrelated to paroxysmal AF in an otherwise healthy population - may thus be a potential indicator of remote paroxysms of AF.

Introduction

Atrial fibrillation (AF) usually occurs in patients with an increased vascular risk profile[1]. AF is called idiopathic AF (iAF) when meticulous phenotyping does not identify any concomitant (cardiovascular) disease. Therefore, these patients are most suitable to study early forms of this arrhythmia without the interference of associated disease[2].

In a recent report, Lind et al.[3] used the OlinkProseek Multiplex Cardiovascular I kit to identify blood markers associated with incident AF in two large-scale population studies. They concluded that N-terminal pro-B-type natriuretic peptide (NT-proBNP), Fibroblast growth factor 23 (FGF23), Fatty acid-binding protein 4 (FABP4), growth differentiation factor 15 (GDF-15), and Interleukin 6 (IL6) are determinants of incident AF, with NT-pro-BNP and FGF-23 remaining significant determinants upon correction for associated cardiovascular diseases. However, as these individuals were recruited from population studies, there were relatively high rates of hypertension and of history of stroke, myocardial infarction and heart failure.

In the study reported here, we used the OlinkProseek Multiplex Cardiovascular I kit to identify markers associated with the presence of AF in otherwise healthy patients, enabling identification of markers associated with AF rather than with underlying vascular disease.

Material and Methods

Of all consecutive patients who underwent cardiac computed tomography angiography (CTA) in the Maastricht University Medical Center between January 2008 and March 2011, all iAF patients were selected (n=115), and all age (±1 year), sex and PROCAM cardiovascular risk score-matched(±2%)[4] patients in permanent sinus rhythm (n=275) were selected as described previously[5]. All patients underwent CTA to detect subclinical coronary artery disease (sCAD). Out of these patients, EDTA-plasma was available for analysis in 180 patients (60 iAF, 120 SR). All patients were in sinus rhythm during blood sampling, which was performed after an overnight fast, at the day of CTA. The study was approved by the Institutional Review Board, all participants gave written informed consent. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Idiopathic AF was defined as the absence of hypertension (antihypertensive drug use, or systolic blood pressure ≥140 mm Hg, or diastolic blood pressure ≥90 mm Hg, or left ventricular hypertrophy [interventricular septum width >10 mm, posterior wall width >10 mm]), diabetes (fasting blood glucose >7.0 mmol/L), or hypercholesterolemia (total fasting cholesterol >7.0 mmol/L). In addition, none of the patients had a history of coronary artery disease, renal dysfunction, chronic heart failure, stroke, malignancy, thyroid disease or pulmonary disease, or evidence of structural cardiovascular disease on echocardiogram, including valvular heart disease (< grade 1).

In all blood samples, the OlinkProseek Multiplex Cardiovascular I kit (Olink Proteomics, Uppsala, Sweden) was used to measure proteins in EDTA plasma by real-timePolymerase Chain Reaction[6]. Due to a low call rate (valid measurement in <85% of patients), Interleukin 4 (IL4), Natriuretic Peptides B (BNP), and Melusin (ITGB1BP2) were not included in further analyses. Values which were below the Limit of Detection (LOD) were replaced by the LOD value (http://www.olink.com/data-you-can-trust/validation/). Ten patients were not included in further analyses based on a large number of values below LOD (valid measurement for <85% of the proteins). The data from the panels were normalised to the median of 0 for each protein. This procedure enabled comparisons between the measurements from different panels. The panel provides NPX-values which allow for relative quantification, values can only be compared for the same protein across samples. NPX values for 2 different proteins are not comparable.

Statistical analyses

Statistical analyses were performed using SPSS statistical software (IBM SPSS statistics version 23.0, IBM Corporation, Armonk, NY). Correlations are presented as Pearson’s r and p-value. Normally distributed continuous variables are presented as mean±standard deviation, non-normally distributed continuous as median[interquartile range], and categorical variables as number of patients and percentage. All biomarkers were tested for an association with iAF using logistic regression, adjusting for age, sex (following the biomarker panel’s instructions) and the presence of CT-angiographic sCAD. All significant biomarkers, age, sex and sCAD were included in multivariable logistic regression. Results were checked for collinearity and interaction among covariates, which were not found. Manual backwards elimination, retaining age, sex and sCAD was used to construct the final models (retention level set at p<0.10). Odds ratios and 95% confidence intervals were calculated. Overall, p<0.05 was considered significant. The first author had full access to all the data in the study and takes responsibility for its integrity and the data analysis

Results

As patients originate from a matched case control study, age (54.3±9.3 vs 54.1±10.3), sex (30.0% vs 31.7% women) and blood pressure (systolic blood pressure 126±9mmHg vs 126±12mmHg, diastolic blood pressure 80±10mmHg vs 78±10mmHg) did not differ significantly between patients with sinus rhythm and idiopathic AF. Echocardiographic parameters did not differ significantly (left atrial diameter 37±4mm vs 39±5mm, interventricular septum thickness 8.8±1.1mm vs 8.6±0.7mm, posterior wall thickness 8.7±1.1mm vs 8.5±0.6mm, and left ventricular ejection fraction 61[4]% vs 62[5]% respectively). The prevalence of sCAD was higher in the iAF group (36.5% vs 46.0%), while statin use was comparable (10.6% vs 11.7%). Full baseline characteristics and imaging parameters were reported earlier[5,7].

The age, sex and sCAD adjusted odds ratios of the 89 biomarkers and their 95% confidence interval (95%CI) are shown in Supplemental [Table 1]. N-terminal pro-B-type natriuretic peptide (NT-proBNP; OR1.70 (1.16–2.49), p=0.006), Cathepsin L1 (CTSL1; OR3.96 (1.27–12.36, p=0.018), Endothelial cell-specific molecule 1 (ESM1; OR2.60 (1.15–5.88), p=0.022), Cancer Antigen 125 (CA-125; OR 1.93 (1.10–3.38), p=0.022), Heat shock 27 kDa protein (HSP27; OR1.46 (1.05–2.02), p=0.024), Galanin peptides (GAL; OR1.93 (1.07–3.47), p=0.029), Proteinase-activated receptor 1 (PAR1; OR2.60 (1.06–6.38), p=0.037), Stem cell factor (SCF; OR3.08 (1.06–8.99), p=0.039), and CD40 ligand (CD40L; OR1.43 (1.00–2.06), p=0.050) were all higher in iAF patients than in SR controls [Figure 1]. The mean expression of these 9 biomarkers is shown in [Figure 2]. There were no significant differences in the levels of the other biomarkers between iAF and controls.

Table 1. Odds ratios, corrected for age, sex and subclinical coronary artery disease (sCAD) on computed tomography, of 89 biomarkers of the idiopathic AF versus the sinus rhythm control group. Shown are OR(95% CI) and corresponding p-value.

	Age, sex and sCAD adjusted
	OR (95% CI)	p
N‐terminal pro‐B‐type natriuretic peptide (NT-proBNP)	1.701 (1.163 – 2.488)	0.006
Cathepsin L1 (CTSL1)	3.961 (1.269 – 12.356)	0.018
Endothelial cell‐specific molecule 1 (ESM1)	2.601 (1.151 – 5.876)	0.022
Cancer Antigen 125 (CA125)	1.927 (1.101 – 3.375)	0.022
Heat shock 27 kDa protein (HSP27)	1.459 (1.052 – 2.023)	0.024
Galanin peptides (GAL)	1.925 (1.068 – 3.468)	0.029
Proteinase-activated receptor 1 (PAR1)	2.601 (1.060 – 6.384)	0.037
Stem cell factor (SCF)	3.082 (1.057 – 8.986)	0.039
CD40 ligand (CD40L)	1.433 (1.000 – 2.055)	0.050
TNF-related apoptosis-inducing ligand (TRAIL)	3.048 (0.952 – 9.759)	0.061
Platelet endothelial cell adhesion molecule (PECAM1)	2.022 (0.943 – 4.334)	0.070
Pentraxin-related protein PTX3 (PTX3)	0.525 (0.248 – 1.113)	0.093
Monocyte chemotactic protein 1 (MCP1)	1.774 (0.933 – 3.375)	0.081
C-X‐C motif chemokine 1 (CXCL1)	1.387 (0.945 – 2.035)	0.095
Spondin 1 (SPON1)	2.249 (0.840 – 6.026)	0.107
Tissue factor (TF)	2.385 (0.809 – 7.029)	0.115
Beta-nerve growth factor (Beta-NGF)	1.972 (0.839 – 4.637)	0.120
Pappalysin‐1 (PAPPA)	1.673 (0.853 – 3.284)	0.134
TNF-related activation‐induced cytokine (TRANCE)	1.531 (0.865 – 2.710)	0.143
E-selectin (SELE)	0.649(0.360– 1.171)	0.151
Leptin (LEP)	0.740 (0.483 – 1.134)	0.167
Interleukin-6 receptor subunit alpha (IL6RA)	1.766 (0.762 – 4.091)	0.185
Fractalkine (CX3CL1)	1.836 (0.740 – 4.559)	0.190
Matrix metalloproteinase 12 (MMP12)	1.371 (0.842 – 2.230)	0.204
Proto-oncogene tyrosine‐protein kinase Src (SRC)	1.332 (0.847 – 2.097)	0.215
Follistatin (FS)	1.502 (0.789 – 2.861)	0.216
Agouti‐related protein (AGRP)	1.493 (0.773 – 2.885)	0.233
Epidermal growth factor (EGF)	1.207 (0.879 – 1.658)	0.245
Fibroblast growth factor 23 (FGF23)	1.301(0.833– 2.030)	0.247
Myoglobin (MB)	0.708 (0.390 – 1.285)	0.256
Adrenomedullin (AM)	1.660 (0.685 – 4.021)	0.261
Interleukin 16 (IL16)	1.457 (0.747 – 2.841)	0.269
Heparin-binding EGF‐like growth factor (HB-EGF)	1.637(0.668 – 4.010)	0.281
Osteoprotegerin (OPG)	1.572(0.650– 3.799)	0.315
Dickkopf‐related protein 1 (DKK1)	1.313(0.764– 2.257)	0.325
Myeloperoxidase (MPO)	1.723(0.578– 5.133)	0.329
NF‐kappa‐B essential modulator (NEMO)	1.234 (0.800 – 1.905)	0.342
Matrix metalloproteinase 3 (MMP3)	1.308(0.751– 2.278)	0.344
ST2 protein (ST2)	1.336 (0.729 – 2.450)	0.348
Interleukin-27 subunit alpha (IL27A)	1.641(0.573– 4.697)	0.356
TNF-related apoptosis-inducing ligand receptor 2 (TRAILR2)	0.615 (0.215 – 1.759)	0.365
P-selectin glycoprotein ligand 1 (PSGL-1)	0.455 (0.082 – 2.528)	0.368
CD40L receptor (CD40)	1.403(0.670– 2.939)	0.369
Placenta growth factor (PIGF)	1.529(0.566– 4.130	0.402
Membrane‐bound aminopeptidase P (mAmP)	0.877 (0.643 – 1.198)	0.410
Tumor necrosis factor receptor 2 (TNFR2)	1.352(0.619– 2.954)	0.450
Thrombomodulin (TM)	1.467(0.520– 4.135)	0.469
Interleukin 8 (IL8)	0.794 (0.422 – 1.495)	0.475
Macrophage colony stimulating factor (CSF1)	1.650(0.408– 6.673)	0.482
Tumor necrosis factor ligand superfamily member 14 (TNFSF14)	1.353(0.562– 3.257)	0.501
Receptor for advanced glycosylation end products (RAGE)	1.298(0.549– 3.071)	0.552
Growth/differentiation factor 15 (GDF-15)	1.212(0.607– 2.419)	0.586
T-cell immunoglobulin and mucin domain 1 (TIM)	0.873 (0.530 – 1.437)	0.592
Angiopoietin-1 receptor (TIE2)	1.287(0.421– 3.935)	0.658
Interleukin 6 (IL6)	0.922 (0.632 – 1.343)	0.672
Tissue-type plasminogen activator (tPA)	1.150(0.593– 2.233)	0.679
Matrix metalloproteinase 1 (MMP1)	0.902(0.544– 1.496)	0.689
Hepatocyte growth factor (HGF)	1.154(0.565– 2.356)	0.694
Eosinophil cationic protein (ECP)	1.112(0.644– 1.919)	0.703
C-C motif chemokine 4 (CCL4)	1.106(0.656– 1.863)	0.706
Vascular endothelial growth factor D (VEGF-D)	0.901(0.505– 1.606)	0.723
Prolactin (PRL)	1.089(0.678– 1.747)	0.725
SIR2-like protiein (SIRT2)	1.038 (0.821 – 1.038)	0.757
C-X‐C motif chemokine 6 (CXCL6)	0.935 (0.587 – 1.489)	0.776
Chitinase‐3‐like protein 1 (CHI3L1)	0.939 (0.602 – 1.463)	0.780
Interleukin 1 receptor antagonist protein (IL1RA)	0.924 (0.519 – 1.654)	0.789
Galectin 3 (GAL3)	0.915 (0.463 – 1.810)	0.799
Tumor necrosis factor receptor superfamily member 6 (FAS)	1.123(0.445– 2.835)	0.806
Urokinase plasminogen activator surface receptor (UPAR)	1.119(0.451– 2.773)	0.808
Cathepsin D (CTSD)	0.914 (0.436 – 1.915)	0.811
Kallikrein 11 (hK11)	0.904(0.390– 2.093)	0.814
Kallikrein 6 (KLK6)	1.097 (0.452 – 2.662)	0.837
Vascular endothelial growth factor A (VEGF-A)	1.095(0.400– 3.000)	0.859
Cystatin B (CSTB)	1.045(0.641– 1.703)	0.860
Fatty acid‐binding protein 4 (FABP4)	1.066 (0.521 – 2.183)	0.861
Lectin‐like oxidized LDL receptor 1 (LOX1)	1.043 (0.605– 1.798)	0.879
C-C motif chemokine 20 (CCL20)	0.978(0.709– 1.349)	0.890
Caspase 8 (CASP8)	1.032 (0.639 – 1.668)	0.897
Resistin (RETN)	0.962(0.519– 1.781)	0.901
Platelet-derived growth factor subunit B (PDGFsuB)	1.015(0.773– 1.334)	0.912
Protein S100-A12 (EN-RAGE)	0.969 (0.553 – 1.700)	0.913
Interleukin 18 (IL18)	1.034(0.530– 2.018)	0.921
Growth hormone (GH)	0.992(0.844– 1.166)	0.924
Matrix metalloproteinase (MMP10)	1.027 (0.583– 1.809)	0.927
C-C motif chemokine 3 (CCL3)	1.044 (0.408– 2.670)	0.928
Tumor necrosis factor receptor 1 (TNFR1)	1.043(0.373– 2.918)	0.937
Renin (REN)	1.004(0.622– 1.618)	0.988
C-X‐C motif chemokine 16 (CXCL16)	0.993 (0.338 – 2.923)	0.990

Figure 1. Shown are age, sex and presence of sCAD-adjusted odds ratios and 95% confidence intervals. Panel A: Nine biomarkers differ significantly in idiopathic atrial fibrillation (iAF) compared to sinus rhythm controls (SR). Panel B: Multivariable regression analysis revealed that both NT-proBNP and CA-125 are independently associated with the presence of iAF.

For all biomarkers shown p≤0.05.

Figure 2. Unadjusted box-plots of 9 biomarkers that were found to be significantly different between patients without overt cardiovascular disease, with and without atrial fibrillation. Shown are mean and standard deviations. ** indicates p<0.01, * indicates p<0.05.

CA-125=Cancer Antigen 125; CD40L=CD40 ligand; CTSL1=Cathepsin L1 ; ESM1=Endothelial cell‐specific molecule 1; GAL=Galanin peptides; HSP27=Heat shock 27 kDa protein; iAF=idiopathic atrial fibrillation, NPX=normalized protein expression, NT-proBNP=N‐terminal pro‐B‐type natriuretic peptide; PAR1=Proteinase-activated receptor 1; SCF=Stem cell factor; SR=sinus rhythm.

Multivariable analysis showed that both CA-125 and NT-proBNP, adjusted for age, sex and sCAD, are associated with a history of iAF (OR for CA-125 2.25 (1.27-3.99), p=0.005 and NT-proBNP 1.79 (1.22-2.62), p=0.002). NT-proBNP and CA-125 do not correlate with each other (r=-0.09, p=0.27). [Figure 3] shows the prevalence of iAF in tertiles of the CA-125 (21.7%, 28.6% and 35.0%; p=0.27) and NT-proBNP levels (19.3%, 26.2% and 39.7%; p=0.04). The highest prevalence of iAF can be found in patients with levels of both CA-125 and NT-proBNP in the highest tertile (50.0%), with an intermediate prevalence in patients with either CA-125 (30.0%) or NT-proBNP (37.2%) in the highest tertile and the lowest prevalence in patients with none of the markers in the highest tertile (18.9%; p=0.005).

Figure 3. Percentages of patients with idiopathic atrial fibrillation (iAF) in tertiles of CA-125 and NT-proBNP levels. SR=Sinus Rhythm, * indicates P<0.05, ** indicates P<0.01.

Discussion

From this study, it can be concluded that even in a healthy population without cardiovascular disease, the levels of NT-proBNP and CA-125 are associated with AF. Results from this study furthermore show that the levels of CTSL1, ESM1, HSP27, GAL, PAR1, SCF and CD40L may contain additional information on the presence of a history of AF.

Multivariable analysis showed that two markers are independently associated with the presence of a history of AF: NT-proBNP and CA-125. NT-proBNP has been recognized as a marker of incident AF in population studies, on top of risk scores based on the presence of cardiovascular disease [8,9]. The data from this cohort confirm that elevated NT-proBNP blood levels are a marker of (remote) episodes of paroxysmal AF even in the absence of concomitant vascular disease.

CA-125is in current clinical use in the follow-up of ovarian cancer, but the value of CA-125 as a biomarker in cardiology is still to be determined[10]. CA-125 has before been linked – in small studies – to new-onset AF in patients with recompensated acute heart failure[11] as well as in post-menopausal women[12], but not yet to prevalent AF in a healthy population. There may be three possible explanations for the association between CA-125 and AF. Firstly, as CA-125 positively correlates with TNF-α, IL-6, IL-10[13] and IL-1β[14], it may be hypothesized to be a marker of the presence of underlying low-grade inflammation in this study population. Secondly: CA-125 is produced in response to mechanical stress or inflammation by different tissues derived from coelomic epithelium, such as the pericardium and pleura [10]. Based on the strong anatomical relation between the posterior left atrium and the pericardium[15] and the dominance of pulmonary vein triggers in early AF, the rise in CA-125 may be hypothesized to be a reflection of the interplay between atrium and pericardium. Lastly, as CA-125 is shown to correlate with pleural effusion [16], it may be hypothesized that the paroxysms of AF have led to – transient – pleural effusion with an associated rise in CA-125.Univariable regression analysis revealed that several pathophysiological processes may be active in patients with early forms of AF, such as angiogenesis (CTSL1[17], ESM1[18]), endothelial dysfunction (ESM1[18], SCF[19]), cellular stress (HSP27[20]), cardiac remodelling (CTSL1[20]), cardiac remodelling (CTSL1[22]) and activation of the coagulation system (GAL[22], PAR1[23], and CD40L[24]). The markers of a hypercoagulable state are in line withprevious reports of a prothrombotic state veryearly in the pathogenesis of AF[25,26]. Clearly, a much larger patient sample would be needed to verify any of these interactions which did not uphold during multivariable testing.

FGF-23 has been found to be associated with incident and prevalent AF in population studies [3,27-29]. We could not replicate this association despite shared methodology[3]. FGF-23 levels are clearly influenced by blood pressure, left ventricular mass, cardiovascular disease in general, and chronic kidney disease[30-33]. As patients with these conditions were excluded from the present analysis, we may conclude that FGF-23 is a marker of advanced vascular disease, while the markers found in this study are associated with very early AF without the effects of confounding vascular disease.

The current categorization of AF in paroxysmal, (long-standing) persistent or permanent may in the long run fall short to adequately describe distinct patient groups [34]. Biomarkers, be it those found in blood or through the use of imaging, could help determine and distinguish between different forms of AF. The results from the present study suggest that, when reporting on biomarkers in AF, underlying processes – such as the presence of other forms of cardiovascular disease - that lead to the creation of a substrate for AF should be taken into account.

Strengths,limitations and clinical implications

The major strength of this study lies in the selection of patients without concomitant vascular disease and blood sampling during sinus rhythm, which enabled us to identify and gain further insight into markers associated with the presence of AF, rather than markers of the vascular conditions that usually accompany AF[35]. However, the study population of this exploratory study is relatively small and highly selected, necessitating confirmation in larger study populations. Next to more insight in the pathophysiology of early AF without concomitant vascular disease, this study should be seen as a means to support investigating the use of NT-proBNP and CA-125 in screening programs for AF incorporating biomarkers[36], which should be tested prospectively.

Conclusion

This study confirms the association between levels of NT-proBNP and a diagnosis of paroxysmal AF even in patients without concomitant cardiovascular conditions, suggesting that NT-proBNP is released by the atria and remains elevated even in the absence of an on-going arrhythmia episode. Furthermore, this data suggests that CA-125 could be a novel marker of the occurrence of (paroxysms of) AF in patients without structural heart disease. Finally, in conjunction with published analyses, our data suggest that elevated FGF-23 levels may be a marker for AF in patients with concomitant cardiovascular disease.