Proteomics and Recurrence of Atrial Fibrillation: A Pilot Study Nested in the PREDIMAR Trial

Atrial fibrillation (AF), a strong risk factor for stroke and heart failure, is the most common sustained cardiac arrhythmia worldwide. It is estimated that more than 33.5 million persons have AF, which is a major health and economic burden [1, 2]. Catheter ablation is the most efficacious therapy in restoring sinus rhythm in AF patients. However, relapses occur in approximately 30% of patients in the first year after ablation [3]. Integrating diverse omics sciences holds the potential to improve the stratification of AF patients by their risk of recurrence [4, 5], improving current clinical management strategies [6].

In this context, the identification of novel biomarkers is needed for an early prediction of AF recurrences and a better management of these patients. Proteomics is a promising field that could provide new insights into the mechanisms of recurrences and prevention strategies. Previous studies on serum proteomics analyses of the incidence of primary AF showed, in the Framingham Heart Study Offspring [7], the GES-Reykjavik [8], and the PIVUS [9] cohorts, that a signature with a low-to-moderate number of proteins, including NT-pro-BNP, may predict the development of AF. These results were partially validated in other cohorts [7‒9]. However, only a few publications on protein analyses related to AF recurrence have been published recently [10, 11], and we have identified just one that employed an array-based approach to analyze the proteome, but with a very small sample size [12].

The “PREvención con DIeta Mediterránea de Arritmias Recurrentes” (PREDIMAR) is a secondary prevention randomized trial to prevent AF recurrence in patients with paroxysmal or persistent AF who underwent an ablation. The aim of this pilot study was to analyze the baseline proteome of subjects included in a case-control study nested within the PREDIMAR trial, to find proteins associated with the risk of AF recurrence.

The design and methods of the PREDIMAR trial (RCT #NCT03053843) were previously described [13, 14]. The PREDIMAR study is a randomized, controlled, single-blinded trial of secondary prevention in patients with AF treated with ablation. It was designed to assess the effect of an intervention based on a Mediterranean diet enriched with extra-virgin olive oil on the prevention of AF recurrences. Briefly, 720 patients diagnosed with AF, who underwent a catheter ablation, were randomly allocated to an intervention group, which received an intervention with a Mediterranean diet, or to a control group that received no dietary recommendations. Fasting serum samples were obtained before the ablation procedure and stored at −80°C.

We did not perform a sample size calculation because, as a pilot study, our aim was to provide preliminary estimates of response levels according to the randomization groups, which will be useful for sample size calculations, and to guide the selection of optimal laboratory conditions for the main study. In addition, we performed the sample handling and profile generation that will be replicated in the full study [15].

The main outcome in the PREDIMAR trial is any documented atrial tachyarrhythmia diagnosed from month 3 after the ablation until month 18 (AF recurrence) [13]. This recurrence had to be documented and confirmed by an electrocardiologist blinded to the intervention. In this pilot study, a case is defined as any participant with paroxysmal AF at baseline (before the ablation) and with a tachyarrhythmia recurrence during the follow-up. After excluding participants without samples at baseline, controls were randomly selected from the at-risk source population at the same time as cases occurred. We randomly selected 18 cases from 2 recruitment centers (Navarra and Madrid), and we matched them by age with 18 controls.

After performing the quality control process of samples and proteomics results, 16 cases (recurrent AF) and 17 controls (non-recurrent) remained in the analyses, and a total of 354 proteins were quantified. All the cases and 88% of the non-recurrent selected participants were allocated to the control group.

Baseline serum samples were diluted 1:1 in lysis buffer. Peptides recovered from in-gel digestion processing were reconstituted. MS/MS data sets for spectral library generation were acquired on a TripleTOF 5600+ mass spectrometer (Sciex, Canada) interfaced to an Eksigent nanoLC ultra 2D pump system (SCIEX, Canada) fitted with a 75 μm ID column (Thermo Scientific). Prior to separation, the peptides were concentrated on a C18 precolumn (Thermo Scientific). MS/MS data acquisition was performed using AnalystTF 1.7, and spectra files were processed through the ProteinPilot v5.0 search engine (Sciex). To avoid using the same spectral evidence in more than one protein, the identified proteins were grouped based on MS/MS spectra by the Progroup algorithm, regardless of the peptide sequence assigned. The protein within each group that could explain more spectral data with confidence was depicted as the primary protein of the group. A false discovery rate (FDR) was performed using a nonlinear fitting method and displayed results were those reporting a 1% global FDR or better [16].

The quantitative analyses were performed using a Sciex TripleTOF 5600+. The resulting ProteinPilot group file from library generation was loaded into PeakView (v2.1, Sciex), and peaks from SWATH runs were extracted with a peptide confidence threshold of 99% confidence and an FDR <1% (filtering false identification of proteins). The MS/MS spectra of the assigned peptides were extracted by ProteinPilot, and only the proteins that fulfilled pre-established quality criteria were validated. Proteins quantified with at least two unique peptides were considered.

The quantitative data obtained by PeakView were further processed using the Perseus software [17]. An unpaired Student’s t test was used for direct comparisons, with statistical significance defined as a p value below 0.05. To account for multiple testing, a permutation-based FDR method was applied. Proteins with a fold change (FC) greater than 1.33 or lower than 0.77 were classified as differentially expressed [17]. Data were normalized using the width adjustment method.

Logistic regression models were used to further analyze the association between differentially expressed proteins (normalized values) and AF recurrence. Crude and multivariable adjusted models were performed for each protein. The covariates included in the multivariable model were age (years), sex (male/female), BMI (continuous), T2D (yes/no), and hypertension (yes/no).

The results were further validated in an AF case series (n = 230) of patients treated with catheter ablation at the Clinica Universidad de Navarra (CUN). The recurrence rate in this case series was 31%. Two out of the ten differential proteins found in PREDIMAR were available for replication in the validation sample. A specific high-throughput methodology-based proximity extension assay (Olink Proseek^® Multiplex cardiometabolic panel) was used to measure the proteins. The same logistic regression models used for the PREDIMAR pilot study were applied to validate the association with AF recurrence (except for the adjustment for BMI).

Stata software version 16.0 (Stata Corp LP, College Station, TX, USA) was used to perform the logistic regression models. Significance testing was considered for p < 0.05.

Among the 33 participants included in the present pilot study, 58% of them had the diagnosis of paroxysmal AF before ablation and 42% persistent AF. Baseline characteristics of the participants according to AF recurrence status are shown in the online supplementary Table 1S (for all online suppl. material, see https://doi.org/10.1159/000543639). AF recurrent subjects were more likely to be men with diagnosed type 2 diabetes and/or hypertension. Mean time until recurrence was 233 (SD 136) days.

A total of 354 proteins were quantified. Differential expression between recurrent cases vs. non-recurrent controls is represented in the volcano plot showed in Figure 1, where the log2-FC against the p value was plotted. We found 10 proteins differentially represented in cases vs. controls: 2 of them were upregulated in cases: keratin type I cytoskeletal 17 (K17) (FC = 2.14; p = 0.017) and H6PD endoplasmic bifunctional protein (FC = 1.65; p = 0.032). On the contrary, eight proteins were downregulated among cases vs. controls: c4b-binding protein alpha chain (C4bpA) (FC = 0.64; p = 0.006), coagulation factor XI (FXI) (FC = 0.83; p = 0.011), chloride intracellular channel protein 1 (CLIC1) (FC = 0.62; p = 0.017), haptoglobin (HP) (FC = 0.37; p = 0.021), immunoglobulin heavy constant alpha 2 (Ig alpha-2 chain C region) (FC = 0.49; p = 0.022), c4b-binding protein beta chain (C4bpB) (FC = 0.73; p = 0.028), N-acetylglucosamine-1-phosphotransferase subunit gamma (GlcNAc-1-phosphotransferase subunit gamma/GNPTG) (FC = 0.61; p = 0.033), and platelet glycoprotein Ib alpha chain (GP-Ib alpha) (FC = 0.84; p = 0.038). In online supplementary Table 2S (online suppl. material), the main function of each of the differential proteins found in this pilot study is described, as well as the potential mechanism that may be underlying the association with AF.

The analyses of the association (crude and multivariate) of each of the differentially expressed proteins and AF recurrence can be observed in Table 1. We found that even after the adjustment for age, sex, BMI, T2D, and hypertension, the direct association between K17 and the inverse association between C4bpA, F11, CLIC1, IGHA2, GNPTG, and GP1BA with AF recurrence maintained the statistical significance. When we repeated the analyses excluding the 2 subjects that were randomized to the intervention group, the statistical significance of the association between GP1BA and AF was lost.

Finally, we were able to compare our results for F11 and GP1BA with those found in the CUN sample. We did not find a statistically significant association with AF recurrence for either protein. However, when we stratified by AF type (paroxysmal vs. persistent), GP1BA showed an inverse association, as observed in the PREDIMAR study, with an OR = 0.48 (95% CI: 0.10–1.28), approaching statistical significance (p = 0.143) for persistent AF recurrence.

In this pilot study nested within the PREDIMAR trial, we found that 10 proteins were differentially expressed in AF recurrent cases vs. non-recurrent controls. Two proteins, keratin type I cytoskeletal 17 (K17) and H6PD endoplasmic bifunctional protein, were upregulated in the cases compared to the controls. On the contrary, eight proteins were downregulated in the cases compared to the controls: C4bpA, coagulation factor XI, CLIC1, haptoglobin, Ig alpha-2 chain C region, C4bpB, N-acetylglucosamine-1-phosphotransferase subunit gamma, and platelet glycoprotein Ib alpha chain. In addition, all proteins downregulated in the cases were also associated with a lower risk of AF recurrence in the multivariable models after adjusting for age, sex, BMI, T2D, and hypertension, except for C4bpB. Finally, we were able to compare the association between GP1BA and AF recurrence with that found in a different population of patients with AF.

Nine out of the ten proteins identified in our study have been previously reported to be associated with AF or related diseases. In relation to the upregulated proteins in cases compared to controls, H6PD enzyme catalyzes the first two steps of the oxidative branch of the pentose phosphate pathway, including hexose-6-phosphate dehydrogenase (H6PD) activity. Interestingly, an H6PD gene variant, associated with higher carotid intima thickness, has also been linked to a higher risk of AF incidence and severeness in large cohorts [18, 19]. On the other hand, K17 promotes epithelial proliferation and tumor growth [20]. However, keratins are typical contaminants in proteomic experiments, so differential keratins in this type of workflow should be considered with cautioning [21].

In relation to proteins downregulated in cases compared to controls, lower circulating levels of both haptoglobin and C4bp have been associated with mitral valve prolapse, which is frequently concurrent with arrhythmias [22]. Moreover, in a small cross-sectional study, circulating levels of F11, C4bpA, and C4bpB appeared to be dysregulated in subjects with concomitant AF, coronary microvascular disease, and heart failure [23]. Another study found that CLICs, including CLIC1, were overexpressed in atrial tissues, but no information about circulating levels was found [24]. Interestingly, gene variants in the gene encoding for GP1BA have been previously associated with stroke [25]. AF is a major risk factor for stroke, and consequently, this protein may be a link between AF recurrence and a higher risk of stroke.

IGHA2 was reported to be a biomarker of subclinical atherosclerosis in the PESA cohort, and it was validated in the AWHS and ILEVAS cohorts [26]. Atherosclerosis and FA are inflammatory diseases that share several risk factors, such as hypertension or diabetes, and can coexist, leading to an increased risk of stroke and myocardial infarction [27]. However, IGHA2 has been previously reported to have a favorable prognosis in breast cancer [28]. Finally, in a proteomics analysis of chronic kidney disease, GNPTG was associated with a lower risk of the disease in very large cohorts [29], and it is well-known that AF is highly prevalent in subjects with chronic kidney disease [30].

Our study presents several strengths and limitations to be acknowledged. As a pilot study, its primary aim was to assess the feasibility of methodological and analytical approaches for analyzing the association between the expression of proteins and the risk of AF recurrence. We identified 9 proteins (out of the ten differential ones) associated with AF recurrence that were previously related to cardiovascular diseases or factors related to cardiac arrhythmias or related diseases. These findings have been observed in the framework of a clinical trial designed to specifically assess the effect of a Mediterranean diet intervention on AF recurrence with the participation of cardiac electrophysiologists. However, we need to be cautious with the interpretation of these results because of the small sample size. Although the associations remained robust after adjusting for multiple variables, a larger study is needed to account for additional potential confounders and to enable stratified analyses. Overall, our study serves as a hypothesis-generating effort to further explore the molecular mechanisms underlying AF recurrence.

This pilot study identifies ten differentially expressed serum proteins associated with AF recurrence, offering potential biomarkers for improved prediction and management. A larger study in the PREDIMAR trial, with a larger sample size and follow-up, will be needed to further analyze the potential predicting role of these proteins and the potential interactions with the intervention.

We thank all the volunteers for the participation and personnel for their contribution to the PREDIMAR trial. We also thank INNOLIVA company for providing the extra-virgin olive oil for PREDIMAR participants.

The study was conducted according to the guidelines of the Declaration of Helsinki. PREDIMAR protocol was reviewed and approved by the Ethics Committees at each of the participating sites. This full list of participating sites and Ethics Committees can be found in the study by Barrio-Lopez et al. [13]. Written informed consent was obtained from all subjects involved in the study.

The authors have no conflicts of interest to declare.

This study was supported by the Spanish Government Official Agency for funding biomedical research – Instituto de Salud Carlos III (ISCIII) – with competitive grants through the Fondo de Investigación Sanitaria y Fondo Europeo de Desarrollo Regional (PI17/00718, PI17/00748, PI17/01870) and ProteoRed; the Regional Government of Navarra (46/2016); and the Spanish Society of Cardiology (FEC/2016).

Conceptualization and project administration: M.R.-C. and C.R. Methodology: J.F.-I., E.S., M.R.-C., M.A.M.-G., and C.R. Software: J.F.-I. and E.S. Validation: C.R., J.F.-I., E.S., and M.R.-C. Formal analysis: C.R., J.F.-I., and M.R.-C. Investigation: M.T.B.-L., P.R., R.M.-R., A.I.C., E.C., J.L.I.-C., L.T., I.G.-L., M.A.M.-G., and J.A. Resources: J.A., M.R.-C., M.T.B.-L., L.T., J.L.I.C., I.G.-C., and E.C. Data curation: L.G., J.F.-I., E.S., and C.R. Writing – original draft preparation: C.R. Writing – review and editing: all the authors. Visualization: C.R. and J.F.-I. Supervision: J.F.-I. and M.R.-C. Funding acquisition: J.A., M.R.-C., M.T.B.-L., B.L., S.R., L.T., J.L.I.C., I.G.-B., E.C., and M.A.M.-G.

The database used for the analyses of this pilot study can be consulted at the link: https://predimar.es/db/pilotstudy_proteomics_PREDIMAR2024_public.xls.