Exercise‐Dependent Modulation of Immunological Response Pathways in Endurance Athletes With and Without Atrial Fibrillation

Abstract

Background

Atrial fibrillation (AF) is a common arrhythmia characterized by uncoordinated atrial electrical activity. Lone AF occurs in the absence of traditional risk factors and is frequently observed in male endurance athletes, who face a 2‐ to 5‐fold higher risk of AF compared with healthy, moderately active males. Our understanding of how endurance exercise contributes to the pathophysiology of lone AF remains limited. This study aimed to characterize the circulating protein fluctuations during high‐intensity exercise as well as explore potential biomarkers of exercise‐associated AF.

Methods and Results

A prospective cohort of 12 male endurance cyclists between the ages of 40 and 65 years, 6 of whom had a history of exercise‐associated AF, were recruited to participate using a convenience sampling method. The circulating proteome was subsequently analyzed using multiplex immunoassays and aptamer‐based proteomics before, during, and after an acute high‐intensity endurance exercise bout to assess temporality and identify potential markers of AF. The endurance exercise bout resulted in significant alterations to proteins involved in immune modulation (eg, growth/differentiation factor 15), skeletal muscle metabolism (eg, α‐actinin‐2), cell death (eg, histones), and inflammation (eg, interleukin‐6). Subjects with AF differed from those without, displaying modulation of proteins previously known to have associations with incident AF (eg, C‐reactive protein, insulin‐like growth factor‐1, and angiopoietin‐2), and also with proteins having no previous association (eg, tapasin‐related protein and α₂‐Heremans‐Schmid glycoprotein).

Conclusions

These findings provide insights into the proteomic response to acute intense exercise, provide mechanistic insights into the pathophysiology behind AF in athletes, and identify targets for future study and validation.

Nonstandard Abbreviations and Acronyms

AHSG

α₂‐Heremans‐Schmid Glycoprotein

FDR

false discovery rate

INOH

Integrating Network Objects With Hierarchies

LLOQ

lower limit of quantification

sE‐Selectin

soluble CD62 antigen‐like family member E

SomaScan

slow off‐rate modified aptamer

sVCAM‐1

soluble vascular cell adhesion molecule‐1

TAPBP

tapasin‐related protein

Research Perspective.

What Is New?

• The mechanisms of atrial fibrillation (AF) in otherwise healthy male endurance athletes are poorly understood but may involve immune modulation and inflammation.

• By assessing the dynamic circulating proteome during acute high‐intensity exercise in highly trained male endurance athletes with AF versus non‐AF controls, known and novel markers were discovered to be associated with AF, potentially linking altered exercise responses to AF pathogenesis.

What Question Should Be Addressed Next?

• This hypothesis‐generating study highlights new markers and pathways that can be interrogated in larger cohorts in the future.

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia in the world, affecting an estimated 46.3 million individuals worldwide (0.4%–1% of the general population). ¹ The development of AF may be secondary to several structural heart diseases ² and is strongly linked to common risk factors including advanced age, hypertension, ³ diabetes, ⁴ sleep apnea, ⁵ and obesity. ⁶ However, even in the absence of disease, the risk of AF, colloquially known as lone AF in this context, can be increased in specific populations, with male endurance athletes having a 2‐ to 5‐fold greater risk than healthy, modestly active, age‐matched men. ⁷ Given a lower relative incidence of well‐established clinical risk factors for AF in endurance athletes, other mechanisms including repetitive hemodynamic load inducing anatomic ⁸ and electrical remodeling (eg, elevated parasympathetic tone) ⁹ as well as the generation of inflammatory proarrhythmogenic substrates ¹⁰ have been suggested; however, these hypotheses remain contentious. While the multifactorial nature may be poorly understood, the mechanisms underpinning lone AF in highly trained athletes are likely attributable to the accepted triad of triggering factors (eg, inflammation), modulators (eg, environmental factors), and proarrhythmic substrates (eg, atrial remodeling).

Among the triad, the link between cardiac inflammation and AF has been extensively examined, with many studies highlighting oxidative and metabolic stress as dynamic regulators of atrial electroanatomical remodeling. ¹¹ , ¹² Notably, the recruitment of inflammatory immune cells and expression of proinflammatory cytokines in the atrial myocardium may induce a proarrhythmogenic state, thereby acting as a precursor to AF. Prior efforts to characterize the role of inflammation in AF have centered largely around the high‐throughput analysis of multimorbid, nontrained populations using genome‐wide association approaches and targeted profiling of the circulating proteome. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ As such, to gain a deeper understanding of the pathogenesis of exercise‐associated AF, it becomes essential to uncover molecular markers specifically associated with a highly trained, comorbid‐free endurance athlete population. In this respect, biochemical profiling of plasma proteins using aptamers has gained recognition for its ability to discover gene–protein interactions, ¹⁷ drug pharmacology, ¹⁸ and biomarkers for individual disease risks such as cardiovascular disease ¹⁹ owing to its dynamic range (≈100 fmol/L–1 μmol/L) and demonstrated high assay reproducibility. Aptamers are short strands of DNA or RNA that bind directly to proteins and, in turn, are used for recognition and measurement. Despite its potential, aptamer‐based profiling has not yet been used to either elucidate the short‐term modulations associated with endurance exercise or for the identification of markers related to AF. In this proof‐of‐concept study based on a cohort of highly trained middle‐aged endurance athletes, we evaluated the ability of aptamer‐based proteomics to simultaneously interrogate the fluctuations of the circulating proteome that occur during a regimen of high‐intensity exercise and subsequently identify biomarkers potentially associated with exercise‐associated AF.

Methods

The authors declare that all supporting data, including reagent identifiers (Table S1), are available within the article and its online supplementary files. Additional methods can be found in Data S1. Information and requests for resources or patient‐level data sets from qualified researchers trained in human subject confidentiality protocols may be sent to the lead contacts, Dr Jack M. Goodman (jack.goodman@utoronto.ca) or Dr Jason Fish (jason.fish@utoronto.ca).

Study Participants

Twelve middle‐aged male endurance cyclists of European descent between the ages of 40 and 65 years, 6 of whom had a documented history of AF (irrespective of their AF treatment or burden) were recruited to participate. A self‐report screening form was used to ascertain exercise experience and medical history, wherein participants were asked about history of exercise, including type, frequency, and intensity, as well as a personal history for details, including alcohol intake, use of prescription and nonprescription drugs, smoking, and any medical conditions. Participants with AF were asked to complete a separate AF history form that included questions on AF diagnosis, burden of disease, symptoms, triggers, relationship to exercise, and any treatments. All athletes had at least 10 years of consistent endurance training, a minimum of 2000 total hours, and were not in a permanent AF state. Participants were recruited using a convenience sampling method, primarily through community‐based outreach, which involved advertising through local cycling and triathlon clubs, as well as word‐of‐mouth. Athletes were excluded if they self‐reported any existing structural heart disease, valvular heart disease, coronary artery disease, heart failure, hypertension, diabetes, kidney disease, obstructive sleep apnea, systemic inflammatory diseases, a history of antibody‐mediated immunotherapy use (biologics), recreational drug use, smoking, or excessive alcohol consumption (exceeding 14 drinks per week). Subjects reporting use of corticosteroids or any other anti‐inflammatory medications (except for nonsteroidal anti‐inflammatory drugs) within 6 months were also excluded. Study participants were asked to refrain from using nonsteroidal anti‐inflammatory drugs within 1 week of experiment participation. All study participants provided written informed consent in accordance with protocols approved by the research ethics boards of the University of Toronto (research ethics board No. 36400) as well as the University Health Network (research ethics board No. 21‐6011.0). The study and all associated protocols abided by the ethical principles for medical research set forth by the Helsinki Declaration. ²⁰ Study reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for human cohort studies (https://www.strobe‐statement.org/).

Exercise Intervention

Graded Exercise

All participants underwent a graded exercise protocol on an electronically braked cycle ergometer (Excalibur Sport, Lode BV, Groningen, Netherlands) until volitional exhaustion or an inability to maintain 60 revolutions per minute. Breath‐by‐breath pulmonary gas exchange (Encore 229, CareFusion, Yorba Linda, CA) was recorded throughout. After a 50‐W warm‐up, exercise intensity increased every minute by 30 W with completion ranging between 8 and 12 minutes; peak rate of oxygen consumption was identified as the highest 30‐second average.

Prolonged Endurance Exercise

Participants returned on a separate visit for a prolonged endurance effort that was completed on their own bicycle, with the use of an indoor cycling trainer. This session was designed to mimic a high‐intensity training session or race. The training session consisted of a 15‐minute warm‐up (self‐guided with gradually increasing power output to a target heart rate of 70%–80% of the maximum achieved during the graded exercise protocol). This was followed by 90 minutes of a sustained high‐intensity effort. The participants performed 8 intervals of 9 minutes each, maintaining an output of 70%–80% of their maximum power output, followed by a 1‐minute interval at an increased intensity of >90% of their maximum power output. In addition to these intervals, participants were also asked to cycle at maximal effort for the last 10 minutes to a maximum tolerated effort to mimic a high‐intensity training or race scenario.

Physiologic Data Collection

Every 10 minutes the participants' exercise heart rate (Polar V800), power output (measured using participant‐provided bicycle trainers when available), and blood pressure (SunTech Medical Tango M2) were recorded. In addition, perceived exertion was collected using the Borg 6 to 20 scale. ²¹ Physiological measures including body weight, hematocrit (point‐of‐care hematocrit; IEC MB Microhematocrit Centrifuge; Block Scientific, Bellport, NY), and temperature (tympanic temperature; Braun Thermoscan Model 6022; Braun Medical, Bethlehem, PA) were measured before and immediately after exercise. Peripheral blood samples (2 mL) were collected from the cubital vein into BD Vacutainer Blood Collection Tubes (BD Bioscience, Franklin Lakes, NJ) containing dipotassium EDTA at 4 time points: (1) immediately before the endurance effort (ie, preexercise), (2) after 1 hour of exercise (ie, midexercise), (3) immediately at the conclusion of the endurance effort (ie, postexercise), and (4) 1 hour after the endurance effort (ie, recovery; Figure 1). A single participant experienced AF exacerbation midtrial and subsequently missed the midpoint sample collection; this participant was included in the validation cohort, but not the SomaScan aptamer‐based profiling cohort. All subsequent methodology pertains to this prolonged endurance exercise session.

Figure 1. Schematic of study design.

Samples were collected before, during, and after high‐intensity exercise and during recovery in male endurance athletes with or without atrial fibrillation. Circulating protein changes were assessed by multiplex immunoassays or aptamer‐based proteomics, and differentially regulated proteins and pathways were identified. AF indicates atrial fibrillation; and RBC, red blood cell.

Targeted Protein Biomarker Analysis

Circulating levels of angiopoietin‐2 (lower limit of quantification [LLOQ], 9.9 pg/mL), soluble CD62 antigen‐like family member E (sE‐Selectin; LLOQ, 4.22 pg/mL), soluble CD54/intercellular adhesion molecule‐1 (LLOQ, 4.1 pg/mL), soluble vascular cell adhesion molecule‐1 (sVCAM‐1; LLOQ, 137 pg/mL), CD105/endoglin, endothelin‐1 (LLOQ, 0.250 pg/mL), interleukin‐6 (LLOQ, 0.41 pg/mL), interleukin‐8 (LLOQ, 0.19 pg/mL), and soluble triggering receptor expressed on myeloid cells‐1 (LLOQ, 4.19 pg/mL) were quantified in platelet‐free plasma samples using the Simple Plex Ella (ProteinSimple, San Jose, CA) platform according to the manufacturer's instructions (reported as the average of triplicate readings).

Aptamer‐Based Proteomics Profiling

The modified aptamer binding reagents, ²² SomaScan assay technique, ²³ performance characteristics, ²⁴ , ²⁵ and analysis ²⁶ have been described previously. In brief, protein levels for ≈7000 analytes (Table S2) in the archived dipotassium EDTA plasma samples were measured by the SomaScan platform, which uses single‐stranded DNA‐based aptamers to translate protein concentrations into DNA signals measurable by standard DNA detection methodologies. Plasma samples had a single controlled freeze–thaw cycle on ice before proteomics profiling, which was required for aliquoting and shipping purposes. Target annotation and mapping of aptamers to UniProt accession numbers as well as Entrez gene identifiers were provided by SomaLogic; normalization information has been provided (Figures S1 through S4).

Analysis of Aptamer‐Based Proteomics Profiling

SomaScan proteomic data are reported in relative fluorescence units, as previously described. ²³ Relative fluorescence units were log‐transformed before statistical analysis to reduce heteroscedasticity, with P values adjusted for multiple testing by false discovery rate (FDR) and subsequently reported as the q‐value. Enrichment analysis using the full proteomics data set was performed using the ConsensusPathDB‐human (http://cpdb.molgen.mpg.de/), ²⁷ a database for integrating human molecular interaction networks from 31 public resources including the Kyoto Encyclopedia of Genes and Genomes, Reactome, Wikipathways, Pathway Interaction Database, and the Integrating Network Objects With Hierarchies (INOH) database as well as Gene Set Enrichment Analysis. ²⁸ Enrichment sets whose hypergeometric q value passed a q<0.05 threshold defined were listed and, for each set, the set size provided.

Statistical Analysis, Data Visualization, and Data Availability

Where appropriate, data were analyzed with either GraphPad Prism version 9.0.1 for MacOS (GraphPad Software, Inc., La Jolla, CA) or the SOMALogic Data Viz platform (https://stats.somalogic.com/). Dot plot visualizations and heatmaps were generated in R Studio version 1.3.1056 using ggplot2, while the study schematic was generated using BioRender. Demographics–Baseline characteristics were summarized using mean and SD for normally distributed continuous variables or median and interquartile range for skewed continuous variables; normality was assessed using the Shapiro–Wilk test (Table S3). Simple Plex Ella ELISA–Multiple comparisons testing of protein expression data was facilitated by mixed‐effects modeling with Šídák's multiple comparisons test. SomaScan–Proteomics data were assessed for changes across exercise time points using a repeated measures ANOVA test. Proteins that were significantly different at the midexercise, postexercise, or recovery time points compared with the preexercise time point (ie, displaying differences between at least 2 groups) were considered to be differentially regulated proteins. Assessing for changes across AF status was accomplished by averaging intraparticipant measurements (ie, preexercise, midexercise, post‐exercise, and recovery) and conducting an unpaired t test. Select data from the Simple Plex Ella ELISA and SomaScan were compared using a Spearman's rank correlation coefficient where available. Although many hypotheses were tested throughout the study, no experiment‐wide multiple‐test correction was applied. Unless indicated otherwise, graphs depict averaged values of technically replicated (indicated as appropriate) independent biological samples and have error bars displayed as mean±SD. Significance thresholds of P<0.05, P<0.01, and P<0.001 were used as standard of practice and indicated in the figures as *, **, and ***, respectively, along with the numerical P value unless indicated otherwise. Further information and requests for resources, reagents, or patient‐level data sets (including access to the deposited raw SomaScan data; https://zenodo.org/records/10079349) from qualified researchers trained in human subject confidentiality protocols may be sent to the lead contacts, Dr Jack M. Goodman (jack.goodman@utoronto.ca) or Dr Jason Fish (jason.fish@utoronto.ca).

Results

Participant Demographics and Physiological Characteristics

The cohort was composed of 12 experienced endurance athletes aged 53±6 years (mean±SD) divided into 2 groups: those with and those without AF (n=6 per group). All participants had considerable background in exercise training, having a mean training experience of 36±13 years, training a median 8 (interquartile range, 7–17) hours per week cumulating to a mean weekly cycling distance of 210±97 km. The median baseline peak rate of oxygen consumption was 50 (interquartile range, 47–62) mL/kg per min (Table 1).

Table 1.

Baseline Demographic and Physiologic Characteristics

Characteristic	Athletes without AF (n=6)	Athletes with AF (n=6)	Combined (n=12)	P value
Age, y	55±8	52±5	53±6	0.57
Average exercise per week, h*	11 (8–19)	8 (7–9)	8 (7–17)	0.21
Endurance experience, y	44±8	28±11	36±13	0.017
Average yearly cycling distance, km	8550±2370	7167±3907	7858±3164	0.48
Exercise per year, h	356±91	594±426	475±319	0.21
Average speed, km/h	24±6	14±11	18±9	0.15
VO2peak, mL/kg per min*	59 (49–62)	48 (46–56)	50 (47–62)	0.12
Average power output, W†	217±21	222±58	219±33	0.86
Resting heart rate, beats/min	65±13	56±4	60±10	0.13
Resting systolic pressure, mm Hg	124±15	126±12	124±12	0.85
Resting diastolic pressure, mm Hg	82±5	80±11	81±8	0.61
Hematocrit before exercise, %	45±2	45±1	45±2	0.99
Hematocrit after exercise, %‡	46±2	45±2	46±2	0.77
Weight before exercise, kg*	73 (66–80)	79 (77–86)	78 (71–81)	0.093
Weight after exercise, kg* , §	71 (65–78)	77 (75–88)	75 (68–78)	0.13
Average exercise heart rate, beats/min*	153 (143–156)	156 (142–163)	151±11	0.56

VO_2peak indicates peak rate of oxygen consumption.

^*Values are presented as the median (25th–75th percentiles); all other values are mean±SD.

^† Power data were available for 8 of 12 participants.

^‡ Data for hematocrit after exercise were available for 11 of 12 participants.

^§ Data for weight after exercise were available for 11 of 12 participants.

In total, 11 of 12 athletes successfully completed the endurance effort, with 1 athlete having to stop prematurely due to the development of symptomatic AF. As intended, the effort was at high intensity with an average heart rate of 151±11 bpm, with all athletes reaching 20 of 20 on the Borg scale for perceived exertion at the end of the effort. Power data were available for 8 of 12 participants, and they sustained an average power output of 219±33 W (Table 1). None of the athletes with AF were on chronic rate control or antiarrhythmic medications at the time of the study. One athlete previously used pill‐in‐pocket antiarrhythmic and 3 had an ablation, while the remainder did not receive therapy (see detailed AF history; Table S4).

Markers of Inflammation and Endothelial Cell Activation Are Dynamically Regulated During High‐Intensity Exercise

Since inflammatory factors have been implicated as instigating triggers for AF, we measured the expression of 8 circulating inflammatory or endothelial cell activation biomarkers in all 12 endurance athletes using a benchtop multiplex enzyme‐linked immunosorbent automated platform. Notably, there were several inflammatory markers (eg, interleukin‐6, interleukin‐8, soluble triggering receptor expressed on myeloid cells‐1) and markers of vascular activation (eg, angiopoietin‐2) that were induced midexercise (Figure 2A). These remained elevated during the postexercise period but returned to near baseline levels during recovery. In contrast, markers of endothelial activation and leukocyte adhesion (eg, soluble CD54/intercellular adhesion molecule‐1, sE‐Selectin, sVCAM‐1) were largely unchanged, except for the downregulation of sVCAM‐1 during recovery. Interestingly, endothelin‐1, a vasoconstrictor and profibrotic molecule, ²⁹ was induced during recovery (Figure 2B).

Figure 2. Plasma concentration of select endothelial and inflammatory markers across exercise time points.

Plasma concentrations of the inflammatory markers angiopoietin‐2, interleukin‐6, interleukin‐8, and soluble triggering receptor expressed on myeloid cells‐1 (A) and the endothelial markers endothelin‐1, soluble CD54/intercellular adhesion molecule‐1, sE‐Selectin, and sVCAM‐1 (B) stratified by collection time point with visual representations of mean±SD. P values for multiple group comparisons were determined by mixed‐effects analysis with Šídák's multiple comparisons test. ANGPT2 indicates angiopoietin‐2; IL, interleukin; sE‐Selectin, soluble CD62 antigen‐like family member; sICAM‐1, soluble CD54/intercellular adhesion molecule‐1; sTREM‐1, soluble triggering receptor expressed on myeloid cells 1; sVCAM‐1, soluble vascular cell adhesion molecule‐1. *, **, and *** indicate P<0.05, P<0.01, and P<0.001, respectively.

Plasma Proteins Associated With Endurance Exercise Among Middle‐Aged Male Endurance Cyclists

To better understand the temporal dynamics of the circulating proteome during exercise and to identify potential distinctions in the proteome of those with a history of AF, we conducted an exploratory analysis in a subset (3 without AF and 3 with AF) of the cohort measuring ≈7000 proteins in collected plasma samples using a multiplexed SOMAmer assay (Table S2). Samples were assessed at 4 time points (ie, preexercise, midexercise, postexercise, and recovery). Since the characterization of the longitudinal change within each participant was a primary end point, all the time points from each subject, along with all subjects, were included in the same plate. Assessment of intraindividual variability by component analysis revealed a proteomic profile effectively separated on the basis of interparticipant measures as opposed to clustering of exercise time points (Figure 3A). This suggests that there were more interindividual differences in the protein expression, as opposed to the acute effects of exercise. Among the entire group, 17 circulating proteins were significantly modulated at the midexercise, postexercise, or recovery time point compared with the preexercise time point (between at least 2 groups), as determined by FDR‐adjusted P values and a fold‐change cutoff of ≥1.5 during an endurance regimen of exercise (Figure 3B). In total, 962 circulating proteins were differentially enriched, according to only FDR‐adjusted P values (ie, no fold‐change cutoff; Table S5 displays the maximum fold change at any time point compared with preexercise).

Figure 3. Pathway analysis of whole plasma aptamer‐based proteomics reveals that exercise modulates immune and angiogenic pathway sets.

A, Principal component analysis demonstrated participant‐specific clustering of samples; grouping of individual patients is represented by the dotted circles. B, Seventeen proteins were significantly differentially enriched across the 4 time points (ie, midexercise, postexercise, or recovery vs preexercise) using filtering thresholds of q<0.05 and fold change >1.5 (dark red), while 962 proteins were differentially enriched across the 4 time points using only a statistical filtering threshold of q<0.05 (light red, increased expression; light blue, decreased expression); fold changes are determined as the maximum fold change between medians of all groups and the preexercise time point. C, Selected KEGG, Reactome, INOH, PID, and BioCarta pathways significantly enriched among proteins significantly increased (q value). D, Selected KEGG, Reactome, INOH, PID, and BioCarta pathways that were significantly enriched among proteins significantly decreased across exercise (q value; between at least 2 time points). Enrichment heatmap of specific proteins throughout the course of the exercise bout scaled by Z score. E, Z scored heatmap of raw count data displaying cardiac and immunoregulatory proteins differentially expressed as a result of the exercise bout. Analyses used the preexercise time point as the referent. EPHA indicates EphA/ephrin‐A; FDR, false discovery rate; GDF‐15, growth/differentiation factor 15; IL, interleukin; INOH, Integrating Network Objects With Hierarchies; KEGG, Kyoto Encyclopedia of Genes and Genomes; MAPK, mitogen‐activated protein kinase; MBP‐C, myosin binding protein C, cardiac; PID, Pathway Interaction Database; and TNF‐α, tumor necrosis factor‐α.

To identify the biological processes associated with exercise, up‐ and downregulated proteins from the proteomics analysis (selected through FDR‐adjusted P values only) were used for enrichment analysis conducted through ConsensusPathDB–human. Genes involved in immune activation, receptor signaling, and atherosclerosis were upregulated, while those involved in scanning of messenger RNAs and initiation of protein synthesis were downregulated (Figure 3C and 3D). Looking at specific subgroups of proteins, it could be seen that traditional cardiac markers, such as NT‐proBNP (N‐terminal pro‐B‐type natriuretic peptide), atrial natriuretic factor, and cardiac troponin T isoform, were significantly modulated throughout the exercise bout but did not cross thresholds that would be considered abnormal in a clinical context (Figure 3E). Interassay concordance between the SomaScan proteomics data set and the Ella immunoassay platform was also assessed using the expression of the 8 circulating proteins measured previously (Figure 2). Spearman correlation coefficients among data points were moderate (average r _s=0.42), with temporal trends closely mimicking the aptamer‐based SomaScan data (Table S6).

Proteins involved in muscle contraction displayed significant modulations across time, with the force‐related proteins α‐actinin‐2, myomesin‐2, myosin‐binding protein C, and myosin light chain 3 displaying an increasing trend with exercise, along with the essential myosin light‐chain elements that regulate force production during muscular cross‐bridge cycles (Figure 3E). Several muscle‐isoform–specific enzymes involved in glycolysis, including β‐enolase, and lactate dehydrogenase‐α were also significantly modulated. Looking at markers typically associated with the endothelium, it could be seen that markers associated with injury such as endothelin‐1, angiopoietin‐2, and vascular endothelial growth factor receptor 1 were significantly modulated, while markers of angiogenesis such as the vascular endothelial growth factors (ie, A, B, C, D) and platelet‐derived growth factor (ie, A, B, C, D) were unchanged; with the exception of platelet‐derived growth factor receptor‐like protein (Table S5). Among immunomodulatory proteins, interleukin‐23 receptor, interleukin‐6, interleukin‐1 adapter protein, interleukin‐16, tumor necrosis factor α (TNF‐α)‐induced protein 8, and interleukin‐17D appeared to have more predominant expression in the period immediately after exercise (ie, postexercise), while interleukin‐17 receptor C and interleukin‐5 receptor subunit α had decreasing expression throughout exercise. Interestingly, during the endurance bout, a considerable number (7/17) of the top differentially expressed proteins (q<0.05 and fold‐change >1.5), were circulating nucleoproteins, including increased histone proteins counts (Table 2).

Table 2.

Differentially Modulated Circulating Proteins Associated With Endurance Exercise

Protein	UniProt ID	Maximum fold change*	FDR
T‐cell surface antigen CD2	P06729	4.9	8.20E‐07
Platelet factor 4	P02776	3.8	9.70E‐03
Histone H2B type 3‐B	Q8N257	3.0	2.20E‐06
Growth hormone variant	P01242	2.9	4.20E‐05
Histone H2B type 2‐E	Q16778	2.6	5.40E‐06
Histone H2B type 1‐K	O60814	2.5	4.90E‐06
Histone H2B type 2‐E	Q16778	2.4	4.00E‐06
Histone H2A type 1‐A	Q96QV6	2.3	7.40E‐06
Histone H2A type 1	P0C0S8	2.1	4.90E‐06
IGF‐binding protein 1	P08833	2.1	4.10E‐04
Histone H2A type 3	Q7L7L0	2.0	8.90E‐06
C‐C motif chemokine 5	P13501	1.9	4.10E‐02
von Willebrand factor	P04275	1.8	2.60E‐03
Metalloproteinase inhibitor 3	P35625	1.7	1.80E‐04
CRSP LCCL domain‐containing 2	Q9H0B8	1.7	6.60E‐03
VEGFR1	P17948	1.6	1.70E‐05
Oligodendrocyte‐myelin glycoprotein	P23515	1.6	1.90E‐04

CD2 indicates cluster of differentiation 2; CRSP, cysteine‐rich secretory protein; FDR, false discovery rate; IGF, insulin‐like growth factor; and VEGFR1, vascular endothelial growth factor receptor 1.

^*Fold change is described as the maximum fold change between medians of all exercise groups, with the preexercise time point as the referent.

Differences in Plasma Proteins Between Subjects With and Without AF

While incident AF‐related proteins have been reliably explored in other cohorts, to date it has been unclear if the proteins uncovered can be reliably seen in otherwise healthy individuals. To address this question, we averaged the time‐course measurements (ie, preexercise, midexercise, postexercise, and recovery) and subsequently separated the groups on the basis of their AF status. Visual inspection of principal component analysis by AF revealed participant separation was at least in part due to their AF status (Figure 4A). There were 20 circulating proteins that were differentially expressed between the 2 groups (AF and non‐AF), as determined by FDR‐adjusted P values and a fold‐change cutoff of >1.5 during an endurance bout of exercise (Figure 4B). This included upregulation of TAPBP ([tapasin‐related protein] in those with AF), an integral interferon‐γ–inducible component of the peptide‐loading complex for efficient peptide loading onto major histocompatibility complex class I molecules. ³⁰ In total, 446 circulating proteins were differentially enriched, according to FDR q values only (Table S7). These included proteins known to be associated with incident AF, including C‐reactive protein, ³¹ sVCAM‐1, ¹³ insulin‐like growth factor‐1, ³² and angiopoietin‐2. ¹⁴ To identify the biological processes associated with exercise, enrichment analysis of up‐ and downregulated proteins (selected through FDR‐adjusted P values only) was again conducted through ConsensusPathDB–human. Gene sets involved in inflammation, hematopoiesis, and immune signaling were upregulated in those with AF, while pathways involved in sulfonation, chemokine receptor interaction and binding, and extracellular matrix organization were downregulated (Figure 4C and 4D). Among the 20 proteins displaying both high fold changes (>1.5) and significant modulation (q<0.05), there was a wide diversity of biological functions observed, including those influencing immunomodulation (ie, paired immunoglobulin‐like type 2 receptor‐α isoforms), inflammation (ie, α₂‐HS‐glycoprotein), and transcriptional regulation (ie, Ripply Transcriptional Repressor 3; Table 3).

Figure 4. Pathway analysis of whole plasma aptamer‐based proteomics reveals modulated inflammatory signaling in endurance athletes with atrial fibrillation.

A, Principal component analysis demonstrated participant‐specific clustering of samples. B, All values from the 4 time points (ie, preexercise, midexercise, postexercise, recovery) in each individual were averaged. A total of 20 proteins were differentially expressed using filtering thresholds of q<0.05 and fold change >1.5 between those with AF and those without (dark red, higher in lone AF; dark blue, lower in lone AF), while 446 proteins were differentially enriched between the 2 groups, q<0.05 with no fold‐change cut‐off (light red, higher in lone AF; light blue, lower in lone AF). C, Selected KEGG, Reactome, INOH, PID, and BioCarta pathways that were significantly enriched among those proteins significantly increased across the endurance bout. D, Selected KEGG, Reactome, INOH, PID, and BioCarta pathways that were significantly enriched among those proteins significantly decreased across the endurance bout. AF indicates atrial fibrillation; FDR, false discovery rate; GPCR, G‐protein coupled receptor; JAK STAT, Janus kinase‐signal transducer and activator of transcription; IL, interleukin; INOH, Integrating Network Objects With Hierarchies; KEGG, Kyoto Encyclopedia of Genes and Genomes; and PID, Pathway Interaction Database.

Table 3.

Differentially Expressed Proteins Across AF Status

Protein	UniProt ID	Fold change*	FDR
α2‐HS‐glycoprotein	P02765	−5.7	1.20E‐02
PIGL type 2 receptor‐α isoform FDF03‐deltaTM	Q9UKJ1	−5.2	2.70E‐02
GMP reductase 1	P36959	−3.4	2.20E‐02
PIGL type 2 receptor‐α isoform FDF03‐M14	Q9UKJ1	−2.8	1.20E‐02
40S Ribosomal protein SA	P08865	−2.0	3.10E‐02
E3 ubiquitin‐protein ligase RNF31	Q96EP0	−1.8	1.20E‐02
Interleukin‐18 receptor accessory protein	O95256	−1.8	1.50E‐02
Killer cell immunoglobulin‐like receptor 2DL5A	Q8N109	−1.7	2.20E‐02
Ripply transcriptional repressor 3	P57055	−1.6	2.00E‐02
Growth‐regulated oncogene beta	P19875	1.6	4.70E‐04
Linker for activation of T‐cell family member 1	O43561	1.6	6.00E‐03
Uracil‐DNA glycosylase	P13051	1.6	1.60E‐02
Muscle, skeletal receptor tyrosine‐protein kinase	O15146	1.6	3.70E‐02
Haptoglobin	P00738	1.7	5.80E‐03
Butyrophilin subfamily 1 member A1	Q13410	1.7	8.80E‐03
Sperm acrosome membrane‐associated protein 3	Q8IXA5	1.8	3.00E‐02
Ectonucleotide family member 7	Q6UWV6	1.8	3.50E‐02
Sulfotransferase 1A1* 2	P50225	2.1	2.10E‐02
Corticoliberin	P06850	2.1	2.60E‐02
Tapasin‐related protein	Q9BX59	2.5	8.80E‐05

FDR indicates false discovery rate; GMP, guanosine 5′‐monophosphate oxidoreductase; HS, Heremans–Schmid; PIGL, paired immunoglobulin‐like; and RNF31, ring finger protein 31.

^*Athletes with atrial fibrillation were compared with athletes without atrial fibrillation (ie, the referent). The values at each time point (ie, preexercise, midexercise, postexercise, and recovery) were averaged for each individual to facilitate a 2‐group comparison.

Within the targeted protein analysis (measured by ELISA), using the entire 12‐patient cohort, there was a significant difference in sE‐Selectin between individuals with and without AF (Figure S5). Levels remained relatively stable across all 4 time points in the AF group; however, there was a significant decrease in the non‐AF cohort immediately after exercise, returning to baseline 1 hour later.

Discussion

We performed an exploratory analysis of ≈7000 proteins in 6 well‐trained, middle‐aged endurance athletes and found that an acute bout of endurance exercise was associated with a marked change in the circulating proteome, a significant increase in the number of circulating histones, and altered signaling in vascular remodeling pathways. Notably, there were also differences in the circulating proteome and the detection of select proteins between subjects with and without a history of AF.

To our knowledge, this study is the first to explore the large‐scale plasma proteome changes in a highly trained endurance population, as well as in those with and without a history of AF. Taking advantage of a high sampling density to capture the immediate and acute effects of an endurance exercise bout, our study highlights the transient nature of inflammatory as well as vascular proteins within a highly trained population of athletes and reveals differences in markers of dynamic immune modulation in the AF subgroup, compared with those without AF.

Proteomic Response to Acute Endurance Exercise

Approximately 13% of the circulating proteins on our platform were modified as a result of the endurance bout, a percentage in line with recent estimates from a 20‐week endurance exercise training program, ³³ including several established hormones (ie, leptin, adiponectin, and insulin‐like growth factors) as well as exercise‐secreted bioactive factors known as exerkines (ie, growth differentiation factor 15 and interleukin‐8). A sharp increase in the plasma concentrations of glycolytic mediators including pyruvate kinase muscle isozyme and l‐lactate dehydrogenase A chain was also observed, which presumably reflect heightened anaerobic metabolism. In line with this observation, oxidative stress signaling was apparent, with elevations of myeloperoxidase throughout all time points during and after exercise. Predominantly released from neutrophils via degranulation, myeloperoxidase is thought to signal skeletal muscle damage or stress and recruit macrophages to damaged sites. ³⁴ Activated neutrophil signaling was apparent throughout the bout, as evident by elevated levels of neutrophil cytosol factor‐1, neutrophil gelatinase‐associated lipocalin, neutrophil elastase, and elevation in histone proteins among others.

Although an acute proinflammatory response was apparent, markers traditionally associated with skeletal muscle damage were not significantly changed over the course of the exercise bout, with none of the myoglobin, creatine kinase, aspartate transferase, or skeletal isoforms of troponin displaying elevations. Well‐described temporal studies support these findings, with elevations of damage markers observed to occur after longer bouts and later after exercise; creatine kinase in particular increased only after 12 hours following exercise and peaked 4 to 6 days later in other studies. ³⁵ , ³⁶ In contrast, 2 markers of cardiomyocyte damage displayed significant elevations throughout the bout, including cardiac troponin T and NT‐proBNP; however, neither exceeded clinical thresholds for overt cardiac injury. The elevations in NT‐proBNP may be instead related to physiological endocrine responses to the myocardial stress induced by exercise, indicative of cardiac fatigue or, additionally, increased cell wall permeability during exercise. ³⁷ Studies conducted by Scharhag et al provide similar observations, highlighting increases in NT‐proBNP after exercise and occurring in the absence of myocardial damage as observed through both echocardiography and magnetic resonance imaging. ³⁸ , ³⁹

Our data demonstrate 2 novel findings within the context of endurance exercise: First, there appears to be a broad increase in the number of circulating histones, the main components of nucleosomes. Largely studied in trauma‐induced multiple organ failure and sepsis, circulating histones appear to function as prothrombotic microbicidal proteins capable of acting on adjacent cells and circulating immune cells via pattern recognition receptors. Intense exercise is known to evoke a rapid and transient increase in circulating cell‐free DNA, largely through regulated cell death dependent on the formation of neutrophil extracellular traps. ⁴⁰ While circulating histones have not been extensively studied in exercise, histones can be released passively from necrotic cells or actively by other modes of cell death during this process. ⁴¹ Increased concentrations of circulating histones could therefore reflect either increased amounts of cell death or an active effector mechanism of innate immunity, namely, the formation of neutrophil extracellular traps by peripheral blood neutrophils. The elevated levels of myeloperoxidase would be in line with this. Considering their possible significance as biomarkers in several pathologies, it would be of interest to understand the potentially different roles of histones in a process such as exercise.

Second, our protein biomarker analysis also supports the notion that a vascular remodeling signaling cascade occurs during exercise. Our findings highlight enrichment in the blood for several members of this process, including vascular endothelial growth factor receptor 1, and various extracellular matrix remodelers such as matrix metalloproteinase 9, matrix metalloproteinase 17, and fibroblast growth factor 7. Increases in angiopoietin‐2, a potent secreted growth factor that mediates its effects on angiogenesis and vascular destabilization by competing with angiopoietin 1 for access to the tyrosine‐protein kinase receptor Tie‐2, were also observed, changes that have been seen in 2 independent studies. ⁴² , ⁴³ These increases were accompanied by an apparent increase in systemic endothelial activation, given increases in interleukin‐6, and endothelial cell–specific molecules, such as von Willebrand factor as well as its proteinase, a disintegrin and metalloproteinase with thrombospondin motif 13. Research on the systemic effects of circulating angiogenic factors is warranted.

Differential Proteomic Profile of Athletes With AF Versus Without AF

Despite a lack of traditional risk factors, athletes have been demonstrated to be at greater risk for the development of AF. In a prospective case–control study, individuals with >2000 hours of lifetime‐accumulated high‐intensity exercise displayed a heightened risk of lone AF (odds ratio, 3.88 [95% CI, 1.55–9.73]), ⁴⁴ with endurance sports such as marathon running ⁴⁵ and long‐distance cycling ⁴⁶ conferring the highest risk. A dose–response has been suggested, but not conclusively demonstrated, and there is no clear threshold for increased risk. In this respect, our sample was representative of highly trained, middle‐aged male endurance athletes with AF with a high maximal oxygen consumption and years of high‐volume, high‐intensity endurance exercise.

While mechanisms that underlie the development of AF in this population are unknown, left atrial enlargement has been implicated and in the general population is associated with the development and persistence of AF. However, while athletes are well known to have increased atrial volumes, ⁴⁷ in one study comparing athletes and healthy controls, larger left atrial volumes were seen in athletes without AF than in nonathletes with AF, suggesting size alone to be an insufficient explanation. Notably, left atrial function in athletes remained preserved. ⁴⁸ Recent studies have also explored the role of atrial fibrosis as a profibrotic substrate in exercise‐associated AF, highlighting substantial increases in fibrotic markers including fibronectin‐1 and matrix metalloproteinase type I. ⁴⁹ Imaging studies have also demonstrated endurance athletes may have increased left atrial fibrosis as seen on magnetic resonance imaging. ⁵⁰ It is unclear if atrial fibrosis is occurring as a result of local mechanical stimulation or as part of a systemic inflammatory response. In our study, targeted protein analysis identified a sharp increase in the levels of endothelin‐1, 1 hour after exercise (Figure 2), a factor that is associated with atrial fibrosis. ⁵¹ Although atrial fibrosis is considered a key element of the AF substrate, with extracellular matrix remodeling, particularly by matrix metalloproteinases, playing a major role in this process, we did not observe direct enrichment of profibrotic pathways (aside from endothelin‐1, as mentioned above); however, many of the proteins modulated are upstream regulators of extracellular matrix processes. It is possible that this signal was not captured by our early sampling and may instead represent a latent process. Importantly, unlike atrial remodeling, immune remodeling is not limited to the atria, and its effects propagate across the peripheral circulation.

Inflammation has also been suggested as a potential mechanism underlying the development and progression of AF in a number of different clinical contexts. While the causal role of inflammation as a bona fide proarrhythmic substrate among the larger AF population remains controversial, there is a clear correlation between inflammatory states (ie, pericarditis, sepsis, or postsurgery) and the incidence of AF, ⁵² as well as an abundance of evidence linking inflammation and AF risk. ⁵¹ Evidence derived from both histologic studies of atrial tissue as well as several epidemiologic studies have supported this hypothesis, finding associations between incident AF with interleukins (ie, interleukin‐6 ¹⁵ , ⁵³ ), myeloperoxidase, ⁵⁴ C‐reactive protein, ³¹ and galectin‐3. ⁵⁵ In the athletic population, the role of inflammation is less clear, and current evidence comes only from animal models. In mice, exercise training has been demonstrated to promote inflammation and atrial fibrosis as well as AF inducibility. The effects of exercise on fibrosis and susceptibility were attenuated when TNF‐α and p38 mitogen‐activated protein kinase signaling were inhibited, underscoring the importance of inflammatory pathways. ⁵⁶

Our pathway analysis revealed an immune‐dominated list of canonical pathways, including immunoregulatory interactions between lymphoid and nonlymphoid cell types as well as both cytokine and interleukin signaling, represented by sustained increases to leukocyte immunoglobulin‐like receptor subfamily B member 5, paired immunoglobulin‐like type 2 receptor‐α isoform expression and decreases in the expression of circulating immunoglobulin E and interleukin‐18 receptor. While no significant differences in TNF‐α expression were observed in our cohort, small but significant increases in TNF‐α–induced protein 3, a potent inhibitor of nuclear factor κ‐light‐chain enhancer of activated B cells signaling and terminator of TNF‐induced signaling, were noted. Additionally, significant upregulation of circulating proteins with C‐X‐C motifs, proteins known to have chemotactic and activating functions on neutrophils, were observed among those with AF including chemokine (C‐X‐C motif) ligand (CXCL) 2, CXCL3, CXCL5, and CXCL16; however, no differences in myeloperoxidase release were observed. To this end, several studies have explored the neutrophil–lymphocyte ratio as a marker of AF, suggesting an imbalance in leukocytes with the dominance of neutrophils over lymphocytes could imbalance the cardiac inflammatory responses. ⁵⁷ While this supports the notion that AF could be viewed as a systemic disease process, the cause–effect relationship between AF and the diversity of immunoregulatory functions makes it difficult to define causation.

Although AF is a commonly studied arrhythmia in athletes, insights are limited to those gained from traditional cardiovascular imaging methodologies. Biomarker data and mechanistic insights are often extrapolated from larger nontrained cohorts who suffer from AF. Longitudinal cohort studies have used similar proteomics approaches to study protein biomarkers, albeit examining incident AF in multimorbid populations. ¹³ , ¹⁴ , ¹⁵ , ⁵⁸ In one recent study: only NT‐proBNP was found to be a strong predictor for the development of AF, with associations of circulating inflammatory cytokines primarily explained by clinical risk factors. ⁵⁸ In our study, we found evidence to support a statistically significant association for some previously reported AF‐related proteins from nonproteomic immunoassays including C‐reactive protein ³¹ and insulin‐like growth factor‐1. ³² However, it is interesting to note we observed either inverse associations (ie, sVCAM‐1 and angiopoietin‐2) or no associations (NT‐pro‐BNP) with vascular markers traditionally associated with nonparoxysmal AF; the characteristics of study cohorts, ascertainment of AF history, and assay methodology (ie, sensitivity, specificity, and spectrum of proteins examined) could contribute to the inconsistency seen across studies. This is especially true given the apparent comorbidity differences between those with lone AF as a result of endurance training when compared with the general population of patients with AF. Given the wide array of clinical scenarios in which AF presents, it is possible that there are multiple underlying mechanisms with different levels of involvement dependent on the clinical context.

Our investigation adds to this body of evidence by identifying novel circulating proteins potentially associated with lone AF. Notably, we identified associations with AHSG (α₂‐Heremans–Schmid glycoprotein) and TAPBP. The latter, TAPBP, is a widely expressed 50‐kDa protein that plays an integral role in the interferon‐γ–inducible peptide‐loading complex. ³⁰ Its primary function is to facilitate efficient peptide loading onto major histocompatibility complex class I molecules. ³⁰ Although TAPBP is typically localized intracellularly, it can enhance peptide exchange on the cell surface–expressed major histocompatibility complex I molecules when overexpressed and leak into the plasma membrane. ⁵⁹ Given TAPBP's capacity to facilitate peptide exchange and the involvement of major histocompatibility complex class I molecules in mediating cytotoxic T‐cell functionality, it would be intriguing to investigate the extent of immune activation in lone AF. On the other hand, AHSG is a glycoprotein synthesized by hepatocytes and adipocytes, found in both serum and bones. ⁶⁰ , ⁶¹ Functionally, AHSG has been shown to hinder the binding of transforming growth factor‐β 1 to cell surface receptors associated with transforming growth factor‐β/bone morphogenetic protein cytokines. ⁶² Our study revealed a significantly lower expression of circulating AHSG in the lone AF subgroups, which could theoretically produce an exaggerated transforming growth factor‐β response ⁶³ ; however, AHSG is an upstream regulator among a plethora of other regulatory proteins. To validate the associations between TAPBP, AHSG, and AF incidence, as well as establish causality, prospective investigations employing direct immunoassays and larger cohorts are warranted.

Limitations

The results of this study should be viewed in the context of its design, sample size, and the sociogeographical background of the study participants. As this study represents a modest sampling of experienced male endurance athletes, the generalizability of the findings will be limited. In line, these findings should not be extended to adolescents, older adults, or those with chronic cardiorespiratory conditions wherein AF may be a result of different pathophysiological processes. More so, much of the research in this field, including the current study, has largely focused on men. A recent meta‐analysis examining the relationship between AF and exercise has suggested there may be a sex‐specific effect, with women performing intensive exercise having a 28% lower risk of developing AF (odds ratio, 0.72 [95% CI, 0.57–0.88]; P<0.001); however, varying definitions of “intensive” or “endurance” make cross‐comparability difficult. ⁶⁴

There are many factors contributing to the development of AF and response to exercise that we were unable to measure or account for including environmental and genetic differences. Any differences observed between AF or non‐AF cohorts could be explained by these confounding variables. More work is required to shed light on the influences of genetics, sex, ethnicity, type of exercise, activity threshold, and other variables on the risk of exercise‐associated AF. Given the merging of time points in the AF analysis, we cannot assess the temporality of the changes (ie, delineating if exercise was responsible or if these changes existed at baseline). Finally, it should be acknowledged that pathway analysis remains an exploratory approach. While hypergeometric testing remains the most widely used method in the field, it can produce false positives; additional pathway analysis using gene set enrichment analysis provided a similar assessment and is therefore presented to enhance the rigor of our study (Table S8). Regardless of these identified limitations, we contend that our findings provide important knowledge relating to both the physiologic changes occurring in endurance exercise as well as processes potentially underlying lone AF in middle‐aged male endurance athletes.

Conclusions

We have demonstrated the utility of an aptamer‐based proteomics platform for the preliminary evaluation of plasma protein dynamics induced by endurance exercise. Additionally, we have identified potential markers and pathways associated with lone AF, laying the foundation for future research endeavors in this domain. Future investigations should aim to scrutinize specific proteins in a larger cohort, incorporating additional time points such as a delayed assessment at 24 hours after exercise.

Sources of Funding

Dr Connelly holds the Keenan Chair in Research Leadership at St. Michael's Hospital and University of Toronto. Dr Fish holds the Tier 2 Canada Research Chair in Vascular Cell and Molecular Biology from the Canadian Institutes of Health Research and has research funding from the Canada Foundation for Innovation and a project grant from the Canadian Institutes of Health Research (PJT‐173489). Dr Goodman was supported by a Canadian Institutes of Health Research project grant (PJT‐180352) and the Heart & Stroke/Richard Lewar Center of Excellence in Cardiovascular Research. Dr Gustafson's salary was supported by Frederick Banting and Charles Best Canada Graduate Scholarship Doctoral Awards from the Canadian Institutes of Health Research, Ted Rogers Centre for Heart Research Education Funding, and an Ontario Graduate Scholarship. This study was made possible, in part, by a generous donation to the Toronto Sports Cardiology Group from the Heart & Stroke/Richard Lewar Center of Excellence in Cardiovascular Research.

Disclosures

None.

Supporting information

Data S1

Tables S1–S8

Figures S1–S5