1

Obesity results in progressive atrial structural and electricalremodeling: Implications for atrial fibrillation

Hany S. Abed, MBBS, B Pharm,* Chrishan S. Samuel, PhD,† Dennis H. Lau, MBBS, PhD,*

Darren J. Kelly, PhD,‡ Simon G. Royce, PhD,† Muayad Alasady, MBChB,* Rajiv Mahajan, MD,*

Pawel Kuklik, PhD,* Yuan Zhang, MD, PhD,‡ Anthony G. Brooks, PhD,* Adam J. Nelson, MBBS,*

Stephen G. Worthley, MBBS, PhD,* Walter P. Abhayaratna, MBBS, PhD,y

Jonathan M. Kalman, MBBS, PhD,k Gary A. Wittert, MBChB, MD,* Prashanthan Sanders, MBBS, PhD, FHRS*

From the *Centre for Heart Rhythm Disorders, Discipline of Medicine, University of Adelaide and Department of Cardiology,Royal Adelaide Hospital, Adelaide, Australia, yDepartment of Pharmacology, Monash University, Melbourne, Australia,zDepartment of Medicine, St. Vincent’s Hospital, University of Melbourne, Melbourne, Australia, yCollege of Medicine,Biology and Environment, Australian National University and Canberra Hospital, Canberra, Australian Capital Territory,Australia and kDepartment of Medicine, University of Melbourne and Department of Cardiology, Royal Melbourne Hospital,Melbourne, Australia.

BACKGROUND Obesity is associated with atrial fibrillation (AF);however, the mechanisms by which it induces AF are unknown.

OBJECTIVE To examine the effect of progressive weight gain onthe substrate for AF.

METHODS Thirty sheep were studied at baseline, 4 months, and 8months, following a high-calorie diet. Ten sheep were sampled ateach time point for cardiac magnetic resonance imaging andhemodynamic studies. High-density multisite biatrial epicardialmapping was used to quantify effective refractory period, conduc-tion velocity, and conduction heterogeneity index at 4 pacing cyclelengths and AF inducibility. Histology was performed for atrialfibrosis, inflammation, and intramyocardial lipidosis, and molecu-lar analysis was performed for endothelin-A and -B receptors,endothelin-1 peptide, platelet-derived growth factor, transforminggrowth factor b1, and connective tissue growth factor.

RESULTS Increasing weight was associated with increasing leftatrial volume (P ¼ .01), fibrosis (P ¼ .02), inflammatory infiltrates(P¼ .01), and lipidosis (P¼ .02). While there was no change in theeffective refractory period (P ¼ .2), there was a decrease inconduction velocity (P o .001), increase in conduction hetero-geneity index (P o .001), and increase in inducible (P ¼ .001)and spontaneous (P ¼ .001) AF. There was an increase inatrial cardiomyocyte endothelin-A and -B receptors (P ¼ .001)and endothelin-1 (P ¼ .03) with an increase in adiposity.

This article was presented in part by Dr Abed, who was awarded theRalph Reader Young Investigator Award by the Cardiac Society of Australiaand New Zealand. It was published as an abstract in Heart Rhythm2012;112:S190, Heart Rhythm 2012;93:S183, and Heart Lung and Circula-tion 2011;20:S2. Dr Abed and Dr Mahajan were supported by the AustralianPostgraduate Award from the University of Adelaide. Drs Abed and Alasadywere supported by the Earl Bakken Electrophysiology Scholarships from theUniversity of Adelaide. Dr Samuel was supported by the RD WrightFellowship jointly funded by the National Heart Foundation of Australia(NHFA) and the National Health and Medical Research Council (NHMRC)of Australia. Dr Lau was supported by an NHMRC Postdoctoral Fellowship.

1547-5271/$-see front matter Crown Copyright B 2013 Published by Elsevier Inc

In association, there was a significant increase in atrialinterstitial and cytoplasmic transforming growth factor b1(P ¼ .02) and platelet-derived growth factor (P ¼ .02) levels.

CONCLUSIONS Obesity is associated with atrial electrostructuralremodeling. With progressive obesity, there were changes inatrial size, conduction, histology, and expression of profibroticmediators. These changes were associated with spontaneous andmore persistent AF.

KEYWORDS Atrial fibrillation; Remodeling; Conduction velocity;Obesity; Fibrosis

ABBREVIATIONS AF ¼ atrial fibrillation; CHI ¼ conductionheterogeneity index; CL ¼ cycle length; CMRI ¼ cardiacmagnetic resonance imaging; CTGF ¼ connective tissue growthfactor; CV ¼ conduction velocity; ERP ¼ effective refractoryperiod; ET ¼ endothelin; ETA ¼ endothelin-A; ETB ¼ endothelin-B; H&E¼ hematoxylin and eosin; LA¼ left atrial; LAA¼ left atrialappendage; LAFW ¼ left atrial free wall; LAP¼ left atrial pressure;MAP¼ mean arterial pressure; PDGF-BB¼ platelet-derived growthfactor; RA ¼ right atrial; RAA ¼ right atrial appendage; RAFW ¼right atrial free wall; TGF-b1 ¼ transforming growth factor b1

(Heart Rhythm 2013;10:90–100) Crown Copyright I 2013Published by Elsevier Inc. on behalf of Heart Rhythm Society. Allrights reserved.

3 3Dr Alasady was supported by a Postgraduate Scholarship from the NHMRC.Dr Mahajan was supported by the Leo J Mahar Electrophysiology Scholar-ship from the University of Adelaide. Drs Kuklik, Brooks, and Sanders weresupported by the NHFA. Dr Sanders served on the advisory board of BardElectrophysiology, Biosense Webster, Medtronic, St Jude Medical, Merck, andSanofi-Aventis. He also received lecture fees or research funding from BardElectrophysiology, Biosense Webster, Medtronic, and St Jude Medical. Addressreprint requests and correspondence: Dr Prashanthan Sanders, MBBS, PhD,Department of Cardiology, Royal Adelaide Hospital, Level 5 McEwin Building,Adelaide, SA 5000, Australia. E-mail address: prash.sanders@adelaide.edu.au.

. on behalf of Heart Rhythm Society. All rights reserved.http://dx.doi.org/10.1016/j.hrthm.2012.08.043

Abed et al Obesity and Atrial Fibrillation 91

IntroductionObesity is recognized to be associated with the development ofatrial fibrillation (AF) and has been proposed as a contributor tothe expanding epidemic of this arrhythmia.1,2 Atrial structuraland electrical remodeling have been implicated in the AFsubstrate associated with many conditions predisposing to thedevelopment of this arrhythmia3–6; however, whether weightgain and obesity result in atrial remodeling is not known.Moreover, induction of this substrate along the adiposityspectrum of normal weight to obesity, and its relationship tohemodynamic disturbances, remains unknown. In this study, byusing a sheep model of progressive weight gain, we aimed tocharacterize the atrial functional, structural, and electrophysio-logical changes accompanying increasing adiposity.

MethodsAnimalsThirty-six sheep (Merino Cross Wethers) were studied inaccordance with guidelines outlined in the “Position of theAmerican Heart Association on Research Animal Use,”adopted in November 11, 1984. This study was approvedby the Animal Ethics Committees of the University ofAdelaide and SA Pathology, Adelaide, Australia.

Study protocolThirty animals underwent ad libitum feeding to induceobesity, as previously described.7 At baseline, 4 months,and 8 months, 10 of the cohort were randomly selected forcardiac magnetic resonance imaging (CMRI) followed byopen chest electrophysiology study. An additional 6 sheepwere studied (3 at each of the 2 time points: 4 months and 8months) as controls.

Ad libitum feeding obese ovine modelA previously characterized model of progressive weightgain, using an ad libitum regimen of hay and high-energypellets, was used to induce progressive weight gain.7 Thismodel showed an approximate increase of 10 kg/month up to36 weeks, after which weight gain reached a plateau. In brief,

Figure 1 Study outline. CMR ¼ cardiac magneti

at baseline, 30 healthy animals were commenced on a high-calorie diet of unlimited supply of high-energy soybean oil(2.2%), molasses, fortified grain, and maintenance hay, withweekly weight measurement. Excess voluntary intake waspredominantly of grass alfalfa silage and hay. For the obesesheep, pellets were gradually introduced at 8% excess basalenergy requirements and rationed to Z70% of the total drymatter intake. Blood samples were periodically collected toensure electrolyte, glucose, and acid-base homeostasis. Tomaintain the 6 controls at their baseline weight, hay wasdistributed for maintenance while high-energy pellets wererationed at 0.75% of live weight daily to maintain weighttightly between 50 and 60 kg. Nutritional content of food andhousing conditions were identical between both groups, butonly the amount was varied. Shorn weight was recordedimmediately prior to surgery. Study outline is illustrated inFigure 1.

Cardiac magnetic resonance imagingChamber volumes were measured by using CMRI (SiemensSonata 1.5 Telsa, MR Imaging Systems, Siemens MedicalSolutions, Erlangen, Germany) with 6-mm slices through theatria and 10-mm slices through the ventricles withoutinterslice gaps. Animals were securely placed in the dorsalrecumbent position for scanning. Mechanical ventilationwas maintained, facilitating electrocardiogram-gated imageacquisition with periodic breath holding. Analyses wereperformed offline by blinded operators by using the proprie-tary software QMass MR (Medis medical imaging systems,Leiden, The Netherlands). Chamber size, ventricular mass,and pericardial fat volumes were measured by using pre-viously described methods.8

Animal anesthesiaIntravenous sodium thiopentone (15–20 mg/kg) was used forinduction before endotracheal intubation. Isoflurane in oxy-gen (2%–4%) was used for maintenance. Invasive arterialblood pressure, heart rate, pulse oximetry, end-tidal CO2, andtemperature were monitored continuously.

c resonance; EPS ¼ electrophysiology study.

92 Heart Rhythm, Vol 10, No 1, January 2013

Hemodynamic recordingsContinuous invasive mean arterial pressure (MAP) monitor-ing was undertaken at the time of electrophysiology study. Inaddition, direct left atrial (LA) catheterization was performedto measure left atrial pressure (LAP).

Electrophysiology studyElectrophysiology study was performed under generalanesthesia. Midline sternotomy was used to facilitate peri-cardial cradle formation and epicardial application ofcustom-designed 128-electrode plaques with 5-mm spacing,spanning the appendage and free wall on each atrialchamber, as previously described.5 Plaques were connectedto a computerized signal digital analyzer (LabSystem Pro,Bard Electrophysiology, Lowell, MA). Surface electrocar-diogram and epicardial electrograms were recorded foroffline analysis. All electrograms were filtered between 30and 500 Hz and measured with computerized calipers at asweep speed of 200 mm/s.

Atrial effective refractory period. Atrial effective refrac-tory period (ERP) was measured at twice the tissue capturethreshold at a cycle length (CL) (S1) of 500, 400, 300, and200 ms from 4 sites: right atrial appendage (RAA), rightatrial free wall (RAFW), left atrial appendage (LAA), andleft atrial free wall (LAFW). Eight basic (S1) stimuli werefollowed by a premature (S2) stimulus in 10-ms decrements.Atrial ERP was defined as the longest S1-S2 interval notresulting in a propagated response. Each measurement wasrepeated 3 times. If there was greater than 10 ms variability,2 further measurements were taken and the total averaged.

Atrial conduction velocity. Conduction velocity (CV) wascalculated from each site during stable capture of S1 pacingtrain and the shortest coupled S2 that captures the atria ateach CL. Activation maps were created by using semiauto-mated custom software, as previously described.5 Eachannotation was manually verified with local activationtiming annotated to the peak of the largest amplitudedeflection on bipolar electrograms. Local CV was calculatedfrom the local vectors within each triangle of electrodes.Mean CV was then derived for each map.

Atrial conduction heterogeneity. Conduction heterogene-ity was assessed by using established phase-mapping tech-niques during S1 pacing.5,6 In brief, the largest activationtime difference between every 4 adjacent electrodes was firstdetermined and divided by interelectrode distances. Thelargest value at each site was then used to create a phase map,with values also displayed as histograms. Absolute conduc-tion phase delay was calculated as the difference between the5th and 95th percentile of the phase difference distribution.Conduction heterogeneity index (CHI) is then determined bydividing the absolute phase delay by the median (P50).

AF induction. Spontaneous and induced AF episodes weredocumented during continuous recording for the duration ofthe study with off-line verification. Spontaneous AF wasdefined as episodes of irregular atrial rhythm lasting Z5seconds occurring during the electrophysiology study in theabsence of pacing after stable plaque application. Inducibility ofAF was determined by ramp burst pacing commencing at CL250 ms with progressive 5-ms decrements until AF was inducedor there was no longer 1:1 capture of the atria (to atrialrefractoriness). This maneuver was repeated 5 times from eachof the 4 per-determined pacing sites and undertaken after thecompletion of electrophysiologic evaluation. When AF wasinduced and persisted for Z5 seconds, the episode and itsduration was recorded. Electrical cardioversion was performedonly when hemodynamic compromise occurred or if AFbecame sustained (Z5 minutes). If cardioversion was requiredor arrhythmia sustained, no further electrophysiologic record-ings were made.

Structural analysisAt study completion, the heart was removed. From eachanimal, 6 sections were harvested from the LA, LAA, RA,and RAA, and fixed in 10% buffered formalin or frozenat �70oC.

Quantitative analysis of wax-embedded specimens forpercentage collagen was performed by using Picrosirius redstaining. Viable sections from each atrium of every animalwere digitally captured (20 sections per animal) with an areaof Picrosirius red selected for its color range and theproportional area of tissue with this range of color quantified.

For lipid content analysis, 4 sections of the frozen myocar-dium from each animal were air dried and formalin fixed aftercareful epicardial fat stripping. Sections 5–7 mm thick wereprepared with Oil red O 0.3% w/v 2-propanol and distilled waterfor 15 minutes and then washed with 60% 2-propanol. Fiverandom non-overlapping fields were captured and digitized byusing a Carl Zeiss microscope attached to the AxioCam MRc5digital camera (Carl Zeiss, North Ryde, Australia) at �200magnification. An area of red (lipid) was selected for its colorrange, and the proportional area of tissue with this range of colorwas then digitally quantified and expressed as proportion perarea. Cellular infiltrates were visualized by using hematoxylinand eosin (H&E) staining and graded by using a previouslydescribed scale of 0–2.9 All analyses were performed blinded.

Profibrotic molecular markers

Western blotting. Western blotting was used to assesschanges in endothelin (ET) receptor expression in the atrialmyocardium between weight groups. Total protein wasextracted from the atria as described before and quantifiedby the Bradford protein assay. To determine changes in ETreceptor expression between groups, protein extracts (con-taining an equal amount of 10–15 mg of total protein perlane) were electrophoresed under reducing conditions on10.5% acrylamide gels. Western blot analyses were

Abed et al Obesity and Atrial Fibrillation 93

performed with polyclonal antibodies to the endothelin-A(ETA) receptor (ab30536; 1:500 dilution; Abcam, Cam-bridge, MA) and endothelin-B (ETB) receptor (AF4496;1:300 dilution; R&D Systems, Minneapolis, MN), in addi-tion to a goat anti-sheep secondary antibody. A monoclonalantibody to the housekeeping protein a-tubulin (1:8000dilution; Millipore Corp., Billerica, MA) was used todemonstrate equivalent loading of protein samples. Densi-tometry of ETA (63 kDa) and ETB (�30 kDa) receptor bandswas performed by using Bio-Rad GS710 Calibrated ImagingDensitometer and Quantity One software (Bio-Rad Labora-tories, Hercules, CA). The density of each receptor was thencorrected for the corresponding a-tubulin and expressed asan absolute optical density (arbitrary units).

Immunohistochemistry. The profibrotic and proliferativeeffect of ET signaling is mediated by a variety of putativemolecules. In order to determine which factors contributed tothe obesity-induced atrial fibrosis observed, the expressionand distribution of transforming growth factor b1 (TGF-b1),connective tissue growth factor (CTGF), platelet-derivedgrowth factor (PDGF-BB), endothelin-1 (ET-1), ETA, andETB in the 3 groups were determined by immunohistochem-istry. Serial sections from paraffin-embedded atrial tissueswere stained by using polyclonal antibodies to (1) TGF-b1(sc-146; 1:200 dilution; Santa Cruz Biotechnology; SantaCruz, CA), (2) CTGF (ab6992; 1:400 dilution; Abcam;Cambridge, MA), (3) PDGF-BB (ab21234; 1:50 dilution;Abcam), (4) ETA (ab30536; 1:100 dilution; Abcam), and (5)ETB (ab50658; 1:100 dilution; Abcam), as well as (6) amonoclonal antibody to ET-1 (E166; 1:1000 dilution; SigmaAldrich, St Louis, MO). Nonspecific protein binding andendogenous peroxidase were blocked by incubating sectionswith antibody diluent-containing protein (DAKO, Carpin-taria, CA) and 3% H2O2, respectively, while antigen retrievalwas performed by heating sections in citrate buffer (pH 6.0).All primary antibodies were incubated overnight, with theexception of TGF-b1, which was incubated for 4 hours.Polyclonal primary antibodies to TGF-b1, CTGF, and PDGFwere detected with the EnVision anti-rabbit kit (DAKO);the monoclonal antibody to ET-1 was detected with theEnVision antimouse (DAKO); and polyclonal antibodies toETA and ETB were detected with an anti-sheep secondary IgGantibody (DAKO) absorbed with normal sheep serum. Sites ofbound antibody were identified by using 3,3-diaminobenzi-dine (DAKO) and sections counterstained with hematoxylin.Negative controls consisted of omission of the primaryantibody. Morphometric analysis was performed by usingImageJ 1.3 software (National Institutes of Health, Bethesda,MD): from 5 to 8 random fields from each section analyzedper sample and per group. In each case, the percentagestaining of each marker analyzed per field was derived.

Statistical analysisContinuous variables are expressed as mean � SD. Hemo-dynamic variables, histologic indices, and AF episode

number/duration are expressed as median and interquartilerange with differences determined by using the Kruskal-Wallis test. The Mann-WhitneyU test was used for pairwisecomparisons. One-way analysis of variance was used todetermine differences in CMRI measures across weightcohorts. Pearson’s correlation coefficient was used to deter-mine the relationship between weight and AF episodes. Thelatter was log transformed owing to its non-normal distribu-tion. To investigate the effect of progressive obesity onelectrophysiological parameters, a linear mixed-effects modelwas used, with each electrophysiological parameter (ERP,CV, and CHI) entered as the dependent variable to determinethe effects of weight cohort for both S1- and S2-derivedrecordings. Fixed factors of weight cohort, CL, and site(appendage and free wall of each corresponding atrium) wereentered for main effects. Animal ID was entered as a randomfactor to account for nested data within each experiment. TheBonferroni method was used for pairwise comparisons ofcontinuous variables. Hemodynamic variables (MAP andLAP) were entered as covariates. To determine any possibletime-related changes in electrophysiology, baseline, 4-month,and 8-month time-point controls were similarly entered into alinear mixed-effects model. Variables were log transformedto satisfy model assumptions. Statistical significance wasestablished at P o .05. All analyses were performed by usingSPSS version 18 (SPSS, IBM corporation, Armonk, NY).

ResultsThere was progressive weight gain with feeding duration from58 � 7 kg at baseline to 77 � 5 kg (“overweight”) at 4months and 105 � 13 kg (“obese”) at 8 months (P o .001).There was no weight change in the control group: 58 � 6 kgat baseline, 50 � 4 kg at 4 months, and 54 � 5 kg at 8 months(P ¼ .2). Electrolyte, acid-base, and glucose levels remainedwithin normal range throughout the overfeeding process.

Functional and structural changesTwenty-two CMRI scans were available for analysis. Theremaining 8 were excluded owing to artifact. Figure 2 showsa significant progressive increase in LA (P ¼ .01) and RA(P ¼ .04) volumes with increasing weight gain and, likewise,a progressive increase in biventricular myocardial mass(P o .001) and pericardial fat volume (P o .001). There wereno significant differences in ventricular volumes or function. Allparameters in the control group remained stable. Interclasscorrelation coefficient between 2 independent observers for LAand RA volumes were 0.87 and 0.92, respectively. For otherCMRI measures, intraobserver and interobserver coefficients ofvariation were 3.5% and 4.9%, respectively. There was a gradedincrease in MAP and LAP (P ¼ .02 and P o .001, respec-tively), with increasing weight gain.

Atrial electrophysiology

Effective refractory period. With increasing adiposity,there were no significant differences in ERP between groups

Figure 2 Cardiac structural and hemodynamic changes with increasing weight. Representative CMRI cines from each cohort are shown. P values werederived from 1-way analysis of variance. Pairwise comparisons were performed by using the Bonferroni method. A: Horizontal long axis. B: Short axis. CMRIdata are presented as mean � SD, and hemodynamic data are presented as median and interquartile range. CMR ¼ cardiac magnetic resonance; LA ¼ leftatrium; LAP ¼ left atrial pressure; LVEDV ¼ left ventricular end-diastolic volume; MAP ¼ mean arterial pressure; RA ¼ right atrium; RVEDV ¼ rightventricular end-diastolic volume.

94 Heart Rhythm, Vol 10, No 1, January 2013

at any site (LAA: P ¼ .2; LAFW: P ¼ .8; RAA: P ¼ .08;RAFW: P ¼ .2). Likewise, there was a nonsignificantincrease in ERP between baseline, 4-month, and 8-monthcontrols (LAA—baseline: 195 � 10 ms; overweight:216 � 18 ms; obese: 217 � 18; P ¼ .4).

Atrial conduction. Figure 3A contains the changes in CVwith increasing weight. Figure 4A shows the progressiveconduction slowing demonstrated by isochronal crowdingand delayed activation. There was progressive conductionslowing with increasing weight, which is most profound inthe obese group (LAA: P o .001; LAFW: P ¼ .001; RAA:P ¼ .001; RAFW: P ¼ .001). Adjusting for hemodynamicchanges, conduction slowing remained significant (LAA:P ¼ .01; LAFW: P ¼ .01; RAA: P ¼ .01; RAFW: P ¼ .03).While baseline and overweight animals demonstrated nosignificant differences in CVs between CLs, obese animalsdemonstrated a marked CL dependence of CV on signifi-cantly greater slowing at shorter CLs; conduction slowingwas significantly greater at CL 200 ms than at CL 500 msonly in the obese cohort (RAFW: P ¼ .03; RAA: P ¼ .004;LAFW: P ¼ .04; LAA: P ¼ ns). Extending these observa-tions, Figure 4B demonstrates the differences in CV atbaseline, overweight, and obese animals in terms of S1 andS2. As expected, the S2 coupling interval demonstratedslower conduction. With increasing obesity, the extent of theconduction slowing with S2 compared with that with S1 wasgreater (P o .001). These features suggest a functionalcomponent to conduction slowing as a result of obesity

and illustrate the impact of premature beats on the underlyingsubstrate.

Associated with atrial conduction slowing was an increasein regional conduction heterogeneity, with increasing adip-osity (Figure 3B). There was an increase in CHI withprogressive weight gain. At slower pacing (CL 500 ms),while CHI was progressive with increasing weight (LAA:P ¼ .008; LAFW: P ¼ .001; RAA: P ¼ .001; RAFW:P o .001) after adjustment for hemodynamic variables(MAP and LAP), at some sites this relationship betweenobesity and CHI was weakened (LAA: P ¼ .3; LAFW:P ¼ .2; RAA: P ¼ .1; RAFW: P ¼ .003). However, withfaster pacing (CL 200 ms), there was a global increase inheterogeneous conduction (P o .05 for all sites); this effectlargely persisted despite statistical adjustment for hemody-namic variables (LAA: P ¼ .2; LAFW: P ¼ .03; RAA:P ¼ .02; RAFW: P ¼ .001).

In the control cohort, there were no significant changes(either site or CL) in CV (LAA—baseline: 1.05 � 0.02 m/s;4 months: 0.99 � 0.04 m/s; 8 months: 1.04 � 0.04 m/s;P¼ .5) and CHI (LAA—baseline: 1.17 � 0.04 m/s; 4 months:1.17 � 0.07 m/s; 8 months: 1.19 � 0.07 m/s; P ¼ .96).

AF burden. Figure 5 shows the AF burden, quantified bynumber of episodes (spontaneous and induced) and durationof arrhythmia. There was a significant association betweenweight and an increase in spontaneous episodes (P o .001),inducible episodes (P ¼ .001), and cumulative duration ofAF (P ¼ .01).

Figure 3 A: Effect of progressive obesity on regional slowing in conduction velocity at the 4 pacing sites. This relationship persisted following adjustment forhemodynamic variables. B: Changes in biatrial conduction heterogeneity index with increasing adiposity. CLs of 500 and 200 ms are presented. CL ¼ cyclelength; LAA ¼ left atrial appendage; LAFW ¼ left atrial free wall; RAA ¼ right atrial appendage; RAFW ¼ right atrial free wall.

Abed et al Obesity and Atrial Fibrillation 95

Fibrosis, cellular infiltrates, and lipidosisFigure 6 demonstrates representative examples of atrialhistology by using Picrosirius, H&E, and Oil red O stainingfor animals at baseline, overweight, and obese. Atrial tissuedemonstrated distorted myocyte arrangement and wideningof the interstitium with obesity. Quantitative histologyshowed increased perivascular collagen deposition withincreasing adiposity, greatest in the obese group (LA: P ¼.02; RA: P¼ .01). Myocardial lipidosis occurred early in theoverweight group and increased progressively (LA: P ¼ .02;RA: P ¼ .07). On H&E staining, with increasing weightthere was an increase in interstitial inflammatory cellularinfiltrates in atrial tissue (LA: P ¼ .01; RA: P ¼ .05).

Profibrotic markersOn Western blot, there was an increase in ETA (P ¼ .001)and ETB (P ¼ .001) receptor levels in atrial tissue(Figure 7A) in the overweight and obese groups relative tothe baseline group, which for both receptor subtypes peakedin the overweight group. Immunostaining, at the cardiomyo-cyte plasma membrane for both receptor protein subtypes,was strongest in the obese and overweight groups relative to

the baseline group (P ¼ .001 for ETA and ETB; Figure 7B).Staining for cytoplasmic ET-1 ligand increased modestly inthe obese group relative to the baseline group (P ¼ .03).There was a significant correlation between increasingweight and atrial interstitial and cytoplasmic PDGF-BB(Figure 8); the highest expression was seen in the obesecohort (P ¼ .02). There was a significant increase incytoplasmic CTGF with early weight gain (P ¼ .03),followed by a nonsignificant reduction in the obese cohort.There was a significant correlation between weight and atrialinterstitial and cytoplasmic TGF-b1 (P ¼ .02).

DiscussionMajor findingsProgressive weight gain resulted in atrial functional, struc-tural, and electrophysiological remodeling characterized bythe following:

1.

Increased atrial volumes, LA and systemic pressures,ventricular mass, and pericardial fat volumes.

2.

Increased atrial interstitial fibrosis, inflammation, andmyocardial lipidosis.

Figure 4 A: Activation isochronal maps of LAA pacing at 400 ms. Demonstrated are representative examples of S1 (top panel) and S2 (middle panel) for ananimal at baseline (left column), overweight (middle column), and obese (right column) cohorts. B: Pacing train (S1) and premature extra stimulus (S2) impact onCV decrement with increasing weight. P values refer to group � condition (S1, S2) interaction. CL¼ cycle length; CV¼ conduction velocity; LAA¼ left atrialappendage.

Figure 5 A: Scatter plot of spontaneous and induced AF episodes (log transformed) for each animal, and bar graph of AF duration (minutes) with weightcohort (median and interquartile range; P¼ .001). B: Example of spontaneous AF observed from the LAA. AF¼ atrial fibrillation; LAA¼ left atrial appendage.

96 Heart Rhythm, Vol 10, No 1, January 2013

Figure 6 Histology from atrial tissue (�200 magnifications). From top panel: Picrosirius red staining, inflammatory infiltrates on hematoxylin and eosin, andmyocardial lipidosis on Oil red O. Values are presented as median and interquartile range. All pairwise comparisons were performed by using the Mann-WhitneyU method. LA ¼ left atrium; RA ¼ right atrium.

Abed et al Obesity and Atrial Fibrillation 97

3.

Progressive conduction abnormalities with slowing ofatrial conduction and increased conduction heterogeneity,which was amplified at shorter coupling intervals and CLswith greater adiposity. The significance of these abnorm-alities with progressive adiposity persisted after adjustingfor potential hemodynamic variables. There was no changein tissue ERP.

4.

Overexpression of atrial cardiomyocyte ETA/B receptorsand a weak association with cytoplasmic ET-1 levels.

5.

Increase in interstitial and cytoplasmic TGF-b1, PDGF-BB, and CTGF levels.

Perhaps as a consequence of these abnormalities, pro-

gressive weight gain was associated with a greater burden ofinduced and spontaneous AF.

Atrial remodeling and the substrate for AFLi et al3 in a rapid ventricular pacing model of heart failureprovided the seminal observations of the importance ofstructural remodeling with atrial fibrosis associated withheterogeneity of conduction to the substrate predisposing toAF compared with that due to arrhythmia itself. Thesedominant components of the AF substrate have since beenconsistently demonstrated in other models of nonischemiccardiomyopathy,6 mitral regurgitation,10 hypertension,5 andmyocardial infarction.11 Recent work has also demonstrated

that obesity promotes diastolic dysfunction and that togetherwith acutely obstructed respiration, to simulate sleep apnea,promotes LA dilation resulting in AF susceptibility. In addition,prevention of LA distension reduced AF inducibility, suggest-ing an augmenting relationship between the 2 conditions inpromoting AF.12 Acute atrial dilatation or stretch in itself hasbeen demonstrated to result in direction-dependent conductionblock in part proposed to reflect an amplification of theanisotropic properties of the atrial myocardium.13

Clinical mapping studies have also observed areas ofelectrical silence, low voltage, fractionated electrograms, andaltered and circuitous conduction associated with a predis-position for AF in heart failure,14 sinus node disease,15

aging, atrial septal defects,16 rheumatic heart disease,17 sleepapnea,18 and hypertension.19 Stiles et al20 have demonstratedsuch findings in patients with lone AF when studied remotefrom episodes of AF. Indeed, the importance of such atrialstructural remodeling has been highlighted not only in thedevelopment of arrhythmia but also in the determination ofablation outcomes21 and has recently been implicated in theprogression of the atrial disease after ablation.22

While population studies have implicated obesity as animportant contributor to the increasing epidemic of AF, thecurrent study provides evidence for the progressive evolutionof remodeling that has been extensively characterized toform the substrate for AF, with increasing weight gain and

Figure 7 A: Western blots of ETA receptor (63 kDa) and ETB receptor (�30 kDa) expression and a-tubulin (55 kDa) demonstrates equivalent loading ofsamples. Two bands are shown per group. B: Top panel shows IHC staining of ETA receptors and middle panel shows IHC staining of ETB receptors on the atrialcardiomyocyte membrane. Lower panel shows IHC staining of the ET-1 ligand in the cardiomyocyte cytoplasm. ET-1¼ endothelin-1; ETA¼ endothelin-A; ETB

¼ endothelin-B; IHC ¼ immunohistochemistry.

98 Heart Rhythm, Vol 10, No 1, January 2013

obesity. The lack of direct evidence in humans is attributableto the confounding effects of the existence of comorbidconditions that cluster in obese individuals and are knownto contribute to atrial remodeling.23 Obese individuals aremore likely to suffer concomitant hypertension, diabetesmellitus, obstructive sleep apnea, coronary artery disease,and heart failure—all associated with AF development.The current study observed marked weight-dependentstructural and electrophysiological changes translating toa greater AF burden. In the absence of the numerousclinically present comorbidities and the persistence of the

abnormalities after controlling for hemodynamic changes,these findings argue in favor of a significant directcontribution of obesity to the AF substrate. In addition,the presence of a greater degree of atrial lipidosis mayrepresent a disease-specific component of the structuralremodeling predisposing to AF.

Atrial tissue fibrosisTissue fibrosis is the critical structural element leading toabnormal conduction and anisotropy. Previous studies have

Figure 8 Immunohistochemical staining of TGF-b1 (top panels), PDGF-BB (middle panels), and CTGF (lowest panels) in atrial tissue and quantitativemorphometry. CTGF ¼ connective tissue growth factor; PDGF-BB ¼ platelet-derived growth factor; TGF-b1 ¼ transforming growth factor b1.

Abed et al Obesity and Atrial Fibrillation 99

focused on the central role of TGF-b1 in promoting fibrosisand AF.24 PDGF-BB has been shown to modulate myofi-broblast persistence and promote local tissue microvasculo-pathy directly and indirectly through TGF-b1, via a feedbackmechanism. Specifically, PDGF signaling may be particu-larly important in the profibrotic response of atrial fibro-blasts.25 Our observation of elevated atrial and systemicpressures with increasing adiposity and the concurrentstepwise elevation in PDGF levels supports the findingsof Iwasaki et al12 in highlighting the role of atrial stretch inpromoting AF in the structurally remodeled atrium.Importantly, while the above study evaluated the impactof acute stretch in a genetically predisposed obesitymodel, the current study evaluates the progressive andchronic evolution of obesity owing to dietary increase incaloric intake. Whether these models share similarmechanisms to result in atrial remodeling requires furtherinvestigation.

ET receptors are expressed in greater abundance on theatrial myocardium than on the ventricular myocardiumand have been implicated as modulators of the load/stretch-induced hypertrophic response.26 Recent studieshave demonstrated increased activity of the ET system inoverweight and obese humans.27 Elevated ET-1 has beenshown to predict AF recurrence following an indexepisode and recurrence following pulmonary vein

isolation.28,29 In this study, we have observed increasedexpression of atrial profibrotic markers, particularlycardiomyocyte ET receptors in the overweight state,persisting at a somewhat lower level with furtherweight gain.

Study limitationsIn this model, independent samples were studied at 3distinct points in time. This may introduce confoundingfactors of intra- and intercohort variability. Nevertheless,our model structure was congruent with a previous modelof progressive weight gain via an ad libitum feedingregimen.7

Finally, clinical AF is recognized to result from thecomplex interaction between triggers, perpetuators, andsubstrates.30 This study focused on the atrial substrate anddid not evaluate other contribution to the development of AF.

ConclusionsProgressive obesity predisposes to a greater burden of AF byforming an electropathological substrate. This is dispropor-tionate to the progressive hemodynamic impact of obesityand suggests a direct pathogenic role of obesity on the AFsubstrate.

100 Heart Rhythm, Vol 10, No 1, January 2013

AcknowledgmentsThe authors thank Ms Samar Babkair and Mr Krupesh Patelfor their assistance with immunohistochemistry and morpho-metric analysis, respectively.

References1. Miyasaka Y, Barnes ME, Gersh BJ, et al. Secular trends in incidence of atrial

fibrillation in Olmsted County, Minnesota, 1980 to 2000, and implications on theprojections for future prevalence. Circulation 2006;114:119–125.

2. Tedrow UB, Conen D, Ridker PM, et al. The long- and short-term impact ofelevated body mass index on the risk of new atrial fibrillation the WHS (women’shealth study). J Am Coll Cardiol 2010;55:2319–2327.

3. Li D, Fareh S, Leung TK, Nattel S. Promotion of atrial fibrillation by heart failurein dogs: atrial remodeling of a different sort. Circulation 1999;100:87–95.

4. Verheule S, Wilson E, Tt Everett, Shanbhag S, Golden C, Olgin J. Alterations inatrial electrophysiology and tissue structure in a canine model of chronic atrialdilatation due to mitral regurgitation. Circulation 2003;107:2615–2622.

5. Lau DH, Mackenzie L, Kelly DJ, et al. Hypertension and atrial fibrillation:evidence of progressive atrial remodeling with electrostructural correlate in aconscious chronically instrumented ovine model. Heart Rhythm 2010;7:1282–1290.

6. Lau DH, Psaltis PJ, Mackenzie L, et al. Atrial remodeling in an ovine model ofanthracycline-induced nonischemic cardiomyopathy: remodeling of the samesort. J Cardiovasc Electrophysiol 2011;22:175–182.

7. McCann JP, Bergman EN, Beermann DH. Dynamic and static phases of severedietary obesity in sheep: food intakes, endocrinology and carcass and organchemical composition. J Nutr 1992;122:496–505.

8. Wong CX, Abed HS, Molaee P, et al. Pericardial fat is associated with atrialfibrillation severity and ablation outcome. J Am Coll Cardiol 2011;57:1745–1751.

9. Geboes K, Riddell R, Ost A, Jensfelt B, Persson T, Lofberg R. A reproduciblegrading scale for histological assessment of inflammation in ulcerative colitis.Gut 2000;47:404–409.

10. Everett TH, IV, Wilson EE, Verheule S, Guerra JM, Foreman S, Olgin JE.Structural atrial remodeling alters the substrate and spatiotemporal organizationof atrial fibrillation: a comparison in canine models of structural and electricalatrial remodeling. Am J Physiol Heart Circ Physiol 2006;291:H2911–H2923.

11. Alasady M, Abhayaratna WP, Leong DP, et al. Coronary artery disease affectingthe atrial branches is an independent determinant of atrial fibrillation aftermyocardial infarction. Heart Rhythm 2011;8:955–960.

12. Iwasaki YK, Shi Y, Benito B, et al. Determinants of atrial fibrillation in an animalmodel of obesity and acute obstructive sleep apnea. Heart Rhythm 2012;9:1409–1416.

13. Eijsbouts SC, Majidi M, van Zandvoort M, Allessie MA. Effects of acute atrialdilation on heterogeneity in conduction in the isolated rabbit heart. J CardiovascElectrophysiol 2003;14:269–278.

14. Sanders P, Morton JB, Davidson NC, et al. Electrical remodeling of the atria incongestive heart failure: electrophysiological and electroanatomic mapping inhumans. Circulation 2003;108:1461–1468.

15. Sanders P, Morton JB, Kistler PM, et al. Electrophysiological and electroana-tomic characterization of the atria in sinus node disease: evidence of diffuse atrialremodeling. Circulation 2004;109:1514–1522.

16. Roberts-Thomson KC, John B, Worthley SG, et al. Left atrial remodeling inpatients with atrial septal defects. Heart Rhythm 2009;6:1000–1006.

17. John B, Stiles MK, Kuklik P, et al. Electrical remodelling of the left and right atriadue to rheumatic mitral stenosis. Eur Heart J 2008;29:2234–2243.

18. Dimitri H, Ng M, Brooks AG, et al. Atrial remodeling in obstructive sleep apnea:implications for atrial fibrillation. Heart Rhythm 2012;9:321–327.

19. Medi C, Kalman JM, Spence SJ, et al. Atrial electrical and structuralchanges associated with longstanding hypertension in humans: implications forthe substrate for atrial fibrillation. J Cardiovasc Electrophysiol 2011;22:1317–1324.

20. Stiles MK, John B, Wong CX, et al. Paroxysmal lone atrial fibrillation isassociated with an abnormal atrial substrate: characterizing the ’’second factor.’’J Am Coll Cardiol 2009;53:1182–1191.

21. Verma A, Wazni OM, Marrouche NF, et al. Pre-existent left atrial scarring inpatients undergoing pulmonary vein antrum isolation: an independent predictorof procedural failure. J Am Coll Cardiol 2005;45:285–292.

22. Teh AW, Kistler PM, Lee G, et al. Long-term effects of catheter ablation for loneatrial fibrillation: progressive atrial electroanatomic substrate remodeling despitesuccessful ablation. Heart Rhythm 2012;9:473–480.

23. Munger TM, Dong YX, Masaki M, et al. Electrophysiological and hemodynamiccharacteristics associated with obesity in patients with atrial fibrillation. J AmColl Cardiol 2012.

24. Verheule S, Sato T, Tt Everett, et al. Increased vulnerability to atrial fibrillation intransgenic mice with selective atrial fibrosis caused by overexpression of TGF-beta1. Circ Res 2004;94:1458–1465.

25. Burstein B, Libby E, Calderone A, Nattel S. Differential behaviors of atrial versusventricular fibroblasts: a potential role for platelet-derived growth factor in atrial-ventricular remodeling differences. Circulation 2008;117:1630–1641.

26. Kerkela R, Ilves M, Pikkarainen S, et al. Key roles of endothelin-1 and p38MAPK in the regulation of atrial stretch response. Am J Physiol Regul IntegrComp Physiol 2011;300:R140–R149.

27. Weil BR, Westby CM. Van Guilder GP, Greiner JJ, Stauffer BL, DeSouza CA.Enhanced endothelin-1 system activity with overweight and obesity. Am JPhysiol Heart Circ Physiol 2011;301:H689–H695.

28. Nakazawa Y, Ashihara T, Tsutamoto T, Ito M, Horie M. Endothelin-1 as apredictor of atrial fibrillation recurrence after pulmonary vein isolation. HeartRhythm 2009;6:725–730.

29. Mayyas F, Niebauer M, Zurick A, et al. Association of left atrial endothelin-1with atrial rhythm, size, and fibrosis in patients with structural heart disease. CircArrhythm Electrophysiol 2010;3:369–379.

30. Allessie MA, Boyden PA, Camm AJ, et al. Pathophysiology and prevention ofatrial fibrillation. Circulation 2001;103:769–777.