Intracardiac expression of markers of endothelial damage/dysfunction, inflammation, thrombosis, and tissue remodeling, and the development of postoperative atrial fibrillation

Authors


Gregory Y. H. Lip, University of Birmingham Centre for Cardiovascular Sciences, Department of Medicine, City Hospital, Birmingham B18 7QH, UK.
Tel.: +44 121 5075080; fax: +44 121 5544083.
E-mail: g.y.h.lip@bham.ac.uk

Abstract

Summary. Background: Atrial fibrillation (AF) is a common complication of coronary artery bypass grafting (CABG), and may have an inflammatory and/or thrombotic etiology. We sought to determine the expression of inflammatory (interleukin [IL]-6), thrombotic (tissue factor and von Willebrand factor [VWF]) and remodeling (matrix metalloproteinase [MMP]-9 and tissue inhibitor of metalloproteinase [TIMP]-1) markers by left atrial appendage (LAA) and right atrial appendage (RAA) tissue in the prediction of postoperative AF. We determined whether the tissue expression of markers of certain different pathophysiologic mechanisms predicted the development of AF after CABG. Methods: LAA and RAA tissue was excised during CABG in 100 patients free of AF and inflammation. Tissue marker expression was quantified by immunohistochemistry and was related to 30-day postoperative AF. Results: Overall, there were no significant differences in staining intensity of any marker between LAA tissue and RAA tissue. However, more intense expression of VWF by LAA tissue predicted the 30 patients with postoperative AF as compared with those free of AF (P = 0.006). IL-6, MMP-9 and TIMP-1 expression by RAA and LAA epicardial tissue was stronger than expression by endocardium or cardiomyocytes (all P < 0.025) but failed to predict AF. Conclusion: In this study, one of the largest to investigate tissue expression of pathophysiologic markers in relation to postoperative AF, we show that more intense expression of VWF by LAA tissue is a significant predictor of postoperative AF. This points towards a possible role of endothelial damage/dysfunction (as reflected by VWF changes) in the pathogenesis of postoperative AF.

Introduction

Atrial fibrillation (AF) is a common consequence of coronary artery bypass grafting (CABG), being present in up to one-third of patients [1–4]. Postoperative AF presents a risk of adverse events, being associated with increased rates of early and late mortality, stroke, respiratory failure, infections, and renal failure, and prolongation of hospital stay [4–9]. Many clinical variables associated with the development of postoperative AF have been assessed, including age, cardiac function, left atrial size, and other comorbidities [1]. A better understanding of pathophysiologic aspects of postoperative AF may help in the development of preventive strategies and the broadening of treatment options.

Inflammation, as defined by increased levels of the plasma biomarkers C-reactive protein (CRP) and interleukin (IL)-6, has been implicated in the pathogenesis of post-CABG AF [10–12]. An alternative pathophysiology to inflammation in cardiovascular disease is atherothrombosis, and the two may be linked as an inflammatory environment within the heart; inflammation may ‘drive’ the prothrombotic state in AF, leading to the increased risk of thrombogenesis and, subsequently, potentially fatal thromboembolism [13,14]. Indeed, perhaps the major consequence of AF is thrombotic stroke, leading to the hypothesis that AF itself is a prothrombotic disease [15]. Evidence in favor of this hypothesis includes increased levels of plasma markers of a prothrombotic state (von Willebrand factor [VWF], fibrinogen, and fibrin D-dimer) in the plasma of patients with AF [13,16]. However, a weakness of this approach is that blood taken from a peripheral vein may not necessarily reflect the situation within the heart, although there is evidence that levels of certain markers in peripheral blood do mirror intracardiac levels [17].

Structural alterations in atria in AF may be a consequence or predictor of this condition [18,19]. Plasma markers of this remodeling include matrix metalloproteinase (MMP) enzymes and their inhibitors, tissue inhibitors of metalloproteinases (TIMPs); of these, MMP-9 and TIMP-1 may be important [20–23]. However, the place of these and allied markers is unclear, especially in predicting which patients are likely to go on to develop AF after CABG. In this respect, we have recently shown that intracardiac plasma levels of IL-6, CRP, and MMP-9, and peripheral blood levels of CRP, predict those patients who go on to develop AF after CABG [24]. Nevertheless, the extent to which the expression of inflammatory, remodeling and prothrombotic molecules by the tissue of the heart is important in the development of AF is unknown.

We tested the hypothesis that the intracardiac tissue expression of the defined markers is indicative of the pathophysiologic processes, that is, that inflammation (IL-6), remodeling (MMP-9 and TIMP-1) and the prothrombotic state (tissue factor and VWF) are predictive of postoperative AF among patients undergoing CABG.

Materials and methods

We recruited 100 patients undergoing elective on-pump CABG at Vilnius University Hospital, Lithuania. Exclusion criteria were any history of AF, pacemakers, surgical treatment of AF (e.g. maze procedure), cancer, chronic inflammatory disease, significant liver failure, significant renal failure (serum creatinine concentration > 200 μm), any hematologic disease, moderate–severe valvular disease, any acute cerebrovascular event or infective/inflammatory disorder within 60 days prior to the operation.

All patients underwent medical history recording, physical examination, coronary angiography, transthoracic echocardiography, 12-lead electrocardiography (ECG) recording, and routine hematologic and biochemical blood tests prior to the surgical intervention. All patients received β-blockers preoperatively, and this was continued following cardiac surgery. The study was approved by the Lithuanian Bioethics Committee, and informed written consent was obtained from each patient prior to inclusion in the investigation.

CABG was performed with identical surgical and anesthetic protocols for each patient. Cardiopulmonary bypass was performed with ascending aortic and two-stage venous cannulation of the right atrium under systemic moderate hypothermia (32 °C). Myocardial protection was accomplished by means of antegrade and retrograde intermittent cold blood cardioplegia. The left internal thoracic artery was routinely grafted to the left anterior descending artery, whereas the saphenous vein was used for other target coronary arteries. The left atrial appendage (LAA) and right atrial appendage (RAA) were harvested with minimal trauma, and processed as described below. Both LAA and RAA samples were obtained, allowing comparison between patients (i.e. AF vs. non-AF) and within patients (that is, is there something intrinsic to the LAA rather than the RAA that predisposes to postoperative AF?). VWF in citrated plasma from a peripheral blood sample was measured by ELISA (Dako-Patts, Ely, UK) [24].

After the operation, all patients were monitored in the intensive care unit (ICU) for at least 24 h. Continuous ECG monitoring was performed only in the ICU, and was followed by routine 12-lead ECG recording every morning or, in the case of clinical symptoms or irregular pulse, at physical examination throughout the hospitalization period. All patients were discharged to our cardiac rehabilitation clinic and followed up for AF occurrence until the 30th postoperative day. Postoperative AF was defined as any electrocardiographically documented AF episode lasting for more than 30 s. Preoperative levels of proinflammatory (IL-6) and prothrombotic (tissue factor and VWF) markers in plasma and in tissues of the LAA and RAA were compared between patients with postoperative AF and those who remained in sinus rhythm (SR).

Immunohistochemistry

Tissues obtained from the LAA and RAA were fixed in buffered 10% formalin, processed overnight, and embedded in paraffin wax. Serial sections were cut at a thickness of 3 μm on charged slides, and were deparaffinized and rehydrated prior to staining. The antibodies were tested at different dilutions on control tissue. Appropriate control tissue (as recommended by the manufacturer’s data sheet) was used to test antibody staining. The control tissue and dilution that gave the desired staining without any background staining were selected. Positive and negative (omission of primary antibody) controls were included in every run. Antigen retrieval was carried out with two methods. Heat-induced antigen retrieval was performed with a microwave oven. Sections were placed in a pressure cooker containing Tris/EDTA buffer (pH 7.8). The pressure cooker was placed in a microwave and heated for different time intervals, according to the antibody. Enzyme antigen retrieval was performed with Protease 1 (Ventana Medical Systems, Tucson, AZ, USA). Table 1 shows the antibodies used and the different retrieval methods.

Table 1.   Antibodies and retrieval techniques
AntibodyManufacturerAntigen retrievalDilutionIncubation (min)Control tissue
Matrix metalloproteinase-9Novocastra, Newcastle Upon Tyne, UKTris/EDTA buffer, 15 min1 : 1032Kidney/liver
Interleukin-6Dako, Glostrup, DenmarkNone1 : 4032Appendix
Tissue factorAmerican Diagnostica, Stamford, CT, USAProtease 1, 10 min1 : 20032Pancreas
von Willebrand factorDako, Glostrup, DenmarkTris/EDTA buffer, 15 min1 : 100032Tonsil
Tissue inhibitor of metalloproteinase-1Dako, Glostrup, DenmarkTris/EDTA buffer, 12 min1 : 10032Pancreas

All immunohistochemical staining was performed on a Ventana NexES automated stainer with a standard streptavidin–biotin–peroxidase technique (iVIEW DAB detection Kits; Ventana Medical Systems), as follows. The primary antibody was applied, and then located with a biotin-conjugated secondary antibody. This step was followed by the addition of a streptavidin enzyme (streptavidin–horseradish peroxidase; R & D Systems, Abingdon, UK) conjugate that binds to the biotin present on the secondary antibody. The complex was then visualized with hydrogen peroxide substrate and 3,3′-diaminobenzidine tetrahydrochloride (DAB) chromogen. Ventana iVIEW DAB detection kits were used for all staining. The kits incorporate an inhibitor that blocks any endogenous peroxidase activity in the tissue. Note that this level of staining cannot differentiate VWF in Weibel–Palade bodies or at the cell surface. After staining, sections were lightly counterstained with hematoxylin, dehydrated, cleared, and coverslipped. Tissue sections were cut and stained together in a single batch by a single technician.

Slides were photographed together in a single run with the same illumination and exposure time, and on the same microscope (Olympus BX51 microscope fitted with an Olympus Colour View Camera [Olympus Soft Imaging Solutions, Gmbh, Southend-on-Sea, UK] and Olympus Cell^B software). The light intensity was kept constant with the ‘Pre-set’ button of the microscope at the start of the session.

Tissues were examined independently by two board-certified consultant histopathologists (U.Z. and M.M.), who both scored each slide as having no staining, weak focal staining, multifocal moderate staining, or diffuse strong staining, giving a score of 0, 1, 2, or 3, respectively. IL-6, MMP-9 and TIMP-1 were assessed on endocardium, epicardium, and cardiomyocytes. VWF and tissue factor were assessed on endocardium. Staining and scoring were performed with the investigators blinded to the AF status of the patient. The means of individual staining scores were used for analysis, and then adjusted by a third party so that pooled mean scores of 0, 0.5 and 1 were rescored as 1 (low-intensity staining), pooled mean scores of 1.5 and 2.0 were rescored as 2 (moderate-intensity staining), and pooled mean scores of 2.5 and 3 were rescored as 3 (high-intensity staining). The kappa statistic for agreement in staining between the observers was 0.58.

Statistical analysis

Following a test of statistical normality, analysis of continuously variable data with parametric and non-parametric tests was performed as appropriate. Data that are normally distributed are expressed as mean (standard deviation), and data that are non-normally distributed are expressed as median (interquartile range). Differences between patient subgroups (e.g. developing AF or not) were determined by t-test for normally distributed variables and by Mann–Whitney U-test equivalent for non-normally distributed variables (Table 1). The chi-square test was used for categorical variables, i.e. the expression of markers in Tables 2 and 3. The relationship between plasma VWF and the tissue expression of VWF in Table 5 was assessed as a continuum by Altman’s linear ordered trend method [25]. A P-value of < 0.05 was considered to be statistically significant. Statistical analyses were performed with Minitab 15 (Minitab, State College, PA, USA).

Table 2.   Patient characteristics
 Postoperative AF
No
(n = 70)
Yes
(n = 30)
P-value
  1. AF, atrial fibrillation; LA, left atrial; LV, left ventricular; MI, myocardial infarction; RA, right atrial; SD, standard deviation. < 0.05 considered to be significant.

Gender, n (%)
 Male58 (82.9)28 (93.3)0.166
 Female12 (17.1)2 (6.7)
Age (years) ± SD63.8 ± 9.265.2 ± 8.80.472
History of MI, n (%)29 (41.4)15 (50)0.429
Hypertension, n (%)59 (84.3)25 (83.3)0.905
Diabetes mellitus or impaired glucose tolerance, n (%)13 (18.6)5 (16.7)0.820
Current smoking, n (%)20 (28.6)11 (36.7)0.422
Echocardiographic data (mm) ± SD   
 Interventricular septum12.4 ± 1.812.7 ± 1.40.347
 LV end-diastolic diameter52.8 ± 5.055.0 ± 5.20.052
  LA diameter52.7 ± 4.653.7 ± 5.70.319
  RA diameter48.2 ± 4.449.3 ± 6.10.303
LV ejection fraction (%) ± SD52.2 ± 7.250.0 ± 7.70.189
Creatinine (μm) ± SD94 ± 22100 ± 210.238
Hemoglobin (g/L) ± SD144 ± 13140 ± 150.201
Leukocytes (×109/L) ± SD6.8 ± 1.67.3 ± 1.60.106
Platelets (×109/L) ± SD236 ± 59237 ± 580.973
Table 3.   Tissue expression of von Willebrand factor (VWF), tissue factor, matrix metalloproteinase (MMP)-9, tissue inhibitor of metalloproteinase (TIMP)-1 and interleukin (IL)-6 by left atrial appendage (LAA) and right atrial appendage (RAA) tissue in 100 patients undergoing coronary artery bypass grafting
 Relative expression (%) by LAA tissue (low–moderate–high)Relative expression (%) by RAA tissue (low–moderate–high)LAA/RAA P-value
  1. –, statistical analysis unsound. Data are number of tissues staining with low, moderate or strong intensity. P-values by chi-squared test. Numbers may fail to sum to 100, owing to insufficient tissue being obtained.

VWF on endocardium18–52–2722–51–260.815
Tissue factor on endocardium64–31–158–38–30.375
IL-6 on:
 Endocardium89–8–091–8–10.409
 Cardiomyocytes88–9–092–8–00.749
 Epicardium50–27–1256–26–120.895
 P-value between tissues< 0.001< 0.001 
MMP-9 on:
 Endocardium100–0–098–1–0
 Cardiomyocytes55–35–666–30–20.186
 Epicardium53–26–955-32-20.078
 P-value between tissues< 0.001< 0.001 
TIMP-1 on:
 Endocardium100–0–0100–0–0
 Cardiomyocytes96–1–0100–0-0
 Epicardium62–23–555–31–40.424
 P-value between tissues< 0.001< 0.001 

Results

One hundred consecutive patients without a prior history of AF within 6 months before the surgical intervention were followed up for the development of postoperative AF, which was documented for 30 days postoperatively. This enabled a case–control comparison of those with AF at 30 days and those who proceeded to 30 days in SR, consistent with published data on the development of post-operative AF [1–4], and justified our recruitment strategy.

Patient clinical characteristics are presented in Table 2: patients who developed AF were slightly older than those free of AF. Table 3 shows the expression of VWF, tissue factor, MMP-9, TIMP-1 and IL-6 in the group as a whole. There was no difference in the expression of the markers between LAA and RAA tissues. Expression levels of the markers by the endocardium of the RAA and the LAA were as follows: VWF > tissue factor > IL-6 > MMP-9 = TIMP (both P < 0.001). Expression of the remodeling markers by endocardium was minimal. However, IL-6, MMP-9 and TIMP-1 staining was significantly higher in both LAA and RAA epicardial tissue than in endocardial tissue (P < 0.001). Notably, MMP-9 expression by cardiomyocytes was equivalent to that by epicardial tissue.

Table 4 shows data according to the presence of AF within 30 days of CABG. More intense expression of VWF by LAA tissue was the only predictor of those patients who went on to have AF as compared with those who were free of AF. Expression in those free of AF was low in 16.7% of patients, moderate in 65.1%, and high in 18.2%, as compared with a profile of 24.2%–31.0%–44.8%, respectively, in those who developed AF (P = 0.006). Figure 1A–C shows typical results of the staining of endocardial tissue for the intensity of VWF. In Fig. 1A, the intensity of staining is low, scoring 1+. In Fig. 1B, the intensity of staining is moderate, scoring 2+. In Fig. 1C, the intensity of staining is high, scoring 3+. The strength of the staining clearly increases across these three figures. In contrast, the counterstaining is of equal intensity in each figure.

Table 4.   Tissue expression of von Willebrand factor (VWF), tissue factor, interleukin (IL)-6, matrix metalloproteinase (MMP)-9 and tissue inhibitor of metalloproteinase (TIMP)-1 by left atrial appendage (LAA) and right atrial appendage (RAA) tissue according to 30-day atrial fibrillation (AF)
 Free of AF (n = 70)With AF (n = 30) 
(1) LAA(2) RAAP-value for LAA vs. RAA (1 vs. 2)(3) LAA(4) RAAP-value for LAA vs. RAA (3 vs. 4)P-value for LAA free of AF vs. LAA with AF (1 vs. 3)P-value for RAA free of AF vs. RAA with AF (2 vs. 4)
  1. –, statistical analysis unsound. RAA: VWF and tissue factor were stained on endocardium only. P-values by chi-squared test. Data are number of tissue blocks staining low–moderate–high. Numbers may fail to sum to 70 or 30, owing to insufficient tissue being obtained.

VWF

Tissue factor
11–43–1218–35–160.6297–9–134–16–100.2070.0060.305
44–22–141–27–20.11320–9–017–11–10.5140.8260.410
IL-6 on:
 Endocardium63–5–064–4–00.73026–3–027–2–10.6400.6240.849
 Cardiomyocytes64–4–066–4–00.96624–5–026–4–00.6760.0780.198
 Epicardium37–18–740–21–50.75713–9–516–5–70.4130.5250.193
 P-value< 0.001< 0.001 < 0.0010.004   
MMP-9 on:
 Endocardium68–0–070–0–029–0–028–1–0
 Cardiomyocytes36–28–447–21–10.11920–6–319–10–10.3660.1420.567
 Epicardium36–21–538–22–20.50218–5–418–9–10.2320.2730.678
 P-value< 0.001< 0.001 0.0160.021   
TIMP-1 on:
 Endocardium68–0–069–0–029–0–030–0–0
 Cardiomyocytes67–1–069–0–029–0–030–0–0
 Epicardium42–18–335–26–20.31820–4–320–5–20.8560.2520.097
 P-value< 0.001< 0.001 < 0.001< 0.001   
Figure 1.

 Typical von Willebrand factor staining of a section of atrial appendage. (A) Low intensity (1+). (B) Moderate intensity (2+). (C) High intensity (3+). All figures are at a magnification × 40.

Expression of IL-6, MMP-9 and TIMP-1 by epicardial tissue, cardiomyocytes and endocardial tissue was approximately the same as in the combined group, and all failed to predict AF.

The relationship between plasma VWF and the continuum of increasing tissue expression of VWF was assessed by Altman’s linear ordered trend method [25], and are shown in Table 5. Mean plasma VWF increased by 13.5% with greater LAA tissue expression (P < 0.05), but only when all subjects were pooled. Although a clear ordered trend of increasing VWF was present in the LAA tissue in the subjects free of AF (rising by 11.2%), the LAA tissue in the subjects who went on to suffer AF (14.8%), and the RAA tissue in the subjects who went on to suffer AF (10.6%), these trends were not statistically significant, probably because of small subject numbers in each group.

Table 5.   Relationship between plasma von Willebrand factor and the expression of von Willebrand factor by the endocardium of the left atrial appendage (LAA) and the right atrial appendage (RAA)
 Staining intensity 1Staining intensity 2Staining intensity 3
  1. AF, atrial fibrillation; n, number of subjects in each group. *P < 0.05 for linear trend. Data presented as mean level of plasma von Willebrand factor (standard deviation).

LAA (all subjects)118 (18)
(n = 18)*
125 (27)
(n = 52)*
134 (27)
(n = 27)*
RAA (all subjects)121 (27)
(n = 22)
130 (26)
(n = 51)
127 (25)
(n = 26)
LAA (subjects free of AF116 (22)
(n = 11)
124 (27)
(n = 43)
129 (29)
(n = 14)
RAA (subjects free of AF)121 (11)
(n = 18)
129 (25)
(n = 35)
121 (21)
(n = 16)
LAA (subjects with AF)121 (11)
(n = 7)
126 (29)
(n = 9)
139 (28)
(n = 13)
RAA (subjects with AF)122 (14)
(n = 4)
132 (26)
(n = 16)
135 (30)
(n = 10)

Discussion

In this study, one of the largest to investigate tissue expression of pathophysiologic markers in relation to postoperative AF, we show, for the first time, that more intense expression of VWF by LAA tissue was a significant predictor of postoperative AF. Although sampling blood from a peripheral vein may reflect intracardiac levels [17], and sampling intracardiac blood may help predict those who will develop AF after CABG [24], the degree to which atrial tissues per se predict AF was previously unknown. Although IL-6, MMP-9 and TIMP-1 expression by RAA and LAA epicardial tissue was stronger than expression by endocardium or cardiomyocytes, these differences all failed to predict postoperative AF. These differences point towards a possible role of endothelial damage/dysfunction (as reflected by VWF changes) in the pathogenesis of postoperative AF.

There is increasing evidence for the contribution of inflammation [10–13,26,27], thrombosis [13–17] and abnormal remodeling [18–23] to the pathogenesis of AF, and although cardiac surgery and cardiopulmonary bypass surgery can themselves cause an inflammatory response, this may not necessarily relate to the development of AF per se [4–9]. There has been a greater focus on more early and comprehensive management of AF [26], with a recommendation in current management guidelines for early interventions with so-called ‘upstream therapy’ in the presence of cardiovascular risk factors for the development of AF [27] that could potentially influence these pathogenic processes (inflammation, remodeling, etc).

Several histologic studies have tested the hypothesis that some structural changes to extracellular matrix metalloproteins or inflammation in cardiac chambers and tissues (such as the RAA and LAA) are associated with AF [28,29]. For example, atrial biopsy specimens in lone AF show signs of myocarditis in 66% of these individuals [29]. Some studies have specifically investigated structural and cellular histologic appearances in relation to postoperative AF [30–32], but immunohistochemical staining of VWF, inflammation or extracellular matrix markers were not reported. For example, Ak et al. [30] reported that the degree of myolysis and increased apoptotic pattern in the right atrial myocardium are significant predictors for the development of postoperative AF. Indeed, atrial histology showed degenerative changes that may correlate with advanced age and left atrial enlargement [31]. Nakai et al. [32] reported quantitative assessment of atrial fibrosis with Sirius red stain, and found that age-related atrial fibrosis rather than cellular hypertrophy may be important in the pathogenesis of AF after CABG.

VWF is an established plasma marker of endothelial damage/dysfunction, and increased levels predict adverse cardiovascular outcomes [33]. In our plasma marker study, neither VWF level nor tissue factor level was associated with postoperative AF [24], suggesting that the general hypercoagulability profile in CABG patients did not predict short-term AF. The discrepancy between the inability of plasma VWF to predict postoperative AF [24] and the ability of tissue expression to detect postoperative AF may be a result of the fact that any VWF shed by a perturbed endothelium would be immediately diluted by flowing blood, whereas cell-bound VWF, as detected by the immunohistochemistry, would be expected to have a higher local concentration. We pursued this in an exploratory analysis, relating the expression of VWF by the LAA and RAA to plasma VWF (Table 5). Levels of plasma VWF increased significantly in parallel with LAA expression only in the combined group of subjects. We do not interpret this as suggesting that the increase in plasma VWF is attributable to expression by LAA alone, but it may be that increased LAA tissue expression reflects whole body increased production of VWF. Nevertheless, we have previously reported that peripheral VWF levels could be correlated with severity of endocardial damage, as visualized by scanning electron microscopy of LAA sections obtained during mitral valve surgery [34]. In a further non-operative study, changes in plasma VWF and ADAMTS13 levels have been related to left atrial remodeling in AF [35]. Furthermore, increased VWF and tissue factor expression in atrial endothelia have been associated with thromboembolism in a cohort of patients with non-valvular AF, but no relationship with the development of new postoperative AF was investigated [36]. Indeed, increased VWF expression in the endocardium may represent a local predisposing factor for enhanced thrombogenesis, given that the greatest VWF expression correlated with grade of thrombus formation as well as with associated mitral valve disease [37].

As mentioned above, we are unaware of any published work relating immunohistochemical tissue expression staining of VWF (and other pathophysiologic markers) to the development of postoperative AF. A possible role of endothelial damage/dysfunction (as reflected by VWF changes) in the pathogenesis of postoperative AF and its complications merits further study, and – as implied earlier – may have therapeutic implications, with drugs that have beneficial effects on the endothelium, such as angiotensin-converting enzyme inhibitors and statins, which are often considered as ‘upstream therapy’ for AF management in patients with cardiovascular risk factors [26,27]. Indeed, (systemic) plasma VWF level is related to cardiovascular diseases and increases with age [33] (and, accordingly, is attracting attention as the potential target of a pharmaceutical agent [38]), and our patients who developed postoperative AF were slightly older than those free of AF. Although there is plausible evidence that plasma VWF directly promotes thrombosis [33], we cannot say whether or not the increased expression of VWF by atrial tissue directly contributes to the increased risk of thrombosis present in AF, or, indeed, whether the increased expression of VWF directly contributes to the development of AF. It may be that increased expression of VWF is simply a non-specific marker of the risk of developing AF, and that it has no major pathophysiologic consequences. Whether increased expression of VWF in atrial tissue represents a similar pathology requires confirmation in a larger population, although our ‘non-AF’ group were broadly similar to the ‘AF group’ with regard to associated comorbidities and concomitant drug therapies.

Limitations

Our study is limited by its modest size, although our 30 patients with postoperative AF and immunohistochemical tissue expression staining represent the largest published series in the literature. Although immunohistochemistry is a very powerful method for detecting localization of proteins within tissue sections, we recognize that it is not very suitable for quantification of protein concentrations, which is beyond the scope of the present study objective. More detailed mechanistic analyses would involve a substantially more complicated study, which would require complex molecular biology investigations with western blotting or additional quantitative analyses to confirm the present observations.

Indeed, given the relatively small numbers, we have opted for a simple cross-sectional comparison of those developing postoperative AF with those free of AF, and have not related the histologic changes to the temporal distribution of incident AF, etc. It is of note that all of our patients were observed by continuous ECG monitoring only within the first 24–48 h after CABG – many postoperative AF episodes are short and self-terminating, with a peak of incidence on the second day following cardiac surgery, and AF may be asymptomatic in some individuals. Thus, the possibility remains that some occurrences of postoperative asymptomatic AF were undetected in our patient population. Our study was also underpowered to confidently determine detailed predictors of postoperative AF, based on age, comorbidities (e.g. extent of coronary disease, cardiac function, and left atrial size), and gender. Tissue samples were also examined and scored independently by two board-certified consultant histopathologists, rather than with imaging software. However, the agreement between observers (kappa = 0.58), although ‘moderate’, is close to ‘good’ (defined by statistical convention as a kappa of ≥ 0.61) [25]. Finally, our data are derived from ‘traumatic’ surgery-related AF, and may not be relevant to the wider AF population.

Conclusion

In conclusion, increased expression of VWF by the endocardium of the LAA is a risk factor for the development of AF. Expression of IL-6, tissue factor, MMP-9 and TIMP-1, although varying between tissues, failed to predict the development of AF. These differences point towards a possible role of endothelial damage/dysfunction in the pathogenesis of postoperative AF. The clinical consequences of these observations are unclear, and do not warrant the targeting of plasma VWF [38] in an attempt to reduce postoperative AF. However, they may prompt larger studies designed to test the hypothesis that the amount of VWF expressed by the LAA marks those at risk of postoperative AF and thus more comprehensive postoperative ECG surveillance of this arrhythmia.

Addendum

D. Kaireviciute: collected and analyzed the samples, managed the project, and contributed to drafting of the paper; U. Zanetto and M. Maheshwari: performed immunohistochemical work, and drafted the paper; A. D. Blann: performed primary data analyses and drafting and revisions of the paper; G. Y. H. Lip: principal investigator, provided the original hypothesis, and drafted and revised the paper. Other authors: collected data and commented on drafts of the paper.

Acknowledgements

We thank L. Fabritz for help with the quality control of the staining score. D. Kaireviciute is funded by a research fellowship awarded by the European Society of Cardiology. We acknowledge the support of the Peel Medical Research Trust for assays and reagents for this project. The support of the Sandwell and West Birmingham Hospitals NHS Trust Research and Development Programme for the Haemostasis Thrombosis and Vascular Biology Unit is acknowledged.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.

Ancillary