Air pollution-associated procoagulant changes: the role of circulating microvesicles

Authors


Marc Hoylaerts, Center for Molecular and Vascular Biology, Herestraat 49, B-3000 Leuven, Belgium.
Tel.: +32 16 346145; fax: +32 16 345990.
E-mail: marc.hoylaerts@med.kuleuven.be

Abstract

Summary. Background: Epidemiological studies suggest an association between exposure to particulate matter (PM) in air pollution and the risk of venous thromboembolism (VTE). Objectives: To investigate the underlying pathophysiological pathways linking PM exposure and VTE. Patients and methods: We assessed potential associations between PM exposure and coagulation and inflammation parameters, including circulating microvesicles, in a group of 233 patients with diabetes. Results: The numbers of circulating blood platelet-derived and annexin V-binding microvesicles were inversely associated with the current levels of PM2.5 or PM10, measured on the day of sampling. Recent past exposure to PM10, up to 1 week prior to blood sampling, estimated at the patients’ residential addresses, was associated with elevated high-sensitivity C-reactive protein (CRP), leukocytes and fibrinogen, as well as with tissue factor (TF)-dependent procoagulant changes in thrombin generation assays. When longer windows of past exposure were considered, up to 1 year preceding blood sampling, procoagulant changes were evident from the strongly increased numbers of red blood cell-derived circulating microvesicles and annexin V-binding microvesicles, but they no longer associated with TF. Past PM exposure was never associated with activated partial thromboplastin time (aPTT), prothrombin time (PT), or factor (F) VII, FVIII, FXII or D-dimers. Residential distance to a major road was only marginally correlated with procoagulant changes in FVIII and thrombin generation. Conclusions: Increases in the number of microvesicles and in their procoagulant properties, rather than increases in coagulation factors per se, seem to contribute to the risk of VTE, developing during prolonged exposure to air pollutants.

Introduction

Ambient environmental air pollutants include gaseous and particulate components. Considering a large body of evidence, the American Heart Association scientific statement on ‘Air Pollution and Cardiovascular Disease’ concluded that both short- and long-term exposure to the particulate component (particulate matter, PM) are associated with increased mortality and cardiovascular disease [1]. In addition to the well-recognized air pollution-related adverse effects on the arterial vascular system [2–8], recent epidemiological evidence also suggests an association between PM exposure and venous thromboembolism (VTE). Thus, a higher risk for deep vein thrombosis (DVT) was associated with increased annual mean levels of PM with a mean aerodynamic diameter smaller than 10 μm (PM10) in the residential area of the study subjects [9]. In the same study population, living near major traffic roads was also associated with an increased risk of DVT, even after controlling for the community-level PM pollution [10]. These initial epidemiological findings by Baccarelli et al. [9] were recently confirmed in a time-series analysis in Chile, demonstrating an association between PM exposure and hospital admission for VTE [11], although also challenged by others [12,13].

The pathophysiological mechanisms explaining the observed link between PM exposure and VTE remain largely unknown. Although increases in the levels of coagulation factors seem the most likely explanation, published data for this interpretation are conflicting and unconvincing. In fact, disappointingly few studies reported on positive associations between air pollution exposure and increased levels of coagulation factors, and estimated effect sizes for the reported associations are relatively small [14–23]. Therefore, the observed increases in coagulation factors are unlikely to be (solely) responsible for an increased venous thrombogenicity.

A potential role for microvesicles (also called microparticles, a term we prefer to avoid in the context of pollution by particles) has been suggested [14,24,25]. Microvesicles are circulating vesicles with a mean diameter smaller than 1 μm that are released from stimulated or apoptotic cells in the vascular bed. Negatively charged phospholipids and tissue factor (TF) on their membranes create a procoagulant surface on which coagulation factors can bind and be activated to promote coagulation [26]. Elevated numbers of circulating microvesicles have been demonstrated in patients with VTE [27,28]. A direct link between air pollution and an elevation in the concentration of circulating microvesicles or their procoagulant potential has hitherto never been shown in humans.

In the present study, we hypothesized that microvesicles, through their procoagulant potential, could represent a missing link between air pollution exposure and VTE. We, therefore, investigated associations between PM exposure and markers of inflammation and coagulation, with a focus on microvesicles and microvesicle-dependent coagulation assays. We investigated these associations in an a priori susceptible population of patients with diabetes, because diabetic subjects are more sensitive to the deleterious effects of PM during air pollution [29].

Patients and methods

Study population

Persons with either type 1 or type 2 diabetes were recruited from the diabetes outpatient clinic at the University Hospital Leuven, Belgium, as a new cohort of patients, different from the cohort of our previous studies [30,31]. These patients visit the diabetes clinic as part of their routine follow-up. They were included if they were 18 years or older, current (for >6 months) non-smokers and not on anticoagulant therapy. Inclusion was done on different days from February 2010 through to April 2010. Of 402 patients contacted, 339 agreed to participate (84% participation rate). We excluded 106 patients because of current smoking (n = 74), anticoagulant therapy (n = 16), accidental lack of blood samples (n = 10) or other reasons (n = 6). Thus, the final study population consisted of 233 patients (Fig. 1). On the study day, patients completed a questionnaire through a personal interview to collect information on age, occupation, socio-economic status, exposure to environmental tobacco smoke, alcohol use, use of medication, use of oral contraception, menopausal status, place of residence and means of transportation to the hospital. Socio-economic status was encoded and condensed into a scale with scores ranging from 1 to 3, based on educational level. The Ethics Review Board of the Medical Faculty of the University of Leuven (KULeuven) approved the study. Participants gave informed consent at recruitment.

Figure 1.

 Flowchart of the study population. The study population was consecutively recruited from the diabetes outpatient clinic at the University Hospital Leuven. The lower right box shows the percentage of samples that were measured per group of analyzes. HbA1c, glycated hemoglobin; FACS, flow cytometric analysis; WBC, white blood cells in blood.

Exposure assessment

Current exposure  A portable laser-operated aerosol mass analyzer (Aerocet 531; Met One Instruments Inc., Grant Pass, OR, USA), previously calibrated against a local monitoring station (Flemish Environmental Agency, Borgerhout, Belgium) [31], was used to measure current PM2.5 and PM10 concentrations in the hospital waiting room, 1 to 2 h before the patient’s participation in the study. In general, patients stayed in the waiting room and the neighboring examination room for at least 1 h.

Subacute, subchronic and chronic exposure  The regional background level of PM10 at each patient’s residential address was calculated using a kriging interpolation method. This model provides interpolated PM10 values from the Belgian telemetric air quality network in 4 × 4-km grids (see Fig. 2). The interpolation is based on a detrended kriging interpolation model that uses land cover data obtained from satellite images (Corine land cover data set, European Environment Agency, 2000) [32]. Regional background levels of PM2.5 are not available in Belgium. Mean residential PM10 values were measured for different time windows, and classified in three categories of exposure: (i) subacute: mean residential PM10 values on the day of blood sampling (‘day 0’), on the first (‘day 1’), the second (‘day 2’) or the third (‘day 3’) day before blood sampling; (ii) subchronic: mean residential PM10 values over the preceding week (‘mean 1 week’) or month (‘mean 1 month’); (iii) chronic: mean residential PM10 values over the preceding 3 months (‘mean 3 months’), 6 months (‘mean 6 months’) or 12 months (‘mean 1 year’). Distances from the home address to major roads (N-road, a major traffic road or E-road, a motorway/highway) also reflect chronic PM exposure and were calculated through geocoding (the shortest distance being set at 10 m).

Figure 2.

 Location of residences of study subjects. Map of Belgium reporting location of the study center (University hospital, Leuven, Belgium) and location of the residences of the study subjects. Background colors represent mean annual PM10 concentrations for 2010 in 4 × 4-km grids.

Clinical measurements

All laboratory tests were performed without knowledge of the subject’s exposure data. A detailed description and validation of the assays used in the present study to identify and characterize microvesicles is given in Data S1. Figure S1 schematically summarizes the physiological role of microvesicles in the coagulation cascade and illustrates which assays were performed in the present study to evaluate microvesicles. In brief, microvesicle quantification was performed by flow cytometry (see below). To evaluate their procoagulant potential, surface expression of negatively charged phospholipids was evaluated by flow cytometry and surface expression of functional TF by thrombin generation assays (TGA, see below).

Blood  All blood samples were collected in a restricted time window in the diabetes outpatient clinic (median hour of sampling: 01.50 pm), thus reducing the possibility of confounding by circadian rhythms of some parameters. Non-fasting blood samples were collected using a 21-gauge needle (Terumo, Leuven, Belgium) on EDTA, on sodium fluoride/oxalate, or on sodium citrate (3.8%) tubes (all BD Vacutainer; BD Biosciences, Erembodegem, Belgium). Analysis of blood cell counts and glucose and glycated hemoglobin (HbA1c) levels were performed on fresh full blood or plasma samples. For all other parameters, plasma was stored immediately at −80 °C for future batch analysis. Citrated samples were centrifuged according to two different protocols: for biochemical analyzes, tubes were centrifuged once at 3000 × g. For the analysis of microvesicles, both by flow cytometry and by TGA, tubes were first centrifuged for 10 min at 1900 × g, followed by a second centrifugation step of 20 min at 1900 × g to obtain blood platelet-depleted but microvesicle-rich plasma. All plasma samples were centrifuged within 1 h after collection.

Blood cell counts and routine biochemical analysis  Blood cell counts, coagulation parameters, glucose levels, glycated hemoglobin (HbA1c) and high-sensitivity CRP (hsCRP) were measured according to standard clinical procedures on automated analyzers. The following ‘traditional’ coagulation parameters were measured: activated partial thromboplastin time (aPTT), prothrombin time (PT), factor (F) VII, FVIII, FXII, fibrinogen and D-dimers.

Thrombin generation assays  Thrombin generation was measured by means of the Calibrated Automated Thrombography (CAT) method using a Fluoroskan Ascent reader (Thermo Labsystems OY, Helsinki, Finland). Thrombinoscope software (Thrombinoscope BV, Maastricht, the Netherlands) was used to calculate thrombin generation curves, from which four parameters were derived: lag time (initiation phase of coagulation), endogenous thrombin potential (ETP, area under the thrombin generation curve), peak height (maximal reaction) and time to peak [33]. In the present study, only the lag time and the ETP are reported as representative parameters. To 80 μL of plasma sample, 20 μL of trigger reagent (see below) was added and thrombin generation recording was started upon subsequent addition of 20 μL FluCa, a mixture of calcium chloride (‘Ca’, 87 mm) and thrombin substrate (Z-Gly-Gly-Arg-AMC, 2.5 mm, Bachem, Weil-am-Rein, Germany). Three different analytical conditions were applied:

First, a contact activator (‘S’, Synthasil, Instrumentation Laboratory, Zaventem, Belgium, final concentration 1/400) was added to trigger the intrinsic coagulation pathway, and ‘lag time(Ca,S)’ and ‘ETP(Ca,S)’ were recorded.

Second, TF (Innovin, Siemens, Hamburg, Germany, final concentration 5 pm) was added to trigger primarily the extrinsic coagulation pathway and ‘lag time(TF)’ and ‘ETP(TF)’ were recorded.

Third, thrombin generation was performed upon simple recalcification of microvesicle-rich plasma in the absence of an exogenous trigger, to investigate the effect of endogenous coagulation triggers in the plasma sample, including microvesicle-bound factors such as TF, and ‘lag time(Ca)’ and ‘ETP(Ca)’ were recorded.

In TGA, especially the lag time is to a large extent determined by the amount of TF present in the assay [34]. Hence, to functionally asses the endogenous TF in the plasma sample, we additionally measured ‘lag time(Ca)’ in the presence of 300 ng mL−1 (final concentration) tissue factor pathway inhibitor (TFPI; R&D Systems, Abingdon, UK). This enabled us to specifically investigate the TF-dependency of associations with air pollution exposure, as explained in detail in Data S1.

All TGA’s were performed in the presence of an excess (4 μm) of exogenous phospholipids (phosphatidylserine 30% and phosphatidylcholine 70%; Sigma-Aldrich, Bornem, Belgium), making thrombin generation independent of the surface expression on microvesicles of negatively charged phospholipids.

Microvesicle analysis by flow cytometry  Microvesicles were analyzed by flow cytometry according to a protocol, standardized by the Scientific and Standardization Subcommittee (SSC) of the International Society on Thrombosis and Haemostasis (ISTH) [35] with some adaptations. In brief, to thawed microvesicle-rich plasma, fluorescein isothiocyanate-labeled mouse anti-CD42a (BD Biosciences), phycoerythrin-labeled mouse anti-glycophorein A (GPA; BD Biosciences) and allophycocyanin-labeled annexin V (AV; Immunotools, Friesoythe, Germany) were added and samples were analyzed on a FACSCantoII flow cytometer (BD Biosciences) to define blood platelet-derived microvesicles (‘BPμV’, CD42a+), red blood cell-derived microvesicles (‘RBCμV’, GPA+) and microvesicles with a procoagulant, negatively charged phospholipid surface, binding annexin V (‘AV+μV’). A detailed description of our microvesicle analysis by flow cytometry is provided in Data S1.

Tissue factor mRNA in circulating white blood cells  Tissue factor mRNA expression measurement was performed by quantitative real-time PCR on the AB 7500 Fast PCR System (Applied Biosystems, Ottignies-Louvain-la-Neuve, Belgium) as described in Data S1.

Statistical analysis

For database management and statistical analysis, we used SAS Software (version 9.1; SAS Institute Inc, Cary, NC, USA). Non-normally distributed data were log transformed. We investigated associations between plasma markers and recent or chronic exposure parameters using multiple linear regression. In all regression models we included the following a priori chosen covariates: gender, age, body mass index, socio-economic status, type of diabetes, physical activity, blood glucose levels, use of insulin, use of statins, use of antiplatelet medication, and temperature and humidity on the day of blood sampling. Potential interactions of type of diabetes, use of statins and use of antiplatelet medication with the association between air pollution and all measured parameters were investigated. Q-Q plots of the residuals were used to test the assumptions of all linear models.

Results

Characteristics of the study population

Characteristics of study participants are shown in Table 1. Of the men, 34 (32%) had type 1 diabetes compared with 57 (45%) for the women. All patients with type 1 diabetes used insulin; among type 2 diabetics, 126 persons (89%) used insulin medication. Mean values ± SD for HbA1c were 7.8 ± 1.1% for type 1 diabetics and 7.2 ± 1.2% for type 2 diabetics. Exposure characteristics are shown in Table 2. A representation of the location of the patients’ residences and of the study center is shown in Fig. 2.

Table 1.   Population characteristics (n = 233)
 Mean (SD) or number (%)
  1. *Antiplatelet medication includes acetylsalicylic acid, clopidogrel, ticlopidine or dipyridamole. Non-fasting values. Data are available for 224 persons. §Reference values (fasting): 55–100 mg dL−1. Data are available for 229 persons. **Reference values: 4.0–6.0%. BMI, body mass index.

Gender (men)107 (46%)
Age (years)57.9 (17.5)
BMI (kg m−2)28.9 (5.5)
Type 1 diabetes91 (39%)
Exposure to environmental tobacco smoke34 (15%)
Socio-economic status
 Low158 (68%)
 Middle55 (23.5%)
 High17 (7.5%)
Unknown3 (1%)
Antiplatelet medication*140 (60%)
Statin154 (66%)
ACE inhibitor129 (55%)
Insulin217 (93%)
Oral antidiabetic medication106 (46%)
Blood glucose, mg dL−1139.2 (63.4)‡,§
Glycated hemoglobin, %7.5 (1.2)¶,**
Table 2.   Exposure characteristics of included subjects (n = 233)
Type of exposureMeanMedianRangeP10–P90IQR
  1. IQR, interquartile range; P10, percentile 10; P90, percentile 90.

Current indoor PM2.5, μg m−34.64.31.4–7.62.4–7.22.6
Current indoor PM10, μg m−322.121.312.0–39.415.0–30.85.6
Residential PM10, μg m−3
 Day 025.726.37.3–52.810.7–36.211.4
 Day-127.628.07.2–55.713.0–40.911.5
 Day-228.027.49.9–85.814.2–42.511.4
 Day-328.326.57.7–72.613.1–47.516.2
 Mean 1 week25.225.811.8–44.414.2–37.816.5
 Mean 1 month26.325.511.6–43.019.9–34.38.5
 Mean 3 months25.925.715.0–38.322.9–29.23.08
 Mean 6 months22.822.613.3–34.620.4–25.52.7
 Mean 1 year22.121.712.5–33.719.3–25.072.7
Residential distance to major road, m68941010–519110–1767755

Clinical measurements

The value distribution of clinical parameters for the study population is shown in Table S1 (Data S1). With the exception of FVIII, mean values for the population were within the normal reference ranges.

Figure 3 shows the associations (adjusted for the aforementioned covariates) between air pollution exposure and all outcome parameters, as determined by the regression analysis for different time windows. The corresponding effect sizes for selected time windows are shown in Figs 4–6. A stratified analysis for type of diabetes, statin use or antiplatelet use is shown in supplementary Figs S5–S7 (Data S1).

Figure 3.

 Representation of significant associations between air pollution exposure and outcome parameters. Regression analysis was performed to determine the associations between outcome parameters and current PM2.5 and PM10 concentrations measured in the diabetes outpatient clinic’s waiting room, residential PM10 concentrations over different time windows preceding blood sampling (as indicated) or residential distance to a major road (N- or E-road). For significant associations, positive and negative slopes are denoted as ‘+’ or ‘−’ respectively. Significant associations that represent increases in inflammatory parameters or blood cells, or procoagulant changes, are highlighted in green striped (P < 0.05) or green plain (P < 0.005) boxes. Significant associations that represent decreases in inflammatory parameters or blood cells, or anticoagulant changes, are highlighted in red striped (P < 0.05) or red plain (P < 0.005) boxes.

Figure 4.

 Effect sizes for significant associations with current PM2.5 and PM10 concentrations. Effect sizes (%, 95% confidence interval [CI]) were calculated for each 10 μg m−3 increase in current (A) PM2.5 and (B) PM10 concentrations, measured in the diabetes outpatient clinic’s waiting room in the hours preceding the blood sampling. Those parameters were selected that showed significant correlations in Fig. 3. hsCRP, high sensitivity CRP; WBC, white blood cells; RBC, red blood cells; BP, blood platelets; ETP, endogenous thrombin potential; BPμV, BP-derived microvesicles; RBCμV, RBC-derived microvesicles; AV+μV, annexin-V binding microvesicles. Analysis adjusted for covariates. *P < 0.05, **P < 0.005.

Figure 5.

 Effect sizes for significant associations with residential PM10 concentrations at different time windows before blood sampling. Effect sizes (%, 95% confidence interval [CI]) were calculated for each 10 μg m−3 increase in residential mean PM10 concentrations measured over (A) 1 week, (B) 1 month or (C) 1 year preceding the blood sampling. Those parameters were selected that showed significant correlations in Fig. 3. hsCRP, high sensitivity CRP; WBC, white blood cells; RBC, red blood cells; BP, blood platelets; ETP, endogenous thrombin potential; BPμV, BP-derived microvesicles; RBCμV, RBC-derived microvesicles; AV+μV, annexin-V binding microvesicles. Analysis adjusted for covariates. *P < 0.05, **P < 0.005.

Figure 6.

 Effect sizes for significant associations with residential distance to a major road. Effect sizes (%, 95% confidence interval [CI]) were calculated for each halving in distance from the patient’s residence to the nearest major road (N- or E-road). Those parameters were selected that showed significant correlations in Fig. 3. hsCRP, high sensitivity CRP; WBC, white blood cells; RBC, red blood cells; BP, blood platelets; ETP, endogenous thrombin potential; BPμV, BP-derived microvesicles; RBCμV, RBC-derived microvesicles; AV+μV, annexin-V binding microvesicles. Analysis adjusted for covariates. *P < 0.05.

Inflammation parameters and blood cells  Significant positive correlations were observed between PM10 exposure at the patient’s residence and hsCRP and white blood cell (WBC) concentrations for PM10 exposure windows within 1 week (Fig. 3), with positive but only borderline significant values (0.05 < P < 0.10) for the longer time windows up to 6 months. Each 10 μg m−3 increase in the mean PM10 concentration over the preceding week at the patient’s residence increased the hsCRP by 23% (95% confidence interval [CI]: 5–45) and the WBC by 7% (95% CI: 2–12) (Fig. 5A).

‘Traditional’ coagulation parameters Current PM10 exposure correlated with a prolongation of the PT (Fig. 3). No significant correlations were found between either current, subacute, subchronic or chronic PM exposure and measurements of aPTT, FVII, FVIII, FXII or D-dimers (Fig. 3). Concentrations of fibrinogen correlated positively with PM10 at ‘day-2’ and ‘day-3’, as well as with the mean PM10 concentration over 1 week (Fig. 3). Each 10 μg m−3 increase in the mean concentration of PM10 over the preceding week at the patient’s residence elevated fibrinogen levels by 4% (95% CI: 1–7) (Fig. 5A). Halving each residential distance to a major road increased FVIII by 2% (95% CI: 1–3) (Fig. 6).

Thrombin generation  In contrast to the paucity of relationships between PM exposure and the above mentioned hemostasis parameters, we found strong significant correlations with various parameters of thrombin generation over many different time windows of exposure (Fig. 3). The strongest correlations were found for the lag time during thrombin generation with a contact activator (‘lag time[Ca,S]’) and for both the lag time and ETP after simple recalcification of microvesicle-rich plasma, in the presence of standardized phospholipid concentrations (‘lag time[Ca]’ and ‘ETP[Ca]’). These significant correlations were consistent for the subacute and subchronic time windows, with the highest effect sizes measured for the mean PM10 concentration over 1 month: each 10 μg m−3 increase in PM10 shortened the ‘lag time(Ca,S)’ by 6% (95% CI: 0–12, P = 0.08), the ‘lag time(Ca)’ by 14% (95% CI: 4–25) and increased the ‘ETP(Ca)’ by 14% (95%CI: 5–24). Interestingly, upon the addition of a specific tissue factor inhibitor (TFPI) to the assay, the positive associations between PM10 and ‘lag time(Ca)’ disappeared for the subacute windows, but not for the subchronic time windows (Fig. 3). Each halving in residential distance to a major road increased the ‘ETP(Ca)’ by 2% (95%CI: 0–4) (Fig. 6).

Microvesicle analysis  The concentrations of current PM2.5 correlated negatively with the concentrations of BPμV (−57%, 95%CI: −78 to −16% per 10 μg m−3 increase in PM2.5) and of AV+μV that express negatively charged phospholipids on their surface (−74%, 95%CI: −85 to −56% per 10 μg m−3 increase in PM2.5), whereas both current PM10 and residential PM10 on ‘day 3’ correlated negatively with the concentration of AV+μV (−27%, 95%CI: −39 to −12% per 10 μg m−3 increase in current PM10) (Figs 3 and 4). In contrast, we found positive correlations between mean PM10 concentrations over the chronic exposure windows (preceding 3 months to 1 year) and the concentrations of RBCμV and AV+μV with the highest effect sizes measured for the 1-year period (+77% 95% CI: 14–176 and +60% 95% CI: 15–123 per 10 μg m−3 increase in yearly PM10, respectively) (Fig. 5C). Similar borderline significant results were found for the BPμV.

TF mRNA in circulating WBC  No significant correlations were observed between air pollution exposure and levels of TF mRNA expression in circulating WBCs (Fig. 3).

Effect modification

Table 1 demonstrates, for our study population of patients with diabetes, the frequent use of different types of medication, some of which (e.g. statins) have been described to influence levels of TF or microvesicles. Therefore, we investigated if type of diabetes, use of statin medication and use of antiplatelet medication induced effect modification on the association between PM exposure and outcome variables. Results are shown in Figs S5–S7 of Data S1.

In general, stratification for type of diabetes (Fig. S5) or medication (Figs S6 and S7) did not considerably influence the non-stratified associations described in Fig. 3. For patients with type 1 diabetes, increased current PM exposure was associated with decreases in inflammatory parameters (hsCRP, fibrinogen), in thrombin generation (‘ETP[Ca,S]’ and ‘ETP[Ca,TF]’) and microvesicles (BPμV, RBCμV, AV+μV) that were not, or to a lesser extent, observed in patients with type 2 diabetes, who rather showed increases in inflammatory parameters (Fig. S5), compatible with the more inflammatory nature of this type of diabetes. Both statins (Fig. S6) and antiplatelet use (Fig. S7) tended to decrease the effect size of associations between chronic (1 week to 1 year) PM exposure and inflammatory (WBC, neutrophils) or procoagulant (‘ETP[Ca,S]’ and ‘ETP[Ca,TF]’) changes, as compared with patients not taking these medications.

Discussion

Parameters of inflammation, coagulation and microvesicles were correlated with measures of current (at blood sampling), subacute (day 0 to day 3), subchronic (mean 1 week to 1 month) and chronic (mean 3 months–1 year) PM exposure and with residential distance to a major road, in patients with diabetes. Type 2 diabetes is especially a chronic inflammatory disease and circulating microvesicles seem to be elevated in these patients [36,37]. It was, therefore, relevant to measure inflammation parameters and microvesicle numbers in patients with diabetes who manifest increased susceptibility to air pollution [29].

Current PM levels were associated with lower numbers of circulating microvesicles and with decreased inflammatory parameters, mainly in patients with type 1 diabetes.

We even found an isolated prolongation of the PT with increased levels of current PM10, in contrast to previous observations [14,38]. However, the lower number of circulating microvesicles should not necessarily be interpreted as an anti-inflammatory response to short-term PM exposure but can be explained by the recruitment of circulating microvesicles to the lung via enhanced expression of adhesive receptors, including P-selectin [39,40] and von Willebrand factor [41], by acutely activated pulmonary endothelial cells.

In contrast to the current PM exposure, consistent pro-inflammatory and procoagulant changes were observed for the longer PM exposure windows.

Subacute and subchronic exposure up to 1 week were associated with a systemic inflammatory status, evidenced by increased hsCRP, total WBC counts and neutrophil counts. In agreement with other studies [15,42,43], and compatible with its role as an acute phase protein, fibrinogen concentrations increased with higher PM exposure levels, within 1 week. We did not measure increases in FVIII, another acute phase protein, but the high mean baseline value for FVIII in a diabetic study population (Table S1 and [44]), could hinder a further increase by PM exposure.

Yet, increased fibrinogen concentrations cannot explain the strong correlations observed here between different thrombin generation parameters and subacute and subchronic PM10 exposure up to 1 month, as TGA are not influenced by fibrin(ogen) levels. Likewise, in the absence of procoagulant changes in any of the other ‘traditional’ coagulation parameters (PT, aPTT, FVII, FVIII, FXII and D-dimers), other processes should be responsible for enhancing thrombin generation with higher levels of PM exposure so consistently.

A role for microvesicles, cellular bodies released from stimulated or apoptotic cells, in VTE has been suggested [27,45]. We assessed the procoagulant potential of microvesicles through measurement of their surface expression of TF in TGA, and through the analysis of the number of red blood cell and blood platelet-derived microvesicles via flow cytometry. In addition, flow cytometric analysis of annexin V-binding was undertaken to measure the surface expression of negatively charged phospholipids (mainly phosphatidylserine) [26]. Specifically the latter measurement has one drawback, for example that freezing-thawing affects the expression of negatively charged phospholipids [46,47], and therefore may not provide an accurate index of phosphatidylserine exposure in vivo. However, the consistency of the strong associations of the number of AV+μV with PM exposure over the different longer time windows indicates that these associations are unlikely to be artificially induced chance findings.

Both subacute and subchronic PM exposure correlated strongly (P-values < 0.0001) with thrombin generation. The most pronounced correlations were found for TGA performed in the absence of an external trigger of coagulation (‘lag time[Ca]’ and ‘ETP[Ca]’), and may therefore depend on the presence of endogenous triggers present in the plasma, such as contact activation or microvesicle-bound TF, the concentration of the latter having a pronounced effect on the lag time [34].

In the subacute time window, associations between PM10 and lag time disappeared both in the presence of an excess of exogenous TF (‘lag time[Ca,TF]’) and upon inhibition of TF by TFPI (‘lag time[Ca]+TFPI’). This points towards exposure-associated increased levels of circulating TF, most likely on microvesicles [48]. The source of TF-bearing microvesicles is a matter of debate, yet with activated monocytes being the most likely candidate [49]. The remarkable coincidence in the subacute time window of associations of PM exposure with inflammatory parameters and with TF-dependent changes in TGA adds to the hypothesis of inflammation-coagulation cross-talk in the first days after exposure to air pollution. In healthy individuals, microvesicles derived from blood platelets and red blood cells account for almost all circulating microvesicles. As WBC-derived microvesicles are extremely low [50], we could not quantify these microvesicles via flow cytometry.

In the subchronic time window, associations between PM exposure and thrombin generation were no longer dependent on TF. Associations with lag times were also present upon exogenous TF addition (‘lag time[Ca,TF]’) and did not disappear upon addition of TFPI. Moreover, procoagulant changes at 1 month occurred in the absence of those inflammatory changes mentioned above. Therefore, the enhanced thrombin generation associated with PM concentrations over several weeks must be explained by (an) other mechanism(s). A recent study in mice suggests that PM promotes early procoagulant changes mostly through a TF-driven extrinsic pathway of coagulation, whereas long-lasting procoagulant effects are predominantly mediated through contact activation of the intrinsic pathway of coagulation by systemically translocated ultra-fine particles [51]. Yet, we found no associations between FXII and exposure to PM. An alternative explanation could be offered by downregulation of the anticoagulant pathways, including protein C and antithrombin, but these markers were not assessed in the present study.

In the chronic PM exposure window, procoagulant changes were no longer obvious from thrombin generation measurements. Yet, a procoagulant tendency was apparent from the higher microvesicle numbers, both blood-platelet derived and red blood cell-derived, and increased microvesicular annexin V binding, reflecting surface expression of negatively charged phospholipids (mainly phosphatidylserine) [26]. In the present study, thrombin generation assays were performed in the presence of an excess of exogenous phospholipids, and are therefore insensitive to the effects by endogenous negatively charged phospholipids. We found the highest effect size on microvesicle number and procoagulant potential for the mean PM10 measurement over 1 year. Interestingly, in the study by Baccarelli et al. [9], a 1-year exposure time window correlated most strongly with the risk of DVT, whereas no significant correlations were found for time windows shorter than 9 months. Hence, upregulation of procoagulant microvesicles could, at least partly, be a pathophysiological mechanism underlying the association between long-term PM exposure and VTE [9,27,28].

Living near a major road has been associated with increased risk for VTE in a case–control study [10]. A recent population-based prospective cohort study [13], also demonstrated an, admittedly, non-significant 16% increase in the risk of VTE for subjects living within 150 m of a major traffic road. In the present study, the correlations with residential distance to a major road were, although following similar trends, fewer and weaker than for the chronic residential PM10 measurements by the land-use interpolation model. This is further discussed in Data S1.

The present study has limitations. First, association studies do not prove causality, and our observations can therefore only suggest that procoagulant changes, consisting of enhanced thrombin generation and higher numbers of procoagulant microvesicles, are induced by exposure to PM. Second, the ‘disappearance’ of associations between subacute PM exposure and thrombin generation in the presence of TFPI in the assays demonstrates a crucial role for circulating TF, which is not necessarily all microvesicle bound. Indeed, TF in plasma is primarily located on microvesicles, but it can also circulate as an alternatively spliced soluble protein [48]. Nevertheless, plasma depletion of microvesicles by filtration through a 0.1-μm filter significantly prolonged the lag time of thrombin generation assays in previous experiments in our laboratory (data not shown), suggesting that microvesicles are indeed the major source for TF. Third, a large number of statistical analyzes were performed in the present study, increasing the possibility of rejecting the null hypothesis too readily. However, correction for multiple testing is not always appropriate, as discussed in larger detail in the online data supplement.

In conclusion, the present study demonstrates for the first time that increases in the number and the procoagulant potential of microvesicles, rather than increases in coagulation factors per se, may contribute to the prothrombotic risk induced by air pollution exposure.

Addendum

J. Emmerechts, L. Jacobs, T. Nawrot, B. Nemery and M. Hoylaerts conceived and designed the research; J. Emmerechts, L. Jacobs, S. Van kerckhoven, S. Loyen and F. Fierens acquired the data; L. Jacobs and T. Nawrot performed epidemiological and statistical analyses; J. Emmerechts and M. Hoylaerts drafted the manuscript; and J. Emmerechts, L. Jacobs, T. Nawrot, C. Mathieu, B. Nemery and M. Hoylaerts critically reviewed the intellectual content of the paper.

Acknowledgements

The authors would like to thank F. Mullier and M. Beerens for their skillful advice in flow cytometric microvesicle analysis and RT-PCR measurements, respectively, and the nurses of the UZ Gasthuisberg diabetes outpatient clinic for their kind assistance with the phlebotomies. J. Emmerechts is holder of a grant from the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen, project 81045). For part of this work, J. Emmerechts received the CSL Behring Heimburger Award in 2011. The C.M.V.B. is supported by the ‘Programmafinanciering KULeuven’ (PF/10/014). The studies of the Flemish Center of Expertise on Environment and Health were commissioned, financed and steered by the Ministry of the Flemish Community (Department of Economics, Science and Innovation; Flemish Agency for Care and Health; Department of Environment, Nature and Energy). T.N. received a grant of the Flemish Fund for Scientific Research (FWO-Vlaanderen, G.0873.11N/1.2.506.07.N.00). C.M. is a clinical fellow of the FWO-Vlaanderen.

Disclosure of Conflict of Interest

The authors state that they have no conflict of interest.

Ancillary