Extracellular vesicle-derived CD14 is independently associated with the extent of cardiovascular disease burden in patients with manifest vascular disease

Abstract

Background

In patients with established cardiovascular disease, high levels of the extracellular vesicle (EV)-derived proteins cystatin C, CD14, and α2-antiplasmin predict recurrent cardiovascular events. We examined whether these proteins are associated with the extent of vascular disease.

Methods

In 1062 patients from the SMART (Secondary Manifestations in ARTerial disease) study, EVs were isolated from plasma at baseline. Cystatin C, CD14, and α2-antiplasmin were measured in these vesicles using a multiplex assay. The extent of vascular disease burden was determined by a sum score that incorporates history and current presence of clinically manifest coronary, cerebrovascular, peripheral arterial, and abdominal aneurysm disease, and parameters of atherosclerosis that were assessed during the SMART screening protocol (ankle–brachial index, common carotid intima–media thickness, carotid stenosis, and aorta diameter). The relation between EV protein levels and extent of vascular disease was evaluated using ordinal multivariable regression models.

Results

EV-derived CD14 was significantly associated with the number of affected vascular territories (OR 2.4, 95% CI 1.4–4.1) as represented by the sum score, independently of cardiovascular risk factors. Cystatin C and α2-antiplasmin EV levels did not show an independent association with vascular disease extent. When investigating parameters of the sum score separately, we did not observe a strong association between any of the EV-derived proteins and the markers of atherosclerosis.

Conclusions

EV-derived CD14 levels are strongly correlated to the extent of vascular disease, but not specifically to markers that reflect atherosclerosis burden, in patients with manifest cardiovascular disease.

Introduction

Extracellular vesicles (EVs) are membrane vesicles including microvesicles, membrane particles, and exosomes ranging from 20 to 1000 nm in size and are shed mainly upon cell activation or apoptosis.^1,2 Besides their role in physiological responses such as intercellular communication³ and different immune responses,² they are also related to pathophysiological mechanisms in cardiovascular disease (CVD).¹ In this aspect, both circulating EVs derived from different cell types and plaque EVs have been shown to play a role in the development and progression of atherosclerosis.^4–7

Recent studies have indicated that high levels of EVs of endothelial origin independently predict recurrent events in patients with heart failure⁸ and, similarly, that different subsets of EVs were predictive for heart failure hospitalization and major bleeding complications in patients with non-ST-elevation myocardial infarction.⁹ Serum EV-derived polymeric immunoglobulin receptor, cystatin C, and C5a concentrations are independently associated with presence of acute coronary syndrome in patients with suspected acute coronary syndrome.¹⁰ Thus, the number, type, and content of EVs contain information about CVD.

In addition, circulating levels of EV-derived cystatin C, CD14, and α2-antiplasmin were previously shown to predict recurrent cardiovascular events.¹¹ It has not been established whether these EV proteins are also associated with the extent of vascular disease.

In this study, we therefore investigated the association between EV-derived cystatin C, CD14, and α2-antiplasmin and the extent of CVD. This was assessed using a sum score combining information on clinically manifest CVD and parameters of CVD during the screening protocol of the Second Manifestations of ARTerial disease (SMART) study: ankle–brachial index (ABI), common carotid intima–media thickness (CCIMT), carotid stenosis on duplex ultrasound, and abdominal aorta (AA) diameter.

Methods

Patient population

We used cross-sectional data from patients enrolled between 2001 and 2005 in the SMART study. The SMART study is an ongoing prospective single-centre cohort study, which has been described in detail previously.¹² Patients who are referred to the University Medical Center Utrecht, Utrecht, the Netherlands, with manifest CVD (coronary artery disease (CAD), cerebrovascular disease (CBVD), peripheral arterial disease (PAD), or abdominal aortic aneurysm (AAA)) were included. Regardless of referral diagnosis, all patients were investigated for history or presence of the aforementioned diseases.

CAD was defined as a history of myocardial infarction, cardiac arrest, or coronary revascularization (coronary bypass surgery or coronary angioplasty). CBVD was defined as a history of a transient ischaemic attack (including amaurosis fugax), cerebral infarction, retinal infarction, or prior carotid endarterectomy. PAD was defined by a resting ABI <0.9 in at least one leg or prior surgery for PAD. Patients were classified to have an AAA when their aorta diameter was ≥3 cm or had a history of AAA surgery.

The SMART study has been approved by the ethics committee of our institution and written informed consent was obtained from all participants.

Data collection and measurement of clinical parameters

All patients who entered the SMART study underwent a screening protocol for detection of atherosclerotic disease and vascular risk factors, details of which have been previously published.¹² In short, participants completed a questionnaire on cardiovascular history, risk factors, and medication use. Physical examination was carried out and fasting blood samples were taken to measure levels of various markers including glucose, lipids, creatinine, and albumin.

ABI was calculated for both legs as the ratio between the highest systolic blood pressure measured in each ankle (posterior tibial and dorsal pedal arteries) and the highest blood pressure in both brachial arteries. The lowest of the ABIs of the left and right leg was used as the outcome variable, as recommended recently for studying CVD risk by the American Heart Association’s scientific statement.¹³

CCIMT was measured in the left and right common carotid arteries using colour Doppler-assisted duplex scanning with a 12–5 mHz linear transducer. Reference point for measurement of CCIMT was the beginning of the dilatation of the carotid bulb, with loss of the parallel configuration of the near and far walls of the common carotid artery. The mean CCIMT of six measurements (anterolateral, posterolateral, and mediolateral direction in both carotid arteries) in each patient was calculated. Carotid stenosis was assessed at both carotid arteries (common, bifurcation and internal) using colour Doppler-assisted duplex scanning and was defined according to clinical guidelines.

AA diameter was measured with ultrasonography and the maximum diameter between infrarenal and the bifurcation was recorded and used as a parameter in this study.

More detailed information about the ABI, CCIMT, and AA measurements have been described previously.¹²

Selection and isolation of extracellular vesicle-derived proteins

Out of 1309 patients in the SMART cohort included between 2001 and 2005, 1062 patients had sufficient EDTA plasma available to measure cystatin C, CD14, and α2-antiplasmin, and these patients were subsequently included in this study. EVs were extracted from thawed EDTA plasma using Exoquick (System Biosciences, Mountain View, USA) and the proteins were quantitatively analysed by a multiplex assay using a Bioplex 200 system (Bio-Rad, Austin, USA). Similarly, the proteins were measured in EDTA plasma in a random subset of 522 patients. All EV protein levels were corrected for total protein amount, as measured by Thermo Scientific Pierce BCA Protein Assay Kit (Pierce Biotechnology, Rockford, USA). Hence, this resulted in protein levels expressed in pg/µg, which was used in all analyses.

We refer to previously published research in this cohort for comprehensive methodology of EV isolation and protein quantification, including selection of the specific proteins in this study.¹¹

Sum score

To analyse the association between EV proteins and the extent of CVD, we constructed a sum score consisting of different parameters for CAD, CBVD, PAD, or AAA. These parameters were either history or presence of clinically manifest disease, and/or abnormalities in atherosclerosis markers ABI, CCIMT, carotid stenosis, and AA diameter (Table 1). Cut-off values for these markers shown in Table 1 were based on guideline recommendations^13,14 and expert opinion. Based on this combination, a more valid estimation of the extent of CVD was obtained, since patients classified as having only CAD, CBVD, PAD, or AAA could also have an abnormal ABI, CCIMT, AA diameter, or carotid stenosis.

Statistical analysis

The sum score (cumulative number of affected vascular areas) was used as dependent variable in ordinal regression. The test of parallel lines was used to confirm that the assumption of proportional odds was not violated.

Linear or logistic regression models were applied to test the relation between EV proteins and atherosclerosis parameters (ABI, CCIMT, carotid stenosis, AA diameter) separately.

We corrected for previously determined (cardiovascular risk) factors (age, gender, body mass index, low-density lipoprotein, smoking, diabetes, hypertension, glomerular filtration rate, platelet aggregation inhibitor use, and statin use) in all analyses to determine the independent relation between EV proteins and outcomes. To investigate which parameters in the sum score were responsible for a potential association, we performed additional analyses in which we separated history/presence of disease from atherosclerosis markers as an outcome.

Potential effect modification by gender was examined by adding a multiplicative interaction term between gender and EV proteins to the adjusted analyses. This revealed no significant effect modification, so no stratification on gender was applied.

In all regression models, naturally transformed EV protein levels (after adding 1) were used because of their right-skewed distribution. After this transformation, a normal distribution resulted for EV protein levels. Likewise, logarithmically transformed EV protein levels were used in the UNIANOVA to compare these levels between different baseline characteristics, adjusting for age and gender.

In a subset analysis in a random selection of half of the cohort, the correlation between the proteins derived from EVs and directly from plasma was tested by Spearman’s correlation coefficient. The association of plasma proteins and vascular disease extent was tested in ordinal regression as above (in quartiles).

A two-sided p-value of 0.05 divided by three (the number of proteins tested) was defined as statistically significant, to correct for multiple testing. To avoid the risk of unnecessary multiple testing, we chose to exclude C1 inhibitor, because it was not significantly associated with secondary events in the previous study in this cohort.¹¹ Statistical analyses were performed using IBM SPSS version 20.

Results

This study comprised 1062 patients, of which 21% were women, with a mean age of 59 years. Overall, 37% of the patients were current smokers, 17% had diabetes, and 55% suffered from hypertension. The majority of patients in this cohort (58%) had a history or presence of clinically manifest CAD. For CBVD, PAD, and AAA, these percentages were 27, 25, and 10% of patients, respectively (Table 2). Cystatin C and CD14 EV protein levels were significantly higher in increasing-age groups, smokers, statin and platelet aggregation inhibitor users, and patients with diabetes and/or impaired kidney function (glomerular filtration rate (Modification of Diet in Renal Disease) ≤60 ml/min/1.73m²). CD14 and α2 antiplasmin EV protein were significantly higher in women, smokers, patients with body mass index >25 kg/m², high low-density lipoprotein cholesterol levels, and/or platelet aggregation inhibitor use. Most associations with risk factors were found for CD14, followed by cystatin C and α2-antiplasmin (Supplementary Table S1, available online).

CCIMT, lowest ABI, and maximum AA diameter were (mean ± SD) 0.96 ± 0.33 mm, 1.03 ± 0.24, and 2.1 ± 0.64 cm, respectively. Carotid stenosis >50% was present in 166 patients (16%). Median and mean sum score were 1 and 1.5 vascular areas, respectively. In 639 patients, only one vascular territory was affected. Two, three, and four affected vascular areas were found in 302, 84, and eight patients, respectively. Because of the low numbers in the last group, patients with three and four affected regions were combined in the analyses.

EV protein levels after natural logarithmic transformation were 2.36 ± 0.41 pg/µg for cystatin C, 2.54 ± 0.28 pg/µg for CD14, and 3.49 ± 0.95 pg/µg for α2-antiplasmin (Table 3).

EV CD14 was associated with the number of affected vascular territories in the sum score (odds ratio (OR) 3.6, 95% confidence interval (CI) 2.2–5.6). This effect was also independent of (cardiovascular) risk factors (OR 2.4, 95% CI 1.4–4.1) (Table 4).

We found similar results when looking at the association within the subgroup of patients with only manifest disease in the past or present (≥2 areas diseased (n = 199) versus 1 area diseased (n = 863): OR for CD14 3.6, 95% CI 1.8–7.1). To assess whether the association between the sum score and EV protein levels was explained by the extent of atherosclerosis, we adjusted for atherosclerosis markers that were assessed during the SMART screening protocol. The ORs and confidence intervals did not substantially change following this adjustment, as shown in Table 4. This suggests that the association may not be simply explained by the extent of atherosclerotic disease but by other factors that contribute to CVD.

Indeed, we did not observe statistically significant associations between expression of EV proteins and atherosclerosis parameters after correction for multiple testing, as shown in Table 5. However, trends could be observed for increasing EV CD14 levels with decreasing ABI, increase of CCIMT, higher degree of carotid stenosis, and increasing AA diameter. EV-derived cystatin C and α2-antiplasmin were not significantly associated with the extent of CVD, as indicated by the sum score (OR 1.1, 95% CI 0.76–1.7 and OR 1.1, 95% CI 0.93–1.3, respectively), or as indicated by history/presence of disease and markers of atherosclerosis burden separately (Tables 4 and 5). Total EV protein content itself was not significantly associated with outcome in any of the multivariable models (data not shown).

EV-derived cystatin C, CD14, and α2-antiplasmin showed a moderate correlation with their plasma equivalents in a subset of 522 patients from the cohort (Spearman’s R 0.47, 0.55, and 0.65, respectively; all p < 0.001)). Only plasma CD14 was found to be associated with vascular disease extent (OR 1.2, 95% CI 1.01–1.5). For cystatin C and α2-antiplasmin, these ORs were 1.1 (95% CI 0.92–1.4) and 1.1 (95% CI 0.93–1.3).

Discussion

Extracellular vesicles are abundant in plasma and involved in intercellular communication. For almost two decades now, they have been recognized to be important in many biological and pathological processes.¹⁵ Thus, isolating these vesicles and studying their content could provide crucial information about different diseases. Indeed, EVs have previously been linked to atherosclerosis progression, and the presence and recurrence of CVD.^4–11

In this cross-sectional study in patients with manifest vascular disease, we have shown that extracellular vesicle-derived CD14, a monocyte marker, is strongly related to the extent of vascular disease burden as expressed by the involvement of disease in different vascular territories. Although all three EV proteins under study independently predict the occurrence of secondary cardiovascular events during follow up in the SMART study,¹¹ EV cystatin C and α2-antiplasmin did not reveal an independent association with either the history and/or presence of vascular disease, or with imaging markers reflecting CVD.

In plasma, cystatin C level is an indicator of kidney function and a strong risk predictor for future cardiovascular events.¹⁶ High levels of α2-antiplasmin, a serine protease inhibitor, could increase the risk of atherothrombosis via fibrinolysis inhibition.¹⁷ Our results suggest that EV cystatin C and α2-antiplasmin levels contain more overlapping information with known risk factors, while CD14 seems to hold additional information regarding the extent of vascular disease in a population with manifest vascular disease. In addition, trends towards an association were observed for ABI, CCIMT, carotid stenosis, and AA diameter as parameters of CVD. Nonetheless, the adjustment for these markers did not influence the strength of the association with the extent of vascular disease. This suggests that the extent of atherosclerotic plaque burden does not fully explain the underlying mechanism by which CD14 is correlated with extent of CVD.

Monocyte-derived vesicles, as indicated by CD14 in our study, have been shown to interact with endothelial cells (ECs) and are associated with EC dysfunction. Tissue factor has an important role in this process, because tissue factor expression is not only increased in ECs following stimulation by monocyte microvesicles in vitro,^18,19 but EV-derived tissue factor also showed increased adherence to ECs. In other studies, circulating tissue factor, with monocytes as its main source, was associated with vascular diseases such as stent thrombosis²⁰ and coronary artery disease.²¹ In addition, monocyte-derived vesicles impair anticoagulant pathways and induce EC death.¹⁹ These procoagulant and proapoptotic pathways could thus be important in the relation between EVs carrying CD14 on their surface and extent of vascular disease.

This study has several limitations. First, we have not identified the amount and origin of the EV proteins, regarding cell type, location (inside the EV or on its membrane), and specific type of EVs (exosomes, microvesicles, or apoptotic vesicles). However, we have confirmed earlier that at least part of the proteins are derived from EVs.¹¹ Furthermore, plasma levels of cystatin C, CD14 and α2-antiplasmin were not consistently analysed in all patients. Plasma levels were only measured in a subset of patients, showing moderate but significant correlations with their EV equivalents. Only CD14 showed a significant association with vascular disease extent; however, this association was much weaker than for EV CD14. Although not measured in the complete cohort, this supports current knowledge that EVs contain valuable information about atherosclerotic disease development and progression.²² In addition, it would be very valuable and interesting to define diagnostic cut-off values for detecting presence of (extent of) vascular disease in specific patient groups in future research, which was not possible in this study. Finally, because our sum score was based on a combination of parameters included in the SMART study and not externally validated, our results cannot be generalized to other cohorts.

In conclusion, EV-derived CD14 levels are strongly correlated to the extent of vascular disease, but not specifically to markers that reflect atherosclerosis burden, in patients with manifest cardiovascular disease.

Funding

This work was supported by the UMC Utrecht Vascular Prevention Project and the Royal Netherlands Academy of Art and Sciences (a strategic grant to the Interuniversity Cardiology Institute of the Netherlands and DPVdK).

Conflict of interest

DPVdK and GP are consultants for Cavadis; a company for the development of biomarker kits. The other authors declare no conflict of interest.

The SMART study group

The members of the SMART study group of UMC Utrecht are: PA Doevendans, Department of Cardiology; A Algra, Y van der Graaf, DE Grobbee, GEHM Rutten, Julius Center for Health Sciences and Primary Care; LJ Kappelle, Department of Neurology; WPTM Mali, Department of Radiology; FL Moll, Department of Vascular Surgery; FLJ Visseren, Department of Vascular Medicine.

References