Association between CD8+ T-cell subsets and cardiovascular disease

Authors


Correspondence: Jan Nilsson, CRC, Jan Waldenströms gata 35, Skåne University Hospital, S-205 02 Malmö, Sweden.

(fax: +46-40-39-12-12; e-mail: Jan.Nilsson@med.lu.se).

Abstract

Background

The findings of experimental studies suggest that the immune system plays a key role in atherosclerosis, but the clinical importance of different immune cells in cardiovascular disease remains poorly characterized. In this study we investigated the association between CD8+ T cells and carotid disease as well as development of cardiovascular disease events.

Methods

The study cohort comprised 700 subjects from the cardiovascular arm of the Malmö Diet and Cancer Study. Peripheral blood mononuclear cells, obtained at the 1991–1994 baseline investigation and stored at −140 °C, were thawed and the different CD8+ T-cell populations analysed by flow cytometry. Baseline carotid intima–media thickness and stenosis were assessed by ultrasonography and clinical events were monitored through validated national registers.

Results

Subjects with a high fraction of CD8+ T cells were characterized by decreased cytokine release from activated leucocytes, metabolic signs of insulin resistance and increased incidence of coronary events; hazard ratios (95% confidence intervals) for the second and third tertiles of CD8+ T cells were 2.57 (1.16, 5.67) and 2.61 (1.19, 5,71), respectively, in a Cox proportional hazards regression model. Correlations were found between the fraction of CD8+CD25+ T cells and the degree of carotid stenosis (r = 0.11, < 0.01), and between the CD8+CD56IFN-γ+ T-cell fraction and the degree of stenosis (r = −0.18, < 0.005). The association between CD8+CD56IFN-γ+ T cells and carotid stenosis remained significant after controlling for major cardiovascular disease risk factors.

Conclusion

This study provides prospective clinical evidence for a role of CD8+ T cells in cardiovascular disease and suggests the existence of CD8+ T-cell subsets with different pathological functions.

Introduction

It is well established that inflammation plays an important role in the development of atherosclerotic plaques as well as in the destabilization of lesions that precedes plaque rupture and arterial occlusion [1]. It has also been demonstrated that plaque inflammation is orchestrated by a complex array of adaptive autoimmune responses against plaque antigens such as oxidized LDL and that both atherogenic and protective immune responses exist [2]. Most interest has focused on the role of CD4+ T cells, the predominant T-cell type in mouse atherosclerotic lesions [3]. However, in contrast to plaques in experimental mouse models of atherosclerosis, advanced human atherosclerotic plaques contain predominantly CD8+ T cells [4]. CD8+ cells can act as cytotoxic T cells that kill cells presenting viral and tumour antigens on MHC class I molecules, but they have also been implicated in autoimmune diseases [5]. There is relatively limited information from experimental studies concerning the role of CD8+ T cells in atherosclerosis. Hypercholesterolaemia induces activation of CD8+ T cells in lymph nodes draining the aorta followed by activation of lymph node CD4+ T cells [6]. Targeting CD8+ T cells towards arterial smooth muscle cells expressing the bacterial transgene β-galactosidase has been shown to aggravate atherosclerosis in Apoe−/− mice [7]. It is interesting that CD8+ T cells have also been shown to inhibit neointima formation induced by vascular injury [8]. CD8+ T-cell deficiency does not affect atherosclerosis development in Apoe−/− mice [9, 10]. However, these genetic knockout studies do not exclude the possibility that there are subpopulations of CD8+ T cells with opposing effects on atherosclerosis. Zhao et al. recently showed that adoptively transferred CD8+ T cells mediate the atheroprotective effects of apolipoprotein B peptide immunization in Apoe−/− mice[11], supporting the existence of an atheroprotective CD8+ T-cell population.

Modulation of adaptive immune responses against vascular antigens represents a novel therapeutic approach to inhibit vascular inflammation in atherosclerosis [12]. However, clinical application of this approach has been limited by poor understanding of the role of immune cells in human atherosclerosis. To investigate the possible involvement of CD8+ T cells in cardiovascular disease (CVD) in humans, we analysed the association between baseline circulating CD8+ T cells and carotid disease in 700 subjects participating in a cardiovascular programme of the Malmö Diet and Cancer Study (MDCS). CD8+ T cells were expressed as a proportion of the total number of CD3+ T cells. We also analysed two subsets of activated CD8+ T cells: circulating CD8+ T cells expressing CD25 and CD8+CD56 T cells expressing interferon (INF)-γ in response to stimulation with phorbol-12-myristate-13-acetate (PMA)/ionomycin and brefeldin A. It has been proposed that expression of CD25 [the α-subunit of the interleukin (IL)-2 receptor] represents a suppressor function of CD8+ T cells in humans [13], but the possible role of CD8+CD25+ T cells in CVD remains unknown. The CD56 gate was used to exclude CD8-expressing natural killer and natural killer T (NKT) cells. We also determined whether baseline levels of different circulating CD8+ T-cell populations predict the development of acute myocardial infarction and ischaemic stroke during a 15-year follow-up period.

Material and methods

Study population

The MDCS is a prospective cohort (= 28 449) study examining the association between diet and cancer [14]. Individuals born between 1926 and 1945 and living in Malmö were eligible for inclusion in the study. Between October 1991 and February 1994, every other MDCS participant was also invited to take part in a substudy focusing on CVD risk (MDCS cardiovascular cohort, = 6103 [15]). In this study, 700 participants aged 63–68 years were randomly selected from the MDCS cardiovascular cohort [16]. Participants were followed from baseline examination until first CVD event, emigration from Sweden, death or 31 December 2008, whichever came first. Methods for retrieval and ascertainment of cases have been shown to be accurate with high validity of the registries used (the Swedish Discharge Registry, the Stroke Register of Malmö and the Cause of Death Registry of Sweden) [17, 18]. A CVD event was defined as a fatal or nonfatal myocardial infarction [i.e. International Classification of Diseases, 9th Revision (ICD-9) code 410)], fatal or nonfatal stroke (ICD-9 codes 430, 431, 434 and 436) or death attributable to underlying coronary heart disease (ICD-9 codes 410–414), whichever came first. A total of 150 incident CVD events (84 coronary events and 66 strokes) were identified during the follow-up period. Eleven cases of haemorrhagic stroke and one of stroke of unknown subtype were excluded from the analysis. Hypertension was defined as blood pressure ≥140/90 mmHg or use of blood-pressure-lowering medication, high cholesterol was defined as >5 mmol L−1 and smoking as current smoking. Blood pressure, body mass index (BMI), smoking status and cholesterol and lipid levels were determined as previously described [15]. One subject was excluded because of incomplete clinical data. The study was approved by the Regional Ethical Review Board in Lund and was conducted in accordance with the Declaration of Helsinki. All participants gave written consent.

B-mode ultrasound

Analysis of common and bulb carotid intima–media thickness (IMT) was performed using an Acuson 128 CT system with a 7-MHz transducer as described previously [15]. Briefly, the right carotid bifurcation was scanned within a predefined window comprising 3 cm of the distal common carotid artery, the bifurcation and 1 cm of the internal and external carotid arteries. All images for measurement of IMT were obtained in the longitudinal projection showing the thickest intima–media complex. Plaque was defined as a focal thickening of the intima–media complex exceeding 1.2 mm. The thickness of the common carotid intima–media complex, i.e. the mean distance between the leading edges of the lumen–intima and the media–adventitia interfaces of the far wall (mean common carotid artery IMT), was measured offline and along a 1-cm section in the longitudinal projection using a specially designed computer-assisted image analysing system based on automated detection of the echo structures, but with the option of operator-controlled manual correction. Common carotid intima–media area (IMA) and carotid stenosis were analysed as described previously [19-21].

Isolation of mononuclear leucocytes

Blood (15 mL) was collected in heparin tubes and layered on top of 15 mL Lymphoprep before centrifugation at 1350 g for 12 min at room temperature. The interface layer was carefully harvested and the cells were then washed twice with 0.9% NaCl (the first centrifugation was at 600 g and the second at 300 g, both for 10 min at room temperature). The cells were resuspended in 1.7 mL autologous serum and 1.6 mL cold RPMI 1640 medium with 20% DMSO was added. The cells were frozen slowly by placing them in a Styrofoam box at −80 °C overnight. Frozen mononuclear cells were stored at −140 °C.

Analysis of cytokine concentrations in cell supernatants

Cells were cultured in complete RPMI and stimulated with CD3/CD28 beads (MiltenyiBiotec, Bergisch Gladbach, Germany) for 72 h at 37 °C in a cell incubator (with 5% CO2). The cell supernatants were then stored at −80 °C until analysis. The concentrations of released cytokines were determined with multiplex technology (MesoScale Discovery, Rockville, MD).

Flow cytometry

Prior to thawing, cells were transferred from storage at −140 °C to liquid nitrogen. Cells were then thawed to room temperature within 2 min; subsequently, 4 mL phosphate-buffered saline (PBS) containing 1% human serum (HS) at 37 °C was added over 1–2 min and a further 4 mL was added over approximately 30 s. All experiments were performed in a cell culture hood under sterile conditions. Gradual dilution of the DMSO in the frozen cells avoided cell damage by osmotic shock, and prewarming the dilution medium helped to actively compensate for changes in osmotic pressure. Cells were centrifuged and suspended in complete medium [RPMI 1640 supplemented with 10% HS (Invitrogen, Stockholm, Sweden), 1% sodium pyruvate, 1% HEPES, 1% penicillin/streptomycin, 1% L-glutamine and 0.1% β-mercaptoethanol (Gibco, Stockholm, Sweden)] at a concentration of 2 × 106 cells mL−1. First, 4 × 105 cells were stained with 7AAD, CD3-PE/Cy7, CD4-PB, CD45-biotin, CD8-AF700 and CD25-PE (antibody panel 1; all from Biolegend, San Diego, CA) for 30 min at 4 °C. Cells were washed in FCM buffer [5% bovine serum albumin (w/v) in PBS with 2 mM EDTA] and stained with streptavidin-CasY for 20 min at 4 °C. For panel 2, 4 × 105 cells/well were incubated with PMA (10 ng), ionomycin (0.2 μg) and brefeldin A (1 μg, all from Sigma) for 4 h at 37 °C. Cells were washed in FCM buffer and incubated with CD56-biotin (Biolegend), CD3-PE/Cy7, CD4-PB and CD8-AF700 (Biolegend) for 30 min at 4 °C. The samples were washed with FCM buffer and incubated with strepavidin-PerCP (Biolegend) for 20 min at 4 °C. Cells were washed and incubated with Fix/Perm solution and permeabilization buffer and subsequently incubated with interferon (IFN)-γ-PE (Biolegend) for 30 min at 4 °C. Samples were acquired on an ADP-Cyan flow cytometer (Beckman Coulter, Brea, CA) and analysis was performed using FlowJo 7.5.5 (Tree Star Inc., Ashland, OR). For both panels, species-specific compensation beads (BD Biosciences, Stockholm, Sweden) coupled to the respective antibodies were used to compensate fluorescent signals and nonstained cells, and FMO (fluorescence-minus-one) control samples were used to set the negative population. Data for CD8+ and CD8+CD25+ T cells were obtained from nonstimulated cells stained with antibody panel 1, whereas CD8+CD56IFN-γ+ T cells were derived from cells stimulated in vitro to induce IFN-γ production and stained with antibody panel 2. CD14++CD16 monocytes were analysed as previously reported [16].

Statistical analysis

A combined strategy including probability–probability plots and measures of skewness and kurtosis was used to test for normal distribution of CD8+ T-cell populations and clinical parameters. Differences between means of normally distributed continuous variables were assessed with independent sample t tests, and differences in proportions between subjects with and without CVD events were assessed using the chi-squared test. Differences between means of nonnormally distributed continuous variables were assessed using the nonparametric Mann–Whitney test. Spearman's rank correlation coefficients were used to examine relationships amongst continuous variables. Linear regression models were used to calculate independent associations. The relation between CD8+ T cells (in tertiles) and incidence of first CVD events (coronary events and ischaemic stroke) during follow-up was assessed by Kaplan–Meier survival curves and quantified by Log-rank test. A Cox proportional hazards regression model was used to assess the hazard ratio (HR, and 95% confidence interval) of first CVD events (coronary events and ischaemic stroke) in relation to tertiles of CD8+ T cells. The basic model included age and gender; a multivariate model included additional adjustment for diabetes, blood pressure, lipid levels, smoking status and use of antidiabetic or blood-pressure-lowering medication. The proportionality of the hazards assumption was confirmed by visual inspection of log–negative-log survival curves.

Results

Baseline clinical characteristics of subjects with and without CVD events are shown in Table 1. Coronary cases were more often male, with diabetes and receiving blood-pressure-lowering medication. They were also characterized by higher fasting blood glucose, LDL/HDL ratio and systolic blood pressure (Table 1). Stroke cases were more often male, treated with antidiabetic and blood-pressure-lowering medication and had higher fasting blood glucose levels.

Table 1. Baseline clinical characteristic of study participants
 All noncases = 549Coronary casesa = 84Stroke casesa = 54
  1. hsCRP, high-sensitivity C-reactive protein; BP, blood pressure.

  2. All values are presented as mean ± standard deviation except triglycerides which are presented as median (range). *< 0.05, **< 0.01, ***< 0.005 for case versus noncase.

  3. a

    Mann–Whitney or chi-squared test for categorical data

  4. b

    History of diabetes, antidiabetic medication or fasting glucose ≥6.1 mmol L−1

  5. c

    Blood pressure ≥ 140/90 mmHg or antihypertensive treatment.

Age at screening65.6 ± 1.165.7 ± 1.265.6 ± 1.2
Gender (% male)38.3%53.6%**53.7%*
BMI (kg m−2)26.3 ± 4.026.6 ± 4.126.4 ± 3.8
Current smoker16.0%25.3%22.0%
Diabetesb11.1%25.0%***20.3%
Hypertensionc79.4%86.9%87.0%
Medication use   
Antidiabetic2.3%4.7%13.0%***
Lipid lowering2.9%4.7%7.4%
Blood pressure lowering20.4%33.3%*38.9%**
Laboratory parameters
Fasting blood glucose5.3 ± 1.35.7 ± 1.8*5.9 ± 2.2**
Triglycerides (mmol/L)1.3 (0.5–6.5)1.3(0.5–5.4)1.5(0.6–3.7)
HDL (mmol L−1)1.4 ± 0.41.3 ± 0.41.3 ± 0.4
LDL (mmol L−1)4.3 ± 1.04.4 ± 1.04.2 ± 1.3
LDL/HDL ratio3.4 ± 1.13.7 ± 1.4*3.6 ± 1.5
Systolic BP (mmHg)150 ± 20156 ± 21*153 ± 19
Diastolic BP (mmHg)88 ± 9.289 ± 890 ± 9
hsCRP (mg L−1)2.9 ± 5.04.5 ± 8.13.7 ± 5.4

CD8+ T-cell subsets were analysed in thawed samples of mononuclear leucocytes that had been frozen in autologous serum/DMSO at the MDCS baseline investigation in 1991–1994 and stored at −140 °C. Flow cytometric analysis revealed limited uptake of 7AAD in thawed CD45+ cells and demonstrated that 95% of the cells remained viable. We have previously shown that there is no difference in 7AAD uptake between cells that had been kept frozen for only a few weeks and those frozen for at least 15 years, suggesting that the duration of freezing at −140 °C has limited impact on cell viability [22]. Comparing cell numbers measured at freezing and thawing also confirmed that no loss of cells had occurred.

CD8+ T-cell analysis

The gating strategies used to define the different CD8+ T-cells populations are shown in Fig. 1. The mean fractions of CD8+ (expressed as percentage of all CD3+ cells), CD8+CD25+ and CD8+CD56IFN-γ+ (both expressed as percentage of all CD8+ cells) T cells in the study cohort were 35% ± 14%, 2.6% ± 2.6% and 14% ± 12% respectively. The viability of the CD8+ T cells, as evaluated by 7AAD staining, was 93.3% ± 3.3%.

Figure 1.

Gating strategy for mononuclear cells. Lymphocytes were gated into CD45 + 7AAD (viable leucocytes), CD3+ (T lymphocytes) and CD8+CD4 (a) or CD3+CD56 (to exclude NKT cells), CD8+CD4 and IFN-γ+ cells (b) and analysed using FlowJo 7.5.5. A shows nonstimulated cells and B shows cells stimulated with PMA/ionomycin/brefeldin A for 4 h at 37 °C.

Association between CD8+ T-cell populations and leucocyte cytokine release

Lymphocyte function was assessed in the leucocyte samples by measuring anti-CD3/anti-CD28 bead-activated cytokine release. The different CD8+ T-cell populations were associated with markedly different patterns of leucocyte cytokine release. Leucocytes from subjects with a high percentage of CD8+ T cells were characterized by a lower release of proinflammatory cytokines such as IL-1β, IL-6 and IL-8 but also by a lower release of the anti-inflammatory cytokine IL-10 (Table 2). Consistent with this observation we also found an association between a high percentage of CD8+ T cells and a lower level of tumour necrosis factor (TNF)-α in plasma (r = −0.12, < 0.005). By contrast, leucocytes from subjects with a high percentage of CD8+CD56IFN-γ+ T cells were characterized by a higher release of IL-1β, IL-6, IL-8, IL-10 and TNF-α. Moreover, there were significant associations between the fraction of CD8+CD56IFN-γ+ T cells and plasma levels of IL-5 (r = 0.15, < 0.001), IL-12 (r = 0.09, < 0.05), IFN-γ (r = 0.09, < 0.05) and TNF-α (r = 0.12, < 0.005). Leucocytes from subjects with a high percentage of either CD8+CD25+ or CD8+CD56IFN-γ+ T cells had similar cytokine release characteristics, but the effects were generally weaker in the former.

Table 2. Correlation (r values) between CD8+ T-cell populations and cytokine release from anti-CD3/anti-CD28 bead-stimulated mononuclear leucocytes
 CD8+CD8+CD25+CD8+CD56IFN-γ+
  1. Mononuclear leucocytes were cultured and stimulated with anti-CD3/anti-CD28 beads for 72 h at 37 °C and the concentrations of released cytokines were determined by multiplex technology. CD8+ T cells are expressed as percentage of all CD3+ cells, and both CD8+CD25+ and CD8+CD56IFN-γ+ T cells are expressed as percentage of all CD8+ cells. The Spearman′s rank correlation test was used to calculate r values. ***< 0.005.

IL-1β−0.12***0.15***0.31***
IL-20.040.01−0.23***
IL-40.050.00−0.08
IL-50.050.02−0.14***
IL-6−0.20***0.16***0.32***
IL-8−0.20***−0.040.19***
IL-10−0.26***0.14***0.25***
IL-12p70−0.050.02−0.02
IFN-γ0.010.020.03
TNF-α−0.050.030.19***

There was a significant inverse association between the fraction of CD8+ T cells and the size of the classical CD14++CD16 monocyte population (r = −0.14, < 0.001), which is believed to be the major monocyte type recruited to sites of inflammation [23]. By contrast, a positive correlation was found between the CD8+CD56IFN-γ+ T-cell fraction and CD14++CD16 monocyte numbers (r = 0.19, < 0.001). There was no significant association between the numbers of CD8+CD25+ T cells and CD14++CD16 monocytes.

CD8+ T-cell populations and CVD risk factors

The fractions of CD8+ and CD8+CD25+ T cells increased (r = 0.12, < 0.005 and r = 0.10, < 0.05, respectively), whereas the fraction of CD8+CD56IFN-γ+ T cells decreased with age (r = −0.22, < 0.001). The proportion of CD8+ T cells was higher in men than in women (38% ± 10% vs. 33% ± 13%, < 0.001). Smokers (= 124) had a smaller fraction of CD8+CD56IFN-γ+ T cells than nonsmokers (= 477; 11% ± 15% vs. 15% ± 13%, < 0.001). The associations between the different CD8+ T-cell populations and other major CVD risk factors are shown in Table 3. A high fraction of CD8+ T cells was found to correlate with several characteristics of insulin resistance including a high waist–hip ratio, high fasting plasma glucose, insulin and triglyceride levels and low plasma HDL cholesterol. However, there was no difference in the fraction of CD8+ T cells between subjects with (= 93) and without diabetes (= 606; 35% ± 14% vs. 35% ± 14%, n.s.).

Table 3. Correlation (r values) between CD8+ T-cell populations and major CVD risk factors
 CD8+CD8+CD25+CD8+CD56IFN-γ+
  1. hsCRP, high-sensitivity C-reactive protein; BP, blood pressure.

  2. CD8+ T cells are expressed as percentage of all CD3+ cells and both CD8+CD25+ and CD8+CD56IFN-γ+ T cells are expressed as percentage of all CD8+ cells. The Spearman′s rank correlation test was used to calculate r values. *< 0.05, ***< 0.005.

Waist–hip ratio0.12***−0.02−0.02
Fasting blood glucose0.11***−0.01−0.03
HbA1c0.030.05−0.12***
Plasma insulin0.15***0.020.01
Triglycerides (mmol L−1)0.09*−0.04−0.02
HDL (mmol L−1)−0.16***0.030.00
LDL (mmol L−1)−0.020.030.00
LDL/HDL ratio0.10*−0.020.02
Systolic BP (mmHg)0.000.00−0.05
Diastolic BP (mmHg)0.000.000.05
hsCRP (mg L−1)0.06−0.040.00

CD8+ T cells and carotid IMT and stenosis

We next determined the association between CD8+ T-cell populations and baseline carotid disease determined as the common carotid artery IMT and IMA, and percentage of carotid stenosis as assessed by B-mode ultrasound. Significant correlations were found between the proportion of CD8+CD25+ T cells and the degree of stenosis (r = 0.11, < 0.01), and between the CD8+CD56-INFγ+ T-cell fraction and both the degree of stenosis (r = −0.18, < 0.005) and carotid IMA (r = −0.09, < 0.05). There were no significant associations between the proportion of CD8+ T cells and carotid IMT or stenosis.

Gender-specific analysis was also performed because average carotid size is different in men and women. In men, the fraction of CD8+CD56IFNγ+ T cells was inversely associated with the degree of carotid stenosis (r = −0.14, < 0.05), IMT (r = −0.14, < 0.05) and IMA (r = −0.16, < 0.01). The CD8+CD56-IFNγ+ T-cell fraction was also inversely associated with the degree of stenosis in women (r = −0.22, < 0.01), whereas the CD8+CD25+ fraction was positively correlated (r = 0.12, < 0.05). In linear regression models including age, LDL and HDL cholesterol, triglycerides, fasting glucose and systolic blood pressure, the proportion of CD8+CD56-IFNγ+ T cells remained independently associated with the degree of stenosis (entire cohort: beta coefficient −0.13, < 0.005; women: beta coefficient −0.18, < 0.005; men: beta coefficient −0.06, n.s.).

Association between CD8+ T-cell subsets and incident CVD events

Next, we assesses whether baseline CD8+ T-cell populations predicted the HR for development of a first CVD event during follow-up. During this period there were 84 incident coronary events and 54 ischaemic strokes, corresponding to 10.0 coronary events and 6.6 ischaemic strokes per 1000 person-years. Kaplan–Meier curves of event-free survival showed a trend towards increased incidence of coronary events in the two highest tertiles of CD8+ T cells (log-rank test: P for trend <0.01, Fig. 2A), whereas no such trend was observed for ischaemic stroke. There was, however, an association between tertiles of CD8+CD56-IFNγ+ T cells and incident ischaemic stroke (log-rank test: P for trend <0.05) with subjects in the second tertile having the highest incidence (Fig. 2B). An increased HR for coronary events was found for both the second and third tertiles of CD8+ T cells (2.56, < 0.01 and 2.49, < 0.05, respectively, Table 4) in the basic Cox proportional hazards regression model adjusted for age and gender. Increased HR values for incident coronary events were also observed for the two highest tertiles of CD8+ T cells when additionally adjusting for fasting glucose, diabetes, LDL/HDL ratio, systolic blood pressure, hypertension, use of blood pressure and diabetes medication and smoking habits (Table 4). Entering tertiles of CD8+CD56-IFNγ+ T cells into the basic model confirmed an increased incidence of ischaemic stroke for the second tertile (HR 2.01, < 0.05). However, this association did not remain significant when also adjusting for other CVD risk factors in the multivariate Cox proportional hazards regression model.

Table 4. Hazard ratios (95% confidence intervals) for incident coronary and stroke events by tertiles of CD8+ T cells
 Tertile 1 referenceTertile 2Tertile 3P for trend
  1. Associations between tertiles of CD8+ T cells and development of a first CVD event were calculated using two different Cox proportional hazards regression models. The basic model adjusted only for age and gender and the multivariate model additionally adjusted for smoking status, LDL/HDL ratio, systolic blood pressure, hypertension, blood pressure medication, diabetes and diabetes medication. Ranges for CD8+ T cells: tertile 1, <27.77%; tertile 2, 27.78%–39.64%; tertile 3 > 39.65%.

Basic model
Coronary event1.002.56 (1.26, 5.18)2.49 (1.23, 5.05)<0.05
Ischaemic stroke1.001.30 (0.66, 2.56)1.02 (0.53, 1.98)n.s.
Multivariate model
Coronary event1.002.57 (1.16, 5.67)2.61 (1.19, 5.71)<0.05
Ischaemic stroke1.001.32 (0.64, 2.72)0.78 (0.38, 1.58)n.s.
Figure 2.

Kaplan–Meier survival curves for CD8+ T-cell populations and development of a first CVD event. Associations between tertiles of CD8+ T cells and a first incident coronary event (a) and CD8+CD56IFN-γ+ T cells and a first incident ischaemic stroke (b). CD8+ T cells are expressed as percentage of all CD3+ cells and CD8+CD56IFN-γ+ T cells as percentage of all CD8+ cells. Ranges for CD8+ and CD8+CD56IFN-γ+ T cells were as follows: tertile 1, <27.77%; tertile 2, 27.78%–39.64%; tertile 3, >39.65% and tertile 1, <7.65%; tertile 2, 7.66%–14.61%; tertile 3, >14.62% respectively.

As shown above there was an inverse association between the fraction of CD8+ T cells and the leucocyte release of IL-10, an anti-inflammatory cytokine known to protect against atherosclerosis [24]. Leucocyte release of IL-10 was also higher in subjects with the lowest CD8+ T-cell tertile as compared with the two highest tertiles combined (675 ± 642 vs. 484 ± 516 pg mL−1, < 0.001). As increased IL-10 release represents a possible protective mechanism that is unlikely to be directly mediated by CD8+ T cells, we also included leucocyte IL-10 release in the multivariate Cox proportional hazards regression model described above. However, CD8+ T-cell tertile remained an independent predictor of incident coronary events when including the leucocyte IL-10 release into the multivariate model.

Discussion

There is convincing evidence that T cells and other components of adaptive immunity play a key role in the development of experimental atherosclerosis [25], but the clinical importance of these findings remains to be fully confirmed. It has been shown in several prospective population studies that a high white blood cell count is an independent risk factor for CVD events [26-29]; however, these studies have not investigated the importance of different T-cell subtypes, and white blood cell count has generally been regarded as a marker of inflammation. Findings of experimental studies have revealed a complex role for immune cells in the atherosclerotic disease process, underlining the need to use detailed phenotypic characterization when investigating the association between these cells and CVD. The findings of this study demonstrate that subjects with a low fraction of CD8+ T cells have a reduced risk of development of acute myocardial infarction. It should be noted that these findings do not prove that CD8+ T cells are directly involved in the disease process; they may simply represent surrogate markers. Nevertheless, our observations point to the existence of a biological association between CD8+ T cells and atherosclerosis in humans. An additional potentially important finding of this study is that the associations with carotid disease vary amongst the different CD8+ T-cell populations. Whereas a large fraction of total CD8+CD25+ T cells was related to a high degree of carotid stenosis, the opposite was observed for CD8+CD56IFN-γ+ T cells. This implies the existence of CD8+ T-cell populations that either have different roles in the disease process or different associations with other proatherogenic or protective processes. However, it should be noted that the associations between CD8+ T-cell populations and carotid disease were weak and therefore need to be interpreted with caution. Evidence for a role of CD8+ T cells in CVD was also recently provided by Bergström et al. [30] who showed that patients with both acute coronary syndromes and stable angina have increased proportions of CD8+CD56+ T cells compared with healthy control subjects. Of interest, they also found higher levels of expression of IFN-γ+ in CD56+ than in CD8+CD56 T cells, further supporting the notion of CD8+ T-cell subsets with potentially different roles in CVD.

It has been established that plaque inflammation aggravates atherosclerosis, and systemic markers of inflammation predict risk of development of CVD events [31]. Of note, we found that a high fraction of CD8+ T cells was associated with reduced leucocyte release of proinflammatory cytokines such as IL-β and IL-6 as well as with lower levels of the CD14++CD16 monocytes that are thought to be recruited to sites of inflammation. Accordingly, it is unlikely that the increased risk conferred by CD8+ T-cell elevation is explained by general inflammatory activation. In contrast to the total CD8+ T-cell population, a larger fraction of IFN-γ-responsive CD8+ T cells was significantly associated with increased leucocyte release of IL-β, IL-6, IL-8 and TNF-α and with higher levels of the CD14++CD16 monocytes. This implies that subjects with a high fraction of IFN-γ+-expressing CD8+ T cells have a more proinflammatory leucocyte phenotype. Notably, this was associated with less carotid stenosis, further suggesting that the association between CD8+ T cells and CVD is unlikely to involve effects on inflammation. An important exception to the general pattern of association between CD8+ T cells and inflammatory markers was IL-10. It has been shown that this anti-inflammatory cytokine has potent atheroprotective effects in experimental models of atherosclerosis [24]. The fraction of CD8+ T cells showed a highly significant inverse association with the leucocyte release of IL-10, which could explain the observation of a correlation between a low fraction of CD8+ T cells and a decreased risk of acute myocardial infarction. However, additional adjustment for IL-10 in the Cox proportional hazards regression model did not influence the results.

Interesting associations were also observed between CD8+ T cells and several CVD risk factors. The fraction of CD8+ T cells was higher in men than women, and increased with age. A high fraction of CD8+ T cells was also related to a high waist–hip ratio, increased levels of fasting glucose, plasma insulin and triglycerides and low HDL cholesterol, indicating an association between CD8+ T cells and the insulin resistance/metabolic syndrome. CD8+ T cells have previously been shown to contribute to adipose tissue inflammation and insulin resistance in obesity [32], providing a possible explanation of these observations. The fraction of CD8+CD56IFN-γ+ T cells was inversely associated with plasma glycated haemoglobin (HbA1c), but was not related to any other metabolic CVD risk factors.

The mechanisms through which CD8+ T cells are associated with CVD in humans remain to be fully elucidated. As discussed above, CD8+ T cells do not appear to contribute to the general leucocyte inflammatory response. We found significant correlations between CD8+ T cells and several factors commonly associated with the insulin resistance/metabolic syndrome, suggesting a possible link with increased CVD risk. However, CD8+ T cells remained independently associated with development of myocardial infarction after adjusting for fasting glucose, diabetes, LDL/HDL ratio and use of antidiabetic medication. Another possible pathological mechanism underlying the effect of CD8+ T cells is cytolysis of vascular cells which in turn would exacerbate inflammation. Evidence for this was provided by Ludewig et al. in a mouse model of beta-galactosidase expression in vascular smooth muscle cells. Following injection of beta-galactosidase-specific DCs the mice developed arteritis and atherosclerosis via CD8+ T-cell-mediated lysis of smooth muscle cells [7]. Induction of smooth muscle cell death has also been shown to result in plaque vulnerability in experimental atherosclerosis, as assessed by thinning of the fibrous cap, loss of collagen, accumulation of cell debris and intense intimal inflammation [33]. We recently reported activation of CD8+ T cells in response to diet-induced hypercholesterolaemia in Apoe−/− mice, and demonstrated that activation of CD8+ T cells precedes that of CD4+ T cells in lymph nodes draining atherosclerotic lesions [6]. Accordingly, CD8+ T cells that specifically recognize antigens in atherosclerotic lesions could be important mediators of tissue destruction and chronic inflammation. Candidate antigens for CD8+ T-cell recognition are mimicry proteins of viral or bacterial origin. Jonasson et al. reported expansion of CD8+ T cells associated with cytomegalovirus infection in patients with coronary artery disease [34]. Furthermore, antibodies against cytomegalovirus cross-react with heat shock protein 60 (HSP60) and are associated with atherosclerosis [35]. Moreover, both CD8+ and CD4+ T cells isolated from human atherosclerotic plaque react with HSP60 and/or Chlamydia pneumoniae [36-38], suggesting the importance of immune responses against these microorganisms in atherosclerosis.

The limitations of this study should be considered. Most importantly, our analyses were performed using cells that had been stored at −140 °C for several years. Compared with initiating new prospective studies, this has the obvious advantage of allowing the study to be completed within a relatively short period of time. However, it remains to be fully established whether thawed cells accurately represent the original cell population. Although we were unable to detect any loss of cells by comparing cell numbers at the time of freezing and at thawing, the possibility of a selective loss of CD8+ T cells cannot be excluded.

In conclusion, the present results provide prospective clinical evidence for a role of CD8+ T cells in CVD and suggest the existence of CD8+ T-cell subsets with different pathological functions. Although the total CD8+ T-cell population was associated with increased risk of myocardial infarction, reduced release of proinflammatory cytokines from activated leucocytes and metabolic signs of insulin resistance, the subpopulation of IFN-γ-producing CD8+ T cells was associated with increased release of proinflammatory cytokines from activated leucocytes. A higher proportion of this subpopulation of CD8+ T cells was inversely correlated with both degree of carotid stenosis and HbA1c level. Further studies are required to characterize the biological mechanisms responsible for the associations between CD8+ T cells and CVD in humans.

Conflict of interest statement

None of the authors has any conflicts of interest to declare.

Acknowledgements

This study was supported by grants from the Swedish Research Council, the Swedish Heart-Lung Foundation, the Swedish Foundation for Strategic Research, VINNOVA, the Crafoord Foundation, the Söderberg Foundation, the Albert Påhlsson Foundation, the Malmö University Hospital Foundation and the Lundström Foundation.

Ancillary