Atherosclerosis and vasomotor dysfunction in arteries of animals after exposure to combustion-derived particulate matter or nanomaterials.

Abstract Exposure to particulate matter (PM) from traffic vehicles is hazardous to the vascular system, leading to clinical manifestations and mortality due to ischemic heart disease. By analogy, nanomaterials may also be associated with the same outcomes. Here, the effects of exposure to PM from ambient air, diesel exhaust and certain nanomaterials on atherosclerosis and vasomotor function in animals have been assessed. The majority of studies have used pulmonary exposure by inhalation or instillation, although there are some studies on non-pulmonary routes such as the gastrointestinal tract. Airway exposure to air pollution particles and nanomaterials is associated with similar effects on atherosclerosis progression, augmented vasoconstriction and blunted vasorelaxation responses in arteries, whereas exposure to diesel exhaust is associated with lower responses. At present, there is no convincing evidence of dose-dependent effects across studies. Oxidative stress and inflammation have been observed in the arterial wall of PM-exposed animals with vasomotor dysfunction or plaque progression. From the data, it is evident that pulmonary and systemic inflammation does not seem to be necessary for these vascular effects to occur. Furthermore, there is inconsistent evidence with regard to altered plasma lipid profile and systemic inflammation as a key step in vasomotor dysfunction and progression of atherosclerosis in PM-exposed animals. In summary, the results show that certain nanomaterials, including TiO2, carbon black and carbon nanotubes, have similar hazards to the vascular system as combustion-derived PM.


Introduction
There is a well-established link between long-term air pollution exposure and increased mortality due to cardiovascular disease (CVD), particularly ischemic heart disease (IHD) and myocardial infarction (Brook et al. 2010). The excess risk in mortality per 10 mg/m 3 increase in fine particles in urban air is 11% (95% confidence interval: 5-16%) for CVDs, particularly IHD, in a large meta-analysis of studies from USA, Asia and Europe (Hoek et al. 2013). This translates into particulate matter (PM) in ambient air contributing to 3.1% of the worldwide global burden of disease, which is dominated by IHD (Lim et al. 2012). Recent meta-analysis of crosssectional studies from urban areas in USA and Europe provide support by findings of associations between long-term exposure to fine particles near residences and carotid intima-media thickness, which is a measure of preclinical atherosclerosis (Adar et al. 2013, Perez et al. 2015, Provost et al. 2015. However, the role of specific sources of ambient air PM, including diesel exhaust (DE), in relation to CVD outcomes has not been sufficiently elucidated in such epidemiological studies. Moreover, despite concern over increasing exposure to nanomaterials (NMs), which due to the particulate nature could have similar effects, has this yet to be established in epidemiological or human exposure studies related to CVD outcomes. Thus, a feasible way to bridge the knowledge gap between exposure to different types of particles and a comparison of CVD outcomes in a quantitative fashion is via the utilization of animal experimental models. As links between exposure to PM and CVD outcomes were first described in relation to inhalation of ambient air pollution, there has been a tendency to view this association as a secondary effect in respiratory toxicology. However, gastrointestinal exposure to NMs may also occur or even directly systemic application by intravenous (i.v.) injection of nanomedicine (De Jong and Borm 2008;Card et al. 2011).
It is generally acknowledged that the underlying cause of PM-induced CVD is atherogenesis and promotion of atherosclerosis, although pro-thrombotic tendency (or imbalance of fibrinolysis) and dysrhythmia are important steps for myocardial infarction and heart failure (Brook et al. 2010). The purpose of the present review is to summarize and integrate studies on ambient air pollution particles and NMs in the same quantitative analysis. We have included results on atherosclerosis and vasomotor function from experiments in animal models.
Vasomotor function refers to the ability of blood vessels to dilate and contract when stimulated with certain vasoactive compounds. This measurement is directly linked to the integrity of the blood vessels, whereas pro-thrombosis tendency occurs in the blood and dysrhythmia is a relatively complex measure of the cardiac electrical conductivity. Endothelium-dependent vasomotor dysfunction is observed in plaque-laden vessel segments, but it is also an early event in atherogenesis (Ross 1999). In general, bioassays for progression of atherosclerosis are often prolonged due to the time required for plaque development. Hence, the long-term exposure period and requirement of special atherosclerosis-prone animals make the assessment of PM-generated plaque progression a costly investigation. The measurement of vasomotor dysfunction is an attractive alternative to detection of progression of atherosclerosis because it is applicable in traditional wild-type animals, although laboratory procedures require considerable training and skill. We have included data from experiments on vasoconstriction and vasorelaxation in terms of endothelium-dependent and independent responses to vasoactive agents. These responses are recorded by treating vessels either in situ by infusion with a vasoactive agent or ex vivo by the isolation of a vessel segment and exposure to the agent in a wire (i.e. without pressure) or pressure myograph. Phenylephrine (PE) is the most commonly used vasoconstrictor in studies on PM-generated vasomotor dysfunction. Endothelium-dependent and endotheliumindependent vasorelaxations are most commonly assessed by the use of acetylcholine (ACH) and sodium nitroprusside (SNP), respectively. The selection of an exact vasoactive agent may depend on the research of the altered function of a specific receptor that promotes a vasomotor response. In addition, the response to a vasoactive agent typically differs between animal species and between blood vessels in the same animal. Thus, there are currently no recommendations regarding the choice or relevance of specific vasoactive agents for assessment of PM-generated vasomotor dysfunction. Here within we have reported observations of myogenic responses in studies of pressurized vessel segments that in the context of the present paper may be interpreted as a ''pre-vasoactive drug treatment condition''. This is typically observed as a reduced ability to sustain normal luminal diameter when the vessel is exposed to increased pressure. In the context of this review, this can be interpreted as a vulnerability of the vessel, but may not affect measurements of drug-induced vasorelaxation or vasoconstriction as these parameters are usually normalized to standardized maximal vasoconstriction responses.
This review encompasses results from 78 publications on exposures to ambient air pollution particles, diluted whole DE, diesel exhaust particles (DEP) and NMs (a description of the literature search is provided in the Supplement) (Figure 1). The figure illustrates that CVD effects following ambient air pollution exposures were assessed first chronologically, whereas there has been a relatively fast accumulation of studies on NMs in recent years. A steady accumulation of studies on DE (or DEP) is also seen, which clearly highlights that the assessment of CVD outcomes related to exposure to PM is an on-going process. Some studies have investigated ambient air pollution particles and DEP that has been obtained on samples that represent ''historic exposures'' in the sense that traffic-related air pollution changes over time due to different fuel and engine technology , McClellan et al. 2012. However, the studies on DE and ambient air pollution particles have been carried out within the last decade and these exposures can be considered to have been caused by ''modern'' fuel combustion technology. Still, it should be recognized that new diesel engines, coming on roads today, emit little PM because of improved combustion technology and use of efficient particle filters.
The introductory section of this review will place PMgenerated CVD outcomes into a broader perspective with regard to classical risk factors of CVD. We have briefly summarized histopathological features of atherosclerosis and the association with oxidative stress and inflammation. The individual studies on PM-generated vascular effects are summarized in sections that have been segregated into ambient air pollution particles, DE (and DEP) and NMs. The studies on air pollution particles have been segregated according to the location of exposure because the particle characteristics (size distribution and chemical composition) differ considerably by region and by proximity to roadways. The subsequent sections ignore the compositional variation in ambient air PM chemical characteristics and integrate the studies in a collective analysis with focus on dose-response relationships and association among pulmonary inflammation, oxidative stress and CVD outcomes. This strategy means that, within each category, the comparison of exposures is inherently Figure 1. Accumulation of studies on authentic air pollution particles (including concentrated ambient air particles (CAPs)), diesel exhaust particles (including whole diesel exhaust) and nanomaterials [the total number of studies is higher than the number of publications because certain papers contain results on different types of PM].
different. However, we do believe that an attempt to bridge observations from different types of PM is meaningful for better understanding of effect of size and physicochemical characteristics that promote the development and exacerbation of CVD.

Cardiovascular risk factors
CVD constitutes an assortment of conditions that affect the heart and the blood vessels. The two main causes of mortality of CVD are IHD and ischemic stroke (i.e. occlusion of blood vessel in the brain) (GBD 2013 Mortality and Causes of Death Collaborators 2015). Interestingly, ischemic stroke and air pollution have received relatively little attention, although epidemiological studies indicate an association between acute exposure to air pollution particles and stroke (Wang et al. 2014, Shah et al. 2015. To the best of our knowledge, there are no studies on vasomotor dysfunction or atherosclerosis in brain arteries of animals that have been exposed to combustion-derived PM or NMs. Coronary heart disease (also called coronary artery disease) is caused by a build-up of plaques in arteries to the heart, which results in the narrowing of the lumen and reduced blood supply. This typically includes IHD (including myocardial infarction and angina pectoris) and hypertensive heart diseases (International Classification of Diseases 9 codes 402 and 410-414). There are numerous risk factors for coronary heart disease, which are interlinked in a highly complex causal pathway. These factors include obesity, physical inactivity, family history of disease, ethnicity, psychosocial factors, elevated blood lipids, hypertension, prothrombotic factors and inflammatory markers . In calculation of the absolute risk of coronary heart disease from the Framingham risk factors, advanced age (score ¼ 7-8) is a stronger risk factor than total cholesterol, hypertension, smoking and plasma glucose (individual scores up to 4) . Similar assessment of risk scores has been conducted in studies on fatal CVDs in European populations (Conroy et al. 2003). Obesity, hypertension, elevated levels of triglycerides, cholesterol and fasting glucose in plasma are also components in the definition of the metabolic syndrome, which is associated with increased risk for developing type 2 diabetes and CVD (Eckel et al. 2005). Table 1 outlines a stratification of risk factors for coronary heart disease in humans, which is driven by atherosclerosis, from observations in epidemiological studies. Certain risk factors have been explored in animal experimental models as intermediate steps in a presumed causal pathway of PM-generated CVD outcomes (e.g. plasma lipids, glucose, systemic inflammation and oxidative stress), whereas others have used animals with predisposition to CVD (e.g. hyperlipidemic or spontaneous hypertensive strains). In this review, we have abstracted results from these wellknown risk factors for coronary artery disease in humans, namely Framingham risk factors, including European equivalents, and risk factors for metabolic disease (outlined in Supplementary Tables 1-4 and discussed below). Plaque burden Promotes progression to more advanced plaques (plaque instability and risk of rupture)

Age
Air pollution exposure b a The classification is based on observations in humans as reviewed by Grundy 1999. b Based on observations of associations between exposure to traffic-related air pollution and sudden death of myocardial infarction (Peters et al. 2004).
Animal models used in studies on cardiovascular effects after PM exposure Table 2 lists general information about the most commonly used animal models in studies of PM-associated CVD outcomes. It is important to state that, within each species, susceptibility to CVD outcomes varies, therefore it is not logical to regard certain species as particularly susceptible to CVD. In addition, none of the non-human Table 2. Description of animal models in studies of association between exposure to particulate matter and cardiovascular outcomes a . Developed from C57BL/6 (black) Weight: normal (slightly lower than wild-type) Dyslipidemic (moderately high LDL levels) Develop spontaneously atherosclerosis in various vessels c Normotensive LDLr À/À mice 1993 Developed from C57BL/6 (black) Weight: normal (slightly lower than wild-type) Dyslipidemic (moderately high LDL levels) Hyperglycermia and impaired glucose tolerance (on Western-type diet) Develop only atherosclerosis in various vessels on high-fat diet c Normotensive a The information has been abstracted from webpages from suppliers and selected articles (Paigen et al. 1990, Doggrell and Brown 1998, Pinto et al. 1998, Yanni 2004, Aleixandre de Artinano and Castro 2009, Shiomi and Ito 2009, Kennedy et al. 2010, Getz and Reardon 2012. Body weights are reported for female (F) and male (M) animals at 10 weeks of age. b The development of high-fat diet induced obesity and atherosclerosis is less pronounced in BALB/c mice as compared to C57BL/6 mice. BALB/c mice also seem to be resistant to high-fat diet induced increases in plasma levels of glucose, insulin and triglycerides (Montgomery et al. 2013). c Plaque progression is typically assessed in the aortic arch or brachiocephalic artery, whereas the strain is resistant to coronary atherosclerosis. species are perfect models for CVD in humans. However, these animal models can be regarded as ideal, to some extent even ''clean'', models of specific mechanisms of human CVD. For instance, people at risk of CVD typically display several Framingham risk factors or descriptors of the metabolic syndrome, which makes it difficult to pinpoint the contribution of a single factor. The use of animals allows for many of these risk factors to be investigated separately as exemplified by studies on obesity (Aleixandre de Artinano and Castro 2009) and diabetes (King 2012).
The animal models can be segregated into strains that are ''resistant'' with regard to development of CVD, strains that are ''susceptible'' to development of CVD and ''disease models'' that have been bred or transgenically altered to develop risk factors of CVD. From these, transgenic mice and rabbits that spontaneously develop atherosclerotic plaques have been mainly used in particle toxicology. Atherosclerosis can be accelerated by utilization of an atherogenic diet containing a high content of cholesterol, lipids or a Western-type diet, although dyslipidemic animal strains also develop atherosclerosis over time at a slower rate on a normal diet (Getz and Reardon 2012). The most widely used experimental model has been the apolipoprotein E knockout mouse (ApoE À/À ) that develops atherosclerosis spontaneously, although this process is accelerated if the mice are fed a high-fat diet. ApoE is a component of lipoproteins, with the exception of low-density lipoprotein (LDL). It functions principally to clear chylomicrons and very low density lipoproteins from the circulation. ApoE also plays a role in macrophage biology, immune function and regulation of adipose tissue. In ApoE À/À mice, atherosclerosis develops in the aortic root as well as the brachiocephalic, carotid and pulmonary arteries. The morphology of these lesions tend to be more ''foamy'' when mice are fed a high-fat diet, whereas ApoE À/À mice on normal chow develop plaques with a more complex cellular morphology (Getz and Reardon 2012). LDL receptor (Ldlr À/À ) knockout mice are the most widely used murine alternative to the ApoE À/À mouse model. This mouse model has impaired lipoprotein clearance from the circulation, but it only develops atherosclerosis on a high-fat diet. In general, mouse models do not span the entire pathophysiology of atherosclerosis as they do not develop vulnerable plaques with risk of rupture, thrombosis and arterial occlusion. Thus, they may not be reliable models for predicting myocardial infarction in humans. A few studies have used rabbits in studies of PM-generated atherosclerosis. As a species, rabbits are sensitive to cholesterol overload (Yanni 2004). However, selective breeding of a mutant rabbit showing hypercholesterolemia has yielded the Watanabe heritable hyperlipidemic (WHHL) rabbit model that has a defect in LDLr, which is associated with very high LDL levels in plasma (Shiomi and Ito 2009). These rabbits develop atherosclerosis in the aorta as well as coronary, cerebral, carotid and pulmonary arteries. In addition, the occurrence of myocardial infarction in WHHL rabbits makes this model more representative of IHD in humans as compared with the murine knockout models. The majority of studies in the present review have investigated plaque progression in ApoE À/À mice (23 studies; Chen and Nadziejko 2005, Sun et al. 2005, Li et al. 2007, Araujo et al. 2008, Sun et al. 2008a, Ying et al. 2009a, Campen et al. 2010, Quan et al. 2010, Vesterdal et al. 2010, Bai et al. 2011, Kang et al. 2011, Mikkelsen et al. 2011, Cassee et al. 2012, Chen et al. 2013a,b, Lippmann et al. 2013, Miller et al. 2013, Pö ss et al. 2013, Vedal et al. 2013, Cao et al. 2014, Rao et al. 2014, Han et al. 2015, Keebaugh et al. 2015; while, three studies have used Ldlr À/À mice on high-fat diet (Niwa et al. 2007, Soares et al. 2009, Li et al. 2013b) and one study has utilized a double knockout for both ApoE and Ldlr (Chen and Nadziejko 2005). The rabbit models encompass one study on New Zealand white rabbits on high-fat diet (Miyata et al. 2013) and three studies on WHHL rabbits (Suwa et al. 2002, Goto et al. 2004, Yatera et al. 2008. The disease models develop certain phenotypes that potentiate intermediate steps in PM-generated CVD. As an example, it has been shown that young ApoE À/À on normal chow developed more pulmonary inflammation than wild-type C57BL/6 mice after intratracheal (i.t.) instillation of nanosized carbon black (Jacobsen et al. 2009). It has also been demonstrated that ApoE À/À on regular chow have higher rate of age-dependent oxidatively damaged DNA in the liver, indicative of a higher rate of oxidative stress as compared with wildtype mice (Folkmann et al. 2007). ApoE À/À mice had elevated number of cells, predominantly macrophages, tumor necrosis factor-a (TNF-a), macrophage inflammatory protein 1a (MIP-1a) and interferon-g (INF-g) in bronchoalveolar lavage fluid (BALF) after high-fat feeding as compared with normal chow (Naura et al. 2009). ApoE À/À mice also develop emphysema after only 10 weeks of Western-type diet feeding, whereas Ldlr À/À mice do not, which has been attributed to altered efflux of cholesterol in macrophages in the lung (Goldklang et al. 2012).
Spontaneous hypertensive rats have been the species of choice in studies of hypertension in animal experiments (Pinto et al. 1998). Male spontaneous hypertensive rats have been shown to produce a higher influx of neutrophils in BALF following i.t. instillation on 3 consecutive days to 1.6-40 mg/kg of fine particles from Shanghai, China (Cao et al. 2007). A more realistic mode of exposure by inhalation of residual oil fly ash (15 mg/ m 3 , 6 h/d) for 3 consecutive days elicited the same influx of neutrophils in BALF in spontaneous hypertensive rats and wild-type Wistar-Kyoto counterparts (Kodavanti et al. 2000). Likewise, inhalation of DE (DEP concentration in the mixture: 500 or 2000 mg/m 3 , 4 h/d for 4 weeks) was associated with the same extent of concentrationdependent increase in neutrophilic influx in BALF in spontaneous hypertensive and wild-type Wistar-Kyoto rats (Gottipolu et al. 2009). Thus spontaneous hypertensive rats may not be specifically sensitive to PM-induced airway inflammation as compared with wild-type rats.
In summary, several animal models have been developed to investigate CVD. The leading causes of CVD mortality in humans are IHD and ischemic stroke. Diminished oxygen supply to the myocardium in patients with IHD gives rise to clinical manifestations such as angina pectoris. In humans, CVD manifests and progresses as a pre-clinical and/or clinical period with a gradual increase of atherosclerosis, endothelial dysfunction, arterial stiffness and hypertension before deterioration and eventual death. Importantly, in transgenic animals, only a number of these manifestations can be mimicked. Thus, these animal models are in all probability not entirely reliable for predicting risk of myocardial infarction in humans. This being said, these animals are still useful for hazard identification of CVD and for comparison of the effect caused by different types of particles. The latter may pertain to vasomotor dysfunction, progression of atherosclerosis or histological composition of plaques. However, once again it should be stated that animals are not perfect models for the pathophysiology or histopathology of IHD or ischemic stroke in humans.

Progression of atherosclerosis
The development and progression of atherosclerosis is the hallmark of coronary heart disease in humans. In animals, it is possible to assess the extent of atherosclerosis with ultrasound imaging, although direct detection in isolated blood vessels is currently the most common method. Atherosclerosis is typically assessed and visualized en face as the area of a blood vessel (visible as white areas). Staining with Sudan IV or Oil Red O is typically used for visualization of lipids in the plaques. The extent of atherosclerosis can also be assessed by histology in cross-section of a blood vessel as the percentage of the lumen that is occupied by the plaque.
The process of atherosclerosis is typically divided into stages based on the morphology of the plaques. Table 3 outlines the American Heart Association's (AHA) classification of disease progression in human coronary arteries and aorta and the corresponding characteristics that are observed in animal models. The AHA lesion types I through to III are regarded as ''early'' precursors to more advanced lesions and most typically present in children's arteries. Lesions I-III mostly affect the intima by the accumulation of extracellular lipids, whereas the adjacent media and adventitia are unaffected. Nevertheless, lesions I-III are regarded as clinically silent, although in these cases there are still increasing levels of cytokines produced, vasoconstriction and decreased endothelium-dependent vasodilation (Stary et al. 1994). Moreover, extracellular deposition of calcium in the arterial wall is intimately related to hypertension, while vascular calcification in the intima is associated with atherosclerosis, whereas medial calcification is typically related to patients with diabetes and renal failure (Kalra and Shanahan 2012). In more advanced lesions, plaques are classified as atheroma (type IV), fibroatheoroma (type V) and complicated (type VI) and produce an increasing narrowing of lumen, obstruction and clinical manifestations.
It is possible to mimic a range of manifestations of atherosclerosis found in humans in transgenically altered murine models, if not necessarily a perfect histopathological match of plaques in humans. WHHL rabbits develop unstable plaques and may, therefore, be a more relevant model for myocardial infarction in humans. However, rabbits are less desirable as experimental model as they are more expensive with regard to housing costs and maintenance. Typically, it takes 11-15 months to develop atheromas in WHHL rabbits, whereas it only takes 14-16 weeks in ApoE À/À mice on high-fat diet. In addition, the studies in rabbits require much more PM material due to the larger body mass (100times higher than mice). This may restrict rabbits as an experimental model in studies of NMs due to limited available quantities of well-characterized material.
In summary, several animal models have been developed to investigate the histopathological condition related to CVD in humans. Atherosclerosis in coronary arteries is the most important cause of IHD and myocardial infarction in humans. Transgenically altered murine models do not develop atherosclerosis in coronary arteries and atheromas do not progress to complicated lesions that are more vulnerable to rupture. However, WHHL rabbits do develop coronary atherosclerosis and complicated plaques, but due to financial constraints, long periods of time required for development of disease and large quantities of PM material required for experiments make these animals a less attractive experimental model as compared with transgenically altered murine.

Inflammation and oxidative stress
In order to explain the association between CVD mortality and air pollution exposure, hypotheses from the early 2000s suggested that pulmonary oxidative stress and inflammation would instigate the release of cytokines (and oxidation products such as oxidized LDL), promote atherosclerosis, autonomic dysregulation and increase blood coagulability (Pope III 2000, Donaldson et al. 2001. The hypothesis hinges on the dogma that inhalation of particles induces pulmonary inflammation that ''spillover'' to the circulation or increase blood levels of signaling factors. In support of this hypothesis, an early study in rats showed that pulmonary exposure to residual oil fly ash was associated with increased plasma levels of fibrinogen (Gardner et al. 2000). However, it is interesting to note that a commentary to this paper highlighted that CVD outcomes have been demonstrated without pulmonary inflammation or injury in healthy individuals (Gordon and Reibman 2000). The systemic inflammation hypothesis, including oxidative stress, has subsequently been scrutinized in animal experimental models as well as studies employing molecular epidemiology in humans. There appears to be different opinions on the importance of systemic inflammation and oxidative stress in development and progression of CVD following exposure of PM. Previously published reviews have discussed systemic inflammation as a link between exposure and CVD outcomes (Campen et al. 2012, Mann et al. 2012, Stapleton et al. 2012b. One review specifically focused on immune responses by macrophages as an important mechanism for PMinduced CVD outcome (Miyata and van Eeden 2011). In another review, progression of atherosclerosis was correlated with pro-oxidant and pro-inflammatory responses as well as the importance of the small size of ultrafine particles (UFPs) and their increasing toxicity in comparison to larger particles (Araujo and Nel 2009). However, Lippmann et al. (2005) argued that long-term Table 3. Types of plaques in human and animal arteries a .
inhalation exposure to concentrated ambient air particles (CAPs) is not associated with persistent pulmonary inflammation due to tolerance to the exposure. Later, Lippmann and Chen (2009) concluded that acute exposure to CAPs was associated with pulmonary inflammation in humans, dogs and rats. The review did not, however, mention pulmonary inflammation in CAPs-exposed mice, which is somewhat puzzling because the authors had conducted several studies on inhalation exposure to CAPs in this species, including atherosclerosis-prone knockout mice. This may be explained by the absence of a positive effect (or conflicting data) or pulmonary inflammation not being investigated as an outcome. In an updated review by the one of the authors, a number of studies on systemic inflammation and oxidative stress were summarized, but there was no definitive conclusion concerning whether the accumulated data provide a convincing association between pulmonary inflammation and CVD outcomes (Lippmann 2014). It should also be emphasized that there are more skeptical views concerning the role of systemic inflammation as a mechanistic link between inhalation of PM and CVD outcomes (Miller 2014). It is our belief that systemic inflammation, i.e. elevated levels of pro-inflammatory cytokines, cannot explain the CVD outcomes that are observed at low-dose exposures to PM. However, there seems to be some evidence linking local inflammation and oxidative stress in the vessel or plaques/atheromas with vasomotor dysfunction and progression of atherosclerosis.

Qualitative description of studies on vasomotor function and progression of atherosclerosis
Within the context of this review, the studies are described based on the mass dose or concentration because it is the only descriptor that can be compared across all publications. Other characteristics of particles include the particle size distribution, shape, agglomeration and chemical components. The latter may be used for source appointment in air pollution studies, descriptor of impurities in NMs or identification of active constituents in particles. The particle size is an important predictor of toxicity of particles, but the size distribution may only be reported for the dry form of the sample (e.g. from electron microscopy). The most important descriptors of particle size, including the size distribution, are count median aerodynamic diameter (CMAD) and mass median aerodynamic diameter (MMAD) for inhalation studies. The hydrodynamic diameter in the exposure vehicle is the most important descriptor of particles that have been administered in a suspension. Information on the particle size range is typically reported as the standard deviation (SD) or geometric SD (GSD), assuming a mono-modal distribution. Certain studies have reported the particle size of the peak number concentrations in multi-modal distributions of PM (Supplementary Tables 1-4 contains information on particle characteristics in all studies highlighted in the review).

Vascular effects after exposure to authentic air pollution
To date, very few studies have investigated the direct effect of authentic air pollution exposure without focus on particulate fractions. This exposure is only applicable in locations with high emissions of PM, i.e. sourced from major roads in urban areas.

Beijing, China
Male ApoE À/À on Western-type diet were exposed for 2 months to air from a major road (average PM 2.5 concentration: 61 mg/m 3 (exposure group) and 18 mg/m 3 (filtered air group), exposure period 18 January-18 March 2010). This exposure was associated with elevated levels of interleukin-6 (IL-6) and TNF-a in the BALF, systemic inflammation (C-reactive protein (CRP) and IL-6) and oxidative stress (decreased superoxide dismutase activity and increased oxidized LDL), whereas plasma triglycerides and high-density lipoprotein (HDL) levels were unaltered. There was also increased aortic plaque area in mice that were exposed to air pollution as compared to filtered air (Chen et al. 2013b).

Sao Paulo, Brazil
Male LDLr À/À mice on either normal or high-fat diet were exposed to air pollution from a busy traffic intersection for 4 months (average PM 2.5 concentration: 20.4 mg/m 3 (exposure group) and 1.6 mg/m 3 (filtered air group)). This exposure did not affect the levels of cholesterol and triglycerides in plasma and there was no increased plaque progression in the aorta, whereas high-fat feeding and PM exposure were associated with increased arterial wall thickness (Soares et al. 2009). This indicates that the exposure was associated with vascular remodeling rather than increased plaque size.
Vascular effects after exposure to concentrated ambient air particles (CAPs) The exposure to CAPs in animal models has been used mainly in studies from USA. These exposures mainly focus on particles in the 0.1-2.5 mm range without affecting the composition (Sioutas et al. 1995(Sioutas et al. ,1997. The health effects of inhalation exposure to CAPs in both humans and experimental animals have been summarized in previous reviews, showing consistency in effects across species (Lippmann and Chen 2009) and a coherence between effects in animal experimental models and risks of CVD from the same locations in epidemiological studies (Lippmann 2014). It seems that pulmonary inflammation has not been routinely assessed or even discussed as a possible mediator of CVD outcomes in the majority of studies on CAPs exposure. There appears to be an inconsistency in observations of pulmonary inflammation following inhalation of CAPs (Ghio and Huang 2004). This may be contributed to the differences between studies in terms of the concentration of CAPs in the exposure chamber and duration of the exposure. In these experiments, a range of concentrations between 100 and 500 mg/m 3 for several hours per day during several months were utilized. It has been argued that lowcytotoxicity particles should not evoke pulmonary inflammation at concentrations below 1000 mg/m 3 (Oberdö rster 1995). On a similar note, it has been shown that the surface area of urban air particles ranged from 150 mm 2 /cm 3 of air next to a freeway to 50-70 mm 2 / cm 3 at urban background sites in Los Angeles, CA (Ntziachristos et al. 2007). Assuming that rats inhale 50 Â 10 3 cm 3 air during a single 6-h session, the total deposited surface area of particles will be 0.005-0.015 cm 2 (20% deposition). It has been argued that infiltration of polymorphonuclear neutrophils (PMNs) in BALF occurs at a lung burden particle surface of 200-300 cm 2 (Tran et al. 2000). As the surface area in CAPs is dominated by small particles, it can be speculated that the inhalation exposure per se has insufficient particle surface area to elicit pulmonary inflammation. In general, it is very difficult to state with any certainty if in CAPs inhalation studies sufficiently large doses have been administered to promote pulmonary inflammation (level of UFPs, mass concentration (i.e. dominated by the non-UFP fraction), exposure frequency and duration). Nevertheless, it is our impression that in the majority of CAPs exposure studies highlighted in this review doses utilized were not high enough to evoke pulmonary inflammation. This is extremely important as a substantial number of CAPs inhalation studies have shown CVD outcomes, which largely can be ascribed to mechanisms that are not secondary events to pulmonary inflammation. Below, the text has been organized according to the location of CAPs exposure in a similar way as that used by Lippmann and Chen (2009).

Tuxedo, NY
The publications from this exposure location contain little information on the characterization of CAPs, although an earlier publication described the experimental setup and showed that the particle number size distribution was practically identical between the CAPs aerosol (135 ± 2 nm) and ambient air (123 ± 12) (Maciejczyk et al. 2005). The first study showed that 6 months exposure (85 mg/m 3 , 6 h/d, 5 d/week) was associated with increased plaque progression in aorta and reduced responsiveness to ACH-induced vasorelaxation and increased response to PE or serotonin vasoconstriction in ApoE À/À mice on high-fat diet, whereas there was no effect in mice on regular chow (Sun et al. 2005). These effects were accompanied by signs of inflammation and oxidative stress in the aorta. Plasma lipid profiles showed unaltered cholesterol and triglyceride levels in filtered air versus CAPs exposed mice (Sun et al. 2005). In a concordant study with similar exposure size and duration, increased plaque size in the aortic arch was noted by ultrasound imaging in CAPs exposed ApoE À/À mice on high-fat diet, whereas the difference was not statistically significant in mice on normal chow (Sun et al. 2008a). Inhalation of CAPs (110 mg/m 3 , 6 h/d, 5 d/week for 5 months) was associated with increased plaque progression in aorta of ApoE À/À mice on high-fat diet, whereas there was no difference in plaque area, assessed by en face staining of aorta with Sudan IV, in ApoE and LDLr double knockout mice (Chen and Nadziejko 2005). The CAPs exposure was not associated with pulmonary inflammation (Lippmann et al. 2005).
A more recent study assessed plaque progression at either 3 or 5 months exposure (105 mg/m 3 , 5 h/d, 4 d/ week) in ApoE À/À mice on normal chow (Quan et al. 2010). The characterization of the aerosol showed a CMAD of 80 nm (GSD ¼ 1.9 nm), whereas the MMAD was 223 nm (GSD ¼ 1.6 nm). First, en face measurement of aorta with Sudan IV staining showed increased plaque progression at 5 months of exposure. The assessment of plaque progression in brachiocephalic artery (BCA) by ultrasound imaging of lumen size and hematoxylin and eosin (H&E) stain of cross-sections also indicated increased plaque progression, especially after 5 months of exposure. The determination of vasomotor function in aorta showed increased PE-induced vasoconstriction, whereas there was an unaltered response to serotonin, ACH and SNP. In addition, the exposure did not cause pulmonary inflammation (i.e. unchanged levels of PMNs in BALF) and IL-6 and IL-10 levels were unaltered in serum (Quan et al. 2010).
Inhalation of CAPs (73 mg/m 3 , 6 h/d, 5 d/week for 128 d) in C57BL/6 mice on high-fat diet was associated with reduced vasorelaxation response to ACH and insulin in aorta rings, which occurred concomitantly with systemic inflammation (TNF-a and IL-6), increased LDL and reduced HDL levels, whereas there were unaltered cholesterol and triglyceride levels in plasma (Sun et al. 2009). This study also showed adipose tissue inflammation after CAPs inhalation exposure (Sun et al. 2009). Additionally, Sprague-Dawley rats exposed to CAPs (79 mg/m 3 , 6 h/d, 5 d/week for 10 weeks, with angiotensin II infusion at the last week) displayed increased PEinduced vasoconstriction in aorta rings, which was prevented by addition of a Rho-associated protein kinase to PE-precontracted aorta rings ex vivo (Sun et al. 2008b). There was also decreased ACHresponse in aorta rings and vasomotor dysfunction was accompanied by an increased expression of NADPH oxidase subunits (Sun et al. 2008b).

New York City, NY
ApoE À/À mice on high-fat diet were exposed to CAPs for 4 months in a laboratory setting located on the fourth floor above a busy street (138 mg/m 3 , 6 h/d, 5 d/week). The exposure was associated with reduced PE-induced vasoconstriction in aorta rings, decreased responsiveness to SNP-induced vasodilatation and abolished vasorelaxation response to a Ca 2+ ionophore (A23187), whereas there was no effect on ACH responsiveness (Ying et al. 2009a). The same study showed a strong plaque burden increase in the aorta (35.4% and 14.8% plaque progression in cross-section of vessels from exposed and controls, respectively), which was accompanied by increased macrophage infiltration, expression of NADPH oxidase subunits, collagen deposition and triglycerides build up in the plaques (Ying et al. 2009a).

Five-city study
In a very interesting study by the National Particle Component Toxicity (NPACT), plaque progression after inhalation exposure to CAPs in Manhattan (NY), Tuxedo (NY), East Lansing (MI), Seattle (WA) and Irvine (CA) was assessed. Initially, ApoE À/À mice on normal chow were exposed to CAPs in Manhattan (NY), Tuxedo (NY) for 3 or 6 months (6 h/d, 5 d/week) and atherosclerotic progression assessed by ultrasound imaging of the left common carotid artery and BCA. There was accelerated plaque progression in both arteries and most pronounced at 6 months of exposure (highest effect was 47% and 23% increased plaque area in the left common carotid artery and BCA, respectively). Subsequently, plaque progression was investigated by ultrasound imaging in ApoE À/À mice in East Lansing, Seattle and Irvine. These mice were exposed to CAPs for 6 months with interim assessment of atherosclerotic progression at 2 and 4 months. The authors noted a slightly increased plaque progression in BCA of mice that had been exposed to CAPs in East Lansing (19% increased plaque progression as compared to filtered air control), while a small statistically significant increase in plaque progression was observed after 2 months CAPs exposure in Irvine, but this was not observed at 4 and 6 months. The study concluded that CAPs inhalation exposure at Manhattan (123 mg/m 3 ), Tuxedo (136 mg/m 3 ) and East Lansing (68 mg/m 3 ) was associated with increased plaque progression, whereas there was no effect in Seattle (61 mg/m 3 ) and Irvine (138 mg/m 3 ). The lack of effect in Seattle and Irvine was confirmed by en face Sudan IV staining of thoracic aorta segments (data from en face staining of aorta were not reported for mice in other cities). The assessment of inflammation markers (CRP, IL-6, IL-10, IL-12, IL-13, TNF-a, vascular endothelial growth factor and granulocyte macrophage stimulating factor) showed inconsistency, therefore, it was concluded that there was no systemic inflammation (Lippmann et al. 2013). The difference in atherosclerotic response could not be attributed to specific PM 2.5 components or the mass concentration across exposure locations.

Los Angeles, CA
ApoE À/À mice were exposed in a whole-body chamber to atmospheres containing UF-CAPs (i.e. less than 0.18 mm) or CAPs (i.e. less than 2.5 mm) in a mobile inhalation laboratory located 300 m from a freeway (Araujo et al. 2008). Inhalation of UF-CAPs (113 mg/m 3 , 5 h/d, 3 d/week for 40 d) was associated with increased plaque progression in the aortic root of male ApoE À/À mice on regular diet, whereas the same exposure to fine CAPs (438 mg/m 3 ) had no effect (Araujo et al. 2008). The particle number concentration (6.6 Â 10 5 versus 4.6 Â 10 5 particles/cm 3 ) and organic carbon content (52% versus 25%) was higher in the UF-CAPs exposure as compared to fine CAPs. Although no results were presented, the authors noted that the level of inflammatory cells in BALF were unaltered in these mice, while increased plasma levels of total cholesterol were observed in the group of mice that had been exposed to CAPs. Thus, a proatherogenic plasma lipid response rather than pulmonary inflammation seems to explain the accelerated plaque progression.
In a later study, UFPs were collected from an area with close proximity to major freeways and subsequently reaerosolized for inhalation experiments (CMAD &50 nm). LDLr À/À mice on high-fat diet were exposed to 360 mg/m 3 , 5 h/d, 3 d/week of UFPs for 10 weeks. This exposure resulted in increased plaque area in the aorta, which was ameliorated by inhibition of nuclear factor-kB activity (NF-kB). The same study also showed systemic inflammation (increased serum amyloid A and TNF levels) and elevated levels of triglycerides in plasma and lower levels of HDL, whereas there were unaltered levels of LDL and cholesterol (Li et al. 2013b). In a parallel study using the same type of re-aerosolized UFPs, concentration and exposure duration increased calcification of atherosclerotic plaques in aortic root sections, concurrently with increased NF-kB staining in LDLr À/À mice was noted (Li et al. 2013a). The data suggest that the exposure to UFPs was associated with both increased plaque area and progression of plaques to advanced stages, although it should be emphasized that there is uncertainty about whether the findings originate from the same study as cross references are missing.
The inhalation of UF-CAPs sourced from a location close to heavy traffic (58 mg/m 3 , 5 h/d, 4 d/week for 8 weeks) was associated with increased plaque area in BCA in male ApoE À/À mice (Keebaugh et al. 2015). In this study, one group of mice was exposed to UF-CAPs after removal of organic compounds using a thermal denuder. The removal of the organic components reduced the mass of organic carbon from 25 to 9 mg/cm 3 , whereas the overall mass of metals and other trace elements was not affected. The removal of organic components rendered the UF-CAPs non-atherogenic after inhalation exposure in ApoE À/À mice (Keebaugh et al. 2015).

Columbus, OH
The publications from this exposure location have little information about particle characteristics, except for some data on the chemical composition of PM. There seems to be some variation in the chemical composition of PM used in different studies. For instance, the concentration of iron and copper display a three-fold difference between publications and contain little discussion about the role of the chemical composition of PM for the observed effects on vascular endpoints. The inhalation exposure to CAPs (74 mg/m 3 , 6 h/d, 5 d/week for 12 weeks) in wild-type C57BL/6, which were treated with angiotensin II for the last 2 weeks of exposure, resulted in increased PE-induced vasoconstriction in aorta rings (Ying et al. 2009b). This effect was blunted when PEprecontracted aortic rings were treated ex vivo with a RhoA/Rho-kinase inhibitor, indicating that the augmented vasoconstriction was mediated by Rho-associated protein kinase signaling (Ying et al. 2009b). In the same location, CAPs inhalation exposure (111 mg/m 3 , 6 h/ d, 5 d/week for 10 weeks) in wild-type C57BL/6 on either normal or high-fat diet showed increased PE-induced vasoconstriction and decreased ACH-induced vasorelaxation response in aortic rings, whereas there was no effect in NAD(P)H p47 phox knockout mice (Xu et al. 2010, Liu et al. 2014. Insulin-induced vasorelaxation was blunted by CAPs exposure in both strains of mice. Nevertheless, systemic inflammation in terms of TNF-a plasma levels was only observed in wild-type animals and this strain also displayed insulin resistance after exposure to CAPs (Xu et al. 2010, Liu et al. 2014. The exposure to CAPs from the same location, (101.6 mg/m 3 , 6 h/d, 5 d/week for 6 months) was also associated with increased plaque progression in aorta of ApoE À/À mice on regular diet (Rao et al. 2014). In another study, 6 months exposure (107 mg/m 3 , 6 h/d, 5 d/week) increased the vasoconstriction response to PE or U-46619 (synthetic analog of prostaglandin H 2 ), and decreased ACH-induced vasorelaxation in mesenteric arteries of C57BL/6 mice on regular diet (Ying et al. 2014). CAPs exposure to BALB/c mice (92.4 mg/m 3 , 6 h/d, 5 d/week) for 20 weeks increased PE-induced vasoconstriction and reduced ACH-induced vasorelaxation in aorta rings (Kampfrath et al. 2011). There were increased levels of TNF-a and MCP-1 in lung tissue and serum of CAPs exposed mice. Toll-like receptor 4deficient mice did not develop pulmonary or systemic inflammation or vasomotor dysfunction following inhalation of CAPs. In the same study, exposure to CAPs increased PE-induced vasoconstriction in the aorta of C57BL/6. However, NOS À/À mice with the same genetic background as C57BL/6 mice did not develop PE-induced vasoconstriction. There was unaltered ACH-induced vasorelaxation in aorta of wild-type C57BL/6 and NOS À/ À mice. The systemic inflammation response was deemed to be mediated by oxidized derivates of 1-palmitoyl-2arachidonyl-sn-glycero-3-phosphorylcholine in BALF, which triggered the systemic inflammation via Toll-like receptor 4 and NADPH oxidase activation (Kampfrath et al. 2011).
In a final study from Columbus, 15 weeks of exposure to CAPs (128 mg/m 3 , 6 h/d and 5 d/week) increased vasoconstriction in aorta rings by PE and U-46199 and reduced ACH-induced vasorelaxation in spontaneous hypertensive rats (Ying et al. 2015). These rats had markedly increased expression levels of TNF-and IL-6 in lung tissue (approximately 40-fold increased), which was persistent (50-fold increased, TNF-a) or reduced (8-fold, IL-6) after a 5 week recovery period that was associated with a normalization of vasomotor function (Ying et al. 2015).

Sao Paulo, Brazil
The exposure of male Wistar rats to CAPs for 14 consecutive days (593 mg/m 3 , 3.3 h daily exposure on average) was associated with a blunted ACH-induced vasorelaxation response and unaltered SNP-response in pulmonary arteries. The exposure also increased TNF-a and oxidative stress (dihydroethidium staining) levels in vessel segments, whereas there were unchanged plasma levels of TNF-a, IL-1b or IL-6 (Davel et al. 2012).

Utrecht, The Netherlands
In the set of trials male Fischer rats were exposed at a roadside location (485 mg/m 3 , 5 d/week, 6 h/d) for 4 weeks to CAPs with a MMAD of 104 nm (GSD ¼ 30 nm). Prior to the CAPs exposure, the rats were exposed to ozone (0.4 ppm for 12 h), to induce mild pulmonary inflammation (Gerlofs-Nijland et al. 2010). It was stated that the CAPs exposure did not generate pulmonary inflammation; while no significant alteration in PEinduced vasoconstriction or vasorelaxation responses to ACH, SNP and isoprenaline in aortic rings was noted (Gerlofs-Nijland et al. 2010).

Vascular effects after exposure to particulate matter in urban air
A number of studies have investigated CVD outcomes after bolus administration of PM into the airways by either i.t. instillation or oropharyngeal aspiration. This mode of administration is non-physiological, but the PM samples are typically better characterized than real-time urban air or CAPs exposure. In addition, some of these samples can be purchased from the National Institute of Standards and Technology as a standard reference material (SRM) or they have been sourced in ''large'' quantities and available to independent researchers.

SRM1648
This sample, also referred to as ''urban particulate matter'', was collected in St. Louis, MO, in a baghouse dust collector over a 12 month period between 1976 and 1977. I.t. instillation of 5 mg of SRM1648 to male Wistar rats was associated with reduced ACH response in intrapulmonary arteries at 12 h post-exposure, whereas there was no effect at 6, 24 or 72 h (Courtois et al. 2008). Likewise, ex vivo exposure of rat pulmonary arteries to the material was associated with vasomotor dysfunction at concentrations between 1 and 100 mg/ml (Li et al. 2005).

SRM1649b
This sample, also referred to as ''urban dust'', was collected in Washington, DC, in a baghouse over a 12-month period between 1976 and 1977. In an experiment, ApoE À/À mice were exposed to SRM1649b by i.t. instillation (0.5 mg/kg twice during 24 h, total dose ¼ 1 mg/kg -hydrodynamic particle size in saline was 224 nm (SD ¼ 44 nm)) with no effect on ACH-induced vasorelaxation in aortic rings or inflammatory response in terms of Ccl2 expression in lung tissue (Vesterdal et al. 2014).

EHC-93
This was recovered from baghouse videlon filters at the Environmental Health Center in Ottawa, Canada, in 1993. EHC-93 is described as PM 10 and is similar to SRM1649 and PM 2.5 samples from the Great Lakes in 1992 (Vincent et al. 1997). The publications on vascular effects describe EHC-93 as having a mean diameter of 0.8 mm (SD ¼ 0.4 mm) with reference to previously published work which described the CMAD (0.35 mm, GSD ¼ 1.7 mm) and MMAD (4.6 mm, GSD ¼ 2.3 mm) (Vincent et al. 1997). Subsequent studies on vascular effects have administered EHC-93 by bolus instillation in animals and the hydrodynamic diameter has not been reported in the suspension vehicle.
Next, the exposure of female WHHL rabbits by intrapharyngeal instillation (5 mg twice weekly for 4 weeks) was associated with increased plaque progression in the aorta and coronary arteries, whereas there was no difference in total cholesterol, HDL and LDL in plasma (Suwa et al. 2002). The scoring of the plaque morphology according to the AHA guidelines indicated a progression from fatty streak lesions (type II) to plaques with extracellular lipids and fibrotic layers (type III) (Suwa et al. 2002). A subsequent study also showed increased plaque progression in the aorta of WHHL rabbits (29% versus 20.9%), although this did not reach statistical significance as compared with unexposed rabbits (Goto et al. 2004). Nevertheless, in another study of EHC-93 exposure in WHHL rabbits, plaque progression in the aorta increased and associated with higher expression of cell adhesion proteins in plaques and higher attachment propensity of monocytes to the endothelium overlaying the arteriosclerotic plaques (Yatera et al. 2008). This indicates that the EHC-93 exposure was associated with accelerated recruitment of monocytes into plaques.
The administration of an inhibitor of 3-hydroxy-3methylglutaryl-coenzyme A reductase (i.e. Lovastatin) reduced the complexity of atherosclerotic lesions in EHC-93 exposed (1 mg/kg, 3 times per week for 4 weeks) New Zealand White rabbits on high-fat diet that were subjected to balloon injury in the abdominal aorta (Miyata et al. 2013). The EHC-93 exposed rabbits displayed increased PE-induced vasoconstriction and reduced ACH-induced vasorelaxation in carotid arteries, whereas the SNP response was unaltered. In the same animals, EHC-93 exposure was associated with increased plasma endothelin 1 (ET-1) levels, whereas cholesterol, triglyceride, HDL and LDL levels were unchanged (Miyata et al. 2013). Lovastatin administration abolished the observed vasomotor dysfunction and reduced ET-1 levels as well as blunted inflammation responses in atherosclerotic lesions. Similarly, utilization of female New Zealand White rabbits showed that exposure by intrapharyngeal instillation (2.6 mg/kg every second day for 5 d or 2.0 mg/kg twice weekly for 4 weeks) was associated with a reduced vasorelaxation response to ACH in carotid arteries, whereas the SNP and PE responses were unaltered (Tamagawa et al. 2008). Additionally, increased pulmonary inflammation, predominantly characterized by the infiltration of macrophages and elevated plasma levels of IL-6, but not ET-1, was also noted (Tamagawa et al. 2008). The role of IL-6 in vasomotor dysfunction was further documented in IL-6 knockout mice, which were not affected by i.t. instillation of EHC-93 (10 or 200 mg), whereas wild-type C57BL/6 mice had reduced responsiveness to ACH-induced vasorelaxation in abdominal aorta rings (Kido et al. 2011b).
Furthermore, studies of vasomotor function in spontaneous hypertensive rats showed that exposure via a single i.t. instillation of EHC-93 (10 mg/kg) was associated with increased vasorelaxation response to ACH and SNP in aorta rings at 4 h post-exposure, whereas there was no effect at 24 h and the PE-induced vasoconstriction response was not affected at both post-exposure times (Bagate et al. 2004b). The same authors also showed that ex vivo exposure to EHC-93 (10-100 mg/ml) in the aorta and mesenteric artery segments from wild-type Wistar Kyoto rats, and aorta segments from spontaneous hypertensive rats was associated with vasomotor dysfunction (Bagate et al. 2004a(Bagate et al. , 2006.

Chapel Hill, NC
Particles less than 150 nm in diameter were collected and described as UFPs. UFPs were collected on filters over 7-d periods in October 2002 (source of PM was not specified). The particles were extracted from the filters by sonication in aqueous solution and lysophilized before suspension in sterile water and administered to ICR mice by i.t. instillation (100 mg/mouse). This exposure resulted in pulmonary inflammation and impaired ACHinduced vasorelaxation in the aorta, whereas the PE-induced vasoconstriction response was unaltered (Cozzi et al. 2006).

Jinchang and Zhangye, China
PM 2.5 was collected simultaneously in a period from 6 March 2009 to 26 March 2010 from a city with a nickel refinery (Jinchang) and a ''control'' city (Zhangye). In these sets of trials, vasomotor function was assessed after repeated exposure (50 mg/mouse, twice/week for 3 weeks by oropharyngeal aspiration) in male FVB/N mice. The exposure to PM 2.5 from Jinchang decreased the responsiveness to ACH in mesenteric arteries, whereas there was no difference in SNP-induced vasorelaxation and PE-induced vasoconstriction. The ACH response was restored by ex vivo addition of a NO synthase inhibitor (L-NAME) or NADPH oxidase inhibitors (apocynin or VAS2870) to vessel rings. Additionally, gene expression analysis of mesenteric arteries showed a larger response to PM 2.5 from Jinchang (increased TNF-, IL-6, superoxide dismutase and NADPH oxidase 4), whereas serum levels of cytokines remained unaltered with no significant difference between PM 2.5 exposure from the two cities. The exposure to PM 2.5 from both cities also evoked pulmonary inflammation and toxicity (increased protein content in BALF). The general impression from these results is that pulmonary and systemic inflammation could not explain the difference in vasomotor dysfunction between PM 2.5 from these two cities (Cuevas et al. 2015).

Summary of studies on exposure to ambient air pollution particles
A review of the literature demonstrated that 17 out of 19 studies on CAPs originate from cities located in USA. All studies on CAPs inhalation exposure from USA have shown vascular effects in terms of vasomotor dysfunction and progression of atherosclerosis. However, the majority of the CAPs exposure studies have only used a single dose, with the exception comparing CAPs exposure at different sites within multiple concentrations (Lippmann et al. 2013). There are only two studies on vascular effects in animals after real-life exposure to air pollution and these have shown mixed results. Importantly, 10 out of 12 studies have reported vascular effects after instillation of urban air PM in the airways. The studies have demonstrated both vasomotor dysfunction and progression of atherosclerosis in animals after exposure to ambient air pollution particles. There is consistent evidence showing CVD outcomes in mice, rats (only vasomotor dysfunction) and rabbits following exposure to ambient air pollution particles.

Vascular effects after exposure to diesel exhaust
A number of studies have investigated CVD outcomes after inhalation exposure of DE to mice or rats. These studies are segregated into the type of diesel engine used, whereas the impact of engine type or fuel has not been assessed.

Single cylinder Yanmar
No significant alteration in responsiveness to PE, ACH and SNP in aorta rings of male C57BL/6 mice was noted following material exposure at doses of 300 mg/m 3 for 6 h (Weldy et al. 2013). The short-term exposure to DE (300 mg/m 3 for 5 h) in Sprague-Dawley rats was associated with increased ET-1 induced vasoconstriction response in coronary arteries, whereas plasma levels of ET-1 and cytokines were unaffected (Cherng et al. 2009). The same exposure also increased ROS levels and reduced ACH-induced vasorelaxation in coronary arteries with effects abrogated by supplementation with sepiapterin (precursor of tetrahydrobiopterin) and inhibition of nitric oxide synthase (Cherng et al. 2011). The short-term exposure to DE (350 mg/m 3 for 4 h) augmented vasoconstriction response to ET-1 and reduced vasorelaxation response to NONOate (a nitric oxide donor) in pressurized mesenteric arteries from male C57BL/6 mice (Knuckles et al. 2008). The same group also showed that ex vivo exposure of mice coronary arteries to DE was associated with an increased vasoconstriction response to ET-1 and decreased responsiveness to SNP-induced vasorelaxation (Campen et al. 2005). Moreover, ex vivo exposures resulted in ACH-induced vasorelaxation which was blunted when aorta segments from naïve mice were incubated with plasma from ApoE À/À mice, exposed to combined gasoline and diesel engine exhaust (300 mg/ m 3 ) for 50 d (Campen et al. 2014). Interestingly, this exposure did not affect protein and lactate dehydrogenase (LDH) levels in BALF, whereas there were augmented responses in terms of macrophage infiltration, superoxide radical generation and matrix metalloproteinase expression in the aorta (Campen et al. 2014). These observations indicate that pulmonary inflammation was not the principal driver of vascular effects.
A study on DE inhalation (436 mg/m 3 , 5 h/d, 4 d/week) assessed plaque progression at either 3 or 5 months exposure in ApoE À/À mice on normal chow (Quan et al. 2010). First, en face measurement of atherosclerosis in aorta with SudanIV staining showed increased plaque progression. Ultrasound imaging demonstrated unaltered lumen size in BCA, whereas there was a marginal decreased lumen size in BCA when assessed by H&E stain of cross-sections at 5 months of exposure. The evaluation of vasomotor function in aorta showed increased PE-induced vasoconstriction, whereas there were unaltered responses to serotonin, ACH and SNP. In addition, the exposure did not cause pulmonary inflammation (i.e. PMNs in BALF) and IL-6 and IL-10 levels were unaltered in serum (Quan et al. 2010). The study also showed that inhalation of DE gases had similar effect as whole DE on pulmonary/systemic inflammation and vasomotor function, whereas whole DE had a more pronounced effect on plaque progression as compared with DE gases.

Cummings
The exposure to DE (109, 305 or 1012 mg/m 3 ) with a MMAD of 100-150 nm for 50 consecutive days in male ApoE À/À mice on high-fat diet did not increase the plaque area in the aortic leaflet region, whereas there was increased macrophage and collagen staining, suggesting that the morphology of plaques was driven toward an advanced and fragile stage (Campen et al. 2010). The same exposure was also used in a study on mixed vehicle exhaust from gasoline and diesel combustion (83.3% of the emission was DE). ApoE À/À mice on a high-fat diet were exposed for 50 consecutive days (6 h/d) to 100 or 300 mg/m 3 , with plaque progression in the aortic leaflet region reported to be statistically nonsignificant, although an increased plaque areas in the DE-exposed mice was noted. The assessment of vasomotor function in aorta rings showed increased PEinduced vasoconstriction and unaltered ACH-induced vasorelaxation. However, in the exposed animals, there was an increased monocyte/macrophage infiltration into atherosclerotic plaques, which was not observed in mice that were only exposed to the gases. In addition, exposure to gases yielded the same level of atherosclerosis and PE-induced vasoconstriction, indicative that particulate fraction of the DE might not be driving the vascular outcomes (Vedal et al. 2013).
Inhalation of DE (200 or 400 mg/m 3 , 6 h/day) for 3 consecutive days or 7 weeks (5 d/week) in 30 week ApoE À/À mice on normal chow showed no effect on ACH-induced vasorelaxation, whereas there was decreased PE-induced vasoconstriction in aorta (Kido et al. 2011a). The authors also showed increased expression of heat shock protein 70 in lung tissue and plasma as a possible link between pulmonary and systemic responses. In addition, it was shown that the level of atherosclerosis was unaltered in the aortic root, whereas assessment of morphology indicated that plaques had a more advanced stage of atherosclerosis following inhalation to 200 mg/m 3 of DE with a MMAD of 104 nm (Bai et al. 2011).

Ingersoll Rand
ApoE À/À mice on Western-type diet were exposed by nose-only inhalation to DE (1700 mg/m 3 , 3 h/d, 5 d/week for 28 d) from regular diesel fuel with or without addition of CeO 2 . The addition of CeO 2 did not affect the MMAD of exhaust particles (&83 nm), whereas the particle number concentration was lower in the exhaust from diesel with CeO 2 (3.6 Â 10 6 particles/cm 3 ) as compared with regular diesel (5.3 Â 10 6 particles/cm 3 ). The exposure to DE from regular fuel accelerated plaque progression in BCA, whereas fuel supplemented with CeO 2 had no proatherogenic effect. There were increased number of pigmented macrophages in lung tissue and the results indicated a statistically non-significant increased plaque progression in BCA, whereas the plaques displayed signs of progression to a more advanced state with a buried fibrous cap (Cassee et al. 2012).

Bredenoord (35 KVA diesel generator under idling condition)
In an inhalation study, male Fischer rats were exposed for 4 weeks (174 mg/m 3 , 5 d/week, 6 h/d) to DE with a MMAD of 76 nm (GSD ¼ 2 nm). Prior to the DE exposure, the rats were exposed to ozone (0.4 ppm for 12 h) to induce mild pulmonary inflammation (Gerlofs-Nijland

Vascular effects after exposure to particulate matter from diesel exhaust
It is important to state that studies on CVD outcomes after exposure to DEP have used standardized material from the National Institute of Standards and Technology. SRM1650 (or later re-bottled samples) and SRM2975 have been used extensively to investigate pulmonary endpoints and oxidative stress. The two samples have similar oxidative stress potential, although they have been obtained from different combustion conditions (Møller et al. 2010a). Another type of material, referred to as A-DEP, was collected from a light-duty diesel engine. To the best of our knowledge, this material has not been used for in vivo studies of CVD endpoints, although it has been shown that it can reduce vasorelaxation responses to ACH and SNP, as well as reducing PEinduced vasoconstriction in thoracic aorta rings from male Sprague-Dawley rats after ex vivo exposure (Ikeda et al. 1995). Additionally, A-DEP can also affect the ACH-induced vasorelaxation response in rabbit aorta segments (Muto et al. 1996).

SRM1650b
This material is representative of heavy-duty diesel engine PM since it has been collected from different four-cycle diesel engines, operating under different combustion conditions. The original batch (SRM1650) has been rebottled and thus superseded to first SRM1650a and subsequently to SRM1650b. The material was produced in 1983 and issued in 1985; hence, it represents a relatively old diesel combustion technology (Risom et al. 2005). In a study, utilizing intranasal instillation (2 mg/mouse, 5 instillations/week for 6 weeks) of SRM1650b in ApoE À/À mice on a Westerntype diet there was increased plaque progression in the aorta (Pö ss et al. 2013).

SRM2975
This material was collected from an industrial dieselpowered forklift and issued in 2000. A short-term exposure (hydrodynamic particle size in saline was 280 nm (SD ¼ 9 nm)) in ApoE À/À mice (0.5 mg/kg at 26 and 2 h before sacrifice, total dose ¼ 1 mg/kg) by i.t. instillation did not alter ACH-induced vasorelaxation in aorta rings and there was no inflammatory response in terms of Ccl2 expression in lung tissue (Vesterdal et al. 2014). The same group also exposed ApoE À/À and wildtype C57BL/6 mice by intraperitoneal injection (0.5 or 5 mg/kg) and showed that only the low dose was associated with reduced ACH-induced vasorelaxation in ApoE À/À mice at 1 h post-exposure, whereas there was no difference in the SNP or PE response in aorta rings (Hansen et al. 2007). In a subsequent experiment, there was no effect on vasomotor function responses by ACH, SNP or PE in aorta rings at 1 h post-exposure to i.v. injection of 0.25 or 0.50 mg/kg (Bai and van Eeden 2013).
Using male Wistar rats, it was shown that a single i.t. instillation of SRM2975 (500 mg/rat) was associated with pulmonary inflammation and systemic inflammation in terms of IL-6 and TNF-a (Robertson et al. 2012). While, intra-arterial injection of these materials in the hind-limb vascular bed did not affect ACH-induced vasodilation in femoral arteries at 6 and 24 h post-exposure. However, the authors stated a reduced vasodilation in response to SNP in exposed rats. In the i.t. exposed rats, assessment of vasomotor function of aorta rings only indicated an augmented vasoconstriction response to PE at 6 h postexposure. In the same vessels, there was no alteration in response to ACH and SNP at 2, 6 and 24 h post-exposure. Likewise, the vasomotor function response to PE, ACH and SNP in femoral and mesenteric arteries were largely unaffected with the exception of an increased vasorelaxation response to SNP in mesenteric arteries at 6 h post-exposure (Robertson et al. 2012). The authors concluded that there was no association between pulmonary/systemic inflammation and vasomotor dysfunction. The modest responses in vasomotor function outcome by SRM2975 was in sharp contrast to effects observed by ex vivo treatment of isolated rat aortic rings (10-100 mg/ml) that showed attenuation of especially the vasorelaxation function (Miller et al. 2009).
Next, repeated exposure to SRM2975 (hydrodynamic size of particles in saline buffer was 257 nm (SD ¼ 46 nm)) by oropharyngeal aspiration (35 mg/ mouse, twice per week for 4 weeks) was associated with increased plaque area in both the aorta and the BCA of ApoE À/À mice on a Western-type diet, whereas there was no effect in wild-type animals (Miller et al. 2013). The histological examination of atherosclerotic plaques in exposed ApoE À/À mice indicated a transition toward a phenotype associated with increased vulnerability to rupture. The same study showed no effect on vasorelaxation response to ACH or SNP in the distal portion of the thoracic aorta or any systemic effects in terms of plasma levels of cholesterol, triglycerides, CRP or fibrinogen (Miller et al. 2013).
In another study, a dysfunction in ACH-induced vasorelaxation was observed in spontaneous hypertensive rats after repeated exposure of SRM2975 by i.t. instillation (0.8 mg/rat, three times per week for 4 weeks), whereas there was no response on SNP-induced vasorelaxation response (Labranche et al. 2012). The hydrodynamic particle size in phosphate buffer saline with Tween 80 was 189 nm (SD ¼ 16 nm). In addition, ex vivo treatment of aorta rings with the particular matter (100 mg/ml for 30 min) attenuated vasorelaxation responses to ACH and SNP (Labranche et al. 2012). Interestingly, this sign of endothelial dysfunction was not observed in wild-type Wistar rats and it correlated with increased expression of p22phox in the aorta tissue of spontaneous hypertensive rats (Labranche et al. 2012).

Summary of studies on exposure to DE and DEP
The studies on pulmonary exposure to DE or DEP have shown mixed outcomes on CVD progression with approximately equal number of studies demonstrating either an altered or no effect on progression of atherosclerosis and vasomotor function. There does not appear to be a relationship between specific types of diesel engines and effect on CVD outcomes. Direct exposure of blood vessels to DEP either via i.v. injection in animals or by ex vivo exposure of vessels has been associated with vasomotor dysfunction, but the relevance of these observations for prediction of in vivo effects is questionable due the higher concentrations utilized.

Vascular effects after exposure to nanomaterials
NMs have various applications either as components in industrial products, additives and drugs. The prominent body of research in the nanoparticle field is primarily focused on the development and exploitation of these materials while a smaller body of work has concentrated on the putative hazardous properties of these NMs. In the first category of studies, NMs have been highlighted for their special properties in drug delivery or diagnostics. This may introduce publication bias because potentially hazardous NMs might have been screened out in projects on development of nanomedicines i.e. development of nanocarriers in drugs or therapeutics. Naturally nanotoxicology research is dominated by studies on NMs with the priority in establishing the potential hazardous nature of the materials. In the present review, we have included publications that have strived to assess toxicological effects to the cardiovascular system after exposure to NMs. Consequently, it is mainly ''classic'' types of particles from inhalation toxicology, e.g. TiO 2 and carbon black, carbon nanotubes (CNTs) and fullerenes C 60 . A few studies also have investigated types of specific NMs that can be found in air pollution PM (i.e. nickel hydroxide) and DE (i.e. CeO 2 ). Hereafter, the term ''nanomaterial'' is used to distinguish these materials from anthropogenic particles in ambient air. These NMs have a primary size in nanometer range, although the actual particle size in air or suspension may be above 100 nm as a consequence of agglomeration. The majority of NMs can be classified as ''spherical'' in comparison with CNTs that are long and thin fibers. Thus, the toxic potential of CNTs is typically compared with other types of fibers such as asbestos (Donaldson et al. 2013).

TiO 2
Exposure to nanosized TiO 2 (primary particle size was reported to be 15 nm, whereas the size distribution in exposure vehicle was not reported) altered ACH-induced vasodilation in second-order branch intralobar pulmonary arteries of male Wistar rats at 21 d post-exposure following i.t. instillation of 100 mg/rat (Courtois et al. 2010). These results were supported by observations that 24 h incubation of pulmonary arteries (200 mg/ml) was not associated with differences in ACH-induced vasorelaxation (Courtois et al. 2008(Courtois et al. , 2010. Nurkiewicz et al. have conducted a number of studies on whole-body inhalation exposure to nanosized TiO 2 in male Sprague-Dawley rats that document impaired vasodilatation response in arterioles of the microcirculation. The primary particle size of these materials was 21 nm -80% anatase and 20% rutile. The aerosol displayed a bimodal size distribution with peaks at 100 and 400 nm (CMAD ¼ 138 nm). In these investigations, different concentrations (1.5-12 mg/m 3 ) and exposure times (2-8 h) were used to produce a range of welldefined deposited doses (4, 6, 10, 19 and 38 mg/rat) for the assessment of vasodilation in situ in pressurized spinotrapezius muscle arterioles at 24 h post-exposure. The data revealed a dose-dependent impairment of A23187induced vasodilation exemplified by a dose of !10 mg/rat (&0.05 mg/kg) associated with reduced vasodilation (Nurkiewicz et al. 2008). A subsequent study showed impaired endothelium-dependent A23187-induced vasodilation in spinotrapezius muscle arterioles, whereas the endothelium-independent SNP-induced vasodilation was unaltered at a deposited dose of 10 mg/rat (Nurkiewicz et al. 2009). This suggests the development of endothelium dysfunction, which occurred concurrently with oxidative stress (i.e. elevated ROS levels) and nitrosative stress in vessels of the microcirculation (Nurkiewicz et al. 2009). The addition of antioxidants (i.e. tempol and catalase) and inhibitors of ROS producing enzymes (i.e. apocynin (inhibitor of NADPH oxidase) and 4-aminobenzoic hydrazide (inhibitor of myeloperoxidase) restored the A23187-induced vasodilation, indicating that the TiO 2 -induced dysfunction in vasodilation might be caused by oxidative stress (Nurkiewicz et al. 2009). Additionally, pretreatment of the animals with cyclophosphamide to deplete neutrophils had no effect on the blunted vasodilation response in the pressurized spinotrapezius muscle arterioles in Fisher 344 rats; while the levels of 26 different circulating cytokines measured were not significantly different from the controls (Nurkiewicz et al. 2011). In another study, reduced vasodilation in spinotrapezius muscle arterioles in response to active hyperemia in Spraque-Dawley rats following inhalation of TiO 2 (corresponding to a deposited dose of 10 mg) was attributed to action via cyclooxygenases, microvascular nitric oxide bioavailability and augmented sympathetic responsiveness (Knuckles et al. 2012). The same authors further documented a blunted vasodilation response to ACH and A23187 in subepicardial arterioles of rats after whole-body inhalation exposure to 6 mg/m 3 for 4 h, corresponding to a deposited dose of 10 mg/rat, whereas endothelial-independent vasodilation response to SNP was unaltered (LeBlanc et al. 2009). In keeping with observations in the spinotrapezius muscle arterioles, coronary resistance arterioles had increased ROS production and impaired ACH-induced vasodilation which was restored by pretreatment with tempol (LeBlanc et al. 2010). The same coronary resistance arterioles displayed only modest vasoconstriction response to arachidonic acid, whereas there was augmented vasoconstriction response to a thromboxane analog (U46619) (LeBlanc et al. 2010). Moreover, in utero exposure to TiO 2 (10 mg/m 3 , 5 h/d for an average of 6.8 d) from gestation day 6 in Sprague-Dawley rats showed a blunted ACH-induced vasorelaxation response in coronary and uterine arterioles (Stapleton et al. 2015). Meanwhile, a similar response was observed for endothelium-independent vasorelaxation by spermine NONOate treatment, whereas the vasoconstriction response was unchanged. These observations are interesting as they suggest a transcended priming of vasomotor function.
The inhalation of a commercial spray product with nanosized TiO 2 (CMAD: 110 nm, 2.62 mg/m 3 for 2 h, 1.72 mg/m 3 for 4 h or 3.79 mg/m 3 for 4 h for 4 d) was associated with pulmonary inflammation, whereas it did not alter the vasomotor function response to ACH or PE in the ventral tail arteries of Sprague-Dawley rats (McKinney et al. 2012).
In another study, administration of nanosized rutile or photocatalytic TiO 2 to ApoE À/À mice (0.5 mg/kg at 26 and 2 h before sacrifice, total dose ¼ 1 mg/kg) by i.t. instillation showed unaltered vasorelaxation response in aorta rings to endothelium-dependent (ACH or calcitonin-gene related peptide) and endothelium-independent vasodilators (nitroglycerin or filodipine) in the presence or absence of the superoxide dismutase mimic tempol (Mikkelsen et al. 2011). The primary particle sizes of the samples were 21 and 12 nm, whereas the hydrodynamic diameter in suspension was well above the nanosize range (518 and 2321 nm) for the rutile and photocatalytic TiO 2 , respectively. However, the same study showed that i.t. instillation of rutile TiO 2 (0.5 mg/kg once a week for 4 weeks, total dose ¼ 2 mg/kg) was associated with a minor increase in plaque area in the aorta, which was not accompanied by pulmonary inflammation as assessed by gene expression levels of monocyte chemoattractant protein-1 (MCP-1) and macrophage inflammatory protein 2 (MIP-2) (Mikkelsen et al. 2011).
Finally, exposure of 5-10 nm anatase TiO 2 by i.t. instillation (5, 25 or 50 mg twice a week for 6 weeks) in ApoE À/À mice on regular diet resulted in a bell-shaped dose-response relationship where a statistically significant increase in plaque progression by en face staining of the whole aorta was observed in the group of mice that had received 25 mg (&2.5 mg/kg per week). On the contrary, assessment of plaque progression by H&E stain of cross-sections of the aortic arch indicated only statistical significance at 50 mg (&5 mg/kg per week). Hence, both methods indicated unaltered plaque progression at the lowest administered dose (&0.50 mg/kg per week). These observations may indicate a threshold for plaque progression following repeated i.t. instillations at about 0.5 mg/kg per week in ApoE À/À mice on regular diet (Chen et al. 2013a). These findings are in line with other observations that a single i.t. instillation of nanosized TiO 2 (primary particle size ¼20 nm), count median diameter ¼ 16 nm, GSD ¼ 1.4 nm) was associated with PMN infiltration in BALF at doses of 25 and 100 mg/ mouse (&1.25 and 5 mg/kg), whereas there was no effect after exposure to 6 mg/mouse (&0.3 mg/kg) (Oberdö rster et al. 2000).

Multi-walled carbon nanotubes (MWCNTs)
MWCNTs consist of multiple sheets of graphine rolled into a hollow tube. It has been suggested that due to their physicochemical characteristics (long, thin and biopersistent), these materials behave more like fibers than spherical particles in the airways.
In one investigation of cardiovascular complications associated with MWCNT (commercially available -5% iron impurity with a fiber length of several microns and a hydrodynamic size in suspension with bimodal distribution of 200 and 1000 nm) in male Sprague-Dawley rats were exposed to 100 mg of materials (&0.27 mg/kg) by i.t. instillation and sacrificed at 24 h post-exposure. The data showed unaltered vasorelaxation response to ACH and SNP in the left anterior descending coronary arteries. There was a statistically non-significant increase in vasoconstriction response to ET-1, whereas the response to serotonin was unaltered (Thompson et al. 2014a). The general impression from the wire myograph data was that the exposure to MWCNTs did not alter the vasomotor function. In contrast, there were signs of augmented vasoconstriction and increased cardiac infarct size following ex vivo ischemia/reperfusion in Langendorf experiments. The exposure did not generate pulmonary inflammation, assessed by total cell counts in BALF, whereas there was increased protein concentration (Thompson et al. 2014a).
MWCNT-7 was the material of choice in one study on inhalation exposure of male Sprague-Dawley rats (5 mg/m 3 for 5 h). The MWCNT was characterized via electron microscopy image analysis, which demonstrated that the fibers were bended with a mean length of 3.9 mm. The rats were exposed to a deposited dose of 13.5 mg (&0.04 mg/kg) and sacrificed at 24-168 h post-exposure. This exposure was associated with pulmonary inflammation and cytotoxicity (increased PMNs and LDH in BALF). The study further indicated endothelial dysfunction, based on a blunted vasodilation response to ACH and A23187 in pressurized subepicardial arterioles and unaltered SNP-induced endothelium-independent dilation (Stapleton et al. 2012a). Furthermore, there was a statistically statistical significant (p ¼ 0.08, ANOVA) reduced PE-induced vasoconstriction (statistically significant in a post-hoc test and considered to be ''additional exploration'' by the authors) (Stapleton et al. 2012a). Nevertheless, the authors suggested that the reduced vasoconstriction response could be a protective mechanism to endothelium-dependent vasomotor dysfunction. Indeed, this somewhat surprising finding is interesting in light of the complex regulation of the vascular tone, which may not be easily and readily manipulated by functional tests with the use of vasoactive compounds.
In another study, two different types of MWCNTs (NM400 and NM402) sourced from the European Commission Joint Research Center Nanomaterials Repository was used to investigate the impact of differences in fiber dimensions on progression of atherosclerosis. These MWCNTs were originally selected to represent a short and a long fiber type, based on the supplier's description of the material. However, a thorough characterization of the samples by the authors showed that the materials had similar dimensions (a fiber length of 0.7-3 mm and 0.4-4 mm) and hydrodynamic diameter in saline (i.e. NM400 and NM402 had a median particle size 116 and 147 nm, respectively). The study showed increased plaque progression in aorta of ApoE À/À mice on a Western-type diet at 24 h after 5 i.t. instillations (once a week for 5 weeks, total dose ¼ 128 mg/mouse), whereas there was only a statistically non-significant increase for one type of MWCNT after a 4-week recovery period (Cao et al. 2014). The exposure was associated with pulmonary inflammation in terms of neutrophilic infiltration in the BALF and elevated cytokine levels, whereas there were unaltered cytokine levels in serum (Cao et al. 2014).
The sub-chronic exposure of MWCNTs in female ApoE À/À mice (40 mg MWCNT by pharyngeal aspiration once a week for 16 weeks and sacrificed at day 1 or 7 postexposure) did not accelerate atherosclerotic lesions in the aorta or affect plasma cholesterol levels, whereas the exposure increased the numbers of PMNs in BALF, total protein and LDH (Han et al. 2015). The hydrodynamic size of MWCNTs in suspension was 98 ± 10 nm with agglomerates of 30-300 nm. The primary dimension of MWCNTs was 20-30 nm with a mean fiber length of 20-50 mm.

Single-walled carbon nanotubes (SWCNTs)
SWCNTs consist of a singular cylindrical graphine sheet that is rolled into a hollow tube. In a study, ApoE À/À mice were exposed to one such SWCNT (less than 1 mm in length -0.5 mg/kg twice during 24 h, total dose ¼ 1 mg/kg). The data showed no effect on ACH-induced vasorelaxation in aorta rings, whereas there was an increased gene expression of CCL-2 (approximately seven-fold) in lung tissue, indicating at a potential inflammatory response (Vesterdal et al. 2014).
Next, the i.t. instillation of SWCNT (20 mg/mouse every other week for 8 weeks, total dose &3.2 mg/kg) increased plaque area in the aorta and BCA of ApoE À/À mice on a high-fat diet, whereas results from animals on regular diet were not quantitated due to changes being too small (Li et al. 2007). There was unaltered plasma level of inflammation markers (MCP-1, IL-12, IL-6, TNF-a and INF-g), yet other signs of inflammation (increased macrophage-3 antigen and vascular cell adhesion molecule-1 protein in BCA) and oxidative stress (increased protein carbonyls and decreased GSH/GSSG ratio in aorta) were noted in SWCNT exposed mice (Li et al. 2007). In this study, the fiber length was not reported for the SWCNTs; however, it was stated that the material had low content of impurities (0.23% Fe).
The i.v. injection of SWCNTs (1.8 mm in length) in Wistar rats (5 mg/rat and sacrifice at 3 h post-exposure) was not associated with vasomotor dysfunction in terms of response to ACH, SNP and PE in aorta rings (Vlasova et al. 2014). Using the same material, it was shown that a concentration of 200 mg/ml increased vasorelaxation response in pressurized mesenteric arteries and blockage of nitric oxide synthesis by N-nitro-L-arginine methyl ester (L-NAME) pretreatment abolished the effect (Vlasova et al. 2014). Other observations from ex vivo exposure studies indicated that SWCNT exposure in a concentration-dependent manner (0.1-10 mg/ml) produced relaxation of PE pre-contracted aorta rings independent of endothelium involvement (Gutierrez-Hernandez et al. 2015).
In a recent study, the injection of Fe-SWCNT (length: 1.5-3.0 mm) or Gd-SWCNT (length: 0.6-0.8 mm) near an arteriole in the left cheek pouch of hamsters or the right cremaster muscle of mice was carried out and vasomotor function by intravital microscopy measured (Frame et al. 2014). In general, there was little difference in response among the SWCNTs, animal models and location of measurement of vasomotor function (arcade or terminal location). It was noted that the suspensions of aggregated SWCNTs caused an immediate vasodilation, whereas non-aggregated suspensions caused vasoconstriction. These effects subsided within 1 min after injection and a blunted ACHinduced vasodilation response was observed at 15 min post-exposure.

Carbon black
An early study on in vivo exposure of a nanosized carbon black with a primary size of 13-14 nm showed unaltered ACH-induced vasorelaxation in second-order branch intralobar pulmonary arteries of male Wistar rats at day 21 after i.t. instillation of 100 mg/rat (Courtois et al. 2010). These results were supported by observations that 24 h incubation of pulmonary arteries (200 mg/ml) did not affect the ACH-induced vasorelaxation response (Courtois et al. 2010). The particle size in exposure vehicle was not reported. However, earlier observations from the same group also had shown unaltered response to ACH-induced vasorelaxation in rat pulmonary arteries after exposure to a high-concentration (200 mg/ml) for 24 h to two different samples of nanosized carbon black with a primary particle size of 13 and 21 nm (Courtois et al. 2008).
A 4 weeks exposure to carbon black with a mean aerodynamic diameter of 85 nm (1.3 Â 10 5 -42 Â 10 5 particles/cm 3 , 4 h/d and 5 d/week) did not cause pulmonary inflammation and unaltered vasomotor response to ACH-induced vasorelaxation and vasoconstriction response to PE and 5-HT in the aorta of Sprague-Dawley (Kim et al. 2011). Still, opposite observations have been obtained in studies using a different sample of carbon black (Printex 90, primary particle size ¼ 14 nm), in which concentration-dependent ACHinduced vasorelaxation ex vivo was decreased in aorta rings after a 30 min exposure period to 100 mg/ml, whereas SNP-induced endothelium-independent vasorelaxation was only marginally altered and PE-induced vasoconstriction was increased (Vesterdal et al. 2012). The same study also showed an increased pressurediameter relationship in mesenteric arteries from rats following a 30 min exposure to 10 mg/ml of Printex 90 (Vesterdal et al. 2012). There was no difference in vasomotor function, assessed by ACH, SNP or PE responses in aorta rings, following a single i.t. instillation in ApoE À/À mice (0.05-2.7 mg/kg), whereas a dosing regimen of two i.t. instillations (26 and 2 h before sacrifice) reduced the responsiveness to ACH-induced vasorelaxation (Vesterdal et al. 2010). Analysis of the particle size in exposure vehicle indicated a bimodal distribution with peaks at 1.4 and 5.5 mm (Vesterdal et al. 2010(Vesterdal et al. , 2012. The highest dose increased neutrophilic influx in BALF (Jacobsen et al. 2009) and MCP-1 gene expression in lung tissue (Vesterdal et al. 2010), without concurrent vasomotor dysfunction, whereas there was no indication of pulmonary inflammation following two i.t. instillations of 0.5 mg/kg (total dose ¼ 1 mg/kg). This suggests that pulmonary inflammation was not necessary for vasomotor dysfunction in aorta following airway exposure to Printex 90. The same study also investigated plaque progression in 48-49 weeks old ApoE À/À mice as a model for advanced CVD. It was planned that the mice should receive repeated i.t. instillations over several weeks, but the exposure was stopped after two i.t. instillations due to death of a couple of mice in each group following exposure. Thus the total received dose (1 mg/kg) was relatively low, which could be the reason for the unaltered plaque progression in aorta and BCA (Vesterdal et al. 2010).
The utilization of carbon black at high doses by i.t. dispersion (1 mg/week for 10 weeks, total dose ¼ 10 mg/ mouse, MMAD ¼ 121 nm) in Ldlr À/À mice on a cholesterol-rich diet was associated with increased plaque progression in aorta, whereas there was no statistically significant effect in mice on a regular diet (Niwa et al. 2007). However, it should be emphasized that the doses in this study were very high and possibly beyond any relevance for human exposures.
The exposure to Printex 90 once a week for 10 weeks via oral gavage (total dose ¼ 0.64 or 6.4 mg/kg, hydrodynamic mean size: 104 ± 53 nm) was associated with a dose-dependent decrease in ACH-induced vasorelaxation of aorta rings from lean and obese Zucker rats, whereas the same total dose administered as a bolus exposure had no effect (Folkmann et al. 2012). The vasorelaxation responses to nitroglycerin and felodipine were unaltered in exposed rats, as was the vasoconstriction response to PE (Folkmann et al. 2012). These results indicated that repeated oral exposure to nanosized carbon black was associated with endothelial dysfunction, which may further depend on sustained exposure (vasomotor dysfunction had subsided after a 13-week recovery period) (Folkmann et al. 2012).

Fullerenes C 60
This material has a bucky-ball structure and is commonly regarded as the classic NM. In a recent study, i.t. instillation or i.v. exposure to fullerenes C 60 (formulated in polyvinylpyrrolidone with a hydrodynamic diameter of 371 ± 1.2 nm) in male and female Sprague-Dawley rats was proceeded by vasomotor function tests carried out in coronary arteries 24 h post-exposure to 28 mg (&0.93 mg/kg) of the materials. The exposure resulted in minimal pulmonary inflammation and slightly increased the protein content in BALF (Thompson et al. 2014b). In general, there was slightly increased vasoconstriction response to ET-1 in coronary arteries after i.t. exposure, which was ameliorated by pretreatment with indomethacin (a non-steroidal anti-inflammatory agent) in male rats. This was supported by a slightly increased vasoconstriction response to serotonin, although it did not reach statistical significance (p ¼ 0.06, ANOVA). There were unaltered vasorelaxation responses to ACH or SNP. The putative augmented vasoconstriction response after i.t. exposure was not accompanied by increases in serum levels of IL-6 or MCP-1, whereas these cytokines were elevated following i.v. exposure without any effect on vasomotor response. It should be noted that only five animals per group were utilized, which could affect the statistical power (Thompson et al. 2014b). Another study using the same formulation and dose of fullerenes C 60 assessed vasomotor function in pregnant (gestational day 17-19) and non-pregnant Sprague-Dawley rats at 24 h after an i.v. injection (Vidanapathirana et al. 2014). Based on the measurements of vasomotor function in aorta, main uterine artery and mesenteric artery, it was concluded that the fullerene C 60 exposure increased vasoconstriction in pregnant rats mediated by Rhokinase activity. It should be stated that the conclusions were based on a thorough assessment of the vasoconstriction response using 2-3 different vasoconstrictors in each blood vessel. The results indicate that the fullerenes C 60 in polyvinylpyrrolidone was not different from the vehicle control, whereas there was a different vasoconstriction response compared to the group of rats that was only exposed to saline. Thus the data indicated an effect induced by the polyvinylpyrrolidone rather than fullerenes C 60 on vasomotor function. Finally, the authors noted a smaller response on ACHinduced vasorelaxation; while the no alteration in cytokine levels in serum were determined after i.v. exposure of fullerenes C 60 .
The effect on vasomotor function following intraperitoneal injection (0.05 or 0.5 mg/kg) was investigated in ApoE À/À mice at different ages (age being important in levels of plaque progression in the aorta) (Vesterdal et al. 2009). The fullerenes C 60 had a small primary diameter of 0.7 nm, but suspension in saline a solution supplemented with BALF resulted in high agglomeration with diameters of above 1 mm. There was reduced ACH-induced vasorelaxation in aortic arch rings of 11-13 weeks old female ApoE À/À mice at both doses of fullerenes C 60 . The same dysfunction in ACHresponse was observed in rings from the descending part of the aorta in 40-42 weeks old female mice, which had a similar plaque level as the aortic arch of the young mice. On the contrary, it was not possible to obtain reliable information about fullerenes C 60generated vasomotor dysfunction in rings from the aortic arch of 40-42 weeks old mice due to low ACHinduced vasorelaxation in the control group. A similar dose-dependent relationship in SNP-induced vasorelaxation response was also observed, although it only reached statistical significance in the group of young mice in the high-dose group (0.5 mg/kg). Additionally, there was no effect on PE-induced vasoconstriction response. The aggregated data thus indicate that intraperitoneal injection of fullerenes C 60 was associated with dysfunction of the vasorelaxation response, whereas this cannot be only attributed to endothelial dysfunction.

Nickel hydroxide NMs
This NM has been investigated due to its use in power and energy industries. One research group has investigated vascular effects after inhalation of nickel hydroxide particles with a primary size of 5 nm (CMAD ¼ 40 nm).
In a series of studies, it has been demonstrated that whole-body inhalation exposure of nickel hydroxide (100-900 mg/m 3 for 5 h/d on 1, 3 or 5 d) was associated with a blunted ACH-induced vasorelaxation response in carotid arteries of C57BL/6 mice (Cuevas et al. 2010).
Earlier studies from the same group of authors showed that the exposure was associated with pulmonary inflammation (Gillespie et al. 2010). It was also shown that a 5 months exposure (79 mg/m 3 for 5 h/d, 5 d/week) resulted in pulmonary inflammation and increased plaque progression in the aorta of ApoE À/À mice (Kang et al. 2011).

Cerium dioxide (CeO 2 )
This material is used as diesel fuel additive to increase the combustion efficiency, although the material may also have applications in food, nanomedical and industrial products (Cassee et al. 2011). Two studies conducted by the same group investigated vascular effects of CeO 2 exposure in animals after i.t. instillation, i.v. injection or gastrointestinal administration. The primary particle size of the material was 4-6 nm, whereas the hydrodynamic diameter in suspension displayed a distribution with peaks at 191, 900 and 5081 nm. The effect of pulmonary exposure on vasomotor function in pressurized coronary and mesenteric arterioles was investigated in male Sprague-Dawley rats following a single i.t. instillation (10,50,100,200 or 400 mg/rat). First, reduced endothelium-dependent vasorelaxation was observed in both coronary and mesenteric arterioles by use of ACH and A23187 at all doses. A similarly blunted response was observed in spermine NONOateinduced endothelium-independent vasorelaxation at all doses; whereas, an augmented serotonin-induced vasoconstriction response was only observed in coronary arterioles at the highest dose (400 mg/rat) and the myogenic response was not affected. In addition, the influx of PMNs and LDH activity in BALF were unaltered at 10 mg/rat, however, these levels were increased at 100 mg/rat (Minarchick et al. 2013). This suggests that the dysfunction in vasorelaxation response occurred at doses that did not generate pulmonary inflammation and cytotoxicity. The same group has also investigated CeO 2 induced effects following i.v. injection and gastrointestinal tract exposures in rats. There was reduced ACH-induced vasorelaxation in mesenteric arteries in rats exposed by i.v. injection (50, 100 and 900 mg/rat), gastrointestinal (600 mg/rat) and i.t. instillation (65 mg/ rat). Moreover, impaired endothelium-independent vasorelaxation was also observed following i.v. injection, whereas gastrointestinal tract exposure was associated with either an augmented (100 mg/rat) or reduced (600 mg/rat) response. There was no difference in myogenic response and assessment of PE-induced vasoconstriction did not provide conclusive results with regard to sign of vascular dysfunction. A further investigation into mechanisms of the endothelial dysfunction by co-incubation of vessel segments with inhibitors of nitric oxide synthase (i.e. N G -monomethyl-L-arginine) or cyclooxygenase (i.e. indomethacin) indicated a dependency of the exposure route of CeO 2 on the vasorelaxation response to ACH (Minarchick et al. 2015).

Silicon-based NMs
These NMs are typically developed for a variety of diagnostic and therapeutic applications in medicine (Napierska et al. 2010). It has been shown that exposure to mesoporous silicon NMs (hydrodynamic diameter ¼ 90 nm) by i.v. injection (5 mg and sacrifice at 3 h post-exposure) was associated with reduced response to ACH in aorta rings of male Wistar rats and unaltered SNP response. There was also an augmented PE-induced vasoconstriction response in aorta rings. On the contrary, there was no effect on vasorelaxation after ex vivo exposure to 200 mg/ml in pressurized PE-precontracted mesenteric arteries (Vlasova et al. 2014). Similarly, in another study ex vivo exposure to amorphous silica NMs (50 nm, 2-50 mg/ml) resulted in a concentration-dependent reduction of ACH-induced vasorelaxation in PEprecontracted mesenteric arterioles from Wistar rats (Nemmar et al. 2014).

Quantum dots
As a group of NMs, quantum dots represent a relatively large group of semiconductor nanocrystals that are applied in biomedical imaging and electronics. The absorption, distribution, metabolism, excretion and toxicity of quantum dots depend on both inherent physicochemical properties and exposure conditions (Hardman 2006). These materials have received much attention with regard to their potential toxicity due to their high-scale manufacture. In one such study, male Wistar rats were exposed to mercaptoundecanoic acidcoated quantum dots by i.v. injection (100 mg or &333 mg/kg) for 2 h. The NMs were described as having a diameter of 5.0 ± 0.9 nm, although transmission electron microscopy images indicated at agglomerated particles with larger diameters. There was unaltered PE-induced vasoconstriction and ACH-induced endothelium-dependent vasorelaxation in pressurized fourthorder mesenteric arteries, whereas there was a blunted endothelium-independent (SNP-induced) vasorelaxation (Shukur et al. 2013). Next, ex vivo exposure of mesenteric arteries to quantum dots (15 mg/ml) showed unaltered level of ACH-induced vasorelaxation, whereas there was a slightly decreased response to SNP-induced endothelium-independent vasorelaxation (statistical significance not reported).
Summary of studies on exposure to NMs CVD outcomes have been assessed after pulmonary or non-pulmonary exposure to a number of different types of NMs. On one hand, pulmonary exposure to TiO 2 , MWCNTs and SWCNTs have been associated with adverse CVD outcomes, with some consistency across laboratories. On the other hand, pulmonary exposure to carbon black has yielded mixed results which might be related to the animal strain and type of carbon black utilized. Interestingly, both TiO 2 and carbon black have been shown to cause vasomotor dysfunction after oral exposure. Moreover, adverse CVD outcomes have been observed after exposure to nickel hydroxide NMs, CeO 2 , silicon-based NMs and quantum dots. However, these observations need to be confirmed in future experiments. The data from the literature collectively indicate that exposure to a range of NMs is associated with vasomotor dysfunction and progression of atherosclerosis.
Quantitative analysis of association between exposure to particulate matter and vascular effects Effect of exposure to particulate matter on vasomotor function Currently, it is not possible to establish the differences in effect size in a systematic analysis of vasomotor function because some studies have not reported actual values in null effect findings. For the analysis of vasomotor dysfunction, we have dichotomized the response to either showing a statistically significant or null effect (Table 4). Furthermore, the percentage of studies with altered vasomotor function in the full dataset as well as for the studies that have investigated effects after pulmonary exposure to PM have been reported. In this analysis, a ''study'' refers to the measurement of one PM sample in an experimental system. Thus, a single reference can refer to more than one study if it incorporates more than one material (e.g. DE and CAPs), strain (e.g. assessment of the same sample in wild-type and spontaneous hypertensive rats) or experimental system (e.g. ApoE À/À mice fed normal chow or high-fat diet). Overall, approximately 40% of the studies have reported altered responses in terms of vasoconstriction, whereas 60% have shown altered endotheliumdependent vasorelaxation (Table 4). It should be noted that there is an uneven distribution of studies that have reported altered vasoconstriction response after Numbers in bracket are type of particulate matter (ambient air pollution particles/diesel exhaust/nanomaterials). Ambient air pollution particles encompass studies on authentic air pollution, concentrated ambient air particles and particulate matter from ambient air. a Different distribution of the type of particulate matter between studies showing altered and unaltered vasorelaxation response (pulmonary exposure: 2 ¼ 11.7, p50.01, all studies: 2 ¼ 11.1, p50.05). b Different distribution of the type of particulate matter between studies showing altered and unaltered response on vasoconstriction response ( 2 ¼ 7.2, p50.05). c p50.05 (there is a non-random distribution in case the 95% lower confidence (95% CI) interval does not include 5%. The 95% confidence interval has been calculated for the binominal distribution). pulmonary exposure to particles ( 2 ¼ 7.2, p50.05), which is mainly driven by an over-representation of studies showing vasoconstriction after exposure to air pollution particles and an over-representation of studies showing unaltered vasoconstriction after exposure to NMs. There were also differences in the endotheliumdependent vasorelaxation response in studies of ambient air pollution particles, DE and NMs (pulmonary exposure: 2 ¼ 11.7, p50.01, all studies: 2 ¼ 11.1, p50.05). This heterogeneity is mainly driven by studies that have shown unaltered response after exposure to DE or DEP and a high number of studies on ambient air pollution particles that have shown vasomotor dysfunction. There appears to be relatively few studies on endothelium-independent vasorelaxation response and the majority of these studies show unaltered effect after PM exposure. It has not been possible to ascertain whether PM-exposure causes endothelium dysfunction because there are relatively few studies that have assessed both endothelium-dependent and endothelium-independent vasorelaxation.

Summary
It appears that exposure to PM is associated with increased vasoconstriction (approximately 40% of the studies) and reduced endothelium-dependent vasorelaxation responses (approximately 60% of the studies).
The majority of studies have investigated pulmonary exposure to PM. Vasoconstriction responses have consistently been observed after exposure to ambient air pollution particles, whereas exposures to DE (or DEP) and NMs have been less consistent with regard to the generated vasoconstriction. There is a similar pattern of endothelium-dependent vasorelaxation after exposure to ambient air pollution particles and NMs, whereas there are fewer studies than anticipated that have shown endothelium-dependent vasomotor response following exposure to DE and DEP.

Effect of exposure to particulate matter on progression of atherosclerosis
The results on plaque progression have been reported in sufficient detail to allow for the use of a meta-analysis tool for assessment of effect size. Therefore, we have calculated standardized mean difference (SMD) with 95% confidence interval (CI) in a random effects metaanalysis using Review Manager (RevMan) version 5.0 (The Nordic Cochrane Center, The Cochrane Collaboration 2008). The SMD is the difference between two groups divided by the pooled SD and used due to the fact the results in different studies have been reported with different scales. A positive value of SMD indicates plaque progression. This effect is statistically significant if the 95% CI does not include zero. The SMD is depicted in a Forest plot, which displays means and 95% CIs for subgroups and the overall analysis with the heterogeneity between studies assessed in a Funnel plot. This depicts the standard error and as function of the SMD. An asymmetrical funnel shape indicates heterogeneity between studies. The effect of exposure to PM and subsequent plaque progression is shown in a Forest plot (Figure 2). The analysis indicates that exposure to DE (and DEP) has the least effect (SMD ¼ 0.36, 95% CI: À0.05 to 0.77), followed by NMs (SMD ¼ 0.67, 95% CI: 0.26-1.09) and ambient air pollution particles (SMD ¼ 1.26, 95% CI: 0.80-1.72). However, it has to be noticed that the group of ambient air pollution particles displays heterogeneity, which is attributed to 4 studies with high effect sizes (Ying et al. 2009a, Chen et al. 2013b, Li et al. 2013b, Rao et al. 2014). The heterogeneity is lost by omission of these publications from the meta-analysis, with the effect size decreased (SMD ¼ 0.70, 95% CI: 0.46-0.94), while there is no difference in plaque progression between exposure to ambient air particles, DE (and DEP) and NMs ( Supplementary Figures 2 and 3).
The quantitation of plaque progression with the AHA classification (Table 3) has been used in a few studies on EHC-93 exposure to rabbits (referred to as ''atherosclerosis score''). The first study showed increased atherosclerosis score in the left main coronary artery (score 2.7 versus 1.9 in exposed and controls) and aorta (score 3.1 versus 2.4) (Suwa et al. 2002). In a later study of New Zealand White rabbits on high-cholesterol diet also showed increased atherosclerosis score in the abdominal aorta after exposure to EHC-93 (score 1.7 and 1.0 in exposed and unexposed, respectively) (Miyata et al. 2013). It has been reported that inhalation of DE was associated with increased collagen content and macrophage infiltration in the aorta leaflet region of ApoE À/À mice on high-fat diet [photographic documentation and description of plaque morphology suggest that DE exposure was associated with a transition from AHA types 4-5 to 5] (Campen et al. 2010). Vedal et al. (2013) showed increased monocyte infiltration in aorta plaques from ApoE À/À mice after exposure to mixed vehicle exhaust [photographic documentation indicates a plaque development in AHA types 3-4, whereas it is not possible to score differences between exposed and unexposed animals]. Furthermore, exposure to DE increased the complexity of atherosclerotic plaques as occurrence of buried fibrous layers in plaques in BCA, whereas there was unaltered lipid content and macrophages/foam cells [this is probably equivalent to an increase in AHA score from 5 to score56] (Cassee et al. 2012). The development of atherosclerosis following exposure to ambient air pollution in LDLr À/À mice on high-fat diet has also been documented [photographic documentation indicates a uniform lipid deposition with numerous cellular structures (Oil Red with hematoxylin counterstain), which differs somewhat from a regular cross-section of plaques]. The plaque morphology was described as a thickening of the arterial wall, which appears to be seen in the intima-media [photographic documentation provides insufficient detail to suggest AHA types] (Soares et al. 2009). It should be emphasized that the quantitative assessment of plaque progression throughout this review is speculative as the assessment is based on a relatively few cross-sections of arteries in different animal models. However, the assessment does suggest that PM exposure is associated with less than one unit AHA score over an exposure period of 1-2 months. This is based on observations from studies on DE and ambient air pollution particles, whereas there are no studies on NMs. As a recommendation for the future it is imperative that investigations into plaque morphology are carried out, although it should be kept in mind that the atherosclerotic histology differs between humans and animal models.

Summary
Exposure to PM is associated with accelerated plaque progression in atherosclerosis-prone mice and rabbits. The studies on ApoE À/À mice have used either normal Figure 2. Forest plot of studies on progression of atherosclerosis after airway exposure to PM. A standardized mean difference (SMD) in the direction toward ''Favors [control]'' indicates plaque progression, whereas ''Favors [experimental]'' indicates no effect. There is statistically significant effect if the 95% confidence interval does not include zero. The corresponding Funnel plot is reported in Supplementary Figure 1. chow or high-fat diet. The latter accelerates progression of atherosclerosis and may also produce a different histopathological composition of the atheroma. To date, only a few studies have assessed histopathological changes in atheromas after PM exposure in groups of high-fat or regular chow fed animals to provide conclusions about differences in plaque morphology. However, there appears to be no difference in plaque progression between animals fed a normal chow and high-fat diet which is not entirely surprising as the high-fat diet merely accelerates the progression of atherosclerosis.
Dose-response relationships for CVD outcomes PM-induced vascular effects have mainly been studied at a singular dose in each study. Hence, there is a lack of dose-response relationships on progression of atherosclerosis. For ambient air pollution particles, there is one study on vasomotor dysfunction after exposure to EHC-93 that has used more than one dose (Kido et al. 2011b). It has been shown that inhalation of DE was not associated with dose-dependent responses of vasomotor dysfunction in the aorta and plaque progression (Campen et al. 2010, Kido et al. 2011a. Furthermore, pulmonary exposure to carbon black, MWCNT, TiO 2 , CeO 2 and nickel hydroxide NMs has not revealed doseresponse relationships on vasomotor dysfunction endpoints or plaque progression (Cuevas et al. 2010, Vesterdal et al. 2010, Chen et al. 2013a, Minarchick et al. 2013, Cao et al. 2014. One study on intraperitoneal injection of fullerenes C 60 was non-conclusive because the effect on vasomotor function was small at all doses investigated (Vesterdal et al. 2009). However, i.v. injection of CeO 2 was associated with a blunted ACHresponse at all doses tested, whereas only the highest dose (600 mg/rat) reduced ACH-induced vasorelaxation in mesenteric arterioles of rats (Minarchick et al. 2015). In another study, oral exposure of nanosized carbon black showed a dose-dependent decrease in ACH-dependent vasorelaxation in the aorta of lean and obese Zucker rats (Folkmann et al. 2012). Table 5 shows the analysis of the dose-response relationship for vasomotor function test in studies on instillation or inhalation of PM. The instilled dose (mg/kg bodyweight per week) was higher in studies that showed altered endothelium-dependent vasorelaxation response (p50.05). There is a positive association between the administered dose and altered vasomotor response (odds ratio ¼ 6.1 (95% CI: 1.2-31) for effect in studies that have used a dose above versus below the geometric mean). These results indicate a dose-dependent relationship on endothelium-dependent vasorelaxation in the instillation studies, whereas there was no difference in inhalation studies. Finally, there were no dose-dependent relationships on endothelium-independent vasorelaxation or vasoconstriction. Figure 3 outlines a comparison of dose-response relationships between inhalation of PM and plaque progression across publications on atherosclerosis-prone mice with the outcome being the daily plaque progression (SMD divided by the number of exposure days) and the exposure calculated as time-integrated weekly dose (mg*h/m 3 ). Interestingly, by using this exposure metric, it has been shown that exposure to 100 mg*h/m 3 of carbon black was associated with pulmonary overload in rats . All studies incorporated in this review have utilized time-integrated weekly doses less than 100 mg*h/m 3 (Figure 3). In general, there is a large spread in the effect sizes between the studies, with three studies showing relative high effect sizes (SMD/ exposure day 40.03); while one has utilized a highexposure concentration of UFPs (360 mg/m 3 ), which may explain the high effect size (Li et al. 2013b). The other studies, originating from Beijing (China) and New York (NY), do not have any striking features that readily explain the high effects (Ying et al. 2009a, Chen et al. 2013b). In general, there is little evidence for doseresponse relationships across studies and there does not appear to be an overt difference between PM originating from DE and ambient air pollution. Figure 3. Relationship between dose and plaque progression in studies that have administered PM by inhalation. All studies have used long-term inhalation exposure in ApoE À/À or LDLr À/À mice. The y-axis is standardized mean difference (SMD) per exposure day. Symbols are studies with exposure to PM from ambient air pollution (diamonds), diesel exhaust (squares) or nanomaterials (triangle). Open and solid symbols represent mice that have been fed a high-fat (solid symbols) or normal chow (open symbols). Figure 4 depicts the dose-response relationship for studies on i.t. instillation (or oropharyngeal aspiration) in atherosclerosis-prone mice and rabbits. The exposure is reported as weekly dose per bodyweight (mg/kg per week) due to the difference in species. To allow for comparison between Figures 3 and 4, a weekly dose of 1 mg/kg bodyweight corresponds to inhalation of 2 mg/ m 3 of PM for 6 h/d and 5 d/week (i.e. 60 mg*h/m 3 ) in mice. The data show a spread in effect sizes between studies, which is not readily explained by differences in types of PM (Figure 4). The results might indicate that a weekly dose of 1 mg/kg bodyweight produces little effect in terms of plaque progression, whereas the effect at higher doses may depend on the type of PM.

Summary
There is little evidence indicating that pulmonary exposure to increasing PM dose leads to vasomotor dysfunction and plaque progression, whereas one study on oral exposure to nanosized carbon black indicates a dose-response relationship. The few studies that have used multiple doses within the same study have not revealed clear dose-response relationships. From the analysis of the data, it is clear that comparison of findings and conclusions are difficult across studies due to the fact that the physicochemical characteristics of PM utilized are too different. The use of disparate methodology and experimental setups further complicates the above mentioned issue.

Particle characteristics and effect on vascular outcomes
The quantitative analysis of the data indicates a scatter pattern in exposure-effect relationships on vasomotor function and atherosclerosis (Figures 3 and 4). This is not surprising since the analysis is based on the mass concentration, which is a relatively inaccurate descriptor of the exposure to nanosized particles. It can be suggested that particle size, shape and chemical composition might be better parameters for prediction of the toxicological potential of particles. However, there is not sufficient information available in the literature on these factors to allow for the assessment of the differences in CVD outcomes within specific types of PM exposure.
It is apparent that the studies on air pollution particles have rather poor description of particle characteristics. Table 5. Dose-response relationship for vasomotor function in animals after instillation or inhalation exposure.
For instance, the publications on CAPs exposure from Columbus, OH, at best only contain information on the chemical composition of the PM (Ying et al. 2009b, Xu et al. 2010, Kampfrath et al. 2011, Rao et al. 2014, Ying et al. 2014. A number of studies of CAPs inhalation from Tuxedo, NY, have not provided any data about particle characteristics (Sun et al. 2005(Sun et al. , 2008a(Sun et al. , 2008b(Sun et al. , 2009, whereas basic description on the elemental composition and particle size is available in other publications (Chen and Nadziejko 2005, Ying et al. 2009a, Quan et al. 2010. The assessment of elemental composition in the studies of air pollution particles seems to be related to source appointment rather than assessment of toxic components in the exposure. The studies on DE and NMs appear to have better description of particle characteristics than the investigations into air pollution particles. Nevertheless, the analysis of the chemical components in the NMs seems mainly to have been used to describe the purity of the sample rather than describing constituents that can explain the exposure-effect relationship. There are some studies on NMs with information on shape and agglomeration of the particles, but this is very limited and cannot be used for comparison of studies on this particle characteristic. Additionally, a number of publications have cited other publications when describing the characteristics of the exposure. For instance, the studies on instillation of EHC-93 in the airways have cited Vincent et al. (1997) as the source of particle size characteristics (Suwa et al. 2002, Bagate et al. 2004a,b, Goto et al. 2004, Tamagawa et al. 2008, Yatera et al. 2008, Kido et al. 2011b, Miyata et al. 2013. However, this publication only contains data on primary size and aerodynamic diameter, whereas the most important and interesting information must be the hydrodynamic particle size and agglomeration due to the fact that the ECH-93 has been administered in a suspension vehicle. A similar example of citation to previously published work is seen in studies on DE, in which the description of exposure is vague with little detail on the equipment used and particle characteristics measured and reported (Knuckles et al. 2008, Campen et al. 2010, Cherng et al. 2011. It is important to note that citation of earlier publications for characterization of particles may not provide reliable information of the particle characteristic in the reported study unless the same vehicle of administration is utilized. The studies on SRM2975 suggest some variation in the reported particles size, although the studies have used a similar vehicle. The hydrodynamic diameter of SRM2975 in saline has been reported as 189 ± 16 nm (Labranche et al. 2012), 257 ± 46 nm (Miller et al. 2013) and 280 ± 9 nm (Vesterdal et al. 2014). The latter also measured the hydrodynamic particles size in water (179 ± 11 nm) and cell medium (197 ± 6 nm) (Vesterdal et al. 2014). There are subtle differences with regard to the addition of compounds that improves the dispersion, but these inter-study differences in particle size may also be due to different equipment and handling. For instance, it has been shown in an intra-laboratory trial that the variation in hydrodynamic particles size for the same NM was dependent on the investigator and equipment (Roursgaard et al. 2014). The magnitude of variation in particle size determination related to differences in vehicle, investigator's handling procedures and equipment has not been assessed in a thorough manner using the same material. Table 6 outlines the segregation of the studies into groups based on thoroughness of the particle characterization into four categories: ''none'' is defined as studies that have only reported the mass concentration/ dose of PM; ''some'' defines publications with additional information on the chemical composition, shape or size in dry form; ''well-characterized samples'' are exposures where the PM has been characterized in dry form either by the supplier (e.g. EHC-93 or the standard reference materials) or investigators. Finally, the highest level of characterization is obtained in the studies that have assessed the particle size distribution in the vehicle, which are either air or a liquid suspension. Furthermore, Table 6 also contains information on the mode of exposure (inhalation or instillation), dose-response relationship, CVD outcome and its significance (vasomotor dysfunction or plaque progression). The data indicate that the studies on air pollution particles have not assessed the dose-response relationship and in general there is poor characterization of the particle size distributions ( 2 ¼ 32.0, p50.001). The studies on DE (or DEP) and NMs have mainly used well-characterized PM exposures. There is, however, an uneven distribution of studies that have used inhalation (mainly air pollution particles) and instillation (mainly DEP and NMs) ( 2 ¼ 11.5, p50.01). Interestingly, the studies on ambient air pollution particles mainly show statistically significant effect on vascular outcomes as compared with the studies on DE (or DEP) and NMs (crude odds ratio: 7.1, 95% CI: 1.9-26.3) ( Table 6). Adjustment for the type of exposure (inhalation or instillation), assessment of doseresponse relationship and CVD endpoint (vasomotor dysfunction or plaque progression) did not affect the odds ratio (7.9, 95% CI: 1.8-35.3, logistic regression with particle characterization dichotomized into low level (i.e. ''none'' and ''some'') and high level (i.e. ''well-characterized'')). Consequently, the quality of the exposure characterization is not a strong independent predictor of the statistical result of the CVD outcome, whereas the locations of ambient air pollution particle exposures are apparently more important for a statistically significant effect (with the majority of the statistically significant effect of exposure observed following exposure to CAPs in Tuxedo and Columbus, and UF-CAPs in Los Angeles). It remains to be determined whether air pollution exposure is more hazardous in these locations as compared with other locations. Finally, observations from the NPACT study indicate differences in hazard across locations, highlighted by CAPs exposure in Tuxedo being more hazardous than exposures in other locations (Lippmann et al. 2013).

Summary
There is substantial difference in the quality of particle characterization across studies. Archetypally, studies on ambient air pollution particles have little information on particle size distribution, although some studies have reported data on the chemical composition of the PM. However, investigation on DE (or DEP) and NMs typically have more information on particle characteristics in either dry form or in the exposure vehicle. Nevertheless, currently, it is not possible to determine whether other metrics than the mass dose/concentration are useful predictors for the dose-response relationship between PM and CVD outcomes in terms of vasomotor dysfunction and progression of atherosclerosis.
Mechanistic link between pulmonary exposure to particulate matter and vascular effects

PM-induced pulmonary inflammation and CVD outcomes
It is a prevailing hypothesis that vascular effects emerge as a consequence of an inflammatory responses after pulmonary exposure to PM. The doses or dose-rates utilized in some studies on CVD outcomes have led to pulmonary inflammation. The most convincing evidence of spillover cytokine-mediated effect from the lung was shown in a study on i.t. instillation of EHC-93 in which increased levels of IL-6 in the lungs and serum were noted concurrently with a blunted vasorelaxation response to ACH in wild-type mice, whereas there were no particle-generated vasomotor dysfunction in IL-6 knockout mice (Kido et al. 2011b). However, there are numerous studies that have exposed animals to PM at doses that are not associated with pulmonary inflammation and still have shown vasomotor dysfunction.
In order to assess the association between pulmonary inflammation and CVD outcomes, the studies on pulmonary exposure of PM have been segregated into groups using doses that have yielded inflammation and doses that have not (Supplementary Table 6). In addition, the studies are further divided into short-term (i.e. less than a week) or long-term exposures (i.e. weeks to months). This analysis demonstrated an uneven distribution of studies that have shown altered vasoconstriction response after repeated low-dose exposures (Table 7); while studies that have used high doses and short duration of exposure have not observed PM-induced vasoconstriction. The distribution of studies that have shown altered endothelium-dependent or endothelium-independent vasorelaxation response was not different between the groups. Next, Table 8 shows a segregation of the studies into three groups based on the susceptibility of the animal model to develop CVD showing an altered vasoconstriction response in susceptible and high-risk models for CVD ( 2 ¼ 7.9, p50.05). Furthermore, endothelium-dependent vasorelaxation has predominantly been observed in studies using CVD-susceptible models ( 2 ¼ 12.3, p50.01). The data in Tables 7 and 8 do not support the hypothesis of pulmonary inflammation exacerbation of vasomotor dysfunction responses in animals that have been exposed to PM in the airways. Eight studies with dose-response 17/9 13/13 a The levels are ''None'': no information except the mass concentration or dose of PM, ''Some'': information on chemical composition, shape or primary particle size in dry form (e.g. elemental composition in CAPs exposure studies), ''Well-characterized sample'': information on both particle size (or shape) and chemical composition in dry sample (e.g. EHC-93), ''Well-characterized sample in vehicle'': information on aerodynamic or hydrodynamic diameter in the suspension vehicle (air or suspension). b There is an uneven distribution of studies with regard to particle characterization ( 2 ¼ 32.0, p50.001, DE and NM studies have been pooled for the analysis). c There is an uneven distribution of studies with regard to inhalation and instillation exposure ( 2 ¼ 11.5, p50.01). Supplementary Table 5 contains additional information on the segregation of the studies.
The studies investigating atherosclerosis have used either regular or high-fat diet and repeated exposures. Therefore, we have separated the studies according to the diet utilized. The low-dose exposures in animals fed a normal chow was associated with the lowest effect (SMD ¼ 0.47, 95% CI: 0.24-0.69), whereas high-dose exposures had higher effect size (SMD ¼ 1.30, 95% CI: 0.80-1.80). The same dose-effect relationship was not observed in the studies using a high-fat diet (low-dose: SMD ¼ 1.59, 95% CI: 0.62-2.56; high-dose: SMD ¼ 0.86, 95% CI: 0.50-1.09) ( Supplementary Figures 5 and 6). However, it should be emphasized that the analysis displayed heterogeneity, which was attributed to few studies with very high effect sizes (Ying et al. 2009a, Chen et al. 2013b, Li et al. 2013b). The omission of these studies from the analysis eliminated the intra-group heterogeneity and reduced the SMD in both the lowdose (SMD ¼ 0.79, 95% CI: 0.21-1.37) and high-dose (SMD ¼ 0.64, 95% CI: 0.28-1.00) groups of animals that were fed on a high-fat diet ( Supplementary Figures 7  and 8). There is also a statistically significant subgroup difference in studies using high doses which were associated with higher effect size than low-dose exposure in animals on normal chow (p50.05). The observations indicate that high-dose PM exposures, associated with pulmonary inflammation, produced a larger effect in animals on normal chow, whereas the exposure displays no dose-response relationship in animals on high-fat diet. Thus, the observations do not indicate that a high-fat diet is a predisposing factor for accelerated plaque progression in PM-exposed animals. The high-fat feeding may facilitate a faster development of atherosclerosis and, therefore, reduce the required period with PM exposure. In addition, the high-fat diet may also increase the lipid content in atherosclerotic plaques, which could make it easier to measure the plaque size and composition.
Collectively, this assessment of the effect of dose and duration of exposure does not indicate a coherent pattern of CVD outcomes. The vasorelaxation response is clearly affected by pulmonary PM exposure, but this appears to occur independently of lung inflammation. In addition, it is striking that high-dose exposure studies, associated with pulmonary inflammation, have not been associated with altered vasoconstriction, whereas lowdose exposure studies predominantly show vasoconstriction. Importantly, pulmonary inflammation is not Group 1 ¼ One/few high-dose exposure(s) that caused inflammation (duration less than one week). Group 2 ¼ repeated high-dose exposures that caused inflammation (duration from weeks to months). Group 3 ¼ One/few low-dose exposure(s) that did not cause inflammation (duration less than one week). Group 4 ¼ repeated low-dose exposure that did not cause inflammation (duration from weeks to months). Supplementary Table 6 outlines studies that were included in the various groups. Data on vasomotor function endpoints are number of studies (altered/unaltered). Differences in distribution have been assessed using 2 -test. a Type of animal model group refers to studies on vasomotor dysfunction (normal/susceptible/disease model). b There is an uneven distribution of studies showing altered and unaltered vasoconstriction response between the four groups ( 2 ¼ 11.8, p50.01). This is mainly driven by an uneven distribution of studies in groups 1 and 4 (highlighted in bold). Plaque progression results are standardized mean difference with 95% confidence interval from the analysis with exclusion of 3 studies with high effect size. The raw data are reported as Forest plot in Supplementary  Figure 7. There are no results on plaque progression following a single/few exposure(s). required for vasomotor dysfunction responses and plaque progression, further highlighted by observations on fine size TiO 2 and residual oil fly ash that showed no association between pulmonary inflammation and vasomotor dysfunction, yet notable local oxidative stress and vasomotor responses in blood vessels (Nurkiewicz et al. 2004(Nurkiewicz et al. , 2006. The studies on plaque progression showed a dose-dependent effect in animals fed a regular chow.

Systemic inflammation as link between PM exposure and CVD outcomes
Chronic low-grade systemic inflammation has been explored extensively as causal risk factor for particleinduced atherosclerosis. The definition of chronic lowgrade systemic inflammation is somewhat elusive, although it is typically a minimum of a two-fold increase in plasma cytokines or acute phase proteins. It has been shown that a 2.7-fold elevated level of CRP (2.4 versus 0.9 mg/l as cutoff values from upper and lower tertile) was associated with an odds ratio of 2.1 (95% CI: 1.4-3.3) of non-fatal myocardial infarction and death from coronary heart disease in humans (Danesh et al. 2000). In another study, increased hazard ratio for coronary heart disease of 1.26 (95% CI: 1.08-1.46) for IL-6 and 1.14 (95% CI: 1.00-1.31) for TNF-a per 1-SD of log-transformed baseline values, which roughly corresponds to two-fold increases from baseline log-values, was noted (Kaptoge et al. 2014). A further re-analysis of data from 52 prospective cohort studies showed that inclusion of CRP levels to the statistical models of age, systolic blood pressure, smoking status, diabetes and cholesterol for prediction of 10-years CVD risk categories (low, medium and high) led to a re-classification of 1.5% of the subjects to new categories (Kaptoge et al. 2012). These findings indicate that changes in plasma cytokines and acute phase proteins (i.e. two-fold increases) is associated with increased risk of CVD outcomes, although the significance in terms of prediction of risk is small compared with other more established risk factors. In comparison, a review of studies on ambient air pollution particles showed a 1.07-fold (95% CI: 1.06-1.51) level of systemic inflammation markers in exposed humans (Møller et al. 2014). The same study showed that animals had a 1.28 (95% CI: 1.10-1.71) higher fold systemic inflammation after exposure to ambient air pollution particles (Møller et al. 2014). In the present review, nine studies showed the same effect on systemic inflammation biomarkers and vasomotor function outcomes, whereas no association was noted in five studies. Likewise, six studies reported a direct link between systemic inflammation biomarkers and plaque progression, whereas four studies did not (Supplementary Table 7). Thus, it appears that only about half of the studies have shown associations between systemic inflammation markers and CVD outcomes. This may further represent an overestimation of systemic inflammatory, due to non-altered plasma levels in some studies being omitted as uninteresting results. This notion is supported by the fact that all studies that have reported dissimilar responses in terms of systemic inflammation and CVD outcomes found unaltered levels of cytokines and acute phase proteins in plasma/serum. It has shown that weekly administration of IL-6 was associated with exacerbated atherosclerosis in ApoE À/À mice on high-fat (SMD ¼ 2.11, 95% CI: 0.31-3.91, 6 weeks exposure) and normal chow (SMD ¼ 1.72, 95% CI: 0.15-3.30, 21 weeks exposure),with a 7.8-fold and 4.9-fold increased plasma concentration of IL-6, and 3.3-fold and 2.0-fold increased TNF-a levels in high-fat diet and normal chow, respectively (Huber et al. 1999). The SMDs are similar to the effects that have been obtained in other studies on PM-induced atherosclerosis, but the fold-increases in plasma concentrations are substantially larger in this particular study. Likewise, ApoE À/À mice displayed accelerated atherosclerosis as compared with ApoE and TNF-double knockout mice (SMD ¼ 1.87, 95% CI: 0.81-2.93) (Ohta et al. 2005). This effect size was similar to that observed in PM-induced atherosclerosis, but it was accompanied by a marked difference in systemic levels of TNF-a.
It has been speculated that particles, which have passed from the airway space to the circulation, can cause systemic inflammation and oxidative stress. However, translocation of particles is very modest as documented from experiments on gold nanoparticles with a diameter of 80 nm that had less than 0.1% translocation in rats following i.t. instillation (Kreyling et al. 2014). This observation is identical to earlier studies showing that 15 and 80 nm iridium particles has less than 0.2% and 0.1% translocation in rats, respectively (Kreyling et al. 2002). Likewise, the translocation of MWCNT-7 to the liver, kidney, heart and brain was less than 0.01% and 0.04% of the deposited dose at 1 or 336 d post-exposure (Mercer et al. 2013). On the contrary, NMs with higher solubility (e.g. CeO 2 or quantum dots) have slightly higher systemic translocation, which can be attributed to passage of dissolved of constituents from the particles (Kermanizadeh et al. 2015). Observations from ex vivo exposure of vessel segments to PM have typically shown that concentrations within 10-100 mg/ml of PM cause vasomotor dysfunction (Bagate et al. 2004a,b, 2006, Hansen et al. 2007, Miller et al. 2009, Vesterdal et al. 2012. It is difficult to envisage that plasma PM concentrations, which are only a small fraction of the deposited pulmonary doses, will ever be sufficiently high enough to cause an effect to endothelial cells in the arterial circulation. As an example, a plasma concentration of 10 mg/ml can be reached from a deposited dose of 250 mg/kg in mice, assuming for 0.1% translocation in a 20 g animal with 1 ml of plasma. This dose is even higher than the doses that have been used in the studies of pulmonary instillation of PM. Collectively, there is no consistent evidence to support a notion that low-grade systemic inflammation is an important link between airway exposure to PM and CVD outcomes in terms of vasomotor dysfunction and accelerated plaque progression. Thus, the best explanation remains as a yet unidentified component in serum acting as a signaling factor between the local effect at the site of external exposure (e.g. lung epithelial cells or gastrointestinal mucosa) and arterial wall. This has been investigated in an ex vivo study where aorta rings were incubated with serum from PM-exposed mice. The serum from mice, which had been exposed to mixed vehicle exhaust or wood smoke, impaired the vasorelaxation response to ACH, whereas dust from roadway surfaces on residential streets and urban thoroughfare roads in Phenix and Tucson had no effect (Aragon et al. 2015).

Systemic levels of lipids and CVD outcomes
Cholesterol levels and to a lesser extent triglycerides in the blood are traditional risk factors for IHD in humans. Lipotoxicity is related to elevated levels of free fatty acids and associated with insulin resistance and endothelial dysfunction (Imrie et al. 2010). However, studies that have measured vasomotor function in long-term PM-exposed animals while simultaneously measuring triglycerides or cholesterol levels in plasma/serum have reported unaltered levels of the lipids (Sun et al. 2005, Folkmann et al. 2012, Miller et al. 2013, Miyata et al. 2013. The same analysis of studies on plaque progression also indicated inconsistent association to responses in terms of cholesterol (four out of 12 studies showing consistent results) and triglyceride (three out of nine studies) levels in blood (summarized in Supplementary  Table 8).
Overall, there seems to be no consistent association between blood lipid levels and CVD outcomes. However, it can be speculated that total cholesterol and triglyceride levels lack sufficient specificity as proxy measures for atherogenic and toxic lipids. In addition, it has been described that mixed whole engine emission exposure to ApoE À/À mice (300 mg/m 3 , 6 h/d for 7 or 50 d) increased expression of lectin-like oxidized LDL receptors (LOX-1) in the aorta (Lund et al. 2011). Similarly, another study showed increased LOX-1 expression in the aorta tissue in male Kyoto rats after exposure to mixed ozone (0.4 ppm) and DE (2200 mg/m 3 ) for 16 weeks (5 h/d and 1 d/week), whereas there was no alteration in the levels of total cholesterol, triglycerides, HDL and LDL (Kodavanti et al. 2011).

PM exposure, inflammation and oxidative stress in vessel walls
A number of studies have investigated inflammation and oxidative stress as mechanism of vasomotor dysfunction in PM-exposed animals. In general, production of ROS has mainly been assessed ex vivo by incubation of tissue cross-sections on microscope slides with dihydroethidium, which is considered to represent superoxide anion radicals (mainly derived from enzymes in inflammatory cells (e.g. NADPH oxidase) or uncoupling of endothelial nitric oxide synthase). The ex vivo incubation means that there is no direct proof of production of ROS in the arterial wall, although it can be considered as an indicator of the potential prooxidant milieu in tissue. These radicals can react with nitric oxide to produce peroxynitrite, which is a highly reactive compound that is typically measured as 3-nitro adducts on tyrosine residues on proteins (3-NT). It has been shown that short-term DE exposure was associated with both vasoconstriction and increased superoxide anion radical production in coronary arteries of Sprague-Dawley rats (Cherng et al. 2011). Similarly, short-term airway exposure to nanosized TiO 2 was associated with increased superoxide anion radical production, increased 3-NT levels and vasomotor dysfunction in spinotrapezius muscle and subepicardial arterioles of Sprague-Dawley rats (Nurkiewicz et al. 2009, LeBlanc et al. 2010). On the contrary, i.t. instillation of nanosized carbon black was associated with vasomotor dysfunction, yet no alteration in 3-NT levels in the aorta of ApoE À/À mice was visible (Vesterdal et al. 2010). Additionally, long-term CAPs inhalation has also been associated with increased activity of NADPH oxidase, 3-NT formation, superoxide anion radical generation and vasomotor dysfunction in the aorta of Sprague-Dawley rats and ApoE À/À mice (Sun et al. 2005, Ying et al. 2009a). Moreover, increased 3-NT staining in the aorta was observed after long-term inhalation of gasoline engine emission and particle-filtered exhaust (60 mg/m 3 , 6 h/d for 7 weeks), which did not generate pulmonary inflammation (Lund et al. 2007). The same authors also showed that 7 d exposure to gasoline engine exhaust (60 mg/m 3 , 6 h/d) or mixed vehicle exhaust (300 mg/m 3 ) was associated with increased production of superoxide anion radicals in the aorta of ApoE À/À mice (Lund et al. 2009(Lund et al. , 2011. Collectively, there is evidence indicating that exposure to PM is associated with oxidative stress in blood vessel walls. Several studies have measured biomarkers of oxidative stress in terms of oxidized lipids, DNA and proteins. Unfortunately, a number of assays for assessment of these biomarkers have poor quality. This pertains especially to certain methods for detection of lipid peroxidation products and nucleobase oxidation products in tissues, blood or urine (Møller and Loft 2010). Therefore, studies that have use these methodologies have not been included in review as they provide little concrete evidence of oxidative stress in the target tissue.

Summary
There is little evidence to support the hypothesis of pulmonary/systemic inflammation as the main causal link between pulmonary exposure to PM and CVD outcomes. The concurrent increases in cytokine levels in lung tissue and vascular dysfunction may only reflect parallel, yet unrelated, processes after exposure to PM. In general, there could be a bias towards reporting a positive association between pulmonary/systemic inflammation and CVD outcomes; with this mechanism of action potentially only discussed in publications with a positive association between inflammation and CVD outcomes. It is, therefore, often difficult to identify publications with selective omission of null results on biomarkers of inflammation. Furthermore, selective citation of publications showing the same association, or support the prevailing hypothesis by focusing only on the positive data, is another obstacle that may bias research (Møller et al. 2010b). The same limitation pertains to conclusions about other systemic biomarkers such as levels of lipids. Nevertheless, there is compelling evidence of oxidative stress and inflammation in the vessel wall and their importance in the development of CVD outcomes. It is very possible that inflammation and oxidative stress are transient effects after a short-term exposure to PM. Still, atherosclerosis is a pathological process characterized by a pro-inflammatory and prooxidant milieu. Therefore, a sudden increase of inflammation and oxidative stress after PM exposure may be associated or at least contribute to the rupture of unstable plaques, which in humans could be the final step in elicitation of a clinical manifestation such as myocardial infarction and ischemic stroke.

Limitations
In the present review, we have only assessed exposure to PM. At least, some studies on DE exposure have indicated that the gaseous components play an important role for vasomotor dysfunction and plaque progression (Quan et al. 2010, Vedal et al. 2013. It is currently uncertain whether the gaseous constituents play a role in the adverse effects of CVD following air pollution exposures, e.g. inhalation of CAPs. It should be stated that oxidizing gases such as ozone and nitrogen dioxide appear to have little direct effect on vascular endpoints or biomarkers in the circulation in controlled exposure studies (Hesterberg et al. 2009, Goodman et al. 2015. However, the inhalation of carbon monoxide (30-100 ppm to mimic fluctuation in urban air) for 4 weeks was associated with reduced vasorelaxation response in coronary arteries of rats, whereas there was no difference in the SNP response (Meyer et al. 2011). Nevertheless, it has been shown that serum from ozone-exposed mice contains vasoactive factors that can directly impair ACH-induced vasorelaxation (Robertson et al. 2013).
Here, the assessment of differences related to vasomotor function is relatively crude as we have used statistical significance rather than effect size as the outcome variable. The concentration-response relationship of vasoactive agents is typically fitted to a sigmoid curve. It is, therefore, possible to obtain the maximal response and the concentration that induces 50% response as outcome measures. Using this information, one can further investigate the mechanisms of vasomotor dysfunction, although re-analysis of original data is most likely necessary to standardize the data.
In this review, the analysis of the characterization of the active components in PM that causes CVD outcomes is imprecise. As previously discussed, many of the publications highlighted contain insufficient information about the particle size, shape and composition in relevant exposure vehicle to compare these metrics across studies. On the contrary, the NPACT studies deserve recognition for their efforts to establish which factors in ambient air pollution contribute to CVD (Vedal et al. 2013, Lippmann 2014. Despite best intentions, the NPACT study did not succeed in identifying the role of PM composition for vascular effects, which emphasizes the challenges in the research of PM-induced adverse effects. Hence, we have heavily focused on the generalizability of PM exposure rather than the cause-effect relationship of specific exposures. As such, there are only few studies that have actually investigated different types of exposures in the same experimental setting. This has been tested in studies on inhalation of CAPs (Gerlofs-Nijland et al. 2010, Quan et al. 2010), the i.t. instillation of carbon black and TiO 2 (Courtois et al. 2010) and SRM2975, SRM1648 and SWCNT (Vesterdal et al. 2014). Interestingly, none of these studies have documented substantial differences in the CVD outcomes by different types of PM. Nevertheless, it may not be possible to extrapolate the findings from the present analysis to other combustion-derived PM such as biomass or environmental exposures. At present, there seems to be relatively few studies on PM exposure from industrial or environmental sources. An example of such investigation is an interesting study on vasomotor dysfunction following exposure to PM from an Appalachian mountaintop mining site (Knuckles et al. 2013).
As a limitation, the analysis has only included vasomotor dysfunction and atherosclerosis as outcomes. Atherosclerosis is the underlying condition for myocardial infarction and ischemic stroke, but it does not necessarily lead to clinical manifestations. In fact, the progression of atherosclerosis can decelerate, stop or even reverse in humans if cardiovascular risk factors are avoided. Vasomotor dysfunction can occur in both healthy individuals and CVD patients after exposure to PM. CVD patients may already have vasomotor dysfunction and, therefore, diminished capacity to regulate blood flow. The additional impairment of vasomotor function by PM exposure may be the final trigger for clinical manifestations such as myocardial infarction or stroke. However, increased thrombosis tendency and dysrhythmia are also important descriptors of the acute risk of myocardial infarction, although they may not contribute to the chronic plaque development and deteriorated vasomotor function.
Finally, the quantitative analysis has been based on the mass concentration as dose metric. It should be stated that particle number concentration or surface area is more accurate dose metrics for UFPs and NMs. However, many publications do not provide sufficient information to assess the dose-response relationships on metrics other than the mass concentration.

Summary
The limitations of this review do not detract from the main conclusion formed based on the securitization of available literature stating that exposure to PM from urban air, DE (or DEP) and certain types of NMs affect vascular system by accelerating atherosclerosis and impair vasomotor function. The animal experiments are not perfect models for atherosclerosis and vasomotor dysfunction in humans. Nevertheless, animals and humans would be expected to show the same ranking of effect by different types of particles (i.e. if exposure to air pollution particles was more hazardous than NMs in animals, one would expect the same pattern in humans, irrespective of species differences in CVD development and progression). The lack of particle characterization data in studies on air pollution particles makes it impossible to compare studies on other metrics than the mass concentration/dose. The studies on NMs have relatively detailed description of particle characteristics, including size distribution, specific surface area and shape. Hence it is possible that future critical reviews, which specifically focus on CVD effects of NMs with a larger number of publications, may identify particle characteristics that are better predictors of CVD outcomes than the mass concentration/dose.

Conclusions
The results from the quantitative analysis in this review show that pulmonary and most likely other exposure routes to PM are associated with vasomotor dysfunction and progression of atherosclerosis. Further conclusions are as follows: (1) Airway exposure to PM is associated with augmented vasoconstriction and reduced endotheliumdependent vasorelaxation responses.
(2) Vasomotor dysfunction is observed in multiple vessels in the arterial tree of PM-exposed animals, including the aorta, mesenteric, coronary and carotid arteries. (3) Exposure to PM is associated with accelerated plaque progression in atherosclerosis-prone animals on either normal chow or high-fat diet. Accelerated atherosclerosis encompasses both increased plaque size and progression of plaques to more advanced stages. This effect is not consistently associated with altered lipid levels in plasma/serum that may promote a more pro-atherogenic environment. (4) Systemic or pulmonary inflammation is not a prerequisite for dysfunction in the vasomotor response and accelerated progression of atherosclerosis in PM-exposed animals. (5) Oxidative stress and inflammation have been observed in the arterial wall of PM-exposed animals with vasomotor dysfunction or plaque progression. (6) There has been a similar effect on atherosclerosis progression and vasomotor dysfunction after airway exposure to ambient air pollution particles and certain types of NMs, including TiO 2 , carbon black and CNTs. The exposure to DE (and DEP) has been associated with lower responses as compared with ambient air pollution particles and NMs. These observations cannot support the suggestion that all types of NMs per se are hazardous to the vascular system.
The observations from the present review point to several challenges in the future. The major limitation with regard to using the findings presented in risk assessment is a lack of dose-response relationship between exposures and CVD outcomes. Pooled analysis of the existing data is possible, although there are obvious challenges to such an approach because of the differences in methodology and exposures. There is a knowledge gap in the mechanism of CVD outcomes following PM exposure, which has been most clearly visible in studies of pulmonary exposure. A coherent linkage between local pulmonary effects and toxicity in the arterial wall (i.e. oxidative stress and inflammation) remains to be established. The role of mediators in the blood also requires further investigation. The few studies on gastrointestinal exposure to NMs have shown fairly clear associations with vasomotor dysfunction. Further investigations of the relationship between intake of NMs and CVD outcomes are warranted because oral exposure is a very likely route of exposure. This could be stimulated by the observation from the present review that NMs and combustion-derived PMs have similar hazards to the vascular system.
In summary, the analysis shows that certain NMs, including TiO 2 , carbon black and CNTs, have similar hazards to the vascular system as combustion-derived PM. In addition, airway exposure by inhalation or instillation to air pollution particles and NMs is associated with similar effect size on atherosclerosis progression, augmented vasoconstriction and blunted vasorelaxation responses in arteries. Exposure to DE (or DEP) is associated with lower vascular responses as compared with air pollution particles and NMs. There is not consistent evidence of pulmonary/systemic inflammation and pro-atherogenic plasma lipid profile as intermediate step in vasomotor dysfunction and progression of atherosclerosis in PM-exposed animals. There is, however, experimental evidence of oxidative stress and inflammation in the arterial wall of PM-exposed animals. From the data, it is clear that exposure to air pollution particles from traffic vehicles is hazardous to the vascular system, leading to clinical manifestations and mortality due to IHD. This implies that NM exposure may also be associated with the same diseases in humans, although this has not yet been investigated in observational studies.