We previously reported that in-vehicle exposure to PM2.5 was associated with increases in markers of inflammation and coagulation, and modulations of heart rate variability in Highway Patrol troopers . Here we demonstrate that most health endpoints were associated to a PM2.5 source factor that reflects speed-changing traffic conditions (dominated by copper, aldehydes and sulfur). Under such driving conditions, copper reflects wear of brakes, aldehydes reflect emissions from accelerating vehicles and sulfur reflects secondary aerosols and possibly diesel combustion products.
In Model A, four principal factors of PM2.5-exposure inside the patrol cars were identified. Their loadings suggest the main in-vehicle sources of PM2.5. Factor 1 reflects exposure to crustal material from the soils in the study region and the road surface ("crustal" factor). Factor 2 represents wear and tear of mechanical automotive parts, mostly chrome-titanium steels ("steel wear" factor). Factor 3 represents components derived from gasoline combustion ("gasoline factor"). Finally, factor 4 is characterized by components expected from speed-changing traffic ("speed-change" factor): copper from brakes  and aldehydes from engine emissions . Note that photochemical processes  are an unlikely source for this factor, since urban background and roadside levels near free-flowing traffic were much lower than in-vehicle levels . The source of the high sulfur loading is unclear. Diesel combustion of accelerating trucks would be a plausible source candidate. However, sulfur is ubiquitous on secondary urban aerosols. It was the most concentrated element on PM2.5 in the study . This prevents the identification of local sources with sulfur as a tracer. A cautious interpretation might be that factor 4 reflects particles from speed-changing traffic mixed with secondary urban particles; a mixture expected on roads in an urban-sprawl area like Raleigh.
Model B proposes only three sources. However, they correspond in principle to the sources from Model A, except that the factors " crustal" and "steel wear" seem to be collapsed into a single source factor "road surface". This notion is supported by the good correlation between the corresponding factors.
The average PM2.5 concentration of ca. 23 μg/m3 inside the vehicles was at a moderate level compared to the 24-hour National Ambient Air Quality Standard for PM2.5 of 65 μg/m3. The two methods used to measure PM2.5 were highly correlated. The differences of their correlations to the components (Table 1) and to the source factors (Tables 2) reflect the fact that two different methods were used to assess the particle mass : PM2.5Lightscatter reflects mostly accumulation mode particles (0.2 to 2 μm), whereas PM2.5Mass includes some coarse dust including fine sand.
The "speed-change" factor (Models A and B) was significantly associated with increased percentage of neutrophil leucocytes in the circulating blood, with decreased percentage of lymphocytes, and with changes in markers of endothelial activation and hemostasis. Endothelial cells are a major storage site for von Willebrand factor , and plasma levels of vWF serve as markers for endothelial activation . Protein C is an antithrombotic agent, it is activated on the endothelium and reduced in the blood after inflammatory stimulation due to protein C consumption . Consequently, endothelial cells may be involved in both inflammatory and coagulatory responses to traffic particles. Blood urea nitrogen was also associated with the "speed-change" factor. This finding would be consistent with the postulated inflammation since blood urea nitrogen increases several hours after an inflammatory stimulus (pig model) . Blood urea nitrogen and vWF lost significance when controlled for all potential confounders together, although the effect estimates were not much changed. It should be noted that including this many confounders into a model with a relatively small number of samples reduces the strength of the statistics considerably.
The "speed-change" factor was significantly associated with changes in MCV (independent of osmolality) and similar to the association for PM2.5Lightscatter with MCV reported earlier . The present analysis suggests that particles originating from speed-changing traffic are an important source of this association with circulating red blood cell mean volume (while other red blood cell indices were not affected). This is consistent with in-vitro blood experiments, where high concentrations of particles caused dose-dependent hemolysis, which was explained by oxidative damage to the membranes . Note that MCV increases with increasing doses of hemolytic chemicals . Future studies might answer the question whether particle-induced oxidative stress caused the association observed between MCV and the "speed-change" source factor.
The heart beat interval MCL increased in association with the "crustal" factor and the "speed-change" factor. Additionally, the "speed-change" factor was associated with significant increases in heart rate variability (SDNN and PNN50) and frequency of supraventricular ectopic beats. This cardiac response suggests increased vagal tone mostly in response to "speed-change" traffic particles. Fluctuations in autonomic tone have been associated with the triggering of atrial arrhythmias . Such fluctuations might also help explain the reported association between air pollution exposure and increases in arrhythmias in patients with an implanted cardioverter defibrillator .
The concentrations of particles and components in this study were low. Direct systemic effects seem therefore unlikely. However, the proposed endothelial activation could provide a link to pathological processes and the associated increase in cardiovascular morbidity and mortality , as follows: Once particles are deposited on the surfaces of the airways or alveoli, toxic products can quickly leach out or be produced on the surface of the particles. Given the small volume of surface liquid, this can result in high local concentrations. Copper and other transition metals can cause oxidative stress  and have been associated with inflammatory lung injury in human subjects  as well as airway epithelial cell injury in vitro . This oxidative stress might induce responses in the adjacent cells. In the alveolar region, the distance to the capillary endothelium is about 100 nanometers. Liberation of pro-thrombotic and pro-inflammatory mediators are well-described consequences of oxidative stress to endothelial and other cells . Inflammatory stimuli also might induce a vagal response . The components copper, sulfur and aldehydes dominated the "speed-change" factor. They seem to merit further attention in future targeted studies on particle toxicology.
Surprisingly, the "steel wear" factor of Model A was not associated to any inflammatory markers, although metal content of particles has been reported to be associated with inflammatory processes [24, 25, 28]. It would be interesting to study such wear particles with regard to size and solubility of metals.
One limitation of this study is the fact that only the association between the mean exposure during the evening shift and the response on the following morning was studied. This design ensured that potential diurnal variations of exposure and health parameters could not mimic a dose-response association, and that the exposure inside the cars was followed by a long unexposed resting period. However, it cannot be excluded that exposures and follow-up at other times of the day could have resulted in different dose-response estimates. Another limitation is the study population, since the troopers were a homogenous group of young, healthy, non-smoking people in excellent physical condition. Consequently, it is possible that the relative response such as the %-increase of inflammatory blood components or ectopic heart beats might be different in the troopers as compared to what could be expected in the general population or in individuals with elevated cardiovascular risks that have higher baseline levels. A final limitation is that the source factor "speed-changing" traffic does not represent a single source but rather a combination of closely related sources such as break wear and engine exhaust products. Answering the question, which of these sub-sources was causing the effects, would require a larger number of subjects or targeted toxicological studies.