Oxidative stress-induced telomeric erosion as a mechanism underlying airborne particulate matter-related cardiovascular disease

Studies providing direct measures of oxidative stress

As noted previously, oxidative stress in this paper has been defined as an imbalance between the generation and removal of free radicals. Air pollution studies have used several direct markers of this oxidative stress response. These include the following:

• circulating biomarkers of erythrocyte antioxidant activity in blood, such as glutathione peroxidase-1 (GPx-1) and copper-zinc superoxide dismutase (Cu, Zn-SOD).

• 8-hydroxy-2’-deoxyguanosine (designated as either 8-OHdG or 8-oxodG) in urine or in lymphocytes, a product of oxidation of the deoxynucleotide pool, and of oxidation of guanine in DNA.

• increases in redox-sensitive transcription factors.

• differences in oxidative stress-related gene activity among subjects exposed to pollutants.

Delfino et al. [74, 75] examined elderly non-smoking retirees having coronary artery disease and living in several centers across the Los Angeles area. Pollution monitors were located both inside and outside residences, and measures of oxidative stress included circulating biomarkers in blood (GPx-1, Cu, Zn-SOD). These are among the few studies that considered secondary organic aerosols (SOA), which in Los Angeles would mainly be oxidation products of carbonaceous gases of vehicular origin, among the different PM_2.5 species being considered. The studies found associations for Cu, Zn-SOD with primary PM_2.5 species mainly of vehicular origin (EC, BC, primary organic carbon, PM_0.25), but not SOA or mixed secondary and primary organic species (OC). Inorganic secondary aerosols, such as sulfate, were not monitored, presumably because major sulfate sources, such as coal-fired power plants, were several hundred miles downwind from the study area.

Loft et al. [77] examined 8-OHdG levels in 57 bus drivers in Copenhagen, 30 who drove urban routes and 27 who drove rural/suburban routes. 8-OHdG excretion was significantly elevated in the former compared to the latter. Although urban atmospheres appeared to be the cause of oxidative stress, this study did not assess specific vehicular-derived emissions.

Wei et al. [79] examined pre- and post-work shift 8-OHdG levels in security workers in Beijing, whose work site was on a heavily trafficked road. Four “clusters” of PM species were examined: PM_2.5 mass, PAHs, metals, and polar organic species. Observed post-shift increases in urinary 8-OHdG were associated with increased levels of PM_2.5 mass, PAHs, and metals, but not polar organic species, at the work site.

Lee et al. [80] studied 28 diesel exhaust inspectors in Taiwan, age and gender matched with office worker controls, for three consecutive work days. Pollutants examined were total diesel PM_2.5 (DEP_2.5) and PAH content. 8-OHdG levels were significantly higher for inspectors vs. controls on days 2 and 3. Increased PAH concentrations in DEP_2.5 were significantly associated with increased 8-OHdG in the exposed group even after adjusting for factors such as smoking.

Lai et al. [81] studied 47 female toll workers in Taiwan, matched with female office workers as controls; pollution concentrations were not measured. Non-smoking toll workers had significantly higher 8-OHdG levels than did non-smoking controls; increases in 8-OHdG were almost 5 times higher per 1000 trucks/buses than per 1,000 cars (comparing car toll lanes with truck/bus lanes), although these results did not reach statistical significance.

Sauvain et al. [78] studied 32 Swiss bus maintenance workers. PM species examined included PM₄, OC, EC, metals (Fe, Mn, Cu), and PAHs. Urinary 8-OHdG concentrations increased within each work shift, and between two consecutive work days. Among non-smokers, increased 8-OHdG concentrations were associated with increased levels of OC and Cu. While secondary OC associations were not found using a different measure of oxidative stress (blood levels of Cu, Zn-SOD) in the Delfino et al. studies [74, 75], primary OC was significantly associated with oxidative stress biomarkers, suggesting that the OC associations in Sauvain et al. reflect mainly fresh (i.e., primary) diesel emissions.

Sorensen et al. [82] used personal monitors to determine levels of PM_2.5 and six metals (V, Cr, Pt, Ni, Cu, Fe) in PM_2.5 fractions to which 49 young, non-smoking students in Copenhagen were exposed. No associations were found with 8-oxodG in urine, but significant positive associations were found for V and Cr with 8-oxodG in lymphocytes.

Kim et al. [76] examined association of urinary 8-OHdG with six PM_2.5 metal species (V, Cr, Mn, Ni, Cu, Pb) among boilermakers repairing oil-fired boilers in the Boston, MA area. Post-work shift levels of 8-OHdG were significantly higher than pre-work shift levels. Increased 8-OHdG was associated with increases in total PM_2.5 and V, Mn, Ni, Pb, but not Cu or Cr. PM levels were considerably higher than ambient; for example, the concentration of V was 1.23 μg/m³, contrasted with contemporary Boston ambient V levels of 3 to 5 nanograms/m³. It is interesting that Cu, but not Mn, was associated with oxidative stress in Sauvain et al. [78], whereas opposite associations were observed for each in Kim et al.

Allen et al. [83] studied 10 adults, aged 18-49, with metabolic syndrome but no history of ongoing medical care for heart disease, hypertension, asthma, diabetes, or similar chronic conditions. Subjects were exposed for two hours either to diesel exhaust (200 μg/m³ PM_2.5) or filtered air. No significant associations were found between urinary 8-OHdG and diesel exhaust exposure. These findings contrast with studies noting increased urinary 8-OHdG with work shift exposure to diesel exhaust in Taiwan [80] and Switzerland [78], or to heavy vehicular traffic in Taiwan [81] and Beijing [79]. The lack of significant findings of oxidative stress in Allen et al. [83] might have to do with shorter exposure times, with differences in makeup of pollution, with differences in levels of activity, i.e., workers presumably were more physically active during exposure than passively exposed subjects in Allen et al., as suggested by the authors, or to other as yet unknown reasons.

Jacobs et al. [84] chose a unique method of assessing pollution exposure, i.e., by measuring the area occupied by carbon in respiratory tract macrophages, in 79 Belgian non-smoking adult diabetics. The authors also examined distance between subject residence and major roads by geocoding. The biomarker measured was oxidized LDL in plasma, a risk factor for atherosclerosis. They found that both increased carbon loading and decreased distance from major roads were significantly associated with higher levels of oxidized LDL. Similarly, using an animal model, namely atherosclerotic apolipoprotein E knockout mice, Lund et al. [93] found that acute exposure to a mixture of gasoline and diesel engine emissions resulted in increased vascular and plasma markers of oxidative stress and the expression of proatherogenic factors.

Kipen et al. [85] exposed healthy young subjects in New Jersey either to diesel exhaust or to a laboratory-generated PM_2.5 secondary organic carbon atmosphere (~ 190 μg/m³ each), with filtered air control for each exposure. Each pollutant atmosphere caused an increase in white blood cell proteasome activity, and the diesel exhaust caused an increase in red blood cell proteasome activity as well, which the authors suggested to indicate oxidative stress. The Delfino et al. studies [74, 75] in Los Angeles did not find oxidative stress associated with secondary organic carbon, but the laboratory-generated SOA atmosphere of Kipen et al. did result in such an association. The SOA generated in the laboratory used vegetative organic gases, while the ambient SOA in Los Angeles presumably is mostly from vehicular-derived organic gases. Whether differences in chemical composition due to the differences in source or differences in methods of assessment of oxidative stress, or whether the higher concentrations in the laboratory SOA compared to ambient levels can explain the disparate findings between the Kipen et al. and Delfino et al. studies, is not clear.

While all of the studies noted above directly examined markers of oxidative stress, the results, however, cannot be neatly summarized for several reasons:

• Several studies, in particular those involving urban emissions, did not measure PM_2.5 species, and only a few measured PM_2.5 metals, which have a high potential for inducing oxidative stress [3].

• Studies were done in several countries, with varying levels and likely composition of urban air pollutants.

• Although most of the studies used urinary 8-OHdG as a biomarker of oxidative stress, some studies used other markers, and it isn’t clear a priori whether oxidative stress impacts of PM on telomere erosion are equivalent for various markers.

Given these limitations, we offer the following assessment of these studies:

• In studies performed in the U.S., primary vehicular emissions, such as EC/BC and primary OC, were associated with increases in biomarkers of oxidative stress, but secondary organic carbon, presumably mostly of vehicular origin, was not.

• In studies performed in Europe and Asia, oxidative stress measures were associated with greater exposure to ambient emissions from roads or at tolls (vs. office worker controls); for those driving in urban vs. suburban/rural areas or living closer to roads; for those with higher levels of carbon in airway macrophages; or for those with greater exposure to emissions found in work locations involving diesels or buses. While exposure to these different atmospheres dominated by vehicular emissions clearly resulted in oxidative stress, little information was generally provided as to which specific components of such atmospheres may have been involved. Only one study examined specific PM_2.5 species, finding associations with OC and Cu, but not EC, Fe, or Mn in bus maintenance workers.

• Exposures to atmospheres particularly high in PM_2.5 metals (boilermakers) or to ambient levels of specific metals such as V were associated with oxidative stress. However, only three studies examined metals, and associations were not always with the same metals.

Overall, the studies reviewed in this section suggest that exposure to specific PM sources, namely vehicular, are associated with oxidative stress, but because few of the studies included specific PM species, it isn’t possible as yet to determine which species were most likely responsible for the oxidative stress. Studies in the next section, using a different proxy for oxidative stress, namely change in heart rate variability (HRV), may provide more details on the potential for several PM species to be so associated.

Studies providing measures of HRV

Heart rate variability is a reflection of the autonomic nervous system control of cardiac function. Altered HRV has been noted in patients with hypertension and heart failure, and has been linked to increased CVD mortality [92]. Cascio [94] noted that increased cardiac-related deaths and decreased autonomic control of cardiac function can be modulated, at least in part, by oxidative stress. Furthermore, an apparently causal relationship between oxidative stress and changes in various measures of HRV has been found with exposure to pollutants in a number of studies, as discussed below.

Schwartz et al. [95] found significant decreases in the HF component of HRV in people having a deletion for the protein GSTM1, part of the cellular defense against oxidative stress, after exposure to PM. For those that did not have the deletion, however, there were no changes in the HF component. Use of statins protected those with this genetic deficiency against oxidative stress associated with exposure to air pollution. Chahine et al. [96] also found that only those people lacking genes protective against oxidative stress, including a different genetic variant than examined by Schwartz et al. [95], experienced reductions in two measures of HRV after PM exposure compared to normal controls. They concluded that oxidative stress was an important pathway for the effects of particles on autonomic nervous function. This is supported by an animal study [97], in which a significant increase in the SDNN component of HRV, which reflects beat-to-beat intervals, and a significant decrease in the ratio of high frequency (HF) to low frequency (LF) components 30 minutes after exposure to urban air particles by instillation was associated with increased oxidative stress in hearts of rats; however, SDNN remained at normal levels when increased oxidative stress was prevented by administration of anti-oxidants prior to exposure (LF/HF was not analyzed in this regard). Finally, Probst-Hensch et al. [98] found that individuals deficient in “oxidant-scavenging” glutathione S-transferase (GST) genes had significant HRV alterations, compared to those not so deficient.

Taken together, these studies indicate that changes (mainly reductions) in HRV after exposure to pollutants can reflect induced oxidative stress. Brook et al. [2] noted that “…pathways that reduced endogenous oxidative stress have a protective effect that mitigates reductions in HRV due to ambient PM exposure.” Overall, these studies suggest that associations between oxidative stress and changes in measures of HRV are causative, and that such HRV changes can be considered to be markers of oxidative stress induced by pollutant exposure, in this sense parallel to increases in 8-OHdG found in studies discussed in previous sections. This section reviews studies of HRV changes in people for whom exposure to pollutants, such as BC, was reasonably well characterized, for instance, with use of personal monitors, with ambient monitors which traveled with subjects, with monitors in close proximity to a road to which nearby subject residences were also in close proximity, or by use of wind back- trajectory analysis.

Suh and Zanobetti [72] used both personal monitors and a central pollution monitor to assess exposure to BC, sulfate, and PM_2.5 in the Atlanta area. The authors found no significant associations with five different measures of HRV when central monitor readings were used. However, risk estimates were 3 to 45 times higher, and BC was significantly associated with changes in all five measures of HRV (Table 2), when personal monitors were used. Thus, for a pollutant such as BC, having large spatial variability across a metropolitan area, use of central monitor readings failed to accurately characterize differences in exposure to BC and understated the extent and significance of HRV changes and, thus, of related oxidative stress, relative to more accurate exposure estimates for BC. No HRV associations were noted for either sulfate (only personal monitor data available) or total PM_2.5.

Adar et al. [86] used a mobile monitor which followed elderly non-smoking retirees during their activities. Both BC and PM_2.5 concentrations were sharply higher when subjects were on buses than at other times. HRV changes (5 min and 24 h means) were essentially monotonic with changes in BC and PM_2.5 concentrations, and were significant in the great majority of cases for both pollutants.

Schwartz et al. [87] examined four different measures of HRV for 1 and 24 h periods. Subjects lived in apartments adjacent to a major Boston roadway; the monitor was also in close proximity to the roadway. BC was consistently associated with changes in all four HRV measures evaluated, while PM_2.5 was only occasionally associated. The authors then used an algorithm to remove BC from PM_2.5 each hour, and termed the remainder “secondary PM,” which contained regional sulfate and SOC but not BC. They found no associations with HRV changes and secondary PM. Furthermore, SDNN declined monotonically with increasing PM_2.5 from the lowest PM_2.5 level to about 18 μg/m³. At these relatively low levels, the investigators noted that PM_2.5 and BC were highly correlated. However, from about 20 μg/m³ to the highest PM_2.5 levels (about 50 μg/m³), the relationship between SDNN and PM_2.5 was flat, i.e., there were no changes in the SDNN measure of HRV. At these higher PM_2.5 concentrations, PM_2.5 was no longer well correlated with BC.

Ebelt et al. [88] examined HRV changes in 16 non-smoking patients with COPD in Vancouver, Canada. Both ambient and personal monitoring were used to obtain measures of PM_2.5; personal monitoring of sulfate was also performed. Non-sulfate PM_2.5 was determined by various algorithms which effectively subtracted personal sulfate exposure from PM_2.5 measures. Of the different measures of PM_2.5, only non-sulfate urban PM_2.5 was associated with changes in HRV (SDNN, R-MSSD). These results are the obverse of those of Schwartz et al. [87], in that Ebelt et al. subtracted sulfur to determine urban non-sulfate, which would include BC. However, in both studies, ambient PM reflecting BC or urban air, but not sulfate or regional air, was associated with altered HRV.

Creason et al. [89] examined the relationship between the HF and LF measures of HRV with PM_2.5 concentrations over 24 consecutive days in 56 elderly, non-smoking men. Sulfate was measured, and reported to average 14% of PM_2.5 concentrations, but was not reported separately from PM_2.5. The authors found a “U” shaped relationship between changes in HRV measures and PM_2.5; HRV changes were most pronounced near the midpoint of PM_2.5 concentrations, and HRV was most normal at the lowest and highest PM_2.5 concentrations. When a two day air mass, containing the highest and 3^rd highest PM_2.5 concentrations was removed from the analysis, for the remaining 22 days a monotonic reduction in HF was observed with increasing PM_2.5, similar to the reduction in SDNN observed with increasing BC by both Adar et al. [86] and Schwartz et al. [87]. The authors used back-trajectory analysis to show that the two day air mass, having PM_2.5 concentrations of approximately 53 and 37 μg/m³, came from rural Pennsylvania, whereas on the other 22 days, the authors suggested that urban or industrial PM_2.5 sources were present. The lack of association with PM_2.5 on days dominated by rural air masses parallels the findings of Schwartz et al. [87] that PM_2.5 in secondary air masses was not associated with changes in HRV, even though PM_2.5 levels were highest on such days.

Park et al. [90] studied 497 subjects using back trajectory analysis to create six different source directions to assess exposure. Of primary interest are the four directions which had very similar BC, PM_2.5, and sulfate concentrations (e.g., PM_2.5 between 14.1 and 15.1 μg/m³; BC between 0.99 and 1.1 μg/m³; and sulfate between 3.8 and 4.2 μg/m³), substantially higher than the remaining two directions (N and NW). These four directions were:

• SW (air mass passed over Philadelphia and New York before following the I-95 corridor toward Boston).

• S (air mass originated in the Carolinas and entered Massachusetts from the ocean to the S).

• W (air mass passed over northern Ohio and middle New York state, mostly north of Pennsylvania).

• Local (slow moving air masses, relatively stagnant over Boston).

On days when air masses were less urban (S and W), there were no HRV changes associated with any emissions. When air masses arrived in Boston from the SW, BC was significantly associated with changes in three of four HRV measures; no other PM_2.5 species were associated with such changes. When the air was slower moving and more local, PM_2.5, sulfate, and BC were all associated with HRV changes, although BC was associated more consistently.

For two other wind trajectories (N and NW), levels of all pollutants were low relative to the other four directions. An association with PM_2.5 was observed for the N wind trajectory, and one for sulfate was seen for the NW wind trajectory. The authors suggested that the effect apparently seen for sulfate in this case, despite low sulfate levels, may actually be due to the presence of transition metals. Lippmann et al. [99] found significant reductions in HRV in mice exposed to ambient air in upstate New York only when the wind trajectory was from the NW, passing over a Sudbury, Ontario nickel smelter.

Two studies that did not monitor PM_2.5 species are included in this section because they demonstrate that HRV changes can be abolished using a face mask that filtered out urban particles, suggesting that the cause of HRV reductions may be traffic-related PM as opposed to associated gaseous vehicular emissions. Langrish et al. [91] recruited 15 young, non-smokers in Beijing for a cross-over study. Subjects walked a predetermined course on separate occasions, either with or without a face mask. Personal PM_2.5 monitors were used. A significant decrease in 24-h SDNN was observed when subjects did not use face masks, suggesting production of oxidative stress. Significant reductions in LF were also observed, but interpretation of this effect wasn’t straightforward, since LF reflects tone of the sympathetic nervous system and it might have been affected to some extent by simply wearing a mask. Beijing PM_2.5 levels when face masks were used were 140 μg/m³ vs. 86 μg/m³ when masks were not used. In another study, Langrish et al. [92] made similar findings, using subjects with coronary artery disease (average age 62). In addition, they also studied other CVD endpoints, such as ST-segment depression and arterial blood pressure, and found that use of the face mask reduced both. Using electron paramagnetic resonance, the authors also established that ambient PM in Beijing has very high oxidative stress potential.

The results of the studies reviewed in this section may be summarized as follows:

• When there were reasonably accurate subject exposure measures, BC was consistently significantly associated with alterations in various measures of HRV. Among these studies, those able to distinguish BC from secondary air masses without BC, rural air masses from those with industrial or urban pollution, urban PM without sulfate vs. those containing sulfate, or personally monitored BC from personally monitored sulfate, HRV measures were influenced by BC or urban or industrial air, but not sulfate, regional air masses containing low levels of BC, or rural air masses, even though high in total PM_2.5. In these studies, PM_2.5 was consistently associated with HRV changes when it was highly correlated with BC, but less so otherwise.

• Although metals were not monitored in any of the HRV studies, the sulfate finding in Park et al. [90] for the air mass with low PM and low sulfate which passed near the Sudbury nickel smelter is suggestive of the Ni association found by Lippmann et al. [99], i.e., HRV was altered in a mouse model only when the wind trajectory was such that the air mass passed near the nickel smelter.

• Those pollutants that were associated with alteration in HRV were also associated with production of oxidative stress which, as noted, can mediate HRV changes.