Association of Biomarkers of Systemic Inflammation with Organic Components and Source Tracers in Quasi-Ultrafine Particles

Background Evidence is needed regarding the air pollutant components and their sources responsible for associations between particle mass concentrations and human cardiovascular outcomes. We previously found associations between circulating biomarkers of inflammation and mass concentrations of quasi-ultrafine particles ≤ 0.25 μm in aerodynamic diameter (PM0.25) in a panel cohort study of 60 elderly subjects with coronary artery disease living in the Los Angeles Basin. Objectives We reassessed biomarker associations with PM0.25 using new particle composition data. Methods Weekly biomarkers of inflammation were plasma interleukin-6 (IL-6) and soluble tumor necrosis factor-α receptor II (sTNF-RII) (n = 578). Exposures included indoor and outdoor community organic PM0.25 constituents [polycyclic aromatic hydrocarbons (PAHs), hopanes, n-alkanes, organic acids, water-soluble organic carbon, and transition metals]. We analyzed the relation between biomarkers and exposures with mixed-effects models adjusted for potential confounders. Results Indoor and outdoor PAHs (low-, medium-, and high-molecular-weight PAHs), followed by hopanes (vehicle emissions tracer), were positively associated with biomarkers, but other organic components and transition metals were not. sTNF-RII increased by 135 pg/mL [95% confidence interval (CI), 45–225 pg/mL], and IL-6 increased by 0.27 pg/mL (95% CI, 0.10–0.44 pg/mL) per interquartile range increase of 0.56 ng/m3 outdoor total PAHs. Two-pollutant models of PM0.25 with PAHs showed that nominal associations of IL-6 and sTNF-RII with PM0.25 mass were completely confounded by PAHs. Vehicular emission sources estimated from chemical mass balance models were strongly correlated with PAHs (R = 0.71). Conclusions Traffic emission sources of organic chemicals represented by PAHs are associated with increased systemic inflammation and explain associations with quasi-ultrafine particle mass.

volume 118 | number 6 | June 2010 • Environmental Health Perspectives Research Cardiovascular hospital admissions and mortality have been associated with ambient mass concentrations of fine particulate mat ter (PM) air pollution ≤ 2.5 µm in aerody namic diameter (PM 2.5 ) (Pope and Dockery 2006). Questions remain regarding the under lying causal chemical components and sources responsible for these associations. A recent timeseries study of 106 U.S. counties showed stronger associations of cardiovascular hospital admissions with countywide averages of PM 2.5 when there were higher fractions of elemental carbon (EC), nickel (Ni), and vanadium (V), suggesting that important sources included fossil fuel combustion, biomass burning, and oil combustion (Bell et al. 2009).
Unlike PM 2.5 , ultrafine particles (UFPs; generally defined as < 0.1 µm in diameter) are not regulated by the U.S. Environmental Protection Agency (EPA), yet this is the size fraction that may have the highest toxic potential because it has magnitudes greater number concentrations and surface area than the larger particles that dominate PM 2.5 mass (Oberdörster et al. 2005). On that large surface area, UFPs carry and deliver redox active organic chemicals, including polycyclic aromatic hydrocarbons (PAHs), to the respira tory tract in disproportionately higher concen trations than do larger particles (Ntziachristos et al. 2007), possibly leading to a cascade of effects related to oxidative stress and inflam mation in the lungs and at extrapulmonary sites . These and other effects could underlie associations of morbid ity and mortality with air pollutants.
Except for some studies with personal or microenvironmental air pollution data (Chan et al. 2004;Delfino et al. 2008Delfino et al. , 2009Folino et al. 2009;Vinzents et al. 2005), regional ambient air monitoring has been the primary data source used in epidemiologic research on the importance of UFP exposure to cardio vascular outcomes and circulating biomarkers in individuallevel studies (de Hartog et al. 2003;Henneberger et al. 2005;IbaldMulli et al. 2004;Lanki et al. 2008;Pekkanen et al. 2002;Rückerl et al. 2006Rückerl et al. , 2007Timonen et al. 2006). These studies of ambient air were all conducted in Europe, and UFPs were measured as particle number concentrations at central regional sites. Exposure error from the use of ambient data is likely, because air moni tors may be far from subject locations and subjects may be exposed to pollutants from local sources, including traffic. UFPs have much higher spatial variability than does PM 2.5 , so exposure error is likely. In addition, UFP mass and particle number do not specifically indicate which particle com ponents or sources are important, although generally in urban areas UFP compositions are dominated by organic chemicals and EC and originate from combustion sources.
We conducted a panel cohort study of elderly subjects with a history of coronary artery disease living in the Los Angeles Basin. This is considered a population that may have among the greatest susceptibility to the adverse effects of air pollution (von Klot et al. 2005). We made repeated measurements of blood biomarkers and air pollutant exposures. To assess the potential importance of UFPs to cardiovascular health, we measured quasi ultrafine particle mass < 0.25 µm in diameter (PM 0.25 ). To address the issue of exposure error, we monitored PM 0.25 at the retire ment communities of subjects. We previously reported positive associations of blood bio markers of inflammation with PM 0.25 but not with larger accumulationmode particles 0.25-2.5 µm in diameter (PM 0.25-2.5 ) (Delfino et al. 2009). However, particle mass alone does not provide sufficient information about com position or sources. We also previously found positive associations between the biomarkers and PM 2.5 EC (Delfino et al. 2009). Based on these findings, we hypothesized that traf fic emission sources of organic chemicals in PM 0.25 would be positively associated with systemic inflammation. In the present analysis, we aimed to better delineate which underlying PM components may be responsible for the associations we observed for EC and PM 0.25 mass using new data on chemical species in the archived PM 0.25 filter samples.

Materials and Methods
Population and design. This was a longitudinal study of repeated measures where each subject acted as his or her own control over time. This limits the impact of confounding by between subject characteristics. We recruited subjects from four retirement communities. Subjects were eligible for participation if they had a con firmed coronary artery disease history and were ≥ 65 years of age, nonsmokers, and unexposed to environmental tobacco smoke. Of 105 vol unteers, 21 were not eligible, 19 dropped out, 2 had too few blood draws or valid biomarker data (> 5 of 12 weeks), and 3 had insufficient biomarker data due to exclusions for frequent infections, leaving 60 subjects. We excluded biomarker measurements during weeks with acute infectious illnesses given their wellknown impact on measured biomarkers. Table 1 lists subject characteristics.
Two retirement communities were stud ied in 2005-2006 and two in 2006-2007. Subjects were followed for a total of 12 weeks with weekly blood draws for circulating bio markers of inflammation in plasma. Each subject contributed 5-12 weekly blood draws (n = 578 total samples).
Each community was studied in two 6week seasonal phases, a warmer period charac terized by higher photochemistry fol lowed by a cooler period characterized by higher air stagnation and lower mixing heights. This seasonal approach was intended to increase the variability in pollutant char acteristics, with higher secondary organic aerosols (SOAs) in the warmer phase and higher primary organic aerosols (POAs) in the cooler phase when trafficrelated air pollutants increase at ground level. POAs are formed during or shortly after the combustion of fossil fuels. SOAs are largely photochemically pro duced from gastoparticle conversion when volatile reactive organic gases from anthropo genic and biogenic sources, and anthropogenic semivolatile organic compounds (SVOCs), are oxidized to form lowvolatility products that condense to produce SOAs. There are few data on the importance of variations in this multipollutant characteristic of PM to human health outcomes. In the present study, POAs are represented by PAHs and hopanes, whereas SOAs are represented by watersoluble organic carbon (WSOC) and organic acids. Most PAHs are considered to be components of POAs. Hopanes are found in the lubricant oils of diesel and gasoline vehicles and are thus tracers of primary vehicular aerosols in the Los Angeles Basin (Schauer et al. 1996(Schauer et al. , 2000. WSOC (Snyder et al. 2009) and organic acids (Robinson et al. 2006) are tracers of SOAs, although a fraction of WSOC comes from biomass burning (Docherty et al. 2008).
The research protocol was approved by the Institutional Review Board of the University of California-Irvine, and we obtained informed written consent from subjects.
Biomarkers. We focused on an informative set of biomarkers of inflammation from the previous analysis of peripheral blood biomark ers and PM 0.25 mass (Delfino et al. 2009). We drew blood samples in ethylenediamine tetra acetic acid tubes on Friday afternoons and processed them and froze the plasma on site within 30 min. Samples were stored at -80°C until assayed. Plasma biomarkers were thawed and assayed using 96well immuno assay kits for the proinflammatory cytokine interleukin6 (IL6) and the cytokine receptorsoluble tumor necrosis factorα (TNFα) receptor II (sTNFRII; Quantikine HS, R&D Systems, Minneapolis, MN). sTNFRII has a longer halflife than TNFα (Aderka 1996) and may thus better reflect sustained or lagged effects. Thawed erythrocyte lysates were assayed spectrophotometrically for activities of the antioxidant enzymes copper/zinc superoxide dismutase (Cu,ZnSOD) and glutathione per oxidase1 (GPx1) (Cayman Chemical, Ann Arbor, MI, USA). Cu,ZnSOD and GPx1 values were normalized to units per gram of hemoglobin. These and related biomarkers are predictive of cardiovascular disease risk (FloresMateo et al. 2009;Kritchevsky et al. 2005;Pai et al. 2004).
Exposures. The methods used to measure components and their relevance to sources of PM 0.25 are described in detail in the Supplemental Material [Chemical Measurement Methods (doi:10.1289/ehp.0901407)] and by Arhami et al. (2010). There we also discuss in detail differences by season and community and describe the relation between indoor and outdoor measurements.
Air sampling occurred in the immediate outdoor environment of each retirement com munity and at an indoor site located in the common areas of the main community build ings. The indoor data are thus representative to some degree of the same indoor environment of each subject. Our main interest here is in the effects of outdoorsource PM components.
More than 5 days before each blood draw, we collected indoor and outdoor sizesegre gated particle samples using Sioutas Personal Cascade Impactors (SKC Inc., Eighty Four, PA, USA) with Zefluor filters (3µm pore size; Pall Life Sciences, Ann Arbor, MI, USA). We evaluated components only in the qua siultrafine fraction (PM 0.25 ). Mass concen trations were determined gravimetrically by weighing the impactor filters and substrates with a microbalance (uncertainty, ± 2 µg; MettlerToledo, Columbus, OH, USA) in a temperaturecontrolled and relative humiditycontrolled room.
Values are mean ± SD or n (%). a Each category is hierarchical and excludes being in the above diagnostic category. b Includes subjects with anginal symptoms relieved with nitrates plus echocardiogram and electrocardiographic evidence of past infarct.
volume 118 | number 6 | June 2010 • Environmental Health Perspectives The first composited quartersection was digested with concentrated acid using micro wave digestion followed by analysis to deter mine 52 trace elements using highresolution inductively coupled plasma mass spectrometry (Finnigan Element 2; Thermo Fisher Scientific, Waltham, MA, USA) (Herner et al. 2006). We focused our analyses of exposure-response relationships on key transition metals that can generate reactive oxygen species by Fenton type reactions: vanadium (V), chromium (Cr), iron (Fe), nickel (Ni), copper (Cu), manganese (Mn), lead (Pb), and zinc (Zn).
The second composited quarter was ana lyzed for WSOC using a General Electric Sievers Total Organic Carbon Analyzer (GE Analytical Instruments, Boulder, CO, USA).
The remaining composited half was analyzed for organic tracer compounds by GC/MS along with field blanks, labora tory blanks, spiked samples, and standard reference material (Urban Dust Standard Resource Material 1649a; National Institute of Standards and Technology, Gaithersburg, MD, USA). Spike recovery after correction for internal standard recoveries was in the range of 96-110% for PAHs, 99-104% for hopanes, and 68-136% for nalkanes. Blank concen trations of MMW PAHs, HMW PAHs, and hopanes were below analytical detection lim its (~ 10 pg/m 3 air). The method detection limits for remaining compounds were limited by field and laboratory blanks. Uncertainties for each measurement were estimated based on analytical uncertainties and uncertainties from the blank correction and were used to determine if each measurement was statisti cally different from zero. The precision of the spike and standard reference material analyses was used to estimate method precision (> 20% for all PAHs, hopanes, and nalkanes).
Statistical analysis. We analyzed relations of repeated (withinsubject) measures of bio markers to air pollutant exposures with linear mixed effects models. Random effects were estimated at the subject level, nested within seasonal phase and community, to account for correlated withinindividual repeated meas ures. To focus estimates of associations at the subject level, we adjusted for between community and betweenphase exposure effects as proposed by Janes et al. (2008) by using exposures that were meancentered across community and phase [see Supplemental Material, Regression Model, Mean Centering Method (doi:10.1289/ehp.0901407)]. We decided a priori to adjust for 5day average temperature. Magnitudes of association from the mixed models are expressed at pollutant interquartile ranges (IQRs; 25th-75th per centile) to allow strengths of association for different pollutants to be compared by limit ing differences due to units of measurement or concentration range.
We evaluated the covariance structure using empirical variograms and found models were best fit as an autoregressive1 correla tion structure. We performed residual analy ses to examine deviations from standard linear   mixed model assumptions and the presence of influential observations. We found four influential high outliers for IL6 > 10 pg/mL that were reset to 10 pg/mL (upper limit of its standard curve) to obtain more representative estimates of association. In a model for 5day average PM 0.25 , including the outliers resulted in an association of 0.41 pg/mL [95% con fidence interval (CI), 0.00-0.82] per inter quartile change in PM 0.25 of 7.37 µg/m 3 , whereas resetting them to 10 pg/mL resulted in an association of 0.26 pg/mL (95% CI, -0.06 to 0.57). It is important that in the previous analysis of PM 0.25 mass (Delfino et al. 2009), the associations with 1day and 3day average PM 0.25 were stronger and had narrower 95% CIs than did the 5day average for both IL6 and sTNFRII.
In exploratory analyses, we retested models for erythrocyte antioxidant enzymes (Cu,ZnSOD and GPx1) from our previ ous publication (Delfino et al. 2009). Random slopes and individual autoregres sive models showed small, highly influential subject clusters (seven subjects) with posi tive associations between air pollutants and antioxidant enzymes, whereas most of the remaining 53 subjects showed inverse associa tions. Details of these clusters and their inter pretation are presented elsewhere (Delfino et al. 2009). We present these datadriven results with the new air pollutant exposure data primarily in the Supplemental Material,  Table 2 provides descriptive statistics for the measured exposures. Seasonal differences were greatest for MMW PAHs, HMW PAHs, and nalkanes, which were higher in the cool sea son, and for WSOC, which was higher in the warm season, as expected because of photo chemistry. Indoor/outdoor ratios were close to 1.0 for PAHs and hopanes, and indooroutdoor correlations were strong (median R was 0.60 for PAH species and 0.74 for hopane species) ). This suggests high penetration of these outdoor PM 0.25 components into indoor environ ments and that measured indoor components were largely of outdoor origin. On the other hand, indoor/outdoor ratios were high for nalkanes and nalkanoic acids, with gener ally low indoor/outdoor correlation coeffi cients ). This suggests that indoor sources influenced the indoor levels of nalkanes and nalkanoic acids. Table 3 shows a correlation matrix for measured outdoor organic components. We found moderate to strong correlations between PM 0.25 mass, PAHs, and hopanes. We also found small negative correlations of these spe cies with organic acids and small positive cor relations with WSOC, suggesting that POA and SOA concentrations are relatively inde pendent of each other at the study sites.

Results
To further improve our understanding of the clearly positive associations of biomarkers with summed PAH compounds presented below, we used the chemical mass balance model (CMB) source apportionment esti mates from Arhami et al. (2010) to evalu ate the possible sources of PAHs. We briefly summarize methods and source apportion ment results in the Supplemental Material, Chemical mass balance (CMB) model (doi:10.1289/ehp.0901407). Table 4 shows a correlation matrix for the relation of PAHs to the CMBestimated sources. Strong correla tions are seen for total PAHs with vehicular emission sources, whereas the apportioned mass from other sources shows weak to null correlations.
In the mixedmodel regression analy ses, we found positive associations of cir culating biomarkers of inflammation (IL6 and sTNFRII) with organic components (Table 5, Figure 1). We found the strongest associations with biomarkers for both indoor and outdoor PAHs, including LMW, MMW, and HMW PAHs. The next strongest asso ciations were for hopanes. Indoor but not outdoor hopanes were associated with IL6, whereas both indoor and outdoor hopanes were associated with sTNFRII.
Outdoor WSOC (a marker of SOAs) was positively associated with sTNFRII, but confidence limits crossed 1.0 (p < 0.14), and we found no other associations with SOA markers. The outdoor organic acids (another marker of SOAs) showed a pattern oppo site to that of the POA markers, with largely negative regression coefficients in relation to biomarkers of inflammation. To assess whether this was due to inverse correlations with PAHs, we coregressed outdoor total PAHs with outdoor organic acids. We found that associations with PAHs and with organic acids decreased in magnitude to small degrees when coregressed, suggesting that the negative regression coefficients for organic acids with  Regression coefficients and 95% CIs are for the expected change in the biomarker among 60 subjects associated with an IQR change in the air pollutant (see Table 2), adjusted for temperature. *p < 0.05, **p < 0.01.  Table 2), adjusted for temperature.  We then tested twopollutant regression models that included both outdoor PM 0.25 mass and total PAHs to assess whether PAHs explained the nominal association with mass. We found that IL6 and sTNFRII associa tions with mass were completely confounded by PAHs in that the regression coefficient for mass decreased to just below zero and the regression coefficient for PAHs was nearly unchanged (Figure 2A,B). We found a sim ilar effect for hopanes, which confounded the nominal association of PM 0.25 mass with sTNFRII ( Figure 2C). The variance inflation factor was < 3.5 for exposures, thus showing little evidence of multicollinearity.
Transition metals were not associated with the biomarkers [see Supplemental Material, As previously shown (Delfino et al. 2009), the analysis of the relation of erythrocyte anti oxidant enzymes (Cu,ZnSOD and GPx1) to air pollutants among all 60 subjects showed regression coefficients were largely negative, suggesting inverse associations, but most upper confidence limits crossed 1.0 (see Supplemental Material, Table 3 (doi:10.1289/ehp.0901407)]. The exploratory analysis showed that among seven subjects previously identified as a "posi tive responder group" (Delfino et al. 2009), we found largely positive associations of Cu,Zn SOD and GPx1 with air pollutants, and lower confidence limits were > 1.0 for out door PM 0.25 mass and several other exposures. In the 53 subjects previously identified as a "negative responder group," we found inverse associations of Cu,ZnSOD and GPx1 with indoor and outdoor total, LMW, MMW, and HMW PAHs and with hopanes (all markers of exposures linked to primary combustion). Indoor WSOC was inversely (p < 0.07) associ ated with Cu,ZnSOD, but we found no other associations with SOA markers in the negative responder group. Confidence limits were wider for GPx1 than for Cu,ZnSOD.

Discussion
To our knowledge, this is the first report from a panel cohort study to show associations of circulating biomarkers of response in human subjects to specific PM organic compound classes. The measured chemicals serve as indi cators and tracers for air pollutant sources and for classes of chemicals with the potential for redox activity in the body. Our prior work has focused on carbonaceous aerosols that pro vided some differentiation between POAs and SOAs by showing associations of biomarkers of inflammation with primary PM 2.5 organic carbon (OC) but not secondary PM 2.5 OC (a marker of SOAs) (Delfino et al. 2008(Delfino et al. , 2009). In the present analysis, we found the stron gest biomarker associations with air pollutant variables for all molecular weight classes of PAHs and specific source markers of vehicular emissions (hopanes) measured in PM 0.25 with GC/MS. Furthermore, twopollutant mod els of the relation between the biomarkers of systemic inflammation and both total PAHs and PM 0.25 mass showed that mass associa tions were completely explained by PAHs. Given the results of the chemical mass balance analysis [see Supplemental Material, Chemical mass balance (CMB) model (doi:10.1289/ ehp.0901407) and Arhami et al. 2010], we infer that the confounding of nominal asso ciations between biomarkers and PM 0.25 mass by PAHs was through a common set of sources. PAHs likely serve here as a surrogate for redoxactive PM chemical components as evidenced in experimental models (Riedl and DiazSanchez 2005). For example, PAHs from diesel exhaust particles and oxidized derivatives of PAHs such as quinones lead to the generation of reactive oxygen species and subsequent oxidant injury and inflammatory responses, including the expression of nuclear transcription factorκB (NFκB) (Riedl and DiazSanchez 2005). NFκB increases the tran scription of cytokines and acutephase proteins that are predictive of coronary artery disease risk (Pai et al. 2004). PAHs can induce oxida tive stress responses after biotransformation to quinones by cytochrome P450 1A1 (Bonvallot et al. 2001), perhaps after delivery from the lungs to systemic targets.
In the Los Angeles Basin, most outdoor PAHs in PM 0.25 are expected to be from mobile sources (Schauer et al. 1996), and the CMB exposure correlations are consistent with this expectation. PAHs were also correlated with source markers of vehicular emissions (hopanes). Hopanes are the most unambiguous source marker of traffic emissions. However, the moderate but not strong correlation between hopanes and PAHs suggests that the measured PAHs include a different subset of mobile sources than that of hopanes. This may in part be due to the variability in PAHs rela tive to hopanes by combustionrelated prob lems in the vehicle fleet (Lough et al. 2007).
Overall, the associations of biomarkers with PAHs and hopanes suggest that our previous findings of positive associations of biomarkers with PM 2.5 , EC, and primary OC (Delfino et al. 2009) were due to PM of mobilesource origin. PAHs are found in greater concentra tions in the quasiUFP range compared with larger particles (Ntziachristos et al. 2007), and this has been hypothesized to explain enhanced prooxidative and proinflammatory effects of urban UFPs in the lungs and peripheral target organs of rodents (Araujo et al. 2008). The increased biological potency of UFPs may be related to the content of organic chemicals that have the capacity to reduce oxygen, such as quinones and nitroPAHs, for which PAHs may act, in part, as a surrogate (Ntziachristos et al. 2007) or as a source after biotransfor mation. From the present results we infer that, although PAHs may have an effect by  Table 2), adjusted for temperature.  themselves, they are also likely surrogates for other causal species we did not measure that are emitted from the same (traffic) sources.
We found little evidence that tracer variables for SOAs and related components (WSOC and organic acids) were associated with the circulating biomarkers in the expected direction. We have no explanation for the negative regression coefficients for organic acids with biomarkers. Although most of the SOAs are expected to be in larger PM > 0.25 µm, the present results are consistent with our finding of few biomarker associations with PM 2.5 secondary OC or accumulation mode particle mass (PM 0.25-2.5 ) in an earlier pub lication (Delfino et al. 2009). In that study, regression coefficients were also negative for IL6 in some models with PM 0.25-2.5 and with secondary OC. We speculate that components in outdoor SOAs estimated by our methods (e.g., organic acids), are mostly water solu ble and highly oxygenated, and dissolve after deposition on the airway epithelium and then quickly react with extracellular macromole cules and cell membrane constituents. Thus, these PM components may not directly inter act with the vasculature, although it has been hypothesized that inhaled particles lead to air way inflammatory responses and subsequent release of activated leukocytes and cytokines into the circulation (Mills et al. 2009).
An important limitation of our characteriza tion of SOAs is that WSOCs and organic acids do not completely characterize the SOA fraction of PM, part of which may come from the pho tochemical oxidation of lowvolatility vapors to form hydrophilic organic components, but whose chemical identity is largely unknown. These precursor vapors include SVOCs that are largely part of POAs. SVOCs evaporate from the particle phase during the process of atmo spheric dilution and subsequently react with oxidant gases to form a significant fraction of SOAs (Robinson et al. 2007).
Lipidsoluble components of PM more closely associated with primary emissions, including PAHs, may become bioavailable after deposition followed by distribution of unmetabolized chemicals to the circulation and to extrapulmonary target sites (Gerde et al. 2001). It is also possible that a small fraction of toxic components is carried via various translocation mechanisms into the circulation on UFPs (Mühlfeld et al. 2008). However, translocation may account for a potentially insignificant amount of the impact of UFPs compared with the high retention of UFPs in the lungs (Möller et al. 2008), which may lead to sustained effects through the gradual transfer of redoxactive compo nents to the circulation over many days.
Although transition metals are known to be redox active, we found no consistent associations with the biomarkers measured, possibly because of low concentrations of these trace elements in the study areas.
Finding positive associations of biomark ers with both indoor and outdoor PAHs and hopanes along with the indoor/outdoor ratios of these organic components being close to 1.0 suggests that, even though people spend most of their time indoors, indoor air qual ity and PM exposures are strongly influenced by PM of outdoor origin. These findings are consistent with our previous analysis for the first half of this panel showing that CMB estimated indoor PM of outdoor origin (par ticle number, EC, and primary OC) were associated with the biomarkers to a similar degree as outdoor PM (Delfino et al. 2008).
Briefly, the exploratory (datadriven) find ings for GPx1 and especially Cu,ZnSOD are consistent with our previous findings for primary OC and EC (Delfino et al. 2009) and suggest antioxidant enzyme inactivation within erythrocytes by trafficrelated pollut ant components, including PAHs, among a subgroup of people. This inactivation is anticipated to increase oxidative stress and thus inflammation. This is potentially impor tant because these enzymes likely represent important intermediate end points that have been linked to the risk of developing coronary artery disease in prospective cohort and other studies (FloresMateo et al. 2009). Given that these findings were far less clear when includ ing the entire 60subject panel (because a small subgroup of seven subjects had positive associations), these results should be viewed as hypothesis generating and retested in other populations. See Delfino et al. (2009) for further details and discussion concerning potential mechanisms of antioxidant enzyme inactivation versus upregulation that may explain group differences.
Strengths of the present study lie in expo sure measurements in each subject's com munity microenvironment and in repeated biological marker assessments in a wellcharac terized patient sample. Limitations include the potential for unmeasured temporal confound ing. However, we performed a priori adjust ment for one of the largest sources of variability in inflammatory mediators that have been doc umented in the literature (infections), and we also accounted for temperature and for com munity and seasonal variability in exposures. We also acknowledge that the present study does perform multiple comparisons, although we did narrow the number of hypotheses being tested based on prior evidence of associations from the work of others and ourselves.
The results of the present study suggest that tracer components of mobile source emis sions in PM 0.25 are associated with increased systemic inflammation in a potentially sus ceptible population of elderly individuals. The measured biomarkers likely represent important intermediate end points (systemic inflammation) that have been linked to the risk of cardiovascular diseases in prospective cohort and other studies (Kritchevsky et al. 2005;Pai et al. 2004). The positive relation between air pollution and cytokine bio markers may also be indicative of acute risk of adverse cardiovascular outcomes related to vascular dysfunction and atherothrombo sis (Mills et al. 2009). We recently reported coherent associations between hourly ambula tory systolic and diastolic blood pressure and hourly air pollutant exposures in the present panel cohort, including stronger associations with primary PM 2.5 OC compared with sec ondary PM 2.5 OC .
We conclude that U.S. EPA-regulated ambient PM 2.5 mass measurements may not adequately represent risk to human health because they are uncharacterized by composi tion, source, or PM size distribution and are not necessarily representative of personal or local exposure. Confirmatory data are needed in other populations using measurements of organic components across several PM size fractions.