Is oxidant activity of PM relevant?
This study was based on the assumption that PM-induced oxidative stress is central to the observed toxicity in vivo
and underlies many of the health effects observed in exposed populations (Donaldson et al. 2003
). Support for this contention, associating particle oxidative components with health effects, has now been published by numerous groups (Donaldson et al. 1996
; Gilliland et al. 1999
; Nel 2005
; Prahalad et al. 2000
; Shi et al. 2003b
; Zielinski et al. 1999
). We used two different measurements of PM-induced oxidative stress: •
OH formation in an oxidant environment (in the presence of H2
) and antioxidant depletion in a synthetic RTLF, reflecting the normal reducing environment at the air–lung interface. In this study, both of these assays were highly correlated, demonstrating that particle suspensions recovered from filters contain components able to cause oxidative stress under very different conditions. Both methods are therefore appealing in investigating PM-related health effects mediated through redox mechanisms. Hence, the capacity of PM to stimulate the production of •
OH in the presence of H2
can be seen as indicative of reactions likely to occur in a diseased/inflamed lung, whereas the RTLF, containing AA, urate, and GSH, is more reflective of a healthy lung scenario (Shi et al. 2003a
; Zielinski et al. 1999
). The importance of these radical-generating processes has been shown in studies demonstrating DNA oxidation in response to PM-induced production of •
OH (Shi et al. 2003b
; van Maanen et al. 1999
Both the PM-stimulated formation of •
OH and the loss of RTLF antioxidants can be attributed to the transition metal content of PM. Several transition metals [Fe, Cu, chromium (Cr), Ti, nickel (Ni), cobalt (Co), and V] will react with H2
to form •
OH via the Haber-Weiss reaction (Halliwell 1999
), whereas the loss of AA and GSH is caused both by their reduction of oxidized metal ions [Fe(III), Cu(II), Cr(VI)] and by the subsequent formation of superoxide (Buettner et al. 1996
). Notably, under aerobic conditions, at neutral pH, Ni, Ti, and Co will not act as catalysts for AA oxidation (Fridman et al. 1979
) such that the two assays of oxidative activity are sensitive to slightly different panels of metals. In addition, AA will undergo catalytic oxidation in the presence of quinone compounds (Roginsky et al. 1999
), such that the antioxidant depletion assay is sensitive to both metal and quinone/hydroquinone components of PM. In contrast, quinones do not yield •
OH in the electron paramagnetic resonance–based assay. The strong correlation between the capacity of PM to generate •
OH and deplete AA would therefore suggest that PM quinone content does not contribute significantly to the oxidative activity in these archive filter PM samples, and that redox active metal components such as Fe and Cu are more important determinants of the observed activity. We therefore conclude that the use of redox activity of PM may greatly advance our understanding of health effects of air pollution and that the use of complementary methods may help identify those components driving the observed activity.
Is oxidative activity correctly measured?
Although we have argued that the capacity of PM samples to cause oxidation reactions is an important determinant of their biologic activity, this assumption holds true only if the PM samples extracted from filters are representative of those breathed in ambient air. Particles collected on Teflon filters or other substrates and stored before sonication may not be a perfect model for PM as encountered by the lung under real-life conditions because of the impact of sonication and aging processes on the PM components. Volatile constituents of PM such as organic fractions that represent a potential source of redox activity in fresh PM may not be fully captured onto filters, and extraction procedures into water may not fully extract those present (Li et al. 2002
; Xia et al. 2004
). We have previously shown that ambient PM extracted into water or methanol has equivalent activity (data not shown), suggesting this may not be a significant problem, although clearly this will depend on the composition of the sampled PM. Mass recovery for PM2.5
in this study was high but not complete (88%), and this may have confounded our data, because ultrafine particles are less well recovered from these filters than larger particles.
Another question not addressed in our PM2.5
-based study is the specific redox relevance of different-size fractions of PM. We and others have shown that, on a per unit mass basis, ultrafine particles are considerably more oxidatively active than are fine and coarse particle fractions (Cho et al. 2005
; Li et al. 2003
; Shi et al. 2003b
). However, the larger fractions of PM2.5
remain important determinants of exposure to redox-active particles (per volume) given that the mass of ultrafine particles in ambient air is rather small (Cho et al. 2005
). The mass is particularly large for the coarse fraction (2.5–10 μm), and health effects and redox activity in this fraction have also been shown to be significant (Brunekreef and Forsberg 2005
; Shi et al. 2003b
Finally, the way that traditional measures of air pollution are reported (mass/volume of air as a measure of air quality) contrasts with the units of oxidative properties [actual measures of particle quality (property/mass of particles)], which needs to be considered when interpreting our results.
We conclude that the approaches chosen in this study are valid to capture a toxicologically relevant feature of ambient PM. Findings should, however, not be generalized beyond the size fraction sampled in ECRHS or the fraction of PM that can be successfully recovered from filters. Studies that compare the currently employed methods to measure PM redox activity are warranted.
Should these assays be used in epidemiologic studies?
Our results show that the oxidative properties of PM vary considerably across Europe and thus may explain part of the large heterogeneity in respiratory health symptoms reported in the ECRHS population (Janson et al. 2001
). However, it appears inevitable from our data that these toxicologically relevant properties need to be measured given that no other PM characteristic served as a reliable surrogate. This was apparent both for short-term temporal patterns within cities and in the cross-community comparison of aggregate annual means. The filter-to-filter correlation between the •
OH formation and the ambient mass concentration of total PM or elemental constituents was not only low (Table 2
) in many cases, but also heterogeneous across cities. This was also true for transition metals (Fe, Cu, V, Mn, and Ti), which are important sources of free radical formation (Aust et al. 2002
; Prahalad et al. 1999
). Factor analyses and other multivariate approaches may elucidate the specific multivariate determinants of PM redox activity. However, a city-by-city approach would be required given the large differences in redox-relevant PM characteristics across Europe (Götschi et al. 2005
). Such analyses are beyond the scope of this article.
We also emphasize that the redox activity of PM-associated metals depends not only on the bulk concentration in the sample but also on bioavailability (for Fenton reactions), chemical speciation, and oxidation state (Shi et al. 2003a
), which are not properly reflected in the elemental mass concentration derived by ED-XRF. This may partly explain the weak associations between metals and redox activity in our study. Correlations between absorbance—a proxy for elemental carbon—and redox activity were also weak, in contrast to recent findings in California showing a correlation coefficient of 0.89 between redox activity and elemental carbon (Cho et al. 2005
). However, the latter study used an assay that is not sensitive to the metal content of PM, and pollution mixtures in southern California are likely to be significantly different from those in Europe; thus, comparability of these results is limited. Moreover, our measurements of light absorbance may in part be determined by the fraction of PM2.5
not recovered by our extraction (see above).
It is noteworthy that sulfur was the best correlate of redox activity in a few cities (Table 2
) as well as across communities (annual means; Table 3
). This underlies the importance of using surrogate measures when examining PM activity, because S content can not be simplistically related to established toxic pathways. The elemental S content of PM reflects both the concentration of secondary sulfate and primary metal sulfates derived from combustion processes in the samples. Secondary sulfate, derived from the oxidation of sulfur dioxide, represents the predominant form of S in PM2.5
but is unlikely to contribute significantly to the toxicity of the samples. Both human and animal challenge studies with sulfate salts have proven unable to duplicate any of the acute/chronic health effects related to PM exposure (Schlesinger et al. 1990
; Utell et al. 1983
). It has been argued, however, that sulfate may represent a proxy for the bioavailability of PM transition metals, because of both the strong association of these metals with acid aerosols and the capacity of sulfate to mobilize insoluble Fe from the surface of particles (Ghio et al. 1999
). In addition to acting as possible ligands for Fe, sulfate can also modify transition-metal–catalyzed oxidative reactions in vivo
by scavenging •
OH to yield less reactive inorganic radicals such as SO4•−
(De Laat et al. 2004
The correlation between the oxidative properties and the other PM markers varied temporally at each of the sampling sites and regionally between cities, as has been shown previously (Shi et al. 2003a
). This may explain differences observed in the toxicity profile across season (Salonen et al. 2004
), location (Schaumann et al. 2004
), or sources of PM (Greenwell et al. 2003
) that have been reported in studies using surrogates rather than a direct measure of redox activity. The results of the two Antwerp locations (Table 2
) also indicate that the heterogeneity of the correlations persists even within the same city, at two locations 11 km apart.
As shown in Table 3
, none of the PM characteristics served as a proxy for the annual mean redox activity of the local PM. Particle oxidative activity was, on average, relatively low in some cities whereas levels of urban air pollution were high. Conversely, other locations appeared to have low mass concentrations, with high redox activity. This emphasizes that the oxidative properties of PM (per mass) are not related to ambient PM mass concentrations (per volume of air).
Several limitations and open questions need to be addressed before call for a large-scale use of these oxidative properties in epidemiologic studies. It is important to establish how PM oxidative activity measured at a central monitor relates personal PM exposures. Similarly, the relationship between activity measured at a single monitor as a marker of personal exposure and its relationship to spatial within-community and indoor/outdoor variation needs to be known. A recent French project and an extensive Dutch investigation showed that differences between personal and fixed site monitor concentrations vary considerably across PM constituents (or other pollutants), as well as across cities (Janssen et al. 2005
; Meng et al. 2005
; Nerriere et al. 2005
; Oglesby et al. 2000
; Sarnat et al. 2005
). For example, although light absorbance and S concentrations showed rather high correlations with personal levels (Spearman rank correlations often > 0.9), outdoor measures were rather poor surrogates of personal exposure to calcium, Cu, or Si (usually ≪ 0.5) (Janssen et al. 2005
). For pollution characteristics with very strong spatial gradients, such as reported for ultrafine particles or for numerous elements, such as Zn, Fe, Ni, or Cu (Zechmeister et al. 2005
), in proximity to traffic arteries, a single monitor does not, therefore, appear to be an informative marker of personal exposure (Zhu et al. 2002
). Given that these constituents are important determinants of redox activity, one has to expect large spatial gradients for oxidant activity, as well. This contention is supported by the data obtained from the two Antwerp monitoring sites (Table 1
). Although average losses of AA from the synthetic RTLF varied 2.5 times across communities, AA depletion was, on average, two times larger in Antwerp City than in Antwerp South.
Therefore, study designs that rely on the assumption that the assigned measures reflect personal exposure of the study participants (e.g., panel studies, or cross-community comparisons of long-term effects) may need personal or at least home-based measurements to thoroughly investigate effects associated with PM oxidant activity. The lack of knowledge of spatial variation of oxidative properties limits the use of this novel information also in the ECRHS cross-community comparisons that are based on one single monitor per center.
In case of time-series studies, this may be less of a concern (Sheppard 2005
; Sheppard et al. 2005
; Zeger et al. 2000
). Our data indicate that the temporal variability of •
OH formation and other traditionally used markers are of comparable size (see CV in Table 1
). Thus, single-monitor daily measurements of PM •
OH generation may be of some value in time-series studies.
The application of these novel measures in the statistical model of epidemiologic investigations needs further clarification. Both •
OH-generating capacity and antioxidant depletion are expressed per standard mass of PM2.5
, whereas mass and elemental content are characterized per volume of air as encountered by the lung under real-life conditions. The association between these radical-generating properties at a standard mass and ambient PM concentrations has been tested but varies between samples (Shi et al. 2003b
). It would be useful to establish this function in a controlled toxicologic model. It is therefore interesting to note that a recent clinical study in which similar masses of PM2.5
were instilled into bronchi of healthy subjects showed different inflammatory responses that appeared to be related to the difference in the oxidative activity between the instilled PM samples (Schaumann et al. 2004
). Redox activity of the standard mass may be used as an independent “exposure” term with or without PM ambient mass concentration as a coexposure. Alternatively, one may use an interaction term of PM and redox activity, assuming that the effect of air pollution on respiratory health depends both on the property of the inhaled particles (redox activity) and on the actual dose of such particles inhaled (PM concentration).