Environ. Sci. Technol. 2008, 42, 7502–7509
Source Apportionment of in Vitro Reactive Oxygen Species Bioassay Activity from Atmospheric Particulate Matter YUANXUN ZHANG,† J A M E S J . S C H A U E R , * ,†,‡ M A R T I N M . S H A F E R , †,‡ MICHAEL P. HANNIGAN,§ AND S T E V E N J . D U T T O N §,| Environmental Chemistry and Technology Program University of Wisconsin - Madison, 660 North Park Street, Madison, Wisconsin 53706, Wisconsin State Laboratory of Hygiene, University of Wisconsin - Madison, 2601 Agriculture Drive, Madison, Wisconsin 53718, and Departments of Mechanical Engineering and Civil, Environmental and Architectural Engineering, University of Colorado at Boulder, 427 UCB, Boulder, Colorado 80309
Received January 14, 2008. Revised manuscript received June 6, 2008. Accepted August 5, 2008.
Recent atmospheric particulate matter health studies have suggestedthattheredoxactivityisanimportantfactorinparticulate matter toxicology, and that reactive oxygen species (ROS) activity may be an important characteristic of particulate matter that is associated with adverse health effects. In this study, associations between atmospheric particulate matter sources and in vitro ROS activities are investigated. Ambient concentrations of fine particle water-soluble elements and total organic and elemental carbon were measured daily in Denver for the 2003 calendar year. The data were used in a multivariate factor analysis source apportionment model, positive matrix factorization (PMF), to determine the contributions of nine sources or factors: a mobile source factor, a water soluble carbon factor, a sulfate factor, a soil dust source, an iron source, two point sources characterized by water soluble toxic metals, a pyrotechnique factor, and a platinum group metal factor. Aqueous leachates, including water soluble and colloidal components, as well as insoluble particles that pass through a 0.2 µm pore size filter, of 45 randomly selected PM samples, were assayed to quantify ROS activity using an in vitro rat alveolar macrophage assay. Results show that PM-stimulated in vitro ROS production was significantly positively correlated with the contributions from three sources: the iron source, the soil dust source and the water soluble carbon factor. The iron source accounted for the greatest fraction of the measured variability in redox
* Corresponding author phone: 608-262-4495; fax 608-262-0454; e-mail:
[email protected]. † Environmental Chemistry and Technology Program University of Wisconsin - Madison. ‡ Wisconsin State Laboratory of Hygiene, University of Wisconsin - Madison. § Departments of Mechanical Engineering, University of Colorado at Boulder. | Civil, Environmental and Architectural Engineering. 7502
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 19, 2008
activity, followed by the soil dust and the water-soluble carbon factor. Seventy-seven percent of the in vitro ROS activity was explained by a linear combination of these three source contributions.
Introduction Atmospheric particulate matter (PM) is an important air pollutant which has been observed to be associated with many adverse human health effects (1-3). Although epidemiology studies have documented positive correlations between atmospheric PM concentrations and mortality, aggravated asthma, chronic bronchitis, chronic cough, respiratory tract infections, ischemic heart disease, and strokes, etc. (4-9), the mechanisms through which PM acts to cause these PM-associated diseases are still not well understood. During the past several decades, numerous investigations into the physicochemical or biochemical properties of ambient aerosols, including concentrations, compositions, sources, atmospheric aging, particle size, redox activities, and toxicities, have been conducted to better define linkages between PM characteristics and associated health responses. Many of these studies suggested that reactive oxygen species (ROS) activity may be a key factor mediating aerosol biological activity in accordance with the hypothesis that many of the adverse health effects may result from oxidant stress. Therefore, PM-associated ROS generation is the focus of much attention in ongoing efforts to understand the mechanisms of the observed adverse health effects associated with PM (10-16). ROS is a collective term encompassing many chemical species such as the oxygen radical and hydroxyl radical, as well as other highly reactive forms of O2 such as hydrogen peroxide and singlet O2 (17). ROS is continually formed in the human body as the natural consequence of aerobic metabolism, and is integral for maintaining tissue oxygen homeostasis. Moreover, ROS is important in the destruction of microbial invaders. However, ROS can also cause significant tissue damage during infection and many diseases (17-19). Generation of ROS and the activity of antioxidant and radical scavenger defenses appear more or less balanced in vivo in healthy organisms (17). Oxidative stress results when ROS concentrations exceed the capacity of the antioxidant systems. Oxidative damage to cellular proteins and lipids may promote carcinogenic activity (18, 20, 21). Therefore, ROS has positive and negative impacts on the disease prevention and promotion paradigm (11). Oxidative stress in the lung can be imposed in many ways, including disease, tissue injury, pathogen invasion, and airborne chemical pollutant exposure (17, 22). Notably, the capacity of PM to generate oxidative stress is associated with two mechanisms: the inherent oxidant-generating properties and the ability to stimulate cellular generation of ROS (23, 24). Previous studies have shown that particle type, size, and surface area are all important factors when considering particle-antioxidant interactions in the airways (25). ROS activity has been found to be higher in the fine size-fraction of ambient aerosol, especially for the PM with an aerodynamic mean diameter less than 2.5 µm (PM2.5). These small particles may contain abundant persistent free radicals, e.g., semiquinone radicals, which can undergo redox cycling resulting in ROS generation and deleterious effects in the lung (14, 26, 27). Transition metals present in fine particles are also thought to be key species involved with the generation of ROS in the lung. In the presence of transition metals, such as iron and copper, hydrogen peroxide can be reduced to 10.1021/es800126y CCC: $40.75
2008 American Chemical Society
Published on Web 09/04/2008
the extremely reactive hydroxyl radical in the metal-catalyzed Haber-Weiss reaction (28). Molybdenum, aluminum, and vanadium were shown to have the capacity to increase ROS generation by polymorphonuclear leukocytes (PMN) in vitro; but chromium, cobalt, nickel, and titanium did not exhibit this activity (29). Other studies, however, indicate that soluble metals may not have a direct impact on the capacity of alveolar macrophages to clear PM via phagocytosis (30), but, many studies have demonstrated that soluble PM components, especially metals, are potent initiators of ROS, key drivers in redox-driven ROS generation mechanisms, and that a large fraction of the oxidant stress of PM may be attributable to soluble species (29, 31). Additionally, real environmental samples, as well as PM emitted directly from sources exhibit very different correlations between their redox activities and chemical composition (24, 27, 29, 31-34). Understanding the relationships between PM composition and PM-associated ROS activity is important for the purposes of health effects prediction and management. However, the complexities of the chemical composition of atmospheric PM make it very difficult with current tools to clearly attribute ROS activity to specific PM components. In this study, we establish associations between aerosol redox activity and PM sources, instead of individual components of PM. Water soluble components of particulate matter, which include colloidal components and insoluble particles that pass through a 0.2 µm pore size filter, were used for the study to focus on an important and reproducibly sampled component of PM that contributes to ROS generation. This goal is achieved for water-soluble components by integrating measurements of in vitro ROS activity with source apportionment results.
Experimental and Statistical Methods Aerosol Sampling and Chemical Analysis. Daily PM2.5 samples (nominal: 27 m3 of air sampled at 18.75 L min-1) were collected for one year in the metro area of Denver, CO in 2003 on parallel-configured precleaned 47 mm diameter Teflon (Teflo-Pall-Gelman) and quartz filters. Total OC and EC were quantified from the quartz fiber filters using the NIOSH thermal-optical method (35). Teflon filters were tared and reweighed on a robotic microbalance to a sensitivity of 1 µg. The median PM mass of the PM2.5 samples used for the ROS evaluation was 133 µg (average ) 151 ( 65 µg, minimum ) 24, maximum ) 348 µg). The Teflon filters were extracted with high purity MQ water in capped polypropylene vials on a shaker table for 6 h. Extracts were filtered through 0.22 µm polypropylene filters and then analyzed for water soluble organic carbon (WSOC) using a high-temperature combustion method (36), and 47 elements by high-resolution ICPMS. Details of the WSOC and ICPMS analysis methods, as well as the analytical results are included in the Supporting Information along with the complete chemical analysis database. Biochemical Analysis, Measurement of in Vitro ROS Production. Macrophages were exposed to aqueous leachates of filter-borne PM and subsequently assessed for viability (membrane integrity) and production of ROS as an indicator of macrophage stimulation. A subset of 45 PM samples were randomly selected for ROS evaluation. Filters (1/2 sections) were extracted with 0.90 mL of MQ water for 16 h on a shaker table, in the dark, and the extract was subsequently processed though a 0.22 µm polypropylene syringe filter. The macrophage exposure solutions were made by adding an aliquot of 10× concentrated solution of salts glucose medium (SGM) to each of the filter leach solutions. SGM was chosen as the medium to buffer the leach solutions because previous results demonstrated that the cells remained viable in this medium and it is a simple medium that allows for chemical speciation determination and accurate exposure assessment.
The viability of the macrophage cells was evaluated by assessing cell membrane integrity using a lactate dehydrogenase (LDH) assay (LDH-Cytotoxicity Assay Kit, BioVision), which probes leakage of LDH from the cell. For the LDH assay, macrophages were seeded into a 96-well plate and allowed to adhere overnight. The following day, the culture medium was removed and replaced with the sample extract exposure solutions diluted into SGM. Cells were allowed to incubate for seven hours after which the medium was removed and assessed for LDH activity as described by the manufacturer’s instructions. Untreated cells incubated in SGM were included as controls. In all sample exposures, the assay results were not significantly different than the controls, indicating viable cells, and no overt toxicity from the samples. For the ROS assay, both floating and suspended macrophage cells (alveolar macrophage cell line NR8383) were harvested, and the cell concentration was adjusted to 1 × 106 cells mL-1 by centrifugation. The culturing media was removed from the cell pellet and cells reconstituted in SGM. Macrophage cells (100 µL ) 100 000 cells) were added to each well of a 96-well plate and incubated for 2 h in a 37 °C incubator. After incubation, the SGM media was removed from the settled cells by pipetting and 100 µL of exposure solution (sample extracts to which dichlorofluorescein diacetate (DCFH-DA), at a concentration of 15 µM, was added immediately prior to cell exposure) was pipetted into the wells. This exposure level (soluble components from 20 to 200 pg of PM per cell), is generally consistent with, but at the low end of the range, of similar experiments reported in the literature. The plate was read (wavelength 485/530 nm), initially at time 0 and subsequently at 30 min intervals, for 2 h. Zymosan, a β-1,3 polysaccharide of D-glucose, which is recognized (binds) by TLR-2 receptors on macrophage cells, activating a strong immuno-chemical response (37), was used as a positive control for ROS induction (38, 39). We also normalized the raw fluorescence data to the zymosan controls, and expressed the ROS response in terms of zymosan units, to account for minor variations in method sensitivity and to aid comparisons with other studies. All samples, and positive (zymosan) and negative (method blanks) controls were run in triplicate and also at three dilutions (100, 66, and 33%) to check and confirm the dose-response. The average response was blank-control corrected and then standardized-normalized to the response of the zymosan positive control. The detailed results are presented in Table S2 in the Supporting Information in units of µg zymosan equivalents per m3. Source Attribution. Factors contributing to atmospheric PM concentrations were determined using the U.S. Environmental Protection Agency (EPA) source apportionment positive matrix factorization (PMF) model, which has been widely used in the past with elemental data, and more recently with organic molecular markers (40, 41). In this study, the U.S. EPA PMF1.1 was used, which allows the user to review input species concentration statistics including signal-tonoise ratio (S/N), and downgrade the importance of species to “weak” or “bad” status. S/N is given by: n
n
i)1
i)1
S/N ) 0.5 × ( Σ Xij2 / Σ Sij2 )1/2 where X is the species concentrations and S is the uncertainties. This definition is different than that described by Paatero and Hopke (44). This version of the PMF model increases the uncertainties of “weak” species by a factor of 3 and removes the “bad” species from computation. Using this protocol, four species (erbium, thulium, ytterbium, and lutetium) with S/N equal to or less than 0.5 were marked as “bad”; and two species (titanium and platinum) with S/N equal to or less than 1.0 were marked as “weak”. To explore the probable rotational ambiguity, PMF2 was also applied VOL. 42, NO. 19, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7503
FIGURE 1. Time series of PMF derived source contributions, Denver, Colorado, 2003. Selected days having ROS measurements are represented by circles. with different FPEAK values to verify that the result derived from PMF1.1 is in accord with the nonrotated result from PMF2. Further detailed discussion of PMF modeling can be found in the literature (41-44).
Results and Discussion Source Apportionment of Water-Soluble Elements and OC/ EC. A series of PMF results were obtained with factor numbers ranging from 8 to 13. The nine-factor result was chosen as the optimum model based upon the computed statistical parameters and a comparison between different solutions using prior-knowledge about potential sources. Additional factorssbeyond the base 9scould not be associated with specific sources or atmospheric factors given existing knowledge of sources and atmospheric processing of carbonaceous 7504
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 19, 2008
PM and water-soluble elements. Contributions to measured water-soluble species plus OC and EC for the nine PMFderived sources are presented in Figure 1. Detailed model results, including source profiles and individual speciated daily apportionment results are shown in Figures S1 and S2 in the Supporting Information. The total apportioned species averaged 96.8% of the measured species, which indicates that residuals are acceptably small for the selected ninefactor case. The characteristics of source contributions and profiles, and the implied source types are briefly discussed below. Mobile Source. The PMF factor with significant high content of carbonaceous components, both water insoluble organic carbon (WIOC) and EC, is thought to be associated with mobile sources (45). Contributions in winter are higher
than in other seasons reflecting higher mobile emission rates at lower temperatures. The mobile source factor had an annual average concentration of 2000 ng m-3, which made up more than half of the total measured species. Water-Soluble Carbon Factor. One of the PMF factors dominated the WSOC, potassium, and boron concentrations. WSOC largely consists of polar organic compounds such as levoglucosan, fatty acids, aromatic acids, polyols, and hydroxy acids, which are derived from both secondary organic aerosol and biomass smoke. The seasonality and the high loadings of water soluble potassium indicated that this factor contains both secondary organic aerosol and biomass smoke. The yearly concentration of this factor was 790 ng m-3, representing 20% of the total mass of water-soluble elements and carbonaceous pollutants. Sulfate Factor. The PMF factor, which contains most of the measured sulfur was identified as a sulfate factor. This factor includes both local and regional sources of sulfate. Its yearly average contribution was 300 ng m-3, accounting for 7.6% of the total mass of the measured species. Soil Dust Source. The soil dust source is identified by the water-soluble resuspended dust elements, including magnesium and calcium. Yearly average contribution of soil dust is 350 ng m-3, 8.9% of the total mass of water-soluble elements plus carbonaceous pollutants. A seasonal comparison shows higher concentrations in the summer and higher concentrations on weekdays as compared to weekends. Iron Source. The PMF factor with high loadings of watersoluble iron and titanium is likely associated with an industrial source, which has not been identified. The iron source represents an important component of the total mass of water-soluble elements plus carbonaceous pollutants, with a yearly average contribution of 290 ng m-3 (7.4%). Relatively high contributions were observed in summer, which may result from redox cycling of iron during photochemical smog (46). Point Sources. Two point sources were derived which dominated the water-soluble lead and cadmium/zinc timeseries, and had yearly average concentrations of 157 and 19 ng m-3, respectively. The lead point source had several large impacts on March 21st, December 21st, and December 28th, whereas the large impacts of the point source of cadmium/ zinc occurred on June 17th, December 6th, and December 11th. The point source of cadmium/zinc also had considerable amounts of associated sodium, potassium and calcium. Pyrotechnics Factor. The pyrotechniques factor was identified by the high levels of water-soluble metals associated with flame reactions such as barium, copper, and strontium, etc. This factor includes fireworks and has a peak contribution on July 4th, as shown in Figure 1, but also includes smaller peaks at other times of the year. The July 4th peak concentration was the source of 520 ng m-3, whereas the yearly average contribution was only 32 ng m-3. Platinum Group Metal Source. The PMF factor, which dominates water-soluble platinum group metals including platinum, palladium, and rhodium, is not well understood but is very distinctive in the PMF model. The metals are associated with catalysis reactions, including catalytic converters on gasoline-powered vehicles, but the source or sources of this factor have not been identified. The yearly average contribution by platinum group metal source is 23 ng m-3, 0.6% of the total mass of water-soluble elements plus carbonaceous pollutants. Correlations Between ROS and Key PM Constituents. The randomly selected days with ROS measurements are indicated by circles in Figure 1. To explore the association of ROS generation with chemical species, we examined the covariances between the measured species concentration and the measured ROS generation. The 12 species with the largest covariances with ROS generation are presented in
Figure 2. The observed univariate correlations between ROS activity and WIOC, EC, magnesium, and barium are poor (Figure 2). In contrast, linear relationships between ROS activity and concentrations of many of the other species are significant, e.g., those with potassium and iron. However, only WSOC, iron, and boron show strong, unambiguous, positive linear correlations. Weak correlations between any individual species and ROS production likely reflects the complexities of ROS activity generation and assessment as discussed previously. To simplify the correlation model, multiple linear regression (MLR) was applied to build statistical relationships between ROS production and the concentrations of the aforementioned 12 potentially significant species. Detailed results of the full statistical analysis are provided in Table S3 in the Supporting Information, where relationships between individual species and ROS activity are detailed. The statistical analysis demonstrated that the ROS generation could not be explained well using these 12 measured species. More details are presented in the Supporting Information of this manuscript. ROS Prediction by Source Contributions. We developed a statistical model between the PMF-identified source contributions and ROS activity using procedures similar to those performed between the species concentration and ROS production metric. Several different statistical models were tested, from which MLR analysis was selected to predict the ROS production from a linear combination of source contributions since it is the simplest of all the available acceptable statistical models which can explain the ROS sources. The statistical metrics for the three MLR models run are listed in Table 1. Regression (a) applied all nine PMFderived sources as independent variables and the complete ROS data set as the dependent variable. As seen in Table 1, all sources except the iron source, had p-values larger than 0.05. Only three sources have positive coefficients: iron source, soil dust, and water-soluble carbon. Multicollinearity of the independent variables is not a problem given that variance inflation factor (VIF) values are all
