Environ. Sci. Technol. 2005, 39, 999-1010
Seasonal and Spatial Relationship of Chemistry and Toxicity in Atmospheric Particulate Matter Using Aquatic Bioassays REBECCA J. SHEESLEY,† J A M E S J . S C H A U E R , * ,†,‡ JOCELYN D. HEMMING,‡ STEVE GEIS,‡ AND MIEL A. BARMAN‡ Environmental Science and Technology Program, University of WisconsinsMadison, Madison, Wisconsin 53706, and Wisconsin State Laboratory of Hygiene, Agricultural Drive, Madison, Wisconsin 53718
In light of current interest in better understanding the environmental impact of atmospheric particulate matter (PM), a new strategy has been employed to screen the relative toxicities of ambient and source aerosols. Shortterm and acute aquatic bioassays using Ceriodaphnia dubia and a green alga (Selenastrum capricornutum) as test organisms have been in use for many years in the regulation of wastewater effluents. These tests have been employed in the present study to compare the toxicity of water extracts of atmospheric particulate matter and dichloromethane (DCM) extracts that have been transferred to dimethyl sulfoxide and diluted in water. Atmospheric PM was collected at four sites located near the south shore of Lake Michigan and one site in Michigan’s Upper Peninsula at discrete events during three seasons. Parallel chemical analyses of the two extracts directly assessed the relation between the chemical composition and the toxicity of the extract. Inductively coupled plasma analysis of the metals in the water extract and gas chromatographymass spectroscopy of the organics in the DCM extract showed a relationship between high toxicity and high watersoluble copper concentration and high secondary organic aerosol tracers in the extracted aerosol. Although previous fractionation studies have not looked at watersoluble copper, significant toxicity has been measured in the semipolar and polar organic fractions of ambient aerosols and diesel exhaust particles, which are the fractions in which secondary organic aerosol components would be expected. For the water extracts, the summer samples were consistently more toxic than the autumn or spring samples. There was not a seasonal pattern for the toxicity of the DCM extracts; however, spatial differences were apparent. The toxicity end points of select samples from one site qualitatively correlate with the high polycyclic aromatic hydrocarbon concentrations. Additionally, high toxicity in the July DCM extracts from another site may be tied to the presence of the insecticide carbaril. The seasonal and spatial variations captured in the toxicity results in this study tend to qualitatively correlate with trace * Corresponding author phone: (608)262-4495; fax: 262-0454; e-mail:
[email protected]. † University of WisconsinsMadison. ‡ Wisconsin State Laboratory of Hygiene. 10.1021/es049873+ CCC: $30.25 Published on Web 01/13/2005
2005 American Chemical Society
(608)-
organic components and metals and not bulk particulate matter composition.
Introduction With increasing interest in the health and environmental effects of atmospheric particulate matter (PM), the ability to correlate the sources and chemical composition of aerosols with their biological impact is of paramount importance. Complete assessment of the biological impact of aerosols would require a broad look at the effects on human health and the terrestrial and aquatic environments. Many established methods exist to determine the detailed chemical composition of atmospheric aerosol (gas chromatographymass spectroscopy (GC-MS), inductively coupled plasma (ICP) mass spectroscopy, X-ray fluorescence, etc). These have been used to demonstrate the extreme diversity of the composition of atmospheric aerosols as a function of temporal and spatial factors. Similarly, biological characterization of atmospheric and emission source PM has been conducted on a variety of species, with rodent inhalation studies (1, 2) and bacterial assays (3-5) constituting the bulk of the in vivo research. In addition, much in vitro work with human and rat cell lines has further expanded the collective knowledge of the biological response mechanisms to PM (6, 7). However, due to the complex nature of biological testing, few of these studies have combined detailed chemical analysis with biological characterization, focusing instead on bulk or physical characterization (elemental and organic carbon, inorganic ions, particle sizing, etc.) or a limited set of detailed analytes such as polycyclic aromatic hydrocarbons (PAHs). Population-based epidemiology studies also tend to correlate human health end points to PM on the basis of bulk and physical characterization of PM. This means that limited connections can be made between the detailed chemical studies of ambient PM and epidemiological and biological characterization studies. To better connect epidemiological studies, which look for indicators of the impact of atmospheric aerosol in specific human health end points (e.g., mortality, emergency room visits) (8-10), and laboratory tests, which look at biological impacts of select ambient and source aerosols on the molecular level (7, 11-14), links need to be established between these different types of research. Methods integrating bioassays with chemical analysis need to be established to study the impact of collected ambient PM and the associated sources and components on organisms in a routine manner. This will build a database of information that can be used to connect epidemiology and toxicology studies through the coordinated measurement of organic compounds and metals relevant to atmospheric emission source attribution and detailed atmospheric PM characterization. Since ambient aerosol is a complex mixture of transformed components of a variety of sources, the integrated chemical and biological analysis will allow the biological assay results to be connected to the source and tracers found in that specific PM sample. In this way, the same components of ambient and source aerosol found to correlate with toxicity in laboratory tests can be tested for correlations with human health end points in epidemiology studies. Thus, the objective of this study is twofold. The first goal is to outline a means of using detailed chemical characterization of water and dichloromethane (DCM) extracts to identify compounds or PM sources which contribute to toxicity as measured by aquatic toxicity tests. By analyzing VOL. 39, NO. 4, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
999
samples at multiple locations in one region collected during three seasons, differences in source contributions and atmospheric conditions can be exploited to identify determining factors in toxicity; this is a plan that can be exploited and expanded upon in future studies using any number of biological assays or human cell lines. Although aquatic toxicity testing cannot be directly related to human health end points, the assays can function as an initial screening tool and provide direction for epidemiological studies and biological characterization assays more closely linked to human response. The secondary aim of this research is to investigate the potential toxic burden of PM which can deposit onto aquatic ecosystems. This is intended to broaden the scope of toxicity studies of atmospheric aerosol to begin to include effects on aquatic systems. Ideally, an integration of chemical and biological assays would allow analysis of chemical constituents in the same matrix tested for toxicity. Following this model, researchers have combined mutagenicity assays and polycyclic aromatic compound (PAC) analysis on the same solvent extract from atmospheric samples (15, 16). While these studies were effective in isolating the mutagenic components of the nonpolar fraction of atmospheric particulate matter, they have not produced similar results with the more polar or water-soluble fraction. By introducing different bioassays and different solvents, including water, into the realm of atmospheric aerosol toxicity, a better picture of the potential impact of atmospheric particulate matter on the environment can emerge. Additional solvents call for the inclusion of a broader spectrum of chemical analysis. More informative conclusions can be made about the source (both primary and secondary) of the toxic fraction in aerosols by expanding the organic characterization beyond PACs and by including the measurement of trace metals. Whole effluent toxicity (WET) testing is used to monitor the toxicity of wastewater effluents released into aquatic environments (17, 18). Standardized procedures involving laboratory aquatic organisms, which are sensitive to a variety of pollutants, offer an initial assessment of potential toxicity of an effluent/substance using short-term and acute tests (19). These methods, however, have not been employed to study atmospheric PM inputs to aquatic environments. Work has been done in assessing the impact of selected components of atmospheric PM including persistent organic contaminants such as polychlorinated biphenyls and pesticides (20-26). Nutrient inputs such as nitrogen and iron from the atmosphere to aquatic environments have also been extensively studied (27-31). As previously noted, little work has been done to test the relative toxicity and potential impacts of particulate matter as a whole on aquatic environments, and thus, the impact of atmospheric PM on aquatic ecosystems is far from understood (32). By using an established aquatic toxicity technique, an initial picture can be established of the relationship of PM sources and chemical composition to aquatic toxicity. Not only aqueous samples or extracts, but also a DCM extract transferred to dimethyl sulfoxide (DMSO), can be analyzed for toxicity using WET tests. This allows integrated chemical and biological analysis of both a DCM extract and a water extract of atmospheric PM samples. To this end, short-term and acute aquatic toxicity tests were used in this study to qualitatively assess the relative toxicities of atmospheric total suspended particulate matter (TSP) samples from four sites around the southern shore of Lake Michigan and one remote site in Michigan’s Upper Peninsula. TSP samples were collected at the five sites, split, and extracted by DCM or reconstituted hard freshwater (17). The DCM extract was used for both toxicity and detailed organic analysis by GC-MS, while the water extract was used for metals analysis by ICP as well as toxicity testing. The two 1000
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 4, 2005
FIGURE 1. Sampling sites located around Lake Michigan: 1, Milwaukee, WI; 2, Waukesha, WI; 3, Porter, IN; 4, Bridgman, MI; 5, Seney National Wildlife Refuge, MI. organisms used in this study were the water flea Ceriodaphnia dubia and a green alga, Selenastrum capricornutum, following U.S. EPA established protocols for these bioassays (17, 18).
Experimental Methods Sampling. Four sites were chosen along the southern shore of Lake Michigan for conducting the atmospheric particulate matter sampling (Figure 1). Each site represented a different land use type, including urban, industrially impacted, and rural. The four sites were located in Milwaukee and Waukesha, WI, Porter, IN, and Bridgman, MI. The Milwaukee site was primarily urban, and was situated on the roof of the Great Lakes Water Institute at an established Wisconsin Department of Natural Resources (WDNR) air-monitoring platform. The Waukesha site was primarily industrial and residential, and was located near a centrifugal casting plant at another established WDNR air-monitoring platform. The Indiana site was located at the Indiana Dunes National Lakeshore Headquarters along the south shore of Lake Michigan in the weather corral. The heavy industry corridor in and around Gary, IN, impacts this area. The last site was located in Bridgman, MI, near Warren Dunes State Park, which is in the southwestern corner of Michigan on the shore of Lake Michigan. The primary land use type was small town/rural. All the sites were located within one mile of Interstate 94. In addition to their representing different land use types, these sites were chosen due to their accessibility, availability of electrical power, proximity to the lake, and sufficiency of open space for air sampling. One final site, located at Seney National Wildlife Refuge in Michigan’s Upper Peninsula, served as a remote site with limited local anthropogenic sources. This site was in the Integrated Monitoring of Protected Visual Environments (IMPROVE) network, which was used to assess regional haze (33). Atmospheric particulate matter was collected using General Metal Works Hi-Volume total suspended particulate matter samplers (GMWL 2000, Thermo Andersen, Smyrna, GA) on 8 × 10 in. (20 × 25 cm) quartz fiber filters and Zefluor filters (quartz filters and Teflon filters from Pall Life Sciences, Ann Arbor, MI). All atmospheric samples described in this project were TSP. TSP was used in this study to enable greater time resolution because larger quantities of aerosol mass could be collected with the high-volume samplers. Additionally, TSP was considered to be more representative of what would actually be deposited into an aquatic system. Table 1 lists sampling details for the five sites including event schedules. Sampling events were designed to capture seasonal and spatial differences. Details of the sampling protocol
TABLE 1. Details of Sampling Events and Filter Compositing for Atmospheric TSP Samples Collected at Four Sites around the Southern Shore of Lake Michigan and One Site at Seney National Wildlife Refuge in Michigan’s Upper Peninsula sampling dates (as composited)
code
Milwaukee, WI
Waukesha, WI
Aug 17-20,a 2000, 72 h Nov 16, 2000
A B
Teflon
Nov 17, 2000
C
Teflon
Nov 18, 2000
D
quartz, Teflon
Mar 9, 2001
E
Mar 10-12, 2001, 48 h Jul 26-28, 2001, 48 h Jul 28, 2001 Oct 25-27, 2001 48 h Oct 25-28, 2001, 72 h Jun 6-8, 2002, 48 h
F G H I J K
quartz,b Teflonc
Porter, IN
Bridgman, MI
quartz Teflon quartz, Teflon quartz, Teflon quartz, Teflon
quartz Teflon Teflon
quartz, Teflon quartz quartz quartz
Seney National Wildlife Refuge, MI
quartz, Teflon Teflon
quartz quartz
quartz, Teflon quartz, Teflon quartz quartz
quartz
quartz
quartz quartz
quartz
a
Each sample was started at noon on the date shown and stopped 24 h later. For multiple-day samples or composites, the dates reflect the start date and the stop date. b Quartz filters were used for mass, organic and elemental carbon, GC-MS, bioassays, ICP, and ion chromatography (when Teflon was not available). c Teflon filters were used for mass and ion chromatography.
have been published previously (34). Briefly, before sampling, the quartz fiber TSP filters were baked individually in aluminum foil packets at 550 °C for 12 h. The samples collected on the quartz TSP filters were sampled at a flow rate of 1.42 m3/min, while the flow rate for the Zefluor TSP was 0.85 m3/min. After sampling, the filters were stored in their aluminum foil packets and/or plastic bags in the freezer until analysis. Chemical analyses and bioassays were all conducted on portions of these filters, with sample splitting occurring at the filter cutting stage or after extraction. Chemical Analyses. Bulk and trace chemical analyses were conducted for all the sampling events shown in Table 1. Bulk analyses included mass, organic and elemental carbon by thermal evolution and combustion (35), sulfate, nitrate, chloride, and ammonium by ion chromatography, and watersoluble metals by ICP. Detailed organic analysis by GC-MS was conducted on all the samples using a DCM extract. Two solvents were used in chemical analyses and bioassays, DCM and reconstituted hard freshwater, synthetic lake water prepared in the laboratory using salts following EPA guidelines (18). The EPA detailed a number of reconstituted waters that could be made in the laboratories for these tests; the freshwater options ranged from very soft to very hard, with the hard freshwater having the second highest final concentration of salts in the water (18). The reconstituted hard freshwater extraction was done using one-quarter of each quartz TSP filter and 30 mL of reconstituted hard freshwater, combined in a 50 mL sterile centrifuge tube and agitated on an orbital shaker for 2 h. An aliquot of 1.5 mL was then removed for analysis by ICP (Thermo Jarrell Ash, Franklin, MA), reserving the remainder of the reconstituted hard freshwater extract for bioassays. The following elements were quantified by ICP: aluminum, antimony, arsenic, barium, beryllium, boron, cadmium, calcium, chromium, cobalt, copper, iron, lead, magnesium, manganese, molybdenum, nickel, potassium, selenium, sodium, thallium, vanadium, zinc, and silver (although only the elements which were consistentlyl detected were included in the results graphed in Figure 5). A second portion of the filters was extracted with 300 mL of DCM using a Soxhlet extractor. After concentration of the DCM extract to 2 mL by rotary evaporation and nitrogen blow down, a 150 µL aliquot was removed for detailed organic analysis by GC-MS, and the remainder was used for bioassays. The samples were analyzed for alkanes, aromatic
and aliphatic carboxylic acids, steroids, substituted and unsubstituted PAHs, hopanes, steranes, levoglucosan, resin acids, and other compounds. The organic analysis method used for these samples has been described previously (36); however, certain details have been modified. Because the DCM extracts were also to be used for bioassays, the internal standard was added to the aliquot designated for GC-MS analysis after the Soxhlet extraction (36). The internal standard was also included in the quantification standards that were run at three dilutions twice during each set of sample runs in GC-MS. Three point calibration curves for each compound in the standard were quantified using the internal standard that most closely mimicked each compound on the basis of structure, polarity, molecular weight, and retention time. Compounds in the samples were quantified using the response curves calculated from the quantification standard and the same internal standard reference. Compounds that were not represented in the quantification standards were calculated by estimating the response curve from the compound in the standard that was most similar in structure, polarity, molecular weight, and retention time. Bioassays. Bioassays were conducted using reconstituted hard freshwater and DCM extracts of the filters. Each DCM extract was exchanged with 1 mL of DMSO and then diluted with reconstituted hard freshwater to 150 mL. The reconstituted hard freshwater extracts were also diluted to 150 mL with additional reconstituted hard freshwater before use in the bioassays. A dilution series composed of five different concentrations was run for both bioassays. To ensure no toxicity was due to the filter material or the method, quartz TSP filter blanks and solvent blanks were included with each bioassay. Positive controls were not included in this study. During the 48 h C. dubia acute test, the five dilutions for each sample were added to 30 mL polycarbonate cups. Each concentration was replicated four times with five neonates (1.7 g L-1 (38), respectively. The total ambient organic mass concentration did not change significantly over the course of the June or July sampling events. Indiana B in November 2000 also had a high ambient mass concentration (50 µg/m3) but a bulk chemical profile very different from that of the July 2001 regional high mass event and was toxic only in the DCM extract (Figure 4). Indiana B had a large contribution from water-soluble
calcium and magnesium, as well as the highest sampled elemental carbon concentration during this study (3.8 µg/m3). A much larger percentage of the mass for the October 2001 event was in the “other species” category, which could include water, additional oxygen and nitrogen in the organic mass component, and crustal materials such as silicates, which were not measured in this project. The calcium, magnesium, and iron were not correspondingly high, however. Since these sites were close to the lakeshore, which is VOL. 39, NO. 4, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1003
FIGURE 5. Select water-soluble metals for atmospheric TSP samples collected at four sites around the southern shore of Lake Michigan and one site at Seney National Wildlife Refuge in Michigan’s Upper Peninsula. predominantly quartz sand, a preponderance of silicates would be expected in the crustal material. The wind speeds during the first day of this event gusted up to nearly 50 mph, resulting in TSP mass loadings ranging from 30 to 50 µg/m3 (the October sampling event is reported as a 3 day composite). The high loadings of crustal material in the October composites did not result in an increase in the toxicity of the sample. The October composites were some of the least toxic of the data set (Figures 3 and 4). This may be due in part to dilution of the toxic species in the atmospheric PM by the more benign crustal material. For aquatic organisms the primary method of toxicity by crustal materials or sediments is light depletion or sedimentation covering of the habitat (39), neither of which could occur with the low concentration introduced from the aerosol extract. In general, the trends in organic mass concentrations do not appear to explain the trends apparent in the toxicity (Figures 3 and 4). The organic mass concentrations were very similar in many of the samples, but the toxicity of the DCM extract did not track with these values. August 2000 samples had the highest percentage of mass as organic mass (from 46% for Michigan A to 32% for Milwaukee A) and high toxicity in both the DCM and reconstituted hard freshwater extracts. But this is not a consistent marker for high toxicity, as the November 2000 Indiana B was also very toxic and had a very low percent organic mass at 13%. The ambient mass concentration does not qualitatively correlate with either the DCM or the reconstituted hard freshwater extract toxicity. The toxicities of the reconstituted hard freshwater extracts for the summer sampling events (June 2002, July 2001, and August 2000) were very similar, while the ambient concentrations of bulk species for these days were very different. This was true regardless of whether the LC50 values are expressed in toxicity per unit mass of aerosol in the test water or toxicity per volume of air in the test water. 1004
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 4, 2005
Toxicity and Detailed Chemistry of Water Extracts. For the detailed chemistry, metals were measured in the water extract and organic compounds were measured in the DCM extract. It should not be inferred that no organic components were present in the water extract, or vice versa. In cases where an organic component has significant water solubility, this will be mentioned in the discussion. Additionally, it should be noted that the detailed constituents discussed in this paper represent what was analyzed in the two extracts, while additional and possibly toxic compounds may remain unextracted by either method. The water-soluble metals measurements showed consistently high ambient concentrations for barium, copper, manganese, and zinc (Figure 5). Trace quantities of other metals may have impacted, but likely not controlled, the toxicity, and were not quantifiable in this study. Copper dominated these metals in all the samples except the manganese mass fraction in Indiana B and C in November. Copper, a known toxicant, was measured in extremely high concentrations on two separate days in two different locations. The July 2001 Indiana H and November 2000 Waukesha D samples were both above 20 mg/g of TSP or 1.2 µg/m3. This was over twice as high as any other sample in this data set and an order of magnitude greater than the average ambient concentration for the remaining samples. For the July event, the water-soluble metals concentration does track with the trend seen in the bulk chemistry with significantly higher concentrations in H than G samples in both mass fraction and ambient concentration (Figure 5). Copper, zinc, and barium are all emitted by motor vehicles (40), and in most of the samples barium and manganese concentrations seem well correlated. The exceptions are the November Indiana samples B and C, especially B, which had a very high manganese mass fraction (2.6 mg/g of TSP). Deviations probably indicate input from a local point source, which is a significant source that emits pollutants at a fixed point in space.
FIGURE 6. Detailed organic characterization by compound class for atmospheric TSP samples collected at four sites around the southern shore of Lake Michigan and one site at Seney National Wildlife Refuge in Michigan’s Upper Peninsula. Comparing the soluble metals data with the corresponding reconstituted hard freshwater extract toxicity (Figures 3 and 5), the extreme copper values for the July 2001 Indiana H sample and the November 2000 Waukesha D sample match up with high toxicity. Most of the summer samples (June 2002, July 2001, and August 2000) had both high water-soluble copper values (>1.5 mg of Cu/g of TSP) and caused mortality at low concentrations (