Mutagenicity and in Vivo Toxicity of Combined Particulate and Semivolatile Organic Fractions of Gasoline and Diesel Engine Emissions

JeanClare Seagrave*,1, Jacob D. McDonald*, Andrew P. Gigliotti*, Kristen J. Nikula*,2, Steven K. Seilkop{dagger}, Michael Gurevich{ddagger} and Joe L. Mauderly*

* Lovelace Respiratory Research Institute, 2425 Ridgecrest Dr. SE, Albuquerque, New Mexico 87108; {dagger} SKS Consulting Services, 3942 Rives Chapel Rd., Siler City, North Carolina 27344; and {ddagger} Office of Heavy Vehicle Technologies, U.S. Department of Energy, 1000 Independence Ave., SW, Washington, DC 20585

Received May 20, 2002; accepted August 22, 2002


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Exposure to engine emissions is associated with adverse health effects. However, little is known about the relative effects of emissions produced by different operating conditions, fuels, or technologies. Rapid screening techniques are needed to compare the biological effects of emissions with different characteristics. Here, we examined a set of engine emission samples using conventional bioassays. The samples included combined particulate material and semivolatile organic compound fractions of emissions collected from normal- and high-emitter gasoline and diesel vehicles collected at 72°F, and from normal-emitter groups collected at 30°F. The relative potency of the samples was determined by statistical analysis of the dose-response curves. All samples induced bacterial mutagenicity, with a 10-fold range of potency among the samples. Responses to intratracheal instillation in rats indicated generally parallel rankings of the samples by multiple endpoints reflecting cytotoxic, inflammatory, and lung parenchymal changes, allowing selection of a more limited set of parameters for future studies. The parameters selected to assess oxidative stress and macrophage function yielded little useful information. Responses to instillation indicated little difference in potency per unit of combined particulate material and semivolatile organic compound mass between normal-emitter gasoline and diesel vehicles, or between emissions collected at different temperatures. However, equivalent masses of emissions from high-emitter vehicles of both types were more potent than those from normal-emitters. While preliminary in terms of assessing contributions of different emissions to health hazards, the results indicate that a subset of this panel of assays will be useful in providing rapid, cost-effective feedback on the biological impact of modified technology.

Key Words: exhaust emissions; diesel; gasoline; comparative toxicity; intratracheal instillation; mutagenicity; particulate material; semivolatile organic compounds.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Air pollution is statistically associated with a wide range of pulmonary and extrapulmonary cancer and non-cancer health effects, including reduced lung function, aggravation of asthma, increased susceptibility to respiratory infections, cardiac arrhythmia, and increased cardiovascular and respiratory mortality (Brunekreef et al., 1997Go; Hoek et al., 2001Go; Pope, III, et al., 2002Go; Samet et al., 2000Go; van Vliet et al., 1997Go). Several studies have specifically implicated fossil fuel combustion products in general, and mobile source emissions in particular, as strong contributors to the health burden (Janssen et al., 2002Go; Laden et al., 2000Go; Mar et al., 2000Go). The particulate material (PM) fraction of diesel emissions has been the focus of many toxicological studies, but combustion emissions also contain semivolatile organic compounds (SVOCs) and gas-phase constituents. Efforts to reduce the total emission rates have led to modifications in fuel, engine, and after-treatment technologies, but little is known about the health impact of these changes. In some cases, modifications of fuel have increased the potential hazard (Mi et al., 1998Go). Rapid, cost-effective methods to screen emissions are needed to ensure that altered composition does not increase toxicity, to compare the potential hazards of competing technologies (e.g., gasoline, diesel, natural gas), and to determine the physicochemical species most strongly associated with health hazards. The many competing emission-control strategies, the large number of laboratories developing and testing new technologies, and the rate at which technologies are evolving preclude inhalation studies in animals for each case. Comparisons of large numbers of emission types are therefore necessarily limited to short-term biological assays of collected emission samples, with further evaluation by inhalation studies done only in selected cases.

Previous comparative toxicity studies of engine emission samples focused primarily on the PM and PM-associated organic material (Bernson, 1983Go; Schuetzle and Lewtas, 1986Go; Ulfvarson et al., 1995Go; Yuan et al., 1999Go) and on carcinogenic/mutagenic endpoints (Bagley et al., 1993Go; Carraro et al., 1997Go; Clark et al., 1982Go; Crebelli et al., 1991Go). However, studies indicating close temporal associations between elevations in PM and health effects (Pope, III, 1999Go; Samet et al., 2000Go) implicate non-cancer effects. These effects are also associated with proximity to vehicle traffic (Brunekreef et al., 1997Go; van Vliet et al., 1997Go), suggesting increased hazard associated with exposure to fresh emissions. Under these conditions, inhaled emissions would include SVOCs with diverse chemical composition overlapping that of the particulate fraction (Lloyd and Cackette, 2001Go; Sera et al., 1994Go). The potential health impacts of SVOCs have received less attention than those from PM or pollutant gases, primarily because SVOCs are infrequently measured. However, given the overlap in chemical composition and concurrent exposure, more appropriate comparisons of toxicity of collected emissions would include both the SVOC and PM fractions.

We have examined the ability of a set of conventional short-term bioassays to characterize responses to collected engine emissions (Seagrave et al., 2000Go, 2001Go). Based on that work, selected assays were used in the present study to compare the potencies of combined PM + SVOC emission samples collected from normal- and high-emitter gasoline and diesel vehicles. The study had two aims: (1) to compare the responses of different endpoints as a step toward selecting response indicators for future studies of collected emission samples; and (2) to evaluate the relative mutagenic, cytotoxic, and inflammatory potencies of combined PM + SVOC emissions from a limited range of different vehicle types. Selection of appropriate methods for assessing the relative toxicity of emission samples and their use to describe the relative potency of selected emission samples, as described herein, are significant steps toward advancing our understanding of the relative contributions of gasoline and diesel vehicles to the health hazards of mobile-source emissions.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Experimental Design
The sample matrix consisted of PM and SVOC fractions collected from light and medium-duty gasoline and diesel vehicles having either normal or higher than normal PM emissions. Separate samples were collected from some vehicles while operating at room temperature and in a cold environment. The two fractions were recombined in the proportions present in the original emissions. The combined material was tested for mutagenicity by determining revertants in Salmonella and for toxicity using intratracheal (i.t.) instillation into rats. Genotoxicity was measured in two strains of Salmonella, with and without rat liver microsomes (S9), as a source of metabolic activation over a wide range of doses. The categories of responses to i.t. instillation of the materials, detected in bronchoalveolar lavage fluid (BALF) and by histopathology, included general toxicity, cytotoxic responses, inflammatory responses, parenchymal changes, oxidative stress, and macrophage functional changes (Table 1Go). Cytotoxicity parameters included lactate dehydrogenase (LDH) and increased protein in bronchoalveolar lavage fluid (BALF), and histopathology indicative of cytotoxicity. Inflammatory parameters included inflammation-associated histopathology and BALF cytokines (MIP-2 and TNF{alpha}), total and differential cell counts, and ß-glucuronidase, an indicator of increased macrophage activity or lysis. The principal measure of parenchymal changes was histopathology, but release of alkaline phosphatase into lavage fluid was also considered in this category, based on evidence that type II cells are the source of this enzyme (Henderson et al., 1995Go). Oxidative stress was assessed by measuring total glutathione and oxidized proteins in BALF, and the levels of superoxide and peroxide production produced by macrophages without additional stimulation. In addition, the ability of macrophages isolated from the exposed rats to produce superoxide and peroxide in response to stimulation with opsonized zymosan was measured. Finally, general toxicity was indicated by increases in the ratio of lung weight to body weight and by a summary score for histopathological changes. These in vivo assays were performed using 3 doses, with responses measured at times ranging from 4 h to 1 week. Doses were selected to produce a range of sub-lethal responses as determined by preliminary range-finding experiments (see Intratracheal Instillation).


View this table:
[in this window]
[in a new window]
 
TABLE 1 Response Categories for Toxicity in Rat Lungs
 
Emission Samples and Control Materials
The samples were provided from a multi-laboratory program sponsored by the Office of Heavy Vehicle Technologies, U.S. Department of Energy (Seagrave et al., 2000Go). PM and SVOC fractions of the emissions were collected at Southwest Research Institute (San Antonio, TX). Samples were extracted from collection media at Desert Research Institute (Reno, NV), and emissions from groups of vehicles were pooled as described below. Aliquots of the PM and SVOC fractions from each sample were reserved for detailed analysis of chemical composition, which, along with emission rates, will be reported elsewhere. Additional aliquots of the separate PM and SVOC fractions for each vehicle or vehicle group were shipped on ice to Lovelace Respiratory Research Institute (LRRI) for evaluation of toxicity as described below.

Automobiles, sport utility vehicles, and pickup trucks ranging from 1976 to 2000 models were selected to include both those having emissions within the normal range expected for the vehicle type and age and those having visibly higher than normal emissions (Table 2Go). Although no effort was made to ensure that the vehicles represented national average vehicle type, age distributions, or emissions, the vehicles comprised a rough cross-section of the in-use gasoline and diesel-powered, light-duty fleet and provided a spectrum of samples useful for the purposes of this study.


View this table:
[in this window]
[in a new window]
 
TABLE 2 Emission Sample Sources, Composition, and Doses for Intratracheal Instillation
 
A normal-emitter gasoline group of five vehicles included a 1982 Nissan Maxima (initial odometer reading 190,203 miles), a 1993 Mercury Sable (70,786 miles), a 1994 GMC 1500 pickup truck (68,325 miles), a 1995 Ford Explorer (76,733 miles), and a 1996 Mazda Millenia (35,162 miles). Emission samples from this group were collected while operating at an ambient temperature of 72°F and pooled into a single, normal-emitter gasoline sample for toxicity evaluation (G). The effect of temperature was investigated by collecting samples from the same vehicles operated at an ambient temperature of 30°F (G30). Two individual high-emitter gasoline vehicles were sampled at 72°F only: a 1976 Ford F-150 pickup truck (199,499 miles) emitting visible black smoke (BG) and a 1990 Mitsubishi Montero (184,583 miles) emitting visible white smoke (WG). A current-technology diesel group of three vehicles included a 1998 Mercedes Benz E300 (47,762 miles), a 1999 Dodge 2500 pickup truck (37,242 miles), and a 2000 Volkswagen Beetle TDI (7,495 miles). Samples collected from these vehicles, operated at 72°F, were pooled into sample D; samples collected at 30°F were pooled into sample D30. These vehicles incorporated combustion chamber, fuel injector, and other technologies representative of light- and medium-duty diesel vehicles current at the time of the study, but were not equipped with more recent after-treatment devices and did not include older, normally higher-emitter, diesel vehicles. A single high-emitter diesel vehicle, a 1991 Dodge 2500 pickup truck (35,455 miles), was sampled at 72°F only (HD).

The vehicles were sampled while operating on chassis dynamometers on the California Unified Driving Cycle (Ho and Winer, 1998Go), with commercial fuel and crankcase oil "as received." Emissions were diluted in a stainless-steel dilution tunnel, and samples were extracted from the tunnel and collected using a positive-displacement, constant-volume sampling system. The PM samples were collected on high-volume, fluorocarbon-coated glass fiber filters, and the SVOC samples were collected using polyurethane foam and polystyrene/divinylbenzene polymer (PUF/XAD) (Westerholm et al., 2001Go) in canisters downstream from the filter. Material from the filter was extracted into acetone by gentle brushing and sonication, resulting in a suspension of PM. Material from the PUF/XAD matrix was extracted into acetone using a Soxhlet apparatus. Both fractions were concentrated by evaporation under nitrogen at room temperature. The SVOC fraction is therefore operationally defined for these studies as the material bound to the matrix but not volatile at room temperature. Acetone was selected as the solvent based on its low toxicity, high volatility, and lack of effect on key parameters of biological responses at concentrations up to 2%. Specifically, these concentrations caused no direct effects in the test systems and did not affect responses to i.t. instillation of the positive controls, diesel soot and silica (Si), as determined by preliminary experiments. The masses of PM and SVOC in the samples were estimated by transferring 100-µl aliquots of the materials into preweighed tubes and allowing the acetone to evaporate at room temperature overnight. Recovery of the PM fraction in the extract, relative to pre- and post-collection weights of the filters, was approximately 65% for D and D30, 77% for BG and HD, 87% for G30, and 98–104% for G and WG. No similar calculation of SVOC recovery was possible since the mass was small relative to the mass of the collection matrix. Some losses of the most volatile components of the SVOC fraction were expected during the concentration phase, but additional losses during sample handling and reconstitution were presumed to be small. PM:SVOC emission ratios were calculated from these masses and known sample flows (Table 2Go). Individual ratios for each vehicle or pooled vehicle group were used to recombine the two fractions into a single sample immediately before toxicity testing.

National Institutes of Standards and Technology Standard Research Material 2975 diesel soot (DS) was included in all experiments as an internal control. The DS sample was suspended in acetone at approximately 150 mg/ml and stored at –20°C to be treated similarly to the emission samples. However, it is important to note that this control material differs from the tested samples in that it was not tested in combination with a SVOC fraction. In the i.t. instillation experiments, Si was used as a strongly proinflammatory and fibrogenic positive control particle (Prasad et al., 2000Go), and FluoresbriteTM YG 0.5-micron polystyrene/latex fluorescent microspheres (FMS) were used as a relatively low toxicity control particle (Brown et al., 2001Go). The emission samples were tested at doses of PM in a range previously shown to induce responses to DS. However, the results are reported as responses per unit of total mass of PM + SVOC.

Measurement of Bacterial Mutagenicity
Combined PM and SVOC samples were transferred to dimethylsulfoxide solvent and shipped overnight at 0°C to Bioreliance (Rockville, MD) for mutagenicity testing, using the Ames Salmonella plate-incorporation assay (Ames et al., 1975Go). Mutagenicity was assessed in Salmonella strains TA98 and TA100, each with and without metabolic activation by the microsomal fraction of Aroclor-induced rat liver homogenates (S9) (Maron and Ames, 1983Go). Strain TA98 revertants indicate mutation via frame-shift mutations; TA100 revertants indicate either frame-shift or base pair substitution mutations. Each sample was tested in duplicate at concentrations ranging from 25 to 5000 µg/plate and in parallel with negative (solvent) and positive controls. The positive controls were 2-nitrofluorene for TA98-S9, sodium azide for TA100-S9, and 2-aminoanthracene for TA98+S9 and TA100+S9. Colonies on plates with sufficient precipitate to interfere with electronic counting were counted by hand; other plates were counted using an automated colony counter. All samples met the response criteria of dose-related increases in mean revertants per plate, rising to at least twice the negative control value. However, high concentrations resulted in toxicity to the bacteria and suppression of the observed mutation rates. Responses were therefore quantified as the slope of the initial linear portion of the dose-response curve fit to the linear model as described under statistical analysis of data below.

Measurement of in Vivo Toxicity
Animals.
Male F344/Crl BR rats were purchased from Charles River Laboratories. The rats, 8 ± 1 weeks old at receipt, were quarantined for 3 weeks before dosing and were 11 ± 1 weeks old at the time of instillation. The rats were housed two per shoebox cage with hardwood chip bedding and filter caps. The light/dark cycle was 12-h light/12-h dark with lights on at 0600 h. Food (Harlan Teklad Lab Blox) and water were provided ad libitum. The room temperature was maintained at 20–22°C, with a relative humidity of 20–50%. LRRI animal facilities are fully accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International.

Reagents and supplies.
General chemicals and reagents were purchased from Sigma Chemical Company (St. Louis, MO), unless otherwise specified. Tissue culture medium and supplies were obtained from Invitrogen (Carlsbad, CA).

Sample preparation.
Samples were tested in pairs in the order received. This study was accomplished over the course of 10 months, using 6 separate sets of instillations. The PM fraction of each sample in acetone was vortexed vigorously to assure a uniform suspension and diluted to 10x the desired final concentration in sterile 0.9% NaCl containing 0.1% Tween-80, a nontoxic surfactant. The suspension of PM was sonicated (uncapped) for 15 min in a low-power sonic bath at room temperature in the dark, to allow evaporation of the majority of the acetone. The volume of the SVOC fraction constituting the appropriate mass ratio to this mass of PM was then added, and the suspension was capped and sonicated at high power for 1 min in a "cup" style sonicator (Branson, Danbury, CT). The material was then diluted with 9 volumes of 0.9% NaCl, and the suspension was sonicated for one additional minute. The final acetone content of this solution was adjusted to 1%. Dilutions were made in 0.9% NaCl with 1% acetone and 0.01% Tween-80. An extraction control was prepared similarly from the concentrated acetone extract of a filter and a PUF/XAD canister not exposed to exhaust materials, using volumes equivalent to the highest dose of the samples.

Intratracheal instillation.
Rats were anesthetized using 5% halothane in oxygen with nitrous oxide. Once anesthetized, each rat was intubated orally with an i.t. cannula and instilled with an emission sample or control material suspended in a total volume of 0.5 ml saline. The rats were allowed to recover and returned to their cages. A group of animals was also instilled with the solvent control at every time point.

Groups of 5 rats for each sacrifice time were instilled with the combined sample at 3 doses (shown in Table 2Go), with the highest dose determined, by a preliminary range-finding experiment, to cause responses without lethality. Previous experiments using DS, additional samples of diesel PM collected at LRRI, the highly proinflammatory mineral dust Si, and ambient air PM suggested that a dose of 3 mg PM/rat would cause substantial responses without lethality. However, for many of the emission samples, this dose of PM, when combined with the appropriate amount of the SVOC fraction, caused unacceptable mortality, and lower doses were therefore used.

Because this study required 10 months and 6 separate sets of instillations, each set included both the solvent (extraction control) and a range of doses of DS as an internal control to assure consistency of the responses. Doses of DS in each series included 0.3/ and 1 mg/rat, as well as either 0.1, 2, or 3 mg/rat depending on the doses of the samples. Si (0.3, 1, and 2 mg/rat) was tested twice, and FMS (0.3, 1, and 3 mg/rat) was tested once.

Euthanasia and processing.
Responses were evaluated at 4 h, 12 h, 24 h, and 1 week following instillation. Early experiments also included a 4-week time point, but this time was eliminated in later experiments, based on observations indicating a return to near-baseline values for all parameters by 1 week. Rats were killed by pentobarbital overdose, and body weights were recorded. The heart and lungs were isolated, the heart was removed, and the trachea was lifted free from the thorax and severed, preventing blood from entering the lung. Lung weights were recorded. The left bronchus was clamped, and the right apical lobe was tied off. A blunt 18-gauge cannula was ligated into the trachea, and the right cardiac, diaphragmatic, and intermediate lobes were then lavaged with 2 aliquots of 3 ml of Dulbecco’s phosphate-buffered saline. The recovered BALFs were combined, the volume was recorded, and the BALF was kept on ice until processed. Fluid recovery was generally between 5 and 5.5 ml. After lavage, the right apical lobe was removed and snap frozen in liquid nitrogen. The left bronchus was then unclamped, and the lungs were filled with neutral buffered 10% formalin, the trachea was ligated, and the lungs were immersed in the same fixative for at least 48 h.

Lung lavage cells and fluid were collected for analyses. BALF samples were analyzed for glutathione at the 4 h and 12 h measurement times only, based on expectations of rapid responses of this antioxidant. For this assay, 0.5 ml of BALF was immediately centrifuged, and the supernatant was treated with sulfosalicylic acid for analysis of glutathione. At all measurement times, an aliquot of BALF was diluted in red blood cell-lysing buffer (Roo’s solution: 9 g/l NH4Cl2 + 1 g/l K2CO3 + 0.0372 g/l EDTA) and 0.02% (final) Trypan blue to determine total cell numbers and cell differentials. The total number of leukocytes was determined using a hemocytometer. Cytocentrifuge preparations were used for evaluation of differential cell counts. The slides were stained with Wright-Giemsa using a Hema-Tek Slide Stainer, and the percentages of macrophages, polymorphonuclear leukocytes (PMNs), and lymphocytes in 300 cells per sample were determined using a 20x objective.

The remaining BALF was centrifuged (10 min at 1000 rpm). LDH (Gay et al., 1968Go) and total BALF protein (Watanabe et al., 1986Go) were measured in the supernatant using a Monarch clinical chemistry analyzer. ß-Glucuronidase enzyme activity was assessed as previously described using p-nitrophenyl-b-D-glucuronide as the substrate (Henderson et al., 1985Go). Alkaline phosphatase activity was measured using the BioRad (Hercules, CA) kit (2 h incubation). Total glutathione in the BALF acid supernatant was measured by the dithionitrobenzoate recycling method (Tietze, 1969Go). Protein carbonyls in samples diluted to the same protein concentration were derivatized with dinitrophenyl hydrazine and assessed using an ELISA (Buss et al., 1997Go) based on the Oxyblot kit (Intergen, Purchase, NY). TNF{alpha} and MIP-2 were measured in cell-free BALF using commercial ELISA kits (Biosource International, Camarillo, CA).

The cell pellet was then resuspended in RPMI 1640 medium with FBS, penicillin, streptomycin, and glutamine, and the cells were plated in 96-well tissue culture plates at 100,000 cells/well to measure production of superoxide and peroxide. The cells were allowed to adhere to the plates for 1 h, then were gently washed with HEPES-buffered Earle’s salts. Production of superoxide (using reduction of acetylated cytochrome C [Pick, 1986Go; Nasrallah, Jr., et al., 1983Go]) and peroxide (using peroxidase-catalyzed luminescence of luminol [Pick, 1986Go]) was measured in the presence and absence of opsonized zymosan (200 µg/ml). Responses were measured in the presence and absence of enzymatic catalysts (10,000 U/ml superoxide dismutase to eliminate superoxide and 3200 U/ml catalase to eliminate peroxide) to define the specificity of the reactions.

Histopathology was assessed in the fixed left lung by light microscopy. Three transverse slices about 3 mm thick (one through the lung cranial to the hilus, one through the main airway just caudal to the hilus, and one near the end of the main axial airway) were embedded in paraffin, sectioned (5-mm), and stained with hematoxylin and eosin. The magnitude, character, and location of the responses to the samples were scored using a scale from 0 (no pathology) to 5 (extreme pathology) for several pathological endpoints (Table 1Go). For analysis, the histopathological scores were pooled into categories of responses indicating cytotoxicity, inflammation, or parenchymal changes. For each animal, the sum of the individual scores for each type of pathology within a category was calculated. The sum of all scores was also analyzed as an overall pathological response.

Statistical analysis of data.
The time course of responses was examined. Maximal treatment-related biological activity was observed at 24 h for all endpoints except for glutathione, oxidized protein, and cytokines, for which the strongest response occurred at 4 h. Responses at these times were plotted as a function of the combined mass (PM + SVOC) of each emission sample. Dose-response functions were fit to these data, and comparisons between the toxicological potencies of the different emissions were based on the slope coefficients from the fitted dose-response curves.

For each endpoint, a single regression model with emission-specific slope coefficients was utilized:

where Yki = endpoint value in the kth emission group at concentration i, concki = concentration i in kth emission, ck = fitted estimate of control group mean, bk = fitted slope estimate for emissions group k, and log = natural logarithm.

For the mutagenicity data, which strongly showed linear initial dose responses up to the level of toxicity to the bacteria, Yki was substituted for log(Yki) in this model, and for variables that could exhibit values of zero, log(Yki + 1) was analyzed. Generalized least-squares regression was employed to estimate the dose-response coefficients for endpoints that exhibited a relationship between sample mean and standard deviation, using reciprocals of the sample variances as weights (Neter et al., 1996Go); ordinary least-squares regression was employed for all other endpoints. The overall fit of this regression model for each endpoint was measured with the multiple correlation coefficient (R2). Statistical significance of individual emission-specific toxicological potency estimates (bk != 0) was assessed using t-tests associated with the regression model. Comparisons between pairs of emission-specific toxicological potencies (bk) were evaluated with F-test contrasts (Searle, 1971Go) calculated by the SAS® software system (Cary, NC). To evaluate differences among emission samples, for each endpoint, p values from the pairwise F-tests were adjusted for multiple comparisons (21 paired comparisons among the 7 samples) using the modified Bonferroni procedure devised by Hochberg (1988)Go. Statistical significance was assessed at p = 0.05.

Endpoints were grouped into response categories (Table 1Go), and the relative rankings of the potency estimates of the samples for the different endpoints within these categories were compared. Potency estimates across endpoints within each category were uniformly scaled by dividing each endpoint’s potency estimates and associated errors by the largest observed potency value for that endpoint. The endpoint within a response category with the smallest scaled standard errors of the potency estimates was selected as the key endpoint. In most cases, this endpoint also produced the highest degree of discrimination, i.e., the largest number of statistically different pairwise comparisons among potency estimates.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Bacterial Mutagenicity
Figure 1Go shows the dose-response curves for mutagenicity in strain TA98 assessed with S9, the condition that provided the standard errors on the uniformly scaled potency estimates. The estimated potencies and statistical significance of differences between mutagenicity in both strains with and without S9 are presented in Table 3Go, where bk is the slope of the linear regression over the initial linear phase of the dose-response curve and represents the mutations per µg of the combined PM + SVOC sample. In this table and all other potency tables, the lower case letters indicate the groups of samples that were not statistically different from each other. For each sample, the samples that were significantly different are also explicitly listed.



View larger version (24K):
[in this window]
[in a new window]
 
FIG. 1. Salmonella mutagenicity test. Emission samples were tested using Salmonella strain TA98 with S9 at a range of concentrations, and the slope of the dose-response curve, revertants/µg, was determined. Data shown are mean and range of duplicate plates. See Table 2Go for abbreviations.

 

View this table:
[in this window]
[in a new window]
 
TABLE 3 Mutagenicity Potency Estimates
 
All samples caused mutations in both strains. The rank order of the samples varied somewhat as a function of the bacterial strain and metabolic activation conditions. In all cases, D30 was the most mutagenic. In TA98, metabolic activation with S9 increased the mutagenicity of three gasoline samples (G, G30, and BG). S9 had much smaller effects on D and WG and did not affect the mutagenicity of HD, but reduced the mutagenic potency of D30 approximately 20%. Metabolic activation reduced the mutation rate in TA100 for WG and HD without affecting the potency of G, BG, G30, D, or D30.

The scaled potencies across mutagenicity endpoints shown in Figure 2Go emphasize the differences and similarities among the different assay conditions. In agreement with the generalization that these materials contain primarily direct-acting mutagens, good correlations were observed between the rankings indicated by TA98 with and without S9 (with the exception of G30, for which mutagenicity was enhanced approximately 3-fold by the presence of S9), and between TA100 with and without S9. However, the assays using TA100 showed relatively greater mutagenicity of the three high-emitter samples (WG, BG, and HD), as well as D, compared to assays using TA98. Emission samples collected in the cold were more mutagenic than samples from the same vehicles collected at 72°F, although the difference was not statistically significant for G30 versus G in TA98 without S9. Combining data from assays run in the presence and absence of S9, for TA98 the ranking of the samples from highest to lowest potency was D30, HD, WG, G, D, BG, with an ambiguous ranking of G30. For TA100 the rank order was D30, WG, HD, D, BG, G30, G, with little discrimination among the samples with intermediate potency.



View larger version (16K):
[in this window]
[in a new window]
 
FIG. 2. Comparisons of mutagenic potencies. Mutagenic potencies as shown in Table 3Go were scaled by the maximal potency for each strain and metabolic activation condition. Relative mutagenicity for each sample is shown with the scaled standard error for each potency estimate: (A) TA98 with and without S9; (B) TA100 with and without S9. Horizontal bars indicate the average of mutagenic potency estimates. Sample order is the order of the key response (TA98 + S9). See Table 2Go for abbreviations.

 
Toxicity in Rat Lungs
Reproducibility of controls.
Comparisons of the in vivo responses to the extraction control and the 1-mg/rat dose of DS during the period required to complete these studies are shown in Table 4Go. These results indicate that the responses were reproducible throughout the study. Furthermore, the extract control produced very minor responses, comparable to instillation of saline (data not shown).


View this table:
[in this window]
[in a new window]
 
TABLE 4 Reproducibility of Internal Controls
 
Time course of in vivo responses.
For most parameters assessed, the responses were maximal at 24 h, although frequently a biphasic pattern was observed of initial increases at 4 h followed by a partial resolution at 12 h, and a second increase at 24 h. Responses had largely returned to near baseline by 1 week, except, interestingly, for the DS control. In contrast, cytokine responses (TNF{alpha} and MIP-2) and the oxidative stress parameters glutathione and oxidized protein were generally maximal at 4 h. The 24-h time was therefore selected for comparisons among samples for histopathology and lavage parameters, except for the cytokines, glutathione, and oxidized protein endpoints, to which the 4-h responses were compared.

Cytotoxicity.
DS caused an intermediate response in LDH and a relatively weak protein response. The positive control material, Si, potently caused release of LDH, but only modestly increased protein leakage, while the low toxicity control material, FMS, caused small increases in both responses only at very high concentrations. None of these control materials caused significant cytotoxicity as indicated by histopathology.

For the emission samples, LDH and total protein in cell-free BALF varied in parallel and provided approximately equal discrimination of cytotoxicity across the emission samples. The histopathological indicators of cytotoxicity provided slightly less discrimination. The potencies as ranked by these three parameters are shown in Table 5Go. Lavage LDH was selected as the key endpoint, and results for this assay are shown in Figure 3Go. The curves shown are the model fit to the data. In all of the assays, the samples from the three high-emitting engines (WG, BG, HD) had the highest potency rankings. The LDH assay indicated that WG was significantly more potent than all other samples. HD and BG were not significantly different from each other by any assay, but were significantly more potent than G, G30, D, or D30 by the lavage protein assay, and BG was more potent than these samples by LDH as well.


View this table:
[in this window]
[in a new window]
 
TABLE 5 Cytotoxicity Potency Estimates
 


View larger version (18K):
[in this window]
[in a new window]
 
FIG. 3. LDH in BALF at 24 h. (A) Gasoline emission samples. (B) Diesel emission samples. Data shown are the mean and standard error of 5 rats per group, except for points marked with # (3 rats) and * (2 rats). These points may underestimate the toxicity, since data were only collected from surviving animals. The curves are the fits to the model. See Table 2Go for abbreviations.

 
In contrast to the two lavage parameters, for which G was the least potent sample, histopathology ranked this sample fourth in potency among the seven samples. There was, however, considerable uncertainty in the ability of histopathology to discriminate between the potency of G and most of the remaining samples. D was significantly more potent than D30, G, and G30 by lavage protein and LDH, although it was not different from G30 or G by the histopathological measures of cytotoxicity. The two samples collected in the cold (G30 and D30) were the least potent by histopathology, and were similar to G by protein and LDH measurements.

Uniform scaling of the potencies (Fig. 4Go) emphasizes the similarity of the cytotoxicity endpoints across the emission samples. Correlation between the two lavage parameters is excellent. This presentation of the data suggests that the histopathological indication of damage caused by WG, and possibly G30 and G, over-represents that observed by the lavage parameters. The only histopathological lesion observed in WG and G30 was alveolar hemorrhage, while G showed mild evidence of necrosis as well as hemorrhage. Both of these lesions would be expected to correlate with increased protein and LDH. In summary, the ranking of the samples for potency in inducing cytotoxicity is WG, (BG and HD), and (D, G, D30, G30), with little discrimination among samples in parentheses.



View larger version (23K):
[in this window]
[in a new window]
 
FIG. 4. Comparison of cytotoxic potencies. Cytotoxic potencies as shown in Table 4Go were scaled by the maximal potency for endpoint. Relative cytotoxic potencies for each sample are shown with the scaled standard error for each potency estimate. Horizontal bars indicate the average of the three relative potency estimates. Sample order is the order of the key response (LDH). See Table 2Go for abbreviations.

 
Inflammatory responses.
DS caused moderate increases in the inflammatory responses, including the total number of leukocytes, macrophages, PMNs/ml of BALF and both cytokines, and a very strong increase in ß-glucuronidase. In contrast, Si caused very strong increases in most of these parameters, especially MIP-2, with only moderate increases in ß-glucuronidase. The negative control material, FMS, caused responses only at very high concentrations.

The ranked potency estimates and discrimination among the emission samples for inflammatory endpoints are shown in Table 6Go. Total leukocytes in BALF provided the greatest discrimination among the samples. The ranking established by the total leukocytes in BALF agreed very well with that of PMNs and macrophages in BALF, as well as with the histopathological indicators of inflammation (Fig. 5Go). WG was the most potent sample in inducing inflammatory responses, as it was for cytotoxicity. HD and BG were the next most potent samples, with comparable rankings according to several of the assays. Similar rankings were also observed with MIP-2 measured at 4 h: WG, BG, and HD caused the greatest MIP-2 response, but were not significantly different from each other. ß-Glucuronidase provided a similar ranking, but only the potencies of HD and D were significantly greater than zero; large standard errors resulted in no significant pairwise differences among the samples. In contrast, TNF{alpha}, measured at 4 h, was suppressed by some samples (WG, BG, G30, D30, and G) and weakly stimulated by others (HD, D), with no clear correlation with any other parameter. For the most part, the assays of inflammatory responses did not discriminate well among the four samples from normal emitter vehicles (G, D, G30, and D30), which caused much weaker inflammatory responses than the three highly potent samples.


View this table:
[in this window]
[in a new window]
 
TABLE 6 Inflammation Potency Estimates
 


View larger version (14K):
[in this window]
[in a new window]
 
FIG. 5. Comparisons of inflammatory potencies. Inflammatory potencies as shown in Table 5Go were scaled to the maximal potency for endpoint. Relative potencies for each sample are shown with the scaled standard error for the 5 potency estimates that provided discrimination and concordant rankings. (Rankings for ß-glucuronidase and TNF{alpha} are not shown.) Horizontal bars indicate the average of the five relative potency estimates. Sample order is the order of the key response (total lavage leukocytes). See Table 2Go for abbreviations.

 
Parenchymal changes.
Neither DS nor FMS caused notable changes in any of the histopathological indicators of parenchymal changes; however, Si caused modest increases, primarily pneumocyte hyperplasia. The parenchymal-change histopathology provided some discrimination among the emission samples, although the differences between samples with similar ranks were not usually significant. The potency estimates are shown in Table 7Go. WG was very potent in causing these changes; pneumocyte hyperplasia was the principal effect. This was also the primary contributor to parenchymal changes induced by BG, HD, D, and D30. However, the other gasoline emission samples, G and G30, caused primarily fibrosis.


View this table:
[in this window]
[in a new window]
 
TABLE 7 Parenchymal Changes Potency Estimates
 
FMS did not affect BALF alkaline phosphatase, and Si actually decreased the activity of this enzyme. However, DS caused a response comparable to the most potent emission samples. Of the emission samples, only the potencies of BG (0.11 ± 0.05), D30 (0.06 ± 0.02), and G (0.01 ± 0.002) were significantly greater than zero, and none of them was significantly different from each other or from the other samples, primarily due to the high variability in this response. No correlation was observed between the responses measured by histopathological evidence of parenchymal change and release of alkaline phosphatase into BALF.

Oxidative stress.
Superoxide responses provided the greatest discrimination among the oxidative stress endpoints (Table 8Go). For nearly all samples, including the DS, FMS, and Si control materials, both superoxide and peroxide production were reduced below levels of cells from control animals. WG was significantly more potent in suppressing superoxide production than any other sample, followed by HD and BG (not significantly different from each other, but significantly more potent than the remaining samples). Weak or insignificant effects, as indicated by the potency parameter estimates in the model, were observed with the other samples. However, it should be noted that sample G, at the lowest dose tested, increased superoxide (to 260% of control levels); at higher doses, this sample produced suppression similar to the other samples. The biphasic dose response resulted in no significant potency being indicated by the model. Similar ranking of potency was observed for peroxide production (Table 8Go), although, in this case, HD was found to be slightly but not significantly more potent than WG, and little discrimination was observed among the other samples.


View this table:
[in this window]
[in a new window]
 
TABLE 8 Oxidative Stress Potency Estimates
 
Glutathione levels at 4 h were strongly decreased by Si, and weakly decreased at the low dose and increased at the highest dose of DS. FMS did not affect this parameter. Few of the emission samples affected glutathione sufficiently to reach statistical significance based on the model used for these comparisons: only the responses of WG and D were significant (Table 8Go), and WG achieved the significant increase due to strong increases at high doses, despite decreases at low doses. HD caused an apparent decrease that did not reach statistical significance. The maximal changes in oxidized protein were also observed at 4 h. Si caused small increases, but neither DS nor FMS affected this response. Significant increases were observed for D and D30, but not for the gasoline samples or HD (Table 8Go). Little correlation among the different responses in the oxidative stress category was observed: WG affected all parameters, although the effect on oxidized proteins was not significant. The other two samples from high-emitter vehicles (HD and BG) suppressed superoxide and peroxide production but did not significantly affect glutathione or oxidized proteins. In contrast, D did not affect superoxide or peroxide production, but increased glutathione and oxidized proteins, while D30 weakly inhibited superoxide and peroxide production, strongly increased oxidized proteins, but did not significantly affect glutathione. The two samples from normal-emitter gasoline vehicles (G and G30) only weakly affected the superoxide and peroxide responses and did not significantly affect glutathione or oxidized proteins.

Macrophage functional changes.
Exposure to these samples affected the ability of the alveolar macrophages to respond to exogenous stimulation with complex dose-response relationships. Si suppressed both superoxide and peroxide production responses, while FMS had little effect on superoxide production but, at low doses, potentiated peroxide. Little or no response was observed for some samples, while increases at low doses and suppression at high doses were observed for others (including DS), and suppression at low doses and increases at higher doses were observed for still others. These complex and biphasic dose-response curves resulted in relatively poor fits to the model (R2 = 0.75 for superoxide, 0.59 for peroxide). The potency estimates were significantly different from zero only for WG (–2.43 ± 1.18), BG (–1.13 ± 0.34), and D30 (–0.71 ± 0.13) for the stimulated peroxide responses, and for HD (–0.15 ± 0.007), G30 (-0.07 ± 0.005), and BG (-0.06 ± 0.01) for the stimulated superoxide responses. However, none of these potencies was significantly different from each other or from the other samples. There was little agreement in the ranking or potency for these two responses.

General toxicity.
Si caused strong increases, while DS caused moderate increases in both total histopathology and lung weight, while FMS had essentially no effect. The histopathology parameter, which provided slightly smaller error estimates and better discrimination, was selected as the key endpoint. Emission sample potencies for these endpoints are listed in Table 9Go, and the uniformly scaled potency comparisons across endpoints are shown in Figure 6Go. As described for the histopathological indications of cytotoxicity, WG appeared to have caused larger increases in histopathology than would be predicted on the basis of the lung weight. This sample was clearly discriminated from all other samples by histopathology, but was not significantly different from any other samples by lung weight (at least partly as the result of a rather large error on the estimate of the potency of WG for changes in lung weight measurements). However, both parameters ranked WG as the most potent emission sample, and G30 as the least potent. The samples from the other high-emitter vehicles (HD and BG) were also quite potent. Similar potencies among the other samples were observed for these responses, although both suggested that the emission samples collected in the cold (D30, G30) were slightly less potent than the samples from the same vehicles collected at room temperature (D, G).


View this table:
[in this window]
[in a new window]
 
TABLE 9 General Toxicity Potency Estimates
 


View larger version (23K):
[in this window]
[in a new window]
 
FIG. 6. Comparisons of general toxicity potencies. Potencies shown in Table 9Go were scaled by the maximal potency for endpoint. Relative potencies for each sample are shown with the scaled standard error. Horizontal bars indicate the average of the relative potency estimates. Sample order is the order of the key response (total histopathology). See Table 2Go for abbreviations.

 

    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
In this study, we used bacterial mutagenicity and i.t. instillation in rats as methods for screening the relative toxicity of a variety of emission samples. Bacterial mutagenicity does not correlate directly with carcinogenicity, but the assay is widely accepted as an indicator of DNA-damaging chemicals. Although responses to instillation and inhalation may differ, instillation, even at supraphysiological doses, is widely accepted as a useful comparative screening tool (Driscoll et al., 2000Go). The primary goals of this study were first, to determine which of the selected conventional bioassays were most useful for assessing differences in potencies of emission samples and second, to evaluate the relative mutagenic, cytotoxic, and inflammatory potencies of this unique set of emission samples consisting of PM and SVOC fractions collected from normal- and high-emitter gasoline and diesel vehicles. The results confirmed that the screening tools described here could discriminate among the samples and should prove useful for the assessment of future modifications in fuels and engine and after-treatment technologies. For some classes of endpoints, a key endpoint adequately indicated the discrimination provided by the majority of the endpoints in that class, suggesting that future comparisons of emission samples might focus on these more informative endpoints. In addition, these results provided a ranking of the relative potency of this particular set of engine emission samples in causing mutagenic, cytotoxic, inflammatory, and parenchymal responses. Caution should be exercised, however, in extrapolating the relative rankings of the samples in this study to the toxicity of emissions from the broader vehicle fleet in the U.S. or elsewhere. Although this sample set provided insight into the potential differences in combined PM + SVOC toxicity due to engine type (gasoline vs. diesel), operating temperature (72 vs. 30°F), and emissions status (normal- vs. high-emitter), the extent to which differences observed in this study may reflect fleet average differences is not known.

Bacterial Mutagenicity
The Salmonella data indicated that all samples were mutagenic, with substantial differences (nearly 10-fold for TA98-S9) between the least and most mutagenic samples. All samples were mutagenic in both strains, indicating that they caused frame shift mutations as well as base pair substitutions. The ranking was affected slightly by metabolic activation conditions. S9 increased the mutagenicity in strain TA98 of the three gasoline samples (G, G30, and BG), which suggests that a portion of the mutagenicity may be due to the presence of polycyclic aromatic hydrocarbons, while the reduction of D30 mutagenicity by S9 suggests that the mutagenicity in this sample might be due to nitroarenes. These relatively small differences indicate that the samples acted primarily as direct mutagens, as is typical for engine emissions (Bagley et al., 1993Go). However, since S9 significantly changed the ranking of the samples, particularly in TA100, the complete screening protocol should be retained for future studies.

Cytotoxicity
There was good agreement in the ranking of the samples by the three cytotoxicity endpoints. Release of protein into the BALF most likely represents loss of epithelial/endothelial integrity, resulting either from cell death or loss of cell–cell contacts. LDH in BALF indicates damage to cell membranes. These two parameters provided an identical ranking of the potencies of the samples and were among the most sensitive assays, showing responses at low doses. Statistically significant potency was observed for all samples. Discrimination among samples was very good, and WG was significantly more potent than all other samples. All other samples were also significantly different from at least four others, with similar potencies for BG and HD, followed by D, then D30, G30, and G. Histopathological indications of cytotoxicity ranked the samples in similar order (although with somewhat less effective discrimination), except that G caused more histopathological damage, but less LDH and lavage protein, than D, D30, or G30. Given the strong agreement between LDH and protein, either LDH or lavage protein are recommended for assessing cytotoxic damage of these types of samples.

Inflammation
Inflammation is an important class of response, given indications of inflammatory effects of inhaled engine emissions in humans (Nel et al., 2001Go). The measures of cellular responses (lavage total inflammatory cells, macrophages, and PMNs) and increased inflammation in histopathological sections provided similar rankings and were reasonably sensitive. The MIP-2 assay showed good sensitivity and moderate discrimination among the samples that produced strong or intermediate responses. The ranking for this parameter was fairly similar to that provided by the other inflammation indicators, and clearly discriminated the high-emitter group (BG, WG, and HD) from the normal-emitter group (D30, G, G30, and D), with no significant differences within the two groups. In contrast, TNF{alpha} provided a very different ranking, with large errors in the potency estimates for most samples. Since TNF{alpha} is an important pro-inflammatory cytokine, it is possible that the times selected were not optimal for its analysis. ß-Glucuronidase did not discriminate among the samples. Total BALF leukocytes are recommended as the primary endpoint for assessment of inflammatory responses, due to the simplicity and good discrimination of this assay. Differential counts of PMNs and histopathological indices of inflammation may be useful secondary indicators.

Parenchymal Changes
Histopathological scores for parenchymal changes clearly placed WG as the most potent sample and D30 as the least, but discriminated poorly among the other samples. As noted under Results, however, the subtypes of parenchymal change indicators differed, with G and G30 producing more fibrosis, and other samples causing more pneumocyte hyperplasia. Although alkaline phosphatase was significantly elevated by some samples in this study, measurements of this enzyme provided little discrimination among the samples. Furthermore, the ranking for ability to increase lavage alkaline phosphatase did not agree well with the parenchymal change responses assessed by histopathology, nor did it correlate with acute cytotoxicity, inflammation, or oxidative responses, possibly suggesting a function of this protein not related to these commonly assessed indicators of toxicity. However, histopathology is the clear recommendation for assessing parenchymal changes.

Oxidative Stress
Oxidative stress may be an important mechanism of pollutant toxicity (Ball et al., 2000Go; Martin et al., 1997Go; Nel et al., 2001Go) and may result from chemical reactions with constituents of the inhaled materials (Squadrito et al., 2001Go), or through activation of the respiratory burst in leukocytes stimulated by interaction with the PM. Acute exposure of macrophages to airborne PM increased oxidant production (Goldsmith et al., 1997Go), but in vivo exposure may suppress oxidant responses (Yang et al., 2001Go). For most of the emission samples, peroxide and superoxide production were reduced following i.t. instillation of the materials. Superoxide production provided better discrimination and ranked the samples in the order WG more potent than HD and BG, followed by the group of normal-emitters, G30, D, D30, and G.

Effects of exogenous agents on changes in extracellular glutathione may be quite complex as a result of cellular modulation following initial chemical responses. Thus, the initial response to toxic insult may be to deplete the glutathione due to conjugation of this antioxidant molecule with electrophilic target molecules via glutathione S-transferase. However, feedback mechanisms frequently cause release of additional glutathione from the cells, resulting in a subsequent increase. WG and D increased glutathione levels in BALF, suggesting that for these samples, the compensatory mechanisms had been activated even at the 4-h time point. Furthermore, several samples caused biphasic dose responses with decreases at low doses and increases at higher doses, possibly indicating that the strong oxidant effects of the high doses had also activated compensatory responses. Such changes are not well described by the curve-fitting approach selected to analyze these data. Proteins that could be derivatized with DNPH were the other indicator of oxidative stress. Changes in this parameter were relatively small and did not correlate well with the glutathione measurements. The differences between these two measures of oxidative stress could have been related to changes in antioxidant status too rapid to be detected at the 4-h time point, or to dilutions of oxidized proteins by influx of serum proteins. Thus, it is unclear whether the two direct measurements of oxidation (loss of glutathione and oxidation of BALF protein) represent oxidation by the exhaust materials themselves, or early activation of the inflammatory cells followed by their inactivation. In summary, the indicators of oxidative stress assessed in the present study proved difficult to interpret. Li et al. (2002)Go have recently described a graded response to oxidative stress induced by organic extracts of diesel exhaust PM, with increased expression of heme oxygenase at low levels of oxidant stress. Given the potential importance of oxidative stress as a mediator of pollutant toxicity, assessment of this enzyme may be useful as a screening tool.

Macrophage Functional Changes
The complicated and frequently biphasic changes in stimulated superoxide and peroxide responses of alveolar macrophages precluded clear ranking of the samples according to these parameters by the method employed for this study. However, individual points on the dose-response curves for some samples showed strong suppression of the ability to respond to exogenous stimulation. Since suppression of reactive oxygen species production by diesel exhaust PM has been implicated in suppression of the ability to clear bacterial infections (Yang et al., 2001Go), these observations may have important implications regarding the effects of air pollution constituents on the function of the innate immune response. Because the current protocols did not provide adequate discrimination among the samples, they cannot be recommended in their present form for future studies. Additional dose-response information, alternative forms of analysis of the dose-response curves, or in vitro exposure of alveolar macrophages to exhaust materials might provide greater information and ability to rank the samples. These recommendations apply both to direct effects of the emission materials, as described in the section on oxidative stress, and to effects on exogenous stimulation of superoxide and peroxide production by exhaust material-exposed cells.

General Toxicity
The sum of all histopathology scores indicates overall toxicity. A second indication of generalized toxicity is the weight of the lungs, relative to body weight, which reflects both tissue changes and edema. Agreement between these two measures of general toxicity was excellent and supported the ranking indicated by the cytotoxicity and inflammation parameters, at least as far as discriminating between the high emitters and normal emitters. However, the discrimination among the samples within the normal- and high-emitter classes was not as good as that achieved with the key parameters of cytotoxicity and inflammation.

Summary
These results demonstrate the ability of this panel of assays, incorporating several categories of biological responses, to discriminate among collected emission samples and permit a preliminary assessment of their relative potency. Because the mechanisms for health effects associated with inhaled emissions are not completely understood, it seems appropriate that a screening strategy should encompass a variety of response types. This panel appeared to provide good discrimination among samples for cytotoxic, inflammatory, parenchymal, and mutagenic potency. These response categories can be adequately evaluated using a reduced panel of parameters, including BALF LDH or protein for cytotoxicity, total BALF leukocytes for inflammation, histopathology for parenchymal changes, and revertant assays in both TA98 and TA100, with and without S9, for mutagenicity. The parameters selected for evaluation of direct or indirect indicators of oxidative damage or macrophage function were less effective for ranking the effects of these samples.

As emphasized above, specific conclusions regarding the contributions of diesel versus gasoline engines to health effects of urban air pollution cannot be drawn from these data alone. However, results from these specific vehicle emission samples suggest several important general conclusions about the toxicity of emissions from light-and medium-duty gasoline and diesel-powered vehicles. First, the relative ranking for mutagenicity was very different from that for the i.t. instillation. In the bacterial strain sensitive to frameshift mutations, the diesel emissions were more potent than gasoline emissions, and, with the exception of gasoline emissions in TA98 without S9, samples collected under cold conditions were more potent than those collected at room temperature. Second, in rat instillation experiments, the toxicity per unit of mass of emissions from normal-emitter diesel and gasoline vehicles was similar in both nature and potency. TNF{alpha} production and several indicators of cytotoxicity and oxidant stress were slightly greater for D compared with G, but no statistically significant differences in any other biological endpoints were observed. Third, in contrast to the mutagenicity results, diesel samples collected under cold ambient temperatures were less potent than emissions from the same vehicles collected at room temperature for cytotoxicity, glutathione, and TNF{alpha} production. Interestingly, the only statistically significant difference for gasoline samples as a function of collection temperature was lung weight, where G was somewhat more potent than G30. Fourth, and perhaps most importantly, these results suggest that high-emitter gasoline and diesel vehicles produce emissions with greater toxicity per unit of PM and SVOC mass than normal-emitters. The indication that vehicles that produce abnormally high levels of emissions may be responsible for an even greater share of the vehicle emissions-related health burden than previously assumed on the basis of mass has important implications for mitigating the public health effects of air pollution. The broad impacts of these conclusions suggest that additional research is warranted to determine whether the findings of this study can be generalized to samples from other vehicles, and to determine the relationship between the physical and chemical properties of these emissions and their biological effects.


    ACKNOWLEDGMENTS
 
This work was supported by the Office of Heavy Vehicle Technologies, U.S. Department of Energy. We gratefully acknowledge the contributions of K. Whitney (SouthWest Research Institute), B. Zielinska, J. Sagebiel, and M. McDaniel (Desert Research Institute) in collection and preparation of the emission samples; J. Berger, D. Olivera, and L. Schutzberger (LRRI) for excellent technical assistance; and James Eberhardt, Director of the Office of Heavy Vehicle Technologies, for contributions to the research strategy and financial support.


    NOTES
 
This article was prepared as an account of work sponsored by an agency of the U. S. Government. Neither the U.S. Government nor any agency thereof, nor any of their employees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe on privately owned rights. Reference herein to any specific commercial product, process, service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the U.S. Government or any agency thereof. The views and opinions of the authors expressed herein do not necessarily state or reflect those of the U.S. Government or any agency thereof. No copyright is asserted in the works of U.S. Government employees.

1 To whom correspondence should be addressed. Fax: (505) 348-4980. E-mail: jseagrav{at}lrri.org. Back

2 Present address: Pharmacia, 800 N. Lindbergh Blvd., St. Louis, MO 63137. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Ames, B. N., McCann, J., and Yamasaki, E. (1975). Methods for detecting carcinogens and mutagens with the Salmonella/mammalian-microsome mutagenicity test. Mutat. Res. 31, 347–364.[ISI][Medline]

Bagley, S. T., Gratz, L. D., Leddy, D. G., and Johnson, J. H. (1993). Characterization of particle- and vapor-phase organic fraction emissions from a heavy-duty diesel engine equipped with a particle trap and regeneration controls. Res. Rep. Health Eff. Inst. 1–121.

Ball, J. C., Straccia, A. M., Young, W. C., and Aust, A. E. (2000). The formation of reactive oxygen species catalyzed by neutral, aqueous extracts of NIST ambient particulate matter and diesel engine particles. J. Air Waste Manag. Assoc. 50, 1897–1903.[ISI][Medline]

Bernson, V. (1983). A comparison of the cellular toxicity of exhausts from cars driven on present and future fuels. Toxicol. Lett. 19, 119–126.[ISI][Medline]

Brown, D. M., Wilson, M. R., MacNee, W., Stone, V., and Donaldson, K. (2001). Size-dependent proinflammatory effects of ultrafine polystyrene particles: A role for surface area and oxidative stress in the enhanced activity of ultrafines. Toxicol. Appl. Pharmacol. 175, 191–199.[ISI][Medline]

Brunekreef, B., Janssen, N. A., de Hartog, J., Harssema, H., Knape, M., and van Vliet, P. (1997). Air pollution from truck traffic and lung function in children living near motorways. Epidemiology 8, 298–303.[ISI][Medline]

Buss, H., Chan, T. P., Sluis, K. B., Domigan, N. M., and Winterbourn, C. C. (1997). Protein carbonyl measurement by a sensitive ELISA method. Free Radic. Biol. Med. 23, 361–366.[ISI][Medline]

Carraro, E., Locatelli, A. L., Ferrero, C., Fea, E., and Gilli, G. (1997). Biological activity of particle exhaust emissions from light-duty diesel engines. J. Environ. Pathol. Toxicol. Oncol. 16, 101–109.[Medline]

Clark, C. R., Henderson, T. R., Royer, R. E., Brooks, A. L., McClellan, R. O., Marshall, W. F., and Naman, T. M. (1982). Mutagenicity of diesel exhaust particle extracts: Influence of fuel composition in two diesel engines. Fundam. Appl. Toxicol. 2, 38–43.[Medline]

Crebelli, R., Fuselli, S., Conti, G., Conti, L., and Carere, A. (1991). Mutagenicity spectra in bacterial strains of airborne and engine exhaust particulate extracts. Mutat. Res. 261, 237–248.[ISI][Medline]

Driscoll, K. E., Costa, D. L., Hatch, G., Henderson, R., Oberdoerster, G., Salem, H., and Schlesinger, R. B. (2000). Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: Uses and limitations. Toxicol. Sci. 55, 24–35.[Abstract/Free Full Text]

Gay, R. J., McComb, R. B., and Bowers, G. N., Jr. (1968). Optimum reaction conditions for human lactate dehydrogenase isoenzymes as they affect total lactate dehydrogenase activity. Clin. Chem. 14, 740–753.[Abstract/Free Full Text]

Goldsmith, C. A., Frevert, C., Imrich, A., Sioutas, C., and Kobzik, L. (1997). Alveolar macrophage interaction with air pollution particulates. Environ. Health Perspect. 105(Suppl. 5), 1191–1195.[ISI][Medline]

Henderson, R. F., Benson, J. M., Hahn, F. F., Hobbs, C. H., Jones, R. K., Mauderly, J. L., McClellan, R. O., and Pickrell, J. A. (1985). New approaches for the evaluation of pulmonary toxicity: Bronchoalveolar lavage fluid analysis. Fundam. Appl. Toxicol. 5, 451–458.[ISI][Medline]

Henderson, R. F., Scott, G. G., and Waide, J. J. (1995). Source of alkaline phosphatase activity in epithelial lining fluid of normal and injured F344 rat lungs. Toxicol. Appl. Pharmacol. 134, 170–174.[ISI][Medline]

Ho, J., and Winer, A. M. (1998). Effects of fuel type, driving cycle, and emission status on in-use vehicle exhaust reactivity. J. Air Waste Manag. Assoc. 48, 592–603.[ISI][Medline]

Hochberg, Y. (1988). A sharper Bonferroni procedure for multiple tests of significance. Biometrika 75, 800–802.[ISI]

Hoek, G., Brunekreef, B., Fischer, P., and van Wijnen, J. (2001). The association between air pollution and heart failure, arrhythmia, embolism, thrombosis, and other cardiovascular causes of death in a time-series study. Epidemiology 12, 355–357.[ISI][Medline]

Janssen, N. A., Schwartz, J., Zanobetti, A., and Suh, H. H. (2002). Air conditioning and source-specific particles as modifiers of the effect of PM(10) on hospital admissions for heart and lung disease. Environ. Health Perspect. 110, 43–49.

Laden, F., Neas, L. M., Dockery, D. W., and Schwartz, J. (2000). Association of fine particulate matter from different sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108, 941–947.[ISI][Medline]

Li, N., Kim, S., Wang, M., Froines, J., Sioutas, C., and Nel, A. (2002). Use of a stratified oxidative stress model to study the biological effects of ambient concentrated and diesel exhaust particulate matter. Inhal. Toxicol. 14, 459–486.[ISI][Medline]

Lloyd, A. C., and Cackette, T. A. (2001). Diesel engines: Environmental impact and control. J. Air Waste Manag. Assoc. 51, 809–847.[ISI][Medline]

Mar, T. F., Norris, G. A., Koenig, J. Q., and Larson, T. V. (2000). Associations between air pollution and mortality in Phoenix, 1995–1997. Environ. Health Perspect. 108, 347–353.[ISI][Medline]

Maron, D. M., and Ames, B. N. (1983). Revised methods for the Salmonella mutagenicity test. Mutat. Res. 113, 173–215.[ISI][Medline]

Martin, L. D., Krunkosky, T. M., Dye, J. A., Fischer, B. M., Jiang, N. F., Rochelle, L. G., Akley, N. J., Dreher, K. L., and Adler, K. B. (1997). The role of reactive oxygen and nitrogen species in the response of airway epithelium to particulates. Environ. Health Perspect. 105(Suppl. 5), 1301–1307.[ISI][Medline]

Mi, H. H., Lee, W. J., Chen, S. J., Lin, T. C., Wu, T. L., and Hu, J. C. (1998). Effect of the gasoline additives on PAH emission. Chemosphere 36, 2031–2041.[ISI][Medline]

Nasrallah, V. N., Jr., Shirley, P. S., Myrvik, Q., and Waite, M. (1983). The use of acetylated cytochrome c in detecting superoxide anion production in rabbit alveolar macrophages. J. Immunol. 131, 2104–2106.[Abstract/Free Full Text]

Nel, A. E., Diaz-Sanchez, D., and Li, N. (2001). The role of particulate pollutants in pulmonary inflammation and asthma: Evidence for the involvement of organic chemicals and oxidative stress. Curr. Opin. Pulm. Med. 7, 20–26.[Medline]

Neter, J., Kutner, M. H., Nactsheim, C. J., and Wasserman, W. (1996). Applied Linear Statistical Methods. Irwin, Chicago.

Pick, E. (1986). Microassays for superoxide and hydrogen peroxide production and nitroblue tetrazolium reduction using an enzyme immunoassay microplate reader. Methods Enzymol. 132, 407–421.[Medline]

Pope, C. A., III. (1999). Mortality and air pollution: Associations persist with continued advances in research methodology. Environ. Health Perspect. 107, 613–614.

Pope, C. A., III, Burnett, R. T., Thun, M. J., Calle, E. E., Krewski, D., Ito, K., and Thurston, G. D. (2002). Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA 287, 1132–1141.[Abstract/Free Full Text]

Prasad, P., Lall, S. B., and Mathur, M. (2000). Effect of intra-tracheal instillation and inhalation of silicon dioxide on some biochemical variables in broncho-alveolar lavage fluid and lung histopathology in rats. Drug Chem. Toxicol. 23, 459–475.[ISI][Medline]

Samet, J. M., Zeger, S. L., Dominici, F., Curriero, F., Coursac, I., Dockery, D. W., Schwartz, J., and Zanobetti, A. (2000). The national morbidity, mortality, and air pollution study: II. Morbidity and mortality from air pollution in the United States. Res. Rep. Health Eff. Inst. 94, 5–70.

Schuetzle, D., and Lewtas, J. (1986). Bioassay-directed chemical analysis in environmental research. Anal. Chem. 58, 1060–1075A.[ISI]

Seagrave, J. C., Berger, J., Zielinska, B., Sagebiel, J., Rogers, C. F., McDonald, J. D., and Mauderly, J. L. (2001). Comparative acute toxicities of particulate matter (PM) and semivolatile organic compound (SVOC) fractions of traffic tunnel air (Abstract). Toxicologist 60, 192.

Seagrave, J. C., Mauderly, J. L., Zielinska, B., Sagebiel, J., Whitney, K., Lawson, D. R., and Gurevich, M. (2000). Comparative toxicity of gasoline and diesel engine emissions. Society of Automotive Engineers Technical Papers 2000–01–2214, 1–4.

Searle, S. R. (1971). Linear Models. J. Wiley and Sons, New York.

Sera, N., Fukuhara, K., Miyata, N., and Tokiwa, H. (1994). Detection of nitro-azabenzo[a]pyrene derivatives in the semivolatile phase originating from airborne particulate matter, diesel, and gasoline vehicles. Mutagenesis 9, 47–52.[Abstract]

Squadrito, G. L., Cueto, R., Dellinger, B., and Pryor, W. A. (2001). Quinoid redox cycling as a mechanism for sustained free radical generation by inhaled airborne particulate matter. Free Radic. Biol. Med. 31, 1132–1138.[ISI][Medline]

Tietze, F. (1969). Enzymic method for quantitative determination of nanogram amounts of total and oxidized glutathione: Applications to mammalian blood and other tissues. Anal. Biochem. 27, 502–522.[ISI][Medline]

Ulfvarson, U., Dahlqvist, M., Sandstrom, T., Bergstrom, B., Ekholm, U., Lagerstrand, L., Figler, B., Nilsen, A., Bjermer, L., Tronnes, T., et al. (1995). Experimental evaluation of the effect of filtration of diesel exhaust by biologic exposure indicators. Am. J. Ind. Med. 27, 91–106.[ISI][Medline]

van Vliet, P., Knape, M., de Hartog, J., Janssen, N., Harssema, H., and Brunekreef, B. (1997). Motor vehicle exhaust and chronic respiratory symptoms in children living near freeways. Environ. Res. 74, 122–132.[ISI][Medline]

Watanabe, N., Kamei, S., Ohkubo, A., Yamanaka, M., Ohsawa, S., Makino, K., and Tokuda, K. (1986). Urinary protein as measured with a pyrogallol red-molybdate complex, manually and in a Hitachi 726 automated analyzer. Clin. Chem. 32, 1551–1554.[Abstract/Free Full Text]

Westerholm, R. N., Almén, J., Li, H., Rannug, J. U., Egebäck, K.-E., and Grägg, K. (2001). Chemical and biological characterization of particulate-, semivolatile-, and gas-phase-associated compounds in diluted heavy-duty diesel exhausts: A comparison of three different semivolatile -phase samplers. Eniviron. Sci. Technol. 25, 332–338.

Yang, H. M., Antonini, J. M., Barger, M. W., Butterworth, L., Roberts, B. R., Ma, J. K., Castranova, V., and Ma, J. Y. (2001). Diesel exhaust particles suppress macrophage function and slow the pulmonary clearance of Listeria monocytogenes in rats. Environ. Health Perspect. 109, 515–521.[ISI][Medline]

Yuan, D., Zhou, W., and Ye, S. H. (1999). Comparison of the mutagenicity of exhaust emissions from motor vehicles using leaded and unleaded gasoline as fuel. Biomed. Environ. Sci. 12, 136–143.[ISI][Medline]