Influence of Physical and Chemical Characteristics of Diesel Fuels

Chem. Res. Toxicol. 1996, 9, 197-207

197

Influence of Physical and Chemical Characteristics of Diesel Fuels and Exhaust Emissions on Biological Effects of Particle Extracts: A Multivariate Statistical Analysis of Ten Diesel Fuels Michael Sjo¨gren,† Hang Li,‡ Carol Banner,§ Joseph Rafter,§ Roger Westerholm,‡ and Ulf Rannug*,† Department of Genetic and Cellular Toxicology, Wallenberg Laboratory, Stockholm University, S-106 91 Stockholm, Sweden, Department of Analytical Chemistry, Arrhenius Laboratory, Stockholm University, S-106 91 Stockholm, Sweden, and Department of Medical Nutrition, Karolinska Institute, Huddinge Hospital F60 Novum, S-141 86 Huddinge, Sweden Received June 7, 1995X

The emission of diesel exhaust particulates is associated with potentially severe biological effects, e.g., cancer. The aim of the present study was to apply multivariate statistical methods to identify factors that affect the biological potency of these exhausts. Ten diesel fuels were analyzed regarding physical and chemical characteristics. Particulate exhaust emissions were sampled after combustion of these fuels on two makes of heavy duty diesel engines. Particle extracts were chemically analyzed and tested for mutagenicity in the Ames test. Also, the potency of the extracts to competitively inhibit the binding of 2,3,7,8-tetrachlorodibenzo-pdioxin (TCDD) to the Ah receptor was assessed. Relationships between fuel characteristics and biological effects of the extracts were studied, using partial least squares regression (PLS). The most influential chemical fuel parameters included the contents of sulfur, certain polycyclic aromatic compounds (PAC), and naphthenes. Density and flash point were positively correlated with genotoxic potency. Cetane number and upper distillation curve points were negatively correlated with both mutagenicity and Ah receptor affinity. Between 61% and 70% of the biological response data could be explained by the measured chemical and physical factors of the fuels. By PLS modeling of extract data versus the biological response data, 66% of the genotoxicity could be explained, by 41% of the chemical variation. The most important variables, associated with both mutagenicity and Ah receptor affinity, included 1-nitropyrene, particle bound nitrate, indeno[1,2,3-cd]pyrene, and emitted mass of particles. S9-requiring mutagenicity was highly correlated with certain PAC, whereas S9-independent mutagenicity was better correlated with nitrates and 1-nitropyrene. The emission of sulfates also showed a correlation both with the emission of particles and with the biological effects. The results indicate that fuels with biologically less hazardous potentials should have high cetane number and contain less PAC and sulfur. The results also indicate that engine factors affect the formation and emission of nitrated PAC.

Introduction When a diesel fuel is combusted, a complex mixture of compounds is emitted with the exhaust (1). These comprise chemical component residues from the fuel and products formed in the combustion. Diesel exhaust pollutants partition to three major parts: a gaseous phase, carbonaceous soot particles, and semivolatile organics that are distributed between the particulate and the gaseous phase. The particles may also contain inorganic salts, e.g., sulfates and nitrates, and traces of metals (2). Some exhaust emission pollutants are in many countries regulated by law (3, 4). Generally, these are the following: carbon monoxide (CO), oxides of nitrogen (NOx),1 hydrocarbons (HC),1 and particulates. The emission of products formed in the combustion process is influenced by the chemical composition of the * Corresponding author; e-mail: [email protected]; fax: +46 8 6124004. † Wallenberg Laboratory, Stockholm University. ‡ Arrhenius Laboratory, Stockholm University. § Karolinska Institute. X Abstract published in Advance ACS Abstracts, December 1, 1995.

0893-228x/96/2709-0197$12.00/0

fuel, engine design, engine load, and presence of an exhaust after-treatment device (5-12). Several studies have demonstrated carcinogenic effects, in rodents, of diesel exhaust particulates (DEP)1 (1322). There are strong indications that the pulmonary carcinogenicity of DEP in rat is caused merely by the presence of particle cores, and not by the organics adhered to them (13, 15-19, 21). It should be noted that this response of DEP has not been observed in other rodent species and that the relevance to man is questionable (18, 19). In rat, the effect is dependent on the particle size (20) and occurs only at lung particle overloads (22) and not at lower concentrations. A formation of pulmonary DNA adducts occurs following the exposure of DEP (23-25). This indicates that the organic phase 1 Abbreviations: TCDD, 2,3,7,8-tetrachlorodibenzo-p-dioxin; PLS, partial least squares regression; PAC, polycyclic aromatic compounds; NOx, nitrogen oxides; HC, hydrocarbons; DEP, diesel exhaust particles; IARC, International Agency for Research on Cancer; PAH, polycyclic aromatic hydrocarbons; PCA, principal components analysis; EHN, ethyl hexyl nitrate; PC, principal component; NIPALS, nonlinear iterative partial least-squares; BaA, benz[a]anthracene; F-SOF, fuelderived soluble organic fraction; 1-NP, 1-nitropyrene; NPAH, nitroPAH.

© 1996 American Chemical Society

198 Chem. Res. Toxicol., Vol. 9, No. 1, 1996

of DEP may have an impact in the initiation of carcinogenesis, whereas the particles per se exert a promotive effect (cf. ref 13). Epidemiological studies on workers occupationally exposed to diesel exhausts indicate associations of lung cancer (26, 27) and bladder cancer (ref 27 and references therein) with the exposure. Based on the observed carcinogenicity in laboratory animals, and the epidemiological indications, the International Agency for Research on Cancer (IARC)1 classifies diesel exhaust emission as being probably carcinogenic to humans (27). IARC also classifies individual diesel emission components as being rodent carcinogens (27, 28). Cancer initiation is assumed to require the reaction of a genotoxic agent with DNA. Based on this assumption, the Ames test (29) has been widely used as a tool for estimating genotoxic potential. By the combination of different bacterial strains and metabolic conditions, mechanisms of mutagenicity of specific substances may be elucidated. Mutational spectra of extracts or fractions of complex mixtures may be used to identify components that contribute to mutagenicity (30). The Ames test has been used to assess the genotoxicity of diesel exhaust emission extracts and of extract fractions (31-40). Much of the genotoxicity of diesel exhaust extracts may be attributed to polycyclic aromatic hydrocarbons (PAH),1 and to their nitrated and oxygenated derivatives (33, 34, 36-40). In mammals, the metabolic activation of premutagens to more reactive species is a key step in a series of events leading to genotoxicity and carcinogenicity. With respect to PAH, this pathway is catalyzed by certain cytochromes, P450 1A1 and P450 1A2 (41). Basal activity of these enzymes is low, but can be greatly induced by certain xenobiotics that bind to the cytosolic aromatic hydrocarbon receptor, the Ah receptor. The formation of ultimate carcinogens may thus be affected by the presence of compounds that have high Ah receptor affinity. Many carcinogens exhibit such affinity. Studies have shown that the chemical substructures that are involved in the carcinogenicity of some PAH are identical to the structural fragments required for Ah receptor affinity (42), but different from the structural fragments involved in mutagenicity of PAH in Salmonella typhimurium (43). In this ongoing study, the goals are as follows: (1) to identify the factors in diesel fuel and its exhaust emission that contribute most to the genotoxicity; and (2) to develop models for optimizing fuels with respect to low genotoxicity of the exhausts. The combination of the Ames test and Ah the receptor affinity assay of diesel exhaust extract provides means for characterization and risk identification. With chemical and physical analysis of the tested fuels, biological assays, and chemical analysis of associated exhaust emissions, this approach generates large amounts of data, making the interpretation of results very complex. It is therefore necessary to apply powerful statistical tools for data analysis. In an earlier study (44) eight diesel fuels and associated particulate exhaust emissions were analyzed using multivariate statistical methods, principal components analysis (PCA),1 and partial least squares regression (PLS).1 It was concluded that the most influential variables, with respect to emissions, were fuel density, final boiling point, total aromatics, and PAH. The same variables were shown to have an impact on the biological responses, i.e., mutagenicity in Salmonella and Ah receptor binding affinity. The present Swedish legislation (4)

Sjo¨ gren et al.

for the environmental classification of diesel fuels was in part based on these results. With the same set of fuels, but with more extensive chemical analysis data, it was later shown that mutagenicity, in the absence of metabolic activation (i.e., -S9), correlated well with the contents of 1-nitropyrene (45). With S9, a higher correlation was observed with PAH, specifically with fluoranthene and anthracene. Westerholm and Li (11), using linear regression, PCA, and PLS, found a good correlation between fuel PAH and emission PAH. These results indicated that a reduction of fuel PAH from 1 g/L to 4 mg/L would decrease PAH in the emission by 80%. New data became available through the inclusion of three new fuels. Initial analysis showed that one Swedish commercial summer fuel was an outlier. This fuel, D6, was much higher with respect to, e.g., PAH than the other fuels. It was therefore excluded from further studies. The ten remaining fuels were described, using PCA, in terms of physical and chemical fuel parameters (46). Physically (26 variables), these ten fuels could be described in a few general terms, e.g., viscosity, density and cetane number. The most descriptive chemical factors of the fuels (45 variables) included PAC, represented by, e.g., 1-methylphenanthrene. Chemical data of the gaseous and particulate exhaust emissions associated with these fuels were submitted to PCA (12). Two engine makes were used for generating particulates from all the fuels. Exhaust after-treatment devices (one per engine make) were in addition used for one of the fuels. A total of 63 chemical descriptors were assessed. PCA results demonstrated correlations between most regulated pollutants, particles, and most particulate associated PAC. The two engines differed with respect to exhaust emission profiles. The emission profile for certain individual combustion products was also influenced by interactions between fuel and engine design (make). The present study is based on the same selection of diesel fuels and exhaust emission samples previously analyzed (12, 46), but includes also the study of biological effects of the diesel exhaust particulates. Models were derived for relating biological effects to fuel physical and chemical data, and to chemical data of the particulate exhaust emissions, respectively. The biological assays were the Ames test and the Ah receptor affinity test. Statistical analyses were performed using PCA and PLS.

Experimental Procedures Notations. Names and notations of physical, chemical, and biological variables are consistent with previous papers in this series (12, 46) and are outlined in Tables 1 and 2. Fuels. Ten different diesel fuels (denoted D1, D2, D4, D5, D7-D10, D14, and D15) from four suppliers were included in the analysis. Fuels D1, D2, D4, D10, D14, and D15 were from the same supplier. D1, D2, and D4 were derived from the same base fuel. Fuel D5 was a blend of kerosenes. D8 was derived from D7 by the addition of 2000 ppm ethyl hexyl nitrate (EHN)1 as ignition improver. Fuel D9 was a blend of highly hydrated fuel composed from cracked gas oils with low sulfur contents. The fuels were characterized with respect to physical and chemical properties as summarized in Table 2. A more extensive description and the correlations of physical and chemical characteristics for these diesel fuels have previously been presented (46). Engines. The fuels were combusted in two makes of heavy duty diesel engine (12). The engines were operated on a chassis dynamometer, in accordance with a well-defined transient “bus” cycle (Stochasticher Fahrzyklus fu¨r Stadtlinien Omnibusse).

Multivariate Analysis of Diesel Particulates

Chem. Res. Toxicol., Vol. 9, No. 1, 1996 199

Table 1. Notations of Variables Analyzed Fuel Physical Variables fp1 fp2 fp3 fp4 fp5 fp6 fp7 fp8 fp9 fp10 fp11 fp12 fp13 fp14 fp15

cetane no. cetane index ASTMa cetane index IPa density, g/L at 15 °C viscosity, kV at 40 °C initial boiling point, °Ca BP 5%, °Ca BP 10%, °Ca BP 20%, °Ca BP 30%, °Ca BP 40%, °Ca BP 50%, °Ca BP 60%, °Ca BP 70%, °Ca BP 80%, °Ca

fp16 fp17 fp18 fp19 fp20 fp21 fp22 fp23 fp24 fp25 fp26 fdp1 fdp2 fdp3 fdp4

fc1 fc2 fc3 fc4 fc5 fc6 fc7 fc8 fc9 fc10 fc11 fc12 fc13 fc14 fc15 fc16 fc17 fc18 fc19 fc20 fc21 fc22 fc23 fc24 fc25 fc26 fc27 fc28 fc29 fc30 fc31 fc32 fc33 fc34 fc35 fc36 fc37 fc38 fc39 fc40 fc41 fc42 fc43 fc44 fc45

Chemical Variables Fuel phenanthrene* ec1 anthracene* ec2 4-methyldibenzothiophene ec3 3-methyldibenzothiophene ec4 3-methylphenanthrene ec5 2-methylanthracene ec6 4 and 9-methylphenanthrene ec7 1-methylphenanthrene ec8 fluoranthene* ec9 pyrene* ec10 1-Me-7-isopropylphenanthrene ec11 benzo[a]fluorene ec12 2-methylpyrene ec13 1-methylpyrene ec14 benzo[ghi]fluoranthene* ec15 cyclopenta[cd]pyrene* ec16 benz[a]anthracene* ec17 chrysene/triphenylene* ec18 benzo[b and k]fluoranthene* ec19 benzo[e]pyrene* ec20 benzo[a]pyrene* ec21 perylene ec22 indeno[1,2,3-cd]fluoranthene ec23 indeno[1,2,3-cd]pyrene* ec24 picene ec25 benzo[ghi]perylene* ec26 coronene* ec27 sum of PAC (14)* (mg/L)a ec28 sum of PAC (27) (mg/L)a ec29 aromatics, vol %, totala ec30 mono-a ec31 a diec32 tri-a ec33 nitrogen, mg/L ec34 sulfur, ppmw ec35 ethyl hexyl nitrate ec36 water, ppmw ec37 aromatics, vol % FIAa ec38 olefins, vol % FIA ec39 a saturates, vol % ec40 naphthenes, W% paraffins, W% aromatic carbon %, NMRa benzene (mg/L) toluene (mg/L)

88+ 0-

TA98 - S9 TA98 + S9 TA100 - S9

Biological Variables 0+ Ah

BP 90%, °Ca BP 95%, °Ca final boiling point, Ca boiling range, °C total distilled, % boiling residue, % CFPP, °C cloud point, °C flash point, °C energy, MJ/kga energy, MJ/L BP, PCA factor 1 BP, PCA factor 2 BP, PCA factor 3 BP, PCA factor 4 Particle Exhaust Emission phenanthrene* anthracene* 4-methyldibenzothiophene 3-methyldibenzothiophene 3-methylphenanthrene 2-methylanthracene 4 and 9-methylphenanthrene 1-methylphenanthrene fluoranthene* pyrene* 1-Me-7-isopropylphenanthrene benzo[a]fluorene 2-methylpyrene 1-methylpyrene benzo[ghi]fluoranthene* cyclopenta[cd]pyrene* benz[a]anthracene* chrysene/triphenylene* benzo[b and k]fluoranthene* benzo[e]pyrene* benzo[a]pyrene* perylene indeno[1,2,3-cd]fluoranthene indeno[1,2,3-cd]pyrene* picene benzo[ghi]perylene* coronene* sum of PAC (14)* (µg/km)a sum of PAC (27) (µg/km)a 1-nitropyrene (µg/km) particle mass organic soluble fraction soluble organic fraction (SOF)a sulfatea SOF, fuel derived SOF, oil derived particle carbon content soluble sulfate sulfate-bound watera nitrate in particle

TA100 + S9 Ah receptor affinity

a Variables excluded due to causes discussed in the Experimental Procedures and references therein. *Denotes the individual PAH included in the sum of 14 PAC (fc28 and ec28).

This driving cycle simulates public transportation within a city (37). It should be noted that the engine models from both engine makes were changed after fuels D1-D9 were run (cf. ref 12). An initial PC analysis was performed on the physical characteristics of a wide range of heavy duty diesel engines. The results indicated that the differences within each engine make were small (data not shown). Hence, for each engine make, the two models were treated as one.

Particulate Samples. Particles from diluted diesel exhaust emissions were collected using Teflon-coated glass fiber filters (Pallflex T60A20, Pallflex Inc., Putnam, CT), in accordance with the specifications of the U.S. Federal Register (3). For fuel D10, additional particulate samples were collected, using exhaust after-treatment devices, i.e., particulate traps. These samples are henceforth denoted D10f. Chemical analyses of the particulates were performed as described elsewhere (37). Data on


Sjo¨ gren et al.

Table 2. List of Selected Fuel Variables, Notations, and Measured and Calculated Values per Fuel variable name cetane no. density at 15 °C viscosity at 40 °C initial BP (IBP) 50% BP (BP50%) final BP (FBP) dist PC1 (≈BP40%) dist PC2 (≈BP95%) dist PC3 (≈FBP%) dist PC4 (≈IBP%) dist. residue CF point cloud point flash point specific energy phenanthrene 3-methylphenanthrene 2-methylanthracene fluoranthene benz[a]anthracene picene sum of 14 PAH sum of 27 PAH nitrogen sulfur ethyl hexyl nitrate water aromatics olefins naphthenes paraffins benzene toluene

fuel unit

g L-1 mm2 s-1 °C °C °C

% °C °C °C MJ L-1 mg L-1 mg L-1 mg L-1 mg L-1 mg L-1 mg L-1 mg L-1 mg L-1 mg L-1 ppmw wt % ppmw vol % vol % wt % wt % mg L-1 mg L-1

abbrev fp1 fp4 fp5 fp6 fp12 fp18 fdp1 fdp2 fdp3 fdp4 fp21 fp22 fp23 fp24 fp26 fc1 fc5 fc6 fc9 fc17 fc25 fc28 fc29 fc34 fc35 fc36 fc37 fc38 fc39 fc41 fc42 fc44 fc45

D1

D2

D4

D5

D7

D8

D9

D10

D14

D15

52.8 50 47.2 47 44.7 55.7 52.8 na 50 56 811.7 821.3 832 831.3 808.3 808.7 813.2 817.5 814.3 817.2 2.11 2.11 2.09 2.26 1.41 1.44 1.96 1.7 2.02 2.57 220 223 221 190 180 176 175 133 164 172 236 239 241 248 206 204 231 219 236 257 261 260 261 323 300 299 301 323 307 337 1.59 2.00 2.29 2.34 -3.78 -4.18 -0.02 -2.79 -0.21 2.77 -2.65 -2.70 -2.63 1.29 -1.08 -1.15 0.71 2.38 1.91 3.91 -0.19 -0.11 -0.12 0.74 0.36 0.23 -0.13 -0.30 -0.63 0.15 0.02 -0.03 -0.07 -0.15 0.17 0.18 -0.24 -0.46 0.33 0.26 1.3 1.8 1.4 1.2 1.7 1.6 1.5 1.5 1 1.1 -40 -39 -39 -32 -40 -40 -40 -40 -39 -20 -34 -35 -34 -24 -40 -40 -40 -25 -37 -17 87 87 92 75 64 64 75 48 59 60 35.1 35.4 35.7 35.7 35 35 35.1 35.2 35.2 35.2 0.17 0.24 0.27 78.10 29.53 29.47 83.71 bd bd bd 0.18 0.78 0.11 52.58 29.07 24.09 86.66 bd bd bd 0.18 0.93 0.13 61.49 35.28 27.89 82.83 bd bd bd 0.14 0.06 0.06 7.80 4.04 3.71 9.82 bd bd bd 0.03 0.02 0.03 0.51 0.50 0.35 0.09 0.30 0.50 0.30 bd 0.01 0.04 0.01 bd 0.02 0.12 bd bd bd 0.73 0.94 0.73 101.8 47.64 41.05 106.5 0.8 1.2 0.6 1.57 4.12 1.39 340.7 231.6 183.5 313 0.8 1.2 0.6 0.21 0.34 3.9 29.2 14.3 207 11 0.13 na na 100 100 2900 200 200 100 100 7 1 1 0 0 0 0 0 0.2 0 0 0 0 15 20 70 50 60 73 58 34 12 17 2.7 14.7 22.5 26 19.8 19.4 16 20.7 7.7 3.7 1.4 2 2.2 1.6 0.9 0.2 0.7 1.8 na na 53.6 38.8 29.6 27.2 32.9 32.9 25.5 32.5 50.7 48.8 45.6 45.5 43.8 47.6 48.5 47.7 58.1 47.1 44.9 47.1 85.5 4.9 10.7 30.7 7.8 7.9 5.9 na na na 215.6 7.4 31.9 1225 59.5 72 299.9 na na na

a The complete set of fuel data have been detailed elsewhere (39). Distillation curve data (fdp1-fdp4) were transformed as described in the Experimental Procedures. bd, below detection level; na, not analyzed.

Table 3. Summary of Biological Responses, Mean Values engine make 1: TA98 - S9‘ TA98 + S9a TA100 -S9a TA100 + S9a Ah receptor affinityb engine make 2: TA98 - S9‘ TA98 + S9a TA100 - S9a TA100 + S9a Ah receptor affinityb

D1

D2

D4

D5

D7

D8

D9

D10f

D10

D14

D15

7.3 8.3 25.7 26.6 0.6

15.7 12.2 52.8 45.8 0.8

28.6 19.9 93.0 66.9 1.3

25.9 26.3 94.0 89.3 0.9

17.6 14.4 63.7 63.7 1.0

6.6 6.5 12.7 31.7 1.0

21.9 19.4 74.8 78.6 1.5

6.0 0.3 6.7 8.6 0.9

na na na na 0.9

7.9 2.5 32.9 1.0 0.7

na na 32.7 1.3 na

24.8 12.0 203.1 82.1 1.1

119.7 28.5 507.3 178.6 1.1

472.3 72.8 1177.7 353.9 na

144.7 48.3 535.9 264.5 1.2

50.6 19.5 257.2 110.5 1.2

52.1 19.3 263.7 115.4 1.1

97.7 58.5 255.1 169.1 1.4

5.2 5.8 30.0 5.5 0.6

29.7 18.4 108.1 57.5 1.1

7.1 10.7 38.7 14.9 0.9

16.2 26.9 52.4 55.5 na

a Numbers show revertants per meter driving distance or bEC , the driving distance sufficient to inhibit the specific binding of 50 radiolabeled TCDD to the Ah receptor by 50%.

the exhaust emissions from these fuels and engines have been detailed elsewhere (12). Chemical and biological characterizations were, in general, performed on triplicate samples per combination of fuel and engine make. Ames Mutagenicity Test. The standard Ames test procedure was used (29), with the modification of histidine and biotin added to the minimal medium instead of to the top agar. The strains used were TA98 and TA100, with and without metabolic activation (S9). Particulate samples were analyzed in triplicate per dose, at three dose levels. The response, expressed as number of revertants per meter driving distance, was calculated from the linear portion of the dose response curves. Ah Receptor Affinity Test. An in vitro ligand binding competition assay (47, 48) was used for the assessment of the Ah receptor affinity, expressed as EC50. This value corresponds to the sample concentration in meter driving distance, that inhibits by 50% the specific binding of radiolabelled TCDD to the Ah receptor. In the statistical analysis, inverted EC50 values were used to simplify the interpretation of the data. This value correlates to both affinity and potential toxicity.

Mean values, per fuel and engine make, of the mutagenicity in Ames test and of the Ah receptor affinity are shown in Table 3. Statistical Pretreatment. The original data matrix consisted of 147 variables and 72 objects. Variables that were calculated from other already included variables, or by definition correlate with other variables were excluded (cf. Table 1). Both variables and objects with high proportions of missing data were manually excluded, as were variables where there was no variation between the objects. Since there was a very good correlation between adjacent distillation curve points, these variables were pretreated using PCA (49-52; discussed below). The 13 original distillation variables were reduced to four orthogonal, i.e., mutually noncorrelated, principal component score vectors. These were added as variables to the data matrix and named fdp1-fdp4 (fdp: fuel distillation parameter). Approximately, fdp1 corresponds to the 40% boiling point (BP 40%), describing 55% of the distillation curve data. Similarly, fdp2 (43%) corresponds to BP 95%, fdp3 (1.1%) to final boiling point (FBP), and fdp4 (0.1%) to initial boiling point (IBP). Together,



Table 4. Summary of PLS Analysis, All Models explained variance (%) sum of squares (%)

TA98

TA100

modela

dimension

X matrix

Y matrix

-S9

+S9

-S9

+S9

Ah receptor affinity

1

1 2 total

30 24 54

24 11 35

9 17 26

27 9 36

15 15 30

35 8 43

25 -1 24

2

1 2 3 total

38 17 14 69

43 21 6 70

30 31 -1 60

50 23 0 72

38 32 0 70

72 12 0 85

14 3 29 46

3

1 2 total

21 32 53

49 12 61

38 22 60

48 4 52

54 21 75

69 6 75

25 -2 23

4

1 2 total

26 15 41

49 17 66

39 35 73

61 6 67

40 30 69

59 11 70

42 0 42

a Model 1: Fuel data, both engine makes. Model 2: Fuel data, engine make 1. Model 3: Fuel data, engine make 2. Model 4: Chemical particle data. All models are related to biological effects of diesel particle extracts. these four vectors accounted for >99% of the distillation curve 1 and 2. When PLS of an X matrix of fuel physical and data. chemical parameters was fitted to a Y matrix of particuMultivariate Statistical Analysis. The data matrix was late emission data, 70% (engine make 1) and 57% (make statistically analyzed using Principal Components Analysis (PCA) (47-52) and partial least squares analysis (PLS) (47, 522) of the emission data could be explained by variations 58). The fundamentals for these methods are the same; i.e., in fuel characteristics (not shown). The residual variboth methods use the stepwise reduction or projection of matrix ances were due to variations with respect to engine data onto orthogonal vectors. factors, sampling, and extraction. It can be assumed that A sample may be described by a number of variables. the genotoxicity of the particulate exhaust emissions Geometrically, this corresponds to a point in a multidimensional similarly would be affected by these factors, and not only space where the variables define the axes. If many objects share by fuel composition. Some fuel characteristics may, this space, correlations between variables may be observed. A 1 however, be either causal or indicative of other causal principal component (PC) or a PLS component is a straight relations (as discussed below). least regression line through the sample points in the multivariable space. The contributions of individual variables to the In order to minimize the confounding effect of engine component are expressed as “loadings”. Sample points are make, separate PLS models were developed for samples projected along the component, and these positions make up a from each motor make. The consensus results of these vector of sample “scores”. For each component, a loadings vector models and a model of both engine makes combined will (p) and a scores vector (t) is obtained. By multiplication of the be commented on below. two vectors, the original data matrix (X) may be approximated. PLS of Physical and Chemical Fuel Data versus The difference between X and p*t yields a matrix of residuals Biological Effects. (A) Engine Makes 1 and 2, (E) from which the next component is extracted. The iterative Combined. When samples from both engine makes component extraction proceeds until no more systematic variawere combined into one dataset, two statistically signifition remains in the residual matrix. The raw data have then been reduced to a loadings matrix (P) and a scores matrix (T) cant PLS components were extracted. Together, they containing systematic variable and object information of the explained 35% of the biological response data, using 54% data set, respectively. Raw data may be approximated from the of the chemical and physical information of the fuels components: X ) 1*X h + TP′ + E. All components are orthogo(Table 4). The cumulative fractions of explained variance nal since they explain a maximum of the remaining variance. for the biological data were low, ranging from 24% to Object relationships and variable correlation patterns can be 43%. The biological effects of samples from D1 to D9 identified by plotting the vectors. Objects with high (positive were generally greater in engine make 2. The Ames test or negative) scores are explained by variables that have high data were on average 1 order of magnitude higher in loadings in the same components. samples from engine make 2. The differences with In PLS, components are extracted from one X matrix (derespect to Ah receptor affinity were less pronounced. The scriptor matrix) and one Y matrix (response matrix). Components are extracted so as to obtain a maximal predictivity of difference in genotoxicity between engine makes with the response data, but also to approximate the X and Y matrices. samples from D10 to D15 was also much lower (Table Since components are extracted for both matrices, scores are 3). Overall, the variables that contributed most to the projected to separate vectors, t (X scores) and u (Y scores). By biological effect were sulfur (fc35), energy in MJ/L (fp26), plotting, for each PLS dimension, t versus u, relationships flash point (fp24), and density (fp4). between X variables and Y variables can be visualized. (B) Engine Make 1. With three significant PLS In the present study, PCA was applied on autoscaled data, components more than 70% of the response data were which means that each variable was mean centered and its modeled by 69% of the fuel physical and chemical variance set to 1. PC’s were extracted using the NIPALS1 (nonlinear iterative partial least-squares) algorithm, with crossinformation (Table 4). The explained variance of muvalidation. The level of statistical significance was 0.05. All tagenicity was in the range 60-85%, for the four comPCA and PLS calculations were performed using version 5.1 of binations of strains (TA98/TA100) and metabolic condithe SIMCA-S for Windows software, developed by UMETRI AB, tions ((S9). All mutagenicity responses were highly Umeå, Sweden.

Results and Discussion PLS of Physical and Chemical Fuel Data versus Exhaust Emission Chemical Data. Engine Makes

intercorrelated. The cumulative fraction of explained variance for the Ah receptor affinity was 46%. The results indicate that, for screening of genotoxicity of fuel particulate exhaust emissions, it would be sufficient to use either strain, but with both metabolic conditions.


Sjo¨ gren et al.

Figure 1. PLS loadings (a) and scores (b, c) of the physical and chemical data of the fuels, engine make 1. Markers in bold text denote the biological response data. Notations are as given in Table 1. The variable’s influence on the projection (VIP), panel d, shows the statistical validity of individual parameters in predicting the biological response. Values above 1 are statistically significant.

The variables that contributed most to the biological response data were energy in MJ/L (fp26), sulfur (fc35), water (fc37), flash point (fp24), and density (fp4, Figure 1a). It is well-known that fuel sulfur increases the mass of emitted particles (8). This is due to the adherence of sulfates to particles and the abrasive effect by sulfate on the filter. Statistically, the content of water in the fuel (fc37) was implicated to have an effect on genotoxicity. This effect is more likely to be associated with a waterrelated factor, than to water itself, which is present in the fuel in trace amounts (12-73 ppm). The specific energy, in MJ/L, is correlated with fuel density (cf. ref 46). The content of aromatics in the fuel, which affects fuel density, is correlated with the carbon/ hydrogen ratio (9, 46), where a higher ratio yields less energy per mass unit. Hence, more aromatic fuels normally have higher density and lower energy per kilogram, forcing the engine to consume more fuel, in mass, in order to obtain the same engine effect. Both the contents of aromatics and fuel consumption, in mass, thus increase the output of particles. The cetane number (fp1) and naphthenes (fc41) are negatively correlated with mutagenicity. It has been proposed that increasing the cetane number is a cost-effective method for reducing the emission of regulated pollutants, including the emission of particles (9). The parameters that best predict the genotoxic potency of exhaust emissions have all, except naphthene, been shown to affect the emission of particles. With seven of the ten fuels in this study, the correlation of biological response with particle emission has previously been observed (44, 45). Fuel D8, which was derived from D7 by the addition of EHN, was 11 cetane number units higher than D7 and yielded samples with significantly lower amounts of

particles. Accordingly, mutagenicity was much reduced in the D8 samples. The addition of nitrate based ignition improvers may have an impact on the formation of nitrated PAH emitted in the exhausts. Some of these products are extremely mutagenic in Ames test (59-65). In this study, however, neither chemical nor genotoxicity data from D8 versus D7 particulate samples indicated any increase of nitrated PAH from the combustion of D8. Statistically, 15 of the 27 PAC exhibited significant modeling powers (Figure 1d). The major impacts were observed for the following: picene (fc25), phenanthrene (fc1), 2-methylanthracene (fc6), 3-methylphenanthrene (fc5), and fluoranthene (fc9). The contributions in fuels D1-D9 of phenanthrene, 2-methylanthracene, 3-methylphenanthrene, and fluoranthene added up to approximately 60% by mass of the 27 analyzed PAC. These compounds were below the detection level (approximately 0.01 mg/L) in D10, D14, and D15. Picene in the fuel correlated well with the mutagenicity data, particularly in the presence of S9. It was, however, below detection level in fuels D1, D7, and D10-D15. It has been shown that picene is both mutagenic and carcinogenic (66). Still, further analysis of the particulates (cf. below) indicated that there was no causal relationship between exhaust emission of particle bound picene and genotoxicity. A previous study (11) also showed that there was no correlation between the fuel picene and the particulate picene. Although no causality can be established for picene and the genotoxic effects, this compound may still be an indicator for other genotoxic factors of the fuel. The particulate samples that had the highest mutagenicity and Ah receptor affinity were derived from fuels D4, D5, D7, and D9 (Figure 1b,c). The latter three fuels



Figure 2. PLS loadings (a) and scores (b, c) of the physical and chemical data of the fuels, engine make 2. Markers in bold text denote the biological response data. Notations are as given in Table 1. The variable’s influence on the projection (VIP), panel d, shows the statistical validity of individual parameters in predicting the biological response. Values above 1 are statistically significant.

contained high amounts of PAC, whereas D4 had a sulfur content that was one order of magnitude higher than any other fuel included in this study. Samples from D14 and D15 were low with respect to fuel PAH, and this was reflected in low genotoxic potency. With this engine make, D10 samples, without a particle trap, were not assayed in Ames test. The only difference seen between the D10 and the D10f samples was regarding the Ah receptor affinity. The reducing effect of the particulate trap regarding Ah receptor affinity is reflected in Figure 1b,c. (C) Engine Make 2. In the PLS model from the second engine make, only two components were extracted. With 53% of the chemical and physical variation, 61% of the biological response variation could be explained (Table 4). The intercorrelations between mutagenicity responses were also observed in this engine make (Figure 2), with cumulative explained variance fractions ranging from 52% to 75%. For the Ah receptor affinity, 23% of the variance could be explained by the fuel data. The best descriptors for biological activity were fuel sulfur (fc35), density (fp4), specific energy in MJ/l (fp26), and flash point (fp24). This model also indicated high impacts of the 40% (fdp1) and 95% (fdp2) distillation points. Although particulate samples from D5 contributed most to the first chemical data vector (t#1, Figure 2b), the D4 samples exhibited the highest biological scores. Both D7 and D9 samples were rated lower in this model; i.e., they had lower activity than D4 and D5 samples (Table 3). Samples from D10, D14, D15, and D1 showed low genotoxic potency. With this engine make there was no significant difference in the amounts of particulate emissions, nor in the genotoxicity, of samples derived from fuels D7 and D8.

The particle trap fitted to engine make 2, reduced the genotoxicity of D10 samples by 70-90%. The Ah receptor affinity was reduced by 50% (Table 3). Using a particle trap in combination with other low sulfur fuels can also be assumed to give reductions in particle emission. Three PAH were statistically significant in this PLS model. Benz[a]anthracene (BaA,1 fc17) is both mutagenic and carcinogenic. In this study, it was, however, negatively correlated with the genotoxicity data. Preliminary studies have shown that BaA in the fuel is negatively correlated with the emitted amounts of fuel derived soluble organic fraction (F-SOF,1 ec35) and the emitted mass of particles (ec31) (not shown). The role of fuel BaA for the genotoxicity of diesel particulate extracts and its potential causality to particle emission remain to be clarified. (D) Consensus of Models from Both Engine Makes. There were significant similarities between the two PLS models for fuel data versus biological effects discussed above (Figure 3a). As seen in Figure 3b, few factors were statistically significant when both engines were taken into account. The biological effects of the particulate exhaust emission samples were mainly due to variations in fuel sulfur (fc35), flash point (fp24), and density (fp4). The latter was closely correlated with energy in MJ/L (fp26). The impact of these factors on the particulate emissions has also been demonstrated in previous studies by us (44, 12) and by others (6, 8, 10). The 40% distillation point (fdp1) and picene (fc25) also correlated with genotoxicity, although no effects on the particulates are known for these factors. The factors that were significantly and negatively correlated with genotoxic potency, in both engine models, were the cetane number (fp1) and the contents of naphthenes (fc41). These factors


Sjo¨ gren et al.

Figure 3. Loadings (a) of the first PLS components from both engine makes PLS models. If both models were exactly equivalent, loadings would form a diagonal line. Deviating factors from this line denote the differences in the two PLS models. The consensus variable’s influence on the projection (VIP) (b) shows what factors are statistically significant in both models individually. Factors with VIP equal to or above 1 are statistically significant.

were mutually orthogonal (r ) 0.14, n ) 9, D10 omitted) in the fuel matrix, so the reducing effect of naphthenes on the emission and/or genotoxicity of particulates cannot be excluded. Although fuels D5, D7, D8, and D9 were most abundant regarding PAC in general, the most pronounced mutagenic effects were seen in the exhaust emission samples from fuel D4. PLS of the Chemical Composition of Particulate Extracts versus Biological Effects. Both Engine Makes. In this model, the variances inferred by engine design, running, and sampling could be accounted for, resulting in variations in the chemical composition of the particles. Therefore, a direct relation between the chemical characteristics of the particles, and the biological effects of the particulate extracts, was obtained. Hence, exhaust emission data and genotoxicity data from both engine makes were combined into one set. In this analysis, two statistically significant PLS components were extracted. Using 41% of the chemical data of the particulates, 66% of the biological variance could be explained (Table 4). The total variance explained was 41% for the Ah receptor assay data, and 6674% for the Ames test data. Close correlations were observed (Figure 4a) for the mutagenicity responses. The Ah receptor affinity was also correlated with the S9dependent mutagenicity. The first PLS component mainly explained the Ah receptor affinity and the responses in the Ames tests with metabolic activation, whereas the second PLS component explained the S9-independent mutagenicity. Results indicate that measuring genotoxicity with either strain and with and without metabolic activation would suffice to characterize these complex mixtures. The chemical components that had the largest impact (Figure 4a) on the first PLS component were 1-nitropy-

rene (1-NP,1 ec30), indeno[1,2,3-cd]pyrene (ec24), particle carbon content (ec37), and fuel-derived soluble organic fraction (F-SOF,1 ec35). Mutagenicity in the presence of S9 was more closely correlated to the contents of PAC and to particulate emission. Without metabolic activation, however, responses were better correlated with 1-NP and to nitrate in particles (ec40). Among the most mutagenic compounds tested in the Ames test are certain nitrated PAH (NPAH;1 52-54). Many of these substances are deactivated in the presence of S9. Except for some nitrated anthracenes, the NPAH are generally more mutagenic in TA98 than in TA100 (59). According to estimates, approximately one-tenth of the emitted PAC comprise NPAH. These 10% accounts for 30-50% of the overall mutagenic activity of particulate extracts (62). The formation of NPAH requires presence the presence of both PAH molecules and of nitrogen oxides and is positively affected by acidic conditions (62). In the present study, a close correlation between activity in both strains, without S9, to particle bound nitrate, was observed (Figure 4a). 1-NP (ec30) and particulate nitrate (ec40) also correlated with the mass of emitted particles (ec31), particle carbon content (ec37), and the particle bound soluble sulfate (ec38). Although only one NPAH, 1-NP, was analyzed in the particulate extracts, the results strongly indicate that a major portion of the mutagenicity was contributed to by other NPAH that have not been analyzed. The samples that exhibited the highest genotoxic potency were high regarding PAC (engine make 2; D9, D5, D2), 1-NP (engine make 2; D4, D9), and sulfate and nitrate in the particles (both engine makes; D4). Specifically, samples derived from fuel D4 combusted on engine make 2 exhibited extreme levels of 1-NP.



Figure 4. PLS loadings (a) and scores (b, c) of the chemical data of the particulate exhaust emission for both engine makes. Markers in bold text denote the biological response data. Notations are as given in Table 1. The variable’s influence on the projection (VIP), panel d, shows the statistical validity of individual parameters in predicting the biological response. Values above 1 are statistically significant.

Calculations based on published mutagenicity data (59-61) and the emissions of 1-NP in these samples indicated that less than 30% of the mutagenicity in TA98S9 was due to 1-NP. It should also be noted that the mutagenicity in TA100-S9 in all samples was higher than that in TA98-S9. This is in contrast to what would be expected if all mutagenicity, without S9, was caused by NPAH. The correlations of 1-NP and particulate nitrate with the mutagenicity indicate a causality of NPAH to mutagenicity. Still, much of the direct acting mutagenicity is possibly caused by chemical components that have not been analyzed. The correlation between mutagenicity, without S9, and nitro-PAH indicates that this unaccounted mutagenicity is contributed to by factors that correlate with NPAH and possibly are derived from the NPAH. The 1-NP levels were very high in D4 samples from engine make 2. It seems plausible that the increase of mutagenicity in these samples is partly contributed to by the acidic conditions obtained when a fuel with high sulfur content is combusted in this engine. This acidity in turn may lead to a higher proportion of NPAH. Indeno[1,2,3-cd]pyrene, considered by the IARC to be carcinogenic and mutagenic (27), has a high positive loading in the PLS model. This compound contains a substructural fragment that in QSAR models has been shown to be associated with high Ah receptor affinity (42), mutagenicity (43), and carcinogenicity (43). The experimentally determined affinity of indeno[1,2,3-cd]pyrene to the receptor is also very high, with EC50 values ranging from 10 nM (Banner)2 to 22 nM (67). The biological activity of indeno[1,2,3-cd]pyrene, together with its high loadings in the PLS model, suggest that indeno2

Carol Banner, personal communication, 1995.

[1,2,3-cd]pyrene has a substantial impact on the genotoxicity and Ah receptor binding of the DEP extracts. In the present analysis, the content of indeno[1,2,3-cd]pyrene is better correlated with the Ames test data than with the Ah receptor data. The two first PLS components demonstrated a difference in the endpoints studied with the two assays used, i.e., Ames test and Ah receptor affinity assay. In the Ames test, compounds in a complex mixture may compete for enzymatically catalytic sites. The metabolism of, e.g., PAH, in the Ames assay is complex. This may lead to antimutagenic effects, inferred by substances that have high enzyme affinity but have metabolites of low genotoxic potency. In the Ah receptor assay, all components of the mixture may compete for a binding site on the receptor. In the ideal case, this test system does not include any metabolic events that affect the composition of the complex mixture. The first PLS component reflects the Ah receptor affinity and the mutagenicity of S9 requiring mutagens, whereas the second PLS component reflects an S9-independent component of mutagenicity in Ames test.

Conclusions The fuel parameters that contributed most to the genotoxic potency of particulate extracts included physical factors that affect the combustion process, and thus the amount of emitted particles. These factors were energy per liter, cetane number, density, and flash point. The content of sulfur in the fuel affects the amounts of emitted particles owing to post-combustion processes and, therefore, also contributes to genotoxicity. The genotoxic potency of the exhaust emission particulates was also shown to be affected by the contents of naphthenes and


picene in the fuel. Comparisons of PLS models derived individually from the two engine makes show that the emission and genotoxicity of particulate matter are much influenced by choice of engine. The results imply that, with a specified engine design (make), the genotoxic potency of particulate exhaust emission from diesel fuels can be predicted by the assessment of the physical and chemical fuel factors discussed above. A reduction of fuel sulfur and an increase in cetane number and possibly naphthenes can be expected to decrease the genotoxicity of the particulate exhausts. The exhaust components that were shown to have the highest impact on genotoxicity were the contents of 1-nitropyrene, particle bound nitrate, fuel-derived soluble organic fraction, and indeno[1,2,3-cd]pyrene. The contents of nitrate, soluble organics, and soluble sulfate correlated with genotoxicity. This indicated that a large portion of the S9-independent mutagenicity was due to the formation of NPAH. The higher mutagenicity observed with tester strain TA100 furthermore indicated the presence of other, undetected chemical species in the particulate extracts. The biological responses that were best described, by chemical variations in all models that were derived, were the mutagenicity data. There were consistently intercorrelations between the Ames test data. TA98 - S9 correlated with TA100 - S9 and TA98 + S9 correlated with TA100 + S9. This indicates that for complex mixtures, similar to those included in this study, testing with either strain and both metabolic conditions would suffice for biological characterization. Ah receptor affinity correlated with mutagenicity in the presence of S9 activation. The data set included in this study is being constantly upgraded as new fuels, samples, and analysis data become available. We intend to constantly refine the models for assessing genotoxic potency of diesel exhaust emissions, based on the combination of proper fuel selection, sampling and analysis techniques, biological testing, and powerful statistical tools, such as PCA and PLS. In this study this approach has been shown to be an effective strategy for obtaining the maximum amount of information regarding the genotoxicity of particulate exhaust emissions. The assessment of genotoxicity of other complex mixtures may benefit from this approach.

Acknowledgment. The authors wish to express their gratitude to Len Elfver, Yvonne Eklund, and Marianne Westman for their skillful technical assistance. KarlErik Egeba¨ck is gratefully acknowledged for technical advice on this paper. This work was supported by grants from the Swedish Environment Protection Agency and the Magn. Bergvalls Foundation.

Sjo¨ gren et al.

(6)

(7)

(8) (9) (10)

(11)

(12) (13) (14) (15) (16)

(17)

(18) (19) (20) (21)

(22) (23) (24)

References (1) Westerholm, R. (1987) Inorganic and organic compounds in emissions from diesel powered vehicles. A literature survey. SNV Report 3389, Swedish National Environmental Protection Board, Stockholm. (2) WHO (in press) Diesel fuel and exhaust emissions. International Programme on Chemical Safety, Environmental Health Criteria 171, World Health Organization, Geneva. (3) US Federal Register (1987) Protection of environment. Code of Federal Regulations Parts 81-99, Revised July 1, 1987, GPO, Washington, DC. (4) SFS (1991) The Swedish fiscal energy tax legislation (in Swedish: “Lagen om allma¨n energiskatt”). Svensk Fo¨ rfattningssamling 1957, 262. (5) Egeba¨ck, K.-E., and Bertilsson, B.-M. (1983) Chemical and biological characterisation of exhaust emissions from vehicles

(25)

(26) (27) (28)

(29)

fuelled with gasoline, alcohol, LPG, and diesel. SNV PM 1635, Swedish National Environmental Protection Board, Stockholm. Callahan, T., Ryan, T., Dietzman, H., and Waytulonis, R. (1985) The effect of engine and fuel parameters on diesel exhaust emission during discrete transients in speed and load. SAE paper 850110, Society of Automotive Engineers, Warrendale, PA, USA. Westerholm, R., Alsberg, T., Strandell, M., Frommelin, Å., Grigoriadis, V., Hantzaridou, A., Maitra, G., Winquist, L., Rannug, U., Egeba¨ck, K.-E., and Bertillson, T. (1995) Chemical analysis and biological testing of emissions from a heavy duty diesel truck with and without two different particle traps. SAE Paper 860014, Society of Automotive Engineers, Warrendale, PA, USA. Baranescu, R. (1988) Influence of fuel sulphur on diesel particulate emissions. SAE Paper 881174, Society of Automotive Engineers, Warrendale, PA, USA. Gra¨gg, K. (1991) Investigation of three diesel fuels regarding the impact of boiling point on the exhaust emissions (in Swedish). MTC 9109B, Motor Test Center, Haninge, Sweden. McCarthy, C. I., Slodowske, W. J., Sienecki, E. J., and Jass, R. E. (1992) Diesel fuel property effects on exhaust emissions from a heavy duty diesel engine that meets 1994 emissions requirements. SAE Paper 922267, Society of Automotive Engineers, Warrendale, PA, USA. Westerholm, R., and Li, H. (1994) A multivariate statistical analysis of fuel-related polycyclic aromatic hydrocarbon emissions from heavy-duty diesel vehicles. Environ. Sci. Technol. 28, 965972. Sjo¨gren, M., Li, H., Rannug, U., and Westerholm, R. (1995) A multivariate analysis of exhaust emissions from heavy duty diesel vehicles. Environ. Sci. Technol. (in press). Scheepers, P. T., and Bos, R. P. (1992) Combustion of diesel fuel from a toxicological perspective. II. Toxicity. Int. Arch. Occup. Environ. Health 64, 163-177. Lewtas, J. (1993) Complex mixtures of air pollutants: Characterizing the cancer risk of polycylic organic matter. Environ. Health Perspect. 100, 211-218. Pepelko, W. E., and Chen, C. (1993) Quantitative assessment of cancer risk from exposure to diesel engine emissions. Regul. Toxicol. Pharmacol. 17, 52-65. Heussen, H. (1994) Toxic effects of indoor and outdoor airborne particles relevant to carcinogenesis. Dissertation Thesis, Department of Toxicology, Agricultural University Wageningen, CIPData Koninklijke Bibliotheek, Den Haag, Holland. Henderson, R. F., and Mauderly, J. L. (1994) Diesel exhaust: An approach for the study of the toxicity of chemical mixtures. In Toxicology of chemical mixtures (Yang, R. S. H., Ed.) pp 119133, Academic Press, San Diego. Mauderly, J. L. (1994) Toxicogical and epidemiological evidence for health risks from inhaled engine emissions. Environ. Health Perspect. 102 (Suppl. 4), 165-171. Hext, P. M. (1994) Current perspectives on particluate induced pulmonary tumours. Hum. Exp. Toxicol. 13, 700-715. Oberdo¨rster, G., Ferin, J., and Lehnert, B. E. (1994) Correlation between particle size, in vivo particle persistence, and lung injury. Environ. Health Perspect. 102 (Suppl. 5), 173-179. Nikula, K. J., Snipes, M. B., Barr, E. B., Griffith, W. C., Henderson, R. F., and Mauderly, J. L. (1995) Comparative pulmonary toxicities and carcinogenicities of chronically inhaled diesel exhaust and carbon black in F344 rats. Fundam. Appl. Toxicol. 25, 80-94. Oberdo¨rster, G. (1995) Lung particle overload: Implications for occupational exposures to particles. Regul. Toxicol. Pharmacol. 27, 123-135. Wong, D., Mitchell, C. E., Wolff, R. F., Mauderly, J. L., and Jeffrey, A. M. (1986) Identification of DNA damage as a result of exposure of rats to diesel engine exhaust. Carcinogenesis 7, 1595-1597. Bond, J. A., Johnson, N. F., Snipes, M. B., and Mauderly, J. L. (1990) DNA adduct formation in rat alveolar type II cells: Cells potentially at risk for inhaled diesel exhaust. Environ. Mol. Mutagen. 16, 64-69. Gallagher, J., Heinrich, U., George, M., Hendee, L., Phillips, D. H., and Lewtas, J. (1994) Formation of DNA adducts in rat following chronic inhalation of diesel emissions, carbon black and titanium dioxide particles. Carcinogenesis 15, 1291-1299. Gustavsson, P., Plato, N., Lidstro¨m, E. B., and Hogstedt, C. (1990) Lung cancer and exposure to diesel exhaust among bus garage workers. Scand. J. Work Environ. Health 16, 348-354. International Agency for Research on Cancer (1989) Diesel and gasoline engine exhausts. IARC Monogr. Eval. Carcinog. Risk Chem. Hum. 46, 41-185. International Agency for Research on Cancer (1987) Overall evaluations of carcinogenicity: An updating of IARC monographs volumes 1 to 42. IARC Monogr. Eval. Carcinog. Risk Chem. Hum. Suppl., Suppl. 7. Maron, D. M., and Ames, B. N. (1983) Revised methods for the Salmonella mutagenicity test. Mutat. Res. 113, 173-215.

Multivariate Analysis of Diesel Particulates (30) Coller, H. A., and Thilly, W. G. (1994) Development and applications of mutational spectra technology. Environ. Sci. Technol. 11, 478-487. (31) Barale, R., Bulleri, M., Cornetti, G., Loprieno, N., and Wachter, W. F. (1992) Preliminary investigation on genotoxic potential of diesel exhaust. SAE Technical Paper Series 920397, Society of Automotive Engineers, Warrendale, PA, USA. (32) 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. (33) Claxton, L. D., Creason, J., Leroux, B., Agurell, E., Bagley, S., Bryant, D.W., Courtois, Y. A., Douglas, G., Clare, C. B., Goto, S., Quillardet, P., Jagannath, D. R., Kataoka, K., Mohn, G., Nielsen, P. A., Ong, T., Pederson, T. C., Shimizu, H., Nylund, L., Tokiwa, H., Vink, G. J., Wang, Y., and Warshawsky, D. (1992) Results of the IPCS collaborative study on complex mixtures. Mutat. Res. 276, 23-32. (34) Bagley, S. T., Stoltz, S. L., Becker, D. M., and Keen, R. E. (1992) Characterization of organic extracts from standard reference materials 1649, ‘urban dust/organics’, and 1650, ‘diesel particulate matter’, using a microsuspension assay. A WHO/IPCS/CSCM study. Mutat. Res. 276, 81-86. (35) Nylund, L., Hakala, E., and Sorsa, M. (1992) Application of a semiautomated SOS chromotest for measuring genotoxicities of complex environmental mixtures containing polycylic aromatic hydrocarbons. Mutat. Res. 276, 125-132. (36) Nardini, B., and Clonfero, E. (1992) Mutagens in urban air particulate. Mutagenesis 7, 421-425. (37) Westerholm, R. N., Almeń, J., Li, H., Rannug, J. U., Egeba¨ck, K.-E., and Gra¨gg, K. (1991) 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. Environ. Sci. Technol. 25, 332-338. (38) Savard, S., Otson, R., and Douglas, G. R. (1992) Mutagenicity and chemical analysis of sequential organic extratcs of airborne particulates. Mutat. Res. 276, 101-115. (39) Arey, J., Harger, W. P., Helmig, D., and Atkinson, R. (1992) Bioassay-directed fractionation of mutagenic PAH atmospheric photooxidation products and ambient particulate extracts. Mutat. Res. 281, 67-76. (40) 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. (41) Conney, A. H. (1982) Induction of microsomal enzymes by foreign chemicals and carcinogenesis by polycyclic aromatic hydrocarbons: G. H. A. Clowes memorial lecture. Cancer Res. 42, 48754917. (42) Rannug, U., Sjo¨gren, M., Rannug, A., Gillner, M., Toftgård, R., Gustafsson, J.-Å., Rosenkranz, H. S., and Klopman, G. (1991) Use of artificial intelligence in structure-affinity correlations of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) receptor ligands. Carcinogenesis 12, 2007-2015. (43) Rosenkranz, H. S., and Klopman, G. (1986) Mutagens, carcinogens and computers. In Genetic toxicology of environmental chemicals, Part A: Basic principles and mechanisms of action (Ramel, C., Lambert, B., and Magnusson, J., Eds.) pp 71-104, Liss, New York. (44) Westerholm, R., and Egeba¨ck, K.-E. (1991) Impact of fuels on diesel exhaust emissions. A chemical and biological characterization. SNV Report 3968, Swedish Environmental Protection Agency, Solna, Sweden. (45) Li, H., Sjo¨gren, M., Rannug, U., and Westerholm, R. (1993) The relationship between PAC emissions and mutagenicitysA multivariate analysis of diesel particulates. In Polycyclic Aromatic Compounds: synthesis, properties, analytical measurements, occurence and biological effects (Garrigues, P., and Lamotte, M., Eds.) pp 159-166, Gordon and Breach Science Publishers, Amsterdam. (46) Sjo¨gren, M., Li, H., Rannug, U., and Westerholm, R. (1995) A multivariate statistical analysis of chemical composition and physical characteristics of ten diesel fuels. Fuel 74, 983-989. (47) Toftgård, R., Lo¨froth, G., Carlstedt-Duke, J., Kurl, R. N., and Gustafsson, J.-Å. (1983) Compounds in urban air compete with 2,3,7,8-tetrachlorodibenzo-p-dioxin for binding to the receptor protein. Chem.-Biol. Interact. 46, 335-346.

Chem. Res. Toxicol., Vol. 9, No. 1, 1996 207 (48) Poellinger, L., Lund, J., Gillner, M., Hansson, L.-A., and Gustafsson, J.-Å. (1983) Physicochemical characterization of specific and nonspecific polyaromatic hydrocarbon binders in rat and mouse liver cytosol. J. Biol. Chem. 22, 13525-13542. (49) Wold, S., Albano, C., Dunn, W. J., III, Edlund, U., Esbensen, K., Geladi, P., Hellberg, S., Johansson, E., Lindberg, W., and Sjo¨stro¨m, M. (1984) Multivariate data analysis in chemistry. In Chemometrics (Kowalski, B. R., Ed.) pp 1-79, Reidel, Dordrecht, Holland. (50) Nordeń, B., Edlund, U., and Wold, S. (1978) Carcinogenicity of polycyclic aromatic hydrocarbons studied by SIMCA pattern recognition. Acta Chem. Scand. 32, 602-608. (51) Albano, C., Blomqvist, G., Coomans, D., Dunn, W. J., III, Edlund, U., Eliasson, B., Hellberg, S., Johansson, E., Nordeń, B., Johnels, D., Sjo¨stro¨m, M., So¨derstro¨m, B., Wold, H., and Wold, S. (1981) Pattern recognition by means of disjoint principals components models (SIMCA): Philosophy and methods. In Proceedings of a Symposium on Applied Statistics, January 22, 1981 (Ho¨skuldsson, A., and Esbensen, K., Eds.) pp 1-35, RECKU, Copenhagen, Denmark. (52) Sharaf, M. A., Illman, D. L., and Kowalski, B. R. (1986) Chemometrics, Wiley, New York. (53) Wold, H. (1982) Soft modelling. The basic design and some extensions. In Systems under indirect observation (Jo¨reskog, K.G., and Wold, H., Eds.) Chapter 1, North-Holland, Amsterdam, Holland. (54) Wold, S., Geladi, P., Esbensen, K., and O ¨ hman, J. (1987) Multiway principal components- and PLS-analysis. J. Chemom. 1, 4156. (55) Wold, S., Albano, C., Dunn, W. J., II, Esbensen, K., Hellberg, S., Johansson, E., Lindberg, W., and Sjo¨stro¨m, M. (1984) Modelling data tables by principal components and PLS: classs patterns and quantitative predictive relations. Analusis 12, 477-485. (56) Wold, S. (1978) Cross-validatory estimation of the number of components in factor and principal components models. Technometrics 20, 397-405. (57) Wold, S. (1991) Validation of QSAR’s. Quant. Struct.sAct. Relat. 10, 191-193. (58) Geladi, P., and Kowalski, B. R. (1986) An example of 2-block predictive partial least squares regression with simulated data. Anal. Chim. Acta 185, 19-32. (59) Rosenkranz, H. S., and Mermelstein, R. (1983) Mutagenicity and genotoxicity: All nitro-containing chemicals were not created equal. Mutat. Res. 114, 217-266. (60) Klopman, G., and Rosenkranz, H. S. (1984) Structural requirements for the mutagenicity of environmental nitroarenes. Mutat. Res. 126, 227-238. (61) Mersch-Sundermann, V., Klopman, G., and Rosenkranz, H. S. (1992) Structural requirements for the induction of the SOS repair in bacteria by nitrated polycyclic aromatic hydrocarbons and related chemicals. Mutat. Res. 265, 61-73. (62) Barale, R., Bulleri, M., Cornetti, G., Loprieno, N., and Wachter, W. F. (1992) Preliminary investigation on genotoxic potential of diesel exhaust. SAE Technical Paper Series 920397, SAE International, Warrendale, PA, USA. (63) Shane, B. S., Squadrito, G. L., Church, D. F., and Pryor, W. A. (1991) Comparative mutagenicity of nitrofluoranthenes in Salmonella typhimurium TA98, TA98NR, and TA98/1,8DNP6. Environ. Mol. Mutagen. 17, 130-138. (64) Jung, H., Shaikh, A. U., Heflich, R. H., and Fu, P. P. (1991) Nitro group orientation, reduction potential, and direct-acting mutagenicity of nitro-polycyclic aromatic hydrocarbons. Environ. Mol. Mutagen. 17, 169-180. (65) Hoyer, M. E., Keeler, G. J., and Ball, J. C. (1992) Detection of oxidative mutagens in an urban air-particulate extract: A preliminary study. Mutat. Res. 283, 295-299. (66) Platt, K. L., Pfeiffer, E., Petrovic, P., Friesel, H., Beermann, D., Hecker, E., and Oesch, F. (1994) Comparative tumorigenicity of picene and dibenz[a,h]anthracene in the mouse. Carcinogenesis 15, 2225-2231. (67) Toftgård, R., Franzeń, B., Gustafsson, J.-A° ., and Lo¨ftoth, G. (1985) Characterization of TCDD-receptor ligands present in extracts of urban particulate matter. Environ. Int. 11, 369-374.

TX950095W

Influence of Physical and Chemical Characteristics of Diesel Fuels

Recommend Documents