StructureâToxicity Relationships for the Effects to Tetrahymena

330

Chem. Res. Toxicol. 2005, 18, 330-341

Structure-Toxicity Relationships for the Effects to Tetrahymena pyriformis of Aliphatic, Carbonyl-Containing, r,β-Unsaturated Chemicals T. Wayne Schultz,† Tatiana I. Netzeva,‡ David W. Roberts,‡,§ and Mark T. D. Cronin*,‡ College of Veterinary Medicine, The University of Tennessee, 2407 River Drive, Knoxville, Tennessee 37996-4543, School of Pharmacy and Chemistry, Liverpool John Moores University, Byrom Street, Liverpool L3 3AF, England, and Chemistry Department, University of Wales Swansea, Singleton Park, Swansea SA2 8PP, Wales Received June 28, 2004

Toxicity data for 82 aliphatic chemicals with an R,β-unsaturated substructure were compiled. Toxicity was assessed in the 2-day Tetrahymena pyriformis population growth impairment assay. Toxic potency [log(IGC50-1)] for most of these chemicals was in excess of baseline narcosis as quantified by the 1-octanol/water partition coefficient (log Kow). The toxicity of the R,βunsaturated aldehydes was modeled well by log Kow in conjunction with the sum of partial charges on the vinylene carbon atoms (QC4 + QC3) and the energy of the lowest unoccupied molecular orbital (Elumo). These electronic descriptors were also successful at modeling the toxicity of R,β-unsaturated ketones. The toxicity of a range of acrylates was constant within about 0.2 of a log unit. Conversely, the toxicity of methacrylates and esters containing the vinylene group varied considerably and was explained by their hydrophobicity. The comparison of the quantitative structure-activity relationship (QSAR) for the methacrylates and esters with that for non-polar narcosis showed little significant difference and hence suggested that substitution on the carbon-carbon double bond in the methacrylates and vinylene unsaturated esters does not enhance toxicity over that of baseline. Substitution on the carbon-carbon double bond in the R,β-unsaturated aldehydes resulted in toxicity that was similar to that for saturated derivatives. Although an excellent hydrophobicity-dependent QSAR was developed for the esters containing ethynylene group, these compounds are considered to act as Michael-type acceptors. Attempts to combine different groups of Michael-type acceptors into a single QSAR, based on mechanistically derived descriptors, were unsuccessful. Thus, the modeling of the toxicity of the R,β-unsaturated carbonyl domain is currently limited to models for narrow subdomains.

Introduction Aliphatic R,β-unsaturated compounds, such as acrylates and methacrylates, are among the more prevalent industrial organic chemicals in the world as defined by the High Production Volume Chemicals list (1). Due to their widespread use and distribution, as well as the lack of suitable toxicity data for them, there is an increased need to obtain information relating to the hazardous effects of such chemicals from molecular structure (2). Since extensive toxicity testing is unlikely (due to cost and time etc), quantitative structure-activity relationships (QSARs)1 could provide a means to predict the toxicity of such compounds. The ability to use QSARs to estimate the relative ecotoxicity of aliphatic R,β-unsatur* Corresponding author. E-mail: [email protected]. Phone: + 44 151 231 2066. Fax: + 44 151 231 2170. † The University of Tennessee. ‡ Liverpool John Moores University. § University of Wales Swansea. 1 Abbreviations: QSARs, quantitative structure-activity relationships; IGC50, 50% growth inhibitory concentration; SAS, statistical analysis system; MOPAC, molecular orbital package; AM1, Austin model 1; SMILES, simplified molecular input line entry specification; log Kow, 1-octanol/water partition coefficient; Elumo, energy of the lowest unoccupied molecular orbital; Ehomo, energy of the highest occupied molecular orbital, EN, electronegativity; AH, absolute hardness; A, acceptor superdelocalizabiliy; D, donor superdelocalizability; Q, atomic partial charge.

ated molecules accurately would be of value to both regulatory organizations and industry. Despite the need for QSARs in this area, few are openly available and previous efforts to model the toxicity, such as to the fish, produced only mixed results (3). Most industrial organic chemicals exhibit a narcotic mode of toxic action (4). Such chemicals are thought to be non-reactive to macromolecules and exhibit acute toxicity that is due merely to membrane perturbation (5, 6). As narcosis depends only on the uptake of chemicals, the modeling of toxic potency of such molecules yields simple log Kow dependent models. However, there are a number of chemical classes (including those that may enter the environment in high quantities) whose toxicity is not dependent on narcosis, and so may not be modeled easily by log Kow (7). Additionally, for some chemical classes, such as the aliphatic isothiocyanates, toxic potency is independent of hydrophobicity (8, 9). Accordingly, for particular chemical classes the modeling of toxicity using a linear regression approach based on generic terms for hydrophobicity and electrophilicity as described earlier (10) may not be appropriate and can lead to substantial inaccuracies. The modeling of the toxicity of aliphatic compounds provides a good example of the problem faced by the increasing use of QSAR. In particular, as use moves from

10.1021/tx049833j CCC: $30.25 © 2005 American Chemical Society Published on Web 02/01/2005

QSARs for R,β-Unsaturated Chemicals

priority setting into risk assessment and chemical classification, the accuracy of the prediction becomes increasingly important. To achieve accurate prediction of toxicity, the capability to identify and establish the domain of a model is seen as being vitally important. This will also assist in the validation and future use of QSARs (11). With the exception of very trivial models, defining the domain of a QSAR is a difficult and complex task. This has resulted in a greater emphasis toward the development of more global models, which can be applied across a range of molecular substructures and mechanisms, as well as a framework within which to place and utilize such QSARs. The toxicity of heterogeneous compounds depends on a multiplicity of physical and/or chemical interactions between a toxicant and an often ill-defined molecular site of action (12). As a result determining the proper QSAR for toxicity prediction is typically not an easy task. Hence, to predict toxicity one can be left using a linear regression model based on generic hydrophobic, electrophilic, and size/shape terms, which may sacrifice accuracy as the structural domain is expanded (10). Thus, for chemicals such as R,β-unsaturated compounds, the development of QSARs for the prediction of their toxicity, must be placed in this more global framework. In other words, it is accepted that these compounds are not well described by a global model for toxicity and that the domains for their QSARs must be highly developed and defined rigorously. One of the best-studied electrophilic mechanisms of toxic action is Michael-type addition. This provides a means of covalent adduct formation at an electrophilic center, without the presence of a leaving group in the molecule (13). This mechanism of action is characteristic of carbonyl-containing polarized R,β-unsaturated aliphatic compounds such as acrylates (14). More specifically, the molecular mechanism of these polarized R,βunsaturates is considered to be addition of a thiol(ate) from the amino acid cysteine to the β-carbon atom in the vinyl moiety via a carbanion intermediate (15). However, methyl substitution on either of the vinyl carbons results in a reduction in toxic potency (14). This reduction is because the alkyl group substitution alters the electron density, and opposes the withdrawing effect of the carbonyl group, which impedes the addition on the β-carbon atom of the olefin. With regards to fish toxicity, Verhaar et al. (16) classified acrylates and methacrylates separately on the basis of their mechanism of toxic action into electrophiles (acrylates) and ester narcotics (methacrylates). Different mechanisms of toxic action for the structurally similar acrylates and methacrylates were also suggested from an analysis of their rate constants toward different nucleophiles (17). Using toxicity data for a small group of aliphatic polarized R,β-unsaturated derivatives of esters, aldehydes, and ketones, a series of six structuretoxicity relationships, each depending on a specific molecular substructure, was described by Schultz and Yarbrough (14). The structure-activity relationships derived by Schultz and Yarbrough (14) give a rationale for classification within that domain. However, the different extent of the toxicity potency provoked by the presence or absence of a given feature varies between the subclasses of molecules. Thus, for example, the toxicity increases more by altering an ester from an internal vinylene (e.g. ethyl crotonate) to internal ethynylene (e.g. ethyl tetrolate) as compared to the same

Chem. Res. Toxicol., Vol. 18, No. 2, 2005 331

change in the structure of a ketone (e.g. 3-buten-2-one vs 3-butyn-2-one). The net result of the assessment of the toxicity of these compounds is that considerable understanding of the alteration in toxicity brought about by structural modifications of the aliphatic, carbonylcontaining R,β-unsaturated compounds is possible. The trends described by Schultz and Yarbrough (14) motivated the acquisition of a larger data set to quantify the toxic effects of such compounds and their rational analysis based on mechanistically sound principles. The aims of this study, therefore, were to report the relative toxic potency to Tetrahymena pyriformis of a sizable series of aliphatic, carbonyl-containing, R,β-unsaturated compounds. Using these data attempts were made to develop empirically a succession of transparent and mechanistically clear ecotoxicity QSARs for different molecular subdomains. The purpose of the QSAR modeling was not necessarily to build the “best” model from a statistical point of view, but to build simple models that would enable further interpretation of possible mechanisms of action.

Materials and Methods Chemicals. The toxicity of 82 aliphatic, carbonyl-containing, polarized R,β-unsaturated compounds was evaluated (Table 1). The collection of compounds represents the majority of such chemicals which are available through traditional commercially sources and includes compounds from three chemical classes aldehydes, esters, and ketones. The esters include a variety of subclasses such as acrylates (CdCR1C(dO)OR2), methacrylates (CC(dC)C(dO)OR1), esters containing vinylene group (R1CdCC(dO)OR2) and esters containing an ethynylene group (R1CtCC(dO)OR2). The common formula of the compounds used in this study is shown in Figure 1. Caution: The following chemicals are hazardous and should be handled carefully. As reported here several of these chemicals have significant acute toxicity as well as being potential mutagens (18, 19) and skin sensitizers (20). The chemicals were obtained commercially (Aldrich Chemical Co., Milwaukee, WI; or Lancaster Synthesis Inc., Windham, NH). In the greater majority of cases purity was > 95%. In all cases, no further purification was undertaken. Biological Data. Population growth impairment testing with the common ciliate T. pyriformis (strain GL-C) was conducted following the protocol described by Schultz (21). This 40-h assay is static in design and uses population density quantified spectrophotometrically at 540 nm as its measured endpoint. The test protocol allows for eight to nine cell cycles in controls. Following range finding, each chemical was tested in three replicate tests. Two controls were used to provide a measure of the acceptability of the test by indicating the suitability of the medium and test conditions as well as a basis for interpreting data from other treatments. The first control had no test material was inoculated with T. pyriformis. The other, a blank, had neither test material nor inoculum. Each test replicate consisted of six to ten different concentrations of each R,βunsaturated chemical with duplicate flasks of each concentration. Only replicates with control-absorbency values of > 0.60 but of < 0.90 were used in the analyses. The 50% growth inhibitory concentration (IGC50) was determined for each compound tested by Probit Analysis of Statistical Analysis System (SAS) software (22). The y-values were absorbencies normalized as percentage of control. The X-values were the toxicant concentrations in milligrams per liter. Molecular Descriptors. Hydrophobicity was quantified by 1-octanol/water partition coefficient values (log Kow). The hydrophobicity values were measured or estimated by the ClogP for Windows (version 1.00) software (23). The data set does not include compounds with log Kow values less than -1.0 or higher

332

Chem. Res. Toxicol., Vol. 18, No. 2, 2005

Schultz et al.

Table 1. Compounds Considered in This Study, Their Toxicity, Residual to the Model (log(IGC50-1) ) 0.645(0.014) (log Kow) - 0.342(0.035) (Elumo) - 1.11(0.05), n ) 353, r2 (adj.) ) 0.859, r2 (pred.) ) 0.857, s ) 0.353, F (not given)), and Selected Descriptors log(IGC50-1) (mM)

Elumo (eV)

QC2 - QO1 (au)

QC4 + QC3 (au)

R,β-Unsaturated Aldehydes -0.139 0.284 1.01b -0.022 0.073 1.53b 0.083 -0.348 2.33b 0.088 0.495 1.01b 0.700 1.426 0.52b 0.754 0.919 1.15b 0.819 0.960 1.45b 0.862 0.688 1.68b 0.911 1.191 1.24a 1.069 0.826 2.05b 1.225 0.368 2.74b 1.335 0.672 2.68b 1.851 0.541 3.69b 1.873 2.942 -0.01a

-0.105 -0.083 -0.113 -0.151 -0.141 -0.594 -0.099 -0.584 -0.088 -0.089 -0.584 -0.130 -0.115 -0.139

0.5031 0.5045 0.5006 0.5020 0.4969 0.4958 0.4963 0.4961 0.4949 0.4950 0.4960 0.4961 0.4970 0.4878

-0.2989 -0.2989 -0.2902 -0.2954 -0.3491 -0.3098 -0.3411 -0.3096 -0.3413 -0.3410 -0.3100 -0.3467 -0.3469 -0.3964

3-methyl-2-cyclopenten-1-one 2-methyl-2-cyclopenten-1-one 4-methyl-3-penten-2-one 3-methyl-3-penten-2-one 3-penten-2-one 2-cyclopenten-1-one 3-hepten-2-one 3-octen-2-one 4-hexen-3-one 3-nonen-2-one 3-hexyn-2-one 3-butyn-2-one 3-buten-2-one 1-penten-3-one 1-hexen-3-one 1-octen-3-one

R,β-Unsaturated Ketones -1.323 -0.545 0.49b -0.826 -0.034 0.49b -0.644 -0.101 0.91b -0.345 0.262 0.82b 0.538 1.441 0.35a 0.638 1.755 -0.03b 0.698 0.819 1.57b 0.740 0.519 2.10b 0.930 1.395 1.04b 0.983 0.420 2.63b 1.319 2.360 0.17b 1.412 3.113 -0.89b 1.506 2.640 -0.01b 1.527 2.327 0.52b 1.656 2.121 1.04b 1.914 1.695 2.10b

-0.047 -0.006 0.059 0.075 0.053 -0.037 0.069 0.068 0.075 0.068 0.118 0.049 0.051 0.075 0.074 0.073

1663-39-4 106-63-8 4245-35-6 140-88-5 141-32-2 925-60-0 2998-23-4 96-33-3 999-61-1 999-55-3 2499-95-8 3066-71-5 2499-58-3 19485-03-1

tert-butyl acrylate isobutyl acrylate isoamyl acrylate ethyl acrylate butyl acrylate propyl acrylate n-pentylacrylate methyl acrylate 2-hydroxypropyl acrylate allyl acrylate n-hexylacrylate cyclohexylacrylate n-heptyl acrylate 1,3-butanediol diacrylate

-0.640 0.291 0.408 0.516 0.519 0.531 0.537 0.547 0.650 0.682 0.741 0.763 1.086 1.144

Acrylates -0.793 -0.017 -0.176 0.788 0.121 0.455 -0.217 1.141 1.565 0.905 -0.354 0.074 -0.351 1.124

2.03b 2.22a 2.65b 1.32a 2.36a 1.86b 2.91b 0.80a 0.30b 1.37b 3.44b 2.83b 3.97b 1.74b

0.135 0.042 0.044 0.041 0.041 0.041 0.040 0.001 -0.005 -0.009 0.040 0.076 0.040 -0.022

126-98-7 80-62-6 97-63-2 4655-34-9 2370-63-0 96-05-9 2210-28-8 4245-37-8 13861-22-8 97-86-9 97-88-1 142-09-6 688-84-6

methacrylonitrile methyl methacrylate ethyl methacrylate isopropyl methacrylate 2-ethoxyethyl methacrylate allyl methacrylate n-propyl methacrylate vinyl methacrylate propargyl methacrylate isobutyl methacrylate butyl methacrylate n-hexyl methacrylate 2-ethylhexyl methacrylate

-1.653 -1.218 -0.935 -0.881 -0.779 -0.677 -0.655 -0.619 -0.372 -0.279 -0.269 1.087 1.569

Methacrylates -0.949 -0.979 -1.044 -1.179 -0.569 -0.635 -0.906 -0.366 -0.043 -0.852 -0.985 -0.177 -0.216

0.68a 1.38a 1.94a 2.25a 1.45b 1.68b 2.16b 1.26b 1.21b 2.66a 2.88a 3.73b 4.54a

0.094 0.055 0.093 0.126 0.102 0.045 0.093 -0.130 -0.001 0.094 0.093 0.093 0.097

924-50-5 623-43-8 18060-770 623-70-1 79218-15-8 6622-76-0 5837-78-5 638-10-8 20474-93-5

methyl-3,3-dimethylacrylate methyl crotonate crotonic acid isopropyl ester trans-ethyl crotonate crotonic acid tert-butyl ester methyl tiglate tiglic acid ethyl ester ethyl-3,3-dimethyl acrylate allyl crotonate

Esters Containing Vinylene Group -1.061 -1.059 1.72b -0.918 -0.666 1.33b -0.782 -1.042 2.16b -0.765 -0.842 1.86b -0.639 -1.138 2.56b -0.617 -0.544 1.63b -0.496 -0.752 2.16b -0.484 -0.813 2.25b -0.481 -0.604 1.90b

0.002 0.000 0.068 0.037 0.124 0.043 0.078 0.036 -0.023

CAS no.

namec

497-03-0 623-36-9 not found 107-86-8 123-73-9 142-83-6 5362-56-1 4313-03-5 922-63-4 1070-66-2 5910-87-2 557-48-2 3913-81-3 107-02-8

trans-2-methyl-2-butenal 2-methyl-2-pentenal 2,4-dimethyl-2,6-heptadienal 3-methyl-2-butenal crotonaldehyde 2,4-hexadienal 4-methyl-2-pentenal trans,trans-2,4-heptadienalc 2-ethylacrolein 2-butylacrolein trans,trans-2,4-nonadienal trans-2,cis-6-nonadienal trans-2-decen-1-alc acrolein

2758-18-1 1120-73-6 141-79-7 565-62-8 625-33-2 930-30-3 1119-44-4 1669-44-9 2497-21-4 14309-57-0 1679-36-3 1423-60-5 78-94-4 1629-58-9 1629-60-3 4312-99-6

residual

log Kow

AC2 (eV-1)

-0.2927 -0.2949 -0.2771 -0.2794 -0.3321 -0.3454 -0.3301 -0.3298 -0.3333 -0.3302 -0.3083 -0.3560 -0.3775 -0.3788 -0.3789 -0.3789

0.3161 0.3180 0.3163 0.3170 0.3155 0.3182 0.3173 0.3152 0.3188


Chem. Res. Toxicol., Vol. 18, No. 2, 2005 333 Table 1 (Continued)

CAS no.

namec

log(IGC50-1) (mM)

residual

log Kow

10371-45-6 1567-14-2 589-66-2 7299-91-4 7367-81-9 111-79-5

Esters Containing Vinylene Group (continued) crotocic acid sec-butyl ester -0.417 -1.018 2.69b methyl-trans-2-methyl-2-pentenoate -0.376 -0.636 2.16b crotonic acid isobutyl ester -0.344 -1.014 2.78b crotonic acid n-butyl ester -0.159 -0.913 2.91b methyl-trans-2-octenoate 0.764 -0.338 3.44b methyl-2-nonenoate 1.039 -0.405 3.97b

23326-27-4 4341-76-8 16205-90-6 18937-79-6 55314-57-3 111-12-6 16930-95-3 10519-20-7 111-80-8 1322-12-9

methyl-2-butynoate ethyl-2-butynoate ethyl-2-hexynoate methyl-2-hexynoate ethyl 2-pentynoate methyl-2-octynoate ethyl-2-heptynoate ethyl-2-octynoate methyl-2-nonynoate ethyl-2-nonynoate

a

Esters Containing Ethynylene Group 0.403 1.271 0.45b 0.509 1.050 0.98b 0.700 0.567 2.04b 0.763 0.957 1.51b 0.825 1.032 1.51b 1.043 0.554 2.57b 1.106 0.632 2.57b 1.233 0.417 3.10b 1.322 0.491 3.10b 1.373 0.215 3.63b

Elumo (eV)

QC2 - QO1 (au)

QC4 + QC3 (au)

0.071 0.067 0.037 0.037 0.020 0.020

AC2 (eV-1) 0.3161 0.3180 0.3168 0.3170 0.3177 0.3177

0.140 0.183 0.214 0.171 0.208 0.172 0.215 0.215 0.172 0.215

Measured. b Calculated. c Technical grade. Table 2. Descriptors Used in the Study and Their Abbreviations abbreviation

Figure 1. The common formula of the compounds used in this study. These include R,β-unsaturated aldehydes (R1 ) H and R2 ) alkyl chain), R,β-unsaturated ketones (R1 ) R2 ) alkyl chain), and R,β-unsaturated esters (R1 ) OR3 and R2 ) alkyl chain).

log Kow Elumo Ehomo EN AH Amax Dmax

than 5. In addition, 22 quantum chemical descriptors were calculated. These were either orbital energies and their derivatives (e.g., superdelocalizabilities of individual atoms) or charge descriptors of single atoms and their derivatives (e.g., sums and differences between partial charges of the atoms from certain functional groups). The molecular orbital descriptors were calculated using the AM1 Hamiltonian, implemented in TSAR version 3.3 (Accelrys Inc, Oxford, England) (24) and the MOPAC93 software (25). Initially SMILES strings were converted into 3-D structures by the CORINA conformation analysis software, as implemented in the TSAR ver. 3.3 molecular spreadsheet (Accelrys Ltd, Oxford, England). The 3D structures underwent energy minimization by means of the COSMIC force field as implemented in TSAR and subsequent full geometry optimization with the AM1 Hamiltonian in the VAMP module of TSAR. The optimized structures were exported to MOPAC93 for the calculation of quantum-mechanical indices. A list of the descriptors used in this study is given in Table 2. Statistical Analyses. QSARs were developed using the regression procedure of the MINITAB statistical software version 13.1 (26). Log (IGC50-1) values reported in millimolar units were used as the independent variable. The appropriate terms listed in Table 2 acted as the dependent variables. Resulting models were measured for fit by the coefficient of determination adjusted for the degrees of freedom (r2 (adj.)). The uncertainty in the model was noted as the square root of the mean square error (s), while the predictivity of the model was described by the r2 (pred.) determined by the leave-one-out method. The statistical significance of the models was compared using Fisher’s criterion (F). Outliers were identified with reference to their residual values being outside the 95% confidence interval of the model.

Results The toxicity to T. pyriformis (log(IGC50-1)) of, along with the relevant chemical descriptors for, the 82 ali-

Aave Dave AC2 AC3 AC4 Qmax QHmax Qmin QO1 QC2 QC3 QC4 QC2 - QO1 QC2 + QO1 QC4 - QC3 QC4 + QO3

descriptor octanol-water partitioning coefficient energy of the lowest unoccupied molecular orbital energy of the highest occupied molecular orbital electronegativity; EN ) (Elumo + Ehomo)/2 absolute hardness; AH ) (Elumo - Ehomo)/2 maximum acceptor superdelocalizability of a single atom in a molecule maximum donor superdelocalizability of a single atom in a molecule average acceptor superdelocalizability in a molecule average donor superdelocalizability in a molecule acceptor superdelocalizability of C2 acceptor superdelocalizability of C3 acceptor superdelocalizability of C4 maximum partial charge of a single atom in a molecule (positive) maximum partial charge of a hydrogen atom in a molecule minimum partial charge of a single atom in a molecule (negative) partial charge of O1 partial charge of C2 partial charge of C3 partial charge of C4 difference between the partial charges of C2 and O1 sum of the partial charges of C2 and O1 difference between the partial charges of C4 and C3 sum of the partial charges of C4 and C3

phatic polarized R,β-unsaturated chemicals containing a carbonyl moiety evaluated in this study are presented in Table 1. Toxicity varied uniformly over a 3.5-fold range (from -1.65 to 1.91) on a logarithmic scale. Hydrophobicity varied over 5 orders of magnitude (from -0.89 to 4.54) on a log scale. A plot of log(IGC50-1) versus log Kow is presented in Figure 2. This plot reveals that most of these polarized R,β-unsaturates exhibit toxicity in excess of that predicted by non-polar narcosis, or baseline. Near the baseline are the methacrylates, and some other esters such as crotonates and tiglates. It is interesting to note that the aforementioned esters generally have negative residuals (i.e. they are below the baseline) for reasons that are discussed later.

334


Figure 2. A plot of toxicity (log(IGC50-1)) versus hydrophobicity (log Kow) for selected polarized R,β-unsaturated compounds; solid circle - aldehydes, empty circle - ketones, solid diamonds acrylates, solid square - methacrylates, empty diamonds esters containing a vinylene group, empty square - esters containing an ethynylene group.

The general response-surface, or generic two parameter QSAR, for the toxicity of aliphatic chemicals was found to be (10)

log(IGC50-1) ) 0.645(0.014) (log Kow) 0.342(0.035)(Elumo) - 1.11(0.05) n ) 353; r2 (adj.) ) 0.859; r2 (pred.) ) 0.857; s ) 0.353; F (not given) (1) where the error of the coefficients is shown in parentheses following the coefficient (in this and other equations in this paper), log Kow is the log of the 1-octanol/water partition coefficient and Elumo is the energy of the lowest unoccupied molecular orbital. The toxicity of the compounds assessed in this study was predicted by eq 1. These predictions are noted in Table 1; in the vast majority of cases, they were relatively poor regardless of class and subgroup. These predictions indicate that such chemicals cannot be modeled by this general QSAR and therefore require separate consideration. To achieve this in a rational manner, QSARs were developed initially for subgroups of compounds selected according to chemical class. The first class considered was the R,β-unsaturated aldehydes. The relationship between toxicity and hydrophobicity is shown in Figure 3. The solid circles in Figure 3 represent a congeneric series of alkyl R,β-unsaturated aldehydes for which there is, as expected, clearly a good relationship with log Kow. One compound (acrolein, the empty diamond in Figure 3) was vastly more toxic than predicted by the relationship between log Kow and toxicity for the other R,β-unsaturated aldehydes. This is thought to be because acrolein is unique in having both a terminal vinyl group and a terminal carbonyl group and can be considered a subset consisting of itself. Figure 3 also indicates that R,β-unsaturated aldehydes with a carbon atom substituted on one of the vinylene carbons (empty squares in Figure 3) are less toxic than predicted by the log Kow dependent model. Efforts to model all the derivatives in a single QSAR focused on using an electronic parameter as a second descriptor in addition to log Kow. Stepwise regression analyses on all the calculated electronic parameters listed in Table 2 with log Kow showed that neither global orbital descriptors (e.g., Elumo) nor local

Schultz et al.

Figure 3. A plot of toxicity (log(IGC50-1)) versus hydrophobicity (log Kow) for selected polarized R,β-unsaturated aldehydes; solid circle - alkyl derivatives, empty square - derivatives with a carbon atom substituted on one of the vinylene carbons, empty diamond - acrolein.

orbital descriptors (e.g., Amax) were important in modeling the toxicity of these compounds. Rather, charge parameters for single atoms, and descriptors derived from them, were determined to be both important and significant as demonstrated in the model

log(IGC50-1) ) 0.279(0.067) (log Kow) 140.7(15.7) (QC2 - QO1) + 70.3(7.8) n ) 14; r2 (adj.) ) 0.868; r2 (pred.) ) 0.755; s ) 0.234; F ) 44 (2) where (QC2 - QO1) is the difference in the atomic partial charges of the carbon and oxygen atoms of the carbonyl group. Interestingly, one can improve the statistical fit of the model by changing the charge parameter and adding a further global orbital energy parameter to the QSAR. (It is appreciated that in doing this we are in violation of the rule of five compounds for each descriptor proposed by Topliss and Costello (27)). More specifically, the sum of the partial charges on the vinylene carbon atoms (QC4 + QC3) was substituted for QC2 - QO1 and resulted in the model

log(IGC50-1) ) 0.301(0.080) (log Kow) 19.4(2.6) (QC4 + QC3) - 6.02(0.90) n ) 14; r2 (adj.) ) 0.816; r2 (pred.) ) 0.741; s ) 0.276; F ) 30 (3) While eq 2 has much poorer fit than eq 3, adding Elumo as a third parameter dramatically improves the fit and predictivity of the model:

log(IGC50-1) ) 0.293(0.034) (log Kow) 21.7(1.18) (QC4 + QC3) - 1.19(0.17) (Elumo) 7.01(0.41) n ) 14; r2 (adj.) ) 0.966; r2 (pred.) ) 0.934; s ) 0.119; F ) 123 (4) Table 3 shows the correlation matrix between the four parameters used in eqs 3 and 4, as expected there is a significant correlation between the two charge parameters, but the three parameters included in eq 4 are not



Table 3. Correlation Coefficients (r) between the Variables Used in Eqs 3 and 4 for the Compounds Considered in Those Equations Elumo QC2 - QO1 QC4 + QC3

log Kow

Elumo

QC2 - QO1

-0.087 0.162 0.202

0.192 -0.286

0.831

significantly correlated. An examination of the plots of observed toxicity versus that predicted by eqs 3 and 4 reveals that, in both cases, adding the third descriptor improves the fit of the multivinylene-containing compounds (Figure 4). The relationship between toxicity and hydrophobicity for the R,β-unsaturated ketones is shown in Figure 5. Examination of Figure 5 shows that the toxicity of these compounds is independent of log Kow. As opposed to toxicity varying with hydrophobicity, the toxicity of vinylsubstituted (empty diamonds in Figure 5) and the vinylene-substituted (solid circles in Figure 5) derivatives are constants at 1.56 ( 0.21 and 0.98 ( 0.17, respectively. Moreover, R,β-unsaturated ketones with a carbon atom substituted on one of the vinylene carbons (empty squares in Figure 5) are less toxic than either of the other two groups. While this subgroup of compounds is small, the toxic potency is lower when carbon substitution is on the vinylene carbon at greatest distance to the

Figure 4. A plot of observed toxicity (log(IGC50-1)) versus predicted toxicity (log(IGC50-1)) based on eq 3 (empty symbols) and eq 4 (solid symbols) for selected polarized R,β-unsaturated aldehydes. Note the better fit of the multivinylene-containing derivatives (triangles) with eq 4.

Figure 5. A plot of toxicity (log(IGC50-1)) versus hydrophobicity (log Kow) for selected polarized R,β-unsaturated ketones; empty diamonds - vinyl-substituted derivatives, solid circle - vinylene-substituted, empty square - derivatives with a carbon atom substituted on one of the vinylene carbons.

carbonyl group (i.e., β-carbon atom) as opposed to when carbon is substituted onto R-carbon atom or the vinylene carbon nearest to the carbonyl group (see Table 1). Stepwise regression analysis of the toxicity of the R,βunsaturated ketones with all calculated electronic descriptors listed in Table 2 lead to the development of the following two-parameter model:

log(IGC50-1) ) -21.4(2.1) (QC3 + QC4) + 9.61(1.68) Elumo - 6.93(0.69) n ) 16; r2 (adj.) ) 0.917; r2 (pred.) ) 0.860; s ) 0.279; F ) 84 (5) The plot of observed toxicity versus that predicted by eq 5 is illustrated in Figure 6. A plot of hydrophobicity versus toxicity for the R,βunsaturated esters is presented in Figure 7. Accordingly, R,β-unsaturated esters were subdivided into four categories for modeling purposes: 1) acrylates (empty circles in Figure 7); 2) methacrylates (empty squares in Figure 7); 3) those with an internal vinylene group (solid squares in Figure 7), and 4) those with an internal ethylnylene group (solid circles in Figure 7). Figure 7 indicates that the log(IGC50-1) values for acrylates are independent of hydrophobicity and are

Figure 6. A plot of observed toxicity versus that predicted from eq 5 for selected polarized R,β-unsaturated ketones.

Figure 7. A plot of toxicity (log(IGC50-1)) versus hydrophobicity (log Kow) for selected polarized R,β-unsaturated esters; empty circles - acrylates, empty squares - methacrylates, solid squares - esters containing vinylene group, solid circles - esters containing ethynylene group, solid down-triangle - diacrylatederivative, solid up-triangle - tert-butyl acrylate, solid diamond - propargyl methacrylate, empty diamond - vinyl methacrylate.

336


Figure 8. A plot of toxicity (log(IGC50-1)) versus hydrophobicity (log Kow) for selected methacrylates; solid circle - alkyl derivatives, open circle - methacronitrile, empty square - vinyl methacrylate, empty diamonds - propargyl methacrylate.

constant at 0.61 ( 0.21. The exceptions are the only diacrylate tested (the solid down-triangle in Figure 7) which is more toxic and the sterically hindered tert-butyl substituted derivative (the solid up-triangle in Figure 7). However, a closer look at Figure 7 offers another interpretation. For the toxicity of the seven linear acrylates only, there is a horizontal line for C1 up to C5 with a mean log inverse of the toxicity of 0.53 ( 0.013 after which there is a strong positive slope with log Kow. It is interesting to compare the observed log(IGC50-1) values for these acrylates with those predicted from the general narcosis equation [log(IGC50-1) ) 0.723 (log Kow) - 1.79; n ) 215, r2 (adj.) ) 0.926, r2 (pred.) ) 0.925, s ) 0.274, F (not given) (10). In such comparisons, the C7 and C6 acrylates both model well by the general narcosis equation. However, the C5 to C1 acrylates are increasingly more toxic than predicted by this relationship. There is a strong relationship between the toxicity of the congeneric series alkyl methacrylates and log Kow. This relationship is shown in Figure 8 and described by the model

log(IGC50-1) ) 0.831(0.079) (log Kow) - 2.35(0.20) 2

2

n ) 11; r (adj.) ) 0.916; r (pred.) ) 0.890; s ) 0.277; F ) 111 (6) Methacronitrile (indicated by the empty circle in Figure 8) is structurally unique, while being a R-carbon atomsubstituted R,β-unsaturated chemical it does not contain a carbonyl group; thus, it is related, in part, to the others in this group. It fits eq 6 whereas compounds with additional unsaturated groups (e.g., vinyl methacrylate and propargyl methacrylate, the empty and solid diamonds in Figure 8 respectively) have toxicity greater than that predicted by eq 6. Similar to the alkyl methacrylates, the toxicity of the vinylene-containing ester derivatives is related strongly to hydrophobicity and modeled by the equation

log(IGC50-1) ) 0.732(0.097) (log Kow) - 2.12(0.024) n ) 15; r2 (adj.) ) 0.799; r2 (pred.) ) 0.728; s ) 0.257; F ) 57 (7) Equation 7 is improved by the addition of the local orbital energy parameter AC2, which is the superdelocalizability of the carbonyl carbon atom. This relationship is noted

Schultz et al.

Figure 9. A plot of observed toxicity versus that predicted by eq 8 for selected vinylene-containing R,β-unsaturated esters.

in eq 8 and observed toxicity versus that predicted by eq 8 is shown in Figure 9. It is probable that the two compounds with the highest toxicity (i.e. in the top right corner of Figure 9) will influence the statistical fit and this is reflected in the drop in leave-one-out coefficient of determination r2 (pred.) from that performed on the complete data set (r2 (adj.))

log(IGC50-1) ) 0.756(0.060) (log Kow) + 190(41) (AC2) - 62.5(12.9) n ) 15; r2 (adj.) ) 0.923; r2 (pred.) ) 0.885; s ) 0.159; F ) 85 (8) Interestingly, combining the toxicity data for methacrylates (without the vinyl- or propargyl-derivative) and the vinylene-containing esters and subsequent reanalysis yields the model

log(IGC50-1) ) 0.795(0.059) (log Kow) - 2.26(0.146) n ) 26; r2 (adj.) ) 0.880; r2 (pred.) ) 0.864; s ) 0.257; F ) 184 (9) The toxicity of the ethynylene-containing compounds is also related to hydrophobicity as described by

log(IGC50-1) ) 0.321(0.029) (log Kow) + 0.240(0.068) n ) 10; r2 (adj.) ) 0.932; r2 (pred.) ) 0.919; s ) 0.088; F ) 124 (10) The slope in eq 10 is markedly less, and the intercept higher, than that for eqs 7-9. The consistency in the significant descriptors used in eqs 3, 4, and 5 led to the combination of the data sets for the R,β-unsaturated aldehydes and R,β-unsaturated ketones and subsequent reanalysis. Of the three descriptors used in eqs 3-5, only log Kow and (QC3 + QC4) were significant in the combined regression though there was relatively poor statistical fit (r2 (adj.) ) 0.737; r2 (pred.) ) 0.695; model not shown). An attempt to include the acrylates in this group resulted in a statistically significant three-parameter (log Kow, (QC3 + QC4), and Elumo) QSAR; however the fit (r2 (adj.) ) 0.67) and predictivity ((r2 (pred.) ) 0.63) were even poorer than the QSAR for the aldehydes and ketones alone. No further efforts were made to model any of the data reported here by combining data within subgroups.



Table 4. Coefficient of Determination (r2) for Eqs 2-10 Following Randomization of the Biological Dataa

a

eq 2

eq 3

eq 4

eq 5

eq 6

eq 7

eq 8

eq 9

eq 10

r1 r2 r3 r4 r5 r6 r7 r8 r9 r10

0.337 0.025 0.311 0.027 0.068 0.067 0.029 0.067 0.021 0.040

0.319 0.068 0.317 0.002 0.049 0.039 0.01 0.050 0.047 0.009

0.323 0.069 0.320 0.056 0.187 0.071 0.026 0.181 0.132 0.280

0.173 0.012 0.177 0.467 0.246 0.153 0.231 0.044 0.397 0.148

0.078 0.072 0.184 0.137 0.075 0.043 0.105 0.052 0.029 0.001

0.001 0.028 0.000 0.025 0.004 0.090 0.011 0.001 0.005 0.056

0.401 0.059 0.015 0.071 0.077 0.136 0.069 0.161 0.082 0.092

0.029 0.054 0.076 0.035 0.040 0.058 0.031 0.025 0.007 0.010

0.078 0.026 0.002 0.067 0.175 0.376 0.108 0.000 0.046 0.329

mean

0.099

0.091

0.165

0.205

0.078

0.022

0.116

0.037

0.121

SD

0.120

0.122

0.112

0.141

0.053

0.030

0.108

0.021

0.133

The random sample generator implemented in MINITAB ver. 14.1 was used; rn ) randomization number.

Figure 10. The mechanism of nucleophilic addition to the R-carbon of the vinyl group of carbonyl-containing, polarized R,β-unsaturated compounds where Nu ) nucleophile, X ) carbonyl moiety, e ) electrophile, I ) intermediate, and p ) product.

The stability of eqs 2-10 was assessed by randomizing the biological activity data and recalculating the QSAR - the so-called process of Y-scrambling. The results of this randomization are shown in Table 4. No significant relationships were formed with the randomized data. The results in Table 4, along with the mechanistic interpretation of the models, provides confidence in their significance.

Discussion There has been growing interest in the prediction of the ecological and environmental effects of chemicals. The capability to predict acute toxicity to aquatic species is fundamental to this goal. Recent approaches have centered on using 2-dimensional structure to delineate the applicability domains of a series of QSARs. For aliphatic compounds this has resulted in, among others, the QSAR (eq 1) presented by Schultz et al. (10). The domain of eq 1 included a wide range of chemical structures which incorporated non-reactive and non-specific electrophilic mechanisms of toxic action. However, it specifically excluded carbonyl-containing R,β-unsaturated aliphatic compounds. This study has assessed the toxicity of this group of compounds. The predicted values for toxicity from eq 1, as well as the information shown in Figure 1, clearly demonstrate that eq 1 is not adequate to model the effects of these chemicals. The poor fit of the toxicity of the carbonyl-containing R,β-unsaturated aliphatic compounds to eq 1 is in accordance with the findings of Schultz and Yarbrough (14). These authors noted that for carbonyl-containing aliphatic R,β-unsaturates: 1) the acetylene-substituted derivative was more toxic than the corresponding olefinsubstituted compound; 2) the terminal vinyl-substituted derivative was more toxic than that substituted by an internal vinylene group; 3) methyl substitution on the vinyl carbon atoms reduces toxicity; 4) compounds with the vinyl group at the terminal position of the molecule are more toxic than those with carbonyl group at the terminal position; 5) homologues containing additional unsaturated moieties were more toxic than homologues having corresponding saturated moieties; 6) homologues

with branched hydrocarbon moieties were less toxic than those with straight-chain ones. The mechanisms of action for the R,β-unsaturated carbonyl compounds require careful consideration to allow for interpretation of the results. The carbonyl moiety can, in the case of aldehydes and certain activated ketones, itself undergo a reaction (i.e., Schiff-base formation) as is thought to be the mechanism of toxic action for saturated aldehydes (15). This reaction is possible for R,β-unsaturated aldehydes, but in most cases (though not all, see below) these react more readily by Michael-type addition. Hence for carbonyl-containing R,β-unsaturated chemicals the most likely molecular mechanism is addition across the carbon-carbon double or triple bond, i.e., Michael-type addition. More specifically, nucleophilic addition to electrophilic olefins or acetylenes is activated by an electron-withdrawing group represented in Figure 10. The charge on the nucleophiles (Nu) changes by -1 in going from the starting electrophile (e) to intermediate (i) prior to forming the stable product (p). For a given nucleophile, the activation energy is approximated by the energy difference between the electrophile and the intermediate. Reactivity differences between different electrophiles depend on the abilities of different Y groups to stabilize, by resonance and inductive effects, the negative charge on the R-carbon of the intermediate. For a constant substitution pattern on the R,β-carbons, the ranking order of Y groups is - ester (COOR) < ketone (COR) < aldehyde (CHO). When the starting material has a triple bond between the R,β-carbon atoms, the intermediate has an sp hybridized R-carbon, which is more electronegative than an sp2 hybridized carbon as in the intermediate form of an olefinic analogue. Hence, triple bonds between the R,βcarbons are more reactive than corresponding double bonds. The degree and pattern of alkyl substitution on the R,βcarbons has a strong impact on the reactivity and toxicity of R,β-unsaturated carbonyl compounds. Alkyl groups on the R-carbon destabilize the negative charge on the intermediate by their electron donating inductive and

338


Figure 11. The reversible conversion of the carbonyl-containing, polarized R,β-unsaturated electrophile (ec) to the zwitterion (ez).

hyperconjugation effects. Alkyl groups on the β-carbon can similarly destabilize the negative charge on the intermediate, although to a lesser extent, but more importantly, they can also stabilize the electrophile by their electron-donating hyperconjugation and inductive effects. This can be conveniently illustrated, using simple valence bond terminology, for an electrophile in which Y is a carbonyl group. The electrophile can be represented as a resonance hybrid between the classical structure and the zwitterionic structure (Figure 11). Alkyl groups on the β-carbon stabilize the positive charge on the zwitterion (cf., the relatively high stability of simple tertiary carbonium ions and the low stability of primary carbonium ions), thereby increasing the contribution of the ion and reducing the energy of the electrophile. Thus, the activation energy for nucleophilic addition to the electrophile is increased by alkyl groups on either the R- or β-carbons. The effect of R-substituted alkyl groups is mainly to increase the energy of the intermediate, whereas, the effect of β-substituted alkyl groups mainly lowers the energy of the electrophile. For a given Y group the ranking order of reactivity is R2Cd CRY < R2CdCHY, RCHdCRY < CH2dCRY, and RCHd CHY < CH2dCHY. As a result, of all the R,β-unsaturated carbonyl-containing substances studied here, acrolein is the most reactive, and the methacrylates are the least reactive. On this basis the whole set of compounds examined in this study can be classified into sub-groups, such that within each sub group all compounds are similarly reactive. For aldehydes we have the subgroups (in descending order of reactivity toward nucleophilic addition): A1) acrolein (sole member of its subgroup); A2) monosubstituted acroleins, which include RCHdCHCHO (i.e., vinylene aldehydes) and CH2)CRCHO, and A3) disubstituted acroleins represented by RCHdCRCHO and R2CdCHCHO. The toxicity pattern for these R,β-unsaturated aldehydes corresponds well with the above-described chemistry-based sub-grouping. Acrolein, the unique member of subgroup 1, is the most toxic, the compounds of subgroup A2 fit a log Kow based QSAR [log(IGC50-1) ) 0.357(0.041) (log Kow) + 0.376(0.087), n ) 9, r2 (adj.) ) 0.904; r2 (pred.) ) 0.793, s ) 0.114, F ) 76] and the compounds of subgroup A3 are less toxic than predicted by this equation. In fact, the compounds of subgroup A3 are modeled well by the QSAR for the toxicity of saturated aliphatic aldehydes [log(IGC50-1) ) -0.503 (log Kow) - 3.386 (Elumo) + 2.06; n ) 17, r2 (adj.) ) 0.919, r2 (pred.) ) 0.898, s ) 0.200, F (not given)] (10) suggesting that for these aldehydes the Schiff’s-base reaction dominates over addition to the olefinic double bond. It is interesting to consider these compounds a little further. In general an R,β-unsaturated aldehyde is expected to be more reactive than a saturated aldehyde in Schiff base formation, since the carbonyl group is activated by the electron-withdrawing sp2 hybridized

Schultz et al.

R-carbon. Hence, for R,β-unsaturated aldehydes acting by the Schiff base mechanism of action, we would expect toxicity to be somewhat under-predicted by the Schiff base SAR derived from saturated aldehyde data. This is the case for 2-methyl-2-butenal, 2-methyl-2-pentenal, and 3-methyl-2-butenal. It is worth noting that 3-methyl-2butenal which, being unsubstituted in the R-position has the most electronegative R-carbon, has the largest excess toxicity. However, for 2,4-dimethyl-2,6-heptadienal the trend appears reversed. This is a compound containing nine carbon atoms, and for compounds with carbon skeletons of this size it has been found that the effect of chain branching on hydrophobicity is better modeled by the position dependent branch factor (PDBF) than by those used in the classical log Kow calculation (28). For 2,4dimethyl-2,6-heptadienal there are two methyl branches, each with a PDBF of -0.43, whereas the classical calculation uses branch factors of -0.22 (for branch on the R-carbon) and -0.13 (for branch on the γ-carbon). If we adjust the calculated log Kow value accordingly, it is reduced by 0.44 log units, and the toxicity, calculated from the saturated aldehyde QSAR becomes -0.204 (i.e., an under-prediction of 0.107), which is in line with the other R-carbon-substituted R,β-unsaturated aldehydes. In the case of the R,β-unsaturated aldehydes, toxicity is influenced by hydrophobicity. However the influence of hydrophobicity in toxicity is small when compared to baseline narcosis or even the general response-surface model (eq 1). To model the toxicity of these compounds, a second descriptor is required and, as would be expected, it is electronic in nature (29). Successful descriptors (e.g. in eq 3) were related to the partial charges on the vinylene carbon atoms (QC4 + QC3) rather than being molecular orbital-based energy-related parameters. This is probably because this parameter is related to the vinylene group or, more specifically, the presence or absence of methyl substitution on the vinylene carbons. Interestingly, the third significant variable in eq 4 was the molecular-based, orbital energy parameter, Elumo. This is acceptable from a mechanistic point of view as Elumo improves the fit of the multi-vinylene-containing derivatives to the model. Further, it is postulated that the reason the difference in the partial charges of the carbon and oxygen atoms of the carbonyl group (QC2 QO1) was found to be significant in eq 2 was not because of the carbonyl group being the site of the reaction but rather due its collinearity with (QC4 + QC3) (r ) 0.831). From a mechanistic point of view, eqs 3 and 4 are considered more acceptable than eq 2. For the R,β-unsaturated ketones we have the subgroups: K1) monosubstituted ketones, which include RCOCHdCH2, K2) disubstituted ketones of the form RCOCHdCHR, and K3) trisubstituted ketones consisting of either RCOCHdCR2 or RCOCRdCHR. Rationalization of the toxicity of the R,β-unsaturated ketones is similar to that described for the aldehydes. Specifically, the presence of a vinyl or vinylene moiety in the compound is important to potency. As with the R,β-unsaturated aldehydes, the atom-based, charge-related parameter (QC4 + QC3) was observed to be significant in modeling the toxicity of the R,β-unsaturated ketones (see eq 5). The fact that (QC4 + QC3) was found to be important for predicting the toxicity of the R,β-unsaturated aldehydes (see above) and ketones supports the hypothesis in the alteration of the primary molecular mechanism of toxicity. A key difference is that the toxicity of the ketones,



Figure 12. The intramolecular hydrogen bonding intermediate form of 1,3-diol diacrylate.

unlike the R,β-unsaturated aldehydes is not dependent on log Kow (see Figure 5). Again, the orbital energy parameter, Elumo was found to be of secondary importance to model toxicity. For these compounds, it explains the difference in potency between the vinyl- and vinylene-substituted derivatives. R,βUnsaturated ketones with a carbon atom substituted on one of the vinylene carbons (the empty squares in Figure 5) are less toxic than other subgroups. Interestingly, the toxicity of these compounds fits the general aliphatic response-surface model (eq 1) well and, with the exception of 3-methyl-2-cyclopenten-1-one, have relatively large residual values when compared to the baseline model (10; see above). R,β-Unsaturated esters examined in this investigation were subdivided into four categories: E1) acrylates; E2) methacrylates, E3) those with an internal vinylene group, and E4) those with an internal ethylnylene group. Of these subgroups, the toxicity (and subsequent QSAR analysis) of the acrylates and methacrylates are the best studied (30). The acrylates are esters which have a terminal olefin attached to a carbonyl moiety to form a polarized R,β-unsaturated compound. They are considered classic examples of chemicals, which have the capacity to act via Michael-type addition (3, 31). The toxicity pattern of the acrylates is rather complex, but can be rationalized as reflecting a diversity of subtle effects on reactivity represented in the data set. Linear acrylates with five carbon atoms or less exhibit toxicity which is independent of hydrophobicity and effectively constant, as would be expected for a set of compounds of closely similar reactivity acting by an electrophilic toxicity mechanism. While the toxicity of higher molecular weight acrylates was found to be related to hydrophobicity, caution is encouraged as only acrylates up to seven carbon atoms in size were examined and we have observed (data not shown) that lauryl acrylate with its 12 carbons atoms is not toxic at saturation. Cyclohexyl acrylate exhibits a toxicity which is greater than predicted by the general narcosis model (10) and is higher than the mean value of 0.53 for the linear one to five-carbon acrylates. This increased toxicity is thought to be because the cyclohexyl group is unable to have any significant steric interaction with the vinyl reaction center due to its fixed conformation. The diacrylate is more toxic (IGC50 lower by a factor of 4) than the linear acrylates. This enhanced toxicity is thought to reflect greater reactivity due to a neighboring group effect, such that the intermediate as well as the transition state leading to it is more stable than for the simple acrylates. An analogous effect is seen in skin sensitization (32). The hydroxypropyl acrylate is somewhat more toxic (IGC50 lower by a factor of 1.3) than the linear simple

acrylates. The higher toxicity can be attributed to intramolecular hydrogen bonding between the hydroxyl and carbonyl groups (Figure 12). This hydrogen bonding reduces the activation energy of the intermediate and transition state. Again, an analogous effect is seen in skin sensitization (32). The isobutyl and isoamyl acrylates are less toxic than the linear analogues; this can be attributed to steric effects at the vinyl reaction center similar to those well documented for esterification and saponification reactions of branched acids, alcohols and esters (33). However, they are somewhat more toxic than predicted by the baseline model indicating that although reacting more slowly than the linear acrylates, they are probably still acting by the electrophilic mechanism. tert-Butyl acrylate is less toxic than the others in the class. It is even more sterically hindered than the other two-position branched acrylates. However, this alone does not completely explain its low toxicity, which is less even than that estimated by the baseline equation. tert-Butyl esters are known to readily undergo hydrolysis by SN1 solvolysis, dissociating into a carboxylate anion and the short-lived but relatively stable (compared to other carbonium ions) tert-butyl cation, which reacts with water to give tert-butanol (33). Following assessment of their mechanism of toxic action in fish, methacrylates are considered to act via ester narcosis (31). With the ciliate T. pyriformis the issue is somewhat different because the ester narcosis mechanism of action is not thought to exist in this organism (34). Due to their reduced reactivity, it is not surprising that the toxicity of the congeneric series of alkyl methacrylates is related strongly to hydrophobicity (eq 6). While the slope of this model is similar to that for baseline narcosis (10), it is surprising the intercept values for eq 6 is less than that of the baseline. Since toxic potency cannot be less than that predicted by the baseline, there are a number of possibilities to be considered to explain the toxicity of the methacrylates and their relationship with log Kow described in eq 6 (in particular the intercept of this model). From a physical perspective, toxicity may be less than that predicted by baseline because of abiotic transformation, biodegradation, or volatility. Of these three choices, the last, volatility, is the most likely explanation. The toxicity of the compounds containing a vinylene functional group was related to hydrophobicity. The log Kow dependent model (eq 7) is very similar to the model from nonpolar (baseline) narcosis described by Hermens et al. (8; see above). Equation 7 was improved following the addition of a local orbital energy parameter, the superdelocalizability of the carbonyl carbon atom (AC2; see eq 8). This subgroup is characterized by having the greatest structural diversity. This diversity includes

340


hydrocarbon branching on both the acid and alcohol sides of the ester. The relationship with AC2 appears to reflect the structural variations in the alkyl parts of these compounds. The toxicity of the ethylnylene-containing esters, although varying over a narrow range, was also related to hydrophobicity (eq 9). In this equation, the slope for log Kow is markedly reduced and similar to that associated with the R,β-unsaturated aldehydes. As it was shown previously, the toxicity of the strong Michael-type acceptors, such as the acrylates and R,β-unsaturated ketones demonstrated a trend to be independent of hydrophobicity and to have a very narrow range of variability (i.e. to exhibit an almost constant toxicity). Although there is no doubt that the ethylnylene-containing esters act via Michael-type addition, the statistically significant correlation with the log Kow could be an indirect result of the correlation of the partition coefficient with other physicochemical descriptors, such as the vapor pressure and Henry’s law constant. Thus, the lower the log Kow, the higher the volatility, and slope and the intercept of the line in eq 9 could be determined by the changes in the volatility rather than by the partitioning behavior of the compounds of this group. The mechanistic organic chemistry literature is replete with cases where individual compounds deviate from general structure-reactivity patterns, sometimes by factors of two or more, due to some structural feature remote (at least in terms of connectivity pathways) from the reaction site having a major effect on the activation energy (e.g. neighboring group effects) or on the activation entropy. Several such examples have been identified in the present data set, in particular with the acrylates. The electronic parameters used in QSAR often fail completely to model such situations. To the non-organic chemist the occurrence of dramatic effects of this type from apparently subtle structural variations can make reactivity-based QSAR modeling seem somewhat of a black art. Certainly the occurrence of such effects does make it very unlikely that reactivity QSARs derived purely by statistical analysis of structural parameters without an underlying transparent mechanistic basis can ever be applied with any confidence. However the science of mechanistic organic chemistry is well developed. We hope we have demonstrated that it also aids in brings clarity to structure-toxicity modeling. In conclusion, toxicity data for aliphatic R,β-unsaturates assessed in the 2-day Tetrahymena pyriformis population growth impairment assay were considered in this study and were, for most part, in excess of that predicted by non-polar (baseline) narcosis. The exceptions to these compounds having excess toxicity were the methacrylates and other esters such as the crotonates and tiglates. R,β-Unsaturated aldehydes are modeled well by the hydrophobic term log Kow in conjunction with the atom-based, charge-related sum of the partial charges on the vinylene carbon atoms (QC4 + QC3). The toxicity of the R,β-unsaturated ketones was independent of log Kow; however, (QC4 + QC3) was again a significant descriptor. The addition of the molecular orbital term Elumo improved the QSARs for both the aldehydes and ketones. The constant toxic potency of the group of acrylates did not permit the development of a QSAR. The toxicity of the methacrylates was modeled well in conjunction with that of the vinylene-containing esters using log Kow alone. The toxicity of the ethylnylene-containing esters was also

Schultz et al.

related to log Kow alone, however, the toxicity of the methacrylates and vinylene-containing esters and that of the ethylnylene-containing esters was related to hydrophobicity to vastly different extents. It is apparent that the for the modeling of the toxicity of aliphatic R,β-unsaturated compounds molecules there is only a small dependence on hydrophobicity and a much greater dependence on reactivity. Moreover, this dependence on reactivity is captured in particular by the atombased partial charges relating to the vinyl or vinylene carbons. Other non-specific electrophilic contributions were related to frontier orbital parameters. Efforts to model the complete set of R,β-unsaturated compounds considered in this study were largely unsuccessful.

Acknowledgment. Dr. Netzeva was supported by the European Union IMAGETOX Research Training Network (Grant HPRN-CT-1999-00015).

References (1) Green, S., Goldberg, A., and Zurlo J. (2001) TestSmart - high production volume chemicals: An approach to implementing alternatives into regulatory toxicology. Toxicol. Sci. 63, 6-14. (2) McKinney, J. D., Richard, A., Waller, C., Newman, M. C., and Gerberick, F. (2000) The practice of structure activity relationships (SAR) in toxicology. Toxicol. Sci. 56, 8-17. (3) Karabunarliev, S., Mekenyan, O. G., Karcher, W., Russom, C. L., and Bradbury, S. P. (1996) Quantum-chemical descriptors for estimating the acute toxicity of electrophiles to the fathead minnow (Pimephales promelas): An analysis based on molecular mechanisms. Quant. Struct.-Act. Relat. 15, 302-310. (4) Bradbury, S. P., and Lipnick, R. L. (1990) Introduction: Structural properties for determining mechanisms of toxic action. Environ. Health Perspect. 87, 181-182. (5) Bearden, A. P., Sinks, G. D., and Schultz, T. W. (1999) Acclimation to sublethal exposure to a model nonpolar narcotic: Population growth kinetics and membrane lipid alterations in Tetrahymena pyriformis. Aquat. Toxicol. 46, 11-21. (6) Roberts, D. W., and Costello, J. F. (2003) Mechanisms of action for general and polar narcosis: a difference in dimension. QSAR Comb. Sci. 22, 226-233. (7) Schultz, T. W., Sinks, G. D., and Bearden, A. P. (1998) QSAR in aquatic toxicology: A mechanism of action approach comparing toxic potency to Pimephales promelas, Tetrahymena pyriformis, and Vibrio fischeri. In Comparative QSAR (Devillers, J., Ed.) pp 52-109, Taylor and Francis, Washington, DC. (8) Hermens, J., Busser, F., Leeuwangh, P., and Musch., A. (1985) Quantitative correlation studies between the acute lethal toxicity of 15 organic halides to the guppy (Poecilla reticulata) and chemical reactivity towards 4-nitrobenzylpyridine. Toxicol. Environ. Chem. 9, 219-236. (9) Schultz, T. W., and Comeaux, J. L. (1996). Structure-toxicity relationships for aliphatic isothiocyanates to Tetrahymena pyriformis. Bull. Environ. Contam. Toxicol. 56, 638-642. (10) Schultz, T. W., Cronin, M. T. D., Netzeva, T. I., and Aptula, A. O. (2002) Structure-toxicity relationships for aliphatic chemicals evaluated with Tetrahymena pyriformis. Chem. Res. Toxicol. 15, 1602-1609. (11) Worth, A. P., Cronin, M. T. D., and van Leeuwen, C. J. (2004) A framework for promoting the acceptance and regulatory use of (quantitative) structure-activity relationships. In Predicting Chemical Fate and Toxicity (Cronin, M. T. D., and Livingstone, D. J., Eds.) pp 429-440, CRC Press: Boca Raton, FL. (12) Nendza, M., and Russom, C. L. (1991) QSAR modeling of the ERL-D fathead minnow acute toxicity database. Xenobiotica 21, 147-170. (13) Lipnick, R. L. (1991) Outliers: their origin and use in the classification of molecular mechanisms of toxicity. Sci. Total Environ. 109/110, 131-154. (14) Schultz, T. W., and Yarbrough, J. W. (2004) Trends in structuretoxicity relationships for carbonyl-containing R,β-unsaturated compounds. SAR QSAR Environ. Res. 15, 139-146. (15) Cronin, M. T. D., Sinks, G. D., and Schultz, T. W. (2001) Modelling of toxicity to the ciliate Tetrahymena pyriformis: the Aliphatic Carbonyl domain. In Forecasting the Environmental Fate and Effects of Chemicals (Rainbow, P. S., Hopkin, S. P., and Crane, M., Eds.) pp 113-124, John Wiley & Sons: Chichester, U.K.

QSARs for R,β-Unsaturated Chemicals (16) Verhaar, H. J. M., VanLeeuwen, C. J., and Hermens, J. L. M. (1992) Classifying environmental pollutants. 1: Structureactivity relationships for prediction of aquatic toxicity. Chemosphere 25, 471-491. (17) Freidig, A. P., Verhaar, H. J. M., and Hermens, J. L. M. (1999) Quantitative structure-property relationships for the chemical reactivity of acrylates and methacrylates. Environ. Toxicol. Chem. 18, 1133-1139. (18) Dearfield, K. L., Millis, C. S., Harrington-Brock, K., Doerr, C. L., and Moore, M. M. (1989) Analysis of genotoxicity of nine acrylate/ methacrylate compounds in L5178Y mouse lymphoma cells. Mutagenesis 4, 381-393. (19) Benigni, R., Passerini, L., and Rodomonte, A. (2003) Structureactivity relationships for the mutagenicity and carcinogenicity of simple and alpha-beta unsaturated aldehydes. Environ. Mol. Mutagen. 42, 136-143. (20) Cronin, M. T. D., and Basketter, D. A. (1994) A multivariate QSAR analysis of a skin sensitization database. SAR QSAR Environ. Res. 2, 159-179. (21) Schultz, T. W. (1997) TETRATOX: The Tetrahymena pyriformis population growth impairment endpoint - A surrogate for fish lethality. Toxicol. Methods 7, 289-309. (22) SAS Institute Inc. (1989) SAS/STAT User’s Guide, version 6, fourth edition, Vol. 2, pp 846, North Carolina. (23) ClogP ver 1.0 for Windows. (1995) BIOBYTE Corp., Claremont, CA. (24) TSARTM ver 3.3 for Windows. (2000) Accelrys, Oxford, England. (25) MOPAC 93 - Stewart, J. J. P. (1993) Fujitsu Ltd., Windows 95/ 98/NT/2k adaptation and MO indices - Kaneti J. (1988-94) MOQC. (26) MINITABTM ver 13.1 for Windows. (2000) Minitab Inc., State College, PA.

Chem. Res. Toxicol., Vol. 18, No. 2, 2005 341 (27) Topliss, J. G., and Costello, J. D. (1972) Chance correlations in structure-activity studies using multiple regression analysis. J. Med. Chem. 15, 1066-1069. (28) Roberts, D. W. (1991) QSAR issues in aquatic toxicity of surfactants. Sci. Total Environ. 109/110, 557-568. (29) Mekenyan, O. G., and Veith, G. D. (1994) The electronic factor in QSAR: MO-parameters, competing interactions, reactivity and toxicity. SAR QSAR Environ. Res. 2, 129-143. (30) Bearden, A. P., and Schultz, T. W. (1997) Relationships of structure-activity for Tetrahymena and comparative toxicity with Pimephales: A mechanism of action approach. Environ. Toxicol. Chem. 16, 1311-1317. (31) Russom, C. L., Bradbury, S. P., Broderius, S. J., Hammermeister, D. E., and Drummond, R. A. (1997) Predicting modes of toxic action from chemical structure: Acute toxicity in the fathead minnow (Pimephales promelas). Environ. Toxicol. Chem. 16, 948967. (32) Rustmeyer, T., Preuss, M., von Blomberg, B. M. E., Das, P. K., and Scheper. R. J. (2003) Comparison of two in vitro dendritic cell maturation models for screening contact sensitisers using a panel of methacrylates. Exp. Dermatol. 12, 682-691. (33) Gould, E. S. (1959) Mechanism and Structure in Organic Chemistry, Holt, Rinehart and Winston, New York. (34) Jaworska, J. S., Hunter, R. S., and Schultz, T. W. (1995) Quantitative structure-toxicity relationships and volume fraction analyses for selected esters. Arch. Environ. Contam. Toxicol. 29, 86-93.

TX049833J

StructureâToxicity Relationships for the Effects to Tetrahymena

Recommend Documents

StructureâToxicity Relationships for the Effects to Tetrahymena