- Email: [email protected]
Human developmental toxicity and mutagenesis
Mutation Research 422 Ž1998. 347–350
Short communication
Human developmental toxicity and mutagenesis Herbert S. Rosenkranz a,) , Ying Ping Zhang a , Orest T. Macina a , Donald R. Mattison a , Gilles Klopman b a
Department of EnÕironmental and Occupational Health UniÕersity of Pittsburgh, Pittsburgh, PA 15238, USA b Department of Chemistry Case Western ReserÕe UniÕersity, CleÕeland, OH 44106, USA Received 30 March 1998; revised 14 July 1998; accepted 23 July 1998
Abstract A previously described SAR model of human developmental toxicity was analyzed further. The model shows a number of mechanistic similarities with SAR models of other toxicological phenomena Žsystemic toxicity, chromosomal and genomic effects.. This implies that there are many targets associated with developmental effects. Surprisingly the analyses revealed no significant mechanistic overlap between developmental toxicity in humans and mutagenicity in Salmonella, a surrogate for the occurrence of point mutations. Our study indicates that this lack of similarity is likely the result of the pre-screening strategies which largely eliminate Salmonella mutagens from among the therapeutics introduced into human medicine. q 1998 Elsevier Science B.V. All rights reserved. Keywords: Developmental toxicity; Mutagenesis; SAR; Salmonella
1. Introduction In a recent report w1x we analyzed, using the expert SAR system CASErMULTICASE w2x, the features of an SAR Žstructure–activity relationship. model describing human developmental toxicity ŽHuDT.. The data for the model were derived from Briggs et al. w3x and TERIS w4x and, although based on human experience and on animal data, it is meant to describe the risk of developmental toxicity in humans. The data used to construct the model were primarily derived from oral administration of therapeutic agents. One of the features associated with the SAR models obtained with CASErMULTICASE, is ) Corresponding author. Tel. q1-412-967-6510; Fax: q1-412624-1289; E-mail: [email protected]
an ability to determine the structural overlaps between different SAR models w5x. The extent of overlap is an indication of mechanistic similarities. Thus, this approach can be used to document whether a complex phenomenon may be the consequence of more than one mechanism. For example, this enabled us to conclude that cell toxicity as well as mutagenicity contributed independently of one another, to carcinogenicity w6x. An examination of the above mentioned human developmental toxicity SAR model suggested that there was little structural overlap between that model and mutagenicity Žin Salmonella.. This was an unexpected finding as 20–30% of birth defects in humans have been attributed to germ cell and somatic mutations w7,8x. Moreover, even if germ cell mutagens induce developmental effects by nonmutagenic Ži.e., epigenetic. mechanisms w9x, there
0027-5107r98r$ - see front matter q 1998 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 7 - 5 1 0 7 Ž 9 8 . 0 0 1 7 0 - 5
348
H.S. Rosenkranz et al.r Mutation Research 422 (1998) 347–350
should still be some structural similarities. In view of these unexpected findings and their possible significance with respect to the causes of developmental effects in humans, we re-examined them. The present report confirms those earlier findings and we suggest a non-mechanistic basis for this lack of overlap.
2. Methods The CASErMULTICASE methodology as well as the rules used to determine structural overlaps have been described w10–13x. The SAR model of HuDT was described previously w1x. It is based upon a computerized version of TERIS w4,14x and a compilation in which FDA definitions were used to quantify risk of developmental toxicity from drug use during pregnancy w3,15x. The SAR models that were used for determination of structural overlaps have also been described previously w10–13x. The significance of the comparisons between the biophores associated with HuDT and those associated with other toxicological phenomena was determined based upon the assumption Žnull hypothesis. that mechanistically no overlap is expected between HuDT and the induction of a 2mnephropathy in male rats, binding to the Ah receptor, or inhibition of cytochrome P4502D.
3. Results The SAR model we used to investigate mechanistic relationships of human developmental effects ŽHuDT. consists of 130 developmental toxicants and 193 non-toxicants w1x. This SAR model shows a concordance of 74.8% between experimental results and predictions for chemicals external to the SAR model Žsensitivity 0.65, specificity 0.81.. This is an acceptable performance for an SAR model. Moreover, the biophores identified were statistically significant and therefore, could be used for the mechanistic studies contemplated w5,16x. By the criteria used to select significant CASE biophores, it was found that the HuDT model exhibited 40 such biophores. This number is significantly greater Ž p s 0.004. than the 20 significant CASE
biophores that were generated from an SAR model of Salmonella mutagenicity containing a similar number of chemicals Ži.e., n s 323. with the same ratio of active to inactive molecules. These findings suggest that the phenomenon responsible for the HuDT is much more complex than that associated with mutagenicity. This conforms with the notion that unlike mutagenicity, developmental toxicity is defined at the organismic rather than at the molecular level w8x. Analysis of the structural overlaps between the HuDT SAR model and SAR models of other toxicological endpoints indicated ŽTable 1. that HuDT overlapped with a wide spectrum of toxicological endpoints including acute toxicity Žcell toxicity, systemic toxicity, e.g., rat LD50 , maximum tolerated dose., chromosomal and genomic events Žsomatic mutationsrrecombinations in Drosophila, induction of sister chromatic exchanges, micronuclei, chromosomal aberrations, mutations in mouse lymphoma Table 1 Structural overlaps between the SAR model of developmental toxicity in humans and other SAR models SAR model
Percent overlap
p-Value
Rodent carcinogenicity—NTP Rodent carcinogenicity—CPDB Mutagenicity Salmonella Unscheduled DNA synthesis Mutagenicityrsomatic recombinations in Drosophila MTD—mouse MTD—rat Cell toxicity Rat lethality ŽLD50 . Sister chromatid exchanges in vitro Chromosomal aberrations in vitro Micronuclei in vivo Sister chromatid exchanges in vivo Mutations in mouse lymphoma cells Malsegregation in yeast a 2m-Nephropathy Žmale rats. a Inhibition cytochrome P4502D6 a Ah receptor binding a Inhibition tubulin polymerization
28 25 5 5 28
0.0009 0.002 0.3 0.3 0.0009
25 20 20 25 25
0.002 0.007 0.007 0.002 0.002
15 23 18 18
0.02 0.003 0.01 0.01
13 3 3 5 18
0.05 0.5 0.5 0.3 0.01
a
Based upon mechanistic considerations, these SAR models were assumed to bear no relationship to HuDT. For statistic analysis, overlaps with these data bases served to test the null hypothesis Žsee text..
H.S. Rosenkranz et al.r Mutation Research 422 (1998) 347–350
ŽMLA., including loss of heterozygosity ŽMLA.. and aneuploidy Žinduction of micronuclei, inhibition of tubulin polymerization, malsegregations.. There was also overlap between HuDT and carcinogenesis. Obviously, this overlap is not to be interpreted as indicating that carcinogenesis is a precursor of HuDT, but rather that some of the mechanisms contributing to carcinogenicity Že.g., DNA damage, mitogenesis. may also contribute to HuDT. The broad and significant overlaps with so many endpoints is not a usual feature of other phenomena studied with this approach. Frequently such analyses indicate mutually exclusive overlaps Že.g., carcinogenicity with either mutagenicity or cell toxicity w6x. but not joint ones, as is evident for HuDT ŽTable 1.. Given this broad spectrum of overlaps, the lack of significant overlap between HuDT and point mutations Žin Salmonella. is remarkable ŽTable 1. even while there were, as pointed out above, significant overlaps with other chromosomalrgenotoxic phenomena. The observation of a lack of overlap between the Salmonella mutagenicity and HuDT SAR models could be the result of the exclusion of chemicals mutagenic in Salmonella from consideration as therapeutics. In view of the widespread use of the Salmonella mutagenicity assay as a pre-screen to identify potential carcinogens, this may be a plausible explanation. In order to test this possibility, we compared the predicted prevalence of Salmonella mutagens among the universe of chemicals Ž37.2%. to the predicted prevalence among the chemicals forming the HuDT data base Ž20.1%.. The difference between the two predicted prevalences is highly significant Ž p s- 0.00001.. Thus, it is likely that the lack of structural overlap between HuDT and mutagenesis is a result of the pre-screening process used. Indeed, a hamster developmental toxicity SAR model based upon a population of molecules that presumably had not undergone such a pre-screen, showed a significant overlap Ž19.2% vs. 5.0%, p s 0.009. between that model and mutagenicity in Salmonella w17x. The present findings indicate that in fact the pre-screening strategies used are effective in preventing the introduction of mutagens or putative mutagens among therapeutics used in human medicine. The findings also indicate that while the HuDT SAR
349
model appears to be highly predictive w1x, it may lack specificity for agents capable of inducing point mutations Žwhich are usually detectable with the Salmonella mutagenicity assay.. Thus, in a strategy that uses the HuDT SAR model to prioritize chemicals for further testing, it might be appropriate to include the SAR model of Salmonella mutagenicity to identify potential mutagens which might be responsible for a proportion of human developmental defects w7,8x.
References w1x M. Ghanooni, D.R. Mattison, Y.P. Zhang, O.T. Macina, H.S. Rosenkranz, G. Klopman, Structural determinants associated with risk of human developmental toxicity, Am. J. Obstet. Gynecol. 176 Ž1997. 799–806. w2x G. Klopman, H.S. Rosenkranz, Prediction of carcinogenicityrmutagenicity using MULTICASE, Mutat. Res. 305 Ž1994. 33–46. w3x G.G. Briggs, R.K. Freeman, S.J. Yaffe, Drugs in Pregnancy and Lactation, 3rd ed., Williams and Wilkins, Baltimore 1990. w4x J.M. Friedman, B.B. Little, R.L. Brent, J.F. Cordero, J.W. Hanson, T.H. Shepard, Potential human teratogenicity of frequently prescribed drugs, Obstet. Gynecol. 75 Ž1990. 594–599. w5x M. Liu, N. Sussman, G. Klopman, H.S. Rosenkranz, Structure–activity and mechanistic relationships: the effects of chemical overlap on structural overlap in data bases of varying size and composition, Mutat. Res. 372 Ž1996. 79–85. w6x H.S. Rosenkranz, G. Klopman, Structural evidence for a dichotomy in rodent carcinogenesis: involvement of genetic and cellular toxicity, Mutat. Res. 303 Ž1993. 83–89. w7x E. Hodgson, P.E. Levi, A Textbook of Modern Toxicology, Elsevier, Amsterdam, 1979. w8x J.B. Bishop, K.L. Witt, R.A. Sloane, Genetic toxicities of human teratogens, Mutat. Res. 396 Ž1997. 9–43. w9x W.M. Generoso, A.G. Shourbaji, E.W. Piergorsh, J.B. Bishop, Developmental response of zygotes exposed to similar mutagens, Mutat. Res. 250 Ž1991. 439–446. w10x A. Labbauf, G. Klopman, H.S. Rosenkranz, Dichotomous relationship between DNA reactivity and the induction of sister chromatid exchanges in vivo and in vitro, Mutat. Res. 377 Ž1997. 37–52. w11x M. Rosenkranz, H.S. Rosenkranz, G. Klopman, Intercellular communication, tumor promotion and non-genotoxic carcinogenesis: relationships based upon structural considerations, Mutat. Res. 381 Ž1997. 171–188. w12x M. Liu, S.G. Grant, O.T. Macina, G. Klopman, H.S. Rosenkranz, Structural and mechanistic bases for the induction of mitotic chromosomal loss and duplication Ž‘malsegregation’. in the yeast Saccharomyces cereÕisiae: Relevance to
350
H.S. Rosenkranz et al.r Mutation Research 422 (1998) 347–350
human carcinogenesis and developmental toxicity, Mutat. Res. 374 Ž1997. 209–231. w13x A.R. Cunningham, H.S. Rosenkranz, Y.P. Zhang, G. Klopman, Identification of ‘genotoxic’ and ‘non-genotoxic’ alerts for cancer in mice: the carcinogenicity potency data base, Mutat. Res. 398 Ž1998. 1–17. w14x J.M. Friedman, J.E. Polifka, Teratogenic Effects of Drugs: A Resource for Clinicians ŽTERIS., The John Hopkins University Press, Baltimore, 1994.
w15x J.L. Schardein, Chemically Induced Birth Defects. Marcel Dekker, New York, 1993 . w16x B. Henry, S.G. Grant, G. Klopman, H.S. Rosenkranz, Induction of forward mutations at the thymidine kinase locus of mouse lymphoma cells: evidence for electrophilic and nonelectrophilic mechanisms, Mutat. Res. 397 Ž1998. 313–335. w17x J. Gomez, O.T. Macina, D.R. Mattison, Y.P. Zhang, G. ´ Klopman, H.S. Rosenkranz, Structural determinants of developmental toxicity in hamsters. in preparation Ž1998..
Short communication
Human developmental toxicity and mutagenesis Herbert S. Rosenkranz a,) , Ying Ping Zhang a , Orest T. Macina a , Donald R. Mattison a , Gilles Klopman b a
Department of EnÕironmental and Occupational Health UniÕersity of Pittsburgh, Pittsburgh, PA 15238, USA b Department of Chemistry Case Western ReserÕe UniÕersity, CleÕeland, OH 44106, USA Received 30 March 1998; revised 14 July 1998; accepted 23 July 1998
Abstract A previously described SAR model of human developmental toxicity was analyzed further. The model shows a number of mechanistic similarities with SAR models of other toxicological phenomena Žsystemic toxicity, chromosomal and genomic effects.. This implies that there are many targets associated with developmental effects. Surprisingly the analyses revealed no significant mechanistic overlap between developmental toxicity in humans and mutagenicity in Salmonella, a surrogate for the occurrence of point mutations. Our study indicates that this lack of similarity is likely the result of the pre-screening strategies which largely eliminate Salmonella mutagens from among the therapeutics introduced into human medicine. q 1998 Elsevier Science B.V. All rights reserved. Keywords: Developmental toxicity; Mutagenesis; SAR; Salmonella
1. Introduction In a recent report w1x we analyzed, using the expert SAR system CASErMULTICASE w2x, the features of an SAR Žstructure–activity relationship. model describing human developmental toxicity ŽHuDT.. The data for the model were derived from Briggs et al. w3x and TERIS w4x and, although based on human experience and on animal data, it is meant to describe the risk of developmental toxicity in humans. The data used to construct the model were primarily derived from oral administration of therapeutic agents. One of the features associated with the SAR models obtained with CASErMULTICASE, is ) Corresponding author. Tel. q1-412-967-6510; Fax: q1-412624-1289; E-mail: [email protected]
an ability to determine the structural overlaps between different SAR models w5x. The extent of overlap is an indication of mechanistic similarities. Thus, this approach can be used to document whether a complex phenomenon may be the consequence of more than one mechanism. For example, this enabled us to conclude that cell toxicity as well as mutagenicity contributed independently of one another, to carcinogenicity w6x. An examination of the above mentioned human developmental toxicity SAR model suggested that there was little structural overlap between that model and mutagenicity Žin Salmonella.. This was an unexpected finding as 20–30% of birth defects in humans have been attributed to germ cell and somatic mutations w7,8x. Moreover, even if germ cell mutagens induce developmental effects by nonmutagenic Ži.e., epigenetic. mechanisms w9x, there
0027-5107r98r$ - see front matter q 1998 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 7 - 5 1 0 7 Ž 9 8 . 0 0 1 7 0 - 5
348
H.S. Rosenkranz et al.r Mutation Research 422 (1998) 347–350
should still be some structural similarities. In view of these unexpected findings and their possible significance with respect to the causes of developmental effects in humans, we re-examined them. The present report confirms those earlier findings and we suggest a non-mechanistic basis for this lack of overlap.
2. Methods The CASErMULTICASE methodology as well as the rules used to determine structural overlaps have been described w10–13x. The SAR model of HuDT was described previously w1x. It is based upon a computerized version of TERIS w4,14x and a compilation in which FDA definitions were used to quantify risk of developmental toxicity from drug use during pregnancy w3,15x. The SAR models that were used for determination of structural overlaps have also been described previously w10–13x. The significance of the comparisons between the biophores associated with HuDT and those associated with other toxicological phenomena was determined based upon the assumption Žnull hypothesis. that mechanistically no overlap is expected between HuDT and the induction of a 2mnephropathy in male rats, binding to the Ah receptor, or inhibition of cytochrome P4502D.
3. Results The SAR model we used to investigate mechanistic relationships of human developmental effects ŽHuDT. consists of 130 developmental toxicants and 193 non-toxicants w1x. This SAR model shows a concordance of 74.8% between experimental results and predictions for chemicals external to the SAR model Žsensitivity 0.65, specificity 0.81.. This is an acceptable performance for an SAR model. Moreover, the biophores identified were statistically significant and therefore, could be used for the mechanistic studies contemplated w5,16x. By the criteria used to select significant CASE biophores, it was found that the HuDT model exhibited 40 such biophores. This number is significantly greater Ž p s 0.004. than the 20 significant CASE
biophores that were generated from an SAR model of Salmonella mutagenicity containing a similar number of chemicals Ži.e., n s 323. with the same ratio of active to inactive molecules. These findings suggest that the phenomenon responsible for the HuDT is much more complex than that associated with mutagenicity. This conforms with the notion that unlike mutagenicity, developmental toxicity is defined at the organismic rather than at the molecular level w8x. Analysis of the structural overlaps between the HuDT SAR model and SAR models of other toxicological endpoints indicated ŽTable 1. that HuDT overlapped with a wide spectrum of toxicological endpoints including acute toxicity Žcell toxicity, systemic toxicity, e.g., rat LD50 , maximum tolerated dose., chromosomal and genomic events Žsomatic mutationsrrecombinations in Drosophila, induction of sister chromatic exchanges, micronuclei, chromosomal aberrations, mutations in mouse lymphoma Table 1 Structural overlaps between the SAR model of developmental toxicity in humans and other SAR models SAR model
Percent overlap
p-Value
Rodent carcinogenicity—NTP Rodent carcinogenicity—CPDB Mutagenicity Salmonella Unscheduled DNA synthesis Mutagenicityrsomatic recombinations in Drosophila MTD—mouse MTD—rat Cell toxicity Rat lethality ŽLD50 . Sister chromatid exchanges in vitro Chromosomal aberrations in vitro Micronuclei in vivo Sister chromatid exchanges in vivo Mutations in mouse lymphoma cells Malsegregation in yeast a 2m-Nephropathy Žmale rats. a Inhibition cytochrome P4502D6 a Ah receptor binding a Inhibition tubulin polymerization
28 25 5 5 28
0.0009 0.002 0.3 0.3 0.0009
25 20 20 25 25
0.002 0.007 0.007 0.002 0.002
15 23 18 18
0.02 0.003 0.01 0.01
13 3 3 5 18
0.05 0.5 0.5 0.3 0.01
a
Based upon mechanistic considerations, these SAR models were assumed to bear no relationship to HuDT. For statistic analysis, overlaps with these data bases served to test the null hypothesis Žsee text..
H.S. Rosenkranz et al.r Mutation Research 422 (1998) 347–350
ŽMLA., including loss of heterozygosity ŽMLA.. and aneuploidy Žinduction of micronuclei, inhibition of tubulin polymerization, malsegregations.. There was also overlap between HuDT and carcinogenesis. Obviously, this overlap is not to be interpreted as indicating that carcinogenesis is a precursor of HuDT, but rather that some of the mechanisms contributing to carcinogenicity Že.g., DNA damage, mitogenesis. may also contribute to HuDT. The broad and significant overlaps with so many endpoints is not a usual feature of other phenomena studied with this approach. Frequently such analyses indicate mutually exclusive overlaps Že.g., carcinogenicity with either mutagenicity or cell toxicity w6x. but not joint ones, as is evident for HuDT ŽTable 1.. Given this broad spectrum of overlaps, the lack of significant overlap between HuDT and point mutations Žin Salmonella. is remarkable ŽTable 1. even while there were, as pointed out above, significant overlaps with other chromosomalrgenotoxic phenomena. The observation of a lack of overlap between the Salmonella mutagenicity and HuDT SAR models could be the result of the exclusion of chemicals mutagenic in Salmonella from consideration as therapeutics. In view of the widespread use of the Salmonella mutagenicity assay as a pre-screen to identify potential carcinogens, this may be a plausible explanation. In order to test this possibility, we compared the predicted prevalence of Salmonella mutagens among the universe of chemicals Ž37.2%. to the predicted prevalence among the chemicals forming the HuDT data base Ž20.1%.. The difference between the two predicted prevalences is highly significant Ž p s- 0.00001.. Thus, it is likely that the lack of structural overlap between HuDT and mutagenesis is a result of the pre-screening process used. Indeed, a hamster developmental toxicity SAR model based upon a population of molecules that presumably had not undergone such a pre-screen, showed a significant overlap Ž19.2% vs. 5.0%, p s 0.009. between that model and mutagenicity in Salmonella w17x. The present findings indicate that in fact the pre-screening strategies used are effective in preventing the introduction of mutagens or putative mutagens among therapeutics used in human medicine. The findings also indicate that while the HuDT SAR
349
model appears to be highly predictive w1x, it may lack specificity for agents capable of inducing point mutations Žwhich are usually detectable with the Salmonella mutagenicity assay.. Thus, in a strategy that uses the HuDT SAR model to prioritize chemicals for further testing, it might be appropriate to include the SAR model of Salmonella mutagenicity to identify potential mutagens which might be responsible for a proportion of human developmental defects w7,8x.
References w1x M. Ghanooni, D.R. Mattison, Y.P. Zhang, O.T. Macina, H.S. Rosenkranz, G. Klopman, Structural determinants associated with risk of human developmental toxicity, Am. J. Obstet. Gynecol. 176 Ž1997. 799–806. w2x G. Klopman, H.S. Rosenkranz, Prediction of carcinogenicityrmutagenicity using MULTICASE, Mutat. Res. 305 Ž1994. 33–46. w3x G.G. Briggs, R.K. Freeman, S.J. Yaffe, Drugs in Pregnancy and Lactation, 3rd ed., Williams and Wilkins, Baltimore 1990. w4x J.M. Friedman, B.B. Little, R.L. Brent, J.F. Cordero, J.W. Hanson, T.H. Shepard, Potential human teratogenicity of frequently prescribed drugs, Obstet. Gynecol. 75 Ž1990. 594–599. w5x M. Liu, N. Sussman, G. Klopman, H.S. Rosenkranz, Structure–activity and mechanistic relationships: the effects of chemical overlap on structural overlap in data bases of varying size and composition, Mutat. Res. 372 Ž1996. 79–85. w6x H.S. Rosenkranz, G. Klopman, Structural evidence for a dichotomy in rodent carcinogenesis: involvement of genetic and cellular toxicity, Mutat. Res. 303 Ž1993. 83–89. w7x E. Hodgson, P.E. Levi, A Textbook of Modern Toxicology, Elsevier, Amsterdam, 1979. w8x J.B. Bishop, K.L. Witt, R.A. Sloane, Genetic toxicities of human teratogens, Mutat. Res. 396 Ž1997. 9–43. w9x W.M. Generoso, A.G. Shourbaji, E.W. Piergorsh, J.B. Bishop, Developmental response of zygotes exposed to similar mutagens, Mutat. Res. 250 Ž1991. 439–446. w10x A. Labbauf, G. Klopman, H.S. Rosenkranz, Dichotomous relationship between DNA reactivity and the induction of sister chromatid exchanges in vivo and in vitro, Mutat. Res. 377 Ž1997. 37–52. w11x M. Rosenkranz, H.S. Rosenkranz, G. Klopman, Intercellular communication, tumor promotion and non-genotoxic carcinogenesis: relationships based upon structural considerations, Mutat. Res. 381 Ž1997. 171–188. w12x M. Liu, S.G. Grant, O.T. Macina, G. Klopman, H.S. Rosenkranz, Structural and mechanistic bases for the induction of mitotic chromosomal loss and duplication Ž‘malsegregation’. in the yeast Saccharomyces cereÕisiae: Relevance to
350
H.S. Rosenkranz et al.r Mutation Research 422 (1998) 347–350
human carcinogenesis and developmental toxicity, Mutat. Res. 374 Ž1997. 209–231. w13x A.R. Cunningham, H.S. Rosenkranz, Y.P. Zhang, G. Klopman, Identification of ‘genotoxic’ and ‘non-genotoxic’ alerts for cancer in mice: the carcinogenicity potency data base, Mutat. Res. 398 Ž1998. 1–17. w14x J.M. Friedman, J.E. Polifka, Teratogenic Effects of Drugs: A Resource for Clinicians ŽTERIS., The John Hopkins University Press, Baltimore, 1994.
w15x J.L. Schardein, Chemically Induced Birth Defects. Marcel Dekker, New York, 1993 . w16x B. Henry, S.G. Grant, G. Klopman, H.S. Rosenkranz, Induction of forward mutations at the thymidine kinase locus of mouse lymphoma cells: evidence for electrophilic and nonelectrophilic mechanisms, Mutat. Res. 397 Ž1998. 313–335. w17x J. Gomez, O.T. Macina, D.R. Mattison, Y.P. Zhang, G. ´ Klopman, H.S. Rosenkranz, Structural determinants of developmental toxicity in hamsters. in preparation Ž1998..