8
Ž . Mutation Research 431 1999 31–38 www.elsevier.comrlocatermolmut Community address: www.elsevier.comrlocatermutres SOS chromotest and mutagenicity in Salmonella: evidence for mechanistic differences Herbert S. Rosenkranz a, ) , Volker Mersch-Sundermann b , Gilles Klopman c a Department of EnÕironmental and Occupational Health, Graduate School of Public Health, UniÕersity of Pittsburgh, Pittsburgh, PA 15261, USA b Institut fur Medizinische Mikrobiologie und Hygiene, Faculty of Clinical Medicine, Heidelberg UniÕersity, 68135 Mannheim, Germany ¨ c Department of Chemistry, Case Western ReserÕe UniÕersity, CleÕeland, OH 44106, USA Received 16 March 1999; received in revised form 23 August 1999; accepted 15 September 1999 Abstract An examination of the relationship of the experimental results obtained with chemicals tested in the SOS chromotest and for mutagenicity in Salmonella indicates that the two assays respond to different genotoxic stimuli. Furthermore, the relationship between results obtained in these assays and in rodents carcinogenicity bioassays suggests that the short-term assays respond to a different spectrum of carcinogens. The same conclusions were reached based upon an analysis of the structural features associated with these three phenomena. With respect to using these short-term assays to predict carcinogens, the present results suggest that they are not equivalent, but complement one another. q 1999 Elsevier Science B.V. All rights reserved. Keywords: SOS chromotest; Salmonella mutagenicity; Carcinogenicity; Mechanism; SAR 1. Introduction The SOS chromotest has been proposed as a simpler and more rapid alternative to the Salmonella mutagenicity assay to identify potential genotoxic carcinogens. The rationale for this proposal is based upon the assumption that the two assays reflect a common mechanism of action. Yet, we know that, in fact, this is not entirely accurate, at best the SOS chromotest reflects one possible aspect of the muta- tional process: error-prone DNA repair. The muta- tional event scored by the Salmonella assay is a more complex phenomenon. Moreover, it must be ) Corresponding author. Fax: q1-412-624-3309. pointed out that even when mutations are induced by a mutagen, it is a relatively rare event while in the SOS chromotest it is assumed that the majority of the exposed cells respond. Given these considera- tions, we undertook a study of the spectrum of mechanistic responses associated with each of these assays. To accomplish this, we took advantage of the earlier recognition that the structural overlaps exhib- ited by SAR models of specific biological phenom- ena indicates the extent of the mechanistic similarity w x between them 1–4 . Additionally, we took advan- Ž tage of a recently developed method the ‘‘Chemical . wx Diversity Approach’’ 5 that is based upon an analysis of the expected and observed frequencies of chemicals jointly possessing multiple toxicological properties. 0027-5107r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. Ž . PII: S0027-5107 99 00155-4

SOS chromotest and mutagenicity in Salmonella: evidence for mechanistic differences

Embed Size (px)

Citation preview

Ž .Mutation Research 431 1999 31–38www.elsevier.comrlocatermolmut

Community address: www.elsevier.comrlocatermutres

SOS chromotest and mutagenicity in Salmonella: evidence formechanistic differences

Herbert S. Rosenkranz a,), Volker Mersch-Sundermann b, Gilles Klopman c

a Department of EnÕironmental and Occupational Health, Graduate School of Public Health, UniÕersity of Pittsburgh, Pittsburgh, PA15261, USA

b Institut fur Medizinische Mikrobiologie und Hygiene, Faculty of Clinical Medicine, Heidelberg UniÕersity, 68135 Mannheim, Germany¨c Department of Chemistry, Case Western ReserÕe UniÕersity, CleÕeland, OH 44106, USA

Received 16 March 1999; received in revised form 23 August 1999; accepted 15 September 1999

Abstract

An examination of the relationship of the experimental results obtained with chemicals tested in the SOS chromotest andfor mutagenicity in Salmonella indicates that the two assays respond to different genotoxic stimuli. Furthermore, therelationship between results obtained in these assays and in rodents carcinogenicity bioassays suggests that the short-termassays respond to a different spectrum of carcinogens. The same conclusions were reached based upon an analysis of thestructural features associated with these three phenomena.

With respect to using these short-term assays to predict carcinogens, the present results suggest that they are notequivalent, but complement one another. q 1999 Elsevier Science B.V. All rights reserved.

Keywords: SOS chromotest; Salmonella mutagenicity; Carcinogenicity; Mechanism; SAR

1. Introduction

The SOS chromotest has been proposed as asimpler and more rapid alternative to the Salmonellamutagenicity assay to identify potential genotoxiccarcinogens. The rationale for this proposal is basedupon the assumption that the two assays reflect acommon mechanism of action. Yet, we know that, infact, this is not entirely accurate, at best the SOSchromotest reflects one possible aspect of the muta-tional process: error-prone DNA repair. The muta-tional event scored by the Salmonella assay is amore complex phenomenon. Moreover, it must be

) Corresponding author. Fax: q1-412-624-3309.

pointed out that even when mutations are induced bya mutagen, it is a relatively rare event while in theSOS chromotest it is assumed that the majority ofthe exposed cells respond. Given these considera-tions, we undertook a study of the spectrum ofmechanistic responses associated with each of theseassays. To accomplish this, we took advantage of theearlier recognition that the structural overlaps exhib-ited by SAR models of specific biological phenom-ena indicates the extent of the mechanistic similarity

w xbetween them 1–4 . Additionally, we took advan-Žtage of a recently developed method the ‘‘Chemical

. w xDiversity Approach’’ 5 that is based upon ananalysis of the expected and observed frequencies ofchemicals jointly possessing multiple toxicologicalproperties.

0027-5107r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0027-5107 99 00155-4

( )H.S. Rosenkranz et al.rMutation Research 431 1999 31–3832

2. Materials and methods

2.1. CASE methodology

The CASE expert system provided the learning toolfor this study. This methodology has been described

w xon a number of occasions 6–9 . Briefly, CASE selectsits own descriptors automatically from a learning setcomposed of active and inactive chemicals. Thesedescriptors are imbedded in a significant number ofchemicals used to derive the model and consist ofreadily recognizable single and continuous structuralfragments, i.e., 2 to 10 connected heavy atoms, i.e.non-hydrogen atoms. The descriptors consist of ei-

Ž . Ž .ther activating biophore or inactivating biophobefragments. The CASE program provides a list of theactivating and inactivating fragments ordered accord-ing to statistical significance levels, i.e., the probabil-

Ž .ity values p-values associated with the distributionof the fragments among the active and the inactivechemicals. These fragments can be examined forstructural overlap comparisons between SAR mod-els. Once the learning set has been assimilated, CASE

can also be used to predict the activity of chemicalsof unknown activity.

The predictive performance of the SAR model forchemicals external to the model is determined by the

w x10-fold cross validation procedure 10 .Structural overlap was determined by comparing

the structural similarities among significant frag-ments obtained from CASE analysis of the data bases.In this study, the comparisons were between theSAR models of carcinogenicity in rodents, muta-genicity in Salmonella, and induction of the SOSchromotest.

Two fragments from different SAR models weredesignated ‘identical’, e.g. CH `CH ` vs.2 2

CH `CH `, or ‘imbedded’, e.g. CH `CH vs.2 2 2 2

CH `CH `CH `.2 2 2

The CASE program is available from MULTI-CASE; 25825 Science Park Drive, Beachwood, OH44122, USA.

2.2. Databases

Ž .The database Ns430 and SAR model of thew xSOS chromotest have been described previously 11 .

The Salmonella mutagenicity data base used in these

studies was generated by the US National Toxicol-w xogy Program 12–14 . A total of 2055 chemicals

were available for generating SAR models. As indi-cated in the text, subsets were selected for compari-

w xson with the SOS chromotest 15 . An SAR modelbased on a subset of the National Toxicology Pro-gram Salmonella mutagenicity data base has been

w xdescribed previously 16,17 .The biophores associated with the SOS chro-

motest were compared to biophores previously iden-tified in SAR models of other genotoxic and carcino-genic phenomena: the induction of chromosomal

Ž .aberrations and sister chromatid exchanges SCE inw x qrycultured CHO cells 18 , of mutations in the tk

w xlocus of cultured mouse lymphoma cells 19 , ofw x w xSCE in vivo 20 and of micronuclei in vivo 21 .

The SAR models of carcinogenicity in rodents werew xalso described previously 8,22,23 .

3. Determination of dependencies

While the determination of structural overlaps hasbeen used successfully to explore mechanistic rela-

w xtionships amongst toxicological phenomena 1–4 .One of the short comings of this approach derivesfrom the fact that most often SAR models are limitedto approximately 300 chemicals which may not berepresentative of the ‘‘universe of chemicals’’. Inorder to overcome this possible limitation, we have

w x Ždevised 5 a complementary procedure ‘‘the Chem-

Table 1Performance of SAR models for SOS chromotest and Salmonellamutagenicity

aSOS chromotest Salmonella mutagenicitybConcordance 87.3% 74.3%

bSensitivity 0.88 0.54bSpecificity 0.87 0.86

a The Salmonella mutagenicity model consists of a subset ofw xchemicals which was also tested in the SOS chromotest 15 .

b Indicates the concordance between experimental results andthe SAR-based predictions of the activity of chemicals not in-

Ž .cluded in the SAR model i.e., external validation . Sensitivity isdefined as correct positive predictionsrtotal number of positivechemicals. Specificity is defined as a correct negative predic-tionsrtotal negative chemicals. Concordance is defined as correctpredictionsrtotal number of chemicals.

( )H.S. Rosenkranz et al.rMutation Research 431 1999 31–38 33

Table 2Structural overlaps between SOS chromotest and other SAR modelsThe summary of these overlaps is shown in Table 3.Abbreviations: SAL, mutagenicity in Salmonella; CAN, rodent carcinogenicityrNTP; CPD, rodent carcinogenicityrCPDB; SCE, sisterchromatid exchanges in vitro; ChA, chromosomal aberrations in vitro, Mnt, in vivo induction of bone marrow micronuclei; SCV, in vivoinduction of SCE; MLA, induction of mutations at the tkqry locus in mouse lymphoma cells. X indicates a structural overlap between an

² :SOS Chromotest biophore and another data base. 2-NO indicates a nitro function attached to the second carbon atom from the left. C.2: Yindicates a carbon atom common to 2 ring systems. C indicates a double bond.

The fragment notations used are illustrated by the 23rd and 22nd fragment from the top. Thus for the 23rd fragment, the 2nd carbon atomfrom the left is shown with a hydrogen atom. That means that it cannot have another substituent at that position. On the other hand, for the

Ž .first carbon atom on the left, there is no substituent shown. That means that this atom can have any substituent e.g., Cl, O, amino, etc.² :except a hydrogen. For the 22nd fragment, the third atom from the left is shown unsubstituted, however, as indicated by 3-0 , it must be

substituted by an oxygen atom.

SOS chromotest biophore SAL CAN CPD SCE ChA Mnt SCV MLA

CO`CH- . X . . . . . XNO`N` X . X X . X . .

YC. `O`C- . . . . . X . .O`C-CH` X X X X X X X XO`CH`O` . . . . . . . .O`CO`CH- . X . . . X . XCO`N`CH ` . . . . . . . .2

YNO `C `O` X . . X . X . .2² :CH-C`C.- 2-NO X . . X . . . .2² :O`C-C` 2-NO X . . X . X . .2² :O`C-CH` 2-NO X . . X . X . .2² :NO `C-CH` 2-O X . . X . X . .2

CH`CH `O`C- . X . X . X . .2

C.-C`CH-CH` X . . . . . . .CH-CH`C.-C.` X . X . . . . .O`CH`O`CH- . . X X . . . XCO`O`CH `CH` . X . X . X . .2

YNO `C `O`C- X X X X . X . .2YNO `C `O`C.- X . . X . X . .2

² :O`C-CH`CH- 2-NO X X X X X X . .2² :CH-C.`C-CH` 3-NO X . . X . . . .2² :CH-CH`C-C.` 3-O . . X . X X . .

C-CH`CH-C.`C- . . . . . . . .C.-N`C.-CH`C- . . . . . . . .

YCH-CH`C. `CO`C.- . . . . . . X XO ^̀ CH`CH `O`C- X X . X X X X X2

NO `C-CH`C.-C.` X . . X . . . .2YNO `C `O`C.-CH` X . X X . X . .2

² :O`C-CH`CH-C` 2-NO X X X X X X X X2Y ² :CH `CH-C`O`C- 3-NO X X X X X X X X2

CH-C`CH-CH`C.-CH` . . X . . X X XO ^̀ CH `CH`CH `O`C- X X X X X X X X2 2

NO `C-CH`CH-C`O` X X X X X X X X2YNO `C `O`C.-CH`CH- X . X X . X . .2

NO-C.`CH-CH`CH . . . X . X . .-CH`C.-NO

YNO `C-CH`C. `CO`C. X . . X . . X X2

-CH`C-CH-CH`C.-CH`CH X . . X . . X X-C.`CH-C.`CH-

( )H.S. Rosenkranz et al.rMutation Research 431 1999 31–3834

.ical Diversity Approach’’ which consists of usingw xvalidated SAR models 24 to predict the activity of

10,000 chemicals representative of the ‘‘universe ofchemicals’’. This allows a comparison of the preva-lence of chemicals predicted to possess multipletoxicological activities with the expected prevalences

Žbased upon the independence of the activities i.e.,.null hypothesis . When the observed prevalence is

significantly greater than the expected prevalence,this may be taken as an indication of mechanisticsimilarities. A lack of significant difference betweenobserved and expected prevalences signifies inde-pendent mechanisms. A significant decrease in theobserved prevalence suggests that the phenomenaunder investigation are antagonistic to one another.

4. Results and discussion

It is of interest to note that a comparison ofchemicals for which both Salmonella mutagenicityand SOS chromotest results are available indicates

Žthat the ratios of activesrinactives differ greatly i.e.,.1.85 and 0.95, respectively thus suggesting that

twice as many chemicals respond in the former thanin the latter. Clearly the SOS error-prone repair assayŽ .chromotest and the Salmonella mutagenicity assaysdiffer in their responses to the same group of chemi-cals.

A number of in-depth SAR studies usingCASErMULTICASE have indicated that the predictiveperformance of an SAR model reflects the complex-ity of the biological phenomenon. More complexphenomena tend to exhibit decreased predictive per-

w xformances. 24 A comparison of the performancesof the SOS and Salmonella SAR models for thesame chemicals indicates that the SOS chromotest

Ž .model performs significantly better p-0.0001Ž .than the Salmonella mutagenicity model Table 1 .

Thus, by these criteria, the SOS chromotest may beconsidered to be less complex than the Salmonellamutagenicity assay. Indeed this is in accord with the

Ž .known mechanisms of these phenomena see above .Ž .As mentioned previously see above , the extent

of structural overlap between SAR models is anindication of the extent of mechanistic similarity.Thus comparison of the major biophores associatedwith the SOS chromotest with those of a series of

mutagenicity, genotoxicity and carcinogenicity mod-Ž .els reveals Tables 2 and 3 that the overlap with the

Salmonella mutagenicity model is 57%, while theoverlap with two carcinogenicity models is 29%–36%. In these respects the SOS chromotest differsfrom the Salmonella SAR model for which the over-lap with rodent carcinogenicity models is 56%–61%Ž .Table 3 . Similarly, there is significant difference inrelationship to the chromosomal aberration SARmodel, i.e., 19% and 39% overlap for SOS chro-motest and Salmonella mutagenicity, respectively.Clearly the responses of the SOS chromotest and theSalmonella assay differ, thereby further indicatingthat they differ mechanistically.

Since the major application of the SOS chro-motest is to identify potential carcinogens, we exam-ined the structural overlaps between the SAR modelfor carcinogenicity with the two short term assays. A

Ž .comparison reveals Tables 4 and 5 that while thecarcinogenicity SAR model overlaps significantlywith both the chromotest and the Salmonella muta-genicity models, the overlap was significantly greaterwith the Salmonella mutagenicity model. This con-firms the findings with the reciprocal comparisonsŽ .Table 3 .

Ž .However, it was also found Tables 4 and 5 thatthe overlaps between rodent carcinogenicity and themodels for these two phenomena was mutually ex-

Ž .clusive. Only two fragments 3.5% overlapped the

Table 3Structural overlaps between the SOS chromotest, Salmonella mu-tagenicity and other SAR modelsb

SAR model SOS Salmonellachromotest mutagenicity

SOS chromotest 100 38Salmonella mutagenicity 57 100

aŽ .Carcinogenicity NTP 29 61aŽ .Carcinogenicity CPDB 36 56

SCE in vitro 60 68Chromosomal aberrations 19 39Induction of micronuclei in vivo 55 52SCE in vivo 24 26

qryMutation at the tk locus 31 24of mouse lymphoma cells

a NTP refers to the rodent carcinogenicity data assembledw xunder the aegis of the US National Toxicology Program 26 while

w xCPDB refers to the Carcinogenic Potency Data Base 27–31 .b Based upon the fragments shown in Table 2.

( )H.S. Rosenkranz et al.rMutation Research 431 1999 31–38 35

Table 4Structural overlaps: carcinogenicity vs. SOS chromotest and Salmonella mutagenicity

aRodent carcinogenicity biophores SOS Salm

N`C- . XNH `C- . X2

O ^̀ CH ` X X2

Cl`CH ` . X2

Br`CH ` . X2

CO`CH- X .YC `O`C- X .

N`C-CH` . XNH `C-CH` . X2

O ^̀ CH`CH ` X .2

CH`CH `O` X X2

Cl`CH `CH` . X2

Br`CH `CH` . X2² :CH-C`C- 2-O X .

Y ² :CH `C-CH` 2-O X .YC `CH-CH`C- X .

O`C-CH`CH- X .CH `CH`CH `O` X .2 2

Br`CH `CH`CH ` . X2 2

NO `C-C`CH- . X2² :CH-CH`C-C`CH- 3-O X .

CH `C-CH`CH-C`C- . X3

CO`C.-CH`CH-C`CH . X3² :CH-CH`C-CH`C-C` 3-CH . X3² :C-CH`CH-C`CH-C` 4-CH . X3² :CH-CH`CH-C.`C-C`CH- 5-N- . X² :CH-CH`CH-C.`C-C`CH-CH` 6-OH . X

CH-C.`CH-CH`CH-CH`C.-C`C-CH` . XYN `C-C.`CH-CH`CH-CH`C.-CH`CH- . X

a Indicates biophores associated with the SAR model of rodent carcinogenicity based upon the NTP data base. X indicates a biophore² :common to two data bases. 2-0 indicates an oxygen atom attached to the 2nd carbon from the left. C. indicates a carbon atom common to

2 ring system. CY indicates a double bond. O ^ indicates an epoxide. For a description on how to interpret the fragments, see the legend ofTable 2.

three phenomena jointly. This suggests that withrespect to carcinogenicity, the two short term assaysreflect different mechanisms, i.e., they may comple-ment one another with respect to their ability to

w xidentify carcinogens 25 .

Table 5Structural overlaps between rodent carcinogenicity and SOS chro-motest or mutagenicityBased upon the overlaps shown in Table 4.

Ž .Overall %

SOS chromotest 19.0Salmonella mutagenicity 34.5Joint SOS and Salmonella 3.5

In fact, a comparison of the sensitivities andspecificities of the two assays with respect to car-cinogenicity indicates altogether different responses

Ž .to carcinogens and non-carcinogens Table 6 . Thus,based upon previous analyses, one would expect thatthe Salmonella mutagenicity assay was most reliable

Table 6Predictivity of SOS chromotest and Salmonella mutagenicity forcarcinogenicity in rodents

a aSensitivity Specificity

SOS chromotest 0.88 0.30Salmonella mutagens 0.55 0.76

a w xBased upon NTP rodent carcinogenicity data base 26 .

( )H.S. Rosenkranz et al.rMutation Research 431 1999 31–3836

when predicting carcinogens while the SOS chro-motest was most predictive for non-carcinogens.Moreover, a battery consisting of the two assayswould complement one another and would be more

w xpredictive than either one alone 25 .While a number of earlier studies have shown that

the mechanistic implications derived from structuralconsiderations were sound and led to testable hy-potheses, we recently developed an independent pro-

Ž .cedure the ‘‘Chemical Diversity Approach’’ forw xconfirming such findings 5 . The method consists of

randomly selecting 10,000 chemicals to represent the‘‘universe of chemicals’’. Using previously validatedSAR models, the biologicalrtoxicological activitiesof these chemicals are then determined. The propor-tion of chemicals that is predicted to display morethan one activity is then evaluated. The data so

Ž .obtained indicate Table 7 that 51% of the putativeSalmonella mutagens are also predicted to be

Ž .carcinogenic 100=1612r3148 while 76% pre-dicted SOS DNA-repair inducers are also predicted

Ž .to be carcinogenic 100=936r1235 . This trend isin accord with the experimentally determined sensi-tivities of the SOS chromotest and the Salmonellamutagenicity assay vis-a-vis carcinogenicity in ro-

Ž .dents Table 6 .ŽThe differences between predicted i.e., ‘‘ob-

.served’’ joint activities and those expected basedŽupon the occurrence of a single activity i.e., null

.hypothesis is an indication of dependence among

Ž .tests i.e., mechanistic similarity . In that context, anŽR value i.e., observed prevalencerexpected preva-. Žlence ratio of unity indicates independence i.e.,

.lack of mechanistic similarity . In fact, the R valuefor chemicals that are putative mutagens and induc-ers of SOS-repair is 2.4, which indicates that the twophenomena are related mechanistically. This con-

Ž .firms the structural overlap 57%, Table 3 that wasfound. These analyses also found that there were, as

Ž .expected from the experimental results Table 6 andŽ .the structural overlap Table 3 , indications of mech-

anistic similarities between rodent carcinogenicityŽ .and induction of SOS repair i.e., Rs2.3 and

between carcinogenicity and the induction of muta-Ž .tions in Salmonella Rs1.5 . However, the most

significant effect was seen for the prevalence of thegroup of chemicals which were predicted to be car-

Žcinogens, mutagens and inducers of SOS-repair R.s5.7 suggesting that a battery of the two assays

would be most predictive of rodent carcinogenicitythan either one alone. This, in fact, confirms theearlier conclusions based upon structural overlapsŽ . Ž .Table 3 and the experimental results Table 6 . Anidentical ranking of the mechanistic similarities wasobtained when the analysis was based upon thedifference between observed and predicted preva-

Ž .lences Table 6, last column .These analyses reveal that the SOS chromotest

and the Salmonella mutagenicity differ from oneanother mechanistically. Moreover, with respect to

Table 7Predictions of the prevalence of chemicals possessing multiple toxicological propertiesRsobserved prevalencerexpected prevalence.

cToxicological properties Observed Expected R P-value 100D rexpecteda bprevalence prevalence prevalence

Rodent Carcinogens 3346Salmonella Mutagens 3148Inducers SOS Chromotest 1235Salmonella and Chromotest 935 389 2.4 -0.0001 140Carcinogen and Salmonella 1612 1053 1.5 -0.0001 53MutagensCarcinogen and Chromotest 936 413 2.3 -0.0001 127Carcinogen, Mutagens and 746 130 5.7 -0.0001 474Chromotest

a The prevalence of chemicals predicted to possess one or more toxicological properties among 10,000 substances representing the‘‘universe of chemicals’’.

b ŽPrevalence of chemicals expected to possess two or more toxicological properties based upon an assumption of independence i.e., null.hypothesis among 10,000 chemicals representing the ‘‘universe of chemicals’’.

c Ž .100 Observed PrevalenceyExpected Prevalence rExpected Prevalence.

( )H.S. Rosenkranz et al.rMutation Research 431 1999 31–38 37

carcinogenicity they appear to detect different mech-anistic responses and thus to complement one an-other.

Previous MULTICASE studies of the SOS chro-motest indicated that, based upon similarities amongbiophores and global quantum chemical parameters,the SOS chromotest and the Salmonella mutagenic-ity assay appeared to operate through overlappingmechanisms and, therefore, were expected to re-

w xspond to the same type of chemical structures 11 .However, in the present analyses using more refinedmethodologies, we find that the SOS chromotest andthe Salmonella mutagenicity differ from one anothermechanistically. Moreover, with respect to carcino-genicity, they appear to detect different mechanisticresponses and thus to complement one another. Thisinformation is undoubtedly of relevance in the selec-tion of predictive short-term assays for carcinogenic-ity.

Acknowledgements

MULTICASE was founded and is partially ownedby Case Western Reserve University, Gilles Klop-man and Herbert S. Rosenkranz.

References

w x1 H.S. Rosenkranz, G. Klopman, Structural evidence for adichotomy in rodent carcinogenesis: involvement of genetic

Ž .and cellular toxicity, Mutat. Res. 303 1993 83–89.w x2 H.S. Rosenkranz, Y.P. Zhang, G. Klopman, Evidence that

cell toxicity may contribute to the genotoxic response, Reg.Ž .Toxicol. Pharmacol. 19 1994 176–182.

w x3 E. ter Haar, B.W. Day, H.S. Rosenkranz, Direct tubulinpolymerization perturbation contributes significantly to the

Ž .induction of micronuclei in vivo, Mutat. Res. 350 1996331–337.

w x4 M. Liu, N. Sussman, G. Klopman, H.S. Rosenkranz, Struc-ture–activity and mechanistic relationships: the effects ofchemical overlap on structural overlap in data bases of

Ž .varying size and composition, Mutat. Res. 372 1996 79–85.w x5 N. Pollack, A.R. Cunningham, H.S. Rosenkranz, Chemical

diversity approach for evaluating mechanistic relatednessamong toxicological phenomena, SAR.QSAR Environ. Res.Ž .1999 in press.

w x6 G. Klopman, Artificial intelligence approach to structure–ac-tivity studies. Computer automated structure evaluation of

biological activity of organic molecules, J. Am. Chem. Soc.Ž .106 1984 7315–7321.

w x7 G. Klopman, MULTICASE 1. A hierarchical computer auto-mated structure evaluation program, Quant. Struct. Act. Rel.

Ž .11 1992 176–184.w x8 H.S. Rosenkranz, G. Klopman, Structural basis of carcino-

genicity in rodents of genotoxicants and non-genotoxicants,Ž .Mutat. Res. 228 1990 105–124.

w x9 H.S. Rosenkranz, Structure–activity relationships for car-cinogens with differing modes of action, in: Vainio, Magee,

Ž .McGregor, McMichael Eds. , Mechanisms of carcinogenesisin risk identification 116 International Agency for Researchon Cancer, Lyon, 1992, pp. 271–277.

w x10 Y.P. Zhang, N. Sussman, G. Klopman, H.S. Rosenkranz,Development of methods to ascertain the predictivity andconsistency of SAR models: application to the U.S. NationalToxicology Program rodent carcinogenicity bioassays, Quant.

Ž .Struct.-Act. Relat. 16 1997 290–295.w x11 V. Mersch-Sundermann, G. Klopman, H.S. Rosenkranz,

Chemical structure and genotoxicity: studies of the SOSŽ .chromotest, Mutat. Res. 340 1996 81–91.

w x12 E. Zeiger, Carcinogenicity of mutagens: predictive capabilityof the Salmonella mutagenesis assay for rodent carcinogenic-

Ž .ity, Cancer Res. 47 1987 1287–1296.w x13 E. Zeiger, B. Anderson, S. Haworth, T. Lawlor, K. Mortel-

mans, W. Speck, Salmonella mutagenicity tests: III. ResultsŽ . Ž .from 255 chemicals, Environ. Mutagen. 9 Suppl. 9 1987

1–109.w x14 E. Zeiger, B. Anderson, S. Haworth, T. Lawlor, K. Mortel-

mans, Salmonella mutagenicity tests: IV. Results from theŽtesting of 300 chemicals, Environ. Mol. Mutagen. 11 Suppl.

. Ž .12 1988 1–157.w x15 V. Mersch-Sundermann, U. Schneider, G. Klopman, H.S.

Rosenkranz, SOS-induction in E. coli and Salmonella muta-genicity: a comparison using 330 compounds, Mutagenesis 9Ž .1994 205–224.

w x16 M. Liu, N. Sussman, G. Klopman, H.S. Rosenkranz, Estima-tion of the optimal data base size for structure-activity analy-ses: the Salmonella mutagenicity data base, Mutat. Res. 358Ž .1996 63–72.

w x17 H.S. Rosenkranz, G. Klopman, The structural basis of themutagenicity of chemicals in Salmonella typhimurium: theNational Toxicology Program database, Mutat. Res. 228Ž .1990 51–80.

w x18 H.S. Rosenkranz, F.K. Ennever, M. Dimayuga, G. Klopman,Significant differences in the structural basis of the inductionof sister chromatid exchanges and chromosomal aberrationsin Chinese hamster ovary cells, Environ. Mol. Mutagen. 16Ž .1990 149–177.

w x19 G. Henry, S.G. Grant, G. Klopman, H.S. Rosenkranz, Induc-tion of forward mutations at the thymidine kinase locus ofmouse lymphoma cells: evidence for electrophilic and non-

Ž .electrophilic mechanisms, Mutat. Res. 397 1998 313–335.w x20 A. Labbauf, G. Klopman, H.S. Rosenkranz, Dichotomous

relationship between DNA reactivity and the induction ofsister chromatid exchanges in vivo and in vitro, Mutat. Res.

Ž .377 1997 37–52.

( )H.S. Rosenkranz et al.rMutation Research 431 1999 31–3838

w x21 W.-L. Yang, G. Klopman, H.S. Rosenkranz, Structural basisof the in vivo induction of micronuclei, Mutat. Res. 272Ž .1992 111–124.

w x22 A.R. Cunningham, H.S. Rosenkranz, Y.P. Zhang, G. Klop-man, Identification of ‘‘genotoxic’’ and ‘‘non-genotoxic’’alerts for cancer in mice: the carcinogenicity potency

Ž .database, Mutat. Res. 398 1998 1–17.w x23 A.R. Cunningham, G. Klopman, H.S. Rosenkranz, Identifica-

tion of structural features and associated mechanisms ofŽ .action for carcinogens in rats, Mutat. Res. 405 1998 9–28.

w x24 H.S. Rosenkranz, A.R. Cunningham, Y.P. Zhang, H.G. Clay-camp, O.T. Macina, N.B. Sussman, S.G. Grant, G. Klopman,Development, characterization and application of predictive-

Ž .toxicology models, SAR.QSAR Environ. Res. 10 1999277–298.

w x25 V. Chankong, Y.Y. Haimes, H.S. Rosenkranz, J. Pet-Ed-wards, The carcinogenicity prediction and battery selectionŽ . Ž .CPBS method: a Bayesian approach, Mutat. Res. 153 1985135–166.

w x26 J. Ashby, R.W. Tennant, Definitive relationships amongchemical structure, carcinogenicity and mutagenicity for 301

Ž .chemicals tested by the U.S. NTP, Mutat. Res. 257 1991229–306.

w x27 L.S. Gold, C.B. Sawyer, R. Magaw, G.M. Backman, M. deVeciana, R. Levinson, N.K. Hooper, W.R. Havender, L.Bernstein, R. Peto, M.C. Pike, B.N. Ames, A carcinogenicpotency database of the standardized results of animal bioas-

Ž .says, Environ. Health Perspect. 58 1984 9–319.

w x28 L.S. Gold, M. de Veciana, G.M. Backman, R. Magaw, P.Lopipero, M. Smith, M. Blumenthal, R. Levinson, L. Bern-stein, B.N. Ames, Chronological supplement to the Carcino-genic Potency Database: standardized results of animal bioas-says published through December 1982, Environ. Health

Ž .Perspect. 67 1986 161–200.w x29 L.S. Gold, T.H. Slone, G.M. Backman, R. Magaw, M. Da

Costa, P. Lopipero, M. Blumenthal, B.N. Ames, Secondchronological supplement to the carcinogenic potencydatabase: standardized results of animal bioassays publishedthrough December 1984 and by the National ToxicologyProgram through May 1986, Environ. Health Perspect. 74Ž .1987 237–329.

w x30 L.S. Gold, T.H. Slone, G.M. Backman, S. Eisenberg, M. DaCosta, M. Wong, N.B. Manley, L. Rohrbach, B.N. Ames,Third chronological supplement to the Carcinogenic PotencyDatabase: standardized results of animal bioassays publishedthrough December 1986 and by the National ToxicologyProgram through June 1987, Environ. Health Perspect. 84Ž .1990 215–286.

w x31 L.S. Gold, N.B. Manley, T.H. Slone, G.B. Garfinkel, L.Rohrbach, B.N. Ames, The fifth plot of the CarcinogenicPotency Database: results of animal bioassays published inthe general literature through 1988 and by the NationalToxicology Program through 1989, Environ. Health Perspect.

Ž .100 1993 65–135.