Chem. Res. Toxicol. 2007, 20, 1359–1363
1359
Verification of the Structural Alerts for Michael Acceptors T. Wayne Schultz,† Jason W. Yarbrough,† Robert S. Hunter,‡ and Aynur O. Aptula*,§ Department of ComparatiVe Medicine, College of Veterinary Medicine, The UniVersity of Tennessee, 2407 RiVer DriVe, KnoxVille, Tennessee 37996-4543, Hunter Systems, 1030 Opal Street, Unit 203, Broomfield, Colorado 80020-7300, and SEAC, UnileVer Colworth, Sharnbrook, Bedford, MK44 1LQ, England ReceiVed June 10, 2007
A diverse series of polarized R,β-unsaturated and related compounds were evaluated for reactivity with a spectrophotometric assay using the sulfhydryl group in the form of the cysteine residue of the tripeptide GSH as a model nucleophile. The reactive end point (RC50) calculations were compared to previously described structural alerts based on conventional organic chemistry. This comparison focused on polarized R,β-unsaturates, including ones containing an aldehyde, ketone, ester, sulfoxide, sulfone, sulfonate, nitro, or cyano moiety as well as ortho- and para-pyridino compounds and ortho- and paraquinones. The alerts were coded by substructure and are available in open-source software (http:// sourceforge.net/projects/chemeval). Comparisons of reactivity between selected analogues revealed that only the polarized R,β-unsaturates were reactive. These results verified the coded structural alerts that define the applicability domain for Michael acceptor electrophiles. Introduction One outcome of the Knoxville Workshop (The University of Tennessee, May 2005) on Reactive Toxicity was the development of a conceptual framework for predicting the toxicity of reactive chemicals with an emphasis on modeling soft electrophilicity (1). The covalent binding of chemicals with critical cellular targets such as proteins is an organizing principle underlying the concept of molecular initiating events for reactive toxicity. Moreover, the mechanistic ability of chemicals to initiate a toxicological event provides a basis to group chemicals into like categories, as exposure to substances with the same mechanism of action has the potential to lead to the same biological effects (e.g., protein denaturation, immunological response, etc.). From this framework arises the need to develop a comprehensive strategy to account for the chemical mechanisms of reactivity (e.g., Michael addition) that lead to the molecular initiating event (e.g., protein binding) that is central to linking organic materials to selected hazards of interest (e.g., skin sensitization). Reactive toxicity (the irreversible interaction of a xenobiotic chemical with endogenous molecules including specific sites on proteins) has been identified as the major gap in our ability to accurately predict key hazards such as skin sensitization (1). It stands to reason that the ability to classify chemicals as reactive or nonreactive to particular nucleophiles will help to segregate sensitizers from nonsensitizers. Among the molecular mechanisms that lead to the molecular initiating event of protein alteration is Michael type addition, which results in the formation of covalent adducts at a soft electro(nucleo)philic center without expulsion of a leaving group in the molecule. Electrophiles acting in this manner are typically organic materials that contain olefinic π-bonds polarized by a neighboring electron-withdrawing substituent (2). Electro(nu* To whom correspondence should be addressed. Tel: +44/1234-264823. Fax: +44/1234-264722. E-mail:
[email protected]. † The University of Tennessee. ‡ Hunter Systems. § SEAC.
cleo)philic interactions of Michael acceptors include the addition of an SH group to the β–carbon atom (β-C atom)1 of a carbon–carbon double (CdC) or carbon–carbon triple (CtC) bond; the decisive molecular substructure, a polarized R,βunsaturated configuration, results in a relatively diffuse and polarizable electron density of the olefinic π-bond, making Michael acceptors among the softest electrophiles (3). The tactical approach used in verifying the structural alerts for identification of chemicals able to undergo Michael addition had two major components, each of which was developed concurrently and refined in an iterative process. The first component is to develop a computer-assisted program based on general conventions of organic chemical reactions to identify reactive toxicants (4) and develop a series of substructure rules that are specific to the Michael acceptor mechanism of chemical reactivity. While it is recognized that limitations exist, it is generally agreed that this is an appropriate starting point as these reactions have been linked to toxicants binding to cellular targets, an important molecular initiating event in toxicology (1). These rules of chemical reactivity have long been known to organic chemists (3, 5); recently, they have been revived, refined, and in some cases expanded (4, 6). More recent efforts (7) have applied them to the mouse local lymph node assay data for skin sensitization in an effort to cluster compounds by the most likely molecular mechanism of reactivity. The second component of this approach is to use in chemico assays to empirically quantify relative reactivity and develop a mechanism-based database (8). In chemico testing is not unique. A cursory examination of the literature reveals that several protocols are available for these types of analyses, including those described by Esterbauer et al. (9), Clarke et al. (10), and Gerberick et al. (11). However, this particular protocol is simple, 1 Abbreviations: RC50, 50% reactive concentration; β–C atom, β–carbonatom; CdC, carbon–carbon double; CtC, carbon–carbon triple; SH, sulfhydryl group; CEF, chemical evaluation framework; SMILES, simplified molecular input line entry system; SMARTS, chemical substructure pattern linear notation parser; DTNB, 5,5′-dithio-bis(2-nitrobenzoic acid); DMSO, dimethyl sulfoxide; SAS, statistical analysis system; CAS no., Chemical Abstracts Services registry number.
10.1021/tx700212u CCC: $37.00 2007 American Chemical Society Published on Web 08/03/2007
1360 Chem. Res. Toxicol., Vol. 20, No. 9, 2007
Schultz et al.
Figure 1. Michael acceptor-related compounds considered in this paper.
rapid, inexpensive, and quantitative (8). While specific protein targets in the cell provide multiple potential targets (12, 13), the model nucleophile cysteine, especially in the form of the tripeptide L-γ-glutamyl-L-cysteinyl-glycine or GSH, is particularly advantageous. Its biological relevance lies with the fact that the cysteinyl moiety provides the critical sulfhydryl group (SH group) that is a key functional element facilitating transport and elimination of many reactive chemicals via conjugation (14). In deciding the nature of a chemical hazard, whether for forming chemical categories or other uses such as risk assessment, several principles are pertinent as follows: The interaction of the toxicant and molecular target must obey the laws of chemistry, the laws of chemistry determine the molecular mechanism of the reaction, and the molecular mechanism of reaction determines the biological mechanism of toxicity and the molecular initiating event. It follows, then, that determining
the most likely molecular mechanism of action by defining the structural limits of the domain of electro(nucleo)philic mechanisms becomes a critical step in predictive toxicology. The effectiveness of the two-prong method used here is that the chemistry, both theoretical and experimental, is driving the process, thus enabling a stringent definition of applicability domains that is crucial to correctly classifying reactive toxicants and better predictions of toxicity. The purpose of this study was to quantify the chemical reactivity between selected compounds and GSH and to compare this reactivity with the predictions based on the structural rules for Michael acceptors previously reported by Aptula and Roberts (2). The specific aims were to (i) code the two-dimensional (2D) structural alerts for Michael acceptors into a simple computeraided system and use this in silico system to predict the ability of selected chemicals to react with GSH via Michael addition
Rules for Michael Acceptors
Chem. Res. Toxicol., Vol. 20, No. 9, 2007 1361 Table 1. Structural Alerts for Michael Acceptors
rule
substructure
message
1 2 3 4A 4B 5 6 7 8 9A 9B 10A 10B 11 12A 12B 13A 13B 14
CtCCdO CdCCdO [CH2]dC(C)CdO OdCCdCdO [CH]dC(C(dO))CdO CdCN(dO)dO CtCS(dO) CdCSdO CdCCtN CtCc1ncccc1 Cd[CH]c1ncccc1 CtCc1ccncc1 Cd[CH]c1ccncc1 Cd[CH]C(dO)[OX1] OdC1CdC[CH]dCC1dO OdC1cc[CH]dCC1dO OdC1[CH]dCC(dO)CdC1 OdC1[CH]dCC(dO)cc1 [CH3]d[CH][CH]dO
ethynylene or acetylenic with a carbonyl vinyl or vinylene with a carbonyl R-C atom alkyl-substituted with a carbonyl R-C atom substituted with a second carbonyl R-C atom substituted with a second carbonyl olefinic nitro ethynylene or acetylenic with a SdO group vinyl or vinylene with a SdO group olefinic cyano ortho-ethynylene azaarene ortho-vinyl azaarene para-ethynylene azaarene para-vinyl azaarene vinylene carboxylic acid ortho-quinone ortho-quinone para-quinone para-quinone acrolein (2-propenal)
and (ii) experimentally determine the reactivity of these selected chemicals with GSH in an effort to verify the structural alert rules. We reasoned that substances possessing the structural alerts for Michael addition would be experimentally reactive, while those lacking the alerts would be nonreactive.
Methods and Materials Structural rules for Michael acceptors were extracted from the scheme of Aptula and Roberts (2). The identifying structural characteristic or structural alert was a CdC or CtC group with a neighboring electron-withdrawing group. Such polarized R,βunsaturates include those containing one of the following as the electron-withdrawing moiety: an aldehyde, ketone, ester, sulfoxide, sulfone, sulfonate, nitro, or cyano moiety as well as ortho- and para-pyridino compounds (olefinic six-member, nitrogen-substituted aromatics). Additionally, ortho- and para-quinones (olefinic diketones) were noted to be Michael acceptors (2). The structural rules have been encoded in software that is available for general dissemination in the Open Source project chemical evaluation framework (CEF) accessible at the SourceForge.net Web site (http://sourceforge.net/projects/chemeval). Briefly, the software system uses simplified molecular input line entry system (SMILES) to encode chemicals for evaluation. The substructures are encoded in SMARTS—chemical substructure pattern linear notation parser. The software uses Open Source libraries to support the basic functionality. To this end, the system is limited to the functionality supported in these libraries (JOELIB2 and CDK), which are also available from SourceForge.net (http:// sourceforge.net/projects/joelib and http://sourceforge.net/projects/ cdk, respectively). All software and libraries are coded in the Java programming language. On the basis of these rules, 27 test chemicals were selected (Figure 1). Eighteen of these compounds contained one of the above-mentioned polarized R,β-unsaturated configurations. The remaining nine contained either a polar group, an unsaturated group, or both groups in a configuration other than an R,β-one. All test substances were purchased from commercial sources (SigmaAldrich.com or Alfa.com) in the highest purity available (95% minimum) and were not further purified prior to testing. A simple and rapid, spectrophotometric-based, concentrationresponse assay for reactivity with the SH group of GSH was conducted as outlined by Schultz et al. (8). With absorbance at 412 nm, the free thiol group was quantified by its reaction with 5,5′-dithio-bis(2-nitrobenzoic acid) (DTNB). Briefly, 1.375 mM GSH was freshly prepared by dissolving 0.042 g of reduced GSH into 100 mL of phosphate buffer at pH 7.4. Stock solutions were prepared by dissolving them in dimethyl sulfoxide (DMSO); subsequently, phosphate buffer was added to the toxicant/DMSO solution. The proper aliquots of GSH solution, toxicant stock
solution, and buffer were added to bring the final concentration of thiol to 0.1375 mM and, at the same time, assuring that the concentration of DMSO in the final solution was always
