QSAR Modeling and Prediction of the Endocrine-Disrupting Potencies of Brominated Flame Retardants Ester Papa,* Simona Kovarich, and Paola Gramatica QSAR Research Unit in EnVironmental Chemistry and Ecotoxicology, DBSF, UniVersity of Insubria, Via J.H. Dunant 3, 21100 Varese, Italy ReceiVed February 2, 2010 In the European Union REACH regulation, the chemicals with particularly harmful behaviors, such as endocrine disruptors (EDs), are subject to authorization, and the identification of safer alternatives to these chemicals is required. In this context, the use of quantitative structure-activity relationships (QSAR) becomes particularly useful to fill the data gap due to the very small number of experimental data available to characterize the environmental and toxicological profiles of new and emerging pollutants with ED behavior such as brominated flame retardants (BFRs). In this study, different QSAR models were developed on different responses of endocrine disruption measured for several BFRs. The multiple linear regression approach was applied to a variety of theoretical molecular descriptors, and the best models, which were identified from all of the possible combinations of the structural variables, were internally validated for their performance using the leave-one-out (Q LOO 2 ) 73-91%) procedure and scrambling of the responses. External validation was provided, when possible, by splitting the data sets in training and test sets (range of Q EXT 2 ) 76-90%), which confirmed the predictive ability of the proposed equations. These models, which were developed according to the principles defined by the Organization for Economic Co-operation and Development to improve the regulatory acceptance of QSARs, represent a simple tool for the screening and characterization of BFRs. Introduction In the past decade, brominated flame retardants (BFRs) 1 have been recognized as an emerging class of organic pollutants. BFRs are incorporated into inflammable polymers to increase the resistance to fire of a variety of consumer products, such as electronic devices, building materials, and textiles. Among the large number of brominated flame retardant compounds, the three most marketed high production volume (HPV) products are tetrabromobisphenol A (TBBPA), hexabromocyclododecane (HBCD), and three commercial formulations of polybrominated diphenyl ethers (PBDEs) (penta-, octa-, and deca-BDE). The widespread production and use of BFRs in the last 40 years has caused them to disperse into the environment. These substances have been found in different organisms of both aquatic and terrestrial ecosystems, and the high concentrations measured at the top levels of different food chains (e.g., birds, whales, polar bears, and human breast milk) have been related to bioaccumulation processes (1–3). Therefore, the ubiquitous presence of BFRs in all environmental compartments (4–6), caused by their high lipophilicity, bioaccumulation, and per- sistence properties, has gradually increased the concerns regard- ing health risks to man and the environment (7, 8). Moreover, the structural similarity of BFRs to other classes of organohalogenated compounds, such as polychlorinated biphenyls (PCBs) and dioxins, has raised concerns about their potential endocrine disruption activity. An endocrine-disrupting (ED) chemical has been defined as “an exogenous agent which interferes with the synthesis, secretion, transport, binding, action, or elimination of natural hormones in the body which are responsible for the maintenance of homeostasis, reproduction, development or behaviour” (9). In recent years, many studies, both in vivo and in vitro, have been conducted to investigate the endocrine disruption activity of BFRs. Experimental evidence has shown that some BFRs may interfere, as agonists and/or antagonists, with steroid receptors (estrogens, androgens, etc.) and interact with the aryl hydrocarbon receptor, AhR (“dioxin-like-activity”) (10–14). The strongest ED activity of BFRs was represented by the effects on the thyroid hormone (TH) system, where thyroid hyperplasia was induced and the TH transport and metabolism was altered (15, 16). These effects could be explained by the structural resemblance of BFRs to the TH thyroxin (T4), especially of those BFRs containing an -OH group, like hydroxylated metabolites of PBDE (OH-BDEs). Several risk assessments have * To whom correspondence should be addressed. E-mail: ester.papa@ uninsubria.it. 1 Abbreviations: BFRs, brominated flame retardants; HPV, high produc- tion volume; TBBPA, tetrabromobisphenol A; HBCD, hexabromocy- clododecane; PBDEs, polybrominated diphenyl ethers; PCBs, polychlori- nated biphenyls; AhR, aryl hydrocarbon receptor; ED, endocrine disrupting; TH, thyroid hormone; T4, thyroxin; OH-BDE, hydroxylated PBDE; REACH, Registration, Evaluation, Authorisation and Restriction of Chemi- cals; CMR, carcinogenic, mutagenic and toxic to reproduction; PBT, persistent, bioaccumulative, and toxic; QSAR, quantitative structure-activity relationships; CH 3 O-BDE, methoxylated PBDE; OECD, Organization for Economic Cooperation and Development; TBBPA-DBPE, tetrabromo- bisphenol-A-bis(2,3)dibromopropyl ether; 246-TBP, 2,4,6-tribromophenol; EC 50 , median effective concentration; IC 50 , median inhibition concentration; RBA, AhR relative binding affinity; TCDD, 2,3,7,8-tetrachloro-dibenzo- p-dioxin; EROD ind , ethoxyresorufin-O-deethylase induction potency; DR ag , AhR agonism; ER ag , estrogen receptor agonism; PR ant , progesterone receptor antagonism; T4-TTR comp , T4-transthyretin competition; T4 REP , T4-TTR relative competition; E2SULT inh , estradiolsulfotransferase inhibition; E2SULT REP , estradiolsulfotransferase relative inhibition; PCP, pentachlo- rophenol; DBDE, decabromodiphenylethane; EBTPI, ethylenebistetrabromo phthalimide; TBE, bis(tribromophenoxy)ethane; OLS, ordinary least-squares; Q LOO 2 , Q 2 leave-one-out; R 2 , coefficient of determination; Q BOOT 2 , Q 2 bootstrap; R/Q YS 2 , R/Q 2 scrambled; Q EXT 2 , Q 2 external; RMSE T/P , root-mean- square of errors for training/prediction sets; AD, applicability domain; HBB, hexabromobenzene; PLS, partial least-squares; SVM, support vector machine. Chem. Res. Toxicol. 2010, 23, 946–954 946 10.1021/tx1000392  2010 American Chemical Society Published on Web 04/21/2010