1752
Chem. Res. Toxicol. 2007, 20, 1752–1759
Article Toxicological Evaluation of Thiol-Reactive Compounds Identified Using a La Assay To Detect Reactive Molecules by Nuclear Magnetic Resonance Jeffrey R. Huth, Danying Song, Renaldo R. Mendoza, Candice L. Black-Schaefer, Jamey C. Mack, Sarah A. Dorwin, Uri S. Ladror, Jean M. Severin, Karl A. Walter, Diane M. Bartley, and Philip J. Hajduk* Global Pharmaceutical Research and DeVelopment, Abbott Laboratories, Deparment R460, Building AP10, 100 Abbott Park Road, Abbott Park, Illinois 60064 ReceiVed September 6, 2007
We have recently reported on the development of a La assay to detect reactive molecules by nuclear magnetic resonance (ALARM NMR) to detect reactive false positive hits from high-throughput screening, in which we observed a surprisingly large number of compounds that can oxidize or form covalent adducts with protein thiols groups. In the vast majority of these cases, the covalent interactions are largely nonspecific (e.g., affect many protein targets) and therefore unsuitable for drug development. However, certain thiol-reactive species do appear to inhibit the target of interest in a specific manner. The question then arises as to the potential toxicology risks of developing a drug that can react with protein thiol groups. Here, we report on the evaluation of a large set of ALARM-reactive and -nonreactive compounds against a panel of additional proteins (aldehyde dehydrogenase, superoxide dismutase, and three cytochrome P450 enzymes). It was observed that ALARM-reactive compounds have significantly increased risks of interacting with one or more of these enzymes in vitro. Thus, ALARM NMR seems to be a sensitive tool to rapidly identify compounds with an enhanced risk of producing side effects in humans, including alcohol intolerance, the formation of reactive oxygen species, and drug-drug interactions. In conjunction with other toxicology assays, ALARM NMR should be a valuable tool for prioritizing compounds for lead optimization and animal testing. Introduction The development of new tools for predicting human drug toxicities is critical to the continued success of drug discovery and development. In fact, drug toxicity is the primary cause of clinical failure (1). With the escalating scientific challenges associated with designing drugs for new targets, there is a real need to enhance the rate at which compounds are evaluated and, in the majority of cases, subsequently discarded (2). Given that toxic side effects constitute the main reason why compounds fail to become drugs (3), advances in toxicology are poised to increase the rate at which we eliminate unsafe compounds and progress safe drugs into the clinic. The ultimate preclinical assessment of drug safety is in vivo study of drug effects followed by clinical and anatomic pathology. Unfortunately, these assessments are costly and timeconsuming, limiting the numbers of compounds that can be evaluated. In addition, these in vivo studies typically capture overt toxicities associated with acute dosing and are less accurate in revealing liabilities from chronic use. For these reasons, batteries of in vitro tests have been developed that correlate with a toxicological outcome but are rapid and cost-effective enough to enable the evaluation of hundreds or even thousands * To whom correspondence should be addressed. Tel: 847-937-0368. Fax: 847-938-2478. E-mail:
[email protected].
of compounds. Most of these in vitro tests attempt to assess the impact of a drug on particular proteins or biological pathways associated with toxicity [e.g., hERG channel binding (4), inhibition or induction of P450 enzymes (5), or mutagenicity (6)]. Given the efforts in genomics (7), metabolomics (8), and proteomics aimed at identifying the relevant pathways associated with drug toxicity, the predictive power of in vitro assays that measure effects on these pathways should continue to increase. However, understanding the toxicities associated with drug chemistry (such as metabolic activation or the formation of protein-drug adducts) is more difficult. For example, numerous chemical structures are commonly avoided in drug discovery due to their propensity to be converted into reactive species that modify proteins and DNA (9). In addition, programs to screen for covalent binding to proteins or glutathione are in place at many pharmaceutical companies to rank order compounds for development (10, 11). Nonetheless, these approaches suffer from the fact that they merely identify the potential for a compound to covalently modify proteins, which is not sufficient to predict whether a toxic response will result. Thus, although drug–protein adducts may account for a significant number of drug-induced toxicities (10, 12), it has remained elusive to connect drug or metabolite reactivity to the biological pathway(s) triggering the toxicology. In addition, many potentially reactive
10.1021/tx700319t CCC: $37.00 2007 American Chemical Society Published on Web 11/15/2007
Thiol-ReactiVe Compounds Identified Using ALARM NMR
functionalities (e.g., R-keto amides, aldehydes, sulfhydryls, etc.) may be sufficiently safe depending on the specific molecular context of a given drug. In other words, toxic liability must be evaluated not simply on the potential for reactivity but on the increased potential to covalently modify one or more proteins involved in key biological pathways associated with drug side effects. We have recently described an in vitro assay called a La assay to detect reactive molecules by nuclear magnetic resonance (ALARM NMR)1 that measures the ability of a compound to covalently modify cysteines in the human La antigen (11). We initially observed that the La antigen exhibits an exceptionally high frequency of reaction with a wide variety of small organic compounds, and this sensitivity was exploited to detect false positives in high-throughput screening (HTS) campaigns arising from nonspecific reactivity with protein targets. However, during the course of evaluating thousands of drug leads over the past several years (13), we also observed that certain drugs whose clinical side effects are associated with covalent modification of proteins were also active in the ALARM assay. These data led us to speculate that reactivity with the human La antigen (La protein) may be a potential in vitro indicator for an increased risk of drug toxicity due to covalent modification of one or more proteins involved in key biological pathways. To investigate this possibility, we assessed the reactivity of a large and diverse number of ALARM-reactive compounds against a panel of proteins implicated in drug toxicity [aldehyde dehydrogenase (14), Cu/Zn superoxide dismutase (SOD) (15, 16), and P450 enzymes (17)]. The results suggest that reactivity in the ALARM NMR assay is associated with a significantly increased risk for alcohol intolerance, free radical formation, and drug-drug interactions.
Experimental Procedures ALARM NMR Assay. La(223–324)-LEHHHHHH was labeled with 13C at the δ-methyl groups of leucine, δ-methyl groups of isoleucine, and γ-methyl groups of valine by including 3-13C-Rketobutyrate and 3,3′-13C-R-ketoisovalerate in the Escherichia coli growth medium and purified by Ni2+ affinity chromatography as described (11). Purified protein was stored at 200–500 µM in 25 mM sodium phosphate, pH 7.0, and 20 mM dithiothreitol (DTT) at -80 °C after snap freezing in liquid nitrogen. Prior to use in ALARM NMR, 1–3 mL of 200–500 µM protein was incubated with an additional 20 mM DTT at 37 °C for 1 h prior to dialysis against 2 × 2 L of 25 mM sodium phosphate buffer, pH 7.0, at room temperature, with constant bubbling of nitrogen in the dialysis buffer. The protein was stored at 4 °C for up to 1 week for use in the assay. Samples were prepared with 25 µM 13C-La(223–324)-LEHHHHHH in 25 mM sodium phosphate buffer, pH 7.0, 10% D2O, and 200-400 µM test compound delivered from dimethylsulfoxide (DMSO) stocks. The final concentration of DMSO did not exceed 4%. A second sample was prepared with 20 mM DTT in the buffer. For samples with DTT, it is not required to use La protein that was regenerated with the dialysis procedure. All samples were incubated at 37 °C for 1 h and then at room temperature for at least 15 h prior to collecting the NMR data. Typical experiments included 40-120 NMR samples that were stored at room temperature on Bruker NMR sample changers while awaiting data collection. The maximum incubation time at room temperature before data collection was 60 h. 1 Abbreviations: SOD, human Cu/Zn superoxide dismutase (F50E, G51E, and E133Q); DTT, dithiothreitol; DMSO, dimethylsulfoxide; RRM, ribonucleotide recognition motif; La protein, human La antigen; ALARM NMR, A La Assay to detect Reactive Molecules by Nuclear Magnetic Resonance; HTS, high-throughput screening; SAR, structure–activity relationships; SRR, structure–reactivity relationships.
Chem. Res. Toxicol., Vol. 20, No. 12, 2007 1753 1 H/13C HSQC spectra of the La protein methyl groups were recorded at 310 K on Bruker DRX500 spectrometers equipped with a cryoprobe. Sixteen to 32 scans were collected with 1024 complex points in F2 and 38 points in F1 with sweep widths of 8333 and 3000 Hz, respectively. Compound reactivity was assessed by overlaying spectra of a test sample with the spectrum of a control sample with an equivalent concentration of DMSO. Compounds were determined to be reactive when they induced chemical shifts of the La protein in the absence of DTT and when these spectral changes were lessened or eliminated in the presence of 20 mM DTT. In our experience, very few compounds caused specific chemical shift changes in the presence and absence of DTT. Qualitative assessment of the reactivity was made by monitoring the magnitude of the NMR spectral perturbations. Mild reactivity was >20% and
