Novel Multiplexed High Throughput Screening of Selective Inhibitors for Drug-Metabolizing Enzymes Using Human Hepatocytes

Selective chemical inhibitors are critical for reaction phenotyping to identify drug-metabolizing enzymes that are involved in the elimination of drug candidates. Although relatively selective inhibitors are available for the major cytochrome P450 enzymes (CYP), they are quite limited for the less common CYPs and non-CYPs. To address this gap, we developed a multiplexed high throughput screening (HTS) assay using 20 substrate reactions of multiple enzymes to simultaneously monitor the inhibition of enzymes in a 384-well format. Four 384-well assay plates can be run at the same time to maximize throughput. This is the first multiplexed HTS assay for drug-metabolizing enzymes reported. The HTS assay is technologically enabled with state-of-the-art robotic systems and highly sensitive modern LC-MS/MS instrumentation. Virtual screening is utilized to identify inhibitors for HTS based on known inhibitors and enzyme structures. Screening of ~4600 compounds generated many hits for many drug-metabolizing enzymes including the two time-dependent and selective aldehyde oxidase inhibitors, erlotinib and dibenzothiophene. The hit rate is much higher than that for the traditional HTS for biological targets due to the promiscuous nature of the drug-metabolizing enzymes and the biased compound selection process. Future efforts will focus on using this method to identify selective inhibitors for enzymes that do not currently have quality hits and thoroughly characterizing the newly identified selective inhibitors from our screen. We encourage colleagues from other organizations to explore their proprietary libraries using a similar approach to identify better inhibitors that can be used across the industry.

Comparison of Tumor Binding Across Tumor Types and Cell Lines to Support Free Drug Considerations for Oncology Drug Discovery

Tumor binding is an important parameter to derive unbound tumor concentration to explore pharmacokinetics (PK) and pharmacodynamics (PD) relationships for oncology disease targets. Tumor binding was evaluated using eleven matrices, including various commonly used ex vivo human and mouse xenograft and syngeneic tumors, tumor cell lines and liver as a surrogate tissue. The results showed that tumor binding is highly correlated among the different tumors and tumor cell lines except for the mouse melanoma (B16F10) tumor type. Liver fraction unbound (fu) has a good correlation with B16F10 tumor binding. Liver also demonstrates a two-fold equivalency, on average, with binding of other tumor types when a scaling factor is applied. Predictive models were developed for tumor binding, with correlations established with LogD (acids), predicted muscle fu (neutrals) and measured plasma protein binding (bases) to estimate tumor fu when experimental data are not available. Many approaches can be applied to obtain and estimate tumor binding values. One strategy proposed is to use a surrogate tumor tissue, such as mouse xenograft ovarian cancer (OVCAR3) tumor, as a surrogate for tumor binding (except for B16F10) to provide an early assessment of unbound tumor concentrations for development of PK/PD relationships.

Defining the Selectivity of Chemical Inhibitors Used for Cytochrome P450 Reaction Phenotyping: Overcoming Selectivity Limitations with a Six-Parameter Inhibition Curve-Fitting Approach

The utility of chemical inhibitors in cytochrome P450 (CYP) reaction phenotyping is highly dependent on their selectivity and potency for their target CYP isoforms. In the present study, seventeen inhibitors of CYP1A2, 2B6, 2C8, 2C9, 2C19, 2D6, and 3A4/5 commonly used in reaction phenotyping were evaluated for their cross-enzyme selectivity in pooled human liver microsomes. The data were evaluated using a statistical desirability analysis to identify (1) inhibitors of superior selectivity for reaction phenotyping and (2) optimal concentrations for each. Among the inhibitors evaluated, α-naphthoflavone, furafylline, sulfaphenazole, tienilic acid, N-benzylnirvanol, and quinidine were most selective, such that their respective target enzymes were inhibited by ~95% without inhibiting any other CYP enzyme by more than 10%. Other commonly employed inhibitors, such as ketoconazole and montelukast, among others, were of insufficient selectivity to yield a concentration that could adequately inhibit their target enzymes without affecting other CYP enzymes. To overcome these shortcomings, an experimental design was developed wherein dose response data from a densely sampled multi-concentration inhibition curve are analyzed by a six-parameter inhibition curve function, allowing accounting of the inhibition of off-target CYP isoforms inhibition and more reliable determination of maximum targeted enzyme inhibition. The approach was exemplified using rosiglitazone N-demethylation, catalyzed by both CYP2C8 and 3A4, and was able to discern the off-target inhibition by ketoconazole and montelukast from the inhibition of the targeted enzyme. This methodology yields more accurate estimates of CYP contributions in reaction phenotyping. Significance Statement Isoform-selective chemical inhibitors are important tools for identifying and quantifying enzyme contributions as part of a CYP reaction phenotyping assessment for projecting drug-drug interactions. However, currently employed practices fail to adequately compensate for shortcomings in inhibitor selectivity and the resulting confounding impact on estimates of the CYP enzyme contribution to drug clearance. In this report, we describe a detailed IC50 study design with 6-parameter modeling approach that yields more accurate estimates of enzyme contribution.

An improved method for cytochrome p450 reaction phenotyping using a sequential qualitative-then-quantitative approach

Cytochrome P450 reaction phenotyping to determine the fraction of metabolism values (fm) for individual enzymes is a standard study in the evaluation of a new drug. However, there are technical challenges in these studies caused by shortcomings in the selectivity of P450 inhibitors and unreliable scaling procedures for recombinant P450 (rCYP) data. In this investigation, a two-step "qualitative-then-quantitative" approach to P450 reaction phenotyping is described. In the first step, each rCYP is tested qualitatively for potential to generate metabolites. In the second step, selective inhibitors for the P450s identified in step1 are tested for their effects on metabolism using full inhibition curves. Forty-eight drugs were evaluated in step 1 and there were no examples of missing an enzyme important to in vivo clearance. Five drugs (escitalopram, fluvastatin, pioglitazone, propranolol, and risperidone) were selected for full phenotyping in step2 to determine fm values, with findings compared to fm values estimated from single inhibitor concentration data and rCYP with intersystem-extrapolation-factor corrections. The two-step approach yielded fm values for major drug clearing enzymes that are close to those estimated from clinical data: escitalopram and CYP2C19 (0.42 vs 0.36-0.82), fluvastatin and CYP2C9 (0.76 vs 0.76), pioglitazone and CYP2C8 (0.72 vs 0.73), propranolol and CYP2D6 (0.68 vs 0.37-0.56) and risperidone and CYP2D6 (0.60 vs 0.66-0.88). Reaction phenotyping data generated in this fashion should offer better input to physiologically-based pharmacokinetic models for prediction of DDI and impact of genetic polymorphisms on drug clearance. The qualitative-then-quantitative approach is proposed as a replacement to standard reaction phenotyping strategies. Significance Statement P450 reaction phenotyping is important for projecting drug-drug interactions and interpatient variability in drug exposure. However, currently recommended practices can frequently fail to provide reliable estimates of the fractional contributions of specific P450 enzymes (fm) to drug clearance. In this report, we describe a two-step qualitative-then-quantitative reaction phenotyping approach that yields more accurate estimates of fm.

Selecting Approaches for Hit Identification and Increasing Options by Building the Efficient Discovery of Actionable Chemical Matter from DNA-Encoded Libraries

Over the past 20 years, the toolbox for discovering small-molecule therapeutic starting points has expanded considerably. Pharmaceutical researchers can now choose from technologies that, in addition to traditional high-throughput knowledge-based and diversity screening, now include the screening of fragment and fragment-like libraries, affinity selection mass spectrometry, and selection against DNA-encoded libraries (DELs). Each of these techniques has its own unique combination of advantages and limitations that makes them more, or less, suitable for different target classes or discovery objectives, such as desired mechanism of action. Layered on top of this are the constraints of the drug-hunters themselves, including budgets, timelines, and available platform capacity; each of these can play a part in dictating the hit identification strategy for a discovery program. In this article, we discuss some of the factors that we use to govern our building of a hit identification roadmap for a program and describe the increasing role that DELs are playing in our discovery strategy. Furthermore, we share our learning during our initial exploration of DEL and highlight the approaches we have evolved to maximize the value returned from DEL selections. Topics addressed include the optimization of library design and production, reagent validation, data analysis, and hit confirmation. We describe how our thinking in these areas has led us to build a DEL platform that has begun to deliver tractable matter to our global discovery portfolio.

Evaluation of Prediction Accuracy for Volume of Distribution in Rat and Human Using In Vitro, In Vivo, PBPK and QSAR Methods

Volume of distribution at steady state (V_ss) is an important pharmacokinetic parameter of a drug candidate. In this study, V_ss prediction accuracy was evaluated by using: (1) seven methods for rat with 56 compounds, (2) four methods for human with 1276 compounds, and (3) four in vivo methods and three Kp (partition coefficient) scalar methods from scaling of three preclinical species with 125 compounds. The results showed that the global QSAR models outperformed the PBPK methods. Tissue fraction unbound (f_u,t) method with adipose and muscle also provided high V_ss prediction accuracy. Overall, the high performing methods for human V_ss prediction are the global QSAR models, Øie-Tozer and equivalency methods from scaling of preclinical species, as well as PBPK methods with Kp scalar from preclinical species. Certain input parameter ranges rendered PBPK models inaccurate due to mass balance issues. These were addressed using appropriate theoretical limit checks. Prediction accuracy of tissue Kp were also examined. The f_u,t method predicted Kp values more accurately than the PBPK methods for adipose, heart and muscle. All the methods overpredicted brain Kp and underpredicted liver Kp due to transporter effects. Successful V_ss prediction involves strategic integration of in silico, in vitro and in vivo approaches.

Optimal designs for pairwise calculation: An application to free energy perturbation in minimizing prediction variability

Pairwise-based methods such as the free energy perturbation (FEP) method have been widely deployed to compute the binding free energy differences between two similar host-guest complexes. The calculated pairwise free energy difference is either directly adopted or transformed to absolute binding free energy for molecule rank ordering. We investigated, through both analytic derivations and simulations, how the selection of pairs in the experiment could impact the overall prediction precision. Our studies showed that (1) the estimated absolute binding free energy ( Δ G ^ ) derived from calculated pairwise differences (ΔΔG) through weighted least squares fitting is more precise in prediction than the pairwise difference values when the number of pairs is more than the number of ligands and (2) prediction precision is influenced by both the total number of pairs and the specifically selected pairs, the latter being critically important when the number of calculated pairs is limited. Furthermore, we applied optimal experimental design in pair selection and found that the optimally selected pairs can outperform randomly selected pairs in prediction precision. In an illustrative example, we showed that, upon weighing ligand structure similarity into design optimization, the weighted optimal designs are more efficient than the literature reported designs. This work provides a new approach to assess retrospective pairwise-based prediction results, and a method to design new prospective pairwise-based experiments for molecular lead optimization. © 2019 Wiley Periodicals, Inc.

Effect of a Novel Engagement Strategy Using Twitter on Test Performance

Introduction

Medical educators in recent years have been using social media for more penetrance to technologically-savvy learners. The utility of using Twitter for curriculum content delivery has not been studied. We sought to determine if participation in a social media-based educational supplement would improve student performance on a test of clinical images at the end of the semester.

Methods

116 second-year medical students were enrolled in a lecture-based clinical medicine course, in which images of common clinical exam findings were presented. An additional, optional assessment was performed on Twitter. Each week, a clinical presentation and physical exam image (not covered in course lectures) were distributed via Twitter, and students were invited to guess the exam finding or diagnosis. After the completion of the course, students were asked to participate in a slideshow “quiz” with 24 clinical images, half from lecture and half from Twitter.

Results

We conducted a one-way analysis of variance to determine the effect Twitter participation had on total, Twitter-only, and lecture-only scores. Twitter participation data was collected from the end-of-course survey and was defined as submitting answers to the Twitter-only questions “all or most of the time”, “about half of the time”, and “little or none of the time.” We found a significant difference in overall scores (p<0.001) and in Twitter-only scores (p<0.001). There was not enough evidence to conclude a significant difference in lecture-only scores (p=0.124). Students who submitted answers to Twitter “all or most of the time” or “about half the time” had significantly higher overall scores and Twitter-only scores (p<0.001 and p<0.001, respectively) than those students who only submitted answers “little or none of the time.”

Conclusion

While students retained less information from Twitter than from traditional classroom lecture, some retention was noted. Future research on social media in medical education would benefit from clear control and experimental groups in settings where quantitative use of social media could be measured. Ultimately, it is unlikely for social media to replace lecture in medical curriculum; however, there is a reasonable role for social media as an adjunct to traditional medical education.