Application of proteomic data in the translation of in vitro observations to associated clinical outcomes

Effects of Chronic Alcohol Intake on the Composition of the Ensemble of Drug-Metabolizing Enzymes and Transporters in the Human Liver

In this study, to better understand the mechanisms of the profound impact of alcohol consumption on drug pharmacokinetics, efficacy, and toxicity, we characterized the alcohol-induced changes in the ensemble of drug-metabolizing enzymes and transporters (DMETs) in the human liver by performing global proteomic analysis of human liver microsomes from 94 donors. DMET protein levels were analyzed concerning alcohol consumption, smoking history, and sex using non-parametric tests, which were further strengthened by correlational analysis. To this end, we used a provisional index of alcohol exposure formulated based on the relative abundances of four marker proteins best correlating with the level of alcohol consumption. Alcohol-induced changes in the cytochrome P450 pool include significant increases in CYP2E1, CYP2B6, CYP2J2, and NADPH-cytochrome P450 reductase levels and the lowering of CYP1A2, CYP2C8, CYP2C9, CYP4A11, and cytochrome b5. Changes in UDP-glucuronosyltransferase (UGT) abundances comprise elevated UGT1A6, UGT1A9, and UGT2A1, and reduced UGT1A3, UGT1A4, UGT2B7, UGT2B10, and UGT2B15 levels. Tobacco smokers showed elevated CYP1A2, UGT1A6, and UGT2B4 and reduced FMO3, FMO4, and FMO5 levels, while in females, CYP1A2, UGT2B17, and UGT2B15 levels were lower, and UGT2A3 and STS were higher compared to males. The alcohol-induced changes in the DMET ensemble at the protein level reported herein provide deep insights into how alcohol impacts drug and xenobiotic metabolism.

Quantitative Proteomics for Translational Pharmacology and Precision Medicine: State of The Art and Future Outlook

Over the past 20 years, quantitative proteomics has contributed a wealth of protein expression data, which are currently used for a variety of systems pharmacology applications, as a complement or a surrogate for activity of the corresponding proteins. A symposium at the 25th North American International Society for the Study of Xenobiotics meeting, in Boston, in September 2023, was held to explore current and emerging applications of quantitative proteomics in translational pharmacology and strategies for improved integration into model-informed drug development based on practical experience of each of the presenters. A summary of the talks and discussions is presented in this perspective alongside future outlook that was outlined for future meetings. SIGNIFICANCE STATEMENT: This perspective explores current and emerging applications of quantitative proteomics in translational pharmacology and precision medicine and outlines the outlook for improved integration into model-informed drug development.

Effects of alcohol consumption and tobacco smoking on the composition of the ensemble of drug metabolizing enzymes and transporters in human liver

We examined the effect of alcohol consumption and smoking on the abundance of drug-metabolizing enzymes and transporters (DMET) in human liver microsomes (HLM) isolated from liver tissues of 94 donors. Global proteomics analysis was performed and DMET protein levels were analyzed in relation to alcohol consumption levels, smoking history, and sex using non-parametric tests (p-value ≤ 0.05; cutoff of 1.25-fold change, FC). The examination of the alcohol-induced changes was further enforced by correlational analysis, where we used arbitrary alcohol consumption grade (ACG) scaling from 0 to 4 to establish a set of protein markers. We elaborated a provisional index of alcohol exposure (PIAE) based on a combination of relative abundances of four proteins (ER chaperone HSPA5, protein disulfide isomerases PDIA3 and P4HB, and cocaine esterase CES2) best correlating with ACG. The PIAE index was then used to find its correlations with the abundances of DMET proteins. Our results demonstrate considerable alcohol-induced changes in composition of the pool of cytochrome P450 enzymes in HLM. We observed significantly increased abundances of CYP2E1, CYP2B6, CYP2J2, and NADPH-cytochrome P450 reductase. In contrast, CYP1A2, CYP2C8, CYP2C9, CYP4A11, and cytochrome b5 protein levels were downregulated. Significant alteration in abundances of UDP-glucuronosyltransferase (UGT) were also detected, comprising of elevated UGT1A6, UGT1A9, and UGT2A1, and reduced UGT1A3, UGT1A4, UGT2B7, UGT2B10, and UGT2B15 levels. Important alcohol-induced changes were also observed in the expression of non-CYP and non-UGT DMET. Additionally, tobacco smoke was associated with elevated CYP1A2, UGT1A6, UGT2A1, and UGT2B4 and decreased FMO3, FMO4, and FMO5 levels.

In Vitro to In Vivo Scalars for Drug Clearance in Nonalcoholic Fatty Liver and Steatohepatitis

In vitro-in vivo extrapolation (IVIVE) allows prediction of clinical outcomes across populations from in vitro data using specific scalars tailored to the biologic characteristics of each population. This study experimentally determined scalars for patients with varying degrees of nonalcoholic fatty liver disease (NAFLD), ranging from fatty liver to nonalcoholic steatohepatitis (NASH) and cirrhosis. Microsomal, S9, and cytosol fractions were extracted from 36 histologically normal and 66 NAFLD livers (27 nonalcoholic fatty liver [NAFL], 13 NASH, and 26 NASH with cirrhosis). Corrected microsomal protein per gram liver (MPPGL) progressively decreased with disease severity (26.8, 27.4, and 24.3 mg/g in NAFL, NASH, and NASH/cirrhosis, respectively, compared with 35.6 mg/g in normal livers; ANOVA, P < 0.001). Homogenate, S9, and cytosolic protein showed a consistent trend of decline in NASH/cirrhosis relative to normal control (post-hoc t test, P < 0.05). No differences across the groups were observed in homogenate, S9, cytosolic, and microsomal protein content in matched kidney samples. MPPGL-based scalars that combine protein content with liver size revealed that the reduction in MPPGL in NAFL and NASH was compensated by the reported increase in liver size (relative scalar ratios of 0.96 and 0.99, respectively), which was not the case with NASH/cirrhosis (ratio of 0.63), compared with healthy control. Physiologically based pharmacokinetics-informed global sensitivity analysis of the relative contribution of IVIVE scalars (hepatic CYP3A4 abundance, MPPGL, and liver size) to variability in exposure (area under the curve) to three CYP3A substrates (alprazolam, midazolam, and ibrutinib) revealed enzyme abundance as the most significant parameter, followed by MPPGL, whereas liver volume was the least impactful factor. SIGNIFICANCE STATEMENT: Nonalcoholic fatty liver disease-specific scalars necessary for extrapolation from in vitro systems to liver tissue are lacking. These are required in clearance prediction and dose selection in nonalcoholic fatty liver and steatohepatitis populations. Previously reported disease-driven changes have focused on cirrhosis, with no data on the initial stages of liver disease. The authors obtained experimental values for microsomal, cytosolic, and S9 fractions and assessed the relative impact of microsomal scalars on predicted exposure to substrate drugs using physiologically based pharmacokinetics.

Epithelial and microbial determinants of colonic drug distribution

A dynamic epithelium and a rich microbiota, separated by multi-layered mucus, make up the complex colonic cellular environment. Both cellular systems are characterized by high inter- and intraindividual differences, but their impact on drug distribution and efficacy remains incompletely understood. This research gap is pressing, as, e.g., inflammatory disorders of the colon are on the rise globally. In an effort to help close this gap, we provide considerations on determining colonic epithelial and microbial cellular parameters, and their impact on drug bioavailability. First, we cover the major cell types found in vivo within the epithelium and microbiota, and discuss how they can be modeled in vitro. We then draw attention to their structural similarities and differences with regard to determinants of drug distribution. Once a drug is solubilized in the luminal fluids, there are two main classes of such determinants: 1) binding processes, and 2) transporters and drug-metabolizing enzymes. Binding lowers the unbound intracellular fraction (f_u,cell), which will, in turn, limit the amount of drug available for transport to desired sites. Transporters and drug metabolizing enzymes are ADME proteins impacting intracellular accumulation (Kp). Across cell types, we point out which processes are likely particularly impactful. Together, f_u,cell and Kp can be used to describe intracellular bioavailability (F_ic), which is a measure of local drug distribution, with consequences for efficacy. Determining these cellular parameters will be beneficial in understanding colonic drug distribution and will advance the field of drug delivery.

Considerations for Physiologically Based Modeling in Liver Disease: From Nonalcoholic Fatty Liver (NAFL) to Nonalcoholic Steatohepatitis (NASH)

Nonalcoholic fatty liver disease (NAFLD), representing a clinical spectrum ranging from nonalcoholic fatty liver (NAFL) to nonalcoholic steatohepatitis (NASH), is rapidly evolving into a global pandemic. Patients with NAFLD are burdened with high rates of metabolic syndrome-related comorbidities resulting in polypharmacy. Therefore, it is crucial to gain a better understanding of NAFLD-mediated changes in drug disposition and efficacy/toxicity. Despite extensive clinical pharmacokinetic data in cirrhosis, current knowledge concerning pharmacokinetic alterations in NAFLD, particularly at different stages of disease progression, is relatively limited. In vitro-to-in vivo extrapolation coupled with physiologically based pharmacokinetic and pharmacodynamic (IVIVE-PBPK/PD) modeling offers a promising approach for optimizing pharmacologic predictions while refining and reducing clinical studies in this population. Use of IVIVE-PBPK to predict intra-organ drug concentrations at pharmacologically relevant sites of action is particularly advantageous when it can be linked to pharmacodynamic effects. Quantitative systems pharmacology/toxicology (QSP/QST) modeling can be used to translate pharmacokinetic and pharmacodynamic data from PBPK/PD models into clinically relevant predictions of drug response and toxicity. In this review, a detailed summary of NAFLD-mediated alterations in human physiology relevant to drug absorption, distribution, metabolism, and excretion (ADME) is provided. The application of literature-derived physiologic parameters and ADME-associated protein abundance data to inform virtual NAFLD population development and facilitate PBPK/PD, QSP, and QST predictions is discussed along with current limitations of these methodologies and knowledge gaps. The proposed methodologic framework offers great potential for meaningful prediction of pharmacological outcomes in patients with NAFLD and can inform both drug development and clinical practice for this population.

Label-Free but Still Constrained: Assessment of Global Proteomic Strategies for the Quantification of Hepatic Enzymes and Transporters

Building and refining pharmacology models require "system" data derived from tissues and in vitro systems analyzed by quantitative proteomics. Label-free global proteomics offers a wide scope of analysis, allowing simultaneous quantification of thousands of proteins per sample. The data generated from such analysis offer comprehensive protein expression profiles that can address existing gaps in models. In this study, we assessed the performance of three widely used label-free proteomic methods, "high N" ion intensity approach (HiN), intensity-based absolute quantification (iBAQ) and total protein approach (TPA), in relation to the quantification of enzymes and transporters in 27 human liver microsomal samples. Global correlations between the three methods were highly significant (R2 > 0.70, P < 0.001, n = 2232 proteins). Absolute abundances of 57 pharmacokinetic targets measured by standard-based label-free methods (HiN and iBAQ) showed good agreement, whereas the TPA overestimated abundances by two- to threefold. Relative abundance distribution of enzymes was similar for the three methods, while differences were observed with TPA in the case of transporters. Variability (CV) was similar across methods, with consistent between-sample relative quantification. The back-calculated amount of protein in the samples based on each method was compared with the nominal protein amount analyzed in the proteomic workflow, revealing overall agreement with data from the HiN method with bovine serum albumin as standard. The findings herein present a critique of label-free proteomic data relevant to pharmacokinetics and evaluate the possibility of retrospective analysis of historic datasets. SIGNIFICANCE STATEMENT: This study provides useful insights for using label-free methods to generate abundance data applicable for populating pharmacokinetic models. The data demonstrated overall correlation between intensity-based label-free proteomic methods (HiN, iBAQ and TPA), whereas iBAQ and TPA overestimated the total amount of protein in the samples. The extent of overestimation can provide a means of normalization to support absolute quantification. Importantly, between-sample relative quantification was consistent (similar variability) across methods.