Chapter 10 Metabolism and pharmacokinetic optimization strategies in drug discovery

P. Ballard, P. Brassil, K H. Bui, H. Dolgos, C. Petersson, A. Tunek, P J H. Webborn

Introduction

Optimization of drug metabolic and pharmacokinetic properties is an integral component of the modern drug discovery process. The objective of the drug metabolism and pharmacokinetics (DMPK) discipline in drug discovery is to aid design and selection of candidate drugs with properties that yield the required efficacy and safety for effective clinical use. The roles of DMPK at the various stages of drug discovery are summarized in Table 10.1. DMPK in vitro and in vivo information are used throughout the drug discovery process to facilitate target validation and safety assessment, and to guide the conversion of early screening hits and leads into drug candidates. Indeed, the frontloading of DMPK in drug discovery has resulted in a reduction of drug attrition rate due to undesirable DMPK properties from approximately 40% in 1990 to 10% in 2000 (Kola and Landis, 2004).

Table 10.1 Roles of drug metabolism and pharmacokinetics (DMPK) in various phases of drug discovery

Discovery phase	DMPK roles
Target identification	Selection and characterization of tool compounds Partner with in vivo pharmacology and toxicology in target validation and safety assessment activities
Hit identification	In silico and in vitro DMPK profiling to help prioritize hit series
Lead identification	Identify DMPK liabilities of lead series Determine whether if DMPK properties in lead series are optimizable (DMPK liabities not linked to pharmacophore) Develop structure (DMPK) property relationship (SPR) Develop ‘in vitro in vivo’ correlation (IVIVC) Guide selection of lead series Contribute to development of PD assay and develop early PK/PD understanding Contribute to development of PD assays and develop early PK/PD understanding
Lead optimization	Guide optimization of DMPK properties toward target profile Comprehensive DMPK characterization of candidate compounds Develop biomarker based and/or efficacy/disease related PK/PD models Prediction of human PK Prediction of efficacious human dose Integration of predicted profile to calculate key margins against side effects

To help drug hunting teams focus on the key issues and goals, in a multitude of screening options, it is important to prescribe, as early as possible, a candidate drug target profile in terms of efficacy, potency, safety and ease of use. DMPK plays a central role in defining this target profile. For instance to be commercially attractive in terms of ease of use, a compound has to be orally active, has a convenient dosing regimen and be able to be administered without effect from food and other medications. The physicochemical and DMPK attributes that will allow a compound to meet this target profile would be: good solubility and permeability, high oral bioavailability, low clearance and reasonable half-life (if PD half-life is not much longer than PK half life), and absence of ‘drug–drug interaction’ potential. Likewise to be orally active, a compound should have good oral bioavailability and able to reach the target organ at high enough concentration to engage the target.

In this chapter, strategies to optimize key DMPK challenges using appropriate in silico, in vitro and in vivo DMPK tools during drug discovery are presented. The rational use of these strategies will help ‘drug hunting’ projects to advance drug candidates with attractive DMPK target profile and with low potential for failure in development due to DMPK issues. In addition, as prediction of human PK and safe and effective dose is probably the most important activity in drug discovery to ensure that the candidate drug has the attribute to test the biological hypothesis in patients, strategy to integrate information in discovery to holistically predict PK properties in man will be discussed.

Optimization of DMPK properties

Optimization principles are described for six key DMPK areas:

• Absorption and bioavailability

• Avoiding PK-based drug–drug interactions

• Achieving/avoiding CNS exposure

• Clearance

• Role of metabolite identification:

Active metabolites

Minimizing risk for reactive metabolites.

The following sections summarize current understanding/available tools and suggest best practice for each of the above. Each section introduces the challenges, outlines tactics for dealing with them and identifies areas requiring caution.

Absorption and oral bioavailability

Introduction

As the preferred route of administration for most indications is oral, it is important to characterize oral bioavailability (F) of a compound during drug discovery. In addition, F must be optimized, as a low F is often associated with poor and variable exposure and lack of efficacy. F is defined as the percentage of dosed drug that reaches the systemic circulation compared to the IV route. As shown below, it can be considered to be dependent on three serial steps: the fraction of dosed drug absorbed (f_a), the fraction escaping intestinal metabolism (f_g) and the fraction extracted by the liver as it passes from the portal vein to the systemic circulation (f_h) (see Rowland and Tozer, 1989):

f_a is influenced by a number of factors including the gastrointestinal (GI) solubility (dose, particle size, pH solubility profile and formulation), the effective permeability (both passive permeability and active transport processes) and GI stability. f_g and f_h are affected by metabolic enzymes in the intestinal wall and liver, respectively. In addition, f_h can be influenced by transporters if a drug is excreted unchanged into the bile. Both metabolic and active transport processes are saturable, generally obeying Michaelis–Menten kinetics. Hence, f_a, f_g and f_h can all be non-linear if relevant concentrations are above the Michaelis–Menten constant (K_m) for the particular enzyme/transporter–drug interaction.

In humans, the combinations of high to low solubility and permeability have led compounds to be characterized according to the Biopharmaceutical Classification System (Amidon et al., 1995). Class 1 compounds, with high solubility and permeability, generally have very good absorption properties. Those in Class 4, with poor solubility and permeability, are likely to present significant formulation challenges and/or variable and poor exposure. It can be important to characterize the maximum absorbable dose (MAD) of a compound relative to its predicted therapeutic dose, as this will determine the risk of being able to deliver an efficacious dose to humans and guide whether high exposure in safety studies is achievable.

Table 10.2 gives guidance on acceptable pharmaceutical properties for typical oral drug candidates.

Table 10.2 Pharmaceutical development candidate drug target profile (CDTP)

Tactics

In theory, it is relatively simple to obtain good absorption and to ensure that solubility and permeability fall within the right ranges. However, the reality is more complex. From an in vivo perspective, the product of absorption and intestinal metabolism is often assessed by accounting for first-pass hepatic clearance in bioavailability estimations:

If f_a x f_g is low, it is important to understand the relative contribution of both f_a and f_g to F so that this can be designed out of the project. Generally it is simplest to estimate the likely f_a and if this does not account for the poor f_a x f_g value, f_g should be investigated. Poor absorption can be a result of slow dissolution rate, low solubility in the GI tract, poor effective permeability (passive or active efflux), or instability in GI fluids or in the wall of the GI tract. If absorption is adequate but bioavailability is poor, hepatic clearance (metabolic or biliary elimination) and/or intestinal metabolism may need to be optimized.

When maximizing the chances of good absorption, the starting point is to ensure that the physicochemical properties of the compound/series are in the optimal space as described by Lipinski and others (Lipinski et al., 1997; Wenlock et al., 2003; Johnson et al., 2009; Waring, 2009). Generally, this requires minimizing the number of H-bond donors and acceptors, restricting lipophilicity in the range LogD_7.4 0 to 3, and limiting molecular weight to <500. Both the solubility and passive permeability of a compound should be assessed prior to any in vivo study. Predictive models for these should be assessed for suitability in each project and considered in compound design if either is an issue for a chemical series.

The primary in vitro tool for assessing absorption is the cell based Caco-2 permeability assay, although MDCK-MDR1, PAMPA (parallel artificial membrane permeability assay), or in silico predictions may also provide valuable information about efflux transporter risk and permeability. This permeability assessment, in combination with a solubility measurement (ideally using crystalline material), is used to estimate f_a using commercially available modelling tools like GastroPlus (www.simulations-plus.com) or SIMCYP (www.simcyp.com). For actively transported (efflux or uptake) compounds, bi-directional permeability assays can be used as a guide to possible in vivo effects. However, it should be noted that it is often difficult to extrapolate the results from these in vitro transport assays to accurately quantify effects in vivo. Because of its reasonable throughput, the Caco-2 assay can be positioned as an early screen if absorption or permeability/efflux is found to be an issue in the project. If further evaluation of absorption or efflux is warranted, it is possible to use more physiological models such as sections of intestinal tissue in an Ussing Chamber (Ungell et al., 1998), or transfected cell lines over-expressing particular efflux transporters (e.g. P-gp in the MDCK-MDR1 cell line). The Ussing Chamber technique can help in understanding cross-species differences and, because the tissue used is enzymatically competent (metabolic and transporters), the output represents the product of f_a and f_g.

Compounds with low hepatic clearance in the rat, good solubility and high effective permeability should exhibit good oral absorption and bioavailability in that species. However, if this is not the case, the troubleshooting decision tree in Figure 10.1 can be used to help determine the cause(s) of poor absorption, identify assays to aid in optimizing compound design and understand if the compound is of sufficient quality to progress in the value chain, despite its non-optimal absorption properties.

Fig. 10.1 Absorption troubleshooting decision tree. Caco-2 ABBA, transport assays; CL, clearance; f_a, fraction of dose absorbed; GI, gastro-intestinal; IVIVC, in vitro to in vivo correlation; LBF, liver blood flow; LI, lead identification; LO, optimization; MAD, maximum absorbable dose; Papp, apparent permeability; SAR, structure–activity relationship.

Table 10.3 is an aid to selecting the assays and techniques to use in the decision tree (Figure 10.1), to explain potential issues and risks, and the parameter (f_a, f_g and/or f_h) the assays impact on.

Table 10.3 Assays and techniques used when troubleshooting absorption

Assay/technique	Potential issue/risk addressed	Impacted
Intestinal microsomes or S9 fraction	Assess cross species differences in intestinal metabolism. Nature and source of metabolites can give key information about potential for gut metabolism, as CYP3A and UGTs account for most gut metabolites	f_g
Absorption profiling on Caco-2 cells	Varying apical to basolateral pH gradient Concentration dependency Use of proteins (e.g. BSA) at various percentages in apical chamber Use of efflux transporter inhibitors	f_a
PAMPA	Passive permeability	f_a
High dose PK studies	Saturation of efflux or metabolism	f_a, f_g, f_h
Ussing chamber technique	Effective permeability (including transporters) Intestinal tissue metabolism	f_a, f_g
GI stability test	Degradation of drug in stomach or intestinal lumen is possible explanation (usually the case if predicted F much greater than measured F)	f_a
Human metabolic phenotyping	Gut metabolism	f_g, f_h
In situ/in vivo portal vein cannulation preparation	Determine amount of drug and metabolites passing through intestine	f_a, f_g
SIMCYP; GastroPlus	Predict absorption rate and extent (software methods)	f_a
Transfected cell lines; vesicles expressing specific transporters; Caco-2 efflux assay with and without specific transporter inhibitors	Determine involvement of specific efflux transporters	f_a, f_h
Knock-out (KO) animals; chemical KO with inhibitors	Determine involvement of specific efflux transporters	f_a, f_g, f_h

Cautions

• As scaling factors for intestinal microsomes or other subcellular intestinal fractions are currently not available, it is difficult to make an accurate quantitative assessment of the contribution of intestinal metabolism to in vivo f_a. However, attempts have been made to use CL_int from human liver microsomes or S9 fraction to estimate the relative contributions of f_a and f_g to F (Gertz et al., 2010).

• It is often very difficult to pinpoint why compounds have poor absorption characteristics and, therefore, to resolve this design issue. Thus, it is often reasonable to prioritize series with good f_a x f_g, in early discovery even if other properties (e.g. potency, clearance) are less attractive.

• Typically, oral doses are formulated as suspensions which, on many occasions, may be derived from amorphous material. However, it is important to assess absorption periodically using crystalline material, as physical form may have substantial effects on absorption profiles.

• Particularly for compounds likely to proceed into development, it is important to determine the effect of the polymorphic solid states on absorption. It is also important to ensure that the formulations used are discussed with Pharmaceutical Development (see Chapter 16), and are appropriate for safety and early clinical development studies.

• If in vitro and in vivo (rat and dog) assessments of f_a do not agree, the risk of an inaccurate estimate of absorption in man will increase. Sometimes other species have been investigated to mitigate this risk, but we caution that there can, for example, be marked discrepancies in f_a x f_g between cynomolgus monkeys and humans (Takahashi et al., 2009).

Avoidance of PK based drug–drug interactions

Introduction

Aid in the design and selection of candidate drugs with a low potential for PK-based drug–drug interactions (DDI) is a key role of discovery DMPK. There are four main forms of PK-based DDI, in which the compound may be a perpetrator or be a victim of DDI:

• Competitive (reversible) cytochrome P450 (CYP) inhibition

• Mechanism based/time dependent CYP inhibition

• Uptake and efflux transporter inhibition

• CYP induction.

CYP, notably CYP3A4 and CYP2D6, based DDI is the most important and most common, and may occur in the liver or intestine. Transporter based DDI is mainly related to renal clearance, although specific issues can arise with CNS compounds and hepatic uptake of statins. The science and regulatory guidance to support risk assessment of CYP-based DDI are well established, but is less advanced for transporter-related issues (Bjornsson et al., 2003; US Food and Drug Administration, 1997, 2006; Huang et al., 2008; International Transporter Consortium, 2010).

Tactics

Competitive (reversible) CYP inhibition

Two main types of CYP inhibition assays with different capabilities are in general use. Fluorescence based assays are relatively cheap and have enhanced throughput, but in a small but significant number of cases can lead to misrepresenting DDI risk (Bell et al., 2008). They are best used for initial profiling of large numbers of compounds, with the data being acceptable in the early phase such as lead generation. LCMS based assays are more expensive, have lower throughput, but are more predictive. They should be used in optimization cycles once CYP inhibition issues have been identified, and for generating compound profiles during more advanced project phases.

Inhibition of the five major CYP isoforms 1A2, 2C9, 2C19, 2D6 and 3A4, should be evaluated in the earliest phases, while later it would be prudent to assess potential interactions with isoforms 2B6, 2C8 and 3A5.

Reduction of CYP inhibition potential is facilitated by the fact that strong QSAR relationships are often obtained. Various computational models that allow prediction of CYP DDI risk are available within most drug development companies. It is well established that lipophilicity, aromaticity and charge type are major drivers for inhibiting various CYP enzymes (Gleeson et al., 2007).

The risk of DDIs based on Phase 2 metabolism (e.g. glucuronidation and sulphation) is usually small, resulting in less than a two-fold increase in area under the concentration versus time curve (AUC), and they are rarely observed, possibly due in part to the nature of the enzymatic reaction (high V_max and moderate to high K_m values). Such DDIs are not generally evaluated in lead optimization (Williams et al., 2004). Evidence suggesting the need to do so at this stage should prompt re-evaluation of the risk.

The decision tree in Figure 10.2 can be used to assess the potential DDI risk of a competitive CYP inhibitor. Hits identified in a fluorescence-based assay should be confirmed with an LCMS-based assay using druglike substrates as probes for the different CYP isoforms. Although the ratio C_max/K_i,u can be used to obtain a preliminary estimate of the DDI risk, more accurate evaluation should be conducted using PBPK modelling (e.g. SIMCYP platform) to predict the potential clinical risk (expanded below in ‘Prediction of DDI risk’ ”).

Fig. 10.2 Decision tree for reversible DDI CYP inhibition. AUC, area under the concentration-time curve; C_max, maximum concentration; CYP, cytochrome P450; DDI, drug-drug interaction; f_u,mics, fraction unbound in microsomes); IC₅₀, concentration producing 50% inhibition; K_i,u, unbound dissociation rate constant; SIMCYP, software platform.

Mechanism-based/time-dependent CYP inhibition

The inhibition of CYP enzymes may be irreversible (due to irreversible or covalent binding to the prosthetic haem or the enzyme) or quasi-irreversible (due to the formation of transient complexes with the iron of the haem prosthetic group). Time-dependent inhibition (TDI) methods can be used to determine this (Riley et al., 2007; Fowler and Zhang, 2008). During early phases of drug discovery, a medium throughput screening assay can be used to screen for TDI. However, for selecting candidates at later phases, the method employed should provide accurate determination of K_inact and K_i to properly evaluate the DDI risks of compounds in which preliminary screening indicated the potential for TDI. A positive TDI finding also suggests that the compound or its metabolites may be reactive, and further evaluation should be conducted as specified by reactive metabolite strategies.

A decision tree to help evaluate the potential DDI risk of a TDI CYP inhibitor is shown in Figure 10.3. If a compound is found to be at risk for CYP inhibition (IC₅₀ <20 µM) or is flagged to have a potential liability for reactive metabolites, it should be screened for TDI. If it is found to have potential TDI risk based on screening data, K_i and K_inact values should be generated to help accurately predict the risk using prediction tools like SIMCYP.

Fig. 10.3 Decision tree to assess time dependent inhibition (TDI) risk. CYP, cytochrome P450; DDI, drug–drug interaction; K_i, dissociation rate constant; K_inact, inactivation rate constant; [I], inhibitor concentration; IC₅₀, concentration producing 50% inhibition; SIMCYP, assay method; TDI, time dependent inhibition.

Uptake and efflux transporter inhibition

Uptake transporter inhibition assays are emerging as being of real value (Ward, 2008), but they are highly dependent on chemotype (e.g. acids being primarily transported by organic anion transporters (OATs)) and likely co-medications (e.g. OCT2 and metformin). Acids and zwitterions should be assessed for inhibition of OATP1B1 during drug discovery. Other inhibition assays (OAT1, OAT3, OATP1B1, OCT1 and OCT2) should be used on a case-by-case basis.

Efflux transporters are believed to serve a protective function, and prevent molecules perceived as foreign from gaining entry in cells or tissues. Of the various efflux transporters, P-glycoprotein (P-gp) is the most prevalent and well understood. As specific locations of the P-gp transporter include small intestine enterocytes, hepatocytes, the kidney and the blood–brain barrier, P-gp can affect oral bioavailability, biliary and renal clearance, and brain uptake of compounds that are substrates of P-gp. In addition, compounds that modulate P-gp can influence the clearance and distribution of drugs that are substrates of P-gp. The bi-directional transport assay using either Caco-2 (P-gp, BCRP, and MRP2) or MDCK-MDRI (specifically for P-gp) cells is widely available. Evaluation of the potential to inhibit P-gp should be evaluated in early discovery using a bi-directional transport assay with a probe P-gp substrate (e.g. digoxin). If a compound is found to be a P-gp inhibitor, its potential impact for P-gp inhibition in the liver or in the gut can be estimated based on its estimated local exposure and the IC₅₀.

Determination of clearance mechanism and CYP phenotyping

The potential for a compound to be a victim of a DDI with a co-medication is greatly reduced if there are multiple clearance mechanisms, particularly involving metabolism by multiple CYP enzymes. This should be a consideration if the therapeutic window is low or if the clinical/marketing disadvantage of the interaction would be significant. Quantitative assessment of multiple clearance mechanisms and phenotyping of CYP metabolism should be established for any candidate drug. Methods for phenotyping the individual CYP enzymes responsible for a drug’s metabolism include the use of (1) specific chemicals or antibodies as specific enzyme inhibitors, (2) individual human recombinant CYPs, and (3) a bank of human liver microsomes characterized for CYP activity prepared from individual donor livers. At nomination of a candidate drug, human phenotyping work should include all eight major CYPs (1A2, 2B6, 2C8, 2C9, 2C19, 2D6, 3A4/3A5), and be followed by other enzymes if necessary. Another way to investigate clearance mechanisms in vitro is a so-called ‘fractional Clint’ assay. In such an assay the rate of parent disappearance is measured in human hepatocytes with and without the presence of an enzyme inhibitor. Most common is the addition of ketoconazole to block out the CYP3A4 contribution to metabolism. The remaining rate of metabolism in such an assay can be attributed non-CYP3A4 pathways like phase 2 enzymes, other CYPs or any other possible mechanism.

CYP induction mediated risk for DDI

Induction of specific CYP enzymes may not only change a drug’s metabolic profile but also have toxicological consequences as CYP enzymes are also involved in the metabolism and synthesis of important endogenous compounds. Although close collaboration with Safety Assessment (see Chapter 15) is needed to evaluate the full impact of CYP induction, it is DMPK’s primary responsibility to predict the DDI potential caused by CYP induction in man. This should be done during optimization with the use of HepaRG cells which are a good surrogate of primary human hepatocytes for AhR-mediated CYP1A induction and PXR- and CAR-mediated CYP3A4 and CYP2B6 induction. If higher throughput is needed during the optimization phase, the PXR reporter gene assay may be used to minimize PXR-dependent CYP3A4 induction liability.