Plasma protein binding - Journal of Pharmaceutical Sciences

Plasma Protein Binding: From Discovery to Development TONIKA BOHNERT, LIANG-SHANG GAN Preclinical PK & In Vitro ADME, Biogen Idec Inc., Cambridge, Massachusetts 02142 Received 5 March 2013; revised 25 April 2013; accepted 25 April 2013 Published online 24 June 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23614 ABSTRACT: The importance of plasma protein binding (PPB) in modulating the effective drug concentration at pharmacological target sites has been the topic of significant discussion and debate amongst drug development groups over the past few decades. Free drug theory, which states that in absence of energy-dependent processes, after steady state equilibrium has been attained, free drug concentration in plasma is equal to free drug concentration at the pharmacologic target receptor(s) in tissues, has been used to explain pharmacokinetics/pharmacodynamics relationships in a large number of cases. Any sudden increase in free concentration of a drug could potentially cause toxicity and may need dose adjustment. Free drug concentration is also helpful to estimate the effective concentration of drugs that potentially can precipitate metabolism (or transporter)-related drug–drug interactions. Disease models are extensively validated in animals to progress a compound into development. Unbound drug concentration, and therefore PPB information across species is very informative in establishing safety margins and guiding selection of First in Human (FIH) dose and human efficacious dose. The scope of this review is to give an overview of reported role of PPB in several therapeutic areas, highlight cases where PPB changes are clinically relevant, and provide drug metabolism and pharmacokinetics recommendations in discovery and development settings. © 2013 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci 102:2953–2994, 2013 Keywords: protein binding; albumin; alpha 1-acid glycoprotein; blood–brain barrier; drug interactions

INTRODUCTION TO FREE DRUG THEORY—WHY IS PLASMA PROTEIN BINDING IMPORTANT? High concentration of proteins in plasma and the propensity of numerous drugs to bind to them have led drug development groups to recognize the importance of plasma protein binding (PPB) in modulating the effective drug concentration at pharmacological target sites. Free drug theory (FDT) is Abbreviations used: BSA, bovine serum albumin; Cl, clearance; CNS, central nervous system; DDI, drug–drug interaction; DMPK, drug metabolism and pharmacokinetics; DSA, dog serum albumin; FDT, free drug theory; HSA, human serum albumin; HTS, high-throughput screen; ICF, intracranial fluid; i.v., intravenous; IVIVC, in vitro–in vivo correlation; NCE, new chemical entity; NSAIDS, nonsteroidal anti-inflammatory drugs; NTI, narrow therapeutic index; PD, pharmacodynamics; Pgp, P-glycoprotein; PK, pharmacokinetics; PK/PD, pharmacokinetics/pharmacodynamics; p.o., oral; PPB, plasma protein binding; RBCs, red blood cells; RSA, rat serum albumin; t1/2 , half-life; TDM, therapeutic drug monitoring; Vdss , volume of distribution. Correspondence to: Tonika Bohnert (Telephone: +617-914-7010; Fax: +617-679-2000; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 102, 2953–2994 (2013) © 2013 Wiley Periodicals, Inc. and the American Pharmacists Association

being increasingly used to explain pharmacokinetics/pharmacodynamics (PK/PD) relationships1 —FDT states that in absence of energy-dependent processes (uptake and efflux transporters, pH gradient), after steady state equilibrium has been attained, free drug concentration in plasma (extracellular space) is equal to free drug concentration in the tissues (intracellular space) and only the free drug in the tissues is available for the target receptor binding and therefore pharmacologic activity (Fig. 1). Hence, unbound drug in plasma should reflect the pharmacologically relevant concentration of unbound drug at the target site in the tissues. Total drug concentration in plasma may be significantly different than total drug concentration in tissues, but for a drug with rapid rate of permeation, intracellular unbound drug concentration equals the systemic unbound drug in circulation. It is thus a common concern that when free drug concentration of drug A increases due to plasma protein binding displacement by another strong-binding drug B or due to changes in endogenous protein levels in certain disease states, the increased free drug A’s concentration could potentially cause toxicity

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

2953

2954

BOHNERT AND GAN

covalent binding of compounds to nonspecific, endogenous proteins, is beyond the scope of this review and is not the intended goal.

MAJOR DRUG BINDING PROTEINS IN PLASMA

Figure 1. Free drug distribution throughout body water at equilibrium.

(due to more free drug available to interact with target receptors) and may need dose adjustment. In addition to target-mediated efficacy/toxicity, in case of metabolism-related drug–drug interactions (DDI), PPB information is also helpful to estimate unbound/effective concentration of drugs that mediate inhibition of drug disposition enzymes (metabolizing enzymes and transporters) that can potentially precipitate an adverse reaction. Disease models are extensively validated in animals to progress a compound into development. Unbound drug concentration, and therefore PPB information across species is very informative in establishing safety margins (based on efficacy, DDI, and tolerability information) and guiding selection of FIH dose and human efficacious dose. Note: Protein binding in general can be specific or nonspecific, depending on whether a compound binds to its intended target receptor (a protein or glycoprotein), to elicit a pharmacological response, referred to as specific binding or nonspecifically binds to endogenous proteins (such as albumin), that it encounters within physiological milieu. An example of specific and desired protein binding is the case of imanitib, designed mainly for chronic myelogenous leukemia (CML).2 Imanitib’s high degree of blood protein binding, also responsible for its long half-life (t1/2 ) is responsible for its efficacy in the treatment of CML. Examples of nonspecific protein binding will be those of nonsteroidal anti-inflammatory drugs (NSAIDS), antimicrobial drugs, central nervous system (CNS) drugs, whose efficacy may be compromised due to binding to plasma proteins. The scope of this paper is to give an overview of reported cases of nonspecific, noncovalent, reversible, protein binding, arising from hydrophobic interactions or hydrogen bonding to plasma proteins, and provide drug metabolism and pharmacokinetics (DMPK) recommendations, based on current understanding of this issue. Desired pharmacology due to specific protein binding to target receptor or undesired pharmacology/toxicity due to JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

Major drug-binding components in plasma are albumin (human serum albumin—HSA) (600 :M), "acid glycoprotein (AAG) (12–30 :M), lipoproteins ((globulin), and erythrocytes of which drug binding to albumin and AAG has been well studied and extensively published over the past several decades. Although HSA has eight binding sites, capable of binding to endogenous compounds as well as xenobiotics with varying affinities, two major sites of HSA are primarily involved in binding ligands and show a bias toward binding acidic drugs. The two HSA-sites are sometimes also referred to as warfarin-azapropazone site, Sudlow site I, and indole-benzodiazepine site, Sudlow site II.3 Despite its large size, HSA is not limited to plasma but is distributed extravascularly, as is AAG. AAG is an acute-phase protein, synthesized in the liver, shows preference to bind basic and neutral drugs, and whose levels could significantly change in disease/inflammation states. In certain disease states, although levels of AAG remained unchanged, yet its binding capacity was reduced.4 HSA levels have also shown to decrease in disease state, although the effect is much less dramatic than those involving AAG. AAG binding can also be dependent on gender, age, obesity, pregnancy, disease state, ethnicity,5–7 and even diurnal changes.4 The lower basal level of endogenous AAG (12–30 :M) than HSA (600 :M), as well as large fluctuations under various physiological and pathological conditions, results in higher variation of AAG from its basal level, as compared with HSA. The higher fluctuation of AAG has been implicated in causing clinical relevant DDI due to drug displacement from this high affinity, low capacity plasma protein (discussed in section SpeciesDependent PPB). For certain drugs such as felodipine, lipoproteins have also been proposed to be the binding component in plasma8 ; however, literature reports on lipoprotein binding and its implications in a clinical setting remain sparse.

SIMPLE, THEORETICAL CONSIDERATIONS OF PPB Plasma protein binding is believed to have a significant influence in the rate of drug diffusion between plasma and tissues (influx and efflux)9 and therefore influence clearance (Cl) and volume of distribution (Vdss ) of drugs. Effect of PPB on Cl is dependent on major route of Cl of the drug and in case of hepatically cleared drugs, on the liver extraction ratio. For renally extracted drugs, Cl due to glomerular DOI 10.1002/jps

PLASMA PROTEIN BINDING

Figure 2. Effect of PPB on hepatic Cl.

Figure 3. Effect of PPB on Vdss .

filtration and active secretion are dependent on fu as only unbound drug can be excreted via these processes. Most renally cleared drugs are hydrophilic and demonstrate low PPB. In case of drugs undergoing hepatic elimination, degree of PPB does not play an important role in hepatic Cl of high extraction drugs (also referred to as perfusion-limited or nonrestrictively cleared where Clhep ∼ Qh ). For a low extraction drug, however, changes in PPB will affect Clhep of the drug (Fig. 2). Volume of distribution, Vdss , which reflects the extent of drug distribution is determined by protein binding in plasma and in tissues (also discussed in section Tissue Binding) (Fig. 3). It can be observed that Vdss is directly proportional to fu and inversely proportional to fu,T . However, changes in fu does not always reflect change in fu,T and therefore it is challenging to accurately predict how changes in both fu and fu,T will affect the Vdss . Because PPB plays a partial role in determination of Cl and Vdss , two key parameters that determine the t1/2 of drugs (PK) as well as pharmacology (FDT—PD), it is very common to evaluate PPB as early as during compound optimization as a primary SAR tool or later stages to be incorporated into PK/PD models and explain disconnect in potency/efficacy-concentration relationships.10 In general, it is desirable to avoid highly plasma protein bound drug as small changes in plasma protein binding of a highly bound drug can lead to significant fluctuation in its free fraction. In a hypothetical situation shown in Table 1, it can be seen, that although in both cases, the percent change of plasma protein binding was same (3%), yet the imTable 1.

2955

pact on the fold change of fu was more significant in case of the highly bound drug A. Although assessment of PPB is routinely carried out both in discovery and development stages, yet, there is no common strategy as to how to use these data for compound selection or risk mitigation. Numerous applications in various therapeutic areas from literature demonstrate how different research teams have adopted strategies to modulate compound binding/affinity to HSA to maximize in vitro potency and attempted to improve in vivo efficacy of existing series or discover newer superior series. Thorough understanding of a compound’s physiological properties, localization of target receptor and how a compound reaches the target is vital in evaluating the role of fu (and fu,T , discussed later) in the compound’s pharmacological action. Potency versus free fraction (FDT) relationships are easier to establish, when pharmacological target resides in the extracellular space (ion channels and G-couple receptors), as cell permeability and active transport do not come into consideration, to add to the complexity of data interpretation. In case where the target receptors are localized in the intracellular space (enzymes and nuclear receptors), unambiguous correlations between free fraction in plasma and in vivo efficacy more challenging to establish. Major reason for this is the unavailability of robust methods to determine free drug concentration inside cell/tissue compared with total/free drug concentration in plasma and establish whether fu = fu,T , which is the major assumption of FDT. It must be borne in mind that the total drug concentration, measured in plasma or in tissues, may not be an accurate representation of pharmacologically active drug at the biological receptor site.

PPB OPTIMIZATION IN DRUG DISCOVERY Endocrinology/Metabolism Plasma protein binding optimization proved to be very useful in the development of depeptidyl peptidase DPP-IV inhibitors, for the treatment of type 2 diabetes mellitus. First-generation lead compound, although very potent, needed much higher than expected plasma concentration (based on its in vitro potency) for efficacy in preclinical model.11 To understand the disconnect between potency and efficacy, when the in vitro assay was performed with mouse and human serum, compound A (Fig. 4) showed a 32-fold potency shift, indicating very high PPB.

Hypothetical Example of Fold Change fu in Case of a High and Low Plasma Protein-Bound Drug

Drug A B

DOI 10.1002/jps

Bound (%)

fu

Bound After a Change (%)

fu,changed

Fold Change from Original fu

99 80

0.01 0.2

97 78

0.03 0.22

3 1.1


2956

BOHNERT AND GAN

Figure 4. Structures of DPP-IV inhibitors and glucokinase activators.

Introduction of fused heterocycles as in compound B, improved the PPB, as seen via the 11-fold in vitro potency shift in presence of serum. Introduction of an additional fluorine, as in compound C, led to a potent, orally active, DPP-IV inhibitor, which had much reduced PPB and where efficacy, in vitro potency, plasma concentration, and free fraction were well correlated.12 Another therapy, directed toward alleviating type 2 diabetes mellitus is via activation of the liver and pancreatic enzyme glucokinase. Increasing potency of the existing lead compound GKA22 (Fig. 4), via decreasing PPB and improving its in vivo efficacy by reducing unbound Cl (decrease intrinsic Cl), to get higher exposure, was the lead optimization approach the team adopted. The most potent glucokinase activator, GKA50,13 with an approximately twofold decreased PPB and an approximately threefold decrease of unbound Cl, was selected for future development. Hematology During development of potent, selective, orally available anticoagulant targeting the serine protease factor Xa, significant efforts to reduce PPB were directed during lead optimization of two series–the aminoisoxazole and the phenylcyclopropyl. Within the aminoisoxazole series, efforts were undertaken to

improve selectivity (factor Xa over trypsin and plasma kallikrein) of the original pyrazole-containing clinical candidate DPC423, which led to a series of potent compounds (e.g., DPC423 analog shown in Fig. 5),14 which showed good potency and selectivity but not acceptable anticoagulant property, attributed to the extremely high PPB of these compounds (>97%). Substitution of the a Me group of one of these potent compounds by–CH2 NMe2 , led to the second-generation lead compound razaxaban, which was potent, selective and had fourfold better anticoagulant activity than the analogs, due to an eightfold lower PPB. Within the phenylcyclopropyl series, increase in fu of the compounds due to introduction of polar group (–CONH2 ,–SO2 Me vs. CF3 ) in compounds B and C resulted in their greater than sevenfold anticoagulant activity compared with compound A. In general, PPB was been implicated in explaining why in vitro Ki s of human Factor Xa (FXa) inhibitors in both series, in whole blood closely reflects in vivo potency (prevent generation of thrombin). Ki s, toward purified human FXa enzyme by similar analogs (one analog had 1% fu , whereas other one had 10% fu ) were 23-fold different but in vivo, the potency was only threefold different, which was reflected by in vitro Ki s determined in whole blood.

Osteoporosis Evaluation of potent and selective, carboxylic-acidcontaining indole antagonists of dual "v $3 / "v $5 intergrins (Fig. 6), revealed that although these indoles demonstrated subnanomolar in vitro affinity toward the "v $3 / "v $5 intergrins (compounds of class A), yet potency was significantly compromised due to high binding to HSA (96.5%–97.5% bound). PPB of the lead indole compound was significantly decreased (40% bound) by incorporating of polar N,Ndimethylaminomethyl group (compound B), while maintaining the subnanomolar target affinity.15

Figure 5. Structures of factor Xa inhibitors. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

DOI 10.1002/jps


Figure 6. Structures of "v $3 /"v $5 intergrins.

Oncology In evaluation of potent inhibitors of Bcl-XL proteins, targeted for cancer chemotherapy, it was observed that these potent aryl acylsulfonamide inhibitors (Fig. 7) (compound A) dramatically lost their potency (>100-fold) in vitro, in presence of HSA.16 . Because HSA-III site is the primary binding site of serum albumin that demonstrates high affinity for small anionic aromatic compounds, the aryl acylsulfonamide inhibitors were likely candidates to be binding to HSA-III site (also validated by potency deactivation in presence of HSA-III). A structure-based approach to decrease binding of lead series to HSA-III via introduction of a polar group (e.g., carbamates, amides, sulfones) on the site that binds to HSA-III, yielded the lead compounds (B and C) with a dimethylamine substitution, that showed efficacy both in vitro and in vivo. Similar efforts were utilized in PPB optimization of reversible MetAP2 inhibitors (Fig. 7).17 Original anthranilic acid sulfonamide compounds (like compounds A), although demonstrated high potency in biochemical assay, yet lost significant activity (10–100-fold) in a human cell proliferation assay. When the biochemical assay was repeated in presence

2957

of physiological concentration of HSA, similar shift in potency was observed as seen in the cellular assays. Structure-based design, based on NMR of lead compound bound to albumin and MetAP2 independently, led to introduction of polar tertiary amines B and C, on the lead series, which successfully improved cellular potency, while reducing HSA binding. In another case study, rigorous structure-based optimization of a potent CDK2/cyclinA inhibitor, cyclopropyl-pyrazolecontaining, PNU-292137 (Fig. 7), was directed toward reducing its very high binding to HSA (>99% bound).18 Introduction of a methyl group on a benzylic position and replacement of the napthyl group by a polar, lactam-phenyl group, on the original candidate, led to another potent compound that possessed both lower solubility and lower plasma protein binding (74% bound to HSA).

Obstetrics and Gynaecology Efforts to optimize human oxytocin receptor antagonists resulted in two second-generation lead compounds with comparable in vitro potency (compounds A and B) (Fig. 8). However, compound B showed approximately 25–40-fold lower efficacy than the other in vivo. This was also reflected in a >11-fold IC50 shift of compound B in presence of HSA, as compared with only a fourfold in vitro IC50 shift demonstrated by the more efficacious (in vivo efficacy model) compound A. Higher PPB binding of compound B compound compared with compound A was proposed to be the culprit for a 25–40-fold lower in vivo efficacy for the

Figure 7. Structures of Bcl-X, MetAP2, and CDK2/cyclinA inhibitors. DOI 10.1002/jps


2958

BOHNERT AND GAN

Figure 8. Structures of oxytocin antagonists.

former compound, although both demonstrated similar in vitro potency.19 Immunology It was observed during development of benzyloxy pyrrolidine-containing inhibitors of TNF-" converting enzyme (TACE), that although several compounds were highly active in vitro, their TACE-inhibitory activity significantly dropped in LPS-stimulated whole blood assay.20–22 Investigation of PPB of these inhibitors revealed very high PPB with fu < 1% for several of these compounds (e.g., compound A) (Fig. 9). PPB optimization, as assessed by measuring IC50 values in whole blood, resulted in a superior compound B with fu of approximately 4% and eightfold higher activity in whole blood assay. In case of rats, with turpentine-mediated inflammation, it was reported that the AAG levels were elevated almost 10-fold, leading to significantly increased protein binding of the neuromuscular drug, atracurium (Fig. 9). Because of this increase in its protein binding during inflammation state, the ED50 (effective dose of drug that causes 50% neuromuscular block) of atracurium was significantly increased,23,24 emphasizing the important role of protein binding in mediating resistance to neuromuscular drugs during systemic inflammation. Cardiology Comparison of in vitro and in vivo data of drugs that are known to cause torsades de points revealed a good

Figure 9. Structures of TACE inhibitors and atracurium. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

Figure 10. Structures of angiotensin II (Ang II) receptor antagonist drugs and digoxin.

correlation between their clinical free plasma concentration and in vitro hERG IC50 values.25 However, authors also pointed out that in vitro IC50 values may overestimate clinical concentration that caused QT prolongation by about 10-fold.25 Differences in PK (Cmax,unbound and AUCunbound ) and PPB between the preclinical species (viz dog) and humans, was attributed to be the major reason for the gap in translating in vitro and preclinical QT prolongation data to a clinical setting. The need to have a wide therapeutic margin (>30-fold of preclinical data) and routine comparison of PPB in preclinical species and in humans dosed with these types of drugs, which potentially can cause QT prolongation, was highly recommended. Crucial role of protein binding in PD of the angiotensin II (Ang II) receptor antagonist drug tasosartan and its potent metabolite enoltasosartan (Fig. 10) was observed.26 The plasma concentration and the AngII-receptor blockade by tasosartan showed a very nice correlation but a significant delay between the PK (plasma levels) and PD (AngII-receptor blockade) was observed in case of its potent enol-metabolite, enoltasosartan. Very high and tight binding (slow koff ) of enoltasosartan to plasma proteins was proposed to be responsible for its significant PK/PD disconnect. Digoxin (Fig. 10), a narrow therapeutic drug, used to treat congestive heart failure, represents an interesting case where, in spite of being approximately 25% plasma protein bound, it is highly informative to monitor free drug levels under certain special circumstances.27 In case of digoxin overdose (and toxicity), when patients (especially renal failure patients) are treated with antidigoxin FAB antibody digiband, and in volume expanded patients (uremia, liver disease, transplant recipients, hypertension, etc.), when there is interference from endogenous digoxinlike immunoreactive factors during serum digoxin DOI 10.1002/jps


Figure 12. Structures dioxopyrazolidines.

of

antibacterial

2959

3,5-

measurements, it is very useful to measure the unbound digoxin concentrations to monitor therapeutic progress.

optimized in a series of nitrofuranylamide-containing antituberculosis agents (Fig. 13).33 It was noted in the particular chemical series that was evaluated, where clog P values greater than 2.4 resulted in compounds that were all greater than 96% plasma protein bound, and incorporation of a terminal morpholine group significantly reduced PPB to 44%.

Anti-infective

In Vitro–In Vivo Correlation/PKPD Modeling

Importance of increasing fu in several anti-infective areas such as antiviral (hepatitis, HIV) antibacterial therapies has been repeatedly emphasized. With Abott’s HIV protease inhibitor A-80987 (Fig. 11) as a model compound, uptake and activity of the PI in HIV-1-infected human peripheral blood mononuclear cells was investigated in the presence of physiological concentration of AAG. It was nicely demonstrated that intracellular concentration of A-80987 (which shows rapid influx via passive diffusion and rapid efflux due to concentration gradient) was directly proportional to its extracellular unbound concentration and its anti-HIV activity was directly related to its intracellular concentration.28 Similar results were obtained with another Abott compound A-77003 (Fig. 11), which showed marked decreased anti-HIV activity in in vitro assays in the presence of physiologically relevant concentration of AAG29 and a Vertex compound VX-950 (Fig. 11), targeted for hepatitis C, which showed a 10-fold decrease in cellular potency, when performed in the presence of 40% HSA.30 A series of antibacterial 3,5-dioxopyrazolidines, with potent inhibitory effect toward peptidoglycansynthetic enzyme MurB (critical enzyme involved in bacterial cell wall biosynthesis) exhibited complete lack of antibacterial activity in the presence of 4% bovine serum albumin (BSA), attributed to very high albumin binding31 (Fig. 12). During validation of therapeutic efficacy with a series of cepham antibiotics in experimentally infected mice, a good correlation was observed between 1/ ED50 , AUCunbound /in vitro minimum inhibitory concentration values but not with the AUCtotal , emphasizing the important role of PPB in this infection model.32 PPB was one of the parameters that were

In predicting human PK/PD from preclinical species it is essential to consider interspecies differences in PPB, especially when the differences are significant, as in the case of etoposide whose fu varies greater than 10-fold across species. In projecting the FIH dose, it is important to factor in the fu to accurately predict a starting dose based on the desired unbound exposure (or Css,unbound ). Comparison of in vitro sandwichcultured rat hepatocyte data with in vivo plasma Cl in rat for a series of biliary excreted compounds revealed a good correlation between in vitro and in vivo intrinsic clearances, when Clint was calculated based on unbound plasma concentration of the compounds. PPB is proposed to play a major role in in vivo intrinsic biliary Cl of highly PPB drugs, as it is commonly believed that only protein-unbound drug can be transported across the canalicular membrane. It was therefore advised to calculate in vivo intrinsic biliary Cl, based on unbound plasma concentration and not total plasma concentration.34 Simulation of counterclock hysteresis due to effect compartment or clockwise hysteresis due to tolerance, were performed to predict the effect of time-dependent protein binding on effect versus drug concentration.35 The results from these simulation studies revealed that in both cases, no hysteresis was observed when free drug concentration was considered, suggesting that free drug provides a better reflection of PK/PD changes. In another case study with a potent, cortocotropin-releasing factor (CFR1)antagonist DMP696 (Fig. 14), it was observed that although the brain concentration of this compound were 150-fold higher than its unbound plasma levels, a very nice correlation was obtained between receptor occupancy in brain and free plasma drug levels.36 PK/PD modeling using unbound drug-plasma level versus brain CRF1 receptor occupancy yielded an

Figure 11. Structures of potential anti-infective drugs.

DOI 10.1002/jps


2960

BOHNERT AND GAN

Figure 13. Structures of cepham antibiotics.

concentration in plasma to contribute to the drug’s pharmacological effect.

RELEVANCE OF PPB IN CLINICAL SETTING Nonsteroidal anti-inflammatory drugs

Figure 14. Structure of DMP-696.

in vivo IC50 value that matched very closely to that determined in vitro. The unbound drug-plasma levels also matched very closely with drug concentrations in CSF. PPB optimization, based on thorough PK/PD assessment has been efficiently demonstrated in the case of matrix metalloprotease (MMP-13) inhibitors.37 Robust in vitro–in vivo correlation (IVIVC), linking plasma and synovial fluid (SF) drug levels were first established that suggested crucial role of PPB and cartilage accumulation in vitro, in the cartilage delivery of compounds in vivo. This in turn supported ranking of lead compounds based on PPB and cartilage accumulation in vitro, projecting efficacious drug concentration required in SF, and optimum dosage requirements in chronic pharmacology studies. In another case with 5HT3 antagonist lerisetron, active metabolites were initially implicated in the drug’s pharmacology but correlations between the in vivo EC50,unbound and in vitro receptor-binding affinity, along with rigorous PK/PD modeling of lerisetron and its hydroxyl L6-OH metabolite, revealed that the pharmacology is due to the parent drug and not its active metabolite, as was previously thought. Reason for this was proposed to be the high Cl of the metabolite, due to which it could not sustain optimal JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

Aspirin, naproxen, piroxicam, ibuprofen, phenylbutazone, indomethacin, flurbiprofen, ketoprofen, methyl salicylate, diflunisal (Fig. 16) are some of the wellknown NSAIDS used in the clinic for the treatment of inflammation-related diseases, for example, rheumatoid arthritis and chronic pain. All NSAIDS, with the exception of aspirin, are extremely highly plasma protein bound, which impacts not only their PK properties (Vdss , Cl, and t1/2 ), but also the duration and magnitude of their efficacy.38 Because NSAIDS have a small Vdss (0.1–0.15 L/kg), changes in fu will not have a significant impact on their Vdss (see section Tissue Binding for more detailed explanation) as observed in the case of naproxen whose Vdss increased marginally in severely renally compromised patients. Because NSAIDS are also low Cl, whose Cl ∼ fu Clint , their Cl is expected to be dependent on their fu (provided their Clint does not change). So, applying basic PK principles, overall an increase in fu is expected to cause a significant decrease in the t1/2 of the NSAIDS, most of which have small Vdss and low Cl, (Fig. 15). Although some positive correlations have been observed between total plasma levels of piroxicam, naproxen, and indomethacin, and their pharmacological effect, for the majority of the NSAIDS, no good correlation has been found between their respective antiinflammatory efficacy or toxicity and corresponding

Figure 15. Effect of fu on t1/2 . DOI 10.1002/jps


2961

Figure 16. Structures of some well-known NSAIDS.

plasma levels. Significant interpatient variability of fu and in the unbound plasma NSAID concentration, along with a relatively subjective, qualitative methods to assess anti-inflammatory efficacy in patients, are implicated to contribute significantly to this disconnect. Because of the variability in therapeutic response and susceptibility to toxicity, unexplainable by observed drug levels in plasma, it is not advisable to practice therapeutic drug monitoring (TDM) for these classes of compounds. However, the narrow safety margin for several of the NSAIDS, along with their high PPB, warrants thorough understanding of the implications of changes in free drug concentration of NSAIDS. Antiretroviral Therapy There is a wide range of protein binding encountered in antiretroviral (ARV) drugs, depending their class—the protease inhibitors and nonnucleoside reverse transcriptase inhibitors are highly bound, whereas nucleoside reverse transcriptase inhibitors are not.39 Some ARV drugs such as ritonavir/ saquinavir are bound to AAG, whereas some, such as efavirenz/nevirapine, are bound to albumin. Success of ARV therapy is dependent upon the ability of the unbound drug to get into the HIV viral-load-bearing cells, so, protein binding will affect the efficacy of ARV drugs. AAG has been found to be the plasma protein that was responsible for extremely tight binding and reduced antiviral potency of several HIV protease inhibitors such as Searle/Monsanto’s SC-52151 and Vertex’s KNI-272 (Fig. 17). SC-52151, a urea-based protease inhibitor, although possessed plasma concentrations much higher than the in vitro IC90 for viral replication, no antiviral activity was observed in the Phase I/II study.40 Investigation into the mechDOI 10.1002/jps

Figure 17. Structures of some ARV drugs.

anism for the failure of its anti-HIV activity revealed that in the presence of physiological AAG concentration, cellular uptake of SC-52151 was remarkably compromised, which in turn, resulted in an approximately 20-fold increase in its EC95 in presence of AAG. KNI-272, a peptide-based protease inhibitor, was found to be 98%–99% bound, predominantly to AAG. Its in vitro anti-HIV activity was markedly reduced in the presence of fetal calf serum (FCS)—approximately 15–25-fold in the presence of 50% FCS and 25–100-fold in the presence of 80% FCS, which suggested that significantly higher KN-272 levels in plasma maybe required in clinical trials compared with those predicted from in vitro studies, performed routinely in the presence of 10%–15% FCS.41 In the similar in vitro assay, 2,3dideoxyinosine, which has low protein binding, did not show any potency shift in the presence of high concentration of FCS. Therapeutic efficacy of ritonavir (Fig. 17) is dependent on its maintaining free, trough, and plasma levels higher than EC90 for HIV replication. Clinical dosages of ritonavir is adjusted, JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

2962

BOHNERT AND GAN

taking into consideration its 90% protein binding.42 It is highly recommended to account for plasma proteinadjusted potencies for HIV drugs for their therapeutic efficacy43,44 —in calculating inhibitory quotients (IQ) of ARV drugs, it is strongly suggested to correlate plasma trough unbound drug levels to the measure of susceptibility of HIV strain to the particular drug. In general, the • wide range of protein binding of the known ARV drugs, • large interindividual variability of PB in patients undergoing ARV therapy, • large interindividual variability of exposure of ARV drugs in patients, • although recently target plasma trough concentrations have been defined for several protease inhibitors toward wild-type HIV, but these ranges are very wide and there is still lack of widely accepted target plasma concentration (trough vs. Cmax or IQ or Cmin /Cmax ) to inhibit HIV strains of different susceptibilites to ARV drugs, • lack of sufficient data to enable clinicians to correlate intracellular drug concentration versus efficacy for optimal viral suppression, and • narrow therapeutic index (NTI) for majority of the ARV drugs have all contributed to an increased need to understand and establish free plasma drug concentration— clinical efficacy correlations, in HIV patients,45–47 before significant efforts are invested in free drug monitoring of antiretroviral drugs. Anticancer Therapy Several factors pertaining to PPB, discussed in case of ARV therapy can be applicable to cancer therapy in a similar fashion as role of PPB has been proposed to be important in anticancer therapy. Degree of PPB of anticancer drugs demonstrates a wide range as well [no binding as in bleomycin (Fig. 18), to highly bound, such as vinblastine]. In a study with acridine derivatives and related antitumor antibiotics, in vitro data strongly suggested that free drug, rather than total drug, is more representative of the cytotoxic potential in tumor cells.48 Usually in cancer patients, AAG levels increase, whereas HSA levels decrease, so it can be expected that anticancer drugs which bind to AAG should be monitored closely, in the likelihood of decrease in efficacy due to less unbound drug availability. For example, in case of tyrosine kinase inhibitor imatinib (Fig. 18), which primarily binds to AAG, several publications suggested good correlation of unbound imatinib with its adverse side effects and correlated both total and unbound trough imatinib concentrations to its therapeutic response in patients with CML and gastrointestinal stromal JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

Figure 18. Structures of some anticancer drugs.

tumor (GIST).49 However, unambiguous confirmation of unbound drug levels and correlation to efficacy and toxicity is yet to be made. Similarly, in case of sorafenib (Fig. 18), a multikinase inhibitor observed to be very highly bound (0.3% fu , primarily to albumin), in both cancer patients and in healthy volunteers, unbound sorafenib levels were not correlated with patient characteristics or laboratory values.50 For the EGFR-inhibitor gefitinib (Fig. 18), which is highly bound to plasma protein (>97% bound and primarily to HSA), although fluctuation of AAG levels in cancer patients have been implicated to be the cause of significant interpatient variability in efficacy and toxicity, a clear correlation has yet to be established.51 Impact of protein binding on pharmacodynamics of (-secretase inhibitor RO4929097 (Fig. 18) was assessed using an in vitro Notch cellular assay and high binding to AAG abolished in vitro Notch-inhibitory activity of RO4929097.52 In clinic, RO4929097 fraction unbound (fu ) exhibited large intra- and interindividual variability and monitoring of unbound RO4929097 was highly recommended, as fluctuations of total RO4929097 levels was anticipated in case of coadministered drugs or AAG fluctuations in patients, but not of the pharmacologically active unbound drug. It is proposed that53 PPB is important only in very few cases of anticancer therapy where: • Drugs demonstrate protein-concentrationdependent binding: epipodophyllotoxins etoposide and teniposide (Fig. 19) are greater than 95% bound to plasma proteins. A wide interpatient variability in free etoposide and DOI 10.1002/jps


2963

• When formulation excipients modulate free drug concentration: for doxorubicin and vincristine (Fig. 21), which are liposomal-formulated and paclitaxel, which is encapsulated in nonionic surfactant Cremephor EL. In both these cases, it is the free drug that is released from the formulation, is the mediator of efficacy. • Drugs that undergo metabolic/chemical interconversion: camptothecin (Fig. 22) undergoes interconversion between its lactone and carboxylate forms and this equilibrium has been proposed to be dependent on HSA binding of the two species. Figure 19. Structures of epipodophyllotoxins etoposide and teniposide.

Figure 20. Structures of cisplatin, carboplatin, and UCN-01.

In general, due to poorly defined relationships between efficacy and systemic drug exposure in chemotherapy and significant interpatient variability in therapeutic response and toxicity, total (not unbound) drug monitoring is used for few agents like methotrexate, carboplatin, and 5-FU (Fig. 22). But, it may be very informative to routinely monitor unbound drug in chemotherapy, as fu values are highly variable amongst cancer patients (who also have elevated levels of AAG) and fu maybe better predictors of therapeutic response and toxic side effects of anticancer drugs, which are notorious to have narrow therapeutic margin, highly variable PK. CNS Disease Therapy

tenoposide concentration (5%–50% free drug concentration) has been reported and it has been suggested that in case of both these anticancer drugs, their unbound plasma concentration correlated better with hematologic toxicity rather than total drug concentration, although a similar correlation between unbound drug concentration and antitumor efficacy is yet to be established. Nonetheless, due to a better toxicity prediction from unbound rather than bound drug concentration, knowledge of unbound drug concentration in these cases is very helpful. • Drugs that bind irreversibly or covalently (or near covalently): cisplatin, carboplatin, and UCN-01(Fig. 20) are drugs for which unbound drug concentration is measured routinely. For cisplatin, the total and free drug concentration versus time profiles are not parallel, whereas for carboplatin, the PPB increases from 10% to 90% within 24 h, warranting monitoring free drug levels in both these cases for optimal safety and efficacy. UCN-01 shows covalent, very extensive binding to AAG, with a very low Cl, in the clinic. This is in sharp contrast to what was expected from preclinical studies (lower binding and higher Cl in animals) and it was deemed important to monitor the extremely low unbound UCN-01 concentration and its very long exposure in humans. DOI 10.1002/jps

Drugs that target the CNS have been highlighted to be extremely sensitive to PPB and PPB displacement situation. Historically, for CNS targets, maximizing compound potency along with the ratio of total brain concentration to total plasma concentration (Kp ) to maximize efficacy, has been a common strategy. However, several examples of compounds that demonstrate high target affinity, high plasma concentration and high brain/plasma (B/P) ratio, and yet which fail to show efficacy, warranted extreme caution in compound optimization based on the total B/P ratio. Acetylcholinesterase inhibitor KA-672 (Fig. 23) presents one such example where it failed to demonstrate any pharmacologic response, in spite of achieving total brain concentrations higher than its in vitro IC50 values. Investigation of KA-672 concentration in CSF suggested that unbound concentration in brain was >100-fold lower than in vitro IC50 , which was

Figure 21. Structures of doxorubicin and vincristine. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

2964

BOHNERT AND GAN

Figure 22. Structures of camptothecin, methotrexate, and 5-fluorouracil.

attributed to its inability to demonstrate any pharmacologic response.54 However, a high brain tissue binding in itself is not a liability for CNS drugs as has been suggested by independent analyses by Maurer55 of 32 marketed CNS drugs in mice and by Summerfield56 of 50 marketed CNS drugs in rats, where a greater than 1000-fold difference in Kp has been observed between some of the drugs that demonstrated lowest (e.g., sertraline, chlorpromazine, fluphenazine, trifluoperazine) (Fig. 23) and highest (e.g., meprobamate, caffeine, ethosuximide, gabapentin, sumatriptan) (Fig. 22) brain unbound fractions. It must be recognized that BBB transporters, permeability (dependent on a compound’s intrinsic permeability), and protein binding to blood versus brain all play equally important role in the rate and extent of drug uptake into CNS.56 Increase in nonspecific binding in brain tissue is considered a driving force to overcome effect of efflux transporters and so compounds which are Pgp substrates can still penetrate the brain because partition into brain tissue acts as a driving force to provide a significant drug concentration gradient at the BBB to overcome effects of efflux transporters. Usually with increase in lipophilicity, not only does CNS penetration increases but also does nonspecific binding to brain tissue, resulting in lesser unbound/pharmacologically active drug available in brain for efficacy. This was nicely illustrated via PK/PD relationship between voltagegated sodium channel in vitro and efficacy in rodent electroshock-seizure model, for a set of struc-

turally diverse anticonvulsants.57 In vitro affinity (Ki ) of the compounds toward Na (V)1.2 channels and the unbound brain concentration (Cu,brain ) required for anticonvulsant efficacy showed a nice correlation. However, with increasing lipophilicity, there was an increase in both, sodium channel blocking potency as well as brain tissue nonspecific binding, where efficacy of the compounds started decrease, due to such a high nonspecific binding, in spite of maintaining potency. Total brain–plasma concentration ratio Kp maybe a useful measure of efficient drug permeation/ distribution into the brain, but this does not truly reflect a drug’s ability to elicit pharmacologic response at the target site. Hence, targeting to achieve high values of Kp in CNS drugs may not be ideal as such drugs may possess high nonspecific binding to brain tissue and will not maintain optimal active-site concentration at the desired receptor/protein. It has been quickly recognized that total brain concentration, which reflects compound in extracellular fluid, intracellular fluid, and bound to lipids and proteins in brain is not reflective of active/ unbound compd at the target pharmacological site of action, which most often is the intracellular ICF (Fig. 24).58–61 Hence significant efforts have been invested to understand Kp,free (also referred to as Kp,uu ) (Fig. 25) rather than Kp , which is the ratio of unbound brain drug concentration to unbound plasma drug concentration. It is important to note that although brain tissue binding does influence total drug concentration in the

Figure 23. Structures of some CNS drugs. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 9, SEPTEMBER 2013

DOI 10.1002/jps


2965

Figure 27. Unbound exposure–receptor affinity ratio.

Figure 24. Schematic of blood–brain–CSF–ICF barriers.

Figure 25. Relationship of Kp and Kp,uu .

brain and hence Kp , it does not affect the unbound brain concentration of drugs, Kp,uu . It is also clear from Figure 25 why Kp does not always accurately represents Kp,uu , as Kp depends on other parameters such as fu, blood and fu,brain , and a low Kp can be due to a low Kp,uu , a low fu,blood , or a high fu,brain .59,61 Kp,uu is dependent on passive diffusion, uptake and efflux transport Cl, brain interstitial bulk flow, and brain metabolic Cl of a drug (Fig. 26), and is independent of protein binding in blood and in brain. Kp,uu value >1 signifies active influx into the brain,