Determining the penetration of drugs across the blood–brain barrier is a significant challenge in central nervous system drug development. The use of a mechanistic physiologically based pharmacokinetic model can predict drug exposures in the brain without needing in situ drug measurements. Clin Cancer Res; 23(24); 7437–9. ©2017 AACR.
See related article by Li et al., p. 7454
In this issue of Clinical Cancer Research, Li and colleagues (1) report on the use of an integrated, quantitative, clinical pharmacology approach incorporating clinical trial data and in vitro–in vivo extrapolation (IVIVE) coupled with physiologically based pharmacokinetic (PBPK) modeling to assess the penetration of AZD1775 across the human blood–brain barrier (BBB) in patients with glioblastoma.
Despite advances in drug discovery and development of targeted therapies, only modest progress has been made on improving outcomes in patients with central nervous system (CNS) tumors. Drug delivery remains one of the major challenges for developing effective therapies for brain cancer. The human brain is a complex and difficult-to-access compartment, with the BBB serving as a physical and physiologic obstacle for delivery of drugs to the CNS. Developing a candidate drug that can overcome this sophisticated protection system to achieve optimal concentrations at the pharmacologic target site in the brain remains a priority. CNS drug development is hampered not only by the dynamic influx/efflux transporter system at the BBB but also by the relatively poor predictive validity of preclinical models, the lack of accepted biomarkers and/or surrogate measures of drug activity/response, and the limited effective strategies to assess drug exposures in the brain. PBPK modeling of the CNS provides the opportunity to predict relevant drug concentrations at the therapeutic target site, and IVIVE linked with PBPK is a strategy to quantitatively bridge in vitro and in vivo data to explore the key mechanisms dictating the pharmacokinetics and BBB penetration of the drug (2).
In an effort to improve the predictive abilities of identifying anticancer agents that will sufficiently penetrate into the brain upon systemic administration, specifically brain tumor tissue, Li and colleagues (1) developed a whole-body PBPK model that utilized not only observed clinical plasma and brain tissue pharmacokinetic data but also the drug's physicochemical properties (lipophilicity, pKa, etc.), protein and tissue binding, and metabolism and transport data obtained in vitro. This model can also account for a healthy brain tissue microenvironment pH range of 7.0 to 7.8 compared with the more acidic brain tumor microenvironment (pH range, 5.8–7.2; refs. 3–5). This pH gradient can change the passive transcellular permeability of drugs across the BBB (based on that molecule's pKa and ionization state), as well as active transporters. In addition, the brain compartment of this whole-body PBPK modeling was structured into a four-compartment mechanistic brain model (6) that describes reversible drug movement from (i) blood to heterogeneous (healthy and tumor) brain; (ii) blood to cranial cerebrospinal fluid (CSF); (iii) brain to cranial CSF; (iv) cranial CSF to spinal CSF, and then both CSF compartments back to blood (Fig. 1).
The small-molecule Wee1 inhibitor AZD1775 (MW 500.6 g/mol) is a BCS class II compound [highly permeable (logP = 3.37), poorly soluble] that is also a chemo- and radiosensitizer to augment treatment of glioblastomas. It is also a dual substrate of ABCB1 and ABCG2, both of which are highly expressed in the BBB and blood–tumor barrier (BTB). To develop, and then validate, this novel whole-body PBPK model, 20 patients with confirmed glioblastoma were enrolled on a phase 0 clinical trial where they received a single oral dose of 100 mg (n = 4), 200 mg (n = 4), or 400 mg (n = 12) of AZD1775. Serial plasma samples and in situ brain tumor resections were collected postdose, with AZD1775 quantitatively measured in both plasma and brain tumor samples from all patients. Using these clinical data, along with in vitro metabolism and transporter data, a whole-body PBPK model was developed. AZD1775 was rapidly absorbed (no lag time), widely distributed (Vdss = 19.2 L/kg), and shown to be cleared through numerous means, including renal (6 L/hour), hepatic (total human liver microsome activity 197 pmol/min/mg; Km = 4.94 μmol/L), and an intestinal intrinsic clearance of 17 μL/min (1).
Pharmacokinetic modeling and simulation has been utilized for decades to predict a drug's disposition following new dose schema or patient population based on prior data. The use of PBPK modeling provides gravitas to simulation results based on inclusion of a plethora of additional data, such as relevant drug metabolizing enzymes and transporters for that drug, as well as relevant pharmacogenetics differences. In this study by Li and colleagues (1), their whole-body PBPK model with mechanistic four-compartment brain disposition was able to successfully capture the measured observed clinical data. The fact that such high concentrations of AZD1775 were measured clinically from the resected brain tumor samples supports evidence that the BTB is altered, likely due to pH where active ABCB1 and ABCG2 efflux clearance is reduced, and with passive transcellular transport and OATP1A2-mediated active uptake unaffected, an overall net influx allows more drug to penetrate into the brain tumor.
The most relevant finding from this study was that slight decreases in pH in the brain tumor microenvironment specifically alters active efflux transporters based on the drug's pKa and ionizability. In the case of AZD1775, which has a pKa of 6.0, the more acidic brain tumor has more ionized AZD1775 in the basolateral (brain tumor) side of the endothelial cells. Therefore, less is unionized and able to be active and/or passively transported toward the apical (blood) end, essentially “trapping” AZD1775 in the tumor microenvironment. On the basis of these inherent properties, different drugs can be used in this brain penetration model to provide a more realistic prediction of BBB and BTB penetration.
The use of gadolinium-enhanced PET, which is a BCS class III drug (low permeability, high solubility) that is largely unionized, may not be the best predictor of brain and/or brain tumor penetration of BCS class II drugs such as AZD1775, which is ionizable in vivo (pKa 6.0), highly permeable, and poorly soluble. This PBPK model can incorporate drug properties with a range of physicochemical properties, binding properties, and affinities for drug-metabolizing enzymes and transporters to more accurately predict the extent of brain penetration compared with current predictors (PET) and without the need for in situ measurements.
Preclinical models of brain penetration are not always predictive of human brain penetration. In the case of AZD1775, preclinical models suggested very poor brain uptake (7). Coupled with the fact that AZD1775 was shown to be substrates for ABCB1 and ABCG2 transporters, both of which are highly expressed in the basolateral-to-apical BBB and BTB, AZD1775 was predicted to penetrate poorly into human brain. The use of PBPK modeling, which incorporated the physicochemical properties of AZD1775 into a physiologically relevant model, allowed a more realistic depiction of how changes in pH in the tumor microenvironment would impact the transport of the drug across the BBB versus BTB. The in situ brain measurements of AZD1775 confirmed the ability of this drug to sufficiently penetrate into brain tumor (glioblastoma) at a therapeutically relevant exposure.
Li and colleagues' modeling approach (1) demonstrates the value of IVIVE–PBPK as a prediction tool throughout the development process of CNS-targeting drugs. Moreover, IVIVE–PBPK simulations in drug development can be used for lead optimization and/or candidate evaluation during early clinical drug development, drug–drug interaction potential prediction, in vivo pharmacokinetic prediction, and so on, thus providing valuable insight to support decision-making in the drug development pipeline. Advances in imaging, analytic, and computational approaches will continue to evolve and increase the predictive power of such methods, allowing for more precise and accurate predictions of drug penetration across the BBB to enable personalized medicine for CNS tumors.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organization imply endorsement by the U.S. Government.
Conception and design: C.H. Chau
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.H. Chau
Writing, review, and/or revision of the manuscript: C.J. Peer, C.H. Chau, W.D. Figg
This work was supported by the Intramural Research Program of the Center for Cancer Research, NCI, NIH.