Role of lipids in prediction of drug distribution: In vitro assessment of drug-lipid binding and implications for prediction modelling

  • Helen Musther

Student thesis: Phd


Descriptions of drug distribution are critical in the drug discovery and development process, with models available to describe and predict relevant parameters (Kp values) for use in physiologicallybased pharmacokinetic (PBPK) models. However, current prediction models have been observed to have limitations that could compromise their continued use. Parameters relating to drug-plasma binding and drug-lipid binding or partitioning are key features in predictions of Kp, however, these are subject to a number of assumptions that cannot easily be challenged due to the lack of supporting in vitro data. Initial comparisons of volume of distribution (Vss) predictions to available in vivo data, indicated that drugs for which distribution is poorly predicted display a range of physicochemical and blood-binding properties, suggesting a number of potential reasons for erroneous predictions. The hypothesis that lipid contributions are not accounted for in equilibrium dialysis measures of drug binding in plasma was challenged by development and validation of a novel assay utilising delipidated serum. Fraction unbound determinations in normal serum and delipidated serum showed a statistically significant 2.88-fold difference for imipramine (strongly basic, logPO:W 4.8), indicating that lipid binding is included when using this technique, and that correction models need not be developed for this contribution. However, for the weakly basic midazolam (logPO:W 3.15), no difference was observed, potentially suggesting mechanisms for lipid binding are related to factors other than just the lipophilicity. Collation of a literature database for plasma binding values for 4 drugs indicated high uncertainty in measured in vitro data, with this uncertainty reflected in subsequent Vss predictions, and care in assessing data incorporated into these models is advised. Explorations of the binding of drugs to individual phospholipid and unique combination cell mimic liposomes was undertaken by development of surface plasmon resonance (SPR) methods, and application of different models for data fitting. The acidic binding phospholipid affinity constant (Ka,AP) was determined with phosphatidylserine for 15 drugs, with the resultant values showing a poor correlation with the values derived from the commonly used blood cell binding calculation. Use of the SPR derived Ka,AP values in predictions of Vss led to an increase in precision with an observed AAFE of 3.14 compared to 3.37 using the blood cell binding calculated values. Binding of 14 drugs to phosphatidylcholine liposomes indicated that basic drugs bind more strongly to the acidic phospholipids, however, zwitterions can behave differently. A case study comparing terfenadine and fexofenadine highlighted the impact of drug structure differences on the lipid-binding capability. Comparisons, conducted here for the first time, of SPR derived binding (Kd) and partitioning (Kp) parameters for phosphatidylcholine, to the 0.3P+0.7 term, used to describe neutral phospholipid binding in the current tissue distribution prediction models, did not indicate a strong correlation. The determination of capture corrected binding isotherms for individual phospholipids and combination liposomes for 3 drugs indicated differences in capacity and affinity between the lipids, including differences between acidic phospholipids, which are not accounted for in the current prediction models. Combinations of Kd and Bmax (binding capacity) indicated that the use of only phosphatidylserine and phosphatidylcholine would not allow the estimation of cell mimic binding although, in the binding isotherms, cell mimic liposomes without cholesterol appeared to behave similarly to neutral phospholipid liposomes. The addition of cholesterol in the cell mimic resulted in a decrease in the binding of drug. Expansion of the use of the cell mimic liposomes to 17, predominantly acidic, uptake transporter substrates gave a poor correlation to observed l
Date of Award1 Aug 2020
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorDavid Hallifax (Supervisor) & Amin Rostami-Hochaghan (Supervisor)


  • Drug distribution
  • Physiologically Based Pharmacokinetic modelling
  • Volume of distribution

Cite this