Optimization of Incubation Time for Accurate Protein Binding Assessment in Biological Matrices Using Equilibrium Dialysis

1. INTRODUCTION

Drug Protein Binding (PB) in biological matrices, such as plasma, is a simple yet intricate phenomenon, particularly when quantifying compounds with low unbound fractions. While Plasma PB (PPB) itself is not typically optimized during drug development [1-3], accurately assessing PB in bioassay media, plasma, tissues, or disease-specific cell lines is critical for extrapolating in vitro potency, in vitro–in vivo correlations (IVIVCs), dose projections for animal and human studies, and estimating safety margins.

The accurate determination of the % unbound of a drug candidate is crucial in the context of the free drug hypothesis [4-6]. At steady-state, unbound drug concentration in tissues is assumed to be in equilibrium with plasma; however, concentrations in tissues could be higher or lower depending on active transport mechanisms [7]. Therefore, it is crucial to assess PB values in multiple biological matrices (tissues, cell lines, and plasma) and animal species for a comprehensive understanding of tissue and plasma distribution. In oncology, PB studies of investigational drugs in tumor cell lines and tissues may provide insights into pharmacokinetic/pharmacodynamic (PK/PD) relationships and support the optimization of therapeutic strategies for disease-specific targets [8]. Tumor cell lines mimic a complex tumor microenvironment, which has high protein content, an altered extracellular matrix, and variable transporter expression profiles. These factors can affect drug binding, drug retention, and cellular drug availability. Likewise, PB in animal tissues, such as liver and brain, differs from plasma due to its unique protein compositions and physiological barriers. Measuring the % unbound in such matrices enables more accurate estimation of intracellular drug exposure, interpretation of preclinical PK data, and prediction of human tissue distribution.

Among the methods used to determine PB in drug discovery, the “workhorse assay” for PB is equilibrium dialysis, performed with Rapid-Equilibrium Dialysis (RED) or High-throughput Dialysis (HTD) devices using undiluted plasma [9]. Given the increasing demand for faster data turnaround in drug-discovery settings and also considering the logistical challenge of total assay run time, equilibrium dialysis for PB estimations performed with short incubation times, such as 2 to 6 hours, is not uncommon [9-11]. An inadequate incubation time that does not ensure complete dialysis equilibrium will not yield accurate % unbound values.

Therefore, it is crucial to determine the appropriate incubation time for dialysis equilibrium in an extended time-course PB study. This study conducted time-course PB assays using diverse biological matrices, including human plasma, cell culture medium containing 10% Fetal Bovine Serum (FBS), and rat brain homogenate, with 15 marketed compounds representing structurally diverse therapeutic classes and varying lipophilicity. Incubation time-course data from these first 15 compounds were later used to study the PB behavior of two additional highly lipophilic drugs, nelfinavir and ritonavir, in two matrices– human/mouse tumor cell lines and mouse tissues. These experiments aim to standardize current PB assays using ED and provide relevant data to drug discovery teams to support their PB assessment strategies.

2. MATERIALS AND METHODS

2.2. Methods

2.2.1. Drug-selection Criteria

Seventeen marketed compounds were selected to represent different drug classes with varying acidic/basic properties, LogD_7.4, and PB characteristics (Table 1). Of these, ritonavir and nelfinavir were highly lipophilic drugs, having high and very high PB characteristics, respectively.

Table 1.

List of compounds and relevant physicochemical properties.

Compound	Structure	Molecular Weight (Free Base)	Class	pKa	LogD (pH 7.4)
Axitinib		386.47	Basic	4.85 (B1), 11.93 (A1)	3.57
Clozapine		326.83	Basic	3.96 (B1), 7.62 (B2)	2.78
Dasatinib		488.01	Basic	3.01 (B1), 6.93 (B2), 10.65 (A1)	2.98
Dexamethasone		392.47	Neutral	NA	1.99
Erlotinib		393.44	Basic	5.37 (B1)	3.21
Fluconazole		306.28	Basic	2.56 (B1), 2.94 (B2), 11.01 (A1)	0.52
Fluphenazine		437.53	Basic	3.57 (B1), 7.63 (B2)	3. 75
Imatinib		493.62	Basic	1.93 (B1), 3.16 (B2), 4.12 (B3), 7.89 (B4)	2. 42
Ketoconazole		551.48	Basic	3.35 (B1), 6.38 (B2)	3.59
Lapatinib		581.06	Basic	4.57 (B1), 5.87 (B2)	3.81
Propranolol		259.35	Basic	9.58 (B1)	1.18
Sildenafil		474.58	Basic	6.5 (B1), 9.2 (A1)	2.81
Sulfamethoxazole		253.28	Acidic	1.55 (B1), 5.62 (A1)	-0.62
Sunitinib		398.48	Basic	8.89 (B1)	2.43
Tolbutamide		270.35	Acidic	5.52 (A1)	0.32
Nelfinavir		567.79	Acidic	9.32 (A1)* 8.18 (B1)*	4.72*
Ritonavir		720.95	Acidic	2.84 (B1)*	5.22*

Open in a new tab

Note: B1, B2, B3, and B4 denote different basic sites within the same molecule; A1 denotes the first acidic site (if more than one is present). pKa for fluconazole and sildenafil were obtained from the study by Corrêa et al. 2012 [38] and Akula & Lakshmi 2018 [39], respectively. *Data sourced from https://go.drugbank.com.

2.2.2. Preparation of Matrices

The study was conducted in 2 parts. The first part determined the optimal incubation time for 15 marketed drugs (axitinib, clozapine, dasatinib, erlotinib, fluphenazine, imatinib, ketoconazole, lapatinib, sildenafil, sunitinib, tolbutamide, dexamethasone, fluconazole, propranolol, and sulfamethoxazole) using human plasma, cell culture medium containing 10% FBS, and rat brain homogenate. In the second part of the study, the PB properties of nelfinavir and ritonavir were determined using tumor cell lines and mouse tissues (brain, intestine, liver, lungs, adipose, muscle, skin, and plasma). The animals were housed in groups of 2-3 animals per cage in a climate-controlled rodent facility with a 12 hours light/dark cycle. The animals had ad libitum access to water and food. Animals were euthanized in an anesthesia chamber using carbon dioxide (CO₂) at a 50% gas exchange rate per minute for 5 minutes. Following confirmation of respiratory cessation, physical methods (cervical dislocation, cardiac puncture, or exsanguination) were applied to ensure death. Tumors and tissues were collected aseptically using sterile scissors and forceps, immediately placed into pre-labeled cryovials, snap-frozen in liquid nitrogen, and stored at -80°C until further processing.

Rat brain homogenates were prepared by adding 3 volumes (mL) of PBS (pH 7.4) to 1 volume (g) of whole brain tissue, followed by homogenization using a tissue homogenizer. Mouse tissue and tumor homogenates were prepared using 100 mM sodium phosphate and 150 mM NaCl buffer (PBS), with tissue-to-buffer ratios of 1:3 for brain, liver, adipose, and tumor tissues; 1:7 for skin, intestine, and lung; and 1:9 for muscle. Homogenization was performed under low-temperature conditions. All homogenates were stored at -80°C and thawed at room temperature immediately before use.

Pooled human plasma (pH adjusted) was obtained from healthy Asian volunteers from a domestic hospital under ethical approvals. To prepare 10% FBS, the frozen FBS stock was thawed overnight at 4°C. Cell culture medium containing 10% FBS was prepared by mixing FBS with 1% penicillin-streptomycin solution in Roswell Park Memorial Institute 1640 (RPMI1640) medium (HyClone, USA). RMPI640 was used as it is a common medium employed for in-house cell culture.

3. RESULTS

4. DISCUSSION

Rapid, high‑throughput screening of compounds in ADME assays is essential in drug discovery. PB measurements are commonly determined during primary screening but may not be re-evaluated until later stages during characterization for regulatory purposes, for which several techniques have been validated recently for robust and relevant estimation of % unbound for highly bound and challenging compounds [7, 25-31]. The accurate determination of the % unbound across matrices, particularly in plasma, tissue homogenates, and relevant in vitro media, is therefore critical early-on because % unbound values are used in the interpretation of in vitro potency, in vitro–in vivo correlations (IVIVC), dose projections for animals and humans, and safety margins, especially for highly bound compounds [32, 33].

When % unbound is measured before achieving dialysis equilibrium by RED and HTD devices, the true % unbound value may be underestimated by several fold. Such underestimation of % unbound values in assay media can overestimate apparent potency, while underestimating the % unbound in plasma can bias hepatic clearance predictions downward or upward, depending on the scaling context [24, 34]. For meaningful cross‑species comparisons and efficacious‑dose projections, total drug concentrations (in vitro or in vivo) must be corrected for % unbound; imprecise % unbound measurements can misdirect lead optimization [29]. Kinetic modeling approaches to estimate the optimum equilibrium time are available [35] however given the impracticality of optimizing the equilibrium time for each compound during early discovery, characterizing the time required for most compounds to reach dialysis equilibrium and implementing it as the standardized assay condition is essential to reduce the risk of % unbound underestimation and to enhance data consistency across compounds, facilitating reliable characterization of lead candidates in drug discovery. In this work, we evaluated PB over time for a panel of marketed compounds spanning diverse physicochemical properties. We profiled multiple matrices that are used routinely in discovery (human plasma, 10% FBS, rat brain homogenate), and extended the assessment to tumor cell lines and selected mouse tissues. Rat brain homogenate was included because its % unbound has been reported to approximate human brain % unbound, enabling a translational utility without interspecies correction [7, 9]. Among the physicochemical descriptors, LogD_7.4 (lipophilicity at pH 7.4) tracked as expected with the extent of PB. Therefore, nelfinavir and ritonavir were chosen for tumor and tissue binding experiments due to their high LogD values.

Both RED and HTD devices meet high‑throughput needs and are widely adopted for PB estimation. Prior reports have suggested that 2–6 hours may be sufficient to reach equilibrium [10-12]. The results in these studies were limited by inadequate incubation times of 6 hours, or the lack of intermediate time points between 6 and 24 hours, and other methodological limitations, such as comparisons made based on the % bound values [10, 11]. Therefore, a plateau curve indicative of true equilibrium and % unbound could not be confirmed. This approach to compare the % bound values could be misleading, particularly for highly bound compounds, where small changes in % bound can mask significant differences in % unbound [10, 11], which is a more relevant parameter. For example, a change from 99.5% to 99% bound appears to represent only a 0.5% or negligible difference, which corresponds to a 2-fold change in the % unbound values (0.5% to 1%).

The matrix composition affected the apparent rate of reaching equilibrium. In 10% FBS, many compounds displayed a higher % unbound than in plasma or other tissues, which can be attributed to the lower protein content and less nonspecific binding. For compounds with <10% unbound in plasma at 2 hours, the % unbound values were on average ~2.5-fold and up to ~6-fold (for imatinib) underestimated compared to those at equilibrium, irrespective of using a RED or HTD device. At the 6-hour incubation, it was still, on average, ~50% and up to a ~2.5-fold underestimation of the % unbound values in the RED device, and ~25% and up to a ~1.7-fold underestimation of the % unbound values in the HTD device. At 16 hours, a difference of <10% in % unbound was noted in both devices. A slightly faster equilibrium was achieved when using the HTD device; however, significant underestimation of % unbound can still occur with an incubation time of ≤6 hours. For compounds with % unboundvalues of >10%,the errors were smaller at the shorter incubation times, but in both RED and HTD devices, % unboundvalues were underestimated up to ~25% for equilibration times at ≤6 hours. As the dialysis incubation time was increased, the measured % unbound values gradually increased until equilibrium was attained. Interestingly, for sunitinib in human plasma, the % unbound values measured at 24 hours using the RED and HTD devices were 8%–9%, which were nearly 2-fold higher than the reported value of 5% [36]. These findings suggest that a longer incubation time (>16 hours) is necessary to maximize the likelihood of achieving dialysis equilibrium. However, due to limited data, definitive correlations between specific compound characteristics and optimal incubation times could not be established.

In 10% FBS, most compounds reached a % unbound plateau by 6 hours, except fluphenazine and lapatinib, which exhibited the lowest % unbound. At 2 hours, the % unbound was generally underpredicted by 2- to 3-fold. These findings suggest that for compounds having a % unbound value of <30% in 10% FBS, it is recommended to use an incubation time of 6–16 hours, consistent with that used for the plasma matrix.

The recently and commonly used diluted plasma approach, in which plasma is diluted up to 20-fold with PBS (pH 7.4) for RED or HTD systems, has yielded quantifiable and reliable % unbound values for highly bound compounds [7]. The equilibrium time may be shorter for matrices such as 10% FBS, which have lower protein concentrations; however, if the % unbound of the compound remains <30%, a longer equilibration time may be required. The initial determination of LogD7.4 is a critical factor for guiding this decision. The diluted plasma approach was not adopted in this study to evaluate compounds such as lapatinib, where a quantifiable concentration of the % unbound could not be acquired at some of the time points. PB studies in both tumor cell lines and brain tissue homogenates are commonly conducted to determine the brain-to-plasma unbound concentration ratio (Kp, uu)[8]. An equilibration time of 16 or 24 hours is suggested for these matrices, as well as for tissues that generally have a higher protein content and a % unbound value than that in plasma_. Despite a limited sample size (n=15), no notable differences or trends were observed in PB and physicochemical properties such as molecular weight, LogD, acidity, basicity, zwitterionic state, or neutral nature in this study. This could be due to the possibility that compounds may bind to different plasma protein components with different binding affinities [9, 37]. All test compounds used in this study were stable and had good recoveries; therefore, unstable or poorly recovering compounds may require additional consideration during their % unbound determination. In a separate experiment, the PB of nelfinavir and ritonavir was assessed in tumor cell lines at 6 and 16 hours. Two distinct cell line variants of each tumor type were selected for this study to capture intratumoral heterogeneity and clinically relevant differences. B16F10 is a metastatic murine melanoma model frequently used for metastasis and immunotherapy-related preclinical studies. For nelfinavir, the extended incubation period confirmed its strong binding, with extremely low % unbound values at 6 hours (0.06% to 0.11%) and a 2- to 3-fold increase by 16 hours (0.13% to 0.23%) across tumor models, demonstrating the importance of prolonged incubation times for capturing the time-dependent binding dynamics in highly bound compounds. Ritonavir also displayed high PB across tumor models, with comparable % unbound values between 6 and 16 hours, indicating faster equilibrium attainment. The longer incubation period (i.e., 16 hours) revealed subtle changes in binding and incremental increases in % unbound compared with the shorter incubations. These findings emphasize the critical role of extended incubation periods in detecting time-dependent changes in highly bound compounds.

The PB properties of both nelfinavir and ritonavir were further evaluated in mouse tissues, including brain, intestine, liver, lung, adipose, muscle, skin, and plasma, exhibiting consistent, tissue-specific binding trends. Nelfinavir displayed exceptionally low % unbound values, reflecting its strong binding to tissues (>99.8% bound). The 16-hour incubation led to substantial fold increases in % unbound, particularly in liver (~4.8-fold) and adipose (~2.9-fold). Ritonavir showed higher % unbound values across all tissues, ranging from 0.07% in the liver to 2.83% in adipose at 6 hours. In comparison to 6 hours, the % unbound increased in all tissues (except skin) at 16 hours incubation time, including a ~3.7-fold rise in liver, and a ~1.4-fold increase in adipose tissue, indicating better equilibrium. These findings underscore the importance of extended incubation times in assessing protein–drug interactions and support the use of longer incubation periods to enhance the predictive accuracy of drug behavior in vivo.

5. STUDY LIMITATIONS

The study was limited by the number of compounds representing different physicochemical classes therefore, a trend analysis cannot be performed for PB. Furthermore, the PB for the tool compound was not assessed using additional techniques such as flux dialysis, diluted plasma, presaturation method, ultracentrifugation, etc. to confirm % unbound using an orthogonal assay.

CONCLUSION

Accurate determination of the % unbound of a compound in plasma or other biological matrices is critical for characterizing compounds in drug discovery. Insufficient equilibrium times during PB assessment can result in significant errors in estimating % unbound values, which may subsequently impact the interpretation of in vitro potency, IVIVCs, dose projections for animal and human use, and the estimation of safety margins for compounds. These errors can ultimately affect the rank order or prioritization of compounds. Based on the current study, an equilibrium time of at least 16 hours is recommended for PB by dialysis using the RED or HTD devices across plasma, tissues, and disease-relevant matrices at relevant concentrations. This approach minimizes underestimation of % unbound values and provides consistent data across compounds, supporting the accurate characterization of lead candidates during drug discovery.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

Frozen human plasma was obtained from healthy Asian volunteers at Liaocheng Second People’s Hospital (Liaocheng, Shandong, China) with ethical approval from the Ethics Committee of Liaocheng Second People’s Hospital (Approval No. KB2023-Ethical Review Report-51-01).

All animal-related procedures for tissue collection were conducted in accordance with current regulations and approved by the Institutional Animal Care and Use Committee (IACUC) of SPF (Beijing) Biotechnology Co., Ltd., Beijing, People’s Republic of China (Approval Nos.: AWE2024-12-27 and AWE2024-12-2702).

HUMAN AND ANIMAL RIGHTS

All procedures performed in studies involving human participants were in accordance with the ethical standards of institutional and/or research committee and with the 1975 Declaration of Helsinki, as revised in 2013.

This study adheres to internationally accepted standards for animal research and follows the 3Rs principle. The ARRIVE guidelines were employed for reporting experiments involving live animals, promoting ethical research practices.

All animal experiments were performed in accordance with the Care and Use of Laboratory Animals guidelines.

CONSENT FOR PUBLICATION

Written informed consent was obtained from all donors prior to sample collection.

AVAILABILITY OF DATA AND MATERIALS

All data generated or analyzed during this study are included in this published article.

FUNDING

This study and writing support were sponsored by IDEAYA Biosciences.

CONFLICT OF INTEREST

The authors declare no conflict of interest, financial or otherwise.

ACKNOWLEDGEMENTS

The authors thank Puja Dhanota, Barbara Saechao, and Kent Wong for proofreading the manuscript.

SUPPLEMENTARY MATERIAL