Manuscript DMD-D-25-00346 The impact of glycocalyx on partitioning and distribution of basic drugs

Abstract

Concentrations of basic drugs are often overpredicted by physiologically based pharmacokinetic models at very early time points after intravenous dosing. An empirical shallow compartment in our recently published permeability- and perfusion-limited PermQ model improved the prediction of this observed early distribution. This study aimed to determine the impact of glycocalyx, a glycoprotein and glycolipid covering on membranes of red blood cells (RBCs) and endothelium, on the partitioning and distribution of basic drugs. Hyaluronidase ($50\mu \text{g}/\text{mL}$), heparinase (5 U/mL), and neuraminidase (0.1 U/mL) were used to remove the glycocalyx from rat or human RBCs. Confocal microscopy was used to confirm glycocalyx disruption. RBC partitioning experiments were conducted to obtain partition coefficients (Kp) in RBCs and PBS buffer in the presence or absence of glycocalyx. Experimental Kp for metoprolol and atenolol were scaled with glycocalyx dimensions. The human PermQ model was used to simulate the concentration-time profiles, and results were compared with simulations with empirical Kp values. Confocal microscopy confirmed RBC glycocalyx modification via sialic acid removal. Significant differences were observed for metoprolol and atenolol Kp on human RBC glycocalyx disruption. Modeling suggests that experimental Kp scaled with glycocalyx dimensions, while underestimating empirical Kp estimates, required a similar scalar for both drugs. This underestimation may be due to various reasons, such as incomplete RBC glycocalyx disruption or the RBC glycocalyx being an inadequate surrogate for the endothelial glycocalyx. Our results support the hypothesis that glycocalyx partitioning may partially explain the early distribution of the basic drugs metoprolol and atenolol.

1. Introduction

In an attempt to better predict drug distribution rates, we recently developed a permeability- and perfusion-based physiologically based pharmacokinetic (PBPK) model termed PermQ.¹ During the model development, concentrations of several drugs were found to be greatly overpredicted at early time points after intravenous bolus or infusion, particularly for hydrophobic bases with a high blood-to-plasma ratio (BP). The addition of a mathematical shallow compartment improved the prediction of early time point concentrations. A hypothesis for the early and marked distribution of basic drugs is the partitioning of positively charged bases into the negatively charged glycocalyx on endothelial cells (ECs). The glycocalyx, also known as the pericellular matrix, is a covering of glycoproteins and glycolipids that surrounds the cell membranes of bacteria, ECs, erythrocytes, and other cells.²

The structure of the glycocalyx is different across different species, vascular beds, organs, and varies with blood flow rates.³ The endothelial glycocalyx is found at the luminal side of vascular ECs⁴ and is mainly made up of glycoproteins bearing acidic oligosaccharides and terminal sialic acids, proteoglycans such as heparan sulfate proteoglycans, syndecans, glypican-1, and glycosaminoglycan (GAG) side chains. The 3 main GAGs found on most cell surfaces are heparan sulfate, chondroitin sulfate, and hyaluronic acid or hyaluronan.^5–7 The glycocalyx has a net negative charge that affects its interaction with plasma components. In ECs, the surface negative charge is mainly attributed to GAG side chain sulfation.⁸ In contrast, the negative charge for red blood cells (RBCs) is mainly dependent on sialic acid residues.^9,10 The human RBCs have approximately 10⁷ negative ionic groups per cell.¹¹ The negatively charged glycocalyx layer creates an electrostatic barrier for plasma cells and proteins. It is an important gatekeeper of vascular permeability.¹² The endothelial glycocalyx thickness in humans can reach up to 500–1000 nm and increases with vascular diameter.³ The RBC glycocalyx is thinner, typically only 10–20 nm, and is primarily composed of sialic acid residues.^13–16 One electron microscopy study has shown that the glycocalyx in rat myocardial capillaries ranges from approximately 182 to 512 nm.¹⁷

Recent studies have evaluated the effect of the glycocalyx on drug partitioning, particularly for large molecules and nanoparticles.^18–23 Olivieri et al²¹ reported that the glycocalyx modulates cellular uptake of albumin-coated nanoparticles, indicating that the presence of glycocalyx can enhance or inhibit nanoparticle internalization according to the surface properties of the nanoparticles. Similarly, Bussin et al²³ found that the glycocalyx reduced the interactions between the anionic nanoparticles and cells, and the magnitude of the effect is dependent on the nanoparticle charge. Another study suggested that cells with a mature glycocalyx internalize cationic nanoparticles and prevent anionic nanoparticles from entering cells.¹⁹ Compared with nanoparticles, small molecules are much smaller and show different distribution kinetics.^24,25 However, the impact of the glycocalyx on the partitioning of small molecules remains unexplored and has not been documented in the current literature.

Compared with in vivo glycocalyx studies, in vitro studies are easier to conduct in a species-specific manner and are simpler and more convenient.^26,27 The goal of the present study was to evaluate whether RBCs can be used as in vitro surrogates for the endothelial glycocalyx to evaluate partitioning, given the relative ease of in vitro experiments with RBCs. We aimed to evaluate the distribution of basic small molecule drugs in rat and human RBCs on disruption of the glycocalyx to characterize the impact of the glycocalyx on drug partitioning. For this purpose, experimentally collected partition coefficients (Kp) for metoprolol and atenolol were incorporated into our human PermQ model. The model was then used to simulate the concentration-time (C-t) profiles with experimental Kp, and results were compared with simulations using empirical Kp values from the shallow compartment within human PermQ.

2. Materials and methods

2.1. Materials

Rat RBCs were obtained from Innovative Research Inc. Human RBCs were obtained as deidentified, leftover samples from Sol Sherry Thrombosis Research Center healthy donors at Temple University and were handled and used in accordance with approved Temple University Institutional Biosafety Committee protocols. Human erythroleukemia cells were obtained from Dr Carlos Barrero's laboratory. Atenolol, chloroquine, labetalol, and neuraminidase from Vibrio cholerae; heparinase I and III blend from Flavobacterium heparinum; hyaluronidase from bovine testes; hyaluronic acid binding protein; Triticum vulgaris lectin or wheat germ agglutinin (WGA); dextran (70 kDa); and HEPES-buffered sodium chloride including 1% bovine serum albumin were obtained from Sigma Aldrich. Chloroquine-d4 was obtained from Santa Cruz Biotechnology. Metoprolol-d7 hydrochloride was obtained from Expert Synthesis Solutions. Atenolol-d7 was obtained from Cayman Chemical Company. Metoprolol tartrate, formic acid, DMSO, acetonitrile (ACN), Invitrogen DAPI and Hoechst nucleic acid stains, streptavidin (Alexa Fluor 647 conjugate), Invitrogen CellMask plasma membrane stains (red and orange), RIPA lysis buffer, Halt protease inhibitor cocktail (100×), Halt phosphatase inhibitor cocktail (100×), EDTA solution 0.5M (100×), N-sulfated heparan sulfate monoclonal antibody, and Tris-glycine NN 10% gel were obtained from Fisher Scientific. PBS (pH 7.4) was obtained from Corning. Running buffer, Precision Plus protein dual color standards, 4%–15% Mini-PROTEAN TGX precast protein gels, and loading buffer were obtained from Bio-Rad. Paraformaldehyde (4%) was obtained from bioWORLD. Ghost RBCs (Bovine) were obtained from Thomas Scientific.

2.2. Optimization of glycocalyx disruption

A total of 5 types of cells were used in this study, including rat RBCs, rat ghost RBCs, bovine ghost RBCs, human RBCs, and human erythroleukemia cells. On preliminary studies, rat ghost RBCs, rat whole RBCs, and human RBCs were evaluated further. Rat ghost RBCs were used for confocal microscopy, whereas whole RBCs were used for partitioning experiments.

Rat ghost RBCs were generated by resuspending RBCs in a 10-fold volume of 5 mM sodium phosphate, pH 8.0 (with 1 mM EDTA), and spun down at 16,000g for 15 minutes. This step was repeated until the supernatant was clear, and the pellet was white. In an attempt to strip the entire glycocalyx from the surface of the membrane (see Fig. 1), cells were treated with a combination of neuraminidase (0.1 U/mL), heparinase (5 U/mL), and hyaluronidase ($50\mu \text{g}/\text{mL}$) for 2 hours at 37 °C with continuous rotation. After 2 hours of incubation, $100\mu \text{L}$ of enzyme-treated cells were washed with PBS to remove enzymes and byproducts by centrifuging at 1000g for 10 minutes at 4 °C.

Fig. 1.

Endothelial cell glycocalyx. Schematic representation of the endothelial glycocalyx. The boxed panels illustrate enzymatic removal of specific components by neuraminidase (sialic acid), heparanase I/III (heparan sulfate), and hyaluronidase (hyaluronic acid).

For rat and human whole RBCs, 5 mL of whole RBCs were washed 3 times with 10 mL of PBS buffer by centrifugation at 1000g for 10 minutes at 4 °C. The supernatant was discarded, and the process was repeated 3 times. Human RBCs were treated with a combination of neuraminidase (0.1 U/mL), heparinase (5 U/mL), and hyaluronidase ($50\mu \text{g}/\text{mL}$) for 2 hours at 37 °C with continuous rotation. After 2 hours of incubation, the treated RBCs were washed with PBS to remove enzymes by centrifugation at 1000g for 10 minutes at 4 °C.

2.3. Confocal microscopy

For rat ghost RBCs and human RBCs, after the enzyme treatment and washing steps, $100\mu \text{L}$ of samples were incubated with N-sulfated heparan sulfate monoclonal antibody and hyaluronic acid binding protein for 1 hour with continuous shaking at 37 °C. Then, the samples were incubated with streptavidin, WGA, and CellMask at 37 °C with continuous shaking. After 1 hour of incubation, the samples were washed with PBS buffer 3 times by centrifugation and seeded in 96- or 384-well glass-bottom plates. Finally, the samples were imaged by the Operetta CLS High Content Analysis System (Revvity). Calcein-AM (Fisher Scientific) was also added after enzyme treatment to assess cell viability. The removal of the sialic acid was also determined by comparing the changes in the ratio of sialic acid intensity to lipid membrane intensity. All experiments were conducted in 2 replicates.

2.3.1. Image acquisition

The parameters used for imaging the cell experiments included a 63× Water objective, 50% power, and 700–1000 milliseconds of exposure for each channel, using an Operetta CLS PerkinElmer automated confocal system. At least 2 independent experiments were performed, including 2–9 replicates. Nine to twenty-five fields were acquired per well, including 3–6 Z planes, and data were quantified from at least 4 fields per replicate.

2.3.2. High content image analysis of sialic acid

Specific parameters for cell identification were performed by using CellMask membrane staining in red or deep red channels. The best method for cell selection was optimized for the RBC area. The sialic acid staining in the green channel was used for membrane detection within the previously identified RBCs. The intensity of sialic acid in the RBC membrane was measured, and the sum of the maximum intensity per cell was used to compare the controls and neuraminidase-treated samples. Data analysis was performed using Harmony 4.8 software from PerkinElmer.

2.4. Partitioning experiments with and without glycocalyx

Four basic drugs were selected based on different apparent permeability (Table 1). Enzyme-treated rat ghost RBCs and human RBCs were prepared using the methods described above. The control group underwent the same procedure, excluding the addition of enzymes. After enzyme treatment and centrifugation, the RBC pellet was resuspended in PBS (pH 7.4). The hematocrit was then measured. Compound spiking solution $(10\mu \text{L})$ was added to $990\mu \text{L}$ of enzyme-treated and control RBCs for a final compound concentration of $1\mu \text{M}$. Samples were then incubated at room temperature for different incubation times with continuous rotation.

Table 1.

Physicochemical and in vitro absorption, distribution, metabolism, and excretion properties of the drugs used in the study

Drug	MW	LogP	${\text{pK}}_{\text{a},\text{a}}$	${\text{pK}}_{\text{a},\text{b}}$	${\text{f}}_{\text{up}}$	${\text{f}}_{\text{um}}$	BP	${\text{P}}_{\text{app}}$ a	${\text{f}}_{\text{e}}$	Reference
Atenolol	266.336	0.16	14	9.6	0.9	0.93	1.08	0.35	0.94	1
Metoprolol	267.364	1.88	14	9.7	0.86	0.83	1.1	28	0	1
Labetalol	328.406	2.70	14	9.3	0.50	NA	1.36	9.44	0.45	28–31
Chloroquine	319.872	4.63	14	10.1	0.61	0.797	3.91	16.0	0.05	32–35

BP, blood-to-plasma ratio; ${\text{f}}_{\text{e}}$, fraction of unchanged drug excreted renally; ${\text{f}}_{\text{um}}$, unbound fraction in the microsome; ${\text{f}}_{\text{up}}$, unbound fraction in the plasma; LogP, the log of the partition coefficient of a solute between octanol and water; MW, molecular weight; NA, not available; ${\text{P}}_{\text{app}}$, apparent permeability; ${\text{pK}}_{\text{a},\text{a}}$, ionization constant for acids; ${\text{pK}}_{\text{a},\text{b}}$, ionization constant for bases.

^aCaco-2 cell permeability, listed as value ×10⁻⁶ cm/s, all values collected at pH 7.4.

The sample preparation protocol was as follows: after incubation, samples were centrifuged at 1000g for 5 minutes at 4 °C. Supernatant was carefully removed and a fixed volume of $100\mu \text{L}$ was pipetted into new tubes. Then, internal standard (IS)/ACN ($200\mu \text{L}$) was added to these supernatant tubes. One hundred microliters of the remaining RBC pellet was also transferred into new tubes. The RBC samples were subjected to 3 freeze-thaw cycles. IS/ACN $(200\mu \text{L})$ was added to the tubes, and all tubes were placed on ice for 5 minutes. Tubes were spun at 10,000g for 10 minutes at room temperature. After centrifugation, $150\mu \text{L}$ of supernatant was submitted to liquid chromatography-triple quadrupole mass spectrometer (LC-MS/MS) analysis. Standards were prepared with the same protocol except the matrix was spiked with nominal concentration of each drug.

These experiments were conducted in PBS because enzyme treatment and washing steps require removal of plasma components and residual enzymes before partitioning measurements. Although partitioning was measured under PBS conditions rather than plasma, a correction based on the unbound fraction in plasma $\left({\text{f}}_{\text{up}}\right)$ was applied to account for differences in protein binding. Although we acknowledge that temperature can influence blood-related partitioning measures and may affect the absolute Kp values,³² incubations were performed at room temperature to enable consistent data collection across the short incubation time of 1–5 minutes.

2.5. LC-MS/MS

Stock solutions of different drugs were prepared in DMSO. IS metoprolol-d7, atenolol-d7, and chloroquine-d4 were prepared at $0.1\mu \text{M}$ in ACN. Working solutions were serially diluted in PBS and then spiked into different matrixes (PBS, rat RBCs, and human RBCs). The calibration standards ranged from $0.03125\mu \text{M}$ to $1\mu \text{M}$. The calibration curve included 6 nonzero standards (0.03125, 0.0625, 0.125, 0.25, 0.5, and $1\mu \text{M}$) and was fit using a $1/x^2$ weighting. Quality control (QC) samples were prepared from a separate stock and used for assay validation at 3 levels (0.1, 0.3, and $0.75\mu \text{M}$). The acceptance criteria were ±15% for calibration standards and QC samples (±20% at the lower limit of quantification). For each QC level, 3 replicates were analyzed, and at least 2 of 3 were required to meet the acceptance criteria. Samples were mixed with IS in ACN to precipitate proteins. Samples were vortexed and centrifuged at 10,000g for 10 minutes at room temperature. Five microliters of supernatant was injected into the LC-MS/MS system for analysis.

An Agilent 1260 high-performance liquid chromatography (HPLC) system and an API 4000 QTRAP mass spectrometer were used to determine the concentrations for different experiments. The mass spectrometer was operated in positive-ion mode. LC-MS/MS data were analyzed using the Analyst software version 1.6. MS parameters for all drugs were obtained by flow injection analysis. The mobile phase for all compounds consisted of water with 0.1% formic acid as the aqueous phase (A) and ACN with 0.1% formic acid as the organic phase (B).

The gradient used for metoprolol and labetalol elution started at 95% A and was maintained for 30 seconds. Then A was ramped down to 5% over 70 seconds and maintained for 140 seconds. Mobile phase A was ramped back up to 95% over 6 seconds and maintained until 7.5 minutes. The flow rate was $400\mu \text{L}/\text{min}$. The analytical column used for HPLC for metoprolol and labetalol was a Luna, $3\mu \text{m}$ C18 (2) 100A, 30 × 2 mm column (Phenomenex) coupled with a C18 guard column (4 × 2.0 mm).

The gradient used for atenolol elution started at 95% A and was maintained for 30 seconds. Mobile phase A was ramped down to 5% over 180 seconds and was maintained for 2 minutes before gradually increasing to 95% over 6 seconds. A then remained at 95% until 8.5 minutes. The flow rate was kept at $1000\mu \text{L}/\text{min}$. The analytical column used for atenolol was a ZORBAX, $5\mu \text{m}$ SB-C18 150 × 4.6 mm column (Agilent) coupled with a C18 guard column (4 × 2.0 mm When using atenolol-d7 as IS, the flow rate was kept at $800\mu \text{L}/\text{min}$.

The gradient used for chloroquine elution started at 95% A and was maintained for 120 seconds. Then A was ramped down to 0% over 420 seconds and maintained for 60 seconds. Mobile phase A was ramped back up to 95% over 6 seconds and maintained until 13 minutes. The flow rate was $700\mu \text{L}/\text{min}$. The analytical column used for HPLC for chloroquine was an Eclipse, $5\mu \text{m}$ C8 150 × 4.6 mm column (Agilent) coupled with a C18 guard column (4 × 2.0 mm).

The columns remained at room temperature for analysis of all the compounds. Declustering potential, collision energy, and collision exit potential were optimized for all compounds, and optimized conditions are listed in Supplemental Table 1.

2.6. Data analysis

The data were described using the means and standard deviation of at least 3 replicate measurements. Statistical $t$ tests and 2-way ANOVA were used to determine the difference between groups, and significance was concluded by a $P$ value of <.05.

The partitioning coefficient (Kp) of partitioning experiments was calculated using the following equation:

$$\text{Kp}=\frac{{\text{C}}_{\text{RBC}}}{{\text{C}}_{\text{PBS}}}$$ (1)

where ${\text{C}}_{\text{RBC}}$ is the concentration in the RBCs, and ${\text{C}}_{\text{PBS}}$ is the concentration in the PBS buffer.

2.7. Modeling and simulation

The average experimental intravenous bolus and/or infusion plasma C-t profiles in humans for metoprolol and atenolol used here were the same as those reported in earlier publications.¹ The observed half-life, clearance, and steady-state volume of distribution (V_ss) were determined by compartmental analysis using standard pharmacokinetic equations with Mathematica version 13.1 (Wolfram Research Inc). One- to three-compartment models were evaluated to predict the C-t profiles from the experimental data. Experimental half-life, clearance, and V_ss values were obtained from compartmental models. The best model fit for the experimental data was determined with the corrected Akaike Information Criterion values.³⁶ All pharmacokinetic parameters were calculated assuming a 70-kg average human body weight.

The human PermQ model, a permeability- and perfusion-limited PBPK model, has already been developed¹ and is briefly described as follows: (1) drug is assumed to reversibly distribute between capillaries and interstitial fluid by fenestra or discontinuities in capillaries or by transcellular diffusion through ECs, (2) passive membrane permeability and transporter-mediated movement across membranes are considered, (3) drugs are also assumed to partition into intracellular phospholipids and neutral lipids depending on their physicochemical properties, and (4) a shallow distribution compartment was added, particularly for basic drugs. In this model, the membranes and the structure of the glycocalyx of erythrocytes were assumed to be similar to ECs; erythrocyte partitioning was evaluated as a descriptor to predict partitioning into a shallow compartment (C0). The following equations were used to model the shallow compartment. The BP due to external association with erythrocytes (${\text{BP}}_{\text{C}0}$) was calculated by subtracting drug inside the erythrocyte (${\text{f}}_{\text{up}}\text{H}$) from experimental BP values:

$${\text{BP}}_{\text{C}0}=\text{BP}-{\text{f}}_{\text{up}}\text{H}$$ (2)

Eq. 2 assumes minimal intracellular binding in erythrocyte for drugs with BP ≤1 based on the lack of intracellular organelles in these cells.³⁷ Detailed derivation of the equation is provided in Supplemental Material.

The associated Kp for external association with erythrocytes (${\text{K}}_{\text{ery,CO}}$) was

$${\text{K}}_{\text{ery},\text{C}0}=\frac{{\text{BP}}_{\text{C}0}-(1-\text{H})}{{\text{f}}_{\text{up}}\text{H}}$$ (3)

The Kp $\left({\text{K}}_{\text{C}0}\right)$, termed “empirical Kp,” for the shallow compartment was calculated as follows:

$$\text{KC}0=\text{c}1\text{Kery},{\text{C}0}^{\text{c}2}$$ (4)

$${\text{k}}_{\text{off}}=\frac{{\text{k}}_{\text{on}}}{{\text{K}}_{\text{C}0}}$$ (5)

where H is hematocrit, BP is the blood-to-plasma ratio, ${\text{f}}_{\text{up}}$ is unbound fraction in plasma of drug, ${\text{c}}_1$ and ${\text{c}}_2$ are optimized constants, ${\text{k}}_{\text{on}}$ is the association rate constant fixed at $1000\mu {\text{M}}^{-1}{\text{min}}^{-1}$, and ${\text{k}}_{\text{off}}$ is the dissociation rate constant. As detailed in our previous publication,¹ the model that best simulated observed early time point concentrations for 14 weak bases was used to optimize the constants ${\text{c}}_1=19.0$ and ${\text{c}}_2=0.69$. Eqs. 3–5 were used in this study to calculate the empirical Kp values for metoprolol and atenolol. For each drug, 2 Kp from partitioning experiments were obtained: ${\text{Kp}}_{+\text{glyco}}$ (with intact glycocalyx) and ${\text{Kp}}_{-\text{glyco}}$ (after glycocalyx disruption). Because only 2 drugs were available, several methods were tested to scale the RBC glycocalyx Kp to ECs. In method 1, a scaling factor for RBC-to-EC glycocalyx $\left({\text{R}}_{\text{glyco}}\right)$ was applied to convert the RBC glycocalyx Kp to ECs, instead of fitting. We assumed that partitioning into glycocalyx should consider the cell glycocalyx surface area (SA) and the glycocalyx thickness (h). The glycocalyx volume (${\text{V}}_{\text{glyco}}$) was simply calculated as SA times h, ${\text{V}}_{\text{glyco}}=\text{SA}\times \text{h}$.

Literature values were for RBC and EC glycocalyx SA and h were collected, and the entire range of these values along with an arithmetic average were used for sensitivity analysis. The reported EC glycocalyx SA and h values resulted in a nonphysiologic volume. Therefore, in method 2, a value of the EC glycocalyx volume measured in human subjects with tracer (dextran 40) methods³⁸ was used to calculate ${\text{R}}_{\text{glyco}}$. Finally, method 3 calculated ${\text{R}}_{\text{glyco}}$ with total SA of RBCs and ECs instead of volume, with the assumption that the glycocalyx net negative charge per SA of RBCs and ECs is comparable.

The scaling factor, ${\text{R}}_{\text{glyco}}$, was calculated by using the following equations:

Method 1:

$${\text{R}}_{\text{glyco}}=\left(\frac{{\text{V}}_{\text{glyco–EC}}}{{\text{V}}_{\text{glyco–RBC}}}\right)\times \left(\frac{{\text{N}}_{\text{EC}}}{{\text{N}}_{\text{RBC}}}\right)$$ (6)

where ${\text{V}}_{\text{glyco-EC}}$ is the glycocalyx volume per EC, and ${\text{V}}_{\text{glyco–RBC}}$ is the glycocalyx volume per RBC; ${\text{N}}_{\text{EC}}$ and ${\text{N}}_{\text{RBC}}$ are the total number of ECs and the total number of RBCs, respectively.

Method 2:

$${\text{R}}_{\text{glyco}}=\left(\frac{\text{Total}{\text{V}}_{\text{glyco}-\text{EC}}}{{\text{V}}_{\text{glyco}-\text{RBC}\times }{\text{N}}_{\text{RBC}}}\right)$$ (7)

Method 3:

$${\text{R}}_{\text{glyco}}=\left(\frac{\text{Total}{\text{SA}}_{\text{EC}}}{\text{Total}{\text{SA}}_{\text{RBC}}}\right)$$ (8)

Because the ${\text{Kp}}_{+\text{glyco}}$ and ${\text{Kp}}_{-\text{glyco}}$ for RBC were measured under PBS conditions with varying hematocrit, a correction was applied with a standard hematocrit (H) (45%) and ${\text{f}}_{\text{up}}$ using the following equation:

$${\text{Kp}}_{+\text{glyco–corrected}}=\frac{\text{H}+{\text{H}}_{\text{experiment}}\left({\text{Kp}}_{+\text{glyco}}-1\right)}{{\text{f}}_{\text{up}}\text{H}}$$ (9)

where ${\text{Kp}}_{+\text{glyco-corrected}}$ is the corrected ${\text{Kp}}_{+\text{glyco}}$ with hematocrit 45%, ${\text{f}}_{\text{up}}$ is the unbound fraction in plasma, and ${\text{H}}_{\text{experiment}}$ represents the hematocrit measured during the experiment. The difference of the Kp of RBC with and without glycocalyx was calculated as follows:

$$\Delta {\text{Kp}}_{\text{glyco–RBC}}={\text{Kp}}_{+\text{glyco–corrected}}-{\text{Kp}}_{-\text{glyco–corrected}}$$ (10)

Then the Kp of the EC glycocalyx (${\text{Kp}}_{\text{glyco-EC}}$) can be calculated by the following equation:

$${\text{Kp}}_{\text{glyco–EC}}=\Delta {\text{Kp}}_{\text{glyco–RBC}}\times {\text{R}}_{\text{glyco}}$$ (11)

Given the hypothesis that disruption of the glycocalyx is expected to decrease early partitioning of basic compounds into the glycocalyx, a 1-sided $t$ test was appropriate to compare ${\text{Kp}}_{+\text{glyco}}$ and ${\text{Kp}}_{-\text{glyco}}$ values. However, given that for drugs partitioning significantly into cell membrane components besides the glycocalyx, it is reasonable to expect an increase in early partitioning rates due to the disruption of the glycocalyx barrier. Therefore, a 2-sided $t$ test was additionally used for completeness of evaluation.

For comparison of the various PBPK model predictions, an exposure overlap coefficient (EOC) was calculated (eq. 12) to determine the accuracy of the C-t profiles. EOC is computed by identifying the overlapping section of the experimental and predicted C-t profile curves and dividing this area by the experimental area under the curve. Using experimental clearance values for all predictions ensured that both the experimental and predicted C-t profiles had the same areas under the curve. This standardized approach enabled the EOC to serve as a direct method for comparing the shapes of these curves.

$$\text{EOC}=\frac{{\int }_0^{\mathit{\text{tlast}}}\text{Min}\left[C_{\text{expt}}\left[t\right],C_{\text{pred}}\left[t\right]\right]}{{\int }_0^{\mathit{\text{tlast}}}{C}_{\text{expt}}\left[t\right]}$$ (12)

3. Results

Preliminary studies comprised a rat RBC sedimentation test and confocal microscopy of human erythroleukemia cells. Detailed methods and results are provided in the Supplemental Material and Supplemental Figs. 1–7.

Confocal microscopy was used for visualization of cells with glycocalyx, as well as glycocalyx disruption. Figure 2A shows the confocal microscopy results for rat ghost RBCs stained with WGA and CellMask. In this study, the cells were treated with different concentrations of neuraminidase for 2 hours with continuous rotation. Compared with control groups, the sialic acid content (indicated by green fluorescence intensity) was reduced in enzyme-treated groups. Figure 2B shows that the sialic acid intensity of enzyme-treated groups was significantly lower than the control group ($P<.0001$, 2-way ANOVA). Figure 3 represents the confocal microscopy images of human RBCs at 2 enzyme concentrations. Negative control groups were included to correct for bleed-through fluorescence from other channels. As expected, no signal for sialic acid was observed without WGA dyes. After neuraminidase and hyaluronidase treatment, the sialic acid intensity was significantly reduced (Fig. 3, A–D), and RBC aggregation was observed (Supplemental Fig. 8).

Fig. 2.

Rat red blood cell ghosts' neuraminidase treatment. (A) Representative confocal microscopy images of rat ghost red blood cells treated with increasing concentrations of neuraminidase (0.05, 0.1, and 0.2 U/mL) for 2 hours. Cells treated under the same conditions without neuraminidase were included as a control. Red blood cell ghost membranes: sialic acid was stained with wheat germ agglutinin (green), and the lipid membranes were stained with CellMask Deep Red (C10046). The merge of sialic acid and membrane staining is shown, along with the Brightfield images. Scale bars, $20\mu \text{m}$. (B) Sialic acid staining was quantified using the sum of the maximum fluorescence intensity per cell (Harmony 4.8 software from PerkinElmer). The bar graph shows the mean (± SD) of 6 datasets per condition. Data were analyzed using 2-way ANOVA for multiple comparisons with Prism GraphPad software (version 10.6.1). $****P<.0001$.

Fig. 3.

Human red blood cell (RBC) neuraminidase treatment. Representative confocal microscopy images of independent assays of human RBCs treated with neuraminidase 0.02 U/mL (A, B) and 0.1 U/mL (C, D) for 2 hours. Human RBCs treated under the same conditions without neuraminidase were included as a control. Membrane sialic acid was stained with wheat germ agglutinin (green), and the lipid membranes were stained with CellMask Deep Red (C10046) in panel (A) and with CellMask Orange (C10045) in panel (C). The merge of sialic acid and membrane staining is shown, along with the brightfield images. Scale bars, $20\mu \text{m}$. Sialic acid staining was quantified using the sum of the maximum fluorescence intensity per cell (Harmony 4.8 software from PerkinElmer). The bar graphs (B, D) show the mean (± SD) of 9 and 4 datasets, respectively, for panels (B) and (D) per condition. Data were analyzed using an unpaired $t$ test with Welch's correction in Prism GraphPad software (version 10.6.1). $****P<.0001$.

For partitioning experiments, 1-minute and 5-minute incubation times were selected for human RBCs, and 1-minute incubation times for rat whole RBCs, based on preliminary experiments. Table 2 shows the average Kp values for all 4 drugs with and without glycocalyx in rat and human whole RBCs. For human RBCs, the results for metoprolol were significantly different between the control and enzyme-treated groups for both 1- and 5-minute incubation times. The 5-minute results for atenolol were significantly different between the control and enzyme-treated groups when using a 1-tailed $t$ test. For labetalol, no statistically significant differences were observed between the control and enzyme-treated groups for both 1- and 5-minute incubation times. For chloroquine, a significant difference was observed at the 1-minute incubation time, whereas no significant difference was observed at 5 minutes. For rat RBCs, no statistically significant differences were observed between the control and enzyme-treated groups for metoprolol and labetalol for 1-minute incubation time. For chloroquine, a significant difference was observed at the 1-minute incubation time in rat RBCs, whereas no significant difference was observed at 30 and 60 minutes.

Table 2.

Average partition coefficient (Kp) values for all 4 drugs with and without glycocalyx of whole rat and human red blood cells at different incubation times

Compound	Time min	N	H %	${\text{Kp}}_{\text{-glyco}}$ (mean ± SD)	${\text{Kp}}_{\text{+glyco}}$ (mean ± SD)	Two-tailed $\mathit{t}$ test $\mathit{P}$ Value	One-tailed $\mathit{t}$ test $\mathit{P}$ Value
Rat
Metoprolol	1	3	18	1.118 ± 0.009	1.173 ± 0.059	.186	.093
	1	5	28	1.167 ± 0.050	1.221 ± 0.065	.178	.089
	1	3	36	0.899 ± 0.161	1.011	.420	.210
	1	2	37	1.042	1.062	NA	NA
	1	4	43	1.324 ± 0.023	1.322 ± 0.051	.946	.473
Labetalol	1	3	NA	4.47 ± 0.37	4.06 ± 0.36	.242	.121
Chloroquine	1	3	43	2.315 ± 0.046	1.945 ± 0.048	.0006	.0003
	30	5	43	2.536 ± 0.053	2.525 ± 0.195	.906	.453
	60	3	43	2.445 ± 0.079	2.493 ± 0.053	.432	.216
Human
Metoprolol	1	5	40	0.978 ± 0.055	1.010 ± 0.010	.063	.032
	5	5	40	0.992 ± 0.023	1.046 ± 0.028	.011	.005
	5	5	32	1.013 ± 0.018	1.049 ± 0.039	.108	.054
Atenolol	5	5	45	0.085 ± 0.003	0.121 ± 0.041	.129	.064
	5	5	34	0.026 ± 0.005	0.054 ± 0.026	.068	.034
Labetalol	1	5	34	3.324 ± 0.122	3.243 ± 0.313	.601	.300
	5	5	34	3.489 ± 0.139	3.379 ± 0.134	.238	.119
	5	5	45	3.436 ± 0.319	3.629 ± 0.288	.344	.172
Chloroquine	1	5	32	2.168 ± 0.058	2.071 ± 0.047	.008	.004
	5	5	32	2.153 ± 0.089	2.129 ± 0.038	.500	.250

H, hematocrit; ${\text{Kp}}_{-\text{glyco}}$, enzyme-treated, without glycocalyx; ${\text{Kp}}_{+\text{glyco}}$, control, with glycocalyx; concentration, $1\mu \text{M}$; N, number of replicate samples per condition; NA, not available.

The effect of glycocalyx on partitioning was also evaluated by using the human PermQ model. Table 3 shows the corrected Kp (calculated with eqs. 9 and 10) for the drugs tested. For metoprolol, atenolol, and labetolol, the shift in Kp on glycocalyx disruption was 5%, 50%, and 5%, respectively. Table 4 shows the range of calculated ${\text{Kp}}_{\text{glyco-EC}}$ for metoprolol and atenolol with 3 scaling methods based on different reported surfaces, thickness, cell numbers, and EC volume. Although the experimentally observed differences between the control and enzyme-treated groups were small (0.061 for metoprolol and 0.031 for atenolol), differences increased after applying the scaling factors. Moreover, due to the variability in the reported values of SA and glycocalyx thickness, the estimated ${\text{K}}_{\text{pglyo-EC}}$ ranges are relatively wide. Table 5 presents the predicted individual EOC with calculated ${\text{KP}}_{\text{glyco}-\text{EC}}$ values for metoprolol and atenolol. Along with the average ${\text{Kp}}_{\text{glyco-EC}}$ values, sensitivity analysis was conducted for the lowest and highest calculated values based on Table 4, for each of the 3 scaling methods. All the scaling methods tested were comparable in prediction of C-t profiles. Comparison of the ${\text{Kp}}_{\text{glyco}-\text{EC}-\text{average}}$ values to the empirical Kp from PermQ revealed a common scalar across methods for both drugs (Table 5), for example, method 1 scalar = 15.5 (metoprolol) versus 17.6 (atenolol). Associated predictions of C-t profiles are shown in Figs. 4 (4A, B, and C for methods 1, 2, and 3 respectively) and 5 (5A, B, and C for methods 1, 2, and 3 respectively). Figures 4 and 5 represent the model predictions for metoprolol and atenolol, using average, highest, and lowest calculated ${\text{Kp}}_{\text{glyco-EC}}$, respectively. For metoprolol, the simulation result with the ${\text{Kp}}_{\text{glyco-EC}}$ value at the high end of the range was close to the result using the empirical ${\text{Kp}}_{\text{glyco-EC}}$ value (EOC = 0.902, Table 5). For atenolol, the model prediction using ${\text{Kp}}_{\text{glyco-EC}}$ values with method 2 provided the best prediction (EOC = 0.924 with ${\text{Kp}}_{\text{glyco-EC-average}}$, Table 5). The predicted Kp values with the RBC glycocalyx disruption method were consistently lower than the empirical Kp (Table 5), thereby underestimating drug distribution into the shallow compartment. All simulations overestimated the early time points for both drugs with the range of calculated ${\text{Kp}}_{\text{glyco-EC}}$ values (Figs. 4 and 5, insets).

Table 3.

Corrected partition coefficients of basic drugs in this study, in human RBCs

Compounds	${\text{f}}_{\text{up}}$	Control			Glycocalyx Disruption			$\text{𝚫}{\text{Kp}}_{\text{glyco-RBC}}$
		${\text{H}}_{\text{experiment}}$	${\text{Kp}}_{+\text{glyco}}$	Corrected	${\text{H}}_{\text{experiment}}$	${\text{Kp}}_{\text{-glyco}}$	Corrected
Metoprolol	0.86	0.41	1.046	1.216	0.40	0.992	1.155	0.061
Atenolol	0.9	0.45	0.054	0.060	0.45	0.026	0.029	0.031
Labetalol	0.50	0.45	3.629	7.258	0.45	3.436	6.872	0.386
Chloroquine	0.61	0.32	2.071	2.888	0.32	2.168	3.001	−0.113

${\text{f}}_{\text{up}}$, unbound fraction in the plasma; ${\text{Kp}}_{+\text{glyco}}$, partition coefficient with glycocalyx; ${\text{Kp}}_{-\text{glyco}}$, partition coefficient without glycocalyx; RBC, red blood cell.

Table 4.

Range of ${\text{K}}_{\text{pglyco-EC}}$ values based on method 1: reported surface areas (SA) and glycocalyx thicknesses (h), method 2: reported glycocalyx volume in humans, and method 3: total surface area of RBCs and EC in humans

Method 13,13,39–45
	RBC			EC
	Surface area $\mathit{\mu }{\mathit{m}}^2$	Thickness $\mathit{\mu }\mathit{m}$	Cell number	Surface area $\mathit{\mu }{\mathit{m}}^2$	Thickness $\mathit{\mu }\mathit{m}$	Cell number
Range (average)	128–144 (136)	0.01–0.02 (0.015)	19–32 × 1012 (25.5 × 1012)	500–2100 (1000)	0.5–1.5 (1.0)	0.6–1.0 × 1012 (0.8 × 1012)
${\text{R}}_{\text{glyco}}({\text{R}}_{\text{glyco-average}})$	1.63–130 (15.4)
Drug, ${\text{Kp}}_{\text{glyco-EC}}\left({\text{Kp}}_{\text{glyco-EC-average}}\right)$	Metoprolol, 0.1–7.9 (0.94)			Atenolol, 0.05–4.0 (0.48)

Method 238
	RBC			EC
	Surface area $\mathit{\mu }{\mathit{m}}^2$	Thickness $\mathit{\mu }\mathit{m}$	Cell number	Total glycocalyx volume $\mathit{\mu }{\mathit{m}}^3$
Range (average)	128–144 (136)	0.01–0.02 (0.015)	19–32 × 1012 (25.5 × 1012)	1.6 × 1015
${\text{R}}_{\text{glyco}}({\text{R}}_{\text{glyco-average}})$	17.4–65.8 (30.8)
Drug, ${\text{Kp}}_{\text{glyco-EC}}({\text{Kp}}_{\text{glyco-EC-average}})$	Metoprolol, 1.06–4.01 (1.88)			Atenolol, 0.54–2.04 (0.95)

Method 344–46
	RBC	EC
	Total surface area	Total surface area
	${\mathit{m}}^2$	${\mathit{m}}^2$
	3468	300–2100
${\text{R}}_{\text{glyco}}({\text{R}}_{\text{glyco-average}})$	0.087–0.61 (0.35)
Drug, ${\text{Kp}}_{\text{glyco-EC}}({\text{Kp}}_{\text{glyco-EC-average}})$	Metoprolol, 0.005–0.037 (0.02)	Atenolol, 0.003–0.02 (0.01)

EC, endothelial cell; ${\text{Kp}}_{\text{glyco-EC}}$, endothelial cell glycocalyx partition coefficient; ${\text{Kp}}_{\text{glyco-EC-average}}$, partition coefficient of endothelial cell using ${\text{R}}_{\text{glyco-average}}$; RBC, red blood cell; ${\text{R}}_{\text{glyco}}$, scaling factor for RBC-to-EC glycocalyx; ${\text{R}}_{\text{glyco-average}}$, factor calculated with average each parameter.

Table 5.

Model predicted individual exposure overlap coefficients (EOC) based on the range of calculated ${\text{Kp}}_{\text{glyco-EC}}$ with different scaling methods for metoprolol and atenolol

	${\text{Kp}}_{\text{glyco-EC}}$	EOC	Kp scalar, empirical/method
Metoprolol
Method 1
Lowest	0.1	0.804
Highest	7.9	0.902
Average	0.94	0.818	15.5
Method 2
Lowest	1.06	0.820
Highest	4.01	0.865
Average	1.88	0.833	7.75
Method 3
Lowest	0.005	0.803
Highest	0.037	0.803
Average	0.02	0.803	738
Human PermQ
Empiricala	14.57	0.915
Atenolol
Method 1
Lowest	0.05	0.873
Highest	4.0	0.906
Average	0.48	0.906	17.6
Method 2
Lowest	0.54	0.909
Highest	2.04	0.929
Average	0.95	0.924	8.87
Method 3
Lowest	0.003	0.868
Highest	0.02	0.869
Average	0.01	0.869	843
Human PermQ
Empiricala	8.43	0.882

Kp, partition coefficient; ${\text{Kp}}_{\text{glyco-EC}}$, endothelial cell glycocalyx partition coefficient.

^aThe empirical Kp values are obtained by using eqs. 2–5.

Fig. 4.

Human PermQ model predictions of C-t profiles with different ${\text{Kp}}_{\text{glyco-EC}}$ for metoprolol. Insets depict C-t data at early time points, 0–100 minutes. The black circles depict mean experimental data,⁴⁷ and the PBPK model predictions for each scaling method are depicted as follows: purple solid line: prediction using the empirical ${\text{Kp}}_{\text{glyco-EC}}$ value, red solid line: prediction using the lowest calculated ${\text{Kp}}_{\text{glyco-EC}}$ value, blue solid line: prediction using the highest calculated ${\text{Kp}}_{\text{glyco-EC}}$ value, and green solid line: prediction using the calculated ${\text{Kp}}_{\text{glyco-EC-average}}$ value. (A) Method 1, (B) method 2, and (C) method 3. C-t, concentration-time profile; ${\text{Kp}}_{\text{glyco-Ec}}$, endothelial cell glycocalyx partition coefficient.

Fig. 5.

Human PermQ model predictions of C-t profiles with different ${\text{Kp}}_{\text{glyco-EC}}$ for atenolol. Insets depict C-t data at early time points, 0–100 minutes. The black circles depict mean experimental data,⁴⁸ and the physiologically based pharmacokinetic model predictions for each scaling method are depicted as follows: purple solid line: prediction using the empirical ${\text{Kp}}_{\text{glyco-EC}}$ value, red solid line: prediction using the lowest calculated ${\text{Kp}}_{\text{glyco-EC}}$ value, blue solid line: prediction using the highest calculated ${\text{Kp}}_{\text{glyco-EC}}$ value, and green solid line: prediction using the calculated ${\text{Kp}}_{\text{glyco-EC}}$-average value. (A) Method 1, (B) method 2, and (C) method 3. C-t, concentration-time profile; ${\text{Kp}}_{\text{glyco-EC}}$, endothelial cell glycocalyx partition coefficient.

4. Discussion

The present work aimed to investigate a possible effect of glycocalyx on the partitioning and distribution of basic drugs with in vitro experimental methods. RBCs were used as a simple surrogate for the endothelium because both RBCs and ECs have a glycocalyx.^3,49–51 We explored the simple idea of testing drug partitioning in RBCs on enzymatic disruption of the glycocalyx. New methods were developed to scale partitioning of drugs into the RBC glycocalyx to endothelial glycocalyx partitioning, and scaled values were compared with PermQ model predicted values.

In vitro models to study the impact of blood flow and shear stress on the endothelial glycocalyx have been developed.^52–54 These include cell culture on microfluidic platforms to recapitulate the physical capillary environment. Evaluation of equilibrium partitioning of drugs into the glycocalyx should require a simpler experimental setup. RBCs used in this study (whole RBCs or ghost RBC preparations) are inexpensive and easy to obtain. Although a rat model would allow in vitro as well as in vivo drug absorption, distribution, metabolism, and excretion/pharmacokinetic evaluation, preliminary results indicated that the efficiency of glycocalyx disruption by enzymes was not similar between rat and human cells. A review of the literature corroborates the species difference in glycocalyx composition.^15,55 Eylar et al¹⁵ found that after neuraminidase treatment, the surface charge of human RBCs decreased by 94%, which suggested that most of the negative charge comes from sialic acid.¹⁵ Another study also reported an 80% reduction in mobility for human RBCs after sialic acid removal. In contrast, for rat RBCs, the reduction was only around 25%, indicating that besides sialic acid, other negatively charged components also contribute to the surface charge in rat RBCs.⁵⁵ The lack of impact of sialic acid removal in rat RBCs on drug distribution (Fig. 2; Table 2) may be due to modest contribution of sialic acid residues to rat glycocalyx negative charges.

All studies with rat RBCs, human RBCs, and human erythroleukemia cells indicate that the neuraminidase effectively removed sialic acid residues, which agrees with the previously published studies.^55,56 In addition, the concentration range of 0.05–0.2 U/mL neuraminidase did not appear to affect the efficiency of sialic acid removal in a concentration dependent manner. Our experiments do not provide evidence that heparinase and hyaluronidase successfully removed heparan sulfate and hyaluronic acid under the conditions used. Confocal microscopy results from human erythroleukemia cells (Supplemental Figs. 6 and 7) confirmed that both heparan sulfate and hyaluronic acid signals were detectable, and their intensities were not reduced in presence of the mixed enzymes. Even with glycocalyx disruption via sialic acid removal, potential RBC aggregation must be considered. RBC aggregation was observed after enzyme treatment (Supplemental Fig. 8). Such aggregation, and related changes in cell shape or surface geometry, may alter the effective cell surface available for drug association and can contribute to variable drug partitioning.

Confocal microscopy results for human RBCs support findings from other laboratories,^15,16,55 confirming that sialic acid is the primary glycocalyx component on the surface of human RBCs (provides 80%–90% negative charge), whereas heparan sulfate and hyaluronic acid are found at lower levels. We assumed that most of the sialic acid was removed by neuraminidase (Fig. 3). The Calcein-AM fluorescence intensity showed no significant reduction, indicating that cell viability remained unaffected by the enzyme treatment, which ensures the reliability of subsequent partitioning experiments.

In human RBC partitioning experiments, metoprolol and atenolol showed statistically significant differences between control and enzyme-treated groups. For labetalol, no statistically significant differences were observed at any incubation time. For chloroquine, no statistically significant difference was observed with human RBCs with the 5-minute incubation, whereas a significant difference was observed at the 1-minute incubation. Labetalol (the $\text{log}$ of the Kp of a solute between octanol and water [LogP]: 2.7–3.1) and chloroquine (LogP: 4.27–5.28) have higher lipophilicity compared with metoprolol (LogP: 1.88) and atenolol (LogP: 0.16), making them easier to partition into cell membranes.⁵⁷ It is plausible that the impact of the glycocalyx is minimal compared with membrane partitioning due to high lipophilicity of these drugs. The chloroquine results are consistent with its high RBC partitioning. Glycocalyx disruption removes a barrier for chloroquine to now bind faster to cellular components, evidenced by distribution equilibrium achieved by 5 minutes but not at 1 minute. Thus, chloroquine is an example where the assumption underlying eq. 2 (minimal binding to intracellular components) may not hold.

Despite the statistically significant differences observed for metoprolol and atenolol, the magnitude of these differences was small. In addition, no significant differences were observed for labetalol and chloroquine. Results can be explained by several possibilities: (1) the disruption of the glycocalyx in the present studies was incomplete, (2) the glycocalyx does not significantly affect drug distribution under the experimental conditions, and (3) negatively charged phosphatidylserine is typically located on the inner side of the plasma membrane and has negative charges.^58,59 After enzyme treatment, RBC damage may possibly expose negative charges in the membrane that partially make up for the loss of surface charge caused by glycocalyx disruption, or (4) the glycocalyx serves as a barrier to membrane partitioning for some drugs; therefore, its disruption may increase the rate of partitioning.

Because the endothelial glycocalyx is more GAG rich and can carry stronger negative charge than the sialic acid-dominated RBC surface,⁶⁰ cation binding may differ, and this may affect the magnitude of the translated glycocalyx-associated partitioning in the model. Nevertheless, whether the observed modest changes in drug distribution into RBCs on glycocalyx disruption would predict distribution rate in vivo was tested with modeling and simulation. Because of the wide variability in reported values for endothelial SA and glycocalyx thickness across different literature, the calculated Kp for ECs showed a wide range with method 1, with a narrower range of scaled values when using method 2 or 3 (Table 4). Furthermore, ${\text{Kp}}_{\text{glyco}}$ values for both drugs with all scaling methods resulted in similar accuracy of C-t predictions (Table 5; Figs. 4 and 5). Similar scalars were calculated for both drugs to correct the underpredicted distribution into the shallow compartment. Additional factors beyond SA and thickness, such as shear-dependent glycocalyx conformation,^61,62 local charge density and microstructural features⁶³ can influence the spatial presentation of negative charges and thereby affect partitioning of cationic drugs into the glycocalyx. Additional data are needed to evaluate these factors.

These results suggest that the proposed methods of calculating ${\text{Kp}}_{\text{glyco}}$ worked reasonably well by incorporating glycocalyx dimensions and combining it with in vitro experimental data. This approach can reproduce results comparable to those obtained through previously reported empirical methods. Our results support the hypothesis that glycocalyx partitioning may explain, at least in part, the early distribution of the basic drugs metoprolol and atenolol. Other hypotheses to explain the early distribution of basic drugs besides partitioning into endothelial glycocalyx have been proposed by others.^64–66 In humans, a difference in the venous sampling site has been proposed as a possible explanation. The addition of a forearm compartment for sampling has been proposed to result in lower initial peripheral venous concentrations relative to central venous sampling.

To better understand and evaluate the effect of the glycocalyx on drug distribution, more accurate and reliable data on endothelial SA and glycocalyx thickness are required. Due to the lack of data on the surface charge of individual ECs, it is currently not possible to directly establish a connection between the ratio of total charge of individual EC to the total charge of individual RBC and propose calculation methods. The data presented here are preliminary, and more basic drugs will need to be tested. The experimental protocol developed in this work is new, and additional in vitro tools to study glycocalyx partitioning will be necessary. Finally, in vivo experimental models for glycocalyx disruption in conditions such as sepsis, shock, and diabetes are available.^67–71 These complex models are difficult to set up, and their use to evaluate drug distribution alone is not ethically justifiable. Collection of drug distribution data as a secondary endpoint in an ongoing study may be feasible. However, these models are designed to induce inflammatory markers and can add confounding variables such as altered plasma protein binding and drug metabolizing enzyme expression.^72–76

In conclusion, RBCs were tested as a potentially facile in vitro surrogate for the endothelial glycocalyx to evaluate partitioning. Given the differences between RBC and endothelial glycocalyx, it is unclear whether RBCs can be used as an in vitro model for the endothelial glycocalyx. Glycocalyx disruption via enzymes exhibited species-specific differences in rat versus human RBCs. With human RBCs with intact versus disrupted glycocalyx, the observed differences in partitioning of metoprolol and atenolol were consistent with PermQ model predictions. These results offer preliminary support for the hypothesis that the endothelial glycocalyx contributes to the early distribution of basic drugs.

Supplementary Material

This article has supplemental material available at dmd.aspetjournals.org

Significance Statement:

We explored whether basic drugs can partition into glycocalyx, as this partitioning may explain the observed early time point distribution profile of basic drugs. Erythrocyte glycocalyx was tested as a possible in vitro model for the endothelial cell glycocalyx. Results indicate that glycocalyx partitioning partially explains the observed data for atenolol and metoprolol in humans.

Abbreviations

ACN

acetonitrile

BP

blood-to-plasma ratio

${\text{BP}}_{\text{C}0}$

BP due to external association with erythrocytes

C-t

concentration-time profile

EC

endothelial cell

EOC

exposure overlap coefficient

${\text{f}}_{\text{up}}$

unbound fraction in the plasma

$g$

relative centrifugal force

GAG

glycosaminoglycan

h

thickness of glycocalyx

HPLC

high-performance liquid chromatography

IS

internal standard

Kp

tissue to plasma partition coefficient or partition coefficient

${\text{Kp}}_{+\text{glyco}}$

partition coefficient with glycocalyx

${\text{Kp}}_{\text{glyco-EC}}$

endothelial cell glycocalyx partition coefficient

${\text{Kp}}_{\text{glyco-RBC}}$

red blood cell glycocalyx partition coefficient

LC-MS/MS

liquid chromatography-triple quadrupole mass spectrometer

logP

the log of the partition coefficient of a solute between octanol and water

PBPK

physiologically based pharmacokinetic

QC

quality control

RBC

red blood cell

${\text{R}}_{\text{glyco}}$

scaling factor for RBC-to-EC glycocalyx

SA

surface area

U

international unit for enzyme

${\text{V}}_{\text{glyco-EC}}$

glycocalyx volume per endothelial cell

${\text{V}}_{\text{glyco-RBC}}$

glycocalyx volume per red blood cell

WGA

wheat germ agglutinin

Data availability

The authors declare that all the data supporting the findings of this study are available within the paper and its Supplemental Material.