Nanocarriers’ repartitioning of drugs between blood subcompartments as a mechanism of improving pharmacokinetics, safety, and efficacy

Abstract

For medical emergencies, such as acute ischemic stroke, rapid drug delivery to the target site is essential. For many small molecule drugs, this goal is unachievable due to poor solubility that prevents intravenous administration, and less obviously, by extensive partitioning to plasma proteins and red blood cells (RBCs), which greatly slows delivery to the target. Here we study these effects and how they can be solved by loading into nanoscale drug carriers. We focus on fingolimod, a small molecule drug that is FDA-approved for treatment of multiple sclerosis, which has also shown promise in the treatment of stroke. Unfortunately, fingolimod has poor solubility and very extensive partitioning to plasma proteins and RBCs (in whole blood, 86% partitions to RBCs, 13.96% to plasma proteins, and 0.04% is free). We develop a liposomal formulation that slows the partitioning of fingolimod to RBCs and plasma proteins, enables intravenous delivery, and additionally prevents fingolimod toxicity to RBCs. The liposomal formulation nearly completely prevented fingolimod adsorption to plasma proteins (association with plasma proteins was 98.4 ± 0.4% for the free drug vs. 5.6 ± 0.4% for liposome-loaded drug). When incubated with whole blood in vitro, the liposomal formulation greatly slowed partitioning of fingolimod to RBCs and also eliminated deleterious effects of fingolimod on RBC rigidity, morphology, and hemolysis. In vivo, the liposomal formulation delayed fingolimod partitioning to RBCs for over 30 minutes, a critical time window for stroke. Fingolimod-loaded liposomes showed improved efficacy in a mouse model of post-stroke neuroinflammation, completely sealing the leaky blood-brain barrier (114 ± 11.5% reduction in albumin leak into the brain for targeted liposomes vs. 38 ± 16.5% reduction for free drug). This effect was only seen for liposomes modified with antibodies to enable targeted delivery to the site of action, and not in unmodified, long-circulating liposomes. Thus, loading fingolimod into liposomes prevented partitioning to RBCs and associated toxicities and enabled targeted delivery. This paradigm can be used for tuning the blood distribution of small molecule drugs for the treatment of acute illnesses requiring rapid pharmacologic intervention.

Introduction

Fingolimod is a sphingosine-1-phosphate receptor (S1PR) modulator that is FDA-approved for the treatment of multiple sclerosis, and has also shown promise for treatment of stroke in preclinical studies and early clinical trials [1,2]. In the context of stroke, fingolimod and other S1PR modulators restore the barrier function of brain endothelial cells, thus decreasing post-stroke brain inflammation and ameliorating damage in animal models of acute ischemic stroke [3]. However, fingolimod has key problems that may limit its application to acute illnesses such as stroke.

First, fingolimod is poorly soluble at physiologic pH and will form micelles under certain conditions, with a critical micellar concentration of 75 uM [4,5]. There are risks associated with intravenous (IV) administration of poorly soluble drugs and micelles, such as disruption of biologic membranes and stability in the physiologic environment [6–9]. IV administration is desirable for treatment of acute diseases because it enables immediate delivery of drug into circulation, where it can then distribute to the site of action and produce a rapid pharmacologic effect. Fingolimod is only clinically available as an oral drug, and the time required for absorption from the intestines delays distribution to the site of action. Treating post-stroke neuroinflammation requires a very short time to peak drug concentration in the brain, in order to stabilize the capillary barrier immediately after removal of the clot in the brain and thereby prevent ingress of injurious plasma proteins and activated leukocytes. In acute ischemic stroke, “time is brain”, and every minute without treatment worsens outcomes [10].

Additionally, fingolimod partitions very strongly to RBCs and plasma proteins, with only 0.04% of the drug being free in whole blood [11]. This partitioning likely delays distribution to tissue [12,13]. In rats, the maximum concentration of fingolimod in the brain was not reached until 24 to 48 hours after IV administration [14]. Finally, the strong partitioning of fingolimod to RBCs (86% of whole blood’s drug mass being contained in RBCs [11]) could perturb RBC rheology, and altered RBC rheology can lead to hemolysis [15], which is particularly dangerous in acute critical illnesses such as stroke. The prescribing information for fingolimod lists hemolytic anemia as a possible adverse reaction, and cases of fingolimod-associated hemolytic anemia have been reported [16–19].

We hypothesized that using nanoparticles such as liposomes for the delivery of fingolimod is a possible approach to overcome the above challenges. First, liposome formulation may avoid potential toxicities associated with intravenous administration of hydrophobic or micellar drugs. Although liposome formulation may not affect the solubility of fingolimod, the change in colloid form from micelles to liposomes may prevent toxicities, such as disruption of biologic membranes, and improve stability in the physiologic environment. Second, nanoparticles such as liposomes and polymer nanoparticles are often used to improve the pharmacokinetics of small molecule drugs via three mechanisms: 1) extending circulation time 2) preventing distribution to off-target tissues and 3) maximizing delivery to the target tissue [20–23]. However, these foci have obscured another challenge that nanomedicine could potentially solve: drug partitioning to the preferred compartments within blood. The partitioning of drug between RBCs and plasma proteins can have major effects on metabolism, distribution to tissue, and ultimately efficacy and safety [24–26]. In general, strong binding to plasma proteins and RBCs increases the retention of drug in the blood, decreases the rate of metabolism and excretion, and decreases the rate and extent of tissue distribution [12,13,27]. Therefore, the strong partitioning of fingolimod to RBCs and plasma proteins may impair its application to acute illnesses, where rapid distribution to the site of action is desirable.

Thus, fingolimod should serve as an ideal test drug to investigate how nanocarriers can repartition drugs between blood compartments, and the potential benefits of this approach. We hypothesized that formulating fingolimod in a nanocarrier could prevent partitioning of therapeutics to RBCs, therefore improving efficacy and safety in the context of acute critical illnesses. Here, we describe a liposome formulation that stably encapsulates fingolimod. The liposome formulation, when incubated in whole blood, significantly delays partitioning of fingolimod to plasma proteins and RBCs. Furthermore, the liposomes prevented negative effects of fingolimod on RBCs and showed improved efficacy in a preclinical mouse model of acute neuroinflammation (a major consequence of stroke).

Materials and Methods

Animal Studies and Ethics Statement

All animal studies were carried out in strict accordance with Guide for the Care and Use of Laboratory Animals as adopted by the National Institutes of Health, and approved by the University of Pennsylvania Institutional Animal Care and Use Committee (IACUC). All experiments were performed in accordance with relevant guidelines and regulations. Experiments were performed on C57BL/6J male mice purchased from Jackson Laboratories, aged 6-8 weeks. Mice were maintained at 22–26°C and on a 12/12 hour dark/light cycle with food and water ad libitum.

Materials

DPPC (dipalmitoyl phosphatidylcholine), cholesterol, soy PC (soy phosphatidylcholine), egg PG (phosphatidylglycerol), TopFluor PC (1-palmitoyl-2-(dipyrrometheneboron difluoride)undecanoyl-sn-glycero-3-phosphocholine), 18:0 PE-DTPA (1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-diethylenetriaminepentaacetic acid (ammonium salt)), and DSPE-PEG2000-azide (1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[azido(polyethylene glycol)-2000] (ammonium salt)) were purchased from Avanti Polar Lipids (Alabaster, Alabama). D4-fingolimod was purchased from Cayman Chemical (Ann Arbor, Michigan). Fingolimod-hydrochloride was purchased from TCI America (Portland, Oregon). Sodium hydroxide, hydrochloric acid, chloroform, tert-butyl-methyl ether, and LC-MS grade acetonitrile, methanol, trifluoroacetic acid, and formic acid were purchased from Fisher Scientific (Hampton, New Hampshire). Plasma (Sprague Dawley, anti-coagulant: K₂ EDTA) was purchased from Innovative Research (Novi, Michigan). A modified Lowry assay kit (DC Protein Assay) was purchased from Bio-Rad Laboratories (Hercules, California). Sepharose 4B-CL size-exclusion chromatography resin was purchased from Sigma Aldrich (St. Louis, Missouri). A Wright Stain Kit was purchased from Volu-Sol (Salt Lake City, Utah). Non-specific rat IgG was purchased from ThermoFisher Scientific (Waltham, Massachusetts). Anti-VCAM antibody (Clone M/K-2.7) was purchased from BioXCell (Lebanon, New Hampshire). ¹¹¹Indium Chloride (¹¹¹In) was purchased from BWXT Medical (Kanata, Ontario, Canada). Silica gel sheets for thin-layer chromatography were purchased from Millipore (Burlington, Massachusetts). All other chemicals and reagents were purchased from SigmaAldrich (St. Louis, Missouri), unless specifically noted. The gamma counter used for measuring radioactivity was a 2470 Wizard² (PerkinElmer, Waltham, MA).

Liposome preparation and characterization

Liposomes were formulated using the thin-film hydration method. Lipids were dissolved in chloroform and combined in a borosilicate glass tube. Fingolimod hydrochloride was dissolved in ethanol and added to lipids in the glass tube, at a drug-to-lipid ratio (mol/mol) of 0.01. For plasma leak experiments, liposomes were fluorescently labeled by adding 0.01 mol% of TopFluor PC lipid. Chloroform and ethanol were evaporated by blowing nitrogen over the solution until visibly dry (~15 minutes) then putting the tube under vacuum for >1hr. Dried lipid films were hydrated with a normal saline and 10mM HEPES buffered saline (HBS) at pH = 6.8 to a total lipid concentration of 20 mM. HBS was used because we observed that fingolimod precipitated when dissolved in phosphate buffered saline (PBS, pH=7.4), but not in HBS (Supplemental Figure 1). The rehydrated lipid solution was vortexed and sonicated in a bath sonicator until visually homogeneous (approximately 1 minute each of vortexing and sonication). The solution was then extruded twenty-one times through a 0.2 μm polycarbonate filter. Liposomes made with DPPC were heated to ~50°C (Just above the phase transition temperature of DPPC) during vortexing and extrusion steps.

To measure drug loading efficiency, drug not encapsulated in liposomes was removed by passing the liposome solution through a Zeba Spin desalting column (7kDa MWCO, 0.5 mL) according to manufacturer’s instructions. Drug concentration before and after purification was measured via UPLC. Drug encapsulation efficiency was calculated as: (Drug concentration after purification) / (Drug concentration before purification) x 100%.

Dynamic light scattering measurements of hydrodynamic particle size, distribution, polydispersity index, and zeta potential were made using a ZetaSizer Advance Series, Pro (Red), equipped with a 10 mW He–Ne laser at a wavelength of 633 nm. (Malvern Panalytical, Malvern UK). Samples for size measurement were prepared by diluting liposomes into PBS. A detection angle of 173° was used for size measurements. Samples for zeta potential were prepared by diluting liposomes into 0.1X PBS.

Antibody modification and conjugation to liposomes

To conjugate antibodies to immunoliposomes, antibodies were functionalized with dibenzocyclooctyne (DBCO) by reacting antibodies with a 5-fold molar excess of DBCO-PEG4-NHS ester for 30 minutes at ambient temperature. The unreacted DBCO-PEG4-NHS ester was removed via centrifugation using an Amicon 10 kDa molecular weight cutoff centrifugal filter. Liposome conjugation to antibodies was carried out using DBCO-azide copper-free “click chemistry”. Azide functionalized liposomes were incubated overnight with DBCO-modified antibodies at 37°C. Immunoliposomes (50 mAbs/liposome) were purified from residual antibodies using size-exclusion chromatography (1.5x20 cm Sepharose 4B-CL column).

UPLC analysis of drug concentration

Drug concentration in liposomes was quantified via ultra performance liquid chromatography (UPLC). A Waters Acquity I-Class UPLC system was used, consisting of a Binary Solvent Manager, Sample Manager-FTN, and Photodiode Array (PDA) Detector set to a detection wavelength of 220nm. Chromatography was performed using a gradient method and a Cortecs C18 column (1.6μm particle size, 2.1x50mm). Mobile phase A was water with 0.1 vol% trifluoroacetic acid (TFA); mobile phase B was acetonitrile with 0.1 vol% TFA. The flow rate was 0.5 mL/min with a gradient of 10% B from 0 to 0.4 min, linearly increasing to 90% B from 0.4 to 0.8 min, holding at 90% B from 0.8 to 1.3 min, then linearly decreasing to 10% B from 1.3 to 1.4 min. The sample injection volume was 8μL, and the retention time for fingolimod was ~1.17min.

A standard curve was created by measuring solutions with known concentrations of fingolimod. A stock solution of fingolimod (5 mg/mL) was prepared in ethanol. Working standards were prepared by diluting the stock solution into a 50/50 (vol/vol) mixture of mobile phases A and B. Liposome samples were prepared by diluting the liposomes 100-fold into a 50/50 (vol/vol) mixture of mobile phases A and B.

Drug release from liposomes in storage and dialysis conditions

Liposomes were prepared via the thin-film method as specified above. Drug not encapsulated in liposomes was removed by passing the liposome solution through a Zeba Spin desalting column. Sequential purifications were performed over the course of 4 days. Liposomes were stored in buffer (normal saline and 10 mM HEPES buffer at pH = 6.8) at 4°C for the duration of the leak study. After each purification step, drug concentration in the liposomes was measured via UPLC. For drug release using the dialysis method, drug-loaded liposomes were added to dialysis tubing (MWCO = 10 kDa). The dialysis tubing was submerged in a PBS volume > 500 times that of the liposome solution and stirred at 25°C. Liposome samples from dialysis tubing were collected at varying time points and analyzed for fingolimod concentration via UPLC.

Drug release from liposomes in plasma

Fluorescently labeled liposomes were diluted into rat plasma to a drug concentration of 0.01 mg/mL. The mixture was incubated at 37°C for various lengths of time. Liposomes were then separated from free fingolimod and plasma proteins using a 1.5x20cm column packed with Sepharose 4B-CL resin (GE Healthcare, Pittsburg PA). Eluent was collected in 1mL fractions (24 fractions total). Protein concentration was measured in each fraction using a Lowry assay per manufacturer’s instructions. Solutions with known concentrations of bovine serum albumin were used as standards. The presence of liposomes in eluent fractions was confirmed via measurement of fluorescence intensity. 200μL of each fraction were pipetted into a black 96-well plate. Fluorescence intensity in each well was measured via a plate reader at 490 nm excitation / 510 nm emission (SpectraMax M2, Molecular Devices, San Jose, CA).

Drug concentration in each fraction was measured via LC-MS. Drug was extracted from plasma samples as follows: 100 μL of sample, 100 μL DI water, 40μL methanol (with internal standard d4-fingolimod), and 800μL extraction solvent (chloroform:methanol:HCl, 83:15:2, v/v/v) were mixed. This leads to separation into an organic layer and aqueous layer with protein precipitate in between. The mixture was vortexed for 1 minute and centrifuged at 15,000 g for 5 minutes. The bottom organic layer was transferred to a clean tube and evaporated under a nitrogen stream for 5 minutes. The residue was reconstituted with 100 μL of 50/50 water/acetonitrile with 0.1% TFA then analyzed via LC-MS.

Drug partitioning in whole blood

Fresh mouse blood was collected from the inferior vena cava of C57BL/6J mice. EDTA was used as anticoagulant. Whole blood was separated into 150 μL aliquots. 3.75 μL of fingolimod solution (0.05 mg/mL fingolimod as free drug or in liposome form) was added to each aliquot for a final drug concentration of 1.25 ug/mL (approximately the initial drug concentration in blood after a 0.1 mg/kg dose in mice). The sample was incubated at 37°C for 30 minutes, then centrifuged at 2000 g for 10 minutes. Plasma and red blood cell pellet samples were analyzed for fingolimod concentration via the following protocol (Adapted from [14]). For red blood cell pellet, 20 μL of pellet was added to 80 μL PBS. For plasma, 50 μL of plasma was added to 50 μL of PBS. To all samples, 40 μL of methanol (with internal standard d4-fingolimod), 40 μL of calibration standard or blank, and 100 μL of 0.1 M NaOH were added to the sample. 2mL of extraction solvent (methyl tert-butyl ether/dichloromethane 75/25 v/v) was added. The mixture was shaken for 1hr then centrifuged at 2000 g for 10 minutes. The top organic layer was transferred to a clean tube and evaporated under a nitrogen stream until dry (~15 minutes). The residue was reconstituted with 100 μL of 50/50 water/acetonitrile with 0.1% TFA then analyzed via LC-MS.

LC-MS for in vitro studies in plasma and blood

The system was a Waters Acquity I-Class UPLC, consisting of a Binary Solvent Manager, Sample Manager, and SQD detector. Chromatography was performed using a gradient method and a HSS C18 column (1.8 μm particle size, 2.1x50 mm). Mobile phase A was water with 0.1 vol% trifluoroacetic acid (TFA); mobile phase B was acetonitrile with 0.1 vol% TFA. The flow rate was 0.5 mL/min with a gradient of 5% B from 0 to 0.5 min, linearly increasing to 95% B from 0.5 to 2.5 min, holding at 95% B from 2.5 to 3 min, then linearly decreasing to 5% B from 3 to 3.1 min. The sample injection volume was 7μL. The retention time for fingolimod was ~2.1 min. The mass spectrometer was operated in positive ion mode and Selective Ion Recording (SIR). The cone voltage was 30V, the capillary voltage was 3kV, the source temperature was 150°C, the desolvation temperature was 350°C, and the desolvation gas flow was 650 L/hr. Fingolimod was detected at m/z = 308.3 and d4-fingolimod was detected at m/z = 312.3.

Calculating fraction of drug in compartments of whole blood

Using the known partitioning of fingolimod between plasma and RBCs, and the known adsorption of liposomes to RBCs, we created the following system of equations:

(1)	$C_{plasma}^{fingolimod,free}+C_{plasma}^{fingolimod,lipo}=C_{plasma}^{Fingolimod,total}$	Mass balance of drug in plasma
(2)	$C_{RBC}^{fingolimod,free}+C_{RBC}^{fingolimod,lipo}=C_{RBC}^{Fingolimod,total}$	Mass balance of drug in RBCs
(3)	$\frac{1}{124}{C}_{plasma}^{fingolimod,lipo}=C_{RBC}^{fingolimod,lipo}$	Known RBC adsorption of liposomes. 0.8% of liposomes adsorb to RBCs. RBC:plasma ratio is $\frac{1}{124}$ (Figure 3D)
(4)	$6.88\ast C_{plasma}^{fingolimod,free}=C_{RBC}^{fingolimod,free}$	Known partitioning of fingolimod. RBC:plasma ratio for fingolimod is 6.88 (Figure 3C)

The four equations contain four unknowns, as follows: $C_{plasma}^{fingolimod,lipo}$ and $C_{RBC}^{fingolimod,lipo}$ are the concentration of liposome-bound fingolimod in plasma and adsorbed to RBCs, respectively; $C_{plasma}^{fingolimod,free}$ and $C_{RBC}^{fingolimod,free}$ are the concentration of fingolimod (free drug) bound to plasma proteins and RBCs, respectively. P_lipo is the RBC:plasma ratio of liposomes, to be determined experimentally. It is measured in Figure 3D of this manuscript as 1/124 (0.8% of liposomes are adsorbed to RBCs). P-_Fingolimod is the RBC:plasma ratio of fingolimod, to be determined experimentally. It is measured in Figure 3C of this manuscript as 6.88.

The equations were then rearranged to fit the format of the Matrix Equation (Ax=b, below) and solved for ‘x’ using the Inverse Matrix Method [28].

Matrix equation:

$$A=\begin{array}{cccc}1& 1& 0& 0\\ 0& 0& 1& 1\\ 0& \frac{1}{124}& 0& -1\\ 6.88& 0& -1& 0\end{array}x=\begin{array}{c}{C}_{plasma}^{fingolimod,free}\\ C_{plasma}^{fingolimod,lipo}\\ C_{RBC}^{fingolimod,free}\\ C_{RBC}^{fingolimod,lipo}\end{array}b=\begin{array}{c}{C}_{plasma}^{Fingolimod,total}\\ C_{RBC}^{Fingolimod,total}\\ 0\\ 0\end{array}$$

Pharmacokinetic dosing and sampling

Fingolimod (either as free drug or in liposomes) was administered to mice (C57BL/6J male, age 6-8 weeks), as a single intravenous bolus via retro-orbital injection, at a dose of 0.3 mg/kg. After 0.5, 1, 4, and 8 hours, they were anesthetized with isoflurane and blood was withdrawn from the inferior vena cava into EDTA-coated plastic syringes. Samples were processed as described previously [14] and adapted as follows: Blood was transferred to a polypropylene tube and centrifuged at 2000 g for 10 minutes. A 400 μL portion of plasma and a 200 μL red blood cell pellet were transferred to borosilicate glass tubes, each sample was brought to a volume of 500 μL with PBS, and 50 μL of internal standard, 50 μL of calibration standard, a quality control standard or blank, and 500 μL of 0.1 M NaOH were added. To each plasma and red blood cell pellet sample (in borosilicate glass tubes), 5 mL of a mixture of tert-butyl-methyl ether and dichloromethane (75:25 v/v) was added. Each sample was then shaken for 1 hour and centrifuged at 2000 g for 10 minutes. The upper organic phase was transferred to a clean borosilicate glass tube, evaporated under nitrogen, and 100 μL of 50/50 water/acetonitrile + 0.1% formic acid was added. This solution was bath sonicated for 5 minutes and transferred to a 350 μL autosampler tube.

LC-MS/MS for in vivo pharmacokinetic studies

Liquid chromatography/tandem mass spectrometry (LC-MS/MS) was performed on an AB Sciex API 4000 electrospray ionization mass spectrometer using a Zorbax SP-C18, 3.5 μm, 1 mm x 150 mm column (Agilent). Solution A was 5% acetonitrile and 0.1 % formic acid in water. Solution B was 0.1 % formic acid in acetonitrile. Solutions were pumped at 0.1 mL/min at 10% B for 1 min, increasing from 10 to 40% B between 1.0 and 2.1 min, and then from 40 to 100% B between 7 and 14 min. Fingolimod eluted at 10.5 minutes.

The mass spectrometer was operated in positive ion mode with a declustering potential of 75V, an ionization energy of 5500V, and a collision gas, curtain gas, and Gas 1 of nitrogen set at 7, 10, and 40 psi, respectively. Fingolimod was detected by monitoring the m/z 308.3 → 255.2 transition, and d4-fingolimod was detected by monitoring the m/z 312.3 → 259.2 transition. Peak areas were measured with the same Analyst software used to control the spectrometer. The LLOQ peaks had a signal-to-noise-ratio greater than 10 (relative to Root Mean Squared (RMS) baseline noise).

Calibration standards and quality control (QC) standards were prepared by weighing 10.0 mg of fingolimod hydrochloride and adding 2.0 mL of ethanol (using a calibrated pipette) to make 5.0 mg/mL solutions. Working solutions were prepared by diluting the stock solution into a 50/50 (vol/vol) mixture of methanol and water. A calibration curve for each sample type was created by adding known amounts of calibration standard to samples from untreated mice. The ratio of fingolimod to internal standard peak areas was plotted against the known drug mass and a linear regression (weighted by 1/(drug mass)² ) was performed. The calibration standards ranged from 67.5 to 2.18 ng/mL. QC samples were used to verify the calibration curve. QC samples were prepared by adding a known amount of QC standards to samples from untreated mice. The QC standards were 50, 12.5, and 2.5 ng/mL.

Non-compartmental analysis of pharmacokinetic parameters

Non-compartmental analysis was performed using the PKNCA package (Version 0.10.2) in R (Version 4.3.3). Area under the curve (AUC) was determined using a sparse NCA algorithm, as each animal contributed one measurement to the dataset. Maximum concentration (Cmax), AUC, and clearance were calculated for each compartment. Volume of distribution at steady state (Vss) was only calculated for whole blood, not for each individual compartment. Fingolimod concentration in whole blood was calculated using a weighted average of fingolimod concentration in plasma and RBC, assuming a hematocrit of 45% (Concentration in whole blood = 0.45 x concentration in RBC + 0.55 x concentration in plasma).

Ektacytometry

Measurements of RBC mechanical properties were performed using a RheoScan-D system (RheoMeditech, Seoul, Republic of Korea). Briefly, 50 μL of RBCs (5% hematocrit) was resuspended in 700 μL of a 5.5% w/v solution of polyvinylpyrrolidone (360 kDa) in phosphate buffered saline. The RBC suspensions were subjected to shear stress varying from ~0-20 Pa in ~190-200 μm depth rectangular channels. In the RheoScan-D software, ellipses were fit to diffraction patterns generated by the RBC suspensions, and the long (L) and short (S) axes of the best fit ellipses were recorded, alongside the shear stress (tau) at which the diffraction patterns were generated. RBC elongation indices (EI) were determined for diffraction pattern ellipses as EI = (L - S)/(L + S). Plots of EI vs. tau were fit to the Streekstra-Bronkhorst model for RBC deformation in flow, as described previously [29] to extrapolate values for maximum EI at infinite shear and shear stress at half maximum EI.

RBC Hemolysis

Fresh mouse blood was collected from the inferior vena cava of C57BL/6J mice. EDTA was used as anticoagulant. Blood was centrifuged at 1000 g for 10 minutes and the plasma and buffy coat were discarded. RBCs were diluted to 1% hematocrit with normal saline. RBCs were divided into 1mL aliquots and heated to 37°C. Then, 25 μL of fingolimod solution (free or liposomal) or vehicle control was added to 1 mL of 1% hematocrit RBCs. The tube was inverted for 5 seconds, then centrifuged at 13,400 g for 1 minute. The absorbance of the supernatant was measured at 410 nm as a measurement of hemolysis. As a control for 100% hemolysis, RBCs were diluted to 1% hematocrit with DI water.

The solution of fingolimod free drug was prepared by diluting a 5 mg/mL stock solution of fingolimod (in ethanol) using normal saline. The vehicle control contained an equal amount of ethanol in normal saline.

Thin smear of red blood cells

Fresh mouse blood was collected from the inferior vena cava of C57BL/6J mice. EDTA was used as anticoagulant. Blood was centrifuged at 1000 g for 10 minutes and the plasma and buffy coat were discarded. RBCs were diluted to 5% hematocrit with normal saline. RBCs were divided into 200 μL aliquots and heated to 37°C. Then, 5 μL of fingolimod solution (0.5 mg/mL fingolimod concentration, free or liposomal) or vehicle control was added to the RBC aliquot. The tube was inverted for 5 seconds, then kept at ambient temperature for the smearing procedure. The smear was prepared using the ‘push’/wedge method [30]. The dried smear was fixed in methanol for 1 minute, then dried and stained using a Wright stain per manufacturer’s instructions. Smears were imaged using light microscopy.

RBC adsorption

Radiotracing was used for studying the absorption of mAb-coated liposomes to RBCs. Non-specific IgG was radiolabeled with Na ¹²⁵I using Pierce Iodogen radiolabeling method. Briefly, tubes were coated with 100 μg of Iodogen reagent. The antibody (1-2 mg/mL) and Na ¹²⁵I (0.25 μCi/μg protein) were incubated for 5 minutes on ice. The materials were purified using a Zeba Spin desalting column. Liposomes were incubated with whole blood for 30 minutes under constant rotation at 37°C. Liposome:RBC suspensions were thrice centrifuged at 100 g for 8 minutes to pellet red cells, remove supernatants, and washed with PBS to remove unattached liposomes.

Liposome biodistribution studies

For biodistribution studies, liposomes were traced with ¹¹¹In as previously described [31]. Liposomes were produced with 0.2 mol% of 18:0 PE-DTPA (a chelator-containing lipid). The dried lipid film was rehydrated with a metal-free buffer of 30 mM citrate, 150 mM sodium chloride, pH = 5.0. Trace metals were removed from the citrate buffer using a Chelex 100 resin, per manufacturer’s instructions, to prevent unwanted occupancy of the chelator. Liposomes were produced as described above (vortexing, sonication, and extrusion). ¹¹¹In chloride was added to the liposomes at a specific activity of 1 μCi of ¹¹¹In per 1 umol of lipid. The mixture was incubated at ambient temperature for 30 minutes. Then, unincorporated ¹¹¹In was removed using a Zeba Spin desalting column. For antibody-targeted liposomes, antibody was conjugated to the liposomes after removal of unincorporated ¹¹¹In.

The removal of unincorporated ¹¹¹In was verified using thin film chromatography (TLC). A 1 μL sample of liposomes was applied to the stationary phase (silica gel strip). The strip was placed in the mobile phase of 10 mM EDTA until the solvent front was 1 cm from the end of the strip (~10 minutes). The strip was cut 1 cm above the initial sample location. ¹¹¹In chelated to the liposome stays at the origin, while unchelated ¹¹¹In travels with the solvent front. The activity in each section was measured using a gamma counter. The percent of ¹¹¹In chelated to the liposomes was calculated as the activity in the origin strip divided by the total activity in both strips. For all experiments, >95% of ¹¹¹In was chelated to the liposomes (Supplemental Figure 8).

For radiotracer biodistribution, TNF-α challenged mice were injected intravenously with radiolabeled liposomes at 16 hours-post TNF-α injury. For all groups, the dose was ~10 mg/kg lipid (6x10¹¹ liposomes/mouse). Prior to injection, the radioactivity of the dose was measured using a gamma counter. The total activity injected per mouse was ~0.5 μCi. All animals were exsanguinated 30 minutes after injection and perfused with 20mL of PBS through the left ventricle prior to collecting organs. The amount of radioactivity in blood and organs was measured using a gamma counter.

TNF-α animal model

A unilateral intrastriatal injection of TNF-α (0.5 μg in 2.5 μL) was performed after placing the anesthetized mice in a stereotaxic frame (coordinates: 0.5 mm anterior, 2.0 mm lateral, −3 mm ventral to the bregma) [31,32]. . Injured animals were injected into the right brain hemisphere with TNF-ɑ. Sham injected mice underwent the same surgical procedure, including insertion of syringe into the striatum, but no injection was performed.

Therapeutic Studies

The effects of fingolimod on TNF-α induced brain edema were assessed as described in our previous publication [31]. TNF-α is the one of the cytokines produced at the highest levels after ischemic stroke, and mediates much of the endothelial activation. In this model, TNF-α is stereotactically injected into the mouse striatum, which is a common region for strokes in humans. TNF-α injection induces brain edema, which serves as a convenient metric of the severity of neuroinflammation. Briefly, 16 hours post-TNF-α injection, mice were dosed IV with either: (1) PBS vehicle control, (2) 0.1 mg/kg fingolimod, or (3) 0.1 mg/kg liposomal fingolimod (either bare or coated with aVCAM antibodies). 20 hours after TNF-α injection, mice were injected with ¹²⁵I-labeled bovine serum albumin (BSA, ~2μCi/mouse), which was then allowed to circulate for 4 hours. After BSA circulation, Blood was collected and mice were perfused with 20 mL of PBS, pH 7.4, over 5 minutes and organs were harvested. Blood was immediately evaluated with complete blood count (CBC) measurements (Abaxis VetScan HM5). Edema was determined by measuring the relative concentration of extravasated BSA in the brain to the concentration in the bloodstream. For calculations of therapeutic efficacy, 0% protection was defined using PBS-treated, TNF-α injured mice and 100% protection was defined using PBS-treated, sham-injured mice.

Alanine aminotransferase (ALT) measurements

Healthy mice were dosed IV with 0.1 mg/kg of fingolimod (free or liposomal) or vehicle control. Blood was collected after 24 hours into EDTA coated tubes and plasma was prepared. Plasma was analyzed for ALT activity using a colorimetric assay per manufacturer’s instructions (Cayman Chemical, Ann Harbor, Michigan, Alanine Transaminase Colorimetric Activity Assay Kit).

Statistics

All results are expressed as mean ± SEM (Standard Error of the Mean) unless specified otherwise. Statistical analyses were performed using GraphPad Prism 8 (GraphPad Software, San Diego, CA). * denotes p<0.05, ** denotes p<0.01, *** denotes p<0.001, **** denotes p<0.0001.

Results

Development and characterization of liposome formulation for fingolimod

To evaluate the ability of nanocarriers to repartition fingolimod between blood compartments, we first needed to develop a nanoparticle formulation for fingolimod. We chose PEGylated liposomes as the nanocarrier, as liposomes are among the most clinically used nanoparticles based on the number of FDA-approved products [33], and fingolimod should load into the lipid bilayer because the drug is lipophilic (LogP = 5.5) and has a lipid-like structure (sphingosine analog) [34].

We produced fingolimod-loaded liposomes and varied the lipid formulation to maximize drug loading and minimize leak. First, we varied the amount of anionic lipid (Egg PG/phosphatidylglycerol) in the lipid formulation. Fingolimod is a weak base with a primary amine that has a pKa of 7.8 [5]. Therefore, at a neutral pH, the majority of drug molecules will be protonated and carry a +1 charge (Figure 1A), We included anionic lipids in the liposome formulation with the hypothesis that electrostatic attraction between drug and lipid would increase drug loading and reduce leak (Figure 1B). Drug loading increased as a function of Egg PG (72.6 ± 7.2% loading at 0 mol% Egg PG vs. 93.5 ± 0.8% loading at 15 mol% Egg PG) (Figure 1C). As expected, zeta potential became more negative with addition of Egg PG (Figure 1C). Liposome size for all formulations was in the range of 130 to 170 nm, and polydispersity index (PDI) was less than 0.2 for all formulations (Table 1, Supplemental Figure 2). This size range is typical for lipidic nanoparticles intended for intravenous administration, as particles between 50 and 200 nm exhibit reduced clearance by the liver, kidneys, and other clearance organs, via various mechanisms [35–37]. We also tested phospholipids with different levels of lipid tail saturation (Soy PC/unsaturated vs. DPPC/saturated) as alterations in phospholipid properties (e.g., fluidity, packing) may alter loading of hydrophobic drugs such as fingolimod. We found no difference in drug loading and leak (Supplemental Figure 2), and we chose to use Soy PC for all subsequent experiments.

Figure 1. Development and characterization of liposome formulation for fingolimod. Addition of anionic lipid improves fingolimod loading and delays leakage.

(A) Schematic of charge distribution of fingolimod at varying pH. The primary amine on fingolimod has a pKa of 7.8; therefore the majority of drug molecules are positively charged at neutral and slightly acidic pH. Species fraction calculated using the Henderson-Hasselbalch equation. (B) Diagram of nanoparticles used for controlled delivery of fingolimod. Liposomes were produced in a buffer of normal saline with 10mM of HEPES at pH=6.8 (> 1 log less than fingolimod’s amine’s pKa), in order to maximize the amount of positively charged drug molecules while keeping a neutral pH. The positive charge of drug and negative charge of the anionic lipids should experience electrostatic attraction, increasing the retention of drug in the lipid membrane. (C) Drug loading and zeta potential of liposomes with varying amounts of Egg PG (anionic lipid). The addition of the anionic lipid Egg PG increases drug loading and decreases zeta potential (n=3). (D) Fingolimod release under infinite dilution conditions, using a dialysis method. Addition of negatively charged lipid Egg PG slowed drug release from liposomes. Comparisons were made by 1-way ANOVA with Dunnett’s post hoc test using the 0% Egg PG formulation as the control (n=3).

Table 1. Characterization of fingolimod-loaded liposomes with different lipid formulations.

Size, PDI, and zeta potential were measured via DLS. Drug concentration and loading efficiency was measured via HPLC. Data presented as mean +/− standard deviation (n=3-4).

Formulation ID	Lipid composition [mol%]	Size [nm]	PDI	Zeta potential [mV]	Drug loading efficiency [%]
1	54% DPPC 0% Egg PG 40% Cholesterol 6% DSPE-PEG2k-azide	152.0 ± 2.1	0.09 ± 0.01	−11.8 ± 10.1	75.8 ± 4.5
2	54% Soy PC 0% Egg PG 40% Cholesterol 6% DSPE-PEG2k-azide	163.5 ± 1.7	0.06 ± 0.02	−6.3 ± 2.3	72.6 ± 7.2
3	49% Soy PC 5% Egg PG 40% Cholesterol 6% DSPE-PEG2k-azide	150.7 ± 1.5	0.06 ± 0.03	−7.7 ± 1.5	88.4 ± 2.7
4	44% Soy PC 10% Egg PG 40% Cholesterol 6% DSPE-PEG2k-azide	135.3 ± 4.8	0.146 ± 0.01	−9.1 ± 0.8	91.0 ± 1.5
Fing-Lipo	39% Soy PC 15% Egg PG 40% Cholesterol 6% DSPE-PEG2k-azide	145.9 ± 9.2	0.06 ± 0.04	−13.6 ± 0.7	93.5 ± 0.8
mAb-Fing-Lipo	Same as Fing-Lipo	161.3 ± 1.4	0.07 ± 0.01	−7.5 ± 0.5	N/A

To evaluate the rate of drug release, we then measured fingolimod release from liposomes under storage conditions (i.e. highly concentrated liposomes in a buffer solution at 4°C). There was negligible drug release after 4 days and addition of Egg PG had no significant effect on fingolimod release rate (Supplemental Figure 2). To further evaluate drug release, we measured release under infinite dilution conditions, using a dialysis method. In this assay, addition of Egg PG significantly reduced drug leak, with the 15 mol% Egg PG formulation performing better than all other formulations (0, 5, and 10 mol% Egg PG) (Figure 1D). We therefore used 15 mol% Egg PG for future experiments.

Liposome formulation delays association of fingolimod with plasma proteins

As a free drug, fingolimod is extensively bound to RBCs and plasma proteins. 86% of fingolimod is bound to RBCs [11], and of the remaining drug not bound to RBCs, 99.7% is protein-bound in plasma [11]. We therefore sought to determine if the liposomal formulation of fingolimod delays partitioning of drug to plasma proteins compared to the unencapsulated fingolimod.

We incubated free fingolimod (Fing), fingolimod liposomes (Fing-Lipo), or antibody-coated fingolimod liposomes (mAb-Fing-Lipo) with plasma at 37°C. The mixture was then passed through a size-exclusion chromatography (SEC) column to separate liposomes from plasma proteins (Figure 2A) [38]. For each fraction of eluent, fingolimod concentration was measured using liquid chromatography/mass spectrometry (LC-MS), protein concentration was measured using a Lowry assay, and liposome amount was measured using fluorescence. Representative elution profiles show that unencapsulated fingolimod predominantly localizes with plasma proteins (Figure 2B), while the liposome-formulated fingolimod remains associated with liposomes (Figure 2C). We then quantified the mass of fingolimod associated with plasma proteins and the mass in non-protein fractions (e.g., the mass encapsulated in liposomes) (Figure 2D).

Figure 2. Liposome formulation delays association of fingolimod with plasma proteins.

(A) Schematic illustration of method for quantifying the fraction of fingolimod bound to plasma protein. Fingolimod (Fing), fingolimod liposomes (Fing-Lipo), or antibody-coated fingolimod liposomes (mAb-Fing-Lipo) were incubated with plasma at 37°C then separated by SEC. For each fraction, fingolimod concentration was measured using LC-MS, protein concentration was measured using a Lowry assay and liposome presence was measured by fluorescence intensity. (B) Representative SEC elution profile of plasma association study of Fing (no nanocarrier) (n=3). More than 98% of fingolimod is associated with plasma proteins. (C) Representative SEC elution profile of plasma association of Fing-Lipo (n=3). Over 90% of the drug remained encapsulated within the liposomes. (D) Quantification of plots like those of B and C, displaying the mass of fingolimod associated with plasma proteins and the mass not associated with plasma proteins (e.g., mass remaining in liposomes) (n=3). (E) Quantification of the fraction of fingolimod associated with plasma proteins for multiple timepoints (n=3). Compared to free drug, there is a 16-fold reduction in association of fingolimod with plasma proteins when fingolimod is loaded into liposomes. The results were comparable for both Fing-Lipo and mAb-Fing-Lipo.

The results show that our liposome formulation delays partitioning of fingolimod to plasma proteins, leading to a 16-fold reduction in protein-bound drug compared to the free drug after 30 minutes incubation in plasma (98.4 ± 0.4% of drug associated with plasma proteins as free drug vs. 5.6 ± 0.4% for mAb-Fing-Lipo) (Figure 2E). Antibody-functionalized liposomes (mAb-Fing-Lipo) showed an increased release rate compared to unmodified liposomes (Fing-Lipo). At 60 minutes incubation time, Fing-Lipo exhibited 8.1 +/− 0.8% of drug associated with plasma proteins, and mAb-Fing-Lipo exhibited 16.1 +/− 0.8% of drug associated with plasma proteins. Comparison of Fing-Lipo to mAb-Fing-Lipo via 2-way ANOVA (using a Sidak’s multiple comparison across all time points) indicates that mAb-Fing-Lipo has significantly more drug release (p<0.001). However, even after 60 minutes incubation, 84% of fingolimod remained associated with mAb-Fing-Lipo, indicating that the liposome formulation is sufficiently capable of retaining fingolimod in presence of plasma to be used for acute delivery.

Liposome formulation changes blood distribution of fingolimod in vitro

Having shown that liposome formulation dramatically delays fingolimod partitioning to plasma proteins, we next sought to evaluate the impact of liposome formulation on fingolimod partitioning to RBCs. We incubated Fing, Fing-Lipo, or mAb-Fing-Lipo with whole blood collected from mice and centrifuged to separate plasma from RBCs. Drug concentration in both plasma and RBCs was measured via LC-MS (Figure 3A). For the free drug condition, 87.2 ± 0.8% of the drug partitioned into RBCs, which matches literature reports (Figure 3B, left bar) [11,39]. To quantify the effect of liposome formulation on partitioning, we calculated the RBC:plasma ratio of fingolimod concentration (Figure 3C). For Fing, the ratio was 6.84 ± 0.46. For Fing-Lipo and mAb-Fing-Lipo, the ratio was 0.49 ± 0.02 and 1.16 ± 0.15, respectively (p<0.0001 for each group compared to Fing). The dramatic decrease in RBC:plasma ratio indicates the liposome formulation delays partitioning of fingolimod to RBCs.

Figure 3. Liposome formulation changes blood distribution of fingolimod in vitro.

(A) Schematic for whole blood partitioning study. (B) Concentration of fingolimod in plasma and RBCs for free drug and liposomal formulations. Note that plasma contains two subcompartments: both drug that is bound to plasma proteins and drug encapsulated in liposomes. (n=3) (C) Calculation of RBC:Plasma ratio from data in (B). The ratio was 6-fold higher for Fing compared to Fing-Lipo or mAb-Fing-Lipo, indicating that the liposome formulation dramatically delays partitioning of fingolimod to RBCs.Comparisons were made by 1-way ANOVA with Tukey’s multiple comparisons test. (n=3) (D) Absorption of liposomes to red blood cells (RBCs) was quantified via radiotracing. For both conditions, minimal adsorption to RBCs (<1%) was observed. (n=3) (E) The plasma from panel (B) was passed through a SEC column to separate liposomes from plasma proteins. Fingolimod concentration in the eluent was measured to determine the fraction of fingolimod remaining in liposomes and the fraction partitioned to plasma proteins. A significant portion remains encapsulated in liposomes. (F) The concentration of drug in three compartments in blood (encapsulated in liposomes, bound to plasma proteins, and bound to RBCs) was calculated using the known partitioning of free drug and the known RBC adsorption of liposomes. The calculation agrees well with experimental results in panel (E). The two outputs were highly similar (p>0.999) as calculated by an ordinary two-way ANOVA, comparing the experimental and calculated results for both Fing-Lipo and mAb-Fing-Lipo.

It is possible for the nanoparticle itself to adsorb onto RBCs [40]. Therefore, we evaluated the adsorption of fingolimod liposomes to RBCs by measured the partitioning of the liposomes (both Fing-Lipo and mAb-Fing-Lipo) in whole blood, using a radiotracer (Figure 3D, Supplemental Figure 3). For Fing-Lipo, only 0.8 ± 0.1% of liposomes associated with RBCs; and for mAb-Fing-Lipo, only 0.6 ± 0.1% of liposomes associated with RBCs. In total, only a very small percentage of the liposomes adsorbed onto RBCs, and addition of a targeting moiety had little effect on RBC adsorption. Therefore, any fingolimod detected in RBCs (Figure 3B) is predominantly drug that was released from the liposomes and partitioned to RBCs.

Since fingolimod liposomes do not extensively adsorb to RBCs, the vast majority of liposomes are located in plasma. Therefore, fingolimod detected in plasma can either be liposome-bound or associated with plasma proteins. To evaluate the fraction of fingolimod in each compartment, we passed the plasma through a SEC column to separate liposomes from plasma proteins. We measured drug concentration in the eluent and calculated the fraction of drug encapsulated in liposomes vs. the fraction bound to plasma proteins (Figure 3E, Supplemental Figure 4). A significant fraction of drug remained encapsulated in the liposomes. For Fing-Lipo, 83.7 ± 13.6% of fingolimod in plasma remained encapsulated in liposomes. For mAb-Fing-Lipo, 58.7 ± 23.7% of fingolimod in plasma remained encapsulated in liposomes.

A technical challenge with using SEC is that the separation significantly dilutes the sample. We intend to do in vivo studies, where the concentration of fingolimod in plasma would be orders of magnitude smaller than in Figure 3B. Separating plasma via SEC would not be feasible, as it would dilute the sample below the limit of detection of the LC-MS methods. To overcome this challenge, we developed a numerical approach to calculate the fingolimod concentration in each compartment in whole blood (encapsulated in liposomes, bound to plasma proteins, and bound to RBCs). Since the distribution of liposomes between RBCs and plasma is known (Figure 3D), and the partitioning of free drug between RBCs and plasma proteins is known (Figure 3B, left column), it is possible to use mass balances and create a system of equations to calculate the concentration of fingolimod in all compartments (equations given in Methods section). Our calculations (Figure 3F) produce nearly identical results to the experimental measurements (Figure 3E). The two outputs were highly similar (p>0.999) as calculated by an ordinary two-way ANOVA, comparing the experimental (Figure 3E) and calculated (Figure 3F) results for both Fing-Lipo and mAb-Fing-Lipo.

Liposome formulation changes blood distribution and pharmacokinetic parameters of fingolimod in vivo

We next sought to evaluate how the liposome formulation affects fingolimod partitioning in vivo. We administered Fing, Fing-Lipo, or mAb-Fing-Lipo to mice via IV injection. We collected blood samples at varying time points after injection and measured the drug concentration in plasma and RBCs using LC-MS/MS (Figure 4A–C). For the Fing-Lipo and mAb-Fing-Lipo conditions, we calculated the concentration in the three compartments in blood (encapsulated in liposomes, bound to plasma proteins, and bound to RBCs) using raw data for RBC and plasma concentration and the numerical approach described previously (Figure 4D–E). To quantify the effect of liposome formulation on partitioning, we again calculated the RBC:Plasma ratio (Figure 4F).

Figure 4. Liposome formulation changes blood distribution of fingolimod in vivo.

(A, B, and C) Fingolimod (free or liposomal) was administered to mice via intravenous injection. Blood samples were taken at varying time points after administration, and drug concentration in plasma and RBCs was measured via LC-MS/MS (n=2-4). (D and E) Using the numerical approach previously validated in Figure 3 known partitioning of free drug and the known RBC adsorption of liposomes, the concentration of drug in the three compartments in blood (encapsulated in liposomes, bound to plasma proteins, and bound to RBCs) was calculated. (F) RBC:Plasma ratio calculated from (A-C). At 30 minutes after injection, the ratio is 5-fold higher for free drug compared to mAb-Fing-Lipo, indicating that the liposome formulation dramatically delays partitioning of fingolimod to RBCs in vivo. At 1 hour after administration, the RBC:plasma ratio was not statistically different for free drug and mAb-Fing-Lipo. Comparisons made by 2-way ANOVA with Sidak’s multiple comparison test.

For fingolimod administered as a free drug, the RBC:plasma ratio of fingolimod was roughly 6, which is in agreement with previous in vitro studies (Figures 4F and 3B). For mAb-Fing-Lipo, the liposomal formulation retained fingolimod and significantly decreased the RBC:plasma ratio during the first 30 minutes after administration. However, at 1 hour after administration, the RBC:plasma ratio was equal for the two conditions (Figures 4D). In total, the data indicates that fingolimod remains encapsulated in liposomes for at least 30 minutes of circulation in vivo. For Fing-Lipo, the liposome formulation retained fingolimod for 8 hours after administration, as judged by the decreased RBC:plasma ratio, compared to Fing (free drug).

To further understand the effect of liposome formulation on fingolimod pharmacokinetics, we estimated PK parameters using non-compartmental analysis (NCA) (Table 2). The PK parameters C_max, Area Under Curve (AUC) and clearance (CL) were determined for each compartment individually (plasma, RBC, and liposome). For Fing-Lipo and mAb-Fing-Lipo, both experimental data (only plasma and RBC) and calculated concentration data (which include plasma, RBC, and liposome) were analyzed. Volume of distribution was only calculated for all compartments (i.e., whole blood), not for each individual compartment. Fingolimod concentration in whole blood was calculated using a weighted average of concentration in plasma and RBC, assuming a hematocrit of 45% [41]. The calculations show that liposome formulation significantly changes the PK parameters of fingolimod. Specifically, Fing-Lipo led to a 20-fold increase in C_max in plasma, a 10-fold increase in AUC of plasma, and an 85% decrease in volume of distribution compared to free drug. These changes are expected for PEGylated liposomes exhibiting extended circulation in plasma [42]. mAb-Fing-Lipo results in similar changes in PK parameters, but to a lesser magnitude, likely because of the antibody coating leading to distribution to target tissue and more rapid clearance from blood [43].

Table 2. Pharmacokinetic parameters for fingolimod determined by non-compartmental analysis.

Pharmacokinetic (PK) parameters for fingolimod (free and liposomal) were determined by non-compartmental analysis (NCA) of data from in vivo studies (Figures 4A–E). Parameters were determined for each compartment individually (plasma, RBC, and liposome). For Fing-Lipo and mAb-Fing-Lipo, both experimental data (only plasma and RBC) and calculated concentration data (plasma, RBC, and liposome) were analyzed. Volume of distribution was not determined for individual compartments. Rather, one value was determined for each formulation. The liposomal formulation (Fing-Lipo) led to a significant increase in Cmax of plasma, increase in AUC for both plasma and RBC (with a larger increase for plasma compared to RBC), decrease in clearance, and decrease in volume of distribution. All of these changes are consistent with PEGylated liposomes leading to extended circulation in plasma. The antibody-targeted formulation (mAb-Fing-Lipo) had a similar effect on PK parameters as Fing-Lipo, but to a lesser magnitude. This is expected given that the antibody coating leads to distribution to target tissue and increased clearance from blood.

Treatment	Compartment	Cmax [ng/mL]	AUC0-t [ng/mL * hr]	CL [mL/hr]	Vss [mL]
Fing	Plasma	56.7 ± 27.9	53.5 ± 6.3	140.1 ± 16.6	N/A
	RBC	311.2 ± 153.4	245.6 ± 33.7	30.5 ± 4.2	N/A
Fing-Lipo	Plasma (Experimental)	994.7 ± 148.4	666.9 ± 40.9	62.5 ± 3.5	N/A
	RBC (Experimental)	562.8 ± 79.9	444.1 ± 19.8	27.3 ± 1.2	N/A
	Plasma (Calculated)	80.7 ± 11.4	63.8 ± 2.9	11.2 ± 0.7	N/A
	RBC (Calculated)	555.4 ± 79.0	439.2 ± 19.6	16.9 ± 0.7	N/A
	Liposome (Calculated)	921.4 ± 140.0	607.9 ± 39.2	187.3 ± 8.5	N/A
mAb-Fing-Lipo	Plasma (Experimental)	235.3 ± 21.9	120.0 ± 6.7	27.3 ± 1.2	N/A
	RBC (Experimental)	179.2 ± 39.7	275.0 ± 12.5	92.9 ± 6.8	N/A
	Plasma (Calculated)	25.9 ± 5.8	40.0 ± 1.8	117.5 ± 5.2	N/A
	RBC (Calculated)	177.1 ± 39.6	274.2 ± 12.5	17.1 ± 0.8	N/A
	Liposome (Calculated)	211.5 ± 16.9	80.7 ± 5.9	12.3 ± 0.8	N/A
Fing	All	171.2 ± 84.4	139.4 ± 18.3	53.8 ± 7.1	156.4 ± 28.9
Fing-Lipo	All	800.4 ± 113.6	566.6 ± 29.5	13.2 ± 0.7	21.3 ± 1.6
mAb-Fing-Lipo	All	210.0 ± 29.3	189.7 ± 8.4	39.5 ± 1.8	100.3 ± 6.3

Liposome formulation reduces toxic effects of fingolimod on red blood cells

Due to the strong partitioning to RBCs, fingolimod has the potential to cause RBC-related adverse effects. In fact, multiple cases of fingolimod-associated hemolytic anemia in the clinic have been reported [43]. Therefore, we investigated the effect of liposome formulation on fingolimod-associated RBC toxicities.

First, we assessed the potential of fingolimod to cause RBC hemolysis. We incubated RBCs with fingolimod (free or liposomal) for 5 seconds, immediately centrifuged to pellet the RBCs, then measured absorbance of the resulting supernatant as a metric for hemolysis. We observed that fingolimod as a free drug led to significant hemolysis as evident by the redness of the supernatant (Figure 5A, right images). This was confirmed by absorbance measurements of the supernatant, which showed that fingolimod as a free drug led to 23.3 ± 3.7% hemolysis of RBCs (using RBCs diluted in DI water as a positive control for 100% hemolysis) (Figure 5A, bar graph). Fingolimod caused significantly more hemolysis than vehicle control (2.2 ± 0.2%) but hemolysis was completely prevented by loading fingolimod into liposomes (3.3 ± 0.3%). As we observed in Figure 3, fingolimod partitions from liposomes to RBCs over the course of 30 minutes. To assess if fingolimod partitioning from liposomes leads to hemolysis at later time points, we measured hemolysis after 30 minutes and 60 minutes incubation of fingolimod with RBCs (Supplemental Figure 6). Again, fingolimod free drug led to RBC hemolysis (46.6 ± 3.8%), but hemolysis was completely prevented by liposomal formulation, even after 60 minutes incubation (3.2 ± 0.4% for Fing-Lipo vs. 3.9± 0.7% for vehicle control). For free drug, there was a trend of increased hemolysis over time, which may be caused by confounding factors other than fingolimod, such as released hemoglobin destabilizing the RBC membrane [44].

Figure 5. Liposome formulation reduces toxic effects of fingolimod on red blood cells.

(A) Fingolimod-induced hemolysis was measured by incubating RBCs with fingolimod (free or liposomal) at a concentration of 12.5 ug/mL and measuring absorbance of the supernatant. Free fingolimod induced significant hemolysis compared to control, as seen by both increased absorbance (left bar graph) and visually red supernatant (right image). Liposomal fingolimod caused no hemolysis compared to vehicle control. Comparisons were made by 1-way ANOVA with Dunnett’s post hoc test using Vehicle as the control (n=5-7). (B) Ektacytometry was performed to measure the relative stiffness of RBCs. Fingolimod treatment stiffens RBCs (black vs red bars), but this effect is largely prevented by loading fingolimod into liposomes. Comparisons were made by 1-way ANOVA with Dunnett’s post hoc test using Vehicle as the control (n=3). (C) A thin smear and Wright stain was performed on RBCs and images were taken using light microscopy. For fingolimod free drug, there is an increased number of oval-shaped RBCs (arrow) and an increased number of RBCs with no zone of central pallor (triangle).

Next, we measured the effect of fingolimod on RBC deformability using ektacytometry. RBCs that are rigid can predispose to clotting and inflammation [45], both of which are very deleterious in stroke [46], which is our target disease for developing fingolimod drug carriers. Fingolimod (free drug) induced a significant degree of rigidification, as measured by the shear stress at half-maximum elongation (Figure 5B, Supplemental Figure 5). However, the liposome formulation abrogated the rigidifying effect of fingolimod on RBCs.

Finally, we assessed the effects of fingolimod on RBC morphology. We incubated fingolimod (free or liposomal) with RBCs, performed a thin smear and Wright stain, and imaged the RBCs with light microscopy. The RBCs treated with fingolimod contained more oval-shaped RBCs (ovalocytes) and a decreased zone of central pallor (Figure 5C, Supplemental Figure 7). Thus, these studies show that liposomal formulation can limit the deleterious effects of fingolimod on RBCs.

Fingolimod-loaded VCAM-targeted liposomes enhance drug delivery and reduce brain edema in neuroinflammation

Having shown that our liposomal formulation slows partitioning of fingolimod to RBCs in vivo, we next sought to apply this formulation to treatment of an acute illness. As a test system, we chose a previously validated mouse model of neurovascular inflammation induced by intra-striatal injection of the cytokine tumor necrosis factor alpha (TNF-α) [31,32,47]. TNF-α is used because it mediates endothelial activation and increased vascular permeability in ischemic stroke [48,49]. Additionally, endothelial activation leads to overexpression of cell adhesion molecules, such as vascular cell adhesion molecule-1 (VCAM) [50]. By coating liposomes with monoclonal antibodies that bind to VCAM, we can deliver nanoparticles to inflamed regions of the brain [31], allowing us to test if targeted delivery can improve the efficacy of fingolimod liposomes.

First, we assessed the brain delivery of both untargeted and VCAM-targeted liposomes (Fing-Lipo and aVCAM-Fing-Lipo). We injected radio-labeled fingolimod liposomes into either naive mice, or injured mice at sixteen hours post-TNF-α injury (Figure 6A). For Fing-Lipo, the liposome concentration in the brain was only ~0.13 %ID/g, with no difference between naive and TNF-α injured mice. In TNF-α-injured mice, the aVCAM-Fing-Lipo treatment produced a liposome concentration of 3.71 ± 0.14% injected dose per gram of tissue (%ID/g) in the ipsilateral (injured) hemisphere of the brain. This is a 4-fold increase compared to naive mice (Figure 6B), confirming that VCAM-targeted fingolimod liposomes exhibit increased delivery to the brain in acute inflammation. Whole-body biodistribution data is given in Supplemental Figure 9. Given the rapid release rate of fingolimod from liposomes observed in Figure 4C, we also tested the time course of aVCAM-Fing-Lipo distribution to the brain. We observed that aVCAM-Fing-Lipo reaches maximum concentration in the brain within 5 minutes after administration (Supplemental Figure 10). Additionally, we performed LC-MS/MS to measure fingolimod concentration in tissues (Supplemental Figure 11). Fingolimod administered as a free drug exhibits a continuous increase in concentration in the brain over 8 hours. This same trend was previously seen in rats, and was explained by permeability-limited distribution of fingolimod to the brain [48]. Fingolimod delivery via aVCAM-Fing-Lipo resulted in greater concentration in the brain compared to free drug for all timepoints. With the liposome formulation, the trend of increased drug concentration over time was also observed. . Together, the data confirms that the VCAM-targeted liposome formulation can deliver fingolimod to the inflamed neurovasculature.

Figure 6. Fingolimod-loaded VCAM-targeted liposomes enhance drug delivery and reduce brain edema.

(A) Experimental timeline. TNF-α injection into the striatum induces brain inflammation and edema. Treatments are administered 16 hours after injury. Edema is quantified by albumin leakage into the brain. Radiolabeled albumin is injected and 4 hours later brains are harvested and radioactivity is measured, as a metric of albumin accumulation. (B) Brain uptake of radio-labeled liposomes in naive and TNF-α injured mice. Organs are harvested 30 minutes after treatment and measured for radioactivity. VCAM-targeted liposomes exhibit significantly higher brain delivery in TNF-α mice compared to naive. In addition, VCAM-targeted liposomes show increased delivery to the ipsilateral (injured) hemisphere, compared to the contralateral (uninjured) hemisphere. Untargeted fingolimod liposomes show minimal brain uptake (n=3-4). (C) Characterization of brain edema in TNF-α injury. Albumin accumulates in the ipsilateral hemisphere of TNF-α injured mice significantly more than in naive mice and significantly more than the contralateral hemisphere of TNF-α injured mice. (n=10-16). Comparisons were made by 2-way ANOVA with Sidak’s multiple comparison test. (D) Measurement of albumin accumulation, normalized such that 100% is the value of a mouse that had a sham injection of TNF-α, and 0% is the value of a TNF-α injured animal given no treatment. Note the 100% therapeutic effect of VCAM-targeted fingolimod liposomes. As controls for VCAM-targeted liposomes, fingolimod free drug or non-targeted liposomes, all at a dose of 0.1 mg/kg of fingolimod were tested, and neither had a statistically significant therapeutic effect compared to PBS treated mice. (n=4-7). Comparisons were made by 1-way ANOVA with Dunnett’s post hoc test, using PBS as the control.

We next assessed the ability of fingolimod liposomes to reduce neuroinflammation-induced brain edema. Edema was measured via intravenous injection of radiolabeled albumin, which accumulates in exudate at sites of inflammation. We first characterized the TNF-α injury (without liposome treatment) and observed significant albumin accumulation in the injured hemisphere (Figure 6C), which confirms results from our previous publication [49]. Visualization of the edema via Evans Blue Dye confirmed that edema is primarily located in the injured hemisphere (Supplemental Figure 12). Finally, we tested the ability of various treatments (Fing, Fing-Lipo, and aVCAM-Fing-Lipo) to reduce albumin leakage into the ipsilateral hemisphere. A trend in reduced edema was observed for fingolimod free drug (38.5 ± 16.5% reduction in albumin extravasation, p=0.43 compared to PBS-treated) and Fing-Lipo (57.3 ± 36.4% reduction, p=0.15 compared to PBS-treated). However, aVCAM-Fing-Lipo produced much greater therapeutic efficacy, resulting in complete protection from brain edema (114.0 ± 11.5% reduction, p=0.003 compared to PBS-treated) (Figure 6D). We additionally tested a lower dose of fingolimod (0.025 mg/kg) in the TNF-α model, and found that neither free drug nor VCAM-targeted liposomes had a therapeutic effect at this lower dose (Supplemental Figure 13).

To further assess the safety of VCAM-targeted liposomes, we investigated potential toxicities to the liver and spleen, which are major clearance organs for nanoparticles. As a metric of splenic toxicity, we chose to assess lymphocyte count because fingolimod is known to prevent lymphocyte egress from the spleen, leading to a decrease in circulating lymphocytes. We observed a decrease in circulating lymphocytes for all fingolimod treatments, but there was no difference between fingolimod free drug or liposome formulations (Supplemental Figure 14), suggesting the liposome formulation does not induce splenic toxicity. As a metric of liver toxicity, we measured the concentration of alanine aminotransferase (ALT) in plasma, and observed no increase due to fingolimod treatment (Supplemental Figure 14). In total, the additional data shows that fingolimod delivery via VCAM targeting does not induce significant splenic or liver toxicities.

Discussion

Fingolimod is an S1PR modulator that has shown promise for the treatment of acute ischemic stroke. However, its application to stroke is limited by hydrophobicity and micelle formation, which may prevent IV administration, and strong partitioning to RBCs and plasma proteins, which slows distribution to the site of action. Here we developed a stable liposome formulation for fingolimod and extensively evaluated how the formulation affects the partitioning of fingolimod between compartments in whole blood (plasma proteins, RBCs, and liposomes).

We optimized the liposome formulation and found the addition of anionic lipids increased drug loading and reduced leak (Figure 1). When incubated in plasma alone, liposomes completely prevent partitioning of fingolimod to plasma proteins (Figure 2). However, when incubated in whole blood, a significant fraction of fingolimod partitioned out of liposomes and into RBCs. The percentage of fingolimod partitioned to RBCs was 56.0 +/− 1.5% for Fing-Lipo and 68.9 +/− 1.5% for mAb-Fing-Lipo (Figure 4), In vivo studies in mice showed a similar trend in fingolimod partitioning (Figure 5). Additional studies showed that liposomes did not associate with RBCs in whole blood (Figure 3D). This agrees with previous publications, which shows the presence of serum protein prevents adsorption of liposomes to RBCs [51]. Together results indicate that RBCs are the major cause of drug release from fingolimod liposomes after IV administration.

Partitioning to RBCs is likely favored as RBCs constitute ~50% of the blood volume and therefore the largest surface area for drug interaction of any organ in the body. Since fingolimod is loaded into the lipid membrane, it is directly exposed to the massive surface area of RBC membranes in the blood, providing a large sink for the drug to partition into. We suspect this problem would be applicable to other lipid nanoparticles in which a drug is loaded into the hydrophobic phase. There is significant interest in loading drugs into RNA-lipid-nanoparticles (LNPs) to expand the functionality of LNPs. Our data suggests that drug partitioning from LNPs to RBCs should be an important consideration during development of these formulations.

For in vivo studies, the liposome formulation reduced partitioning of fingolimod to RBCs for the first 30 minutes, However, at 1 hour after liposome administration, the RBC:plasma ratio for Fing and mAb-Fing-Lipo were not statistically different. This suggests that fingolimod was completely released from antibody-coated liposomes after 1 hour in circulation. The rapid release of fingolimod may actually be beneficial for our current application, as it will result in rapid availability of fingolimod to enact a pharmacologic effect. Importantly, nanoparticles targeted to endothelial markers, such as VCAM, can accumulate in the target organ less than 30 minutes after administration, as the target is directly accessible from the bloodstream (Supplemental Figure 10) [31,52,53]. However, the rapid partitioning of fingolimod to RBCs may be a challenge for other applications that require extended circulation, such as tumor delivery via the Enhanced Permeability and Retention (EPR) effect [36,54].

Finally, we tested the organ distribution and efficacy of fingolimod-loaded liposomes in a mouse model of neuroinflammation (Figure 6). We observed that VCAM-targeted liposomes accumulated in the inflamed brain and the spleen. Analysis of drug concentration in tissue via LC-MS/MS (Supplemental Figure 11) confirmed that VCAM-targeted liposomes deliver fingolimod to target tissue. Fingolimod delivery via aVCAM-Fing-Lipo resulted in greater concentration in the brain compared to free drug for all timepoints. Additionally, aVCAM-Fing-Lipo exhibited a continuous increase in drug concentration in the brain over time, which was also seen for fingolimod free drug and in previous studies in rats, and was explained by permeability-limited distribution of fingolimod to the brain [53]. We speculate that for aVCAM-Fing-Lipo, two modes of fingolimod distribution to the brain are occurring. First, there is rapid delivery of fingolimod to endothelium in the brain due to targeting of aVCAM liposomes (occurring in the first 5 minutes after injection), leading to the overall higher concentration in the brain for aVCAM liposomes compared to free drug. Second, fingolimod that is released from liposomes slowly permeates into brain tissue and accumulates over time, leading to increasing concentration over time.

For therapeutic studies, untargeted liposomes did not alter efficacy compared to free fingolimod, but liposomes targeted to VCAM completely resolved brain edema (Figure 6). Although the efficacy in reducing brain edema provides proof-of-concept, there are multiple paths that could be pursued to validate the therapeutic effect. First, a more comprehensive dose-response would better characterize effect and determine the optimal dosage of fingolimod. Second, testing fingolimod liposomes in a more translationally relevant animal model, such as middle cerebral arterial occlusion (MCAO) would increase confidence in the translatability of the therapeutic efficacy [55]. Finally, a more comprehensive pharmacokinetic/pharmacodynamic (PK/PD) understanding could be established by measuring the PD output (albumin leakage) at multiple timepoints after fingolimod administration, therefore validating the impact of rapid delivery. However, a clear PK/PD relationship may be difficult to establish given multiple mechanisms of action of fingolimod, including acting on endothelial cells to reduce vascular permeability and acting on lymphocytes to reduce egress from lymph nodes [54].

Although we initially chose VCAM targeting because it has been shown to target inflamed endothelium in the brain, the significant spleen delivery (Supplemental Figure 11) suggests an additional possible mechanism for improved therapeutic efficacy of fingolimod liposomes. The canonical mechanism of action for fingolimod in multiple sclerosis is to reduce lymphocyte egress from lymph nodes and therefore prevent their infiltration into the central nervous system [56]. By delivering fingolimod to the spleen, VCAM-targeted liposomes may prevent lymphocyte infiltration to the brain, which is known to contribute to edema during ischemic stroke [57]. Although it is an important question, dissecting the relative contributions of brain and spleen delivery is outside the scope of the current work and is an area for future investigation.

There have been other studies on the delivery of fingolimod using nanoparticles [58–62]. However, there are several aspects that differentiate our current work from previous research. First, we show that the addition of anionic lipids can significantly reduce fingolimod release from liposomes. Previous studies have not investigated the impact of nanoparticle formulation on nanoparticle properties such as drug release. Second, we are the first to compare fingolimod release from nanoparticles in buffer, plasma, and whole blood. We observed that drug release from our liposome formulation was insignificant in buffer and plasma, but drug release was appreciable in whole blood. The RBC:plasma ratio in the whole blood studies closely matched in vivo studies, highlighting the usefulness of testing drug release in whole blood. Third, we show that fingolimod itself has significant toxicity on red blood cells, but liposomal formulation can prevent these toxicities. Finally, we show in vivo efficacy of fingolimod liposomes in a mouse model of acute neuro inflammation, while the majority of previous studies focus on oncology applications.

While the drug partitioning approach we demonstrated shows promise, it is prudent to consider limitations. First, the approach will not be applicable to all drugs. We expect only drugs that bind very strongly to plasma proteins or RBCs to benefit from changing partitioning within the blood. Somewhat surprisingly, 24% of drugs approved by the FDA from 2003-2013 had plasma protein binding of >99%, which is considered very high protein binding [63]. Therefore, a significant minority of drugs may still benefit from the approach described in this manuscript. Second, a different nanoparticle formulation needs to be developed for each drug to ensure stable encapsulation and minimal release to plasma proteins and RBCs. Formulation development can be time and resource-intensive, especially for novel formulations such as nanoparticles. Third, for improved efficacy in acute illnesses, a suitable targeting moiety is needed to target the nanocarrier to the diseased cells. To date, numerous targeting determinants have been studied that can target nanoparticles to inflamed endothelial cells in the brain, lung, and tumors [52,64]. However, it is likely that the best target determinant needs to be empirically studied for each drug and each disease application.

There are also general challenges to the successful application of targeted nanoparticles [36,65]. First, nanomedicines are extremely complex compared to traditional pharmaceuticals. The complexity makes all areas of product development more difficult, including formulation, manufacturing, characterization, and understanding mechanisms of action. Another major challenge is biocompatibility and safety, particularly activation of the innate and adaptive immune system after intravenous administration [66,67]. This is an active area of research, and multiple groups, including ours, are developing technologies to prevent nanomedicine-induced immune-activation and related side effects [68,69]. Despite these limitations, the present studies show a useful paradigm for using nanoparticles to control drug partitioning within blood and tune PK for treatment of acute critical illnesses.

Conclusions

A large fraction of drugs partition within blood to the RBC- or protein-bound compartments. This distribution is disadvantageous for acute illnesses, like stroke, which need rapid distribution of drug to target tissues and cannot tolerate the side effects caused by drug-mediated disruption of RBCs. Here we show that nanocarriers can resolve these problems, using as our example the FDA-approved drug fingolimod, which holds promise for treating stroke, but is limited by its strong partitioning to RBCs and plasma proteins. Liposomal formulation of fingolimod allowed for rapid, IV delivery of fingolimod, without acute partitioning to RBCs and plasma proteins or associated side effects, such as hemolysis. Further, by targeting the liposomes using antibodies directed at the target tissue, this approach led to rapid delivery and was able to completely resolve a leaky blood-brain barrier. Thus, targeted fingolimod-liposomes appear to be a promising therapeutic for stroke. Additionally, this approach of using nanocarriers to partition drugs away from the RBC- and protein subcompartments of blood may be a generalizable approach to the rapid drug delivery needed for acute illnesses.