Exosome is a Mechanism of Intercellular Drug Transfer: Application of Quantitative Pharmacology

Abstract

Purpose

Exosomes are small membrane vesicles (30–100 nm in diameter) secreted by cells into extracellular space. The present study evaluated the effect of chemotherapeutic agents on exosome production and/or release, and quantified the contribution of exosomes to intercellular drug transfer and pharmacodynamics.

Methods

Human cancer cells (breast MCF7, breast-to-lung metastatic LM2, ovarian A2780 and OVCAR4) were treated with paclitaxel (PTX, 2–1000 nM) or doxorubicin (DOX, 20–1000 nM) for 24–48 h. Exosomes were isolated from the culture medium of drug-treated donor cells (Donor cells) using ultra-centrifugation, and analyzed for acetylcholinesterase activity, total proteins, drug concentrations, and biological effects (cytotoxicity and anti-migration) on drug-naïve recipient cells (Recipient cells). These results were used to develop computational predictive quantitative pharmacology models.

Results

Cells in exponential growth phase released ~220 exosomes/cell in culture medium. PTX and DOX significantly promoted exosome production and/or release in a dose- and time-dependent manner, with greater effects in ovarian cancer cells than in breast cancer cells. Exosomes isolated from Donor cells contained appreciable drug levels (2–7 pmole/10⁶ cells after 24 h treatment with 100–1000 nM PTX), and caused cytotoxicity and inhibited migration of Recipient cells. Quantitative pharmacology models that integrated cellular PTX pharmacokinetics with PTX pharmacodynamics successfully predicted effects of exosomes on intercellular drug transfer, cytotoxicity of PTX on Donor cells and cytotoxicity of PTX-containing exosomes on Recipient cells. Additional model simulations indicate that within clinically achievable PTX concentrations, the contribution of exosomes to active drug efflux increased with drug concentration and exceeded the p-glycoprotein efflux when the latter was saturated.

Conclusions

Our results indicate (a) chemotherapeutic agents stimulate exosome production or release, and (b) exosome is a mechanism of intercellular drug transfer that contributes to pharmacodynamics of neighboring cells.

Graphical abstract

Introduction

Inadequate drug delivery is a major cause of treatment failures in solid tumors [1]. After entering the systemic circulation, e.g., via an intravenous injection, the drug encounters multiple transport barriers before reaching and exerting its action on the intended targets. Recent intraoperative intravital microscopy findings in patients further show that about one-half of vessels in human tumors are not patent or functional [2, 3]. These issues highlight the need to better understand the mechanisms of interstitial drug transfer. The present study examined the potential role of exosomes.

Cells utilize exocytosis to sort intracellular substances into exosomes that are subsequently released to the extracellular space [4]. Exosomes are small membrane vesicles with an average diameter of between 30 and 100 nm. They originate from the inward budding of endosomal lumen layer and carry cellular components including lipids, proteins (e.g., heat shock proteins, transcription factors, enzymes, major histocompatibility receptors and tetraspanins), and nuclei acids (e.g. DNA, mRNA, microRNA and long non-coding RNA) [4–7].

The life-cycle of exosomes comprises endosome biogenesis, trafficking, release, and re-uptake via endocytosis [4, 7, 8]. Biogenesis begins with internalization of plasma membrane as early endosomes, which later become multivesicular bodies and form intraluminal vesicles (pre-exosomes) that mature into exosomes. Contents of exosomes are sorted and loaded through ESCRT-dependent and -independent mechanisms. In the latter, a sphingolipid ceramide is involved in the loading of microRNA and lipid rafts into endosomes, and the initiation of exosome biogenesis [9, 10]. Several Rab proteins, including Rab-27a/b, Rab-11 and Rab-35, are known molecular motors that drive multivesicular bodies towards plasma membrane [11–13]. Release of exosomes into extracellular space is mediated by exocytosis, which involves fusion of exosome membrane with plasma membrane using SNARE (soluble N-ethylmaleimide sensitive fusion protein attachment receptors)-dependent and -independent mechanisms. Re-uptake of exosomes into cells primarily uses receptor-mediated endocytosis, with plasma membrane fusion and phagocytosis as minor pathways [6, 14].

Cancer cells generally produce higher levels of exosomes compared to normal cells [15]. Exosomes derived from cancer cells are involved in distal metastatic niche initiation [16, 17], intercellular communications (e.g., during drug resistance development [18, 19]), and immune system modulation [20, 21]. Cancer cells enhance their exosome secretion in response to environmental changes including pH [22], ion [23], temperature [24], and treatment by cytotoxic agents [25]. For example, liver HepG2 cells, when treated with cytotoxics (PTX, etoposide, irinotecan, carboplatin), release exosomes containing elevated level of heat shock proteins [25] triggered as a response to stress and as a survival mechanism [26].

Most exosome studies have focused on characterizing their contents and biological functions [13, 15–18, 27, 28]. The current study used in vitro experiments and in silico studies to investigate the intercellular drug transfer via exosomes and the quantitative relationship between this process and pharmacodynamics (PD) in solid tumors. Paclitaxel (PTX) and doxorubicin (DOX) were the test drugs as they are commonly used in first-line therapy of multiple types of major solid tumors including, e.g., ovarian, breast, lung, and prostate cancers [29]. Our results indicate exosomes is a mechanism of intercellular drug transfer with significant pharmacological consequences.

Materials and methods

Reagents

PTX and DOX (purity >99.5%), and cell culture grade dimethyl sulfoxide (DMSO) were purchased from Sigma-Aldrich (St. Louis, MO). Drug stock solutions were prepared in DMSO and diluted to desired concentrations such that the DMSO concentration was below 0.5%.

Cell culture

Human breast adenocarcinoma MCF7 cells were purchased from ATCC (Manassas, VA). LM2 cells, a highly breast-to-lung metastatic subline of human breast cancer MDA-MB-231 cells [30], were a gift from Dr. Y. Kang (Princeton University, NJ). Human ovarian cancer A2780 and OVCAR4 cells were provided by Dr. D. Dhanasekaran (University of Oklahoma Health Sciences Center, OK). MCF7 and LM2 cells were maintained in DMEM (Mediatech, VA), and A2780 and OVCAR4 cells in RPMI-1640 (ATCC). The medium for cell growth (Growth Medium) was supplemented with 10% fetal bovine serum (FBS, Atlanta Biologicals, GA), whereas the medium for exosome isolation (Conditioned Medium) was supplemented with 10% exosome-depleted-FBS (System Biosciences, CA). All medium contained 100 IU/mL penicillin and 100 μg/mL streptomycin (Mediatech, VA). Cells were cultured in a humidified incubator at 37°C with 5% CO₂.

Effects of PTX and DOX on exosomes

Effects of PTX and DOX on exosome production and release were studied using cytotoxic drug concentrations that produced 20, 50 and 80% of their maximal cytotoxicity (EC₂₀, EC₅₀ and EC₈₀; see below for their determination). Exosomes collected from cells treated with drugs (Donor cells) are denoted as Drug-Exo_conc,time, e.g., PTX-Exo_500nM,24h denotes exosomes from Donor cells treated with 500 nM PTX for 24 h. PTX-Exo was used to study its pharmacological effects (Cytotoxicity_recipient and Anti-migration_recipient) on drug-naïve recipient cells (Recipient cells).

Exosome isolation, characterization, and quantification

Exosomes were isolated using a previously reported serial centrifugation protocol [31], with the following modifications. In brief, cells were cultured in T175 flasks, without or with drug. The post-incubation exosome-containing medium (Conditioned Medium) was collected at preselected time points, centrifuged twice, first at 2,000 × g for 10 min and then 10,000 × g for 30 min, to remove dead cells and debris. The supernatant was transferred and ultracentrifuged at 100,000 × g for 18 h at 4°C using a type 50.2 Ti rotor (Optima™ L-100 XP, Beckman Coulter Instruments). The pellet was washed once with 4 mL phosphate-buffered saline (PBS, pH 7.4), then ultra-centrifuged at 100,000 × g for 70 min at 4°C. For studying biological activity of exosomes, the resulting pellet was re-suspended in Growth Medium supplemented with 10% exosome-depleted FBS after PBS wash. For the remaining studies, the pellet was re-suspended in 150 μL PBS.

Exosome morphology was characterized using transmission electron microscopy. In brief, an exosome sample was diluted 1:100 in PBS, loaded on a 400 mesh, Formvar-coated, glow discharged copper grid using the single drop method. The liquid was removed 60–90 s later by wicking with filter paper and the exosome-loaded grid washed with deionized water for 10 s. After removing the water again with filter paper, the sample was stained with 4% uranyl acetate (pH 4.4, negative staining solution) for 60 s, washed in deionized water and air dried at room temperature. The grid was then viewed on a transmission electron microscope (Hitachi H7600, equipped with a 2k × 2k AMT digital camera) at 80 kV. A second aliquot of exosome suspension was used to determine exosome size using nanoparticle tracking analysis (NanoSight NS300; Malvern Instruments, Inc., UK). A third aliquot of exosome suspension was used to determine the total protein amount by BCA assay (Pierce kit, ThermoFisher Scientific Inc., CA). Another aliquot was used to determine the activity of acetylcholinesterase (AChE), an enzyme found in exosomes [12, 23], by the Ellman colorimetric assay (Sigma-Aldrich) [32]. Briefly, exosome suspension (20 μL) was added to an aqueous reaction system (80 μL) containing 1 mM 5,5′-dithio-bis(2-nitrobenzoic acid) and 3 mM acetylthiocholine chloride (substrate for AChE), per well in a 96-well plate. After 40 min incubation at room temperature, the absorbance at 412 nm was determined using Synergy HT microplate reader (BioTek Instruments, VT). Standard curves of AChE activity were established using plasma exosome standards calibrated by NanoSight analysis (System BioSciences) and used to calculate the exosome quantity.

The levels of exosome protein markers TSG101 and CD63 were analyzed by Western blotting. Samples of exosomes or Donor cell lysates containing 60 μg proteins in RIPA lysis buffer (ThermoFisher Scientific) were mixed with a protease and phosphatase inhibitor cocktail (ThermoFisher Scientific), and the mixture was loaded on 10% Mini-PROTEAN^® TGX pre-cast gel (BioRad Laboratories, CA). After electrophoresis, proteins were transferred to a 0.2 μm nitrocellulose membrane (BioRad Laboratories) and probed sequentially with primary antibodies (mouse IgG anti-CD63 and anti-TSG101 from Santa Cruz Biotechnology (Dallas, TX), rabbit IgG anti-beta actin and anti-calnexin from Cell Signaling Technology (Danvers, MA)), and then with fluorescent secondary antibodies (IRDye^® 800CW goat anti-mouse and goat anti-rabbit, LI-COR Inc., NE). Fluorescence intensity was quantified using Odyssey CLx system (LI-COR Inc.).

Biological activities of PTX, DOX and PTX-Exo

We measured the cytotoxicity of PTX and DOX on donor cells (Cytotoxicity_donor) and the cytotoxicity of PTX-Exo on Recipient cells (Cytotoxicity_recipient) using the sulforhodamine B (SRB) colorimetric assay [33]. In brief, cells (5000 per 100 μL Growth Medium per well in 96-well plate) were seeded overnight, treated with drugs or PTX-Exo for 48 h, fixed with trichloroacetic acid (10% w/v, 100 μL per well, 30 min at 37°C) after removing the medium, washed gently with tap water three times, air-dried overnight, and stained with SRB (0.1% in 1% acetic acid (v/v), 100 μL per well, 30 min at room temperature) with gentle agitation. After removing the excess SRB by washing four times with 1% acetic acid, the cell-bound SRB was dissolved with Tris base solution (10 mM, 100 μL per well) and the absorbance measured at 510 nm using Synergy HT microplate reader. The concentration-effect relationships were analyzed with a sigmoidal Hill equation using nonlinear least square regression [34] (Prism 7, GraphPad Software, CA). For PTX, the Hill equation was modified with a residual unaffected fraction R_e as we previously described [34]. The analysis provided the drug concentration producing 20%, 50% and 80% of the maximal cytotoxicity (EC₂₀, EC₅₀ and EC₈₀, respectively).

Anti-migration_recipient was measured using the wound healing assay that measures cell migration [35]; the assay used serum-free medium to minimize cell proliferation. Results of pilot studies indicated only one of the four cell lines (LM2) remained viable in serum-free medium after 48 h and PTX-Exo_{1000nM,24–48h} induced shrinkage and fragmentation in Recipient cells. Therefore, the study used LM2 cells as Recipient cells and PTX-Exo_{5–500nM,24–48h}. Briefly, cells were cultured in Growth Medium supplemented with 10% FBS on a tissue culture-coated (vacuum gas-plasma, hydrophilic and negative charge surface) 6-well plate (Corning Incorporated, NY) and allowed to grow to confluence, after which time the medium was replaced with serum-free Growth Medium. The growth surface was scratched with a 200 μL pipet tip to create an open wound area, and cell debris was carefully washed off with pre-warmed DMEM. Afterwards, a suspension of exosomes (containing 100 μg proteins in 2 mL serum-free medium) was added to each well, and microscopic images of a fixed field were obtained before and after scratching (0 h, 24 h and 48 h). The open wound area was measured as fraction of area not covered by cells using TScratch, a stand-alone Matlab application that measures the covered pixels vs. the uncovered pixels of an image [36] with default detection threshold settings.

Quantification of PTX in PTX-Exo and cell lysates

Cells were treated with PTX for 24 h. PTX-Exo was collected from Conditioned Medium as described above. The remaining cells were washed with PBS and trypsinized, collected, and stained with trypan blue. The number of trypan blue-excluding cells were counted and lyzed with RIPA lysis buffer. Drug concentrations in PTX-Exo and cell lysates were measured using high performance liquid chromatography-tandem mass spectrometry (LC-MS/MS). First, PTX in the sample was extracted with 10x volume of methyl t-butyl ether containing 10 ng/mL of the internal standard docetaxel. The organic phase containing paclitaxel was separated, evaporated, and reconstituted with 10 μL of 0.1% aqueous formic acid/methanol (40/60, v/v). A calibration curve was performed using seven calibration standards covering a dynamic range of 1 – 1,000 ng/mL, prepared in the same biological matrix and run in duplicates. The best-fit line was obtained using linear regression with 1/x² weighting. Quality control standards were prepared in quintuplet at each of three levels (low (3 ng/mL), mid (50 ng/mL), high (800 ng/mL)), and back-calculated against the best-fit line to ensure accuracy and precision of the assay.

For LC-MS/MS analysis, samples (10 μL) were injected onto a Symmetry Shield RP18 column (2.1×50 mm, 3.5 μm; Waters, MA), using a gradient elution scheme of 0.1% aqueous formic acid in methanol, where methanol was increased from 40% to 100% in 7 min before returning to 40% by the end of the 10 min run (flow rate 0.2 mL/min). The column eluent was directed into an electrospray ionization triple quadrupole mass spectrometer (Micromass Quattro Premier XE; Waters) for detection based on multiple reaction monitoring of a structural fragment in the positive ion mode. PTX was monitored by the mass transition from precursor ions to product ions of m/z 854.0➔287.0 (collision energy 16 V), and docetaxel by transition of m/z 806.8➔526.3 (collision energy 9 V). Universal mass spectrometer settings included capillary voltage of 3500 V, cone voltage of 25 V, source temperature 120°C, desolvation temperature 400°C, and desolvation gas flow (N2) of 600 L/hr.

Quantitative pharmacology modeling of exosome-mediated intercellular drug transfer and biological effects: Overview

The experimental results indicated exosomes collected from PTX-treated Donor cells contained high drug concentrations, served as a mechanism of intercellular transfer and conferred biological activities on drug-naïve Recipient cells. We established predictive quantitative pharmacology (QP) models to depict these processes, as follows. First, we modified our previously published PTX cellular pharmacokinetic (PK) models [37–39] to include the sorting and release of exosomes and their re-uptake via receptor-mediated endocytosis. These models were used together with previously published model parameters and experimental results from the current study to obtain values of the exosome-related model parameters. The resulting equations plus parameter values were used to simulate the drug concentrations derived from PTX-Exo, in cells and medium. Next, we established PD models to depict the relationships between total extracellular/intracellular drug concentrations (including the pharmacologically active tubulin-bound moiety) and drug-induced cytotoxicity. Model simulations and data fitting (nonlinear least-squares algorithm) were performed using Matlab Simbiology (Release 2016a, Mathworks, MA).

Development of cellular PK models

We previously established cellular PTX PK models that have since been adopted by multiple investigator groups [34, 38–43]. The current study extended these earlier models to account for exosome-related processes. The model assumptions are (a) unbound drug in medium enters cells by passive diffusion, (b) saturable drug binding to proteins in extracellular fluid and to intracellular tubulin, plus non-saturable drug binding to other intracellular organelles, (c) drug efflux from cells uses a combination of passive diffusion, p-glycoprotein (Pgp)-mediated saturable efflux, and release of drug-containing exosomes, (d) first order sorting of intracellular unbound drug into exosomes through the endosomal transport system, (e) first order release of exosomes from the cell into extracellular space, and (f) internalization of drug-containing exosomes through saturable receptor-mediated endocytosis [14, 44]. The exosome-related processes in (c) through (f) are new, whereas the remaining processes describe the cross-membrane drug transportation and intracellular drug distribution as established previously [37–39]. To simplify the model, we assumed negligible PTX transfer between C_medium,free and C_medium,exo.

The above processes are summarized in Figure 1. In general, subscripts are used to denote the various drug entities, their locations and transport mechanisms. For drug concentrations C, the first subscript indicates if the drug is located intracellularly or extracellularly (cell vs. medium) and the second subscript indicates if it is bound to proteins or exosomes, e.g., C_cell,tubulin is tubulin-bound drug concentration in cells, and C_medium,exo and C_cell,exo are the respective drug concentrations in medium and cells derived from exosomes. Note that because C_medium,exo and C_cell,exo were calculated as amount divided by volume, C_medium,exo does not equal C_cell,exo due to difference in medium and cell volumes.

Figure 1. Cellular pharmacokinetic/pharmacodynamic models of paclitaxel.

C_medium,free or C_cell,free is free (unbound) drug concentration in medium or cells, respectively. C_cell,tubulin is the tubulin-bound drug concentration. D_fd is the rate constant of passive diffusion of free drug. Jmax_Pgp is the maximum Pgp-mediated drug efflux rate. Kd_Pgp is the dissociation constant of drug from Pgp. C_medium,exo or C_cell,exo is the drug concentration in extracellular or intracellular exosomes, respectively. Jmax_inter,exo is the maximum rate of receptor-mediated internalization of exosomes. Kd_inter,exo is the dissociation constant of exosome from receptor for internalization. k_sort,exo is the rate constant of drug sorting into exosomes. k_rel,exo is the rate constant of exosome release into extracellular space. B_tubulin,max is the maximum available drug binding sites in tubulin. k_tubulin,on or k_tubulin,off are the rate constants of drug association and disassociation with tubulin, respectively. NSB is the proportion constant for the linear nonsaturable drug binding in cells. For PD parameters, IC₅₀ is the tubulin-bound drug concentration needed to generate 50% of maximum drug effect, k_kni is the maximal rate constant of cell kill. n is Hill exponent.

For transport, Jmax is maximum rate and Kd denotes dissociation constant, and their subscript denotes the transport mechanism, e.g., Jmax_Pgp is maximum Pgp-mediated efflux rate per cell and Kd_Pgp is dissociation constant of drug from Pgp, whereas Jmax_inter,exo is maximum rate of receptor-mediated internalization of exosomes per cell and Kd_inter,exo is dissociation constant of exosome bound to the receptor for internalization. k_sort,exo and k_rel,exo denote the rate constant for sorting and release of exosomes, respectively. For intracellular drug binding to its molecular target tubulin, B_tubulin,max is the maximal available binding sites, and k_tubulin,on and k_tubulin,off are the respective association or disassociation rate constant. NSB is the proportionality constant for the linear nonsaturable drug binding in cells. D_fd is the diffusion rate constant of the free drug (i.e., unbound to macromolecules) through cell membrane. V_cell and V_medium are volume of a single cell and extracellular medium, respectively. ICN and TCN are respectively the initial and total cell number, and were experimentally counted.

Eq. 1–2 depict C_cell,total and C_medium,total as sums of various intracellular and extracellular drug entities, respectively.

$$C_{\mathit{cell},\mathit{total}}=C_{\mathit{cell},\mathit{free}}+C_{\mathit{cell},\mathit{exo}}+C_{\mathit{cell},\mathit{tubulin}}+\mathit{NSB}\cdot C_{\mathit{cell},\mathit{free}}$$ (1)

$$C_{\mathit{medium},\mathit{total}}=C_{\mathit{medium},\mathit{free}}+C_{\mathit{medium},\mathit{bound}}+C_{\mathit{medium},\mathit{exo}}$$ (2)

Eq. 3–5 describe the time-dependent changes of C_cell,free, C_cell,exo, and C_cell,tubulin. Based on the reported value of 0.99 ± 0.04 ligand per alpha/beta tubulin dimer [45, 46], we used a 1:1 PTX-tubulin binding stoichiometry (i.e., binding of C_cell,free to tubulin increases linearly with B_tubulin,max). As reflected in these equations, there are three drug efflux mechanisms: passive diffusion (equals D_fd · C_cell,free), and active efflux via Pgp (equals to (Jmax_Pgp · C_cell,free)/(Kd_Pgp + C_cell,free)) and exosomes (equals k_rel,exo · C_cell,exo · V_cell).

$$\frac{d{C}_{\mathit{cell},\mathit{free}}}{\mathit{dt}}=\frac{\left(D_{\mathit{fd}}\cdot C_{\mathit{medium},\mathit{free}}-D_{\mathit{fd}}\cdot C_{\mathit{cell},\mathit{free}}\frac{\mathit{Jma}{x}_{\mathit{Pgp}}\cdot C_{\mathit{cell},\mathit{free}}}{K{d}_{\mathit{Pgp}}+C_{\mathit{cell},\mathit{free}}}+\frac{\mathit{Jma}{x}_{\mathit{inter},\mathit{exo}}\cdot C_{\mathit{medium},\mathit{exo}}}{K{d}_{\mathit{inter},\mathit{exo}}+C_{\mathit{medium},\mathit{exo}}}\right)}{V_{\mathit{cell}}}-k_{\mathit{sort},\mathit{exo}}\cdot C_{\mathit{cell},\mathit{free}}+k_{\mathit{tubulin},\mathit{off}}\cdot C_{\mathit{cell},\mathit{tubulin}}-k_{\mathit{tubulin},\mathit{on}}\cdot C_{\mathit{cell},\mathit{free}}\cdot (B_{\mathit{tubulin},\mathit{max}}-C_{\mathit{cell},\mathit{tubulin}})$$ (3)

$$\frac{d{C}_{\mathit{cell},\mathit{exo}}}{\mathit{dt}}=k_{\mathit{sort},\mathit{exo}}\cdot C_{\mathit{cell},\mathit{free}}-k_{\mathit{rel},\mathit{exo}}\cdot C_{\mathit{cell},\mathit{exo}}$$ (4)

$$\frac{d{C}_{\mathit{cell},\mathit{tubulin}}}{\mathit{dt}}=k_{\mathit{tubulin},\mathit{on}}\cdot C_{\mathit{cell},\mathit{free}}\cdot (B_{\mathit{tubulin},\mathit{max}}-C_{\mathit{cell},\mathit{tubulin}})-k_{\mathit{tubulin},\mathit{off}}\cdot C_{\mathit{cell},\mathit{tubulin}}$$ (5)

We reported B_tubulin,max increases linearly with time (from an initial value B_{tubulin,initial}) and the corresponding rate constant kB_tubulin,max changes with drug concentration [39]; these kinetic processes are captured in Eq. 6 and 7. Eq. 7 was obtained by nonlinear regression of the plot of kB_tubulin,max vs. C_cell,free (r² = 0.998), where the experimental results were obtained from our earlier study [39].

$$\frac{B_{\mathit{tubulin},max}}{\text{dt}}=B_{\mathit{tubulin},\mathit{initial}}\cdot k_{B\mathit{tubulin},\mathit{max}}$$ (6)

$$k_{B\mathit{tubulin},\mathit{max}}=408.48\cdot {C_{\mathit{cell},\mathit{free}}}^{0.2941}$$ (7)

Eq. 8–10 describe changes of C_medium,free, C_medium,exo, and C_medium,bound with time, where B_medium,max is maximum saturable drug binding sites in medium.

$$\frac{d{C}_{\mathit{medium},\mathit{free}}}{\mathit{dt}}=\left(-D_{\mathit{fd}}\cdot C_{\mathit{medium},\mathit{free}}+D_{\mathit{fd}}\cdot C_{\mathit{cell},\mathit{free}}+\frac{J_{\mathit{max},\mathit{Pgp}}\cdot C_{\mathit{cell},\mathit{free}}}{K{d}_{\mathit{Pgp}}+C_{\mathit{cell},\mathit{free}}}\right)\cdot \frac{\mathit{TCN}}{V\mathit{medium}}$$ (8)

$$\frac{d{C}_{\mathit{medium},\mathit{exo}}}{\mathit{dt}}=(k_{\mathit{rel},\mathit{exo}}\cdot C_{\mathit{cell},\mathit{exo}}\cdot V_{\mathit{cell}}-\frac{\mathit{Jma}{x}_{\mathit{inter},\mathit{exo}}\cdot C_{\mathit{medium},\mathit{exo}}}{K{d}_{\mathit{inter},\mathit{exo}}+C_{\mathit{medium},\mathit{exo}}})\cdot \frac{\mathit{TCN}}{V\mathit{medium}}$$ (9)

$$C_{\mathit{medium},\mathit{bound}}=\frac{C_{\mathit{medium},\mathit{free}}\cdot B_{\mathit{medium},\mathit{max}}}{K{d}_{\mathit{medium},\mathit{bound}}+C_{\mathit{medium},\mathit{free}}}$$ (10)

For initial conditions (before treatment), C_medium,exo equals zero, and C_medium,free and C_medium,bound were calculated from C_medium,total using Eq. 11 (obtained by substituting Eq. 10 into Eq. 2, followed by rearrangement).

$$C_{\mathit{medium},\mathit{free}}=\frac{-(K{d}_{\mathit{medium},\mathit{bound}}+B_{\mathit{medium},\mathit{max}}-C_{\mathit{medium},\mathit{total}})+\sqrt{(K{d}_{\mathit{medium},\mathit{bound}}+B_{\mathit{medium},\mathit{max}}-C_{\mathit{medium},\mathit{total}})-4\cdot (C_{\mathit{medium},\mathit{total}}\cdot K{d}_{\mathit{medium},\mathit{bound}})}}{2}$$ (11)

Development of PD models

PD models describe cytotoxicity as a function of drug treatment (concentration and time). Eq. 12 depicts the net change in TCN due to (a) cell growth over time at a rate constant k_g until confluence, and (b) drug-induced cell kill as function of concentration of tubulin-bound drug. TCNss is maximal cell number at confluence. k_kill is maximum cell kill rate constant. Note that Eq. 12 does not account for potential concentration-dependent changes in k_kill. IC₅₀ is tubulin-bound drug concentration that generates 50% maximum cell kill. n is the Hill exponent. In view of the time- and dose-dependent PTX cytotoxicity [34] and the development of drug resistance over time in MCF7 cells [47], we used Eq. 13 to depict the time-dependent changes in IC₅₀, where IC_50,initial is the IC₅₀ value at time zero and γ_IC50 is the rate of IC₅₀ change per unit time t.

$$\frac{\mathit{dTCN}}{\mathit{dt}}=k_{\mathit{growth}}\cdot \text{TCN}\cdot \left(1-\frac{\mathit{TCN}}{\mathit{TCNss}}\right)-k_{\mathit{kill}}\cdot (\frac{{C_{\mathit{cell},\mathit{tubulin}}}^n}{{C_{\mathit{cell},\mathit{tubulin}}}^n+I{C}_{50}^n})\cdot \mathit{TCN}$$ (12)

$$I{C}_{50}=I{C}_{50,\mathit{initial}}+{\gamma }_{IC50}\cdot t$$ (13)

Model parameterization

The parameters for exosome-independent processes were obtained as follows. D_fd, and Kd_Pgp were taken from our published results [37, 39]. The calculation of PTX sequestration into exosomes required the C_cell,free-time profiles, which in turn required the rate constants of drug-tubulin interactions (k_tubulin,on, k_tubulin,off). PTX binding to tubulin depends on their nucleotide contents (GTP or GDP) and can be high affinity (Kd of 15 to 60 nM) or low affinity (Kd of 2 to 3 μM) [46, 48–51]. We used the reported parameters for high affinity binding (k_tubulin,on of 2 nM⁻¹s⁻¹, k_tubulin,off of 30 s⁻¹, Kd of 15 nM [48]). The current models also required B_{tubulin,initial} and kB_tubulin, which were obtained in two-steps: we first modified Eq. 1 to 6 to remove the exosome-related components because drug binding to tubulin is independent of exosomes, and then we fitted the modified equations to our previously obtained experimental results (C_cell,total vs. time plots after treatment with 0.1–1000 nM PTX [39]) to obtain the best-fitting parameters values.

To obtain the rate constants for the exosome-related PK processes (sorting, release, endocytosis), we used the above parameters together with the current models (Eq. 1–9 that incorporated exosome-related processes) to fit four sets of experimental data in PTX-treated Donor MCF7 cells obtained in our previous studies [27, 29], i.e., (a) C_cell,total vs. time plots, (b) time-dependent depletion of C_medium,total as function of initial C_medium,total (1 to 1000 nM), (c) Cytotoxicity_donor vs. treatment duration plots, and (d) ratio of A_medium,exo [drug amount in exosomes in culture medium, equals C_medium,total times V_medium] to A_cell,total [drug amount in Donor cells, equals C_cell,total times V_cell times TCN] at 24 h. The first three data sets were obtained from our earlier studies [34, 38, 39], whereas the last data set was from the current study (see Table 2 in Results). For example, k_sort,exo, k_rel,exo, Jmax_inter,exo, and Kd_inter,exo were obtained by simultaneously fitting the plots (a), (b) and (d) (see Figures 5A, 5B and 5D in Results) with Eq. 1–5, Eq. 2 and 8–10, and Eq. 1 and 4, respectively.

Table 2. PTX cellular pharmacokinetic and pharmacodynamic model parameters.

Some parameter values were obtained from the literature and some were obtained from analyzing the experimental results of the current study (indicated as Analysis). Mean values of parameters obtained from Analysis are shown with % CV in parentheses.

Category	Parameter	Annotation	Value (% CV)	Source
A. Cellular pharmacokinetic model
Pgp-mediated efflux	JmaxPgp (pmoleh−1·cell−1)	Maximum Pgp efflux rate	2.8 × 10−6	[37]
	KdPgp (nM)	Dissociation constant of drug from Pgp	13.9	[37]
Diffusion	Dfd (μL·h−1·cell−1)	Diffusion rate constant of free PTX across cell membrane	3.34×10−3	[39]
Intracellular microtubule binding	Btubulin,initial (μM)	Maximal initial PTX binding sites on tubulin	59.2	[39]
	ktubulin,off (s−1)	Rate constant of PTX dissociation from tubulin	30	[48]
	ktubulin,on (nM−1·s−1)	Rate constant of PTX association with tubulin	2	[48]
Binding to cell or extracellular fluid	Bmedium,max (μM)	Maximal extracellular drug binding sites	3.94	[39]
	Kdmedium,bound (nM)	Dissociation constant of PTX from extracellular binding sites	781	[39]
	Vcell (μL)	Volume of single cell	2.06 × 10−6	[39]
	Vm (μL)	Medium volume	1000	Experimental set-up
Exosome release and internalization	Jmaxinter,exo (pmole·h−1·cell−1)	Maximum rate of endocytosis of extracellular exosomes	0.038 (23)	Analysis
	Kdinter,exo (nM)	Dissociation constant of exosomes from endocytosis receptor	63 (45)	Analysis
	ksort,exo (h−1)	Rate constant for sorting intracellular PTX into exosome	31 (67)	Analysis
	krel,exo (h−1)	Rate constant for exosomes release into extracellular fluid	0.105 (12)	Analysis
B. Pharmacodynamic model
Drug cytotoxicity	kgrowth (h−1)	Cell growth rate constant	0.0288 (8)	Analysis
	kkill (h−1)	Maximum cell kill rate constant	0.0202 (11)	Analysis
	IC50,initial (nM)	Tubulin-bound concentration to produce 50% Emax at time zero	826 (27)	Analysis
	γIC50 (nM·h−1)	Change in IC50 over treatment time	7.2 (48)	Analysis
	n	Hill exponent of PTX concentration- cytotoxicity curve	1.67 (4)	Analysis
	Nss	Maximum cell number at confluence	1.1 × 106	Experimentally determined
	ICN	Initial cell number	5 × 103	Experimental set-up

Figure 5. Obtaining model parameters.

(A) Donor cell C_cell,_total vs. time plots. (B) Time-dependent changes in C_{medium, total} at four initial C_medium,total (1 to 1000 nM). (C) Cytotoxicity_donor vs. treatment concentration and duration plots. (D) Changes of A_medium,_exo as a fraction of A_cell,_total in Donor cells with time. The experimental results in (A), (B) and (C) were from our earlier studies ([34, 39] with permission), and the results in (D) are from the current study (calculated from Table 1; 24 h results only). Fitting Eq. 1–13 to the experiment data points (symbols) yielded the best-fitting curves and model parameter values (see text). Note the different x and y scales (i.e., 24 h in (A), (B) and (D) vs. 96 h in (C); logarithmic in (A) and (B) vs. linear in (C) and (D)).

PD parameters (k_kill, n, IC_50,initial, γ_IC50) were obtained using experimental results from our previous study [34] and the current study, by fitting Eq. 1–13 to the above plot (c) (see Figure 5C in Results). k_g was calculated from the cell doubling time [52]. TCN_ss was experimentally counted.

Evaluate model performance

QP models were used to predict the changes in Cytotoxicity_recipient of PTX-Exo and Cytotoxicity_donor of PTX due to exosome-mediated drug efflux. The model-predicted data were then compared with experimental results to evaluate model performance. The deviation between model-predicted results and experimentally observed results, as percentage, was calculated using Eq. 14 where O_i is the observation value and P_i is the predicted value.

$$M=\frac{100}{\mathrm{n}}\cdot {\displaystyle {\sum }_{i=1}^n\left(\frac{{\mathrm{O}}_{\mathrm{i}}-{\mathrm{P}}_{\mathrm{i}}}{{\mathrm{O}}_{\mathrm{i}}}\right)}$$ (14)

Sensitivity analysis

The four parameters on exosome-mediated drug transfer processes (k_sort,exo, k_rel,exo, Jmax_inter,exo, Kd_inter,exo) were evaluated for their effects on three cellular PK and PD endpoints (C_cell,total, C_cell,tubulin, Cytotoxicity) in both Donor and Recipient cells. Sensitivity analysis was performed by (a) changing the value (10-fold increase or decrease) of a selected parameter while keeping all other parameters constant in Donor cells, and (b) comparing the differences in the simulated outcomes without or with the parameter value change. These exosome-related effects were further compared to the effects caused by changes in tubulin binding capacity, a parameter we have demonstrated to significantly alter the cellular PTX PK [37, 39]. Simulations of Cytotoxicity_recipient of PTX-Exo used equal cell numbers, i.e., PTX-Exo derived from a selected number of Donor cells were applied to the same number of Recipient cells.

Statistical analysis

Experimental results were analyzed for statistical significance using Student’s t-test (unpaired, two-tailed) or Dunnett’s test (Prism 7). p<0.05 was considered as significant. Data are presented as mean ± SD (standard deviation) or mean ± SEM (standard error of the mean).

Results

Characterization of exosomes

Transmission electron microscopy (Figure 2A) showed the spherical shape of exosomes (from MCF7 cells) obtained after drying and processing. The average particle diameter, based on nanoparticle tracking analysis, was 108 nm (n=100, Figure 2B). Exosomes contained high levels of CD63 (a tetraspanin and marker of exosomes) and TSG101 (a component of ESCRT-I complex or endosomal sorting complexes required for transport), but undetectable levels of skeleton protein β-actin and endoplasmic reticulum protein calnexin, whereas the reverse was found for Donor cell lysates (Figure 2C).

Figure 2. Characterization of cancer cell-derived exosomes.

(A) Transmission electron microscopy. Representative image of MCF7 exosomes. (B) Size distribution of MCF7 exosomes by nanoparticle tracking analysis. Black: rolling average. Red: SD of results from three repeated measurements. (C) Western blot results. Exosomes and Donor cell lysates, collected from A2780 cells and LM2 cells, were analyzed for CD63, TSG101, calnexin and β-actin.

Cytotoxicity_donor of PTX and DOX

Figure 3A shows Cytotoxicity_donor of PTX and DOX. PTX yielded incomplete cytotoxicity in all 4 cells; MCF7 cells were the least sensitive with the highest residual unaffected fraction R_e. In contrast, DOX yielded complete cytotoxicity in the two cells studied, with ~8-times higher activity in A2780 cells compared to MCF7 cells. Subsequent studies on the effects of PTX and DOX on exosomes used drug concentrations that corresponded to their EC₂₀, EC₅₀, and EC₈₀ values in individual cell lines.

Figure 3. Cytotoxic concentrations of paclitaxel and doxorubicin induced exosome production and/or release.

(A) Cytotoxicity_donor of PTX and DOX. MCF7, LM2, A2780, and OVCAR4 cells were treated for 24 or 48 h with PTX or DOX, and the remaining cell numbers were determined using SRB assay. Results were analyzed for the values of drug concentrations producing 20%, 50% and 80% of the maximal cytotoxicity (EC₂₀, EC₅₀ and EC₈₀, respectively). (B) Cytotoxics enhanced exosomes. Cells were treated for 24 h or 48 h with PTX or DOX at C_medium,_total equivalent to their EC₂₀, EC₅₀ and EC₈₀ for 48 h treatment. Number of exosomes in culture medium after drug treatment for 24 h (circles) and 48 h (open triangles) were standardized by cell numbers (mean ± SE, n = 3 experiments, four replicates per experiment). *p<0.01 for 24 h vs. 48 h (unpaired Student’s t-test).

PTX and DOX stimulated exosome production and/or release

Figure 3B shows the levels of exosomes in Conditioned Medium. Without drug treatments, the four cells during exponential growth phase (over 24 to 48 h) yielded comparable exosome levels, ranging from 170 ± 35 to 280 ± 68 exosomes per cell (mean ± SEM). Treatments with cytotoxic concentrations of PTX or DOX generally enhanced the exosome levels, in drug concentration- and time-dependent manners. The data also showed drug- and cell-specific differences. For example, compared to untreated control, treatments with PTX at EC₈₀ for 48 h increased the exosome level by ~3-folds in the two breast cells (LM2 and MCF7) and ~10-folds in the two ovarian cells (A2780 and OVCAR4), and the exosome level in MCF7 and A2780 cells continued to increase with time (50–200% higher after 48 h treatment compared to 24 h) whereas the maximal increase was reached at 24 h with no further increases in LM2 or OVCAR4 cells. Similarly, treatment with DOX at EC₈₀ enhanced the exosome level in MCF7 by 4- to 6-folds after 24 or 48 h, but required longer treatment duration to produce changes in A2780 cells (no change after 24 h and 7-fold increase after 48 h).

Quantification of PTX in PTX-Exo and cell lysate by LC-MS/MS

Table 1 summarizes the LC-MS/MS results of PTX concentrations or amounts in PTX-Exo or cell lysates, after 24 h treatment. In general, the four cancer cells showed comparable PTX concentrations in cell lysates or exosomes (e.g., ~27 pmole in lysates and ~1.5 pmole in exosomes per 10⁶ trypan blue-excluding cells at 100 nM C_medium,total), and increasing PTX concentrations/amounts with increasing C_medium,total. However, the ratios of PTX concentration in PTX-Exo and cell lysate indicate (a) substantial cell-specific differences, i.e., up to two-fold difference between cells (e.g., 4.3% in LM2 cells vs. 8.3% in OVCAR4 cells at 100 nM C_medium,total) and (b) nonlinear uptake of PTX and nonlinear drug sorting into PTX-Exo (e.g., the PTX concentration increases in PTX-Exo and cell lysate were not proportional to either the increase in C_medium,total (average of ~4-folds vs. 10 folds) or to each other (e.g., average of ~2.5-folds vs. ~4-folds)).

Table 1. Amount of PTX in PTX-Exo or cell lysates.

Cells were treated with 100 or 1000 nM PTX for 24 h. PTX-Exo was collected from Conditioned Medium. Cells were washed and collected with trypsinization, counted and lyzed. Drug concentrations in exosomes and cell lysates were analyzed using LC-MS/MS, and standardized to cell number. The numbers of exosomes per LM2, MCF7, A2780 and OVCAR4 cell were, respectively, 1170, 955, 2236 and 2457 at 100 nM PTX and 1232, 1060, 2513 and 2794 at 1000 nM PTX. Results are mean ± SD (three independent experiments with three replicates per experiment).

Cell line	PTX, pmole per 106 Donor cells
	Treated with 100 nM PTX			Treated with 1000 nM PTX
	Exosome	Lysate	Exosome-to-Lysate ratio, %	Exosome	Lysate	Exosome-to-Lysate ratio %
LM2	1.13±0.02	25.0±0.10	4.27±0.22	6.18±0.65	62.2±0.19	9.29 ± 1.08
MCF7	1.60±0.04	29.3±0.54	5.12±0.36	8.33±0.16	69.7±0.53	11.59 ± 0.79
A2780	1.44±0.08	23.8±0.23	6.67±1.12	5.38±0.27	73.4±3.49	7.67 ± 1.64
OVCAR4	2.38±0.06	28.6±0.47	8.28±0.69	6.71±0.46	71.7±2.72	9.65 ± 0.33
Average	1.64±0.27	26.7±1.34	6.08±1.77	6.65±0.62	69.3±2.46	9.55 ± 1.61

The amount of PTX in PTX-Exo was between 6–10% of the amount in Donor cell lysate (Table 1). We further calculated the PTX concentration in PTX-Exo from MCF7 cells treated with 100 nM C_medium,total for 24 h, as [Amount_medium,exo of 1.6 pmole per 10⁶ cells] divided by [Volume of exosomes excreted by 10⁶ cells, calculated using an average exosome radius of 54 nm and release of 10³ exosomes per cell (see Table 1)]; the resulting concentration in exosomes equaled 2.4 mM, which is ~60-times the C_cell,total and ~30,000-times the C_medium,total (experimentally determined to be 39 μM and 83 nM, respectively, after treatment with 100 nM for 24 h [39]).

PTX-Exo exhibited Cytotoxicity_recipient and Anti-migration_recipient

Figure 4A compares the Cytotoxicity_recipient of PTX to that of PTX-Exo. Each treatment condition had its own control, i.e., no drug for the PTX treatment and exosomes collected from untreated cells (Untreated-Exo) for the PTX-Exo treatment. The amount of PTX-Exo added was calculated based on the drug concentration in PTX-Exo as determined by LC-MS/MS, whereas the amount of Untreated-Exo added was calculated to contain the same protein amount as in PTX-Exo. No cytotoxicity was observed in all control groups, confirming Untreated-Exo had no cytotoxicity (not shown). In contrast, PTX-Exo_{10–1000nM,24–48h} induced significant Cytotoxicity_recipient in all four cells; the extent increased with the drug treatment concentration and duration of Donor cells (e.g., greater cytotoxicity for PTX-Exo_48h compared to PTX-Exo_24h) and followed the rank order of the chemosensitivity of individual cells (LM2>MCF7>A2780>OVCAR4).

Figure 4. PTX-Exo conferred dose-dependent Cytotoxicity_recipient (A and B) and Anti-migration_recipient (C and D).

For (A) and (B), drug-naïve cells (MCF7, LM2, A2780, OVCAR4) were treated with PTX or PTX-Exo collected from Conditioned Medium of their corresponding Donor cells. The Recipient cell number was determined using SRB assay. Data are Mean ± SD (n=4 experiments with triplicates per experiment). (A) Cytotoxicity_recipient of PTX-Exo. PTX-Exo_24h (circles) and PTX-Exo_48h (open triangles), both containing 10 μg proteins, were incubated with Recipient cells for 48 h. (B) Comparison of Cytotoxicity_recipient of PTX-Exo and PTX. Recipient cells were treated with PTX (solid symbols) or PTX-Exo (open symbols) collected from Donor cells after 24 h PTX treatment. The x-axis denotes the PTX-equivalent concentration added to the culture medium, where the amount of the added PTX-Exo was calculated based on the LC-MS/MS results. Note the different scales in (A) and (B). For (C) and (D), LM2 cells were cultured in 6-well plates, without drug treatment (control), or treatment with exosomes collected from untreated Donor cells (Untreated Exo) or PTX-Exo for 48 h. The exosomes added to each well contained 100 μg proteins. (C) Inhibition of cell migration. The initial open areas (i.e., not covered by cells) in untreated control and Untreated-Exo groups were equal (62 ± 2%; mean ± SD, n=6 experiments, duplicates in each experiment). (D) Concentration- and time-dependent Anti-migration_recipient by PTX-Exo_24h (open bars) and PTX-Exo_48h (filled bars) harvested from Donor cells treated with 5, 50 or 500 nM PTX. Statistically significant differences between (a) Control and other groups (p<0.01, Dunnett’s test), (b) Untreated-Exo and PTX-Exo groups (p<0.01, Dunnett’s test), (c) among the three PTX-Exo groups at different C_medium,_total (p<0.01, Dunnett’s test), and (d) between PTX-Exo_5nM,_24h and PTX-Exo_5nM,_48h (*p<0.01, unpaired Student’s t-test).

We further compared Cytotoxicity_recipient of PTX and PTX-Exo, at equal PTX-equivalent concentrations (Figure 4B); the overlapping concentration-effect relationships of PTX and PTX-Exo in all 4 cells indicate Cytotoxicity_recipient of PTX-Exo was due to its PTX content.

Figure 4C shows Anti-migration_recipient of PTX-Exo. After 48 h incubation, the untreated control showed a much smaller scratch-induced open-wound gap (from 62% to 9%, equivalent to 85% reduction), indicating substantial cell migration. In comparison, Untreated-Exo slightly but significantly promoted the cell migration as indicated by a yet smaller open-wound gap (>90% reduction, p<0.05 compared to untreated control); this minor effect is consistent with the reported invasion and pre-metastasis functions of tumor-derived exosomes [16, 53]. In contrast, PTX-Exo_{5–500nM,24–48h} significantly inhibited the migration, as indicated by a larger open wound area (e.g., decreased from 62% to 43%, equivalent to 31% reduction, for PTX-Exo_500nM,48h). As observed for Cytotoxicity_recipient, Anti-migration_recipient of PTX-Exo generally increased with PTX treatment concentration and duration of Donor cells (Figure 4D); the differences between C_medium,total of 5, 50 and 500 nM were significant (p<0.05 compared to untreated control or Untreated-Exo groups). However, prolonging treatment from 24 to 48 h yielded a significant difference only at the lowest concentration and not at the two higher concentrations; this concentration-dependence may be due to the different rates for C_exo,total to equilibrate with C_cell,total.

Quantitative pharmacology models of exosome-mediated effects

Computational models were constructed to capture the above cellular PK and PD findings (Eq. 1–13). Figures 5A–5D show the best-fitting curves for changes in C_cell,total over time at different initial C_medium,total, depletion of C_medium,total as function of initial C_medium,total, Cytotoxicity_donor as function of drug treatment duration and initial C_medium,total, and changes in % of A_medium,exo as a fraction of A_cell,total with treatment time at initial C_medium,total of 100 and 1000 nM, respectively. Table 2 summarizes the best-fitting PK and PD model parameter values. The coefficients of variations (CV) for most parameters were between <10% to 25%; the three exceptions are the parameters for exosome transport Kd_inter,exo and k_sort,exo and for resistance development over time γ_IC50 where the CV was between 45% and 67%; these high CV were in part due to the inability to experimentally measure C_cell,exo.

Comparisons of model parameter values provided the following insights regarding exosome-mediated drug transfer and PD. The value of k_sort,exo of ~31 h⁻¹ corresponds to a half-life of <2 min, indicating rapid sorting of C_cell,free into exosomes. Internalization of PTX-Exo into Recipient cells was also rapid, as demonstrated by a relative high Jmax_inter,exo value. In comparison, the release of PTX-Exo into extracellular medium was much slower (k_rel,exo of 0.105 h⁻¹ per cell, which corresponds to a half-life of ~6–7 h). Hence, release of PTX-Exo from Donor cells is rate-limiting for their appearance in extracellular fluid and Recipient cells. With respect to PD, the γ_IC50 of 7.2 nM per h indicates a doubling of the initial IC₅₀ of 826 nM in about 115 h, which is consistent with the finding of a separate study showing that continuous PTX treatment led to a 2-fold higher IC₅₀ in human prostate PC3 cells after 120 h (unpublished results).

Evaluation of model performance

Figure 6 compares the QP model-simulated Cytotoxicity_recipient of PTX-Exo with the experimental results, at four initial C_medum,exo (3.5 to 45 nM) and four treatment durations (24 to 96 h). The good agreement between these results (average deviation of <10% for all data, <15% deviations for individual data points) indicates the QP models in Eq. 1–13 successfully depicted the pharmacological activity of PTX-Exo.

Figure 6. QP model-simulated pharmacodynamics of PTX-Exo: Comparison with experimental results.

Model-simulated results (curves) vs. experimentally observed results (symbols) on Cytotoxicity_recieint of PTX-Exo as functions of drug treatment concentration and duration. Drug concentrations in culture medium at time zero were calculated as [drug amount in PTX-Exo] divided by extracellular fluid volume (equaled 1000 μL). For experimental data, the remaining cell number was measured using SRB assay (n = 3 experiments, four replicates per experiment). Deviations between simulated and experimental results were 12 ± 7% (mean ± SD).

Sensitivity analysis

The sensitivity analysis results identified the different key determinants of three PK and PD endpoints (C_cell,total, C_cell,tubulin and Cytotoxicity) in Donor and Recipient cells; these endpoints are expressed as area-under-curve from 0 to 24 h (Table 3). The three model parameters (k_sort,exo, k_rel,exo, Jmax_inter,exo) exerted different effects at low and high initial C_medium,total (1 vs. 1000 nM) and in Donor vs. Recipient cells, as follows.

Table 3. Effect of exosome production and internalization on PTX accumulation and cytotoxicity in cells: Model-simulated results.

Simulations were performed for Donor cells treated with 1 or 1000 nM PTX for 24 h. The resulting values of C_cell,total, C_cell,tubulin and Cytotoxicity (E, rate constant of drug-induced cell death) are represented as area-under-curve from 0 to 24 h. E was calculated as $k_{\mathit{kill}}\cdot \left(\frac{{C_{\mathit{cell},\mathit{tubulin}}}^n}{{C_{\mathit{cell},\mathit{tubulin}}}^n+I{C}_{50}^n}\right)$. The original values of model parameters are 31 h⁻¹ for k_sort,exo, 0.105 h⁻¹ for k_rel,exo, 0.038 pmole·h⁻¹·cell⁻¹ for Jmax_inter,exo, and 59.2 μM for B_{tubulin,initial}. With the exception of B_{tubulin,initial} where the low and high values were one-half or 2-times the original values, the low and high values of all other parameters were one-tenth or 10-times the original values. Signs in the table correspond to minimal change of <20% (−), increase or decrease of 20–50% (↑ and ↓), 51–100% (↑↑ and ↓↓), 101–200% (↑↑↑ and ↓↓↓), and >200% (↑↑↑↑ and ↓↓↓↓).

Parameter				A. Effects on Donor cells						B. Effects on Recipient cells
				1 nM			1000 nM			1 nM			1000 nM
Ccell,total	Ccell,tubulin	E	Ccell,total	Ccell,tubulin	E	Ccell,total	Ccell,tubulin	E	Ccell,total	Ccell,tubulin	E
Exosome production	ksort,exo	Low	–	–	–	↑	–	–	↓	↓	↓↓	↓	↓	↓↓
		High	–	–	–	↓	–	–	↑	↑	↑↑	↑↑	↑↑	↑↑
	krel,exo	Low	↑	↑↑	↑↑ ↑↑	↑	↑	–	↓	↓	↓↓	↓↓	↓↓	↓↓↓
		High	↓	↓↓	↓↓ ↓↓	↓	↓	–	↑↑	↑↑	↑↑ ↑	↑↑	↑↑	↑↑↑ ↑
Exosome internalization	Jmaxinter,exo	Low	–	↓	↓	–	–	–	↑	↑	↑↑ ↑	↑↑	↑↑	↑↑↑ ↑
		High	–	↑	↑	–	–	–	↓	↓	↓	↓	↓	↓↓
Tubulin binding capacity	Btubulin,initial	Low	↓	↓↓	↓↓ ↓	↓↓	↓↓↓	–	↑	↑	↑	↑	↑↑	↑↑
		High	↑	↑↑	↑↑ ↑	↑↑	↑↑↑	–	↓	↓	↓↓	↓	↓	↓↓

For Donor cells treated with 1 nM C_medium,total, a 10-fold increase or decrease in k_rel,exo produced small-to-moderate changes in C_cell,total (20–50%) and C_cell,tubulin (51–100%), but significantly greater changes in Cytotoxicity_donor (241%), whereas similar changes in k_sort,exo or Jmax_inter,exo yielded minor or no changes in all three endpoints (<20%). At 1000 nM C_medium,total or when drug binding to tubulin would be saturated and consequently maximal cytotoxicity would have been reached, 10-fold changes in k_sort,exo or k_rel,exo produced 20–50% changes in C_cell,total and/or C_cell,tubulin, whereas Jmax_inter,exo had little or no effects, and none of the three parameters affected Cytotoxicity_donor. These findings indicate (a) exosome release played a greater role on C_cell,tubulin and C_cell,total compared to exosome sorting or re-uptake, especially at low C_medium,total of 1 nM, and (b) exosome release played a major role in Cytotoxicity_donor at low C_medium,total but had no effect at high C_medium,total. These concentration-dependent outcomes indicate complex interplay between exosome-mediated processes and other competing linear and nonlinear processes that also determine C_cell,tubulin and hence Cytotoxicity_donor.

Recipient cells showed changes in the opposite direction with substantial quantitative differences. Because <10% C_cell,total in Donor cells was released in PTX-Exo, which was the drug source for Recipient cells, C_cell,total in Recipient cells was much lower compared to Donor cells. In general, changes in exosome-related model parameters had greater effects on Cytotoxicity_recipient than on Cytotoxicity_donor, especially at the high C_medium,total of 1000 nM. Changes in Jmax_inter,exo also affected C_cell,total and C_cell,tubulin in Recipient cells more than in Donor cells.

We further compared the effect of exocytosis of PTX-Exo to the effect of a known paclitaxel resistance mechanism, i.e., tubulin alteration resulting in reduced drug binding as reflected by a lower B_{tubulin,initial} [39, 54]. The simulation results indicate a 10-fold increase in k_rel,exo yielded effects that are comparable to a 2-fold decrease in B_{tubulin,initial} on all three endpoints in Donor cells. Changes in B_{tubulin,initial} had less impact on Recipient cells compared to Donor cells, presumably due to nonlinear exocytosis and endocytosis of PTX-Exo.

Comparison of drug efflux rates by different mechanisms

We used simulations to compare the drug efflux rates via passive diffusion and the two active efflux by Pgp and exosomes; the simulations were for cells treated with PTX C_medium,total of 0.1–1000 nM for 24 h or after C_cell,total has reached a plateau level. The results, shown in Figure 7, indicate passive diffusion was the dominant efflux mechanism at all C_medium,total, accounting for at least 92% of total efflux. At C_medium,total exceeding 30 nM or when PTX binding to tubulin was saturated, C_cell,free increased nonlinearly with C_medium,total and resulted in greater efflux by all three mechanisms such that efflux by passive diffusion increased from ~12-times to 44-times the active efflux at 1000 nM. The contributions of two active efflux mechanisms changed with C_medium,total due to saturation of Pgp-efflux, i.e., Pgp-efflux initially exceeded exosome-efflux (e.g., 235%–88% higher at up to 100 nM C_medium,total) but became less efficient at higher C_medium,total (e.g., one-fourth the exosome-efflux at 1000 nM).

Figure 7. Contribution of drug efflux mechanisms.

Drug efflux rates by passive diffusion and two active efflux mechanisms (Pgp and exosomes) were simulated using Eq. 3–4 for cells treated with 0.1–1000 nM PTX for 24 h.

Discussion

The present study indicated several new findings regarding the potential roles for exosomes in pharmacological effects of cytotoxics in solid tumors.

We observed that cytotoxics such as DOX and PTX stimulated the production and/or release of exosomes containing high drug levels. The sorting of intracellular drug content into exosomes and their release allows the cell to reduce the intracellular drug concentration (e.g., 15–50% reduction of C_cell,tubulin in the current study). Hence, drug-stimulated exosome production/release may reduce drug activity, rendering this effect a potential new chemoresistance mechanism. This is supported by our QP model-simulated results indicating that in PTX treatments, a 10-fold increase in k_rel,exo was as effective in reducing Cytotoxicity_donor as a 2-fold reduction in drug binding to tubulin, a known mechanism of PTX resistance [54]. Note the current study was conducted using MCF7 cells which have relatively low Pgp level; under this setting the exosome-efflux accounted for up to 80% of total active drug efflux at initial C_medium,total of 1000 nM, whereas Pgp-efflux accounted for the remaining 20%. It is conceivable that drug elimination via exosomes may become less pronounced in cells with higher Pgp levels. Studies to investigate the relative contribution of exosome- and Pgp-mediated drug depletion in cells with different Pgp expression are ongoing in our laboratory.

Second, we calculated the drug concentration in PTX-Exo per unit volume (i.e., 2.4 mM in PTX-Exo from MCF7 cells treated with 100 nM C_medium,total) was 60-times the C_cell,total and 30,000-times the C_medium,total, indicating PTX-Exo represents an important drug depot. This high PTX level in exosomes is not likely due to pH trapping in acidic exosomes as reported for weakly basic drugs such as DOX and cisplatin [55–57], since the pKa of paclitaxel is 11.5 and is unionized at physiological pH [58]. A possible cause is exosomes contain tubulin [59, 60], which, based on the extensive PTX binding to tubulin and the following calculation, would readily account for the high drug concentration in PTX-Exo. Based on our previously reported C_cell,total of 39 μM at 100 nM C_medium,total [39] and cell volume of 2 μL per 10⁶ cells [39], and since nearly all C_cell,total is due to macromolecule-bound drug, we calculated the maximal bound drug would be 78 pmole PTX per 10⁶ cells. With tubulin being the primary binding site and because tubulins constitute ~1% of total proteins in neuroblastoma and HeLa cells (3 vs. 300 pg per cell [61]), the amount of tubulin-bound PTX would be 26 μmol per gram tubulin (equals 78 pmole divided by 3 × 10⁶ pg). Using the experimentally determined protein amount in PTX-Exo (1.3 μg per 10⁶ MCF7 cells) and assuming the same 1% protein-to-tubulin ratio as in cells, the tubulin amount in PTX-Exo would be 13 ng per 10⁶ MCF7 cells, corresponding to a maximal tubulin-bound drug concentration of ~3.4 mM in PTX-Exo. This calculation shows that preferential sorting of tubulin and tubulin-bound PTX would lead to the high drug concentration in PTX-Exo observed in the present case. We further evaluated whether inclusion of this process (i.e., sorting of tubulin-bound drug into exosomes) in the QP model would significantly alter the PK/PD in Donor cells; the simulations indicated the model modification yielded negligible changes in the C_cell,total-time profiles (~5% compared to without this additional sorting step).

Third, our results indicate exosome is a mechanism of intercellular drug transfer. Unlike diffusion which is driven by the concentration gradient, exosomes, which use receptor-mediated endocytosis for internalization [14], enable the drug sequestered in exosomes to enter cells irrespective of the concentration gradient. An earlier study shows that exosomes isolated from prostate cancer cells and loaded with PTX extracellularly enter cells via endocytosis [62]. Another notable property is that exosomes, as nanoparticles, are less readily removed from tumor interstitium compared to free drug and thereby provide sustained drug exposure in tumors. This property may have therapeutic importance as our experimental results indicate substantial Cytotoxicity_recipient by PTX-Exo.

Fourth, our study used PTX-Exo collected from epithelial cancer cells. Their pharmacological effects, together with an earlier report that PTX-treated dendritic cells yielded exosomes capable of inducing cytotoxicity in mdr1-transfected canine kidney cells [63], suggest release of pharmacologically active PTX-Exo is a general property that occurs in different cell types/lineages.

Fifth, the finding that exosome release, rather than sorting or internalization via receptor-mediated endocytosis, is the rate-limiting step of exosome appearance in Recipient cells suggests that its perturbation, e.g., by agents that block exosome release, will be a critical determinant of intercellular drug transfer.

Finally, the current study established the utility of QP models to delineate the complex interplay between exosome-related processes and other competing intracellular and extracellular processes, and the pharmacological consequences of these various processes. The current, first-generation model does not account for cell doubling leading to dilution of drug concentration, concentration-dependent cell kill mechanisms (e.g., apoptosis, necrosis), potential cytostasis (which would decrease the k_growth), nor PTX transfer between C_medium,free and C_medium,exo. Such dynamic changes may be considered in future model development. Additional areas that warrant further studies include spatial distribution and residence of exosomes within tumor interstitium, their transport into systemic circulation, and factors that alter their production and release.

Conclusion

The present study demonstrated exosomes as a mechanism to reduce intracellular drug concentration as well as a mechanism of intercellular drug transfer, with significant pharmacological consequences. We further provided the first QP models that (a) captured the PK of cellular PTX processing and drug release through exosomes, and (b) successfully described the effects of exosome exocytosis on PD in Donor cells and drug-naïve Recipient cells. These findings and computational tools may be used to interrogate exosomes as a resistance mechanism and as a means to deliver drugs to the hard-to-reach regions in solid tumors, and to predict the contributions of these opposite effects of exosomes to therapeutic outcomes.

Abbreviations

AChE

acetylcholinesterase

Conditioned Medium

culture medium collected after incubating cells under preselected conditions

CV

coefficient of variation

Cytotoxicity_donor

cytotoxicity in Donor cells

Cytotoxicity_recipient

cytotoxicity in Recipient cells

DMSO

dimethyl sulfoxide

Donor cells

cells that provide drug-containing exosomes

DOX

doxorubicin

Drug-Exo_conc,time

exosomes collected from Donor cells treated with a drug at a preselected concentration for a preselected duration

E_max

maximum effect

ECn

drug concentration producing n% of E_max

ESCRT

endosomal sorting complexes required for transport

Growth Medium

cell culture medium supplemented with 10% fetal bovine serum (either unaltered or exosome-depleted) and antibiotics

IC₅₀

tubulin-bound paclitaxel concentration that generates 50% maximum cell kill

LC-MS/MS

liquid chromatography–tandem mass spectrometry

PBS

phosphate-buffered saline

PD

pharmacodynamics

Pgp

P-glycoprotein

PK

pharmacokinetics

PTX

paclitaxel

QP

quantitative pharmacology

Recipient cells

drug-naïve cells receiving treatment with drug-containing exosomes

SD

standard deviations

SEM

standard error of the mean

SRB

sulforhodamine B

For abbreviations used in equations

A

denotes drug amount

B

denotes drug binding sites

C

denotes concentration

D

denotes diffusion

J

denotes flux

Kd

denotes dissociation constant

k

denotes rate constant

NSB

denotes non-saturable binding

t

denotes time

subscripts are used to denote location (extracellular, intracellular, exosomes), drug moieties with respect to binding status (free, bound, total) to macromolecules/organelles (tubulin, exosomes)