PBPK Modeling-based optimization of site-specific chemo-photodynamic therapy with far-red light-activatable paclitaxel prodrug

Abstract

Photodynamic therapy (PDT) is a clinically approved therapeutic modality to treat certain types of cancers. However, incomplete ablation of tumor is a challenge. Visible and near IR-activatable prodrug, exhibiting the combined effects of PDT and local chemotherapy, showed better efficacy than PDT alone, without systemic side effects. Site-specifically released chemotherapeutic drugs killed cancer cells surviving from rapid PDT damage via bystander effects. Recently, we developed such a paclitaxel (PTX) prodrug that targets folate receptors. The goals of this study were to determine the optimal treatment conditions, based on modeling, for maximum antitumor efficacy in terms of drug-light interval (DLI), and to investigate the impact of rapid PDT effects on the pharmacokinetic (PK) profiles of the released PTX.

PK profiles of the prodrug were determined in key organs and a quantitative systems pharmacology (QSP) model was established to simulate PK profiles of the prodrug and the released PTX. Three illumination time points (DLI = 0.5, 9, or 48 h) were selected for the treatment based on the plasma/tumor ratio of the prodrug to achieve V-PDT (vascular targeted-PDT, 0.5 h), C-PDT (cellular targeted-PDT, 48 h), or both V- and C-PDT (9 h).

The anti-tumor efficacy of the PTX prodrug was greatly influenced by the DLI. The 9 h DLI group, when both tumor and plasma concentrations of the prodrug were sufficient, showed the best antitumor effect. The clearance of the released PTX from tumor seemed to be largely impacted by blood circulation. Here, QSP modeling was an invaluable tool for rational optimization of the treatment conditions and for a deeper mechanistic understanding of the positive physiological effect of the combination therapy.

1. Introduction

An effective regimen for controlling local and regional tumors, for which surgery is not possible, is essential for cancers such as esophageal, head and neck, bladder, inflammatory breast, and peritoneally metastasized ovarian cancers. Photodynamic therapy (PDT), a clinically approved visible and near IR-light-based approach to treat certain types of cancers, is a viable option to treat those local tumors due to high site-specificity. PDT generates reactive oxygen species from photosensitizer (PS), oxygen, and light. It damages tumors via three well-known mechanisms: direct cancer cell killing, vascular damage, and immune activation [1, 2]. PDT can directly kill cancer cells by the induction of necrosis or apoptosis. It can also damage the endothelial cells of surrounding microvessels during the light treatment, which leads to vessel stasis or thrombosis. This can further produce an acute inflammatory response that attracts leukocytes, such as dendritic cells (DCs) and neutrophils. Matured DCs will present antigens to the T lymphocyte. Activated T cells will migrate to the tumor and start killing cancer cells [2].

The time of illumination, also called drug-light interval (DLI), in PDT is important for its tumor damage mechanism and antitumor efficacy. PDT has unique mechanisms and pharmacodynamics (PD) by transient active species, ROS, particularly singlet oxygen (SO) [3]. SO has an extremely short half-time (< 40 ns) and diffusion distance (~ 20 nm) in biological conditions [4]. Thus, SO is generated only during the illumination and at the site with a photosensitizer. Two of the major mechanisms of tumor damage, direct cell kill and vascular damage, occur rapidly while a tumor is illuminated. Thus, the concentration of PSs in blood and tumor at the time of illumination, which dictates the selection of DLI, is a critical factor for tumor damage mechanisms and the antitumor efficacy of PDT [5–7]. In typical PDT protocols, a tumor is illuminated either at a very early time point (DLI < ~ 1 h) for anti-vascular PDT (V-PDT, e.g., TOOKAD) or a late time point (DLI = ~ 1–3 days) for effective cancer cell killing (C-PDT, e.g., Photofrin) [8, 9].

We have developed a unique prodrug system in which visible and near IR-light activates prodrugs; this system takes advantage of PDT and site-specific chemotherapy [10–12]. The prodrugs are comprised of a photosensitizer and anticancer drug(s), which are conjugated with a singlet oxygen cleavable linker (Fig. 1A) [10, 13, 14]. The anticancer drug(s) can be site-specifically released at the tumor site by the controlled illumination, to prevent the systemic toxicity that arises from systemically circulating anticancer drugs (Fig. 1B) [11].

Figure 1.

A. Structure of FR-targeted prodrug (FA-PEG2K-Pc-L-PTX) and B. Schematic representation of light-activated & singlet oxygen-mediated release of PTX.

The released drug effectively kills cancer cells surviving rapid PDT damage via bystander effects [10, 15]. In in vivo conditions, the released drug is cleared from the tumor by circulating blood. With the combination of PDT and chemotherapy, direct PDT damage by SO occurs within seconds to minutes, whereas anticancer drugs cause sustained damage over hours to days, killing cells that remain after PDT damage. Therefore, effective and complete killing of bystander cells is critical for complete ablation of tumor. Based on our in vitro QSP study, we showed that the clearance rate of the released anticancer drug from tumor could be an important factor for effective cell killing by the released drugs [15]. The released PTX cytotoxicity was significantly enhanced by lowering the clearance value. Rapid vascular damage by PDT could delay the clearance of the released drug from tumor. Hence, we expect it could enhance the bystander effect of PTX.

The current study tested the hypothesis that the antitumor activity of light-activatable prodrug is enhanced by optimizing DLI. We used our folate receptor (FR)-targeted paclitaxel prodrug [13]. We performed PK studies of the PTX prodrug and developed a QSP model to describe (1) tissue distribution of PTX prodrug, (2) intracellular delivery of PTX prodrug, and (3) biodistribution of released PTX. The results and the model were then used to choose different DLIs to achieve anti-vascular, cytotoxic, or both anti-vascular and cytotoxic PDT effects, and the resulting in vivo antitumor efficacy was determined. We further used our model to predict PK of the released PTX and experimentally determined the concentration of the released PTX in tumor and plasma to validate our predictions. With the help of this multi-scale model, we optimized the treatment protocol of PTX prodrug to achieve maximum antitumor efficacy and were able to explain the DLI-dependent responses of the treated tumor at both cellular and systemic levels.

2. Methods

2.1. Reagents and Materials

Reagents and analytical-grade solvents were obtained from Sigma-Aldrich, VWR, or Fisher Scientific, and were used without further purification. FA-PEG2K-Pc-L-PTX was synthesized as previously reported by our laboratory [13]. UV–Vis absorbance was measured on the UV–VIS spectrometer LAMBDA 25 (PerkinElmer) using 10-mm optical path length quartz cuvettes. The fluorescence was recorded on a plate reader (SpectraMax Gemini EM, Molecular Devices) using a 96-well plate, 200 uL in volume, bottom reading, and excitation-emission at 605 – 680 nm.

2.2. Preparation of PTX prodrug

PTX prodrug stock solution was prepared fresh in DMSO for use each day. The concentration of the prodrug was determined (in triplicate) by absorbance at 678 nm (extinction coefficient = 209,967 cm⁻¹·M⁻¹) and adjusted to stock solutions of around 4 mM in DMSO [13]. For in vitro uptake, in vivo PK, and antitumor efficacy studies, the formulation of PTX prodrug was prepared in Cremophor EL. Cremophor (5% by volume) and absolute ethanol (5% by volume) were added to PBS (90% by volume) followed by vortexing for 30 s. The PTX prodrug stock solution in DMSO was then added to the Cremophor /Ethanol/PBS (1:1:18 v/v/v) solution, and the mixture was vortexed for 30 s. The mixed solution was then filtered through the 0.45-μm Millex-HV membrane (Durapore PVDF membrane; Merck Millipore Ltd). The final concentration of PTX prodrug after filtration was determined by measuring absorbance of aliquots in DMSO at 678 nm, as mentioned in section 2.1.

2.3. In vitro uptake of PTX prodrug

Colon 26 cells (NCI DCTD Tumor repository) were seeded in 96-well plates at a density of 10,000 cells/well in 180 μL complete medium (RPMI-1640, 10% fetal bovine serum, 1% penicillin /streptomycin) and were incubated at 37°C for 24 h. The PTX prodrug was diluted to 50, 100, and 200 μM with complete medium, and the diluted solutions (20 μL) were added to each well to achieve final concentrations of 5, 10, and 20 μM. After incubation for 0, 1, 3, 6, 9, and 12 h, the medium was removed, and the wells were washed with 0.2 mL clean medium three times to remove the prodrugs outside of the cells. DMSO (200 μL) was added to each well to digest the cells. The prodrug concentration was quantified by its fluorescence intensity using a fluorescence plate reader (SpectraMAX Gemini EM, Molecular Devices) with excitation at 605 nm and emission at 680 nm and bottom reading option. Data were obtained using SoftMaxPro software version 5.4.1.

To determine folate receptor (FR)-mediated and non-specific uptakes, additional uptake assays were performed following the same procedures with 0.5 mM FA. Cells were preincubated with 0.5 mM FA for 1 h before the prodrug was added to the wells.

2.4. Subcellular localization of PTX prodrug

Colon 26 cells were seeded at 30,000 cells/well in 24-well plates containing 12-mm diameter cover slips. Cells were then preincubated for 24 h. The filtered formulation of PTX prodrug was diluted with medium to a concentration of 500 nM and incubated for 1, 9, and 24 h. The cells were stained with LysoTracker™ Green DND-26 (Invitrogen, Carlsbad, CA) for 2 h. After the staining, the medium was removed and the cover clips were washed three times with PBS. The cells were imaged under the confocal microscope (Leica SP8 Confocal White Light Laser system). Images were captured with a 63 × (1.4 N.A.) oil objective. The LysoTracker™-stained lysosome was excited by 496 nm and 515 nm lasers. PTX prodrug was excited by 616 nm and 669 nm lasers. The images were processed with LAS X software (Leica Microsystems, Inc., Buffalo Grove, IL, USA) and ImageJ [16].

2.5. Pharmacokinetics of PTX prodrug in tumor bearing mice

All in vivo plasma PK, tissue distribution, and anti-tumor efficacy studies were approved by the Institutional Animal Care and Use Committee at University of Oklahoma Health Sciences Center. Eight-week-old female BALB/c mice (~20–22 g, Charles River Laboratories, Inc.) were used for the plasma PK and tissue distribution studies. The mice were implanted with 1 × 10⁶ colon 26 cells in PBS (100 μL) subcutaneously on the lower back neck. Once tumors reached diameters of 4–6 mm, mice received a single intravenous (IV) dose of 2 μmol/kg prodrug via retro-orbital injection. Blood samples were collected from each animal at the following time points (three mice per time point): 15 min, 0.5, 1, 3, 6, 9, 12, 24, and 48 h. Various tissues, including tumor, liver, kidney, spleen, muscle, and skin, were collected at 0.5, 1, 3, 6, 9, 12, 24, and 48 h. Samples were processed to collect serums and tissue homogenates to determine prodrug concentration using the fluorescence of the prodrug (Fig. S1). Non-compartmental PK analysis was performed using Phoenix WinNonlin 8.1 (Certara, Princeton, NJ).

2.6. Intratumoral distribution of PTX prodrug

To evaluate the prodrug distribution within the tumor, mice were euthanized 1, 9, and 48 h after prodrug (2 umol/kg) administration. Tumors were collected and embedded in Tissue-Tek® O.C.T. compound (Sakura Finetek USA, Inc.) for cryosectioning. Cryosections were incubated with combinations of markers of Alexa Fluor® 594 anti-mouse CD31 (BioLegend, Inc., San Diego, CA) and DAPI (4’,6-diamidino-2-phenylindole) buffer. The immunofluorescence samples were imaged using the Leica SP8 Confocal White Light Laser system (Leica Microsystems, Inc., Buffalo Grove, IL). The DAPI-stained nucleus was excited by 406 nm and 459 nm lasers. The Alexa Fluor® 594 anti-mouse CD31-stained vessel was excited by 511 nm and 569 nm lasers. The PTX prodrug was excited by 616 nm and 669 nm lasers. The images were processed with LAS X software (Leica Microsystems, Inc., Buffalo Grove, IL, USA) and ImageJ [16].

2.7. Quantification of released PTX in tumor and plasma

The released PTX concentration upon illumination (DLI = 9 h) was quantified in tumors and plasmas. Nine hours after prodrug administration, mice were treated at the fluence rate of 75 mW/cm² for 30 min with a collimated circular beam of 8-mm diameter using a frontal light distributor (model FD1 Medlight SA, Larges-Pièces B, 1024 Ecublens, Switzerland). Blood was drawn at pre-illumination, 5 min, 0.5,1, 3, 6, 12, and 24 h post-illumination. At the same time points, tumor biopsies were collected, rinsed with PBS, weighed, and frozen until analysis. The concentration of the released PTX in both plasma and tissue samples was determined using a previously validated assay by ultra-high performance liquid chromatography with tandem mass spectrometric detection, with a calibration range of 1–1,000 ng/mL in plasma and 5–10,000 pg/mg in 100 mg/mL tissue homogenate [15]. Briefly, 10× volume of methyl-tert-butyl ether (MTBE) was added to plasma or tissue homogenate in a liquid-liquid extraction. The mixture was vortexed and centrifuged, and the supernatant was transferred to a collection plate to be dried down under nitrogen. Samples were reconstituted with (60/40/0.1, v/v/v) water/methanol/formic acid. The plate was then vortexed and centrifuged for 5 minutes at 2,000 rpm at 4°C. Finally, the reconstituted solutions were injected onto a Waters Symmetry Shield® RP18 column (2.1 × 50 mm, 3.5 μm) before mass spectrometric detection in the positive ion mode.

2.8. QSP model of the prodrug and released PTX

A QSP model was built based on in vitro and in vivo data of the PTX prodrug and PTX either measured or collected from the literature [20]. The model scheme is shown in Fig. 2. In brief, an i.v. bolus administration of PTX prodrug is given to mice. After administration, PTX prodrug starts to distribute throughout the body. PTX prodrug circulating in the bloodstream will overcome the tumor interstitial barrier to reach cancer cells. Prodrug can be internalized into cancer cells either through FR or non-specific endocytosis. Once the tumor area is irradiated, PTX will be released in cancer cells, tumor interstitium, or tumor capillary. The released PTX may bind to microtubules or travel through the system circulation.

Figure 2.

Model scheme of (1) intracellular uptake of PTX prodrug, (2) PBPK of PTX prodrug, and (3) PBPK of released PTX. Q represents tissue blood flow rate. k_p is the tissue-specific partition coefficient for PTX prodrug. k_PTX is the tissue-specific partition coefficient for released PTX. k_int is the internalization rate of PTX prodrug. QT is influenced by vascular damage after illumination.

Tissue distribution of PTX prodrug

Based on tissue distribution profile, PTX prodrugs are likely to accumulate in well-perfused tissues, such as the liver, spleen, and kidney, where the mononuclear phagocyte system (MPS) presents. MPS can digest PTX prodrug once it is internalized. Other tissues with less MPS (e.g., muscle, G.I.) demonstrated fast removal of prodrug from tissue. In the model, we used tissue-specific partition coefficients $(K_p^{tissue})$ to describe its transportation from bloodstream to tissue. Cellular internalization rate $(k_{int}^{tissue})$ was used to describe its intracellular internalization.

Tumor intracellular trafficking

The tumor is the target for prodrug delivery. Once PTX prodrug reaches the cancer cell membrane, it can either bind to FR, which presents on the cancer cell membrane, or it can be internalized through non-specific endocytosis. The prodrug that is internalized into cells will slowly accumulate in lysosome.

Tissue distribution of released PTX

During the illumination, the linker of PTX prodrug was cleaved and PTX was released in cancer cells, tumor interstitium, and tumor capillary. We assumed a 60% release of PTX based on PTX yield. PTX that is released in cancer cells can either bind to microtubules or diffuse out of cells. PTX that is released in interstitium and capillary can either diffuse into cancer cells or travel through the systemic circulation.

Model parameter estimation and validation

The physiological parameters, including tissue weight, fractions of vascular, interstitial and cellular space in tissues, and plasma flow rates to organs, were either measured or obtained from the literature [17, 18]. Tumor- and drug-specific parameters, such as protein binding fraction, folate receptor expression level, and binding affinity, were obtained or estimated from in vitro assays. Tissue-specific partition coefficients $(K_p^{tissue})$ and cellular uptake rate constants $(k_{uptake}^{tissue})$ for tissue compartments were estimated from PK data. In the experiments, light intensity and light dose (duration of exposure) were fixed. All model fittings used the maximum likelihood algorithm in ADAPT 5 (Biomedical Simulations Resource, CA). A best model was selected based on the Akaike Information Criterion (AIC), the goodness-of-fit, weighted residual plots, and reliability of parameter estimations. For simulations, Berkeley Madonna (version 9.1.9, University of California at Berkeley, CA) was used. The detailed model equations and parameters are listed in the supplementary materials.

Model simulation

Microvascular damage that results in blood flow stasis is a frequent consequence of photodynamic therapy [19]. Vascular damage observed after PDT included enhanced vessel permeability and leakage at an early phase, and platelet aggregation followed by occlusion at a later phase [20]. Therefore, we simulated the influence of vascular damage on released PTX by altering tumor blood flow rate (Q_T). We assume that Q_T is the normal blood flow rate before illumination. During illumination, Q_T may subject to increase 2 or 3 fold, due to vasodilation. After PDT, Q_T may remain 100% if no vascular damage occurred, or it could reduce to 50%*Q_T for partial vascular damage, and 0%*Q_T for complete vascular destruction.

2.9. Antitumor efficacy of PTX prodrug in vivo

Eight-week-old male BALB/c mice (22–25 g, Charles River Laboratories, Inc.) were used for the antitumor efficacy study. The mice were implanted subcutaneously with 1 × 10⁶ colon 26 cells in PBS (100 μL) on the lower back neck. Once tumors reached 4–6 mm in diameter, mice were then randomly divided into four groups of four mice per group: control, DLI=0.5 h, DLI=9 h, and DLI=48 h. The intravenous administration of the prodrug 1 μmol/kg in the filtered formulation (100 μL) was performed retro-orbitally. Before the light illumination for each group, mice were anesthetized by intraperitoneal injection of 80 mg/kg ketamine and 6 mg/kg xylazine (Henry Schein, Inc.). Tumors were illuminated at 0.5, 9, or 48 h post dose, with a 690-nm diode laser connected with the frontal collimated light distributor at 75 mW/cm² for 30 min. The illuminating beam diameter (8–10 mm) was larger than the tumor diameter in order to cover the whole tumor area. Mice in all groups were kept in the dark without light treatment. Tumor volume (V=width² × length/2) and body weight were measured five times a week. Mice were euthanized when tumor volume reached 1000 mm³.

Results

3.1. In vitro cellular uptake and subcellular localization of PTX prodrug

We evaluated the uptake kinetics and mechanisms (FR-mediated and non-specific uptakes) of PTX prodrug. The intracellular prodrug concentration was adjusted by cell number and single cell volume. The intracellular prodrug concentration gradually increased till 12 h (Fig. 3A), a rate that is much slower than that of PTX uptake (equilibrium at ~ 3 h) [15]. The PTX prodrug was expected to be internalized by FR expressed on the cell membrane. By adding FA, the intracellular uptake of PTX prodrug was significantly reduced (Fig. 3B). Around 60–70% of prodrug was delivered into cells through FR-mediated uptake (Fig. 3C), which is consistent with our previous findings [13].

Figure 3.

In vitro uptake of PTX prodrug to colon 26 cells at concentrations of 5, 10, and 20 μM. A. Total uptake of PTX prodrug. B. Non-specific uptake of PTX prodrug in the presence of 5 mM FA. C. FR-mediated and non-specific uptake percentage at 5, 10, and 20 μM doses. D. Confocal images (64 X, zoom) of colon 26 cells after incubation with PTX prodrug (500 nM) for 1, 9, and 24 h. LysoTracker Green (50 nM) was added to each well 2 h before images were taken.

The uptake of PTX prodrug occurred mainly through FR-mediated endocytosis [21–23]. Endocytosis-mediated transport accumulates mostly in lysosomes [24]. Fig. 3D shows the confocal microscopy image of fluorescent PTX prodrug and lysosomal probe (LysoTracker Green). At later times of 9 and 24 h, PTX prodrug signals co-localized with the lysosomal signals, but there was little co-localization at the early time of 1 h, indicating that PTX prodrug was taken up via endocytosis and accumulated in lysosomes.

3.2. Pharmacokinetics of PTX prodrug in mice

PTX prodrug showed a moderate circulation half-life in blood (T_1/2=8.6 h) and restricted volume of distribution at steady state (V_ss=0.14 L/kg). The V_ss is less than that of the mice extravascular fluid (0.22 L/kg), indicating that PTX prodrug did not enter into the tissue cells of most organs [25]. However, we observed a high accumulation of PTX prodrug in certain well-perfused tissues, such as liver, spleen, and kidney tissues (Fig. 4A). The liver and spleen are the two major organs containing MPS that engulf and digest pegylated PTX prodrug. In tumors, the slow increase and long retention of prodrug was probably caused by its slow intracellular uptake via endocytosis, consistent with in vitro cellular uptake kinetics.

Figure 4.

A. Concentration-time profiles of PTX prodrug in plasma, tumors, and various tissues after a 2 μmol/kg IV dose of PTX prodrug. B. PTX prodrug concentration ratio in tumor versus plasma, plasma versus tumor, tumor versus skin, and tumor versus muscle, respectively.

We then compared the tumor/plasma, plasma/tumor, tumor/skin, and tumor/muscle PTX prodrug concentration ratio at each time point (Fig. 4B). Skin and muscle near the tumor area is likely to be damaged during irradiation if a high concentration of drugs remains at illumination time. Tumor and plasma concentration will contribute to the level of SO damage and vascular damage, respectively. Based on different levels of the tumor/plasma ratio, we selected three DLIs: 0.5, 9, and 48 h. Prodrug concentrations in tumor and plasma were 0.3 vs 22, 2.3 vs 4.7 and 1.8 vs 0.2 μM, respectively, giving tumor/plasma ratios of 0.015, 0.5, and 8.2.

3.3. Intratumoral distribution of the prodrug

It was expected that V-PDT, both V- and C-PDT, and C-PDT effects after selected DLIs would be based on tumor/plasma concentration ratio. To confirm that the ratio trend is consistent in tumor microvessel versus tumor interstitium, immunofluorescence images were taken to visualize the prodrug distribution in the colon 26 tumor (Fig. 5). Intratumoral distribution of the prodrug 1, 9, and 48 h after IV injection was investigated. Consistent with the PK data, most of the prodrug was found in the blood vessels of the tumor at 1 h, with limited distribution into extravascular spaces. At 9 and 48 h, more prodrug was seen in areas other than the blood vessels, likely in the interstitial and cellular spaces of the tumor.

Figure 5.

Representative images of intratumoral distribution of the PTX prodrug. The nuclei and blood vessels were stained with 4’,6-diamidino-2-phenylindole (DAPI, blue) and Alexa Fluor® 594 anti-mouse CD31 (red), respectively. The prodrug emitted green fluorescence. Scale bar=25 μm.

3.4. QSP model and evaluation of model performance

A QSP model was constructed to describe the PK profiles of the prodrug and predict the kinetics of the PTX released upon illumination, in tumor and plasma. Fig. 2 shows the model, which comprises (1) the intracellular uptake of PTX prodrug, (2) physiologically based pharmacokinetic model (PBPK) of PTX prodrug, and (3) PBPK of site-specifically released PTX after illumination. In vitro uptake data (Fig. 3A) were used to estimate uptake rate of the PTX prodrug (Table 1A). The current uptake model is an extension of our earlier model to include the FR-mediated cellular uptake of PTX prodrug [15]. Model fitting results of in vitro uptake are provided in Fig. S2. PK data of PTX prodrug (Fig. 4A) was used to estimate tissue-specific partition coefficients $(k_p^{tissue})$ and the cellular internalization rate $(k_{int}^{tissue})$ (Table 1B). The fitting result is shown in Fig. 4A. For released PTX, we assumed 60% release of PTX from our prodrug based on our calculation:

$$Release\%=\frac{C_{PTX}^{plasma}\cdot V_{plasma}+C_{PTX}^{tumor}\cdot V_{tumor}}{C_{Prodru{g}_{-}hv}^{tumor}\cdot V_{tumor}}$$

where $C_{PTX}^{plasma}$ and $C_{PTX}^{tumor}$ is the PTX concentration in plasma and tumor measured immediately after light irradiation. $C_{Prodru{g}_{-}hv}^{tumor}$ is the prodrug concentration remained in tumor area at the time of illumination.

Table 1A.

Parameters of the intracellular uptake model of PTX prodrug.

Parameters	Definitions	Value (CV%)	Source
fbound	Non-specific bounded/unbound ratio	0.25	Measured
ksyn (10−3 pmol/hr/106 cells)	Folate receptor synthesis rate constant	9.9	[27]
kdeg (h−1)	Folate receptor degradation rate constant	0.54	[27]
Rt(0) (nM)a	Receptor at steady state	9200	Calculated
Kd (nM)	Disassociation rate constant	47	[28–30]
Qint (μL/h/cell)	FR-mediated uptake rate of PTX prodrug	132 (5.8)	Estimated
Qin (μL/h/cell)	Non-specific uptake rate of PTX prodrug	7 (11)	Estimated
Qout (μL/h/cell)	Efflux rate of PTX prodrug	0.007 (18)	Estimated
Vsingle cell (μL)	Single cell volume	2×10−6	[31]
kg (h−1)	Cell growth rate constant	0.028	Calculated
Nss	Plateau cell number	1×107

^aParameter was originally calculated as an amount, and was converted to concentration by single cell volume.

Table 1B.

Physiological and estimated parameters of the PTX prodrug and released PTX PBPK model.

	flow rate fraction (%)a	kp (CV%)	kint (CV%)	kPTXb
muscle	0.159	0.036 (22)	N/A	N/A
skin	0.058	0.047 (60)	0.66 (32)	N/A
spleen	0.00625	1.1 (17)	2.2*10e−3 (30)	2.98
GI tract	0.1125	0.18 (13)	N/A	1.9
liver	0.161	0.73 (22)	0.07 (32)	9.71
kidney	0.091	1.08 (22)	9.4*10e−3 (46)	3.25
tumor	0.0125	2.8*10e−4 (40)		0.82
remainder	0.39975	8.1*10e−4 (42)	N/A	2.1

k_p: partition coefficient of PTX prodrug in different tissues; k_int: internalization rate of PTX prodrug in different tissues. k_ptx: partition coefficient of released PTX in different tissues; N/A: not applicable

^amice total blood flow=480 mL/h [17]

^bparameters were adopted from the literature [32]

PK data of PTX from the literature were used to estimate its partition coefficients $(k_{PTX}^{tissue})$ [26] (Table 1B). The model structure of released PTX was adopted from the literature with minor modification to describe vascular damage [17]. The fitting result is provided in Fig. S3.

3.5. Predicted PK profiles of released PTX by the QSP model

We used our developed model to predict the kinetics of released PTX in tumor and plasma. These simulated data were used to evaluate the extent of PDT-mediated vascular damage and its influence on tumor PTX retention. We implemented the anti-vascular effect of PDT on the blood flow to and from tumor (Q_T), which will remove the PTX released from the tumor to the circulation, and the remaining blood flow ranges from normal (100%), partial damage (50%), and complete shutdown (0%). Predicted PK profiles of the released PTX in the tumor and plasma with different blood flow rates are shown in Fig. 6A. The model predictions showed higher and more prolonged exposure of PTX with complete shutdown than with normal blood flow rate, suggesting that slower clearance could improve the cytotoxic effects of PTX with vascular damage. However, plasma concentration of PTX was predicted to be minimum, but with increased vascular damage, the systemic exposure of PTX was even more reduced (Fig. 6B).

Figure 6.

A. Predicted time-concentration profile of PTX release in tumor. B. Predicted time-concentration profile of PTX release in plasma. With complete shutdown, no PTX is released in plasma. C. Simulated time-concentration profile of released PTX in tumor during/post illumination. D. Simulated time-concentration profile of released PTX in plasma during/post illumination.

To validate the predictions, we determined the PK profiles of released PTX in the tumor and plasma at 5 min, 0.5, 1, 3, 6, 12, and 24 h after the illumination (Fig. 6C and D). First, the impact of vascular damage to PTX clearance from the tumor was indeed quite fast. PTX_tumor concentration did not decrease 1 h after the illumination and Q_T = ~ 0 from 1 h post illumination to 36 h (Fig. 6C). Second, there might be a rapid clearance phase of the released PTX from the tumor based on higher PTX_plasma at 5 min and 0.5 h, which might happen during the illumination (Fig. 6C). Although this rapid clearance was not captured in the PK data in the tumor, it can be assumed based on the PTX PK profile in plasma (Fig. 6D). There was minimal PTX_plasma (0.0007 nM, < 1% of the prodrug), suggesting that PTX is not released before illumination (9 h-post IV administration), at least for 9 h (Fig. S4). A spike of PTX in plasma after illumination (10 nM, 5 min post-illumination) should come from the illuminated tumor. We postulate that the rapid clearance of PTX from the tumor was due to the increased blood circulation during the illumination, caused by rapid PDT damage [33]. Third, the AUC_plasma of PTX was much smaller than the AUC_plasma of Taxol at the therapeutic dose (10 mg/kg): 35 vs. 0.66 μM·h (Fig. 6D inset). These findings strongly support our hypothesis that systemic side effects are not expected from the PTX released in our prodrug therapy.

3.6. Antitumor efficacy of PTX prodrug with different DLIs

We evaluated the antitumor efficacy of PTX prodrug with different drug-light intervals (DLIs) of 0.5, 9, and 48 h. Without illumination, the prodrug itself did not influence the tumor growth (Fig. S5). There was no sign of systemic toxicity (body weight change and abnormal behaviors) in the treated mice. Different DLIs made dramatic impacts on antitumor efficacy. The log rank test was used to evaluate the difference between the treatment and control groups. The 9-h DLI showed much better antitumor efficacy than did the 0.5-h and 48-h DLIs (Fig. 7A). The 9-h DLI resulted in rapid tumor shrinkage within 1–2 days after treatment (Fig. 7C inset) and a long-term cure for all mice up to 90 days without recurrence (Fig. 7B). In the 0.5-h DLI group, tumor growth was suppressed to unmeasurable size up to 21 days in 2 mice, but those tumors grew back. The time required for the median tumor volume to reach endpoint was 35 days, compared with 20 days for the control group (P value= 0.017), which is a statistically significantly difference. Compared with the rapid tumor shrinkage of 9-h DLI group, the tumors of 0.5-h DLI group showed very minor shrinkage at early time periods (Fig. 7C inset). In the 48-h DLI group, the tumors showed an initial gradual shrinkage in 1 to 2 days, similar to the 9-h DLI group, but tumors rapidly regrew from ~ day 7, indicating incomplete tumor damage from the treatment. The re-growing tumors in the group were donut-shaped, suggesting that cancer cells in the periphery of the initial tumor survived from the treatment and regrew. This DLI-dependent antitumor effect is consistent with our hypothesis that 9-h DLI exerts better antitumor effect, presumably through both V-PDT and C-PDT effects, based on the PK profile of the prodrug, in addition to the site-specific chemotherapeutic effect by the released PTX.

Figure 7.

A. Photographic images of the tumors from four treatment groups: 1) PTX prodrug + no hv, 2) PTX prodrug + hv (DLI=0.5 h), 3) PTX prodrug + hv (DLI= 9 h), and 4) PTX prodrug + hv (DLI= 48 h), light illumination (hv) = 690 nm, 75 mW/cm² for 30 min. Prodrug was intravenously administered once at a dose of 1 μmol/kg. B. Kaplan–Meier plot comparing survival percentages of different groups. C. Individual tumor growth curves after treatment.

Discussion

The current study is a proof-of-concept for drug efficacy optimization of our prodrug system. The developed QSP model provides a quantitative prediction about prodrug distribution and released PTX tumor retention under different treatment conditions, which helped us to understand the significance of DLIs in drug efficacy.

The PK properties of our prodrug are different from those of free PTX. The PTX prodrug has a longer plasma half-life, but restricted tissue distribution, compared with Taxol. These differences could be attributed to its physicochemical properties. While PTX is a highly lipophilic small molecule with log P of 3.96 and molecular weight (MW) of 854 g/mole, the prodrug is much larger and more hydrophilic, with logD_7.4 of 0.09 and MW of ~ 4226 g/mol due to the PEG2K moiety [13]. PTX distributed throughout the body (V_ss = 2.6 L/kg) because its lipophilicity makes it easily transport through cell membranes. However, the prodrug showed a specific preference to accumulate in organs having a mononuclear phagocyte system (MPS), such as the liver, spleen, kidney, and skin [34]. MPS is involved in the endocytic uptake of particles, which is the major mechanism for the prodrug uptake [35]. Other tissue with less MPS (muscle, G.I. tract) showed lower prodrug accumulation. The longer retention of the prodrug in the tumor area could be due to lysosomal trapping after FR-mediated endocytosis [13].

Distribution of photosensitizer (here prodrug) during the illumination is one of the major determining factors. Among the three DLIs, the DLI 9 h showed far better antitumor effect than the other two groups with DLI 0.5 or 48 h, likely due to the effective damage to both tumor vasculature and cancer cells. However, DLI can also be chosen based on specific therapeutic goal or location of tumors by adjusting other PDT parameters. For example, if muscular damage surrounding the tumor is a concern, DLI 48 h can be chosen with an escalated drug dose. PBPK model and PK data in key organs enabled us to optimize treatment conditions with a certain degree of flexibility.

In typical combination strategies, two or more therapeutic components attack common targets, such as cancer cells or tumors. The combination effects are judged as synergistic, additive, or antagonistic based on the therapeutic outcome of the combination treatment compared with the sum of outcomes of each component. However, in certain combination therapies, one component can affect the other component’s PK and/or PD. In this prodrug case, the rapid vascular damage by PDT effects slowed the clearance rate (Q) of released PTX from the tumor, a physiological modification-based synergistic effect between PDT and locally released PTX. A deeper understanding of the PK and PD of all components could provide better interpretation of the therapeutic outcome of the combination effect and further new ideas for better strategies. In addition, PBPK modeling could provide newer insights about the experimental data, as we could postulate that the brief rapid clearance of PTX mostly during the illumination was due to the enhanced circulation at the illuminated tumor [33, 36]. This effect might have been overlooked without PBPK modeling-based analysis.

There were differences in antitumor response dynamics among three groups (Fig. 7C). Tumors in the DLI 9 h group showed rapid and completed remission within 1–2 days, without regrowth for up to 90 days (equivalent to 25 years in humans) [37]. These tumors can be considered cured. Tumors in the DLI 0.5 h group showed delayed growth up to about 7 days, and then two tumors grew back. The other two tumors shrank to undetectable size, but grew back starting at 21 days later. The variation in the DLI 0.5 h group may be attributed to vascular heterogeneity. Interestingly, tumors in the DLI 48 h group showed rapid shrinkage at a rate close to that of the DLI 9 h group. However, all tumors grew back quickly from 3–6 days post-treatment. Those dynamic patterns are consistent with V-PDT and C-PDT effects. After V-PDT, no rapid shrinkage and slow recovery are expected because the vascular compartment volume in tumor is much smaller than the cancer cell compartment volume, and recovery of the vasculature could be slower than cancer cell growth [38]. After C-PDT, rapid volume reduction and fast regrowth are expected due to the rapid direct cell killing of cancer cells occupying most of the space in the tumor. Tumor recovery could be fast because of rapid proliferation of the surviving cancer cells.

The complete ablation in the DLI 9 h group could be the outcome of the combined effects of direct cell killing, vascular damage, cytotoxic effects of PTX, and, possibly, stimulated immune response. Considering the rapid reduction of tumor volume, the major damage might have been made by direct cell killing and vascular damage, causing tumor damage mostly during the illumination. The antitumor effects of the released PTX and stimulated immune responses might contribute to suppressing recurrence by killing cancer cells surviving from the early PDT effects. The concentration of the released PTX in tumor (14 nM for > 24 h) (Fig. 6B) might be able to kill cancer cells. Fourteen nM is above the general IC50 of PTX [39]. Because Balb/c mice have a complete immune system, it is likely that their immune system might be activated to damage surviving cancer cells via both innate and adaptive immune responses from the PDT damage [40, 41]. The relative contribution of the four mechanisms could depend on many factors in both treatment conditions and tumors. Understanding this contribution and establishing its best utilization are topics of our ongoing investigations.

To achieve the desired site-specific chemotherapeutic effects from stimuli-responsive systems, a sufficient AUC of released drug in target is desired. Our PBPK model could be used for other stimuli-responsive drug delivery systems [42–46] with appropriate adjustments in some key parameters, such as drug release rate and kinetics, cellular uptake kinetics, tumor clearance rate of drugs, and the impact of vascular damage on the clearance rate. Also, our PBPK model can incorporate SO generation parameters to describe different light dosimetry, intensity and exposure [47].This will expand the application of our PBPK model to evaluate PDT and chemo-damage in different conditions. The corresponding PBPK model can be used to predict the AUC of released drug and antitumor efficacy of the drug delivery system. A well-established PBPK model could be further used to find key parameters for further improvement of the drug delivery system.

In our experimental setting, we used small tumors to demonstrate our concept. Small tumors may underestimate heterogeneity and its physiological barriers to drug delivery. In bigger tumors, the heterogeneity of prodrug and photon distributions would be more significant. Some studies claimed that chemotherapy can benefit from pre-PDT treatment by overcoming delivery barriers [48, 49]. Thus, appropriate adjustments will be needed to accommodate tumor-size-specific components in the model.

Conclusion

The addition of systemic chemotherapy to PDT is a sensible approach to kill cancer cells surviving from rapid PDT damage; indeed, this approach showed improved anticancer activity [50–52]. However, the systemic administration of anticancer drug could cause systemic side effects due to unnecessary exposure of the whole body to the cytotoxic drugs. Thus, our light-activatable prodrug strategy for the combination of PDT and site-specific chemotherapy is very rational, by avoiding unnecessary exposure to systemically circulating anticancer drug at high concentrations to treat local tumors. In our previous reports, we demonstrated the superior antitumor efficacy of these prodrugs over PDT alone [10, 53]. In the present report, we used for the first time a QSP approach to optimize treatment conditions and to demonstrate the PK profiles of locally released anticancer drug PTX. In addition, we demonstrated the positive effect of the rapid pharmacodynamic effects of PDT, most likely through vascular damage, to the PK profile of PTX in tumor. Rapid PDT damage slowed the clearance rate (Q) of PDT from the tumor, enhancing its AUC in tumor. Treatment with DLI 9 h not only achieves both V-PDT and C-PDT effects, but also takes advantage of this pharmacodynamic-based synergistic effect in PDT and site-specific chemotherapy to yield outstanding tumor control. QSP was utilized here as an excellent tool to provide rational decision making for treatment optimization and an in-depth quantitative understanding of the complex dynamic PK and PD processes of the combination therapy.

Highlights.

• Pharmacokinetics and tissue distributions of the paclitaxel prodrug and released paclitaxel were characterized via developing a physiologically based PK model.

• Quantitative systems pharmacology was successfully applied to optimize treatment condition of the paclitaxel prodrug.

• The clearance of released paclitaxel in tumors was slowed by PDT-induced vascular damage.

• The maximum antitumor effect was achieved at DLI 9 hr, when the prodrug concentrations were high at both plasma and tumor, through effective vascular damage and direct cancer cell killing.

Abbreviations

PDT

photodynamic therapy

PS

photosensitizer

DLI

drug-light interval

PK

pharmacokinetics

PD

pharmacodynamics

ROS

reactive oxygen species

SO

singlet oxygen

V-PDT

anti-vascular photodynamic therapy

C-PDT

cellular-targeted photodynamic therapy

QSP

quantitative systems pharmacology

PTX

paclitaxel

FR

folate receptor

FA

folate

MPS

mononuclear phagocyte system

V_ss

volume of distribution at steady state

G.I.

gastrointestinal tract

PBPK

physiologically based pharmacokinetic model