* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Signaled drug delivery and transport across the blood
Survey
Document related concepts
Polysubstance dependence wikipedia , lookup
Compounding wikipedia , lookup
Pharmacognosy wikipedia , lookup
Psychopharmacology wikipedia , lookup
Pharmacogenomics wikipedia , lookup
Theralizumab wikipedia , lookup
Blood–brain barrier wikipedia , lookup
Neuropsychopharmacology wikipedia , lookup
Pharmaceutical industry wikipedia , lookup
Prescription drug prices in the United States wikipedia , lookup
Prescription costs wikipedia , lookup
Drug discovery wikipedia , lookup
Drug interaction wikipedia , lookup
Plateau principle wikipedia , lookup
Drug design wikipedia , lookup
Transcript
Signaled drug delivery and transport across the blood-brain barrier Peter Hinow∗ John N. J. Reynolds§ Ami Radunskaya† Sean M. Mackay‡ Morgan Schroeder¶ Eng Wui Tan‡ Ian G. Tuckerk September 30, 2015 Abstract We use a mathematical model to describe the delivery of a drug to a specific region of the brain. The drug is carried by liposomes that can release their cargo by application of focused ultrasound. Thereupon, the drug is absorbed through the endothelial cells that line the brain capillaries and form the physiologically important blood-brain barrier. We present a compartmental model of a capillary that is able to capture the complex binding and transport processes the drug undergoes in the blood plasma and at the blood-brain barrier. We apply this model to the delivery of levodopa (L-dopa, used to treat Parkinson’s disease) and doxorubicin (an anticancer agent). The goal is to optimize the delivery of drug while at the same time minimizing possible side effects of the ultrasound. Keywords: liposomal delivery, focused ultrasound, mathematical modeling. 1 Introduction The challenge for drug delivery science is summarized by the following: to deliver the drug specifically (or at least selectively) to the target site, in the required amount, at the required rate, at the appropriate time and for the required time period. So called responsive nanocarriers loaded with drug are of interest in meeting the challenge [1]. Ideally, responsive nanocarriers are designed to release some or all their drug load when stimulated by an external trigger (e.g. ultrasound, near infrared) or by an internal ∗ Department of Mathematical Sciences, University of Wisconsin - Milwaukee, PO Box 413, Milwaukee, WI 53201, USA; Corresponding author; email: [email protected], phone: (++1) 414 229 6886 † Department of Mathematics, Pomona College, 610 N. College Ave., Claremont, CA 91711, USA ‡ Department of Chemistry, University of Otago, Dunedin, New Zealand § Department of Anatomy and The Brain Health Research Centre, University of Otago, Dunedin, New Zealand ¶ Department of Biology, University of Oregon, Eugene, OR 97403, USA k New Zealand’s National School of Pharmacy, University of Otago, Dunedin, New Zealand 1 trigger (e.g. local pH, a specific enzyme) [2]. If the external stimulus is focused at the target site for drug release, it will trigger release of drug from the responsive drugloaded nanoparticles as they are carried by the blood through that region. Sophisticated targeted delivery using magnetic resonance guided high intensity focused ultrasound (MR-HIFU) has been used for targeted release of cytotoxic drugs from heat-sensitive liposomes [3, 4]. Although it may be useful to use focused ultrasound (FUS) to trigger release of drug from circulating liposomes in a target region when the drug is required, research is required to ensure therapeutic concentrations are achieved in the tissue after application of FUS. The relationship between intravenous drug and the drug concentration in the tissue after FUS treatment may lead to adjustments to clinically prescribed dosages [5]. The dose that is delivered to the target site will depend on the liposomes (concentration, drug loading, responsiveness), the FUS characteristics, the properties of the drug, the transfer processes to the target, and processes interfering with transfer to the target. Safety considerations limit the intensity and choice of frequency of the FUS [6]. Delivery of drugs to the brain has special challenges because of the selective barrier function of the blood brain barrier (BBB). The neurovascular unit of the BBB consists of the capillary endothelial cells, astrocytes, pericytes and neurons. Because of the presence of tight junctions and efflux transporters, this structure is far less permeable than non-brain capillaries [7] although it has carriers for the uptake of selected molecules (e.g. glucose, large aromatic amino acids). This selective permeability poses a major challenge to the delivery of drugs, such as those used to treat dementia, movement disorders and brain cancers [8]. Drug released from responsive liposomes into the blood in the target region must cross the BBB. If the drug is transferred by an active process (e.g. as for L-dopa), the local high concentration may saturate the transporter compromising transfer. Conversely, a localized high concentration may be beneficial for transfer of a drug which is subject to P-glycoprotein (P-gp) efflux at the BBB, but plasma protein binding of released drug may reduce brain uptake. Further, HIFU, particularly in the presence of microbubbles, has been demonstrated to increase the permeability of the BBB [9, 10]. In this paper we describe a model of a single capillary to predict the uptake of drug into the tissue supplied by the capillary, where the drug is released from ultrasound (US)-sensitive liposomes. We anticipate that this model will form the basis of a larger model for the cerebral vasculature [11, 12]. Previous mathematical models have focused on delivery of oxygen to the brain [13, 14] and drugs in single capillaries [15]. Using a spatial advection-diffusion type model (partial differential equation), Richardson et al. [15] investigated the delivery of a drug with the help of a magnetic carrier particle. Compartmental (ordinary differential equation) models for drug transfer across the blood-brain barrier were proposed recently by Liu et al. [16] and Ball et al. [17]. In the latter paper, a full body pharmacokinetic model was constructed in order to scale results obtain in rats to humans. It is also worthwhile to mention the work of Gugat [18] who used a system of hyperbolic partial differential equations to model contaminant transport in a water distribution network. 2 Mathematical modeling has long proved its usefulness in pharmaceutical science, in particular when concerned with the understanding of drug delivery, release and transport processes; see [19, 20, 21] for some recent reviews. Our earlier work addressed predicting the release kinetics of matrix tablets [22, 23] and the modulation of the permeability of liposome membranes by binding of bile salts [24]. An important contribution of mathematical modeling is to help understand the release from the delivery device [1] and drug transport in the tissue [25, 26] where experiments are often difficult, expensive and possibly fraught with ethical problems. The mathematical tools include ordinary differential equations, partial differential equations, and cellular automata, depending on the problem at hand [21]. While it is desirable to have experimental data available to test the predictions of a mathematical model, the process of modeling and numerical simulations themselves are valuable as they often raise questions and drive future research. In this paper, we use ordinary differential equations to describe a compartmental model of blood flow in a capillary. This modeling framework is commonly used as a first step in the development of a mathematical description of a complex network describing blood flow in the brain. The differential equations of the model are derived by balancing gain and loss terms for each substance due to transport into or out of the compartment and to chemical reactions among the constituents. The system of equations is then solved numerically using open source software. In many cases, when parameters in the model are unknown, experimental data can be used to fit the model to the data [24]. Ordinary differential equations are also amenable to tools from control theory and optimization [27], which has found many applications for example in improvement of cancer chemotherapy protocols. Here we propose a compartmental ordinary differential equation model that describes the delivery of a drug to the brain from responsive liposomes. The model considers the interactions of the free drug with proteins in the plasma and the brain tissue [28] and nonlinear active transport mechanisms [17]. We focus on the transport of two drugs, L-dopa, a drug that readily crosses the BBB via a carrier [29] and doxorubicin that transfers poorly. L-dopa, a precursor of dopamine, is used to treat Parkinson’s disease. However, long term exposure to the drug can produce fluctuations that reduce the beneficial effects of the treatment [30]. It is being investigated for possible targeted delivery [31]. Doxorubicin, used to treat various cancers, is delivered in liposomes to diminish cardiac side effects. Recent work by Aryal et al. [32] and Graham et al. [33] has used FUS in the presence of microbubbles to increase the permeability of the BBB to doxorubicin thereby improving the survival of rats with implanted gliomas. Previously, Enden and Schroeder [1] studied drug release from liposomes subjected to low frequency ultrasound. Herein we take the further step to consider the subsequent transport and clearance of the drug in the vascular network serving the brain. This paper is organized as follows. In Section 2 we present a general ordinary differential equation model for the delivery of a drug to the brain by way of carrier particles and their subsequent triggered release. This model is intended as a “blueprint” for future work not only by our own group but hopefully also by other researchers. We then 3 give further details and specify this model for the two drugs L-dopa and doxorubicin in Section 3. In Section 4 we describe in vitro experiments to determine the permeability of liposomes exposed to ultrasound and we discuss the parametrization of our mathematical model. The results of this research are presented in Section 5. First we report the release of carboxyfluorescein from sonosensitive dioleoylphosphatidylethanolamine (DOPE) liposomes [34] under different ultrasound parameters. The remainder of that section contains numerical simulations of liposomal drug delivery for both L-dopa and doxorubicin. The discussion in Section 6 ends the paper. 2 The general model For the purpose of mathematical modeling, we assume that 1. the transport of the drug across the blood-brain barrier happens in a capillary network that is replaced by a single “surrogate” vessel, 2. this vessel serves a volume of brain tissue with a well-defined boundary, 3. the injection of the liposomes happened sufficiently long ago such that they are uniformly distributed in the blood, and 4. the blood vessel is entirely contained in the volume to which the ultrasound signal is applied. A schematic depiction of the ultrasound mediated drug release is given in Figure 1. Let u denote the concentration of liposomes and u0 its baseline blood concentration, which we at least initially assume to be constant. The capillary volume is sufficiently small so that the destruction of liposomes in this volume does not alter significantly their concentration in the total blood volume. The liposomes and their contents are identified as one, i.e. whether the ultrasound destroys the liposomes or merely causes them to leak is insignificant for the purpose of modeling, as only the source of drug, u, gets depleted. We denote by f the inverse of the transition time of the blood through the capillary network (in s−1 ). The differential equation for u is given by du = f (u0 − u(t)) − h(s(t))u(t), dt (1) where s(t) ≥ 0 is the time-dependent ultrasound signal. The function h is a non-negative monotone increasing function that models the effect of the ultrasound on the liposomes, which is discussed further in Section 3. If s ≡ 0, then this equation has the steady state solution u = u0 . We also take u0 as initial condition for u. Let α ≥ 1 be the number of drug molecules per liposome, which we call the “drug loading coefficient” of the liposomes. We assume that α is constant initially. Once released from the liposome carriers, the drug may bind to plasma proteins in the blood and is subsequently unable to cross the BBB [28]. Note that this loss is independent 4 cell uptake, clearance brain parenchyma passive transport active transport blood-brain barrier capillary blood flow ultrasound Figure 1: The general mechanism of ultrasound-mediated drug release and transport across the blood-brain barrier. Yellow circles represent drug-loaded liposomes, green ellipses represent plasma protein and purple circles represent the drug molecules. of the flow rate. We let k1 denote the binding rate constant of drug to plasma protein and assume that there is always enough plasma protein, so that no competition for binding occurs. Transport across the BBB can occur either by passive mechanisms due to concentration differences only, or by active transport. We denote the rate constant for the passive transport in Fick’s law by k2 > 0. The active transport occurs at a maximal rate k3 and K is the drug concentration at which the active transport occurs at half the maximal rate. Then we have the following equation for the drug concentration v in the capillary dv k3 = αh(s(t))u(t) − (f + k1 )v(t) − k2 (v(t) − w(t)) − v(t), dt K + v(t) (2) where w(t) is the concentration of free drug in the brain tissue. We assume that there is no active transport in the direction from the brain to the blood. We note, however, that active transport from brain tissue to the blood could be incorporated into future models of drugs that are subject to efflux at the blood-brain-barrier. Let V denote the ratio of the brain tissue volume to the volume of the capillary. Then we have for w dw k3 V = k2 (v(t) − w(t)) + v(t). (3) dt K + v(t) The equation for w contains no terms for metabolism or elimination of free drug from the brain. These can be added if desired and will be further discussed in Section 3. The 5 initial conditions for all variables are u(0) = u0 , v(0) = 0, w(0) = 0. The triggered delivery of drug with liposomes and ultrasound has to be compared with the direct delivery, where the drug is either injected into the bloodstream or taken orally and then absorbed. The equation for this direct delivery method dv k3 = f v0 − (f + k1 )v(t) − k2 (v(t) − w(t)) − v(t), dt K + v(t) (4) where v0 is the concentration of free drug in the blood when it arrives in the brain. The equation for the drug concentration in the brain is the same as (3). 3 Details and Special Cases: L-dopa and Doxorubicin Ultrasound can trigger the release of a drug from liposome carriers via thermal and non-thermal effects. A careful study of the effect of ultrasound on release rates from dierucoylphosphatidylcholine (DEPC) liposomes is reported in [35], where liposomes loaded with fluorescent molecules were subjected to focused ultrasound at different frequencies and intensities, for varying amounts of time. Release rates were estimated as the ultrasound parameters were varied. In Section 4 we describe a calibration experiment with a type of liposome that is sensitive to sonication below the inertial cavitation threshold. Here we report the strength of the ultrasound signal as the peak negative pressure, s. In our experiments we use a fixed ultrasound frequency of 1 M Hz, so that s is equivalent to the mechanical index. Due to potential tissue damage, it is generally agreed that a mechanical index of 1.9 should not be exceeded [6]. As a consequence of this, all in vitro ultrasound release experiments were conducted at acoustic pressures less than 1 M P a, see Figure 2 below. We assume the following about the general response of these liposomes to ultrasound. 1. There is a threshold smin , below which the lipid molecules are in a gel phase. In this phase the liposomes exhibit a minimum baseline rate of leakage. 2. For a given population of liposomes, the release rate increases linearly with ultrasound strength above this threshold. Based on these assumptions, we use the following piecewise linear function 0 if 0 ≤ s < smin h(s) = a0 + a1 (s − smin ) if smin ≤ s. (5) We therefore have to determine three parameters that will be specific to the particular liposomes used, namely the ultrasound threshold smin , the release rate below this threshold a0 , and the release response slope a1 . In order to model the background release of drug, we let the loading coefficient, α, in equation (2) satisfy the equation dα = −a0 α dt 6 (6) so that α(t) = α0 e−a0 t , where α0 is the initial drug loading coefficient. This plays a role since the duration of the treatment is comparable to the time scale of drug loss from the liposomes (see Section 5.3 for further discussion). The activation threshold smin may be lowered and the slope a1 may be increased in the presence of microbubbles. Ultrasonication of microbubbles in close proximity to cell membranes has been proposed to cause transient openings in the lipid bilayer (sonoporation) [36], possibly through inertial cavitation in some cases, although the exact mechanism is unclear. The clearance of liposomes from the blood also necessitates the addition of a differential equation describing the evolution of u0 d u0 = −k6 u0 (t) (7) dt If the permeability of the blood-brain barrier changes under the influence of ultrasound, then we introduce a differential equation d 0 if 0 ≤ s < σmin k2 (t) = −c1 k2 (t) + (8) c2 (s − σmin ) if σmin ≤ s. dt Here σmin is the ultrasound threshold that causes the opening of the blood-brain barrier. The constants c1 and c2 denote the closing and opening rates respectively, of the bloodbrain barrier. The value for k2 from this equation is then used in equations (2) and (3). Note that in absence of ultrasound, equation (8) has only the equilibrium solution 0; this is also the initial condition for k2 . In some cases it may be of interest to track the metabolism or clearance of the drug that has reached the brain. In this work, we will follow the metabolism of L-dopa which is converted to dopamine. To this end, we introduce a new variable z which is the concentration of metabolite in the brain. There are many ways in which the metabolism can be modeled. We use the simplest of these, namely first-order kinetics. Thus equation (3) is augmented and an equation for z is introduced as follows V dw k3 = k2 (v(t) − w(t)) + v(t) − V k4 w(t), dt K + v(t) dz = k4 w(t) − k5 z(t), dt (9) where k4 is the rate constant of conversion, and k5 is the clearance rate constant for the metabolite. The initial condition for z is z(0) = 0. To summarize, in the following simulations we will use equations • (1), (2), (6), (7) and (9) for L-dopa and equations • (1) - (3) and (8) for doxorubicin, with the obvious modifications. The transfer function of the ultrasound is given by (5) throughout the paper. 7 4 4.1 Materials and Methods Liposome Experiments All chemical reagents for liposome preparation were of analytical grade, obtained from Sigma-Aldrich, and used without further purification. 1,2-Distearoyl-sn-glycero3-phosphocholine (DSPC), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), and 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)2000] (ammonium salt) (DSPE-PEG2000) were obtained from Avanti Polar Lipids Inc. Milli-QTM water with a resistivity of 18 M Ω cm−1 was used in the preparation of all solutions. Sodium phosphate buffer solution (20 mmol L−1 ) was prepared with Na2 HPO4 adjusted to pH 7.4 with HCl and degassed under vacuum. Polycarbonate membranes and mini-extruder were obtained from Avanti Polar Lipids Inc. and used for the preparation of all liposomes. Fluorescence measurements were performed using a PerkinElmer LS50B luminescence spectrometer. Dynamic light scattering was conducted on a Malvern Zetasizer Nano. Sonosensitive DOPE liposomes were prepared using the lipid formulation DOPE:DSPC:DSPE-PEG2000:Cholesterol in a 25 : 27 : 8 : 40 mol% ratio as previously described by Evjen et al. [34]. Liposomes were prepared by adding chloroform solutions of DOPE (233 µL, 8 mg mL−1 ), DSPC (152 µL, 14 mg mL−1 ), DSPE-PEG2000 (321 µL, 7 mg mL−1 ), and cholesterol (774 µL, 2 mg mL−1 ) to a round-bottom flask. The solvent was removed in vacuo over one hour to produce a dry thin lipid film on the surface, which was rehydrated under sonication with 1 mL of a carboxyfluorescein (CF) solution (100 mmol L−1 CF, Na2 HPO4 buffer 20 mmol L−1 , pH 7.4) to give a final lipid concentration of 5 mmol L−1 . The lipid suspension was heated to 50◦ C, then first R extruded 11 times through 1000 nm polycarbonate membranes using an Avanti MiniExtruder, and subsequently extruded a further 11 times through a 200 nm membrane to give a unilamellar liposome suspension with an average diameter of approximately 200 nm. The unentrapped solute was removed by size exclusion chromatography using a R Sephadex G-100 gel column. The liposome suspension was collected and the phospholipid concentration was determined using the Stewart Assay [37], whereupon the concentration was adjusted to 1 mmol L−1 . To confirm that the liposomes were formed with minimal polydispersity and at the correct diameter, the suspension was measured using dynamic light scattering (DLS). Fluorescence measurements were performed using an excitation wavelength of 465 nm and an emission of 520 nm, slit widths of 2.5 nm, and an integration time of 0.5 s. The fluorescence intensity measurements were recorded in a quartz fluorescence cuvette at a scattering angle of 90◦ . The quartz cuvette was filled with 2.94 mL of pH 7.4 buffer and 60 µL of liposome suspension. The initial fluorescence intensity of the resulting suspension was monitored for a duration of 30 s, after which the cuR vette was removed, coupled to the ultrasound transducer using Aquasonic ultrasound transmission gel, and irradiated across the length of the cuvette. Ultrasound was applied at various intensities, producing acoustic pressures ranging from 0 to 0.64 M P a, 8 determined using a Precision Acoustics Ltd. 0.5 mm needle hydropohone with a 9 µm PVdF membrane. A continuous wave was applied for a pulse duration of 3 s at 1 M Hz (i.e. 3 · 106 cycles of ultrasound), after which the fluorescence intensity was measured over 30 s. This was repeated for 5 pulses. To obtain the final fluorescence intensity, 100 µL of 10% Triton-X100 was added to the suspension. The fluorescence intensity after each pulse was averaged and converted to carboxyfluorescein concentration using a standard curve, which was created by measurement of the fluorescence intensity of carboxyfluorescein at various concentrations ranging from 0 to 16 µmol L−1 . 4.2 Model Parametrization The average diameter of a liposome is d = 2 · 10−7 m and so its volume is Vlip = πd3 = 4.2 · 10−18 L. 6 The solubility of L-dopa hydrochloride in water is cLD = 5 · 10−2 mol L−1 . The number of L-dopa molecules encapsulated in one liposome by passive loading is therefore αLD = NA · Vlip · cLD = 6 · 1023 mol−1 · 4.2 · 10−18 L · 5 · 10−2 mol L−1 = 1.26 · 105 . Since commercially prescribed doxorubicin is loaded into liposomes by an “active” loading method, a 100-fold increase in the concentration of the drug in the liposome is achievable [38], we therefore work with αDXR = 1.3 · 107 . For the simulations presented in this paper, we use the baseline concentration of liposomes in a human’s blood volume, which is u0 = 1.25 · 1015 = 420 pmol L−1 , NA 5L where 1.25 · 1015 is a typical dose. Results reported in [39] suggest that a half-life of liposomes in the circulation of 24 hours is reasonable, yielding k6 = 8.3 · 10−6 s−1 . According to Bär [40], the vascular volume fraction in the rat cerebral cortex is 0.015, which is largely independent of brain size between mammals [41, Figure 3C]. Thus we work with V = 65 (which is dimensionless). Kleinfeld et al. [42, Table 1] report a mean red blood cell velocity in the capillaries of the rat cortex of 0.7 mm s−1 . Even though capillary lengths in rat brains may vary greatly, we assume a “network length” of 350 µm. Thus we obtain f = 2 s−1 . The drug is subject to plasma binding from the moment of release from the liposomes. The equilibrium values of free drug have been reported by Khor and Hsu for L-dopa [43] and by Greene et al. [44] for doxorubicin. The actual binding rate constants are difficult to measure, and those for doxorubicin are not available in the literature. Work of Huang et al. indicates that the half life time of radioactively marked L-6[18 F]Fluoro-DOPA is 30 s, [45, Figure 3], giving k1 = 0.02 s−1 . However, binding rates that have been measured vary dramatically. For example, Sevilla et al. [46] estimate values of k1 for emodin to be between 250 and 500 s−1 , while Paul et al. [47] find a much 9 K (µmol L−1 ) k3 (µmol L−1 s−1 ) 28.2 0.3 34.2 0.35 101.6 1.1 Table 1: Parameter values governing the active transport of L-dopa. slower binding rate of approximately 1.5 · 10−3 s−1 for the anti-HIV drug 1-phenylistan (1-Pl). In the drug-specific simulations we use k1 = 0.6 s−1 for L-dopa and k1 = 1 s−1 for doxorubicin. In terms of first-order kinetics, the rate of transport across the endothelial wall separating the brain from the blood compartment is given by Pθ SCblood , where Pθ is the permeability coefficient, S the surface area and Cblood the concentration of drug in the blood. It must be emphasized that this is a combination of active and passive transport which are presently indistinguishable. However, we assume that if a transporter has been identified for a drug (such as L-dopa), then active transport dominates over passive transport. Pardridge et al. [48, Table 5] report Pθ S = PS = 0.61 mL min−1 g −1 (per gram of rat brain). Pardridge et al. also report the radioactivity of [3 H]-labeled L-dopa used in their study to be 5 µCi ml−1 and the specific activity of L-[3 H]dopa as 38.9 µCi nmol−1 [48, Table 1]. The concentration in the blood can therefore be determined as µCi 1 nmol Cblood = 5 ≈ 0.13 µmol L−1 . mL 38.9 µCi Given that the experiment reported in [48] represented a very short elapsed time of 18 s, we assume that the concentration of drug in the blood during the experiment is constant, and equal to the initial concentration. We can then equate the active transport term from Equation (2) to the general uptake term dv k3 v =− ≈ −PS Cblood dt K +v and solve this for k3 . Del Amo et al. [49, Table 5], report values of K ranging from 28.2 to 101.6 µmol L−1 compiled from various experimental sources. An intermediate value of K from [29] is 34.2 µmol L−1 . Using the value of the initial concentration estimated above for Cblood = v, we can estimate k3 , using a density of brain tissue as 1.05 g cm−3 [50]. The ranges for K and k3 for L-dopa are summarized in Table 1. Parameters for the uptake of L-dopa in the brain and the subsequent metabolism and clearance of dopamine are taken from [51], Model 3. To parametrize equation (8) for doxorubicin, it should be noted that doxorubicin has a relatively small molecular weight of 543 g mol−1 . Chen and Konofagou [9] report that blood-brain barrier disruption can occur at acoustic pressures below 0.3 M P a for molecules below 3 kDa. We select σmin = 0.1 M P a. Liu et al. [10] found that focused ultrasound at 1 % duty cycle delivered for a mere 120 s resulted in accumulation of epirubicine over a period of 3 − 6 h after the event. Additionally, Hynynen et al. [52] found that the blood-brain barrier permeability, which was initially increased due to ultrasound, decreased to about 15% of its original maximal value after 3 h. We use these results to estimate values for c1 and c2 . 10 parameter u0 f α0 α V k1 k1 k3 K k4 k5 k6 value 420 pmol L−1 2 s−1 1.3 · 105 1.3 · 107 65 0.6 s−1 1 s−1 see Table 1 see Table 1 5 · 10−4 s−1 4.8 · 10−4 s−1 8.3 · 10−6 s−1 interpretation circulating liposome concentration loss due to blood flow initial drug loading coefficient, L-dopa drug loading coefficient, DXR brain volume served by capillary binding rate of L-dopa in plasma binding rate of doxorubicin in plasma active transport rate, L-dopa active transport saturation, L-dopa L-dopa metabolism rate dopamine clearance rate liposome clearance rate reference this paper [42] this paper this paper [40] see text see text [49, 48, 29] [29] [51] [51] [39] Table 2: Parameter values used in this work. Note that u0 is an initial condition, not a constant in the L-dopa simulations. Numerical simulations were done with the open source packages scilab [53] using the Euler backward scheme and in R [54] using an LSODA solver. All codes are available from the authors upon request. 5 5.1 Results Determination of the Response Function In vitro experiments were performed in which carboxyfluorescein was released from DOPE liposomes when exposed to varying intensities of ultrasound, the results of which are described in Figure 2 (left). For the concentration of carboxyfluorescein released, vCF (t), we have as in equation (2), dvCF = αh(s)u. dt In order to determine the total amount of carboxyfluorescein encapsulated within the liposomes, and therefore the theoretical maximal concentration able to be released, the liposomes were lysed with Triton-X100. Only a small amount of the encapsulated content was released during the short accumulated time of ultrasound exposure (15 s), thereby justifying the assumption that u is roughly constant in vitro for short times (not shown). Moreover we find that the terminal concentration of carboxyfluorescein is αu = 6.7 mmol L−1 . The transfer function parameters displayed in Table 3 were obtained through fitting the function in (5). Note that the unit of a1 is M P a−1 s−1 . 5.2 General Numerical Simulations To demonstrate the capabilities of our model, we begin with simulation of triggered 11 Figure 2: (Left) Release of carboxyfluorescein from DOPE liposomes at seven different ultrasound strengths with the corresponding linear fits. Release curves with an ultrasound strength below 0.45 M P a are difficult to distinguish. (Right) The slopes as functions of maximal negative ultrasound pressure with the optimal fit of the piecewise linear response function from equation (5). parameter a0 a1 smin σmin c1 c2 value 2 · 10−4 s−1 6 · 10−3 M P a−1 s−1 0.45 M P a 0.15 M P a 2 · 10−4 s−1 1 M P a−1 s−2 interpretation background release rate release rate slope lower ultrasound threshold BBB opening threshold, DXR BBB closing rate, DXR BBB opening rate, DXR reference this paper this paper this paper [9] [10, 52] [10] Table 3: Ultrasound parameter values used in this work. delivery of an “abstract” drug. The idea is that even without precise knowledge of parameter values, it is possible to obtain qualitative results that will guide further investigations. Here we consider an example where the drug is subject to active transport only and vary parameters that are associated with either the liposome formulation, namely α, or the dosage of the liposomes delivered, namely u0 . In Figure 3 the area under the curve of w is shown, when applying an ultrasound signal of 1 min duration, pressure 0.6 M P a and a total time of 2 min. We use the parameters from Tables 1 and 2, except we have divided K by 10. This simulates the case where the transporters saturate at low concentrations of drug. We observe a saturation at high drug loading coefficients and liposome loads. Moreover, the level sets of the total delivered drug coincide roughly with the hyperbolas αu0 = const. This indicates that, to a first approximation, increasing the loading capacity of the liposomes is equivalent to increasing the dose of liposomes delivered. 12 Figure 3: Area under the curve of w when α and u0 vary. Parameter values are taken from Tables 1 and 2, except that K from Table 1 was divided by 10. The level curves of equal area under the curve are roughly hyperbolas. 5.3 Delivery of L-dopa L-dopa is a drug used to treat Parkinson’s disease (PD), a condition characterized by marked deficits of dopamine in the striatum of the brain due to progressive death of dopamine-producing cells in the midbrain [55]. Dopamine itself does not cross the bloodbrain barrier, however L-dopa is a substrate of the large neutral amino acid transporters LAT1 and LAT2 and therefore readily enters the brain, particularly through the action of LAT1 [56, 49]. In dopamine neurons, L-dopa is converted to dopamine through the action of dopa decarboxylase in the surviving dopamine neurons. It is also converted in the periphery, hence is administered along with a dopa decarboxylase inhibitor to maximize the availability for brain transport and minimize peripheral side effects [57]. The plasma half-life of L-dopa is 50-90 min [58, 59, 60], and storage of dopamine is altered in the Parkinsonian brain [61], hence L-dopa is usually administered to PD patients in multiple doses over the course of a day. As the disease progresses, fluctuations in the levels of brain dopamine occur, which are thought to lead over time to the side effects of motor fluctuation and abnormal movements (dyskinesias). This has led to recent attempts to administer L-dopa and other dopamine agonists in a manner that leads to more continuous stimulation of dopamine receptors [62]. These include continuous enteral infusion of L-dopa or parenteral infusion of dopamine agonists [57]. One recent approach which has shown promise has been administration of L-dopa and analogues using liposomes. Liposomes containing L-dopa when injected systemically in a rat model of PD elevated dopamine in the striatum to much higher levels than can be achieved using equimolar systemic L-dopa [63]. In addition, targeting the liposome to the brain vasculature was utilized in order to release L-dopa specifically into areas of interest, resulting in increased dopamine levels in a mouse model of PD. This increased 13 dopamine level resulted in enhanced behavioral performance on the rotarod apparatus, compared with free L-dopa alone [31]. Thus there is evidence that maintaining the levels of dopamine in the striatum for as high and for as long as possible will improve the treatment of PD using L-dopa. In the present study, we determine if frequent release of L-dopa from liposomes using ultrasound, with an intermittent pattern of stimulation to reduce the unwanted heating effects of ultrasound, results in good achievable L-dopa levels in the brain. Figure 4 shows a simulation with a 10% duty cycle, with ultrasound pulses of width 10 ms (i.e. 10,000 cycles at 1 M Hz) and a total ultrasound delivery time of 1 s (100 pulses) over a time period of 10 seconds. The ultrasound is initiated one minute after the liposomes are injected and is stopped 10 s later, 70 s after the liposomes are injected. Figure 4 shows the results of a simulation when there is no leakage of drug from the liposomes (a0 = 0 in equation (5)). This simulation suggests that, with “ideal” liposomes that release no drug in the absence of ultrasound, small amounts of drug can be delivered to a targeted area. In the upper right panel of the figure, the concentration of free drug in the capillary, v(t), is shown. This concentration quickly drops to zero when the ultrasound is turned off. The lower panel shows the kinetics of the drug concentrations in the tissue. The concentration of free drug in tissue, w(t), has a sharp peak two seconds after the ultrasound is turned off, while the concentration of metabolized or bound drug, z(t), does not reach its peak for more than 30 minutes. If the liposomes are assumed to have a baseline leakage, the effect of very short periods of ultrasound is relatively small compared to the effect of the baseline release. In the subsequent simulations, we use longer sonication periods and include the baseline release as measured in the calibration experiment described in Section 5.1. In order to determine the effect of fractionated dosing of ultrasound on liposomal release, treatment simulations with duty cycles decreasing from 50% to 1%, and pulse widths decreasing from 1 min to 100 ms were performed, the results of which are shown in Figure 5. We note that changing the duty cycle decreases the maximum drug in the brain, but also evens out the delivery. Fractionating the doses without changing the duty cycle has little effect. In other words, it is the total amount of sonication delivered in a fixed amount of time that affects the delivery of drug to the brain. However, fractionating the ultrasound may decrease harmful thermal effects. Once calibrated, the mathematical model can be used to predict concentrations of bound or free drug when repeated doses of liposomes are administered. In Figure 6 we show two representative simulations. In these simulations we model repeated injections of liposomes that are cleared from the system. The concentration of metabolized or bound drug in tissue can be maintained within a prescribed range by specifying an ultrasound application frequency and by repeated doses of liposomes. In Figure 6 we compare two different delivery protocols. In both cases, the total sonication time is two minutes per dose. In one case, the liposome boosts are given every two hours, with 5 ms pulses of ultrasound, and in the other case the liposome boosts are given every two and a half hours, with 10 ms pulses. We see that the former delivery protocol decreases 14 v(t): free drug in plasma 5 10 µmolL−1 1.0 0 0.0 0.5 MPa 1.5 15 2.0 s(t): ultrasound intensity 60 62 64 66 68 70 60 62 64 time (s) 66 68 70 time (s) w(t) z(t) 0.000 0.005 0.010 0.015 0.020 µmolL−1 Free drug, w(t), and bound drug, z(t), in brain 0 100 200 300 400 time (min) Figure 4: (Top left) Ultrasound signal pressure during the application of focused ultrasound: 10% duty cycle, with 100 pulses of width 10 ms, separated by 90 ms with the ultrasound off. The ultrasound frequency is 1M Hz, and assuming a peak rarefactional pressure of 1.5 M P a, the mechanical index is 1.5. (Top right) Concentration of free L-dopa in plasma during the same time period. For this simulation, we assume that there is no baseline release from the liposomes: a0 = 0 in equation (5). (Bottom) The concentration of free drug in the brain, w(t), peaks relatively quickly at 72 seconds. The concentration of drug taken up and metabolized, z(t), peaks much later at approximately 35 min. All parameters used in this simulation, with the exception of a0 , are shown in Tables 1 and 2. 15 1.5 2 3 50% duty cycle 10% duty cycle 1% duty cycle 0 1 z(t) µmolL−1 1.0 0.5 0.0 s(t) (MPa ) 50% duty cycle 10% duty cycle 1% duty cycle 0 50 100 150 200 0 200 time (min) 400 600 800 2.5 time (min) 1.5 1.0 0.5 0.0 pulse width 1 min pulse width 10 s pulse width 1 s pulse width 0.1 s z(t) µmolL−1 2.0 pulse width 1 min pulse width 10 s pulse width 1 s pulse width 0.1 s 1 2 3 4 0 200 400 600 800 time (min) time (min) Figure 5: (Top) A comparison of three duty cycles, 50%, 10% and 1%. Note that the graph does not show the full duration of delivery at 1% duty cycle. In all three cases, the pulse width is 1 min and total sonication time is 8 min. The pulse repetition frequency is 0.5, 0.1 and 0.01 min−1 , respectively. Lowering the duty cycle decreases the total amount of drug delivered to the brain, and also flattens out the concentration profile of drug in the brain. (Bottom) A comparison of four pulse-widths. In all cases, the duty cycle is 10% and the total sonication time is 8 min (the full signal is not shown). The fractionation of the ultrasound pulses has very little effect on z(t), the concentration of bound drug in the brain. 16 1.5 1.0 0.5 z(t) (µmolL−1) 0.0 10 ms pulses, injections every 2.5 hours 5 ms pulses, injections every 2 hours 0 5 10 15 20 time (h) Figure 6: A comparison of two different protocols for the administration of repeated doses of liposomes and ultrasound. In both cases the total sonication time is 2 minutes per liposome dose, and the duty cycle is 2%. If liposomes are re-injected every two hours, the concentration of bound drug in the brain tissue, z(t), can be maintained between 0.91 and 1.25 µmolL−1 . As the time between injections is increased, the pulsewidth of the ultrasound must be increased in order to maintain similar concentrations of bound drug. the amplitude of the fluctuations in concentration. The simulation results in Figure 3 suggest that the mass of drug delivered to the brain depends largely only on the total dose (drug in liposomes × number of liposomes) of drug, and not on the concentration of drug in the liposomes. This can be tested by making batches of liposomes with two different drug loadings (low and high) and administering doses of liposomes such that the total mass of drug is the same, then measuring brain concentrations over time, in the presence and absence of ultrasound. This can be done at different ultrasound intensities and duty cycles (Figure 5) to test this aspect of the model. It could be done with an actively absorbed drug first and if an effect is seen then look at a passively absorbed drug to test if the effect is due to localized “saturation” of the carrier. The model considers the issue of the kinetics of drug binding to plasma proteins. This is an issue which is usually considered to be of no importance since free drug in plasma is in equilibrium with protein bound drug, within the time frames of standard pharmacokinetic processes. However, in the case of triggered drug release, the drug 17 Figure 7: Uptake of L-dopa after direct delivery to the blood, together with data points from [45]. molecules move from a protein-free environment within liposomes into plasma during flow through a capillary. Thus the relative kinetics of protein binding, permeation (active and passive) across the endothelium, and blood flow in the capillary become important. More research is required to measure the kinetics of protein binding before this aspect of the model can be tested experimentally. When delivered directly into the bloodstream, the drug will immediately bind to the plasma protein and therefore will only achieve a much lower concentration in the brain vasculature. In Figure 7 we present simulations of the system of equations (4) and (3). Huang et al. [45] show the uptake of radioactively labeled L-6-[18 F]Fluoro-DOPA in the striatum and cerebellum after an intravenous bolus injection. They show that after 20 − 30 min, the maximal activity, corresponding to maximal concentration, has been reached. We obtain a good qualitative agreement with this observation. However, we have not attempted to match this too closely since we do not have enough information how these experiments were carried out, nor do we have a calibration curve between concentration and radioactivity. 5.4 Delivery of Doxorubicin Doxorubicin and its derivatives are anticancer drugs that interfere with DNA replication and thereby lead to cell death. Their adverse side effects include suppression of white blood cell production in the bone marrow and cardiomyopathy, possibly leading to heart failure. Over the past years, the delivery of doxorubicin encapsulated in liposomes, known as Doxil, has become increasingly common. In the case of brain cancer treatment, there are no known active transporters for doxorubicin to cross the BBB, making the treatment of brain tumors with this drug particularly difficult. Recent works have 18 shown that the application of focused ultrasound in combination with microbubbles can locally and temporally increase the permeability of the BBB, increasing the potential for treating malignant gliomas [9, 32, 64]. Aryal and collaborators [32] report experimental work in which rats implanted with gliomas showed improved survival when treated with liposomal doxorubicin and focused ultrasound. In fact, only rats in the group treated with both doxorubicin and ultrasound survived, compared with those treated with the drug or ultrasound alone, or those who received no treatment. In these experiments, ultrasound is delivered in the form of brief pulses that are repeated over a few seconds, in order to avoid overheating the brain tissue. Mathematically, we can model the effect of ultrasound on the temperature of nearby tissue using an equation similar to that describing the change of permeability of the blood-brain barrier, Equation (8). The change in the permeability constant, k2 , can therefore be used as a surrogate to monitor thermal effects in an attempt to avoid damage to nearby tissue. In Figure 8, bursts of ultrasound (20 s on, 10 min apart) are simulated. The rapid opening and slow restoration of the blood-brain barrier may result in a reverse transport of drug from the brain to the blood, if the concentration of drug becomes higher in the brain. Hence, at steady state, the concentrations in both compartments are equal. 6 Discussion The application of liposomes to the site-specific delivery of therapeutic agents has grown over the past few decades. Recently, liposomes have also been applied to the treatment of neurological disorders, such as Parkinson’s disease and glioblastoma multiform, where systemic administration of the required therapeutic agent is either ineffective, or results in debilitating side effects. Even with the advances in targeted delivery, due to the complexity of the in vivo experiments required to assess drug delivery to the brain, precise measurements of the drug concentrations delivered are scarcely reported in the literature. With the growing success of mathematical modeling in pharmaceutical applications, it is our goal to complement the complex nature of in vivo animal drug delivery and disease models with predictive mathematical modeling tools for the optimization of delivery technologies. In regards to ultrasound assisted drug delivery, the relative simplicity of conducting in silico experiments compared with their in vivo counterparts allows for easy, time efficient optimization of the ultrasound parameters required to achieve successful and meaningful in vitro and in vivo results. Our present physiologically-based compartmental model has the advantage of great conceptual simplicity, to which a user may add gain or loss terms, or even additional variables to fit a given situation of interest. The numerical implementation of the model is equally simple, as many libraries of ordinary differential equations exist and can be implemented in Open Source software. As an example of a drug delivery system sensitive to ultrasound, the response function of liposomes containing 25% DOPE (DOPE25) to ultrasound was determined under relatively low acoustic pressures. DOPE25 was used as DOPE has previously been re- 19 Figure 8: Concentrations of doxorubicin in the blood, in the brain and permeability of the blood-brain barrier. The ultrasound signal s is plotted in red in the fourth panel, together with the threshold σmin for opening of the BBB (dashed line). ported to act as a sonosensitizing agent in liposome formulations, due to its cone-shaped geometry and subsequent non-bilayer forming nature [65]. A clear threshold is evident for the acoustic pressure required in order for the liposomes to release their encapsulated solute in response to ultrasound, see Figure 2. This threshold likely coincides with the energy required to induce a lamellar to reverse hexagonal phase transition, which is thought to be the mechanism behind the sonosensitivity of DOPE [65]. The release of L-dopa and doxorubicin from liposomes and subsequent delivery to the brain has also been simulated. The model has been carefully parameterized from available literature. Furthermore, L-dopa and doxorubicin highlight the differences in transport mechanisms across the BBB; L-dopa relies on active transport and readily crosses the blood-brain barrier whereas doxorubicin relies solely on passive diffusion and therefore does not ordinarily cross. Substrate-receptor binding rate constants are scarcely reported in the literature, likely due to the experimental difficulty in obtaining them. Therefore in cases where there were no direct reports of the constants required for the model, estimates were made from studies reporting similar parameters. 20 We have demonstrated a simple compartmental model of drug delivery to the brain utilizing sonosensitive liposomes as a delivery vehicle. The present work provides a mathematical setting that is amenable to the optimization of ultrasound-mediated drug delivery under a range of optimality criteria. In general, overheating of sensitive brain tissue should be avoided to minimize tissue damage and achieve a successful therapeutic outcome. Optimization to achieve this is possible by employing a total energy constraint (the area under the curve of s(t)), or by a “reporter variable” such as temperature (for which k2 was taken as a surrogate). The resulting simulations suggest that lowering the duty cycle of the ultrasound applied decreases the total amount of drug delivered to the brain, whilst fractionation of the ultrasound pulses has very little effect on the concentrations of free (w(t)) and bound (z(t)) drug in the brain. In the case of L-dopa, the goal is to achieve a relatively constant level of drug or its metabolites in the brain, for which the model suggests that ultrasound should be applied at low duty cycles over a long time period. However this may be difficult to achieve in a clinical environment due to the mechanical constraints of delivering this level of ultrasound over prolonged periods. The criteria for the design of an optimal liposome formulation is also evident for the delivery of a therapeutic agent on demand in response to ultrasound. Ideally, sonosensitive liposome should possess a high drug loading capacity, thermal stability, a low ultrasound threshold to induce release, and minimal passive leakage below this threshold. In future work, experimental validation of the model will be undertaken using the case of L-dopa delivery to the brain and its subsequent conversion to dopamine, by measuring these enhanced dopamine levels in vivo. The physiologically-based compartmental model also does not yet take into account the spatial extent of the brain region at which the drug is released. Introduction of a spatial component requires a formulation of the model using partial differential equations, thereby greatly increasing the mathematical complexity; a similar example of which can be seen in work by Gugat [18]. Nevertheless, numeral simulations of flow through networks are available and may be beneficial, making it possible to account for the dynamical regulation of blood flow to different regions of the brain depending on the local demand for oxygen and nutrients. Conflict of interests The authors declare no conflict of interest. Acknowledgments We are grateful for financial support from the grant “Collaborative Research: Predicting the Release Kinetics of Matrix Tablets” (DMS 1016214 to Peter Hinow and DMS 1016136 to Ami Radunskaya) of the National Science Foundation of the United States of America. Morgan Schroeder received an award from the Support for Undergraduate Research Fellows (SURF) program at the University of Wisconsin - Milwaukee. John Reynolds received a Rutherford Discovery Fellowship from the Royal Society of New 21 Zealand. Sean Mackay received a doctoral scholarship from the University of Otago. We acknowledge the kind hospitality of the University of Otago during several collaborative visits and during a sabbatical stay of Ami Radunskaya. Peter Hinow was partly supported from the Simons Foundation grant “Collaboration on Mathematical Biology” during a visit to Pomona College. The funding bodies had no influence on study design, collection, analysis and interpretation of data. References [1] G. Enden, A. Schroeder. A mathematical model of drug release from liposomes by low frequency ultrasound. Ann. Biomed. Eng., 37:2640–2645, 2009. [2] S. Mura, J. Nicolas, P. Couvreur. Stimuli-responsive nanocarriers for drug delivery. Nat. Mater., 12:1–9, 2013. [3] H. Grüll, S. Langereis. Hyperthermia-triggered drug delivery from temperaturesensitive liposomes using MRI-guided high intensity focused ultrasound. J. Control. Release, 161:317–327, 2012. [4] R. Staruch, R. Chopra, K. Hynynen. Localised drug release using MRI-controlled focused ultrasound hyperthermia. Int. J. Hyperthermia, 27:391–398, 2011. [5] A. Burgess, K. Hynynen. Drug delivery across the blood-brain barrier using focused ultrasound. Expert Opin. Drug Del., 11:711–721, 2014. [6] G. ter Haar. Ultrasound bioeffects and safety. Medicine), 224:363–373, 2010. P. I. Mech. Eng. H (J. Eng [7] R. Cecchelli, V. Berezowski, S. Lundquist, M. Culot, M. Renftel, M.-P. Dehouck, L. Fenart. Modelling of the blood-brain barrier in drug discovery and development. Nat. Rev. Drug Discov., 6:650–661, 2007. [8] I. Tucker, L. Yang, H. Mujoo. Delivery of drugs to the brain via the blood brain barrier using colloidal carriers. J. Microencapsul., 29:475–486, 2012. [9] H. Chen, E. E. Konofagou. The size of blood-brain barrier opening induced by focused ultrasound is dictated by the acoustic pressure. J. Cerebral Blood Flow Metab., 34:1197–1204, 2014. [10] H.-L. Liu, M.-Y. Hua, H.-W. Yang, C.-Y. Huang, P.-C. Chu, J.-S. Wu, I.-C. Tseng, J.-J. Wang, T.-C. Yeh, P.-Y. Chen, K.-C. Wei. Magnetic resonance monitoring of focused ultrasound/magnetic nanoparticle targeting delivery of therapeutic agents to the brain. Proc. Natl. Acad. Sci. USA, 107:15205–15210, 2010. [11] T. David, T. van Kempen, H. Huang, P. Wilson. The geometry and dynamics of binary trees. Math. Comput. Simulat., 81:1464–1481, 2011. 22 [12] N. Safaeian, M. Sellier, T. David. A computational model of hemodynamic parameters in cortical capillary networks. J. Theor. Biol., 271:145–156, 2011. [13] T. Hayashi, H. Watabe, N. Kudomi, K. M. Kim, J.-I. Enmi, K. Hayashida, H. Iida. A theoretical model of oxygen delivery and metabolism for physiologic interpretation of quantitative cerebral blood flow and metabolic rate of oxygen. J. Cerebral Blood Flow Metab., 23:1314–1323, 2003. [14] S. N. Jespersen, L. Østergaard. The roles of cerebral blood flow, capillary transit time heterogeneity, and oxygen tension in brain oxygenation and metabolism. J. Cerebral Blood Flow Metab., 32:264–277, 2012. [15] G. Richardson, K. Kaouri, H. M. Byrne. Particle trapping by an external body force in the limit of large Peclet number: applications to magnetic targeting in the blood flow. Eur. J. Appl. Math., 21:77–107, 2010. [16] K. Liu, B. J. Smith, C. Chen, E. Callegari, S. L. Becker, X. Chen, J. Cianfrogna, A. C. Doran, S. D. Doran, J. P. Gibbs, N. Hosea, J. Liu, F. R. Nelson, M. A. Szewc, J. Van Deusen. Use of a physiologically based pharmacokinetic model to study the time to reach brain equilibrium: An experimental analysis of the role of blood-brain barrier permeability, plasma protein binding, and brain tissue binding. J. Pharmaco. Exp. Therap., 313:1254–1262, 2005. [17] K. Ball, F. Bouzom, J.-M. Scherrmann, B. Walther, X. Declèves. Development of a physiologically based pharmacokinetic model for the rat central nervous system and determination of an in vitro-in vivo scaling methodology for the bloodbrain barrier permeability of two transporter substrates, morphine and oxycodone. J. Pharm. Sci., 101:4277–4292, 2012. [18] M. Gugat. Contamination source determination in water distribution networks. SIAM J. Appl. Math., 72:1772–1791, 2012. [19] P. H. van der Graaf, N. Benson. Systems pharmacology: bridging systems biology and pharmacokinetics-pharmacodynamics (pkpd) in drug discovery and development. Pharm. Res., 28:1460–1464, 2011. [20] N. A. Peppas, B. Narasimhan. Mathematical models in drug delivery: How modeling has shaped the way we design new drug delivery systems. J. Control. Release, 190:75–81, 2014. [21] P. Hinow, A. E. Radunskaya. The Mathematics of Drug Delivery. In A. Eladdadi, P. S. Kim, D. Mallet, editor, Mathematical Modeling of Tumor-Immune System Dynamics, pages 109–123, Berlin, Heidelberg, New York, 2014. Springer Verlag. [22] B. Baeumer, L. Chatterjee, P. Hinow, T. Rades, A. Radunskaya, I. Tucker. Predicting the drug release kinetics of matrix tablets. Discr. Contin. Dyn. Sys. B, 12:261–277, 2009. 23 [23] E. Buchla, P. Hinow, A. Nájera, A. Radunskaya. Swallowing a cellular automaton pill: predicting drug release from a matrix tablet. Simulation, 90:227–237, 2014. [24] P. Hinow, A. Radunskaya, I. Tucker, L. Yang. Kinetics of bile salt binding to liposomes revealed by carboxyfluorescein release and mathematical modeling. J. Liposome Res., 22:237–244, 2012. [25] M. Kim, R. Gillies, K. A. Rejniak. Current advances in mathematical modeling of anti-cancer drug penetration into tumor tissues. Front. Oncol., 3:278, 2013. [26] K. A. Rejniak, V. Estrella, T. Chen, A. S. Cohen, M. C. Lloyd, D. L. Morse. The role of tumor tissue architecture in treatment penetration and efficacy: an integrative study. Front. Oncol., 3:111, 2013. [27] U. Ledzewicz, H. Schättler. Geometric Optimal Control - Theory, Methods and Examples, volume 38 of Interdisciplinary Applied Mathematics. Springer Verlag, Berlin, Heidelberg, New York, 2012. [28] R. V. Smith, R. B. Velagapudi, A. M. McLean, R. E. Wilcox. Interactions of apomorphine with serum and tissue proteins. J. Med. Chem., 28:613–620, 1985. [29] H. Uchino, Y. Kanai, D. K. Kim, M. F. Wempe, A. Chairoungdua, E. Morimoto, M. W. Anders, H. Endou. Transport of amino acid-related compounds mediated by l-type amino acid transporter 1 (LAT1): Insights into the mechanisms of substrate recognition. Mol. Pharmacol., 61:729–737, 2002. [30] B. Thanvi, N. Lo, T. Robinson. Levodopa-induced dyskinesia in Parkinson’s disease: clinical features, pathogenesis, prevention and treatment. Postgrad. Med. J., 83:384–388, 2007. [31] Y. Xiang, Q. Wu, L. Liang, X. Wang, J. Wang, X. Zhang, X. Pu, Q. Zhang. Chlorotoxin-modified stealth liposomes encapsulating levodopa for the targeting delivery against the parkinson’s disease in the MPTP-induced mice model. J. Drug Target., 20:67–75, 2012. [32] M. Aryal, N. Vykhodtseva, Y.-Z. Zhang, J. Park, N. McDannold. Multiple treatments with liposomal doxorubicin and ultrasound-induced disruption of bloodtumor and blood-brain barriers improve outcomes in a rat glioma model. J. Control. Release, 169:103–111, 2013. [33] S. M. Graham, R. Carlisle, J. J. Choi, M. Stevenson, A. R. Shah, R. S. Myers, K. Fisher, M.-B. Peregrino, L. Seymour, C. C. Coussios. Inertial cavitation to non-invasively trigger and monitor intratumoral release of drug from intravenously delivered liposomes. J. Control. Release, 178:101–107, 2014. [34] T. J. Evjen, E. Hagtvet, E. A. Nilssen, M. Brandl, S. L. Fossheim. Sonosensitive dioleoylphosphatidylethanolamine-containing liposomes with prolonged blood circulation time of doxorubicin. Eur. J. Pharm. Sci., 43:318–324, 2011. 24 [35] M. Afadzi, C. de L. Davies, Y. H. Hansen, T. Johansen, Ø. K. Standal, R. Hansen, S. Måsøy, E. A. Nilssen, B. Angelsen. Effect of ultrasound parameters on the release of liposomal calcein. Ultrasound Medi. Biol., 38:476–486, 2012. [36] S. P. Wrenn, S. M. Dicker, E. F. Small, N. R. Dan, M. Mleczko, G. Schmitz, P. A. Lewin. Bursting bubbles and bilayers. Theranostics, 2:12, 2012. [37] J. C. Stewart. Colorimetric determination of phospholipids with ammonium ferrothiocyanate. Anal. Biochem., 104:10–14, 1980. [38] G. Haran, R. Cohen, L. K. Bar, Y. Barenholz. Transmembrane ammonium sulfate gradients in liposomes produce efficient and stable entrapment of amphipathic weak bases. BBA Biomembranes, 1151:201–215, 1993. [39] T. Ishida, H. Harashima, H. Kiwada. Liposome clearance. Bioscience Rep., 22:197– 224, 2002. [40] Th. Bär. The Vascular System of the Cerebral Cortex, volume 59 of Advances in Anatomy, Embryology and Cell Biology. Springer Verlag, Berlin, Heidelberg, New York, 1980. [41] J. Karbowski. Scaling of brain metabolism and blood flow in relation to capillary and neural scaling. PLoS One, 6:e26709, 2011. [42] D. Kleinfeld, P. P. Mitra, F. Helmchen, W. Denk. Fluctuations and stimulusinduced changes in blood flow observed in individual capillaries in layers 2 through 4 of rat neocortex. Proc. Natl. Acad. Sci. USA, 95:15741–15746, 1998. [43] S. P. Khor, A. Hsu. The pharmacokinetics and pharmacodynamics of levodopa in the treatment of Parkinson’s disease. Curr. Clin. Pharmacol., 2:234–243, 2007. [44] R. F. Greene, J. M. Collins, J. F. Jenkins, J. L. Speyer, C. E. Myers. Plasma pharmacokinetics of adriamycin and adriamycinol: Implications for the design of in vitro experiments and treatment protocols. Cancer Res., 43:3417–3421, 1983. [45] S.-C. Huang, D.-C. Yu, J. R. Barrio, S. Grafton, W. P. Melega, J. M. Hoffman, N. Satyamurthy, J. C. Mazziotta, M. E. Phelps. Kinetics and modeling of L-6[18 F]Fluoro-DOPA in human positron emission tomographic studies. J. Cerebral Blood Flow Metab., 11:898–913, 1991. [46] P. Sevilla, J. M. Rivas, F. Garcı́a-Blanco, J. V. Garcı́a-Ramos, S. Sánchez-Cortés. Identification of the antitumoral drug emodin binding sites in bovine serum albumin by spectroscopic methods. Biochim. Biophys. Acta, 1774:1359–1369, 2007. [47] B. K. Paul, D. Ray, N. Guchlait. Unraveling the binding interaction and kinetics of a prospective anti-HIV drug with a model transport protein: results and challenges. Phys. Chem. Chem. Phys., 15:1275–1287, 2013. 25 [48] W. M. Pardridge, D. Tiguero, J. Yang, A. C. Pasquale. Comparison of in vitro and in vivo models of drug transcytosis through the blood-brain barrier. J. Pharmaco. Exp. Therap., 253:884–891, 1990. [49] E. M. del Amo, A. Urtti, M. Yliperttula. Pharmacokinetic role of L-type amino acid transporters LAT1 and LAT2. Eur. J. Pharm. Sci., 35:161–174, 2008. [50] T. W. Barber, J. A. Brockway, L. S. Higgins. The density of tissues in and about the head. Acta Neurol. Scand., 46:85–92, 1970. [51] P. Cumming, A. Gjedde. Compartmental analysis of dopa decarboxylation in living brain from dynamic positron emission tomograms. Synapse, 29:37–61, 1998. [52] K. Hynynen, N. McDannold, N. A. Sheikov, F. A. Jolesz, N. Vykhodtseva. Local and reversible blood–brain barrier disruption by noninvasive focused ultrasound at frequencies suitable for trans-skull sonications. NeuroImage, 24:12–20, 2005. [53] Scilab Enterprises. Scilab: Free and Open Source software for numerical computation. Scilab Enterprises, Orsay, France, 2012. www.scilab.org. [54] R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, 2014. www.R-project.org. [55] H. Bernheimer, W. Birkmayer, O. Hornykiewicz, K. Jellinger, F. Seitelberger. Brain dopamine and the syndromes of parkinson and huntington. clinical, morphological and neurochemical correlations. J. Neurol. Sci., 20(4):415–455, 1973. [56] D. M. Killian, P. J. Chikhale. Predominant functional activity of the large, neutral amino acid transporter (LAT1) isoform at the cerebrovasculature. Neurosci. Lett., 306:1–4, 2001. [57] T. H. Johnston, S. H. Fox, J. M. Brotchie. Advances in the delivery of treatments for Parkinson’s disease. Expert Opin. Drug Del., 2:1059–1073, 2005. [58] G. Fabbrini, M. M. Mouradian, J. L. Juncos, J. Schlegel, E. Mohr, T. N. Chase. Motor fluctuations in Parkinson’s disease: central pathophysiological mechanisms, Part I. Ann. Neurol., 24:366–371, 1988. [59] W. Olanow, A. H. Schapira, O. Rascol. Continuous dopamine-receptor stimulation in early Parkinson’s disease. Trends Neurosci., 23(10 Suppl):S117–126, 2000. [60] K. S. Kang, Y. Wen, N. Yamabe, M. Fukui, S. C. Bishop, B. T. Zhu. Dual beneficial effects of (-)epigallocatechin-3-gallate on levodopa methylation and hippocampal neurodegeneration: in vitro and in vivo studies. PLoS One, 5:e11951, 2010. [61] J. Lotharius, P. Brundin. Pathogenesis of Parkinson’s disease: dopamine, vesicles and alpha-synuclein. Nat. Rev. Neurosci., 3(12):932–942, 2002. 26 [62] T. N. Chase. The significance of continuous dopaminergic stimulation in the treatment of Parkinson’s disease. Drugs, 55(Suppl 1):1–9, 1998. [63] A. Di Stefano, M. Carafa, P. Sozio, F. Pinnen, D. Braghiroli, G. Orlando, G. Cannazza, M. Ricciutelli, C. Marianecci, E. Santucci. Evaluation of rat striatal Ldopa and DA concentration after intraperitoneal administration of L-dopa produgs in liposomal formulations. J. Control. Release, 99:293–300, 2004. [64] L. H. Treat, N. McDannold, N. Vykhodtseva, Y. Zhang, K. Tam, K. Hynynen. Targeted delivery of doxorubicin to the rat brain at therapeutic levels using MRIguided focused ultrasound. Int. J. Cancer, 121:901–907, 2007. [65] T. J. Evjen, E. A. Nilssen, S. Barnert, R. Schubert, M. Brandl, S. L. Fossheim. Ultrasound-mediated destabilization and drug release from liposomes comprising dioleoylphosphatidylethanolamine. Eur. J. Pharm. Sci., 42:380–386, 2011. 27