Flow Arrest Intra-arterial Delivery of Small TAT-decorated

Flow arrest intra-arterial delivery of small TAT-decorated and neutral micelles to gliomas Abstract The cell-penetrating trans-activator of transcription (TAT) is a cationic peptide derived from human immunodeficiency virus-1. It has been used to facilitate macromolecule delivery to various cell types. This cationic peptide is capable of crossing the blood–brain barrier and therefore might be useful for enhancing the delivery of drugs that target brain tumors. Here we test the efficiency with which relatively small (20 nm) micelles can be delivered by an intra-arterial route specifically to gliomas. Utilizing the well-established method of flow-arrest intraarterial injection we compared the degree of brain tumor deposition of cationic TAT-decorated micelles versus * Shailendra Joshi sj121@cumc.columbia.edu 1 Department of Pharmaceutical Sciences, University at Buffalo, State University of New York, Buffalo, NY, USA 2 Institute for Computational Engineering and Sciences, University of Texas at Austin, Austin, TX, USA 3 Department of Molecular Cardiology, Texas Heart Institute, Houston, TX, USA 4 Department of Anesthesiology, Columbia University Medical Center, New York, NY, USA 5 Department of Neurological Surgery, Columbia University Medical Center, New York, NY, USA 6 Department of Electrical Engineering, Boston University, Boston, MA, USA 7 Department of Biomedical Engineering, Boston University, Boston, MA, USA 8 Department of Pharmacology & Therapeutics, Roswell Park Cancer Institute, Buffalo, NY, USA 9 Department of Anesthesiology, College of Physicians and Surgeons, Columbia University, 630 West 168th Street, P&S Box 46, New York, NY 10032, USA neutral micelles. Our in vivo and post-mortem analyses confirm glioma-specific deposition of both TAT-decorated and neutral micelles. Increased tumor deposition conferred by the positive charge on the TAT-decorated micelles was modest. Computational modeling suggested a decreased relevance of particle charge at the small sizes tested but not for larger particles. We conclude that continued optimization of micelles may represent a viable strategy for targeting brain tumors after intra-arterial injection. Particle size and charge are important to consider during the directed development of nanoparticles for intra-arterial delivery to brain tumors. Keywords Adjuvant therapy · Blood–brain barrier · Brain tumor · Chemotherapy · Drug delivery · Glioblastoma Introduction Intra-arterial (IA) drugs have been used to treat gliomas over the last five decades but the technology has remained largely investigational [1, 2]. In flow arrest intra-arterial (FA-IA) delivery, small boluses of drugs are injected into an artery during transient flow stoppage either using an occluding balloon catheter [3] or adenosine-induced cardiac pause [4, 5]. This approach overcomes several issues associated with conventional IA chemotherapy to the brain in which drugs are either diluted by the blood flow during infusion or are not absorbed due to rapid transit through the cerebral circulation [6–8]. In contrast, when drugs are injected during transient flow arrest, they can momentarily displace blood from the cerebral circulation [6]. This increases the regional arterial blood drug concentration and drug transit time. Blood cell and protein binding as well as hemodynamic shear stress on the delivered drug molecules greatly decrease. We have previously reported robust uptake of relatively large 80–200 nm cationic liposomes when delivered by the FA-IA method to the brain and more specifically to gliomas [9]. Increased macromolecule interaction with the cerebral vascular endothelium appeared to favor absorption of these larger cationic liposomes. However, brain tumor uptake was not always consistent possibly resulting from the large particle sizes utilized. Therefore in this study, we have endeavored to investigate the feasibility of brain tumor delivery using optimized molecules on a smaller size scale. Here we compare tumor uptake of 20 nm micellar particles decorated with the cationic human immunodeficiency virus-1 trans-activator of transcription (TAT) versus neutral micelles in a rat glioma model. Additionally, we performed simulations of a broad range of particle sizes to further inform our results. Materials and methods Micelle preparation TAT-modified cysteine (GRKKRRQRRRPQC) was added to PEG-DSPE(2000) Maleimide dispersed in phosphate buffered saline (PBS) at pH 7.4 and incubated at room temperature for 4 h under constant shaking. Unreacted TAT peptide was removed using a molecular weight cut-off column (Pierce Concentrator). Micellar carriers were prepared as described previously [10]. Briefly, DSPE-mPEG(2000) (Avanti Polar Lipids, Alabaster, AL) or a 20/80 mixture of DSPE-PEG(2000)-TAT/PEG-DSPE(2000)-amine were mixed with 1,1′-dioctadecyl-3,3,3′,3′tetramethylindodicarbocyanine, 4-chlorobenzene-sulfonate salt at a molar ratio of 98/2. The PEGylated lipid mixture was then dried to a thin PEG-lipid film using a rotary evaporator followed by rehydration of the lipid film in phosphate buffered saline (PBS) and sonication for 10 min at room temperature Fig. 1 Spectroscopy and fluorescence concentration measurements. Measurement of concentrations by diffuse reflectance spectroscopy (left) and total photon count (right) in a tissue phantom are shown. A linear correlation is seen between measured and known increasing concentrations of 20 nm TAT-decorated and neutral micelles using both techniques until fully dispersed. The particle sizes of the TAT-decorated DSPE-PEG(2000)-amine and methoxy-PEG-DSPE (mPEG-DSPE) micelles were measured by dynamic light scattering (DLS) using the Nanobrook Omni (Brookhaven Instruments Corp., Holtville, NY). To determine the effect of pH on zeta potential, the micelles were dispersed in PBS at pH 7.4 and in 20 mM of 2-(N-morpholino) ethanesulfonic acid (MES) buffer at pH 6.5, 6.0, and 5.5 [11]. Three measurements were taken per sample and represented as mean ± standard deviation. In vitro spectroscopy DiD-labeled micelles were dissolved in 10% intralipid and concentrations were compared to concentrations measured by diffuse reflectance spectroscopy (DRS) and semi-quantitative fluorescence imaging [12–15]. A custom built device (Optimum Technologies Inc., Boston, MA) employing a xenon light source to generate white light (450–850 nm) delivered to the target site via a fiber optic cable was used for DRS. In this setup, the backscattered light is relayed to a spectroscope via another cable and is subjected to spectral analysis to determine tracer and drug concentrations. For semi-quantitative fluorescence imaging, a 635 ± 5 nm LED light excitation was imaged through a 680 ± 5 nm bandpass filter using an electron multiplying charge-coupled device (EMCCD, Evolve 512, Photometrics Inc., Huntington Beach, CA). The camera corrects for background noise and can quantify the photon yield. The dissolved concentrations were compared to concentrations measured by DRS and semi-quantitative fluorescence imaging, which were closely correlated (Fig. 1). In vivo studies Institutional Review Board and Animal Care and Use Committee approvals were obtained for all studies. Rodent gliosarcoma cells (9L) were cultured to confluence in T25 flasks in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum and penicillin/streptomycin and kept humidified at 37 °C with 5% C ­ O2. The 9L tumor model has similar blood–brain barrier permeability properties to human gliomas making it a reasonable model for these IA delivery studies [16–18]. On the day of implantation, cells were trypsinized, washed, and re-suspended to 4 × 105 cells (85–95% viability) in 5 µl of basal DMEM. Stereotactic implantation was performed using a 33-G Hamilton needle 2 mm posterior, 3 mm lateral, and 3.5 mm deep to the bregma over the right hemisphere. Power analysis based on a conservative 50% effect yielded a necessary sample size of 6 animals to demonstrate a difference. Experiments were conducted on two groups of Fisher 344 rats: Group 1 rats (n = 14) had tumors growing for a mean of 15 ± 1 days prior to injection of a high dose of 0.6 ml stock micelles, and Group 2 rats (n = 8) had tumors growing for a mean of 13 ± 1 days prior to injection a lower dose of 0.2 ml stock micelles. All animals were confirmed to be asymptomatic from their brain tumor prior to IA injection of micelles. Surgical procedures were carried out under general ketamine and isoflurane anesthesia. A tail vein line, tracheostomy, and right common carotid catheter were placed, followed by ligation of all branches of the external carotid artery. The DRS probe was placed over the right parietal region in Group 1 and close to the site of tumor implantation in Group 2. To monitor cerebral blood flow (CBF), a laser Doppler probe was placed just in front of the coronal suture over the right frontal region. In addition, the inspired and exhaled gas composition, skin blood flow, electrocardiogram, pulse oxygen saturation and volume, and rectal temperature were monitored. All physiological data were recorded in real-time by Mac-Lab software (AD Instruments, Boston, MA). DRS measurements recorded tissue tracer concentrations with each micelle injection over a period of approximately 30 min. The first twelve baseline measurements were taken over 1 min, then 300 measurements were taken 500 ms apart, and finally 300 measurements were 5000 ms apart. IA drugs were injected using a pneumatic 1 ml syringe with a Parker III injector controlled by an Agilent II Signal generator (Applied Signal Inc., Sunnyvale, CA). Hypotension was induced with a bolus intravenous injection of adenosine (2–3 mg), esmolol (2–3 mg), and cold saline (4 °C, 1.5 ml). Pulsed intracarotid injections (60–70 µl) were delivered every 2 s such that micelles were injected within 30 s of peak hypotension (Fig. 2). Micelle characterization Micelle size was determined by dynamic light scattering (DLS). TAT-decorated micelles displayed a hydrodynamic diameter of 21.8 ± 6.4 nm and the micelles composed of PEG-lipid-containing non-ionizable methoxy groups displayed a hydrodynamic diameter of 21.46 ± 2.98 nm. Zeta potential measurements were used to characterize the pHdependent surface charge of the neutral micelles and TATdecorated micelles (Table 1). Simulation studies Considering the nanoparticles as passive scalars, the mass transport of nanoparticles in blood flow is governed by the linear advection-diffusion-reaction equation [19]. As detailed in previous work [20], the adhesion of blood-borne spherical nanoparticles to the vessel wall is directly influenced by the local wall shear rate S, the particle diameter dp and the probability of adhesion Pa. The probability of adhesion Pa is a measure of the strength of adhesion and can be thought of as the propensity of nanoparticles, liposomes, or micelles to stably adhere to the vessel wall overcoming the hemodynamic forces that tend to displace them. Pa is modeled as a function of nanoparticle design parameters such as size, shape and surface characteristics including charge density and the avidity between the cationic nanoparticle and anionic blood vessel wall; as well as certain physiological parameters such as the local wall shear rate and the disease state of the vessel wall. The particle adhesion model was validated against a range of spherical particle sizes and wall shear rate. Readers are referred to previously published work for further details [20, 21]. Particle adhesion is known to be strongly modulated by the deposition parameter Π defined as Π = PaSdp/2 and consequently, the probability of adhesion [20]. A MATLAB (8.4.0.150421 Release 2014b, The MathWorks Inc., Natick, Massachusetts) code was used to estimate the particle adsorption rate (deposition parameter) and probability of adhesion for a range of relevant particle sizes and wall shear rate mimicking the experimental conditions in order to enhance our understanding of the underlying physics and gain further insight into experimental observations. Data analysis Stat View 5.2 (SAS Institute, Cary NC) was used to perform ANOVA and factorial ANOVA where appropriate. Bonferroni correction was used when multiple comparisons were made. Statistical significance was set at p ≤ 0.05. Results In vitro phantom tissue experiments The dissolved concentrations were compared to concentrations measured by diffuse reflectance spectroscopy (DRS) Fig. 2  Uptake of TAT-decorated and neutral micelles by gliomas. Data from both low dose (a, b) and high dose (c, d) injections are shown. The concentration–time curves (left panel) indicate tumor surface kinetics with micelle clearance recorded for over 30 min (1800 s) after injection (0–200 s). Corresponding gross, fluorescence, and histology images are shown (right panel). The localized high intensity fluorescence signal indicates glioma-specific uptake of the micelles Table 1  pH-dependent zeta potential of TAT-decorated micelles and neutral micelles (Fig. 1). Therefore the optical pharmacokinetic strategies utilized for in vivo and post-mortem experiments provide an accurate representation of directly measured tissue concentrations. pH TAT micelles (mV) Neutral micelles (mV) 5.5 6.0 6.5 7.4 16.8 ± 7.7 13.0 ± 4.4 8.4 ± 3.0 7.1 ± 1.4 −5.9 ± 1.7 −2.6 ± 2.5 −4.8 ± 1.3 −3.6 ± 1.2 and semi-quantitative fluorescence imaging. Both diffuse reflectance spectroscopy and fluorescence measurements correlated well with the dissolved micelle concentrations In vivo experiments In vivo experiments were conducted on 22 rats split into high injection dose (0.6 ml micelles) and low injection dose (0.2 ml micelles) groups as previously described. Technical issues with carotid cannulation necessitated one animal from the high-dose group be excluded from further analysis. As expected, the high-dose injection animals had somewhat larger tumors than the low-dose injection animals since tumors were allowed to grow for longer in the high dose group. No clinical difference was observed in animals injected with neutral versus cationic TAT-decorated micelles. The hemodynamic data, heart rate, skin blood flow, and cerebral blood flow showed similar changes between TAT-decorated and neutral micelles. There was an expected decrease in hemodynamic parameters during flow arrest that subsequently recovered (Table 2). Diffuse reflectance spectroscopy Tumor deposition of both TAT-decorated and neutral micelles was noted after intra-arterial injection. Surface measurements of tracer concentrations of the two micelle formulations showed higher values for TAT-decorated micelles. This was evident in both the high and low dose animal groups. However, the observed trend toward increased uptake of TAT-decorated micelles was not statistically significant. Thus, the quantitated parameters including peak concentration, end concentration, and area under the concentration–time curve were not statistically different between the TAT-decorated micelles and the neutral ones (Table 2). Table 2 Tumor size, tissue concentrations and hemodynamics Postmortem image analysis Postmortem image analysis showed effective, consistent, and targeted uptake of both neutral and TAT-decorated micelles by the gliomas (Fig. 2). No obvious differences were noted in the deposition pattern between TAT-decorated micelles and neutral micelles. There was also deposition of both micelle types along the blood vessels, particularly around the Circle of Willis. Expected increased delivery concentration in the high injection dose group was seen. Simulations The simulations performed show how nanoparticles size and surface charge are predicted to affect regional brain deposition. A wide range of nanoparticle sizes from 10 to 500 nm for a variety of relevant hemodynamic condition in terms of wall shear rate were tested. At 80 nm or larger, there is increased deposition at low shear stresses which can be clinically achieved with flow arrest. However as the wall shear stress increases the deposition of larger particles also decreases dramatically. Cationic charge increases both the deposition rate and probability of adhesion under FA-IA for these larger particles. However, the significance High dose Neutral (n = 6) Tumor sizes Tumor volume gross image 175 ± 161 ­mm3 Tumor volume histology 102 ± 95 ­mm3 Diffuse Reflectance spectroscopy Peak Conc.(mMol) 1.91 ± 0.87 End Conc (mMol) 0.31 ± 0.20 AUC (mMol.s) 293 ± 212 Heart rate (BPM) Base 291 ± 34 Injection 135 ± 85 5 min 283 ± 27 End 287 ± 18 LD brain (%∆ baseline) Base 100 ± 0 Injection 17 ± 3 5 min 95 ± 25 End 73 ± 33 Laser Doppler skin (%∆ baseline) Base 100 ± 0 Injection 14 ± 6 5 min 108 ± 30 End 106 ± 33 Low dose TAT (n = 7) 145 ± 90 ­mm3 93 ± 59 ­mm3 Neutral (n = 4) 34 ± 35 ­mm3 41 ± 27 ­mm3 TAT (n = 4) 59 ± 33 ­mm3 42 ± 27 ­mm3 2.44 ± 1.12 0.32 ± 0.19 318 ± 175 0.44 ± 0.11 0.05 ± 0.02 37 ± 10 0.59 ± 0.23 0.15 ± 0.13 115 ± 60 260 ± 43 110 ± 45 266 ± 37 277 ± 23 222 ± 22 59 ± 12 228 ± 22 233 ± 16 238 ± 23 77 ± 16 233 ± 26 254 ± 8 100 ± 0 13 ± 5 125 ± 41 42 ± 30 100 ± 0 25 ± 27 115 ± 22 85 ± 26 100 ± 0 14 ± 27 115 ± 22 85 ± 26 100 ± 0 11 ± 3 98 ± 38 49 ± 38 100 ± 0 13 ± 7 127 ± 30 120 ± 11 100 ± 0 14 ± 7 129 ± 45 87 ± 33 $ Significant difference from baseline, # Significant difference between TAT and neutral micelles of charge lessened with small nanoparticle sizes. Indeed the computational simulations show that particle charge effects an almost negligible consequence for the deposition of 20 nm nanoparticles such as the micelles we injected (Fig. 3). Discussion Contrary to our previous findings that cationic surface charge plays a major role in regional liposome (large nanoparticle) deposition [7, 22, 23], here we detected no benefits from the application of a cationic charge to smaller 20 nm micelles. Indeed, at this particle size both neutral and cationic micellar preparations penetrated tumors effectively after intra-arterial injection. We also noted deposition along blood vessels, particularly the Circle of Willis, with both preparations. Ante-mortem measurements of tissue concentrations by DRS showed no significant difference between the two particles delivered. Fig. 3  Computational modeling IA nanoparticle delivery to the brain. Nanoparticle interaction and consequently the probability of particle adhesion to the endothelial surface are determined by a range of physical characteristics as noted in the schematic of a cerebral blood vessel (a). Using a range of experimental hemodynamic data during flow-arrest drug delivery, we calculated wall shear stress (b). Rates of particle uptake (c) and the probability of adhesion (d) to the Flow arrest drug delivery has been long used to treat cancers in organs such as the liver, breast, and eye. However, it is a relatively new method of drug delivery for the treatment of brain tumors. To date, only a single clinical case utilizing this method in a brain tumor patient has been reported [3]. In other scenarios this technique has gained some traction including transient cardiac pause with intravenous adenosine to treat high flow arteriovenous malformations and local flow arrest with temporary clipping during aneurysm, carotid, cerebral bypass and other neurovascular surgeries [4, 5, 24]. Thus there exists an ample safety record for the clinical use of transient flow arrest to the brain in humans. The routine utilization of transient flow arrest during neurovascular surgery makes a compelling case that momentary flow interruptions are both feasible and safe and could be used to improve drug delivery to the brain. The impact of flow arrest drug delivery on the pharmacokinetic paradigm of IA drug delivery is often overlooked. Brain tissue has a very high basal blood flow, which greatly vascular endothelium as a function of wall shear rate for cationic and neutral particles in sizes ranging from 10 to 500 nm are shown. These simulations show that decreased wall shear stress (<500 1/s) such as during FA-IA drug delivery will increase the probability of adhesion of larger particles. This effect is seen to be minimal for small particle sizes approximating 20 nm as noted during in vivo delivery analysis decreases the benefits of IA injections. IA bolus injections during flow arrest increase arterial blood concentration, increase drug transit time, prevent or minimize non-targeted binding with blood proteins and cells, and decrease the shear stress on particles to improve targeted uptake. Thus, flow arrest can greatly improve targeted drug delivery. We recently modeled flow arrest-IA drug delivery by elaborating the Dedrick model and simulation showed that the benefits of FA-IA drug delivery were significant when there is rapid uptake and retention by tumor issue [25–27]. Computer simulation reported here shows how decreasing shear stress facilitates the delivery of large cationic particles (>80 nm). As the local wall shear stress increases, these large particles experience greater drag force compared to smaller particles. At the same time, the larger the particle size, the greater the surface area available for adhesive interaction with the vessel wall. As a result, the attractive forces between the cationic particle and anionic vessel wall can better overcome the relatively low hydrodynamic stresses prevailing during FA-IA delivery, leading to firm particle adhesion. In this scenario, the attractive forces between the positively charged particles and the negatively charged endothelium are critical to promoting adhesion and absorption. This is relevant for particles between 80 and 500 nm in size. With decreasing particle sizes (≤20 nm), the area of interaction decreases, reducing the adhesive forces between the particle and endothelium. As a result, the probability of adhesion, and consequently the rate of deposition, is lower for smaller particles during FA-IA delivery. According to the simulations, the effect of surface charge on particle adhesion would be largest for size ranges between 80 and 500 nm. For the small 20 nm particles utilized in this set of experiments, the effects due to surface charge seemed to be minimal. Prior to conducting these insightful computational studies, we hypothesized that decorating micelles with TAT would greatly improve drug delivery due, in part, to its positive charge and result in enhanced regional drug deposition. We were initially surprised to find that this was indeed not the case with improvements in regional delivery of TAT-decorated micelles being much less robust than was previously noted in experiments utilizing larger liposomes. A posteriori, our simulation studies support the experimental data. Perhaps a better strategy for improving tumor delivery of micelles is to trap them in the acidic pH of the tumor through ionization after uptake as we have previously shown [27]. The question of whether tumor uptake efficiency is influenced b y t umor s ize w as n ot d irectly a ddressed i n this report. Although the Group 1 and Group 2 experiments had different g ross t umor v olumes, a ll s tatistical comparisons were made within each independent group experiment. This will be an important avenue for investigation in future experiments since we believe that: (1) large tumors with a complex vascular supply may show uneven, heterogeneous uptake, (2) large tumors may have higher interstitial hydrostatic pressures that could affect d rug u ptake a nd c learance a nd ( 3) l arge tumors might have more physiological variability including regions of ischemia, hemorrhage, and necrosis with microcirculatory failure all of which can affect u ptake efficiency [28, 29]. Several different m echanisms h ave b een p roposed t o explain TAT facilitated macromolecule uptake into cells [30]. This includes adsorptive endocytosis and direct fusion with the cellular membrane. The exact mechanism of TAT-decorated micelle uptake, however, remains unclear. Our ongoing studies have shown rapid uptake of TAT-FITC by cancer cells and after FA-IA delivery [31]. At the molecular level, TAT-conjugated FITC is robustly taken up by brain tumors when delivered by FA-IA injection but when TAT is conjugated to larger proteins such as GFP, TAT-GFP proteins are taken up less than the equimolar amounts of TAT and GFP when co-administered in cell cultures [32]. It seems that both the size and rigidity of the molecules might play a role in TAT-enhanced drug delivery. Micelle formulations might be taken up by brain tumors utilizing other mechanisms, such as by increased lipid solubility. Because of their surfactant-like behavior, micelles are capable of undergoing lipid transfer and exchange with cellular membranes. As a result, the membrane dye DID—which is incorporated into the micelles—may have been transferred to the cellular membrane after FA-IA delivery. This alternate mechanism of uptake is equally applicable to TAT-decorated and neutral micelles and therefore could explain the lack of an observed difference in uptake between the two formulations. We conclude that both neutral and cationic micelles may be formulated to facilitate drug delivery to gliomas by the FA-IA method. At the small particle size (20 nm) tested, cationic micelles show modest delivery superiority. Future studies are warranted in additional models and to assess the surface properties of micellar carriers to gauge their ability to affect the delivery of therapeutic cargo into glioma cells.