Background
Progress towards the effective clinical treatment of malignant gliomas has been hampered due to ineffective drug delivery across the blood-brain tumor barrier (BBTB), in addition to the inability to simultaneously image drug permeation through tumor tissue [
1‐
3]. The current paradigm for treating malignant gliomas is the placement of implantable 1,3-bis (2-chloroethyl)-1-nitrosourea (BCNU, also called carmustine) wafers in the tumor resection cavity followed by administration of oral temozolomide, an alkylating agent, with concurrent radiation [
4‐
7]. BCNU, a low molecular weight nitrosourea, is able to cross the BBTB, but is unable to accumulate within malignant glioma cells at therapeutic levels due to a short blood half-life [
8]. Intra-operative placement of polymeric wafers impregnated with BCNU along the tumor resection cavity has resulted in improved patient outcomes, and significantly decreased toxicity compared to that associated with intravenous BCNU treatment [
9,
10]. Since this local method of BCNU delivery circumvents the BBTB and allows for sustained release of BCNU from the polymer, there are higher steady-state BCNU concentrations within the tumor resection cavity[
11]. However, a major limitation of this delivery method is that the placement of the BCNU polymer wafers may only be performed at the time of initial tumor resection [
12]. Temozolomide, like BCNU, has a low molecular weight and a short blood half-life which limits its ability to accumulate within malignant glioma cells [
5,
13].
The sizes of traditional chemotherapeutics, such as BCNU and temozolomide, are commonly reported as particle molecular weights since these particles are usually smaller than 1 nm in diameter [
13]. In contrast, the sizes of nanoparticle-based therapeutics are commonly reported as particle diameters since these particles usually range between 1 and 200 nm in diameter [
14,
15]. Particle shapes and sizes determine how effectively particles can be filtered by the kidneys [
16‐
18]. Spherical nanoparticles smaller than 5 to 6 nm and weighing less than 30 to 40 kD are efficiently filtered by the kidneys [
17]. Spherical nanoparticles that are larger and heavier are not efficiently filtered by the kidneys; therefore, these particles possess longer blood half-lives [
19]. The BBTB of malignant gliomas becomes porous due to the formation of discontinuities within and between endothelial cells lining the lumens of tumor microvessels [
20]. Nanoparticles smaller than the pores within the BBTB, with long blood half-lives, could function as effective transvascular drug delivery devices for the sustained-release of chemotherapeutics into malignant glioma cells.
Even though fenestrations and gaps within the BBTB of malignant gliomas allow for unimpeded passage of low molecular weight therapeutics [
21], these pores are narrow enough to prevent the effective transvascular passage of most nanoparticles [
22‐
25]. If the upper limit of the therapeutically relevant pore size of the BBTB could be accurately determined, then intravenously administered nanoparticles, with long blood half-lives, could serve as effective drug delivery vehicles across the BBTB of malignant gliomas.
By performing intravital fluorescence microscopy of xenografted human glioma microvasculature in the mouse cranial window model, Hobbs et al. [
26] observed perivascular fluorescence 24 hours following the intravenous infusion of rhodamine dye labeled liposomes of 100 nm diameters. Since then several classes of nanoparticles have been designed to be less than 100 nm in diameter for the purposes of effective transvascular drug delivery across the BBTB. These classes of nanoparticles include metal-based (i.e. iron oxide) [
27], lipid-based (i.e. liposomes) [
28], and biological-based (i.e. antibodies, viruses) [
29,
30].
Yet another class of nanoparticles are the polymer-based dendrimers [
2,
31]. Polyamidoamine (PAMAM) dendrimers [
32] are multigenerational polymers with a branched exterior consisting of surface groups that can be functionalized with imaging [
33,
34], targeting [
35], and therapeutic agents [
35,
36]. PAMAM dendrimers functionalized with low molecular weight agents remain particularly small, typically ranging between 1.5 nm (generation 1, G1) and 14 nm in diameter (generation 8, G8) [
32,
33]. Particle shapes are spherical and sizes are uniform within a particular generation. With each successive dendrimer generation, the number of modifiable surface groups doubles while the overall diameter increases by only 1 to 2 nm [
37].
We hypothesized that the major reason for the ineffectiveness of metal-based, lipid-based and biological-based nanoparticles in traversing the BBTB of malignant gliomas is the large size of these particles relative to the physiologic pore size of the BBTB. In this work, using the RG-2 malignant glioma model [
38,
39], we also investigated how the transvascular transport of dendrimer nanoparticles is affected by tumor volume-related differences in the degree of BBTB breakdown.
The hyperpermeability of the BBTB of malignant gliomas results in contrast enhancement of brain tumor tissue on magnetic resonance imaging (MRI) scans following the intravenous infusion of gadolinium (Gd)-diethyltriaminepentaacetic acid (DTPA), a low molecular weight contrast agent [
40,
41]. To visualize the extravasation of PAMAM dendrimers across the BBTB of rodent malignant gliomas by dynamic contrast-enhanced MRI, we functionalized the exterior of PAMAM dendrimers with Gd-DTPA. Using dynamic contrast-enhanced MRI, we measured the change in contrast enhancement of malignant gliomas for up to 2 hours following the intravenous infusion of successively higher Gd-dendrimer generations up to, and including, Gd-G8 dendrimers. To verify that dendrimer size, and not dendrimer generation, is the primary determinant of particle blood half-life, we studied Gd-G4 dendrimers of two different sizes. One was a lowly conjugated Gd-G4 weighing 24.4 kD and the other was a standard Gd-G4 weighing 39.8 kD. The Gd concentration, a surrogate for the amount of Gd-dendrimer within tumor tissue, was determined by measuring the molar relaxivity of Gd-dendrimers
in vitro in combination with the change in the blood and tissue longitudinal relaxivities (T
1) before and after Gd-dendrimer infusion [
42]. Based on comparisons of the contrast enhancement patterns of malignant gliomas for up to 2 hours, within a particular Gd-dendrimer generation as well as across Gd-dendrimer generations, we determined the physiologic upper limit of BBTB pore size.
In addition to the in vivo dynamic contrast-enhanced MRI experiments with Gd-dendrimers, we performed in vitro and ex vivo fluorescence microscopy experiments using rhodamine B labeled Gd- dendrimers to confirm that the impediment to the cellular uptake of functionalized dendrimers is the BBTB. The observations made in this study, using functionalized dendrimers, are to serve as a guide for designing nanoparticles that are effective at traversing the pores of the blood-brain tumor barrier and accumulating within individual glioma cells.
Methods
PAMAM dendrimer functionalization and characterization
Bifunctional chelating agents and gadolinium-benzyl-diethyltriaminepentaacetic acid (Gd-Bz-DTPA) functionalized PAMAM dendrimers were synthesized according to described procedures with minor modifications, as were the corresponding rhodamine-substituted conjugates [
43‐
45]. Gd-dendrimers, with the exception of lowly conjugated Gd-G4, were prepared by using a molar reactant ratio of ≥ 2:1 bifunctional chelate to dendrimer surface amine groups. For lowly conjugated Gd-G4 a lower molar reactant ratio of 1.1:1 was used to limit conjugation. The duration of the chelation reaction for the lowly conjugated Gd-G4 was 24 hours as compared to the standard 48 hours for chelation of all other dendrimers. Rhodamine B labeled Gd-dendrimers were prepared by stirring rhodamine B isothiocyanate (RBITC) and PAMAM dendrimers at a 1:9 molar ratio of RBITC to dendrimer surface amine groups in methanol at room temperature for 12 hours. Isothiocyanate activated DTPA was then added in excess and reacted for an additional 48 hours. Gadolinium was then chelated after the removal of the
t-butyl protective groups on DTPA. The percent by mass of Gd in each Gd-dendrimer generation was determined by elemental analysis to be: Gd-G1 (15.0%), Gd-G2 (14.8%), Gd-G3 (12.9%), lowly conjugated Gd-G4 (12.3%), standard Gd-G4 (12.0%), Gd-G5 (11.9%), Gd-G6 (11.9%), Gd-G7 (12.2%), Gd-G8 (10.2%). The Gd percent by mass for the rhodamine B Gd-dendrimers was determined to be: rhodamine B Gd-G2 (9.6%), rhodamine B Gd-G5 (9.8%), rhodamine B Gd-G8 (9.3%). Gd-G1 through Gd-G5 dendrimer molecular weights were determined by matrix assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectroscopy (Scripps Center for Mass Spectrometry, La Jolla, CA). Gd percent by mass of the Gd-dendrimer, in its solid form, was determined with the inductively coupled plasma-atomic emission spectroscopy (ICP-AES) method (Desert Analytics, Tucson, AZ). Gd-dendrimer infusions were normalized to 100 mM with respect to Gd, while rhodamine B Gd-dendrimer infusions were normalized to 67 mM with respect to Gd, in order to guarantee proper solvation.
In vitro scanning transmission electron microscopy
For
in vitro transmission electron microscopy experiments, a 5 μl droplet of phosphate-buffer saline solution containing a sample of Gd-dendrimers from generations 5, 6, 7 or 8 was absorbed onto a 3 nm-thick carbon support film covering the copper electron microscopy grids. Lacey Formvar/carbon coated 300 meshcopper grids supporting an ultrathin 3 nm evaporated carbon film were glow-discharged an air pressure of 0.2 mbar to facilitate Gd-dendrimer adsorption. After adsorption for 2 minutes, excess Gd-dendrimer solution was blotted with filter paper. The grids were then washed 5 times with 5 μL aliquots of deionized water, and left to dry in air. Annular dark field scanning transmission electron microscope (ADF STEM) images of the Gd-dendrimers were recorded using a Tecnai TF30 electron microscope (FEI, Hillsboro, OR, USA) equipped with a Schottky field-emission gun and an in-column ADF detector (Fischione, Export, PA) [
46].
In vitro fluorescence experiments
For in vitro fluorescence experiments, RG-2 glioma cells were plated on Fisher Premium coverslips (Fisher Scientific, Pittsburgh, PA) and incubated in wells containing sterile 3 ml DME supplemented with 10% FBS (Invitrogen, Carlsbad, CA). The RG-2 glioma colonies were allowed to establish for 24 hours in an incubator set at 37°C and 5% CO2. Rhodamine B Gd-G2, rhodamine B Gd-G5 or rhodamine B Gd-G8 dendrimers were added to the medium by equivalent molar rhodamine B concentrations of 7.2 μM and the cells were incubated in the dark for another 4 hours. Following incubation, cells were washed 3 times with PBS, then 50 μl DAPI-Vectashield nuclear stain medium (Vector Laboratories, Burlingame, CA) was placed on the coverslips for 15 minutes. Coverslips were then inverted and mounted on Daigger Superfrost slides (Daigger, Vernon Hills, IL) and sealed into place. Confocal imaging was performed on a Zeiss 510 NLO microscope (Carl Zeiss MicroImaging, Thornwood, NY). Slides were stored in the dark while not being analyzed.
In vitro magnetic resonance imaging for calculations of Gd-dendrimer molar relaxivity
Gd-dendrimer stock solution (20 μl of 100 mM) and rhodamine B Gd-dendrimer stock solution (30 μl of 67 mM) for the particular generation, used for
in vivo imaging, was diluted using PBS into 200 μl microfuge tubes at 0.00 mM, 0.25 mM, 0.50 mM, 0.75 mM and 1.00 mM with respect to Gd. As an external control, Magnevist (Bayer, Toronto, Canada), a form of Gd-DTPA, was also diluted at the above concentrations into 200 μl microfuge tubes. The microfuge tubes were secured in level and upright positions within a plastic container filled with deionized ultra pure water. The container was placed in a 7 cm small animal solenoid radiofrequency coil (Philips Research Laboratories, Hamburg, Germany) centered within a 3.0 Tesla MRI scanner (Philips Intera; Philips Medical Systems, Andover, MA). Gd signal intensity measurements were made using a series of T
1 weighted spin echo sequences with identical T
E (echo time, 10 ms) but different T
R (repetition time, 100 ms, 300 ms, 600 ms and 1200 ms). Using the measured Gd signal intensity, in addition to the known values for T
R and T
E, the T
1 and equilibrium magnetization (M
0) were calculated by non-linear regression [
42].
In vitro and
in vivo Gd-dendrimer molar relaxivities were assumed to be equivalent for the purposes of this work.
Brain tumor induction and animal preparation for imaging
All animal experiments were approved by the National Institutes of Health Clinical Center Animal Care and Use Committee. Cryofrozen pathogen-free RG-2 glioma cells were obtained from the American Type Culture Collection (Rockville, MD) and cultured in sterile DME supplemented with 10% FBS and 2% penicillin-streptomycin in an incubator set at 37°C and 5% CO
2. The anesthesia and route for all animal experiments was isoflurane by inhalation with nose cone, 5% for induction and 1 to 2% for maintenance. On experimental day 0, the head of anesthetized adult male Fischer344 rats (F344) weighing 200–250 grams (Harlan Laboratories, Indianapolis, IN) was secured in a stereotactic frame with ear bars (David Kopf Instruments, Tujunga, CA). The right anterior caudate and left posterior thalamus locations within the brain were stereotactically inoculated with RG-2 glioma cells [
47]. In each location, either 20,000 or 100,000 glioma cells in 5 μl of sterile PBS were injected over 8 minutes, using a 10 μl Hamilton syringe with a 32-gauge needle. With this approach the majority of animal brains developed one large and one small glioma. On experimental days 11 to 12, brain imaging of re-anesthetized rats was performed following placement of polyethylene femoral venous and arterial cannulas (PE-50; Becton-Dickinson, Franklin Lakes, NJ), for contrast agent infusion and blood pressure monitoring, respectively. After venous cannula insertion, 50 μl of blood was withdrawn from the venous cannula for measurement of hematocrit.
In vivo magnetic resonance imaging of brain tumors
All magnetic resonance imaging experiments were conducted with a 3.0 Tesla MRI scanner (Philips Intera) using a 7 cm solenoid radiofrequency coil (Philips Research Laboratories). For imaging, the animal was positioned supine, with face, head, and neck snugly inserted into a nose cone centered within the 7 cm small animal solenoid radiofrequency coil. Anchored to the exterior of the nose cone were three 200 μL microfuge tubes containing 0.00 mM, 0.25 mM and 0.50 mM solutions of Magnevist to serve as standards for measurement of MRI signal drift over time. Fast spin echo T2 weighted anatomical scans were performed with TR = 6000 ms and TE = 70 ms. Two different flip angle (FA) 3-D fast field echo (3D FFE) T1 weighted scans were performed with TR = 8.1 ms and TE = 2.3 ms, for quantification of Gd concentration. The first FFE scan was performed at a low FA of 3° without any contrast agent on board. The second FFE scan was performed with a high FA of 12°. For this scan, the dynamic scan, each brain volume was acquired once every 20 seconds, for 1 to 2 hours. During the beginning of the dynamic scan, three to five baseline brain volumes were acquired prior to Gd-dendrimer infusion. Gd-dendrimers were infused at doses of 0.03, 0.06 or 0.09 mmol Gd/kg bw depending on the experiment. Gd-dendrimer was infused as a bolus over 1 minute in order to accurately measure the contrast agent dynamics in blood during the bolus. Following completion of the 1 or 2 hour dynamic contrast-enhanced MRI scan, another 15 minute dynamic contrast-enhanced MRI scan was performed during which Magnevist was infused at a dose of 0.30 mmol Gd/kg bw over 1 minute. Tumor regions of interest were drawn based on the Magnevist dynamic scan data.
Dynamic contrast-enhanced MRI data analyses and pharmacokinetic modeling
Imaging data was analyzed using the Analysis of Functional NeuroImaging (AFNI;
http://afni.nimh.nih.gov/) software suite and its native file format [
48]. Motion correction was performed by registering each volume of the dynamic high FA scan to its respective low FA scan. Alignments were performed using Fourier interpolation. A baseline T
1 without contrast (T
10) map was generated by solving equation
1 (the steady-state for incoherent signal after neglecting T
2* effects) voxel-by-voxel for T
1, at both low and high FA's, before contrast was infused [
42].
(1)
where
(2)
After determining the T
10 value at each voxel, T
1 map was calculated using equations
1 and
2 for each voxel of each dynamic image during the high FA scan after contrast infusion [
42]. Datasets were converted to Gd concentration space [
42]. Whole tumor regions of interest were drawn on the basis of the dynamic contrast enhancement pattern of tumor tissue observed following the infusion of Magnevist. These data were important for the drawing of accurate whole tumor regions of interest for minimally enhancing gliomas, especially for all malignant gliomas within the 0.03 mmol Gd/kg bw Gd-dendrimer dose category and those in the 0.09 mmol Gd/kg bw Gd-G8 dendrimer dose sub-category. Normal brain regions of interest were spherical 9 mm
3 volumes in the left anterior caudate.
The pharmacokinetic properties of Gd-G1 through lowly conjugated Gd-G4 dendrimers were modeled using the dynamic contrast-enhanced MRI data from the groups of animals receiving 0.09 mmol Gd/kg bw Gd-dendrimer infusions. The change in blood Gd-dendrimer concentration over time was obtained by selecting 2 to 3 voxels within the superior sagittal sinus, a large caliber vein that is minimally where influenced by in-flow and partial volume averaging effects. Since the transit time of blood movement between an artery and a vein within the brain is approximately 4 seconds, while the image acquisition rate was once every 20 seconds, the superior sagittal sinus was used for generation of the vascular input function for pharmacokinetic modeling [
41]. Animal brains from which an optimal vascular input function could not be obtained were excluded from being analyzed by pharmacokinetic modeling. The voxels chosen had peak blood Gd concentrations closest to the calculated initial Gd-dendrimer volume of distribution, based on the blood volume of a 250 gram rat being 14 ml [
49]. Blood concentration was converted to plasma concentration by correcting for the hematocrit (Hct) as shown in equation
3 [
40].
(3)
The 2-compartment 3-parameter generalized kinetic model (equation 4) [
40,
50] was employed for pharmacokinetic modeling by performing voxel-by-voxel nonlinear regression over all time points.
(4)
Constraints on the parameters were set between 0 and 1 calling on 10,000 iterations. Least squares minimizations were performed by implementing the Nelder-Mead simplex algorithm. Prior to statistical analysis, voxels with poor fits or non-physiologic parameters were censored.
Ex vivo fluorescence microscopy and histological staining of brain tumor sections
Six additional rats received 0.06 mmol Gd/kg bw of rhodamine B Gd-G5 and two additional rats received 0.06 mmol Gd/kg bw of rhodamine B Gd-G8. Subsequent to the standard 2 hour dynamic contrast-enhanced MRI study, the brains of these animals were harvested and snap-frozen. On the day of cryosectioning, two 10 μm sections of tumor bearing brain were cut onto each Daigger Superfrost slide with a Leica Cryotome (Leica, Bensheim, Germany). The first of two slides was prepared for fluorescence microscopy by application of DAPI-Vectashield nuclear stain medium and coversliping. Confocal imaging was performed on a Zeiss 510 NLO microscope. The second slide was stained with Hematoxylin and Eosin for visualization of tumor histology.
Statistical analysis for pharmacokinetic modeling
Vascular parameter pharmacokinetic values for individual tumor voxels were averaged in order to yield one value per parameter per tumor per rat, with tumors within a rat being treated as correlated. On the basis of the range of individual tumor volumes within Gd-G1, Gd-G2, Gd-G3 and lowly conjugated Gd-G4 dendrimer study groups, a dichotomous variable for tumor size was generated by using 50 mm3 as the cut-off between large and small tumors. Multivariate analysis of variance (MANOVA) models were used to examine the effect of dendrimer generation and tumor size. Prior to the MANOVA, it determined that there was no interaction between dendrimer generation and tumor size on any of the three parameters. The covariance structure was considered to be compound symmetric and the Kenward-Roger degrees of freedom method was used. Post-hoc comparisons between lowly conjugated Gd-G4 and each of the other generations were conducted. The significant P-values we report are following Bonferroni correction for multiple comparisons. Analyses were implemented in SAS PROC Mixed (SAS Institute Inc., Cary, North Carolina) with α = 0.05.
Discussion
Effective transvascular delivery of therapeutics into malignant glioma cells remains challenging. Although conventional low-molecular weight chemotherapeutics can easily cross the pores within the BBTB of malignant gliomas [
21,
54], these drugs do not achieve and maintain effective steady state concentrations within malignant glioma cells because of short blood half-lives.
Ultrastructural studies of brain tumor microvasculature have shown that fenestrations and gaps exist within the BBTB ranging from 40 to 90 nm and 100 to 250 nm, respectively [
20,
55]. Using intravital microscopy, Hobbs et al. [
26] have reported that there is primarily perivascular fluorescence in xenografted human malignant gliomas 24 hours after the intravenous infusion of long-circulating rhodamine labeled liposomes 100 nm in diameter. Using MRI, Moore et al. [
25] and Muldoon et al. [
56] have reported that there is minimal contrast enhancement of rodent gliomas 24 hrs after the intravenous infusion of various long-circulating dextran coated iron oxide (also known as LCDIO) nanoparticles with a mean diameter of 20 nm [
57,
58]. These findings indicate that the therapeutically relevant upper limit of the BBTB pore size should range between 20 nm and 100 nm. However, the effective transvascular delivery of nanoparticle-based drug carriers across the BBTB into malignant glioma cells has remained elusive, to date. We reasoned that the physiologic upper limit of BBTB pores size would be less than 20 nm in diameter. We were aware that PAMAM dendrimers are particularly small multigenerational nanoparticles of uniform sizes within a generation [
31,
37]. Functionalized PAMAM dendrimer particle sizes typically range between 1.5 nm (G1) and 14 nm (G8) in diameter following the conjugation of low molecular weight imaging compounds to the dendrimer exterior [
33]. In order to probe the physiologic upper limit of BBTB pore size in RG-2 malignant glioma microvasculature with dynamic contrast-enhanced MRI, we functionalized PAMAM dendrimers G1 through G8 with Gd-DTPA (charge -2) [
33,
34,
45]. As a result of the conjugation of Gd-DTPA to approximately half of the surface amine groups, the positive surface charge on the PAMAM dendrimer exterior was neutralized. In order to confirm that the barrier to cellular entry of Gd-dendrimers is at the level of the BBTB, and that permeable functionalized dendrimers with long blood half-lives can accumulate in malignant glioma cells, we used rhodamine B labeled Gd-dendrimers for fluorescence imaging
in vitro and
ex vivo. Based on these studies, we report here that the physiologic upper limit of BBTB pore size ranges between approximately 11.7 and 11.9 nm. We also report that permeable functionalized dendrimers with long blood half-lives can accumulate within glioma cells.
We observed that there was virtually no contrast enhancement of malignant glioma tissue over 2 hours on dynamic-contrast enhanced MRI following the intravenous infusion of Gd-G8 dendrimers. We found this to be the case at both Gd-dendrimer doses investigated, one being the standard 0.03 mmol Gd/kg bw dose for pre-clinical dynamic contrast-enhanced MRI and the other being 0.09 mmol Gd/kg bw [
33]. These dynamic contrast-enhanced MRI findings demonstrate that Gd-G8 dendrimers are larger than the upper limit of the physiologic pore size of the BBTB of RG-2 gliomas. Using ADF STEM, we measured the diameters of a population of our Gd-G8 dendrimers to be 13.3 ± 1.4 nm (mean ± standard deviation) and that of Gd-G7 dendrimers to be 11.0 ± 0.7 nm. Based on these ADF STEM data, the range of the physiologic upper limit of BBTB pore size in RG-2 malignant gliomas is between 11.7 and 11.9 nm.
To confirm that the limitation to functionalized G8 dendrimer entry is not at the cellular level, we performed fluorescence microscopy of cultured RG-2 glioma cells following the application of rhodamine B labeled Gd-dendrimers to the media. We found that rhodamine B labeled Gd-G2, -G5 and -G8 dendrimers accumulated in the cytoplasm of all RG-2 glioma cells; however, we found it particularly interesting that, in some cases, rhodamine B labeled Gd-G2 dendrimers also accumulated in the RG-2 glioma cell nuclei. This finding suggests that it may also be possible for other smaller nanoparticles (i.e. molecular weight ≤ 11.2 kD) to cross nuclear pores.
Irrespective of dose, we found that Gd-G1, Gd-G2, Gd-G3 and lowly conjugated Gd-G4 (molecular weight 24.4 kD) dendrimers had short blood half-lives because particle sizes of these lower generation Gd-dendrimers are small enough that particles can be efficiently filtered by the kidneys [
17]. Therefore, Gd-G1 through lowly conjugated Gd-G4 dendrimers only remain temporarily within the tumor extravascular extracellular space. We also found that as the Gd-dendrimer generation and particle size increased, the transvascular flow (
K
trans
) rate decreased; and that the lower transvascular flow rate of lowly conjugated Gd-G4 dendrimers resulted in the more focal distribution of particles within brain tumor tissue. Therefore, since lower generation dendrimers have short blood half-lives, the transvascular flow rate across the BBTB is the primary determinant of how widespread particle distribution was within the extravascular extracellular tumor space. These findings suggest that nanoparticles with higher molecular weights,
yet particle sizes small enough to
still be effectively filtered by the kidneys, do not remain within the extravascular tumor space sufficiently long to effectively permeate through tumor tissue. Therefore, such nanoparticles would remain within close proximity of tumor microvessels, and would not reach malignant glioma cells located within tumor regions that are poorly vascularized.
We found that standard Gd-G4 dendrimers (molecular weight 39.8 kD) had a longer blood half-life than the lower generation Gd-dendrimers because the particle size of standard Gd-G4 dendrimers is at the threshold of effective renal filtration [
17]. Irrespective of dose, Gd-G5 through Gd-G8 dendrimers maintained steady state blood concentrations over a minimum of 2 hours because particle sizes of these generations of Gd-dendrimers are clearly above the threshold of effective renal filtration [
17]. As a result of the long blood half-lives, Gd-G5 and Gd-G6 were able to slowly extravasate across the BBTB of even the smallest gliomas that we studied. Based on these findings, we conclude that it may be possible to effectively deliver permeable nanoparticles with long blood half-lives across a minimally compromised BBTB, including across the BBTB of the microvasculature supplying emerging malignant glioma colonies.
To verify that only permeable functionalized dendrimers with long blood half-lives accumulate within malignant glioma cells, we infused rhodamine B labeled Gd-G5 dendrimers and rhodamine B labeled Gd-G8 dendrimers to separate groups of rats. The dose of rhodamine B Gd-dendrimers was 0.06 mmol Gd/kg bw, since in pilot experiments we observed that the anesthetic effect of isoflurane was potentiated at the 0.09 mmol Gd/kg bw rhodamine B Gd-dendrimer dose [
59,
60]. Fluorescence microscopy of RG-2 glioma specimens demonstrated extensive subcellular localization of rhodamine B Gd-G5 dendrimers, confirming that functionalized G5 dendrimers accumulate within malignant glioma cells, due to long blood half-lives.
We observed with both fluorescence microscopy and dynamic contrast-enhanced MRI that there was some accumulation of rhodamine B Gd-G8 dendrimers in RG-2 gliomas (Figure
5C and
5E), as well as some non-selective accumulation of rhodamine B Gd-G5 and rhodamine B Gd-G8 dendrimers in tumor-free brain regions (Additional file
5). We suspect that rhodamine B labeled Gd-G5 and Gd-G8 dendrimers are toxic to the BBTB in addition to the otherwise healthy blood-brain barrier. This toxicity is likely due to the introduction of additional positive charge to the Gd-dendrimer surface from the attachment of rhodamine B, a cationic and lipophilic fluorescent dye [
61‐
64]. Therefore, the extravasation of rhodamine labeled nanoparticles [
26,
65] and other charged nanoparticles [
66‐
69] across the barrier may be from direct charge induced damage to endothelial cells of the barrier and disruption of the barrier. Our proposed mechanism for the increased barrier permeation of rhodamine labeled Gd-dendrimers is analogous to the mechanism recently proposed by Herce and Garcia [
70,
71] for the movement of cell-penetrating peptides across cell membranes. We plan to clarify, in the future, with additional
in vivo imaging experiments, the relationship between charge on the dendrimer surface and disruption of the blood-brain barrier.
Conclusion
In this study, we identified the precise physiologic upper limit of blood-brain tumor barrier pore size, and demonstrated that nanoparticles of diameters smaller than this upper limit can effectively traverse the pores of the blood-brain tumor barrier; in addition, we validated the importance of prolonged nanoparticle blood half-life for the effective accumulation of nanoparticles within brain tumor cells. Therefore, based on these findings, we conclude that effective drug delivery across the BBTB of malignant gliomas, and potentially the BBB of other neuropathologies, can be accomplished with non-toxic nanoparticles that are smaller than 11.7 to 11.9 nm in diameter and have prolonged blood half-lives.
In the broadest sense, our findings will serve as general guidelines, for the future design and development of multifunctional transvascular delivery devices, based on nanoparticles (i.e. liposome-, quantum dot-, or iron oxide-based) and biological particles (i.e. antibody- or viral-based), that are particularly effective at crossing the diseased BBB and accumulating in neuropathologic tissues.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
HS conceptualized, designed, and supervised the overall study; performed the dynamic contrast-enhanced MRI experiments, analyzed the data, interpreted the overall study results, and prepared the manuscript. ASK performed the dynamic contrast-enhanced MRI experiments, analyzed the data, and assisted with the preparation of the manuscript. HW synthesized and performed the preliminary characterization of the functionalized dendrimers. KRB assisted with the confocal fluorescence microscopy experiments. SHF performed the initial dynamic contrast-enhanced MRI experiments. KS assisted with the preparation of the manuscript. AAS characterized the higher generation functionalized dendrimers by electron microscopy. SA performed the statistical data analysis. CMW assisted with the synthesis of the functionalized dendrimers. MAA assisted with the characterization of the higher generation functionalized dendrimers by electron microscopy. RDL supervised the electron microscopy-based characterization of the functionalized dendrimers. GLG supervised the synthesis and preliminary characterization of the functionalized dendrimers, and contributed to the design of the overall study. MDH conceptualized, designed, and supervised the confocal fluorescence microscopy experiments; assisted with the interpretation of the overall study results, and prepared the manuscript.