Distribution of Polymer Nanoparticles by Convection-Enhanced Delivery to Brain Tumors

Abstract

Glioblastoma multiforme (GBM) is a fatal brain tumor characterized by infiltration beyond the margins of the main tumor mass and local recurrence after surgery. The blood-brain barrier (BBB) poses the most significant hurdle to brain tumor treatment. Convection-enhanced delivery (CED) allows for local administration of agents, overcoming the restrictions of the BBB. Recently, polymer nanoparticles have been demonstrated to penetrate readily through the healthy brain when delivered by CED, and size has been shown to be a critical factor for nanoparticle penetration. Because these brain-penetrating nanoparticles (BPNPs) have high potential for treatment of intracranial tumors since they offer the potential for cell targeting and controlled drug release after administration, here we investigated the intratumoral CED infusions of PLGA BPNPs in animals bearing either U87 or RG2 intracranial tumors. We demonstrate that the overall volume of distribution of these BPNPs was similar to that observed in healthy brains; however, the presence of tumors resulted in asymmetric and heterogeneous distribution patterns, with substantial leakage into the peritumoral tissue. Together, our results suggest that CED of BPNPs should be optimized by accounting for tumor geometry, in terms of location, size and presence of necrotic regions, to determine the ideal infusion site and parameters for individual tumors.

Graphical Abstract

Introduction

More than 17,000 new cases of brain cancer are diagnosed in the U.S. each year [1]. Glioblastoma multiforme (GBM), a primary high-grade tumor that originates from glial cells, is the most common, aggressive, and deadly form [2]. The current standard of care includes surgical resection, when possible, followed by radiotherapy and chemotherapy with temozolomide, an alkylating agent that is administered systemically [2-4]. Despite this aggressive treatment, the median life expectancy is approximately 14 months, with modest improvement over the past 35 years [5]. Tumor recurrence is primarily due to tumor cell infiltration into tissue surrounding the bulk of the tumor [6], difficulty in delivery of drugs through the blood-brain barrier [7], and resistance of tumor cells to many chemotherapy drugs [8].

Various strategies have been tested to circumvent the BBB, enhance the delivery of drugs to the local tumor site, and reduce systemic toxicities. For example, controlled-release vehicles, such as the carmustine-loaded Gliadel® wafer, can be placed in the tumor resection cavity at the time of surgery [9, 10]. Although Gliadel® increases survival of GBM patients [11, 12], the survival benefit appears to be limited by the volume of drug exposure in the brain, which is determined by drug diffusion from the wafer [13, 14]. Direct infusion of drugs via convection-enhanced delivery (CED) induces fluid convection in the brain by use of an external pressure gradient, providing volumes of distribution much larger than achievable by diffusion [15]. Hence, the benefits of CED include both the ability to locally target the tumor area, and to deliver agents that cannot penetrate the BBB over a large volume of tissue [16, 17]. Recent clinical trials showed that CED is safe and feasible [18], but fails to significantly improve survival in GBM. This is likely due to multiple technical factors, such as catheter placement, but more importantly, most drugs have short half-lives in the brain and are metabolized soon after the infusion stops. Encapsulation of therapeutic agents into nanocarriers such as polymeric nanoparticles or liposomes can protect them from clearance by protecting them from transport and metabolism until their release from the carrier. Polymer nanoparticles [19] and liposomes [20] can be administered by CED, offering the potential for sustained drug release in the days to weeks after infusion has ended [21-23]. However, of the two classes of materials, polymer nanoparticles allow for easier control over sustained release of agents than liposomes [24], and more versatility in design permitted by synthetic polymer chemistry [25].

Recently, we described the infusion of “brain-penetrating” nanoparticles (BPNPs) that diffuse readily in the healthy brain when administered by CED, thanks to their small size and lack of aggregation [22]. In previous studies, we have shown that CED of drug-loaded polymer nanoparticles and drug-loaded polymer BPNPs enhance survival in animals with intracranial tumors and, in some cases, eliminate tumors completely [21, 22]. Recent analysis of GBM patients revealed that while 80% of patients receive some form of surgical resection, 19% received only biopsy [26], indicating that there is a need for better treatment of partially or non-resectable tumors. CED might be useful for this subset of patients in which complete resection is not possible. However, little is known about the dynamics of CED in the presence of tumors, and almost nothing is known about the transport of nanoparticles when delivered by CED in regions of the brain where tumor is present. Indeed, brain tumor tissue is different than normal brain tissue, presenting dense cellular patterns, resulting in highly tortuous pathways [27] as well as potentially higher interstitial fluid pressures within the tumor core [28]. Previous reports examining the intratumoral infusion of liposomes and other small molecules via CED into rat brain tumors have shown distribution of varying degrees, with some resulting in widespread tumor coverage [29-34]. A better understanding of intratumoral transport of polymer nanoparticles during CED is needed.

In this report, we investigated the effect of intracranial tumors on the distribution of PLGA BPNPs delivered by CED. We examined the distribution of BPNPs in two intracranial glioma models, U87 and RG2, which were selected for their differing characteristics and frequent use in the literature. U87 tumors, although derived from a human GBM, have limited infiltrative capacity and necrosis, with enhanced vascularization [35]. RG2 tumors, derived from a rat cell line, demonstrate some infiltrative properties with the presence of necrotic tissue [36, 37]. We determined the distribution patterns of BPNP both within and around the tumor to measure the capacity of these particles to deliver drugs in the bulk tumor as well as the associated tumor microenvironment.

Materials and Methods

Materials

Ester-terminated poly(lactide-co glycolic acid) (PLGA, 50:50 copolymer ratio, 0.55-0.75 dL/g inherent viscosity) was acquired from LACTEL Absorbable Polymers, Durect Corporation (Birmingham, AL, USA). Dichloromethane was ordered from Fischer Scientific (Fair Lawn, NJ). Ethyl Acetate was purchased from J.T. Baker (Center Valley, PA, USA). Nile Red, sterile phosphate buffered saline (PBS), Dulbecco's Modified Eagle Medium, and penicillin-streptomycin were ordered from Life Technologies (Eugene, OR, USA). Polyvinyl alcohol (PVA), trehalose, and paraformaldehyde were ordered from Sigma-Aldrich (St. Louis, MO, USA). U87 cells were purchased from American Type Tissue Culture (ATCC, Manassas, VA, USA). RG2 cells were a generous gift from Dr. Fahmeed Hyder (Magnetic Resonance Research Center, Yale University, New Haven, CT, USA). Fetal bovine serum was acquired from Atlanta Biologicals (Flowery Branch, GA, USA). pSicoR Plasmid (#11579) was ordered from Addgene (Cambridge, MA, USA).

Fabrication of PLGA BPNPs

BPNPs were fabricated from PLGA using a double emulsion technique as previously described [22]. Briefly, 100 mg PLGA polymer was dissolved in ethyl acetate at a concentration of 50 mg/mL. In order to be able to visualize and quantify the presence of BPNPs in the brain after CED, Nile red was encapsulated by adding it in the polymer solution at a concentration of 2 μg/mg of polymer. The aqueous phase consisted of 2 mL of 5% polyvinyl alcohol (PVA). An emulsion of the solutions was created through dropwise addition of the polymer solution to the aqueous solution under vortex, followed by three cycles (10 sec each) of sonication using a probe sonicator. BPNPs were separated from the larger nanoparticle population through centrifugation at 11,500 rcf. Ultracentrifugation steps were used to further separate and wash BPNPs. Trehalose was added as a cryoprotectant at 30% wt/wt before lyophilization.

Characterization of BPNPs

Sizing and morphology

NPs were sized using both imaging techniques (SEM) and dynamic light scattering (DLS). For SEM imaging, dry PLGA BPNPs (without trehalose) were placed on carbon tape and sputter coated with gold for 30 sec under 40 mA current (Sputter Coated 180aute, Cressington). Images were acquired using a XL-30 ESEM-FEG (FEI Company, Hillsboro, OR, USA) under 10 kV acceleration voltage. Average nanoparticle size was quantified using the ImageJ software (National Institutes of Health, Bethesda, MD, USA), where sample populations of at least 1,000 NPs were included for statistical analysis.

Hydrodynamic size analysis was completed using DLS. BPNPs were resuspended at 0.05 mg/mL in DI water and sized using a Malvern Nano-ZS. Results are reported as the Z-average diameter, corresponding to the hydrodynamic diameter of the particles. In all cases, the polydispersity index was less than 0.2. To measure surface charge (zeta potential), BPNPs were diluted in DI water at a concentration of 0.5 mg/mL; 750 μL of solution was loaded into a disposable capillary cell (Malvern, UK) and the charge was measured using a Malvern Nano-ZS.

Fluorescent dye loading in BPNP

Nile red loading in BPNPs samples was determined through fluorescence quantification. Five mg of BPNPs were diluted in 1 mL of DMSO and left to dissolve for 1 h. Samples were quantified by comparing to a standard curve of free nile red in DMSO in a 96 well plate, using a SpectraMax M5 plate reader.

In vivo CED in normal brain

All animal work was completed at Yale University in accordance with Yale Animal Resource Center (YARC) and the Institutional Animal Care and Use Committee (IACUC) guidelines. Male Sprague Dawley rats (200-250 g, Charles River, Willimantic, CT, USA) were placed under ketamine (75 mg/kg) and xylazine (5 mg/kg) anesthesia, and Meloxicam SR analgesia (4 mg/kg) until a surgical plane was achieved. The heads were shaven and rats were placed in a stereotactic frame. After sterilization of the scalp with alcohol and betadine, a mid-line incision was created and a 1.5 mm burr hole was drilled in the skull at 1 mm anterior and 3 mm lateral to bregma, in order to reach the right striatum. PLGA BPNPs were sonicated and vortexed to ensure proper resuspension at a concentration of 100 mg/mL in PBS, and loaded in a 50 μL Hamilton syringe with a stepped tip of polyamide tubing. The syringe was then inserted into the burr hole at a depth of 5 mm from the top of the brain, and left to equilibrate for 7 min before infusion. A micro-infusion pump (World Precision Instruments, Sarasota, FL, USA) was used to infuse 20 μL of BPNPs at a rate of 0.667 μL/min (30 min of infusion). The syringe was left in place for 7 min post infusion for tissue equilibration, before catheter removal and euthanasia. Brains were immediately harvested and frozen for further tissue processing.

As a control, Evan's blue-labeled albumin was infused using the same methods as described above to determine distribution of an ideal small molecule in our system. Evans Blue at 1.25 mg/mL was added to a solution of albumin at 20 mg/mL in 1X PBS and stirred for 1 h. The solution was then filtered at 0.22 μm prior to infusion.

In vitro cell culture and GFP transduction

U87 and RG2 cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S), and maintained in a humidified incubator at 37°C and 5% CO₂. U87-GFP and RG2-GFP cells were generated through transduction of normal U87 and RG2 cells with artificial lentiviruses produced using pSicoR-GFP as previously described [38].

In vivo intracranial tumor implantation

Male nude rats (U87-GFP studies) or male Fischer 344 rats (RG2-GFP studies) were prepped for intracranial tumor implantation as described previously for the infusion study. After induction of anesthesia and shaving, animals were placed in the stereotactic frame and their scalps were sterilized with alcohol and betadine. A mid-line incision was created and a small burr hole was drilled in the skull at 1 mm anterior and 3 mm lateral to bregma to reach the right striatum. U87-GFP or RG2-GFP cells were trypsinized, washed and resuspended in sterile PBS. A 10 μL Hamilton syringe was loaded with cell suspension and inserted through the burr hole at a depth of 5 mm from the top of the brain. After 5 min of tissue equilibration, 2.5×10⁵ cells were injected in 3 μL of PBS at an infusion rate of 1 μL/min. At the end of the infusion, the needle was left in place to allow the tissue to equilibrate for an additional 5 min before careful removal. Bone wax was used to fill the burr hole, and skin was stapled and cleaned with antibiotic ointment. Animals were placed in a recovery cage until sternal. Tumors were allowed to grow in the animals for either 8 or 13 days (RG2-GFP tumors), or 13 or 26 days (U87-GFP tumors) before a second surgery for BPNPs infusion was performed as described above for normal brains. Differences in infusion time points were selected based on the growth rates of the individual tumor models. Brains were harvested immediately after ending CED and tumor volume was determined by measurement of the tumor length and width at its largest points on histological slices and the tumor volume (V) was estimated using the formula V = (length × width²)/2 in mm³ [39]. The site of infusion of BPNPs was confirmed on histological slices as residing within the tumor interior for inclusion in all further analysis.

Brain tissue processing and histological analysis

Fluorescent imaging of brain cryosections

Frozen brains were cryosectioned in 50 μm slices on the coronal plane. All sections were collected and stored on charged Superfrost plus glass slides (Fisher Scientific, Pittsburgh, PA, USA). Individual slices were imaged to determine fluorescent BPNPs distribution or Evan's blue albumin distribution using a Zeiss Lumar.V12 stereoscope. Using a MATLAB script developed in our laboratory [19, 21, 22], image reconstruction was completed to determine the volume of distribution of BPNPs or Evan's blue in each brain, including both white and grey matter volumes. Results were expressed as V_d/V_i, where V_d corresponds to the volume of distribution of the BPNPs after CED, and V_i to the volume of infusion (which was 20 μL in every case). 2D tumor and BPNP distribution projections were constructed by displaying slices every 500 μm either anterior or posterior to the site of infusion: fluorescent images of BPNPs and tumors were merged and the image background was removed using an image processing software. Scaled images were placed on a representative drawing of a coronal brain slice for display.

Analysis of co-localization of nanoparticles and tumor

Further histological analysis to determine tumor penetration and co-localization of NPs was completed using a Leica DMI6000 B microscope. Brightfield and fluorescent images were acquired at 5X magnification, and reconstructed to display whole brain hemispheres using automated image tiling and reconstruction software (Leica Microsystems, Buffalo Grove, IL, USA). Additional images were acquired at 10X magnification. Co-localization of BPNPs within tumors was analyzed using MetaMorph Image Analysis Software (Molecular Devices, Inc., Sunnyvale, CA, USA). Separate channels displaying BPNPs and tumor fluorescent images were thresholded and analyzed for overlap using the pre-installed co-localization application. 50 μm slices at the site of infusion (designated as 0 μm), as well as slices 500 μm anterior (designated −500 μm) and posterior (designated 500 μm) were averaged for all samples in each treatment group to determine an overall average value at each location.

Hematoxylin and eosin (H&E) stain of paraffin embedded sections and cryosections

Non-CED infused brains containing either RG2-GFP (13 day) or U87-GFP (26 day) tumors were fixed in 4% PFA and cryoprotected in 30% sucrose. Samples were embedded in paraffin and sliced at 5 μm. Cryosections at 50 μm on slides were fixed in 10% NBF. All samples were stained with H&E by Yale Research Histology Services to examine general tumor histology and necrotic tissues. Samples were imaged using an Axioimager A1 microscope with an Axiocam mHRC color camera (Carl Zeiss, Jena, Germany) at 5X magnification. Composite H&E images of whole brain slices were created using Microsoft Image Composite Editor software. H&E samples were further analyzed by a board certified neuropathologist. Concentration maps were produced from fluorescence images using an ImageJ script, where pixel count from 0-255 defined the intervals from low (blue) to high concentration (red).

Statistical Analysis

Brain volume distributions were compared using a standard one-way ANOVA.

Results

Nanoparticle characterization

PLGA NPs (Fig 1A) were characterized for size using SEM (Fig 1B) and DLS (Table 1). These NPs qualify as BPNPs, according to our prior definition [22], as they were around 80 nm in diameter (by microscopy) and non-aggregating (by DLS). PLGA BPNPs were observed to be spherical in morphology and homogeneous in size (Fig 1B). They were found to have negatively charged surfaces by zeta potential analysis (Table 1). Additionally, PLGA Nile Red NPs released less than 3% of dye after 25 days of incubation in physiological conditions (Supp Fig 1).

Figure 1.

(A) Schematic of PLGA NPs loaded with Nile Red. (B) PLGA NPs imaged via SEM displayed spherical particles with homogenous populations. (C) Overlay of a representative image of an injection site and area of distribution of nanoparticle on a rat striatum. (D) Reconstructed 3-D volume of distribution (V_d) of fluorescent NPs.

Table 1.

PLGA NP characteristics

Diameter, SEM/TEM (nm ± SD)	74 ± 37
Hydrodynamic diameter (nm ± SD)	166 ± 43
Zeta Potential (mV ± SD)	−23.1 ± 8
Vd/Vi ± SD	2.53 ± 0.27

In vivo volume of distribution of PLGA BPNPs in the normal rat brain via CED

The fluorescently-labeled PLGA BPNPs were infused via CED into the striatum of Sprague Dawley rats; volumes of distribution were determined from reconstructed fluorescence images (Fig 1C, D). Brains were cryosectioned and imaged to detect BPNP distribution through volumetric image reconstruction. Volume of distribution (V_d) was ~ 50 mm³, corresponding to a V_d/V_i of ~ 2.5 (Table 1). Evan's blue-labeled albumin has been frequently used as a tracer for infusion via CED, as it can be easily detected due to its blue color or fluorescence [40]. As a basis of comparison to our BPNPs distributions, Evan's blue-labeled albumin was injected via CED using the same surgical delivery system and imaging methods used for our BPNP, yielding a V_d/V_i of 4.2.

U87-GFP and RG2-GFP rat brain tumor implantation, detection, and characterization

Prior to testing BPNPs distribution in brains containing tumors, we first characterized the growth of two different intracranial tumors in rats. U87 and RG2 cells were transduced with GFP to allow imaging via fluorescence either in culture or after implantation into the rat brain (Supp Fig 2). Intracranial tumors were examined at early and late times after infusion of cells into the brain. After 8 days of growth for the RG2-GFP cells or 13 days of growth for the U87-GFP cells, the tumors presented a volume of ~ 5 mm³ (Supp Fig 2C, D, G, H), and were referred as “small” tumors. On the other hand, after 13 days of growth for the RG2-GFP cells or 26 days for the U87-GFP cells, the tumors presented significantly larger volumes of 80 and 100 mm³ respectively (Supp Fig 2E, F, I, J), and were considered as “large” tumors. Sections from brains with large RG2-GFP or large U87-GFP tumors that were not infused with NPs were stained with H&E (Supp Fig 3) in order to characterize histopathological features of these tumors. RG2-GFP tumors displayed dense cellular populations with multiple locations of tumor necrosis and indications of infiltration (Supp Fig 3A, C), whereas U87-GFP tumors presented a dense cellular core with limited indications of necrosis and a circumscribed border (Supp Fig 3B, D). Necrotic areas were not readily observed in small RG2 or U87 tumors (data not shown). These characteristics are consistent with previous reports [36] and compared to histopathological characteristics of primary human GBM in Table 2 [41, 42].

Table 2.

Characteristics of human GBM and intracranial tumors in rats

Tumor	Cellular Density	Infiltration	Necrosis
Human Glioblastoma*	High	Extensive	Extensive
U87	High	Circumscribed	Limited
RG2	High	Limited Infiltration	Extensive

^*References for human glioblastoma characterization (1, 2)

BPNP CED in RG2-GFP and U87-GFP tumors

To investigate if BPNP distributions were altered due to the presence of a tumor, the BPNPs were infused into the caudate of rats bearing either small or large RG2-GFP or U87-GFP tumors. Sections from the brains of animals with small RG2-GFP tumors (Fig 2A), large RG2-GFP tumors (Fig 2B), small U87-GFP tumors (Fig 2C), and large U87-GFP tumors (Fig 2D) were imaged by fluorescence microscopy and reconstructed in three dimensions (Supp Fig 4). The results of infusions into three animals are shown, illustrating the observed range of tumor growth and BPNP distribution. Images from infusions into small tumors (Fig 2A,C and Supp Fig 4A,C) show significant quantities of BPNPs in the peritumoral space, such that most of the tumor is surrounded by particles which are distributed along the tumor margins. Images from infusions into large tumors (Fig 2B,D and Supp Fig 4B,D) show variable BPNP coverage of the tumor, and the margins of these larger tumors are only partially surrounded by particles.

Figure 2.

Projections of A) RG2-GFP 8 day tumors, B) RG2-GFP 13 day tumors, C) U87-GFP 13 day tumors, and D) U87-GFP 26 day tumors with NP CED infusions. Three typical GFP tumor infusions with NP (red) and tumor (green) are displayed. Presented coronal sections are at 2 mm intervals. Black lines indicate the approximate site of NP infusion during CED. Scale bar = 4 mm

The volumes of distribution of the BPNPs were determined after CED into brains with RG2-GFP and U87-GFP tumors of different size. Interestingly, the presence of a small or large tumor did not affect V_d, as all groups yielded similar values, which were similar to V_d measured in brains with no tumor (Fig 3A). To further study the effect of the tumor on BPNP distribution, the area of BPNP spread was examined for each coronal brain section. Although the final V_d of BPNPs was similar in the normal brain and brains with small and large tumors, the presence of a tumor tended to increase the anterior-to-posterior breadth of distribution, in particular in the case of large tumors (Fig 3B). In small tumors, the volume of infusion (20 mm³) and the volume of BPNP distribution (~50 mm³) are both larger than the volume of the tumors (5 mm³). In large tumors, the volumes of infusion and distribution are smaller than the tumors (80-100 mm³).

Figure 3.

(A) CED of NP volume distribution in tumor-burdened rat brains. Unmodified PLGA BPNPs were injected into both small and large U87-GFP or RG2-GFP tumors. Total volume distribution of the BPNPs in the brain indicated that the presence of a tumor, large or small, does not significantly change Vd/Vi. In addition, the Vd/Vi in the presence of a tumor is similar to that in normal brain (for normal brain and U87-GFP large, N = 4; for U87-GFP small, and RG2-GFP large and small, N = 5). (B) Area of distribution of NPs via CED in the presence of a brain tumor. The distribution of unmodified NPs in normal brain (black curve) was compared to that of NPs infused by CED in the presence of a small RG2-GFP, large RG2-GFP, small U87-GFP and large U87-GFP tumors. Curves were aligned based on the site of infusion at 0 μm with the average of all infusions calculated anterior and posterior to the site of infusion (for normal brain and large U87-GFP, N = 4; for large and small RG2-GFP, as well as small U87-GFP, N = 5).

Histological analysis of nanoparticles and tumor environment

H&E staining of intracranial tumors was performed after CED of BPNPs, in order to investigate the influence of BPNP infusion on histopathological features of both tumor models (Supp Fig 5). In representative brain slices at the infusion site, large RG2 tumors (Supp Fig 5A) were found to present numerous necrotic regions (Supp Fig 5B, C), whereas large U87 tumor samples (Supp Fig 5D) exhibited limited or no necrosis (Supp Fig 5E, F). Overall, these results indicate that the infusion of BPNPs did not influence the histopathological findings in either tumor type.

For small RG2-GFP tumors, with a volume of ~ 5 mm³, tumor masses were localized to the caudate and, typically extended along the axis of catheter tract (Fig 4A), resulting in an elongated shape. Brightfield images merged with fluorescence images from the tumor (Fig 4B) and BPNPs (Fig 4C,D) formed composite images (Fig 4A,E), which revealed a diffuse but nonuniform pattern of intratumoral BPNPs distribution. BPNPs were concentrated at the site of infusion, spreading in certain directions over the tumor mass and accumulating around the tumor margins (Fig 4F). BPNPs were distributed throughout the tumors (Fig 4E), penetrating laterally around the site of infusion (Fig 4F). At sites away from the point of infusion, BPNPs infiltrated around the tumor margin (Fig 4G,H), but were present at substantially lower concentration than in the tumor core (Fig 4D).

Figure 4.

RG2-GFP 8 day intratumoral infusion. (A) Brightfield image of an 8 day (B) RG2-GFP tumor (green) with (C) infused NPs (red). (D) Concentration analysis indicates medial-lateral NPs spread through the tumor, at high concentration near the site of infusion. (E) Microscopic analysis of (F) NP infiltration indicates particles around the infusion site. (G) Particles are observed to enter the tumor from the external tumor borders (H) away from the site of infusion. Scale bar = 2 mm. Images F, G and H taken at 10X magnification. Colored boxes on (E) demonstrate locations of images (F-H). White line denotes tumor border.

For large RG2-GFP tumors, with a volume of ~ 80 mm³, the central tumor mass was primarily in the caudate, but also present in the overlying cortex, and resulted in disruption of the corpus callosum. There was evidence of slight ventricular occlusion, which produced a midline shift due to the large tumor size (Fig 5A). Brightfield images of the hemisphere were merged with fluorescence images from the tumor (Fig 5B) and BPNPs (Fig 5C,D) to form composite images (Fig 5A,E), revealing substantial distribution of BPNP within the tumor core. Tumor lobes away from the site of infusion, such as in the cortex, typically did not contain BPNP (Fig 5A). Concentration analysis demonstrated high but diffuse distribution of BPNPs in the tumor mass, with leakage and accumulation at lower concentration in the lateral peritumoral regions (Fig 5D), which was a consistent finding in tumors in separate animals (Fig 2B). Microscopic analysis of tumor and BPNP overlap (Fig 5E) suggests an accumulation of BPNPs at the lateral tumor periphery (Fig 5F), which was less pronounced at the superior and inferior tumor margins (Fig 5G). Substantial and diffuse coverage by BPNPs was observed in the tumor mass near the infusion site (Fig 5H).

Figure 5.

RG2-GFP 13 day intratumoral CED infusion. Coronal cryosection at site of infusion were imaged via (A) merged brightfield with (B) tumor (green) and (C) NPs (red). Concentration analysis demonstrates (D) widespread and high concentration of NPs within the tumor. Microscopic analysis of tumor and NP co-localization (E) demonstrates (F) infiltration at the tumor borders near the site of infusion, however (G) limited penetration away from the site of infusion. (H) Widespread distribution of NPs is seen within the tumor. Colored boxes on (E) demonstrate locations of images (F-H). White line denotes tumor border. Scale bar = 2 mm. Images F, G, and H taken at 10X magnification.

Small U87-GFP tumors, with a volume of ~ 5mm³, were confined to the caudate (Fig 6A). Brightfield images merged with fluorescence images from the tumor (Fig 6B) and BPNP (Fig 6C,D) formed composite images (Fig 6A,E) that revealed the presence of BPNP at the site of infusion, as well as extensive peritumoral coverage, nearly surrounding the tumor at the margin. This peritumoral coverage was consistent among animals (Fig 2C). BPNP were present at high concentration within the tumor core, and at lower concentrations in the peritumoral region (Fig 6D). BPNP appeared diffusely distributed throughout the majority of the tumor (Fig 6D). Higher magnification analysis revealed BPNP penetration into the tumor from the site of infusion (Fig 6F) and accumulation of BPNPs in the volume surrounding the tumor (Fig 6G, H).

Figure 6.

U87-GFP 13 day intratumoral infusion. Small U87 tumors were imaged via (A) brightfield with co-localization of (B) tumor (green) and (C) NPs. (D) High intratumoral concentrations were found at the site of infusion, with diffuse NP concentration throughout the tumor. (E) Analysis of NP and tumor overlap indicated (F) spread of NP from the site of infusion (G) with limited NP penetration from the peritumoral borders. (H) Extensive NP coverage of the peritumoral regions was observed. Colored boxes on (E) demonstrate locations of images (F-H). White line denotes tumor border. Scale bar = 2 mm. Images F, G, and H taken at 10X magnification.

Large U87-GFP tumors, with a volume of ~ 100 mm³, encompassed the majority of the hemisphere resulting in a mid-line shift and infiltration of the cortex (Fig 7A). Brightfield images merged with fluorescence images of tumor cells (Fig 7B) and BPNP (Fig 7C,D) provided composite images (Fig 7A,E). Significant deposition of BPNP at the site of infusion was observed, with BPNPs radiating in preferred directions toward the tumor periphery. BPNP were rarely observed in tumor regions that extended into the cortex (Fig 7A). BPNP concentration was highest at the infusion site of, but high concentrations were also observed in parts of the peritumoral region (Fig 7D), and were typically confined to the lateral aspects of the tumor (Fig 7D, E). Microscopic analysis revealed some tumor boundaries with BPNP accumulation in the peritumoral tissue, and low BPNP presence in the adjacent tumor (Fig 7F). High concentrations of BPNP were found near the site of infusion (Fig 7G, H).

Figure 7.

U87-GFP 26 day intratumoral CED infusion. Coronal cryosection at site of infusion was imaged via (A) merged brightfield with (B) U87-GFP 26 day tumor (green) and (C) NPs (red). (D) Particle concentration is highest at the site of infusion, with limited NP penetration in the tumor and leakage to the peritumoral regions. (E) Analysis of the tumor periphery indicates defined borders lining NP infiltration. (G, H) High concentrations of intertumoral NPs are observed near the site of infusion. Colored boxes on (E) demonstrate locations of images (F-H). White line denotes tumor border. Scale bar = 2 mm. Images F, G and H taken at 10X magnification.

Co-localization of NPs in tumor tissue

Image analysis was used to quantify the percentage of BPNP localized in the tumor tissue as compared to the surrounding normal brain tissue (Fig 8). CED infusions led to an intracranial volume of distribution of ~50 mm³, which was consistent for all infusions in animals with tumors (Fig 3). This infused volume of BPNP was distributed through tumor and non-tumor tissue; the fraction residing in the tumor depended on tumor type and size. In small tumors (5 mm³), a substantial fraction of the tumor was co-localized with BPNP: 4 mm³ of RG2 tumors (80%) or 3 mm³ of U87 tumors (60%). But the majority of the infused BPNP volume (36-40 mm³) was in the tissue surrounding the tumor. In large tumors (80-100 mm³), the infused BPNPs were co-localized in roughly equal proportions with tumor and non-tumor tissue. Only a fraction of the large tumors were covered by infused BPNP: 25 mm³ of the 80 mm³ RG2 tumors (31%) and 18 mm³ of the 100 mm³ U87 tumors (18%).

Figure 8.

Calculated co-localization of NP distribution in tumors. Imaging software was utilized to analyze the overlap of tumor and NPs and the percent co-localization of NPs within each tumor was measured. V_t is the volume of the tumor in mm³ and the sizes of the inner circles represent the tumors relative sizes. Using the total volume of distribution measured, calculated volumes V_in and V_out were derived. The total coverage of NPs in the tumors demonstrates that small tumors have more total tumor coverage by NPs than large tumors. However, it indicated that large tumors, both U87 and RG2, display a higher amount of administered NPs in the tumor as compared to smaller tumors.

Discussion

In this report, we investigated BPNP distributions in the presence of intracranial tumors after administration by CED. We observed that when injected directly into an intracranial tumor mass, BPNP distribution after CED is non-uniform: the extent of non-uniformity varies with size, degree of necrosis, and other properties of the tumor. Our most important finding is that a significant fraction of BPNPs escaped from the tumor volume during CED, accumulating in the peritumoral region, which resulted in partial tumor coverage by BPNPs. This phenomena was observed even in cases where the tumor volume was substantially smaller than the BPNP volume of distribution.

Over the past decade CED has moved from the laboratory to the clinic, and multiple clinical trials involving CED for the treatment of brain tumors are currently in progress [43]. The vast majority of these trials use CED of drugs in solution, with the exception of one trial currently testing liposomes loaded with irinotecan (ClinicalTrials.gov Identifier: NCT02022644). CED of polymeric nanoparticles has the potential to improve therapeutic outcome over CED of drugs in solution, by allowing the infusion of nanoparticles that slowly release the therapeutic agent, providing the opportunity for a dramatic increase in the duration of tumor cell exposure to the drug [21, 22, 44, 45]. Compared to liposomes, polymeric nanoparticles can take advantage of versatility in design permitted by synthetic polymer chemistry, which allows for control of factors that are important for the penetration and distribution of nano-scale materials in the brain, such as particle size or surface chemistry.

A major challenge in the practice of CED is to understand the factors controlling distribution of agents during infusion, and their dependence on the characteristics of the tissue. The potential value of a CED treatment is directly linked to the volume of distribution (V_d), which quantifies the volume of tissue covered by a detectable amount of tracer or drug after infusion. V_d can be measured by MR imaging (which is the method of choice in CED clinical trials [34]), autoradiography [15, 46], or image reconstruction of a fluorescent tracer [34]. For a range of infusion volumes (V_i) with drugs in solution, V_d has been found to be linearly dependent on V_i, [15, 47], and the ratio V_d/V_i, is commonly reported in the literature, to allow for comparison between studies using different V_i. It is important to acknowledge that the linearity of V_d/V_i, obtained for CED of drugs, has not been confirmed for polymer nanoparticles. A theoretical value of ~ 5 has been estimated for agents that freely distribute through the extracellular space of the brain, such as most small molecules, as well as for non-binding macromolecules such as albumin, due to the brain extracellular porosity of ~ 0.2 [48]. We verified here that the infusion of Evans blue labeled albumin yielded a V_d/V_i of 4.2, consistent with the theoretical value of 5. Downward deviations in V_d/V_i can be due to characteristics of the infused agent, such as molecular weight or particle size, as well as properties of the tissue, including relative proportion of grey and white matter, or the presence of a tumor mass [20]. Measured values of V_d also depend strongly on technical parameters, such as catheter positioning or the presence of backflow during the infusion [49], and on the imaging technique and the analysis method. In our previous study, we showed that PLGA nanoparticles distributed homogeneously after CED, and that formulation with trehalose to prevent particle aggregation further increased V_d to yield a V_d/V_i of up to 3.7 for particles with a diameter of 70 nm. Here the infused PLGA nanoparticles presented a diameter of 80 nm and yielded a V_d/V_i of 2.4, slightly lower than previously obtained values. However, in this previous study, a high concentration of trehalose (50%) was used as a cryoprotectant before lyophilization, possibly providing enhanced distribution of the particles due to the generation of osmotic pressure in the interstitial space. Here, only 30% trehalose was used in order to provide non-aggregating properties, and the diameter of the particles was 10 nm larger, which could account for the slightly decreased distribution. Overall, our particles appeared to distribute with a V_d/V_i in the range of the values obtained previously for particulate systems.

Interestingly, when we examined the distribution of polymeric BPNP after CED within brains bearing one of two intracranial tumors, RG2 or U87, at two different stages of growth, we obtained V_d/V_i values that were similar to what we obtained in the healthy brain, and V_d/V_i did not change significantly with tumor size (Fig 3). However, the distribution of our BPNP was heterogeneous throughout each tumor mass, with a preferential accumulation in the peritumoral space. These findings were similar to those obtained for liposomes infused in different preclinical models of intracranial tumors (C6, 9L-2, and U87) [32, 34], where volumes of distribution in brains with tumors and without tumors were comparable. These results suggest that as long as NPs can penetrate through healthy tissue, the final volume of distribution will be dictated by the volume injected (V_i).

The heterogeneity we observed in BPNP distribution in tumors appears to be due to heterogeneity in the tumor itself, and the presence of regions with high resistance to flow, resulting in particle movement preferentially through some pathways in the tumor mass, modifying the shape of the distribution volume. This is particularly true in the case of large tumors, which show decreased nanoparticle accumulation in the tumor together with an increased penetration into healthy tissue, anterior and posterior to the tumor (Fig 3B). Reduced accumulation in the tumor may be due to its high cellular density (Supp Fig 2 and 3), inducing elevated interstitial hypertension [50], and increased tortuosity of flow pathways in local environment [27]. All of these features potentially increase resistance to convection in the tumor environment and direct flow into healthy surrounding brain regions, where there is a decreased resistance. We also observed that once in the peritumoral region, the BPNP poorly penetrated the tumor mass, particularly in U87 tumors. This apparent resistance to BPNP movement from peritumoral region into the tumor tissues might also be due to the presence of a capsule-like structure around the tumor, which would hinder external penetration [51] in addition to the tumor interstitial hypertension.

These observations raise the issue of adequate tumor coverage by infused BPNPs. To address this question, image analysis was used to calculate the percentage of NPs located within the tumor, peritumoral, and normal brain regions (Fig 8). Large tumors received a higher amount of administered BPNP as compared to small tumors, but also demonstrated low fractional coverage of the tumor due to their large size and limited BPNP penetration. This finding suggests that in the case of large tumors, multiple injections at different sites might be necessary to cover the whole tumor mass. But even in the case of small U87 and RG2 tumors, which had volumes significantly smaller than the volume of distribution of BPNP, only 59% and 77% of the tumors were covered by particles, respectively (Fig 8). BPNP penetration into the peritumoral space was observed in most tumors, providing for a high concentration of NPs at the tumor periphery (Fig 4 and 6). Although accumulation of BPNP in the peritumoral region is probably beneficial to kill invasive satellite tumor cells which tend to promote tumor recurrence post-treatment [33], these results suggest that strategies are needed to improve the BPNP distribution in the tumor tissue and allow more complete coverage of the tumor mass. Finally, infusions into RG2 tumors with heavily necrotic cores demonstrated widespread distribution of NPs within the tumor interior (Fig 5 and Supp Fig 5). Necrotic zones, with low hydraulic resistance due to decreased cell density, provide pathways that might lead to increased accumulation of BPNP, producing reservoirs for sustained release of therapeutic agents inside the tumor mass. This factor may be particularly relevant for the treatment of non-resectable human GBMs, which often contain necrotic regions [52].

Predictive mathematical models of nanoparticle distribution during CED would facilitate translation of this technology into clinical practice. Modeling algorithms, such as the CED modeling software iPlan Flow, are already used clinically to plan the infusion of drugs in solution [53]. As a result, multiple injection sites are used to allow the full coverage of the desired volume of tissue, which is personalized to each patient. Similarly, infusion at multiple locations, strategically selected with respect to the tumor geometry, would produce more significant coverage of large tumors, and balance nanoparticle accumulation in the tumor periphery and core [33]. Recent mathematical models also provide some understanding of intratumoral and peritumoral transport of small molecules during CED [54], but these models must be refined in order to be applicable to the transport and distribution of infused nanoparticles. This work represents a first step in identifying the mechanisms of particle transport in tumors, which we hope will soon be translated into useful models.

Conclusions

The present study investigated the presence of a tumor mass on the distribution of polymer BPNP after CED. Intratumoral infusion of BPNP produced a similar volume of distribution as in normal brain, despite the presence of a tumor. However, BPNP demonstrated both anisotropic distribution at the site of tumor infusion, as well as limited tumor infiltration in RG2 and U87 tumors revealing the need for new strategies allowing for homogeneous distribution and full coverage of the tumor mass. With the availability of imaging technologies, such as pre-surgical and intra-operative MRI scans to determine tumor structure and location, as well as an understanding of nanoparticle deposition by CED from studies such as those presented here, future pre-clinical and clinical studies may better take advantage of the potential therapeutic benefits of polymeric NP and CED.