Ultrasmall Core-Shell Silica Nanoparticles for Precision Drug Delivery in a High-Grade Malignant Brain Tumor Model

Despite breakthroughs in the treatment of solid tumor malignancies, treatment options for primary gliomas and metastatic central nervous system (CNS) tumors remain limited due, in part, to variable and poor transport of small-molecule drugs across the blood–brain barrier (BBB), a significant obstacle to the delivery, penetration, diffusion, and retention of drugs within CNS tumors (1, 2). Significant off-target effects contribute to dose-limiting toxicities of targeted agents that could otherwise offer significant therapeutic potential (3, 4). In an attempt to overcome these technical hurdles, a variety of nanodelivery vehicles have emerged that aim to abrogate acute toxicity while improving precision drug delivery, efficacy, and therapeutic index (5).

The introduction of targeted small-molecule inhibitors (SMI) has revolutionized cancer treatment for tumors with genomically defined sensitivities. However, as these patients live longer with successful treatment of systemic disease, there has been an increased incidence of brain metastasis and cases in which the CNS is the only site of uncontrolled progression (3, 4). Newer SMIs offer the promise of higher selectivity and potency, but BBB penetration and dose-limiting toxicity remain major limitations to successful therapy (6–13). This suggests that improving BBB penetration and distribution may augment the therapeutic potential for drugs previously thought to lack efficacy in the CNS. Moreover, current standard-of-care treatment approaches for glioblastoma, the most common primary brain malignancy, have led to a stagnation in cancer survival rates, with a median survival of less than 15 months and a 5-year survival rate of only 5% (14). Small-molecule pathway inhibitor drugs with promising preclinical data have failed to show efficacy against glioblastoma multiforme (GBM) in clinical trials of unselected populations to date (15, 16). For instance, dasatinib is a highly potent second-generation adenosine triphosphate–competitive inhibitor shown to be effective against multiple protein tyrosine kinases, including platelet-derived growth factor receptor (PDGFR) and the Src family of kinases, in vitro. However, clinical trial results with dasatinib have been disappointing, failing to demonstrate a survival benefit at clinically acceptable doses (17). Possible reasons include a poor understanding of Src signaling in GBM making enrichment of patient populations difficult (18), as well as the presence of drug efflux pumps (i.e., p-glycoprotein, breast cancer resistance protein) expressed in the CNS leading to low levels of dasatinib accumulation (19). For the long list of receptor tyrosine kinase inhibitors whose clinical use is limited by off-target or wild-type receptor toxicities, it is critical to explore mechanisms to improve the tumor delivery and therapeutic index of existing drugs (9–11).

The growing concern over limitations driving inefficient therapeutic delivery to CNS malignancies has led to an explosive growth in nanodelivery systems, with the hope that such agents may overcome impediments of limited drug transport across the BBB, high interstitial pressures, and diffusion of drugs beyond sites of initial deposition (20–23). Given their versatility and tunability, particle-based delivery vehicles can offer a viable solution—not only to potentially overcome these limitations but also to enhance the accumulation, distribution, and retention (ADR) of therapeutics at sites of disease. Specifically, the physicochemical characteristics, including size, shape, charge, and surface chemistries, of such materials during particle synthesis are known to modulate these and other biological properties. However, despite the enormous body of research focused on the development of a variety of CNS drug delivery vehicles including polymeric nanoparticles (24–26), iron oxide nanoparticles (27), gold nanoparticles (28–30), liposomes (31), micelles (32, 33), quantum dots (34), and silica nanoparticles (35), consensus has yet to be achieved on clinically promising design features (Supplementary Table S1). Nearly all nanodelivery vehicles are larger than 20-nm diameter, which can limit extravasation from pores of the tumor neovasculature and impede diffusion through the tumor interstitium (36). Furthermore, the imaging information derived from many of these systemically administered probes is largely of a qualitative nature, thereby precluding accurate assessments of probe uptake at sites of disease and off-target sites (e.g., liver) in a time-dependent manner to inform particle designs and tailor treatment planning efforts.

Although the introduction of targeting moieties on the surfaces of particle-based probes in the form of antibodies (or their fragments; ref. 37), peptides (38), and aptamers (39) has been shown to enhance target tissue penetration and distribution in brain tumor models, sufficiently detailed and time-dependent biological investigations needed to improve understanding of structure–function relationships for different classes of therapeutic platforms at disease sites are surprisingly scarce. Studies demonstrating increased accumulation and distribution of therapeutic platforms are also often limited to in vitro assessments and/or lack quantitation with respect to in vivo particle distributions throughout the bulk tumor area (40–42). In addition, it is not well understood how changes in particle surface chemistry affect therapeutic product accumulation, retention, tissue diffusion, and distribution over extended time intervals (>24 hours). It is therefore important to explore new high-resolution in vivo imaging strategies that can carefully assess the influence of surface modifications, such as the attachment of therapeutics or the conjugation of targeting ligands, on particle ADR and pharmacokinetics (PK) using clinically relevant models of CNS disease.

Herein, we employed a genetically engineered mouse model of glioblastoma (mGBM), known to effectively recapitulate the heterogenous breakdown of the BBB observed in patients (43). We aimed to determine the physicochemical factors impacting the ADR by modifying the surface chemistry of a clinically promising ultrasmall fluorescent core-shell silica nanoparticle, termed Cornell prime dots (or C' dots; refs. 44, 45). We assessed the feasibility of this particle probe as a dual-modality (PET-optical) platform to determine whether its distinctly favorable physicochemical features, pharmacologic properties, and modular nature could be exploited to overcome key biological limitations plaguing native small-molecule drugs. We chose to functionalize its surface with varying densities of cyclized Arg-Gly-Asp (cRGD) peptides to create cRGD-PEG-Cy5-C' dots (abbreviated cRGD-C' dots) for targeting α_v integrins, a well described target of tumor neovasculature and gliomas (46). Comparison of ADR across a range of targeting ligand densities attached to C' dots, as well as to nonbinding cyclic Arg-Ala-Asp scrambled peptide control particles (cRAD-C' dots), demonstrated an enhanced penetration, diffusion, and ADR for high ligand density cRGD-C' dots at extended time frames. Expanding on these findings, cRGD-C' dots were subsequently engineered for drug delivery, using previously described methods for the conjugation of SMIs to the C' dot surface (47). Dasatinib was utilized as a prototype SMI drug for the inhibition of PDGFR signaling and illustrated how the PK and other key biological properties of the free drug could be favorably altered by conjugating drug-linker constructs to the particle surface while preserving its activity. The work presented highlights the successive tailoring of particle surface chemistry to achieve a first-in-kind renally clearable, targeted delivery system—one that exhibits controlled pharmacologic properties, high intratumoral ADR, and in vivo target inhibition in a genetically engineered glioma model for precision drug delivery and treatment.

RCAS/tv-a derived glioma (DXFM) cells were plated at a density of 5 × 10⁵ cells per well in a 6-well plate and allowed to attach overnight in DMEM-high glucose media supplemented with 10% FBS, 1.8-g/L sodium bicarbonate, and 1% penicillin/streptomycin. Cells were subsequently treated with 100 nmol/L solutions of cRAD-C' dot or cRGD-C' dot containing varying ligand numbers (n = 6 and 15–20 and n = 6, 14, and 18, respectively) and allowed to incubate for 4 hours at 37°C. Following incubation, cells were washed twice in full media, transferred to flow cytometry tubes, and stained with 0.05 mg/mL DAPI for live/dead cell determination. Flow cytometry analysis was carried out in the MSKCC Flow Cytometry Core Facility on a BD LSRFortessa flow cytometer (BD Biosciences). Results of triplicate samples were displayed as percentage of Cy5+ (C' dot+) cells or median fluorescent intensities using Prism 7 software (GraphPad).

Similar methods to those described above were used for temperature-dependent C' dot uptake studies. cRGD-C' dots with varying ligand densities were again incubated with DXFM cells in 6-well plates. However, individual plates were incubated at 4°C, 25°C, or 37°C for 4 hours. Following incubation with the C' dots, cells were prepared for analysis in a manner analogous to that mentioned previously.

Whole brains mounted in Tissue-Tek O.C.T were sectioned at a thickness of 10 μm on an Avantik Cryostatic Microtome (Avantik Biogroup). Sections were evaluated for tumor based on altered cellular density and vascular permeability determined by visualization of the Hoechst and FITC-Dextran fluorescence signals. Tumor was confirmed present on sections via hematoxylin and eosin (H&E) staining. Images were captured on a BX60 fluorescence microscope (Olympus America Inc.) equipped with a motorized stage (Prior Scientific Instruments Ltd.) and CC12 camera (Olympus) at 10× magnification. Whole-brain images were obtained by acquiring numerous fields that were aligned using MicroSuite Biologic Suite (version 2.7, Olympus America Inc.). Brightfield and fluorescent images were acquired using the appropriate filter cube sets.

Tumor diffusion calculations were carried out using ImageJ software (Ver. 2.0-rc-43/1.51h). Briefly, both H&E and fluorescence images were first set to the proper scale. Regions of interest (ROIs) were drawn around the entire tumor (H&E) as well as any detectable Cy5 signal (optical). Dividing the measured Cy5 area by the total tumor area and multiplying by 100 provided the percentage of tumor diffused with particle. These computed values were then plotted using Prism 7 (GraphPad).

For PET imaging, adult brain tumor–bearing mice (n = 6) were intravenously injected with 200 to 300 μCi (7.4–11.1 MBq) ⁸⁹Zr-cRGD-C' dots. Approximately 5 minutes prior to the acquisition of PET images, mice were anesthetized via inhalation of 2% isoflurane (Baxter Healthcare)/oxygen gas mixture and placed on the scanner bed; anesthesia was maintained using 1% isoflurane/oxygen gas mixture. PET imaging was performed in a small-animal PET scanner (Focus 120 microPET; Concorde Microsystems) at 0.5 and 24 hours postinjection. An energy window of 350 to 700 keV and a coincidence timing window of 6 ns were used. Data were sorted into two-dimensional histograms by Fourier rebinning, and transverse images were reconstructed by filtered back projection into a 128 × 128 × 63 (0.72 × 0.72 × 1.3 mm³) matrix. The PET imaging data were normalized to correct for nonuniformity of response, dead time count losses, positron branching ratio, and physical decay to the time of injection; no attenuation, scatter, or partial-volume averaging corrections were applied. The counting rates in the reconstructed images were converted to activity concentrations (percentage injected dose per gram of tissue, %ID/g) by use of a system calibration factor derived from the imaging of a mouse-sized water-equivalent phantom containing ⁸⁹Zr. ROI analyses of the PET data were performed using IRW software.

At 72 hours post–intravenous administration of ⁸⁹Zr-cRGD-C' dots, whole-brain tissue samples were collected and frozen in Tissue-Tek O.C.T for sectioning. As described previously, the presence of brain tumors was first confirmed via H&E, followed by imaging of ⁸⁹Zr-cRGD-C' dot accumulation via fluorescence microscopy (Cy5) and autoradiography (⁸⁹Zr). Autoradiography signal intensity and total tumor area were calculated by ROI analysis using ImageJ software (Ver. 2.0-rc-43/1.51h) across all subjects and plotted using Prism 7 software (GraphPad).

Frozen sections were air dried for 30 minutes at room temperature (RT), followed by fixation in 3% paraformaldehyde for 15 minutes at RT in a humidified chamber. Sections were then washed 2× with wash buffer (1X TBS/0.1% Tween-20) for 5 minutes. Endogenous peroxides were blocked using BLOXALL (SP-6000, Vector Labs) for 10 minutes at RT. Sections were then washed 2× with wash buffer for 5 minutes, prior to the addition of the blocking solution (5% normal goat serum/wash buffer). Blocking solution was incubated with sections for 1 hour at RT. Blocking solution was removed from sections and anti-pS6RP (#2211, Cell Signaling) primary antibody was added at a dilution of 1:200 in blocking solution. Sections were incubated overnight at 4°C in a humidified chamber. Sections were then washed 3× with wash buffer, followed by addition of antirabbit secondary antibody (BA-1000, Vector Labs) at a dilution of 1:300 in blocking solution and incubated at RT for 30 minutes. Tissue sections were then washed 3× for 5 minutes. Staining detection was performed using Vectastain ABC HRP Kit (PK-4000, Vector Labs) and ImmPACT DAB Peroxidase Substrate (SK-4105, Vector Labs) according to the manufacturer's instructions. Sections were then counterstained using VECTOR Hematoxylin QS (H-3404, Vector Labs), rinsed 2× in dH₂O, and coverslips mounted using VectaMount AQ Aqueous Mounting Medium (H-5501, Vector Labs). Whole-section brightfield images were acquired as described previously in these methods. Signal intensity quantification was carried out on threshold-adjusted images using ImageJ analysis software and adaptation of previously reported methods (48). Briefly, the sum of signal intensities across five high-power fields within the tumor area, as delineated by H&E staining, was acquired. These values were then averaged and corrected for the area of the ROI. Finally, the data for treated sections (45 μmol/L cRGD-Das-NDC) were normalized to the untreated control group.

Newer generation Cornell Prime dots (C' dots) manufactured in water-based environments (45, 49, 50), enable exquisite control over particle architecture and composition. Ultrasmall C' dots are core-shell hybrid silica nanoparticles composed of a rigid silica core encapsulating deep-red/near-infrared fluorescent dyes. The particle surface is covalently modified by polyethylene glycol (PEG) chains to confer in vivo stability and facilitate renal clearance (refs. 50, 51; Fig. 1A). Targeting moieties such as cyclic arginine-glycine-aspartic acid-d-tyrosine-cysteine (cRGD) are conjugated to the surface of C' dots for active cancer targeting (refs. 44, 52; Fig. 1A). C' dots exhibit homogeneous particle morphology (ref. 45; Fig. 1B), narrow size dispersion (ref. 53; Fig. 1C and D), and high product purity (refs. 45, 54, 55; Fig. 1C). The average hydrodynamic size of C' dots is around 6 to 7 nm by fluorescence correlation spectroscopy (FCS, Fig. 1D), endowing C' dots with unique pharmacokinetic properties, as previously described (44, 52). C' dots can be surface-conjugated with radioisotopes, e.g., ¹²⁴I or ⁸⁹Zr, for serial, quantitative PET–CT imaging of time-varying particle distributions and ADR (ref. 55; Fig. 1A). Small-molecule drugs have been attached to C' dot surfaces via a peptide-based enzyme-cleavable linker group (47). Resulting C' dot NDCs exhibit efficient drug release kinetics in the presence of model proteases and are stable in various biological media, rendering C' dots a promising drug-delivery platform for cancer therapeutics (Fig. 1A). However, in vivo results of such C' dot based NDCs have not been reported to date.

Integrins are well-known transmembrane receptors recognized to be overexpressed in a number of tumor types, as well as neoangiogenic vessels (56, 57). To demonstrate integrin expression in the RCAS/tv-a–driven model of glioblastoma, we conducted immunofluorescence staining on murine brain tumor tissue sections obtained from mice with confirmed GBM. The results depicted in Fig. 2A demonstrate a high level of α_v integrin (red) expression throughout the tumor area, as well as along the endothelium of the tumor neovasculature, defined by CD31 expression (green). Utilizing mGBM single-cell suspensions derived from murine RCAS/tv-a GBM, we next sought to understand (i) whether an integrin-targeting peptide was necessary to increase cellular binding and internalization of the particles and (ii) optimal ligand densities needed to maximize targeted uptake. To answer these questions, C' dots were functionalized with varying densities of the α_v integrin-targeting cyclic-arginine-glycine-aspartic acid peptide (cRGD; n = 6–18) or scrambled peptide control, cyclic-arginine-alanine-aspartic acid (cRAD; n = 6–20). Results demonstrated that high cRGD ligand densities (n = 14, 18) led to an increase in the percentage of cells staining positively for cRGD-C' dots (Cy5+; Fig. 2B) and an amplified cellular median fluorescence intensity (Supplementary Fig. S1A) as determined by flow cytometry. Minor increases in positively staining cells (Fig. 2B) and median fluorescence values (Supplementary Fig. S1A) were observed for cRAD-C' dots as a function of increasing ligand density; however, the percentage and median intensity values were significantly less than those observed with cRGD-C' dots. The enhanced cellular binding of cRGD-C' dots demonstrated the importance of an active targeting moiety and thus became the focus of subsequent studies.

We next sought to determine whether cRGD-C' dots were internalized nonspecifically in mGBM cells, or through an energy-dependent pathway, such as receptor-mediated endocytosis. To demonstrate that C' dots were internalized via endocytosis, cells were incubated with 100 nmol/L solutions of C' dots with varying cRGD ligand densities. Incubation was performed for 4 hours at 4°C, 25°C, or 37°C, and subsequently analyzed via flow cytometry. Results indicated that reductions in incubation temperature led to a decrease in both the number of Cy5+ cells (Fig. 2C) and the cellular median fluorescence intensities (Supplementary Fig. S1B), suggesting a receptor-mediated mechanism of cRGD-C' dot uptake (46). Finally, visualization of cRGD-C' dot (n = 18) intracellular accumulation was performed via high-resolution confocal microscopy. mGBM cells treated with 100 nmol/L cRGD-C' dots for 4 hours showed robust accumulation of Cy5 signal in cells, likely localized to endolysosomes as shown for prior cancer cell types (58), with a predominant perinuclear distribution (Fig. 2D; Supplementary Fig. S1C) when compared with cRAD-C' dot (n = 15–20)–treated GBM cells (Fig. 2E; Supplementary Fig. S1D). Together, these results confirmed that cRGD-C' dots underwent internalization, and that this occurs primarily through energy-dependent mechanisms of uptake.

Overall, these data indicated that the inclusion of an integrin targeting peptide (cRGD) on the C' dot platform led to an increase in the level of cellular binding and internalization in mGBM cells. Observed increases were found to be dependent on targeting ligand density, where 18 cRGD peptides per particle were shown to maximize intracellular accumulations. This ligand density was therefore selected for in vivo studies.

Utilizing an experimental design that employed concurrent intravital stains and timed injections to create a real-time map of the BBB, we were able to evaluate particle distributions relative to the BBB at various time points posttreatment. This allowed for a noninvasive, quantifiable, and time-dependent method for assessing distribution and brain tumor penetration beyond that attributable to the enhanced permeability and retention (EPR) effect, a property characterized by preferential delivery of drugs to tumors (59). Animal models of GBM were created using well-established RCAS/tv-a methodologies (ref. 43; Supplementary Scheme 1A). Mice with confirmed mGBM were treated systemically with cRAD- or cRGD-C' dots and sacrificed 3 hours posttreatment (Supplementary Scheme 1B). We utilized a 70 kDa FITC-labeled dextran molecule (FITC–dextran) that is of comparable size to the C' dot as a measure of BBB permeability (60, 61) At 3 hours postinjection, tumor-specific Cy5 signal (red) distribution closely corresponded to that observed for the intravenously administered permeability marker, FITC–dextran (green), demonstrating strong colocalization in regions of disrupted BBB breakdown (Supplementary Fig. S2A and S2B). These results were equivalent for both cRAD- (Supplementary Fig. S2A) and cRGD-C' dot–treated animals (Supplementary Fig. S2B), suggesting that the in vivo distribution at this early time point is largely dependent on the EPR effect without additive effects from integrin targeting. High magnification imaging (60×) was performed for subcellular localization of cRGD-C' dots (Supplementary Fig. S2C). These high-resolution images demonstrated strong Cy5 signal (red) in close proximity to nuclei stained by Hoechst (blue), suggestive of intracellular localization of cRGD-C' dots. Furthermore, colocalization of FITC-dextran (green) with particles suggested that cellular internalization was occurring through the endolysosomal pathway (62).

A qualitative comparison of particle uptake between cRGD- and cRAD-C' dot–treated mGBM showed an increase in the ADR of the targeted particle at 96 hours postadministration (Fig. 3A and B, respectively). Fluorescence microscopy demonstrated enhanced particle distribution (Fig. 3A2 and A6) across the total tumor area, as well as increased distribution beyond areas of BBB breakdown (Fig. 3A3 and A7), for cRGD-C' dots in comparison to cRAD-functionalized particles (Fig. 3B). Furthermore, ¹²⁴I-labeled cRGD-C' dots (Fig. 3A4) demonstrated increased particle accumulation relative to cRAD-C' dots (Fig. 3B4) in areas of disease, as shown by MRI (Fig. 3A8 and 3B8, respectively). Together, these results provided qualitative support for the use of a cRGD targeting moiety on the C' dot surface as a means of increasing particle ADR.

A quantitative approach utilizing region of interest (ROI) analysis was employed to measure fluorescence signal in relation to the total tumor area. Clear enhancement of cRGD-C' dot distribution and retention within the defined tumor area was observed when compared with that shown for cRAD-C' dot–treated tumors at 96 hours postinjection (Fig. 4A). Comparison of ROI areas for cRAD- and cRGD-C' dot signal (Cy5), in relation to the total tumor area, as defined by H&E staining, demonstrated a significant enhancement of coverage offered by cRGD-C' dots over that of cRAD-C' dots (Fig. 4B).

We next examined the extent to which cRAD- and cRGD-C' dots distributed beyond areas of BBB breakdown, as demonstrated by the diffusion pattern of FITC–dextran, within GBM tumors (Fig. 4A, right column). Highlighting areas where fluorescent particles were free to diffuse as a result of an altered BBB made it possible to measure any enhanced distribution of C' dots. Results of this analysis revealed a trend in the data, with cRGD-C' dots distributed more uniformly throughout the tumor volume than that observed for FITC–dextran alone, which showed a more locally distributed, punctate pattern of particle signal. In comparison, the distribution of cRAD-C' dots closely mimicked that of the permeability marker across all mice (Fig. 4C). The slope of each line depicted in Fig. 4D was then calculated, allowing for a measure of enhanced coverage beyond that of BBB breakdown across groups. This analysis demonstrated a >4.5-fold increase in cRGD-C' dot coverage (slope = 45.9 ± 10.4) relative to that seen for cRAD-C' dots (slope = 9.8 ± 8.2). This indicates that cRGD-C' dots distribute to a much greater degree, and beyond areas that are freely diffusible, than that found for cRAD-C' dots. Finally, the percentage of Cy5 coverage was evaluated on the basis of the average FITC–dextran coverage across each treatment group (Fig. 4E). This result again demonstrated that the addition of cRGD to the C' dot platform enhanced coverage of the total tumor area to a greater extent than that observed for the scrambled (cRAD-C' dot) control.

These results provide quantitative support to the observed increase in cRGD-C' dot coverage relative to that found with cRAD-C' dots within mGBM. In addition, it was shown that α_v integrin-targeting C' dots were capable of distributing beyond areas of BBB breakdown, a property critically needed for augmenting particle-based therapeutic delivery and penetration of CNS malignancies.

Implementing newly described methods for radiolabeling C' dot platforms with ⁸⁹Zr (55), a radionuclide that residualizes stably within cells after internalization (63), we next evaluated the utility of ⁸⁹Zr-labeled DFO-cRGD-C' dots (i.e., ⁸⁹Zr-cRGD-C' dots) as a particle-based imaging tracer for serial PET imaging using the RCAS/tv-a mGBM model (Fig. 5). Representative maximum intensity projections of a single mouse at an early (0.5-hour) and a late (24-hour) time point demonstrated rapid renal clearance and retained tumor uptake (Fig. 5A; Supplementary Video S1A and S1B). High specific uptake (∼7% ID/g) was noted in mGBM at 24 hours postinjection (Fig. 5A, bottom row), with a dramatic decrease in whole-body signal when compared with imaging performed at 0.5 hours (Fig. 5A, top row). Whole-brain ex vivo analysis of ⁸⁹Zr-cRGD-C' dot–treated mice demonstrated specific particle accumulation in areas of tumor visualized by both Cy5 and ⁸⁹Zr signal localization (Fig. 5B and C). Finally, particle activity observed on autoradiography showed a positive correlation with overall tumor area (Fig. 5D) and was dramatically higher than that found in normal brain (Fig. 5E).

Together, these data demonstrate that ⁸⁹Zr-cRGD-C' dots can be utilized for molecularly targeted PET imaging of the RCAS/tv-a mGBM model. Radiolabeled particles demonstrated “target-or-clear” capabilities with rapid renal clearance (55), relative lack of reticuloendothelial system uptake (i.e., liver), and retained brain tumor uptake. Favorable pharmacokinetic and biodistribution profiles show increased uptake of ⁸⁹Zr-cRGD-C' dots in areas of tumor, thus resulting in enhanced tumor-to-normal brain uptake ratios. These results highlight the utility of ⁸⁹Zr-cRGD-C' dots as a viable PET tracer for the imaging of CNS malignancies.

Implementing previously described synthetic methods to link SMIs to the C' dot surface (47), dasatinib was successfully conjugated to the integrin-targeting particle probe to create a theranostic, cRGD-C' dot nanoparticle drug conjugate, or cRGD-Das-NDC. cRGD-Das-NDC was synthesized utilizing a methylethylenediaminocarbamate attached to a dipeptide (Phe-Arg; Supplementary Figs. S3 and S4A), such that a protease can liberate the terminal primary amine of the methylethylenediamine (Supplementary Fig. S4A2). Next, an intramolecular cyclization occurs (Supplementary Fig. S4A3), leading to the subsequent release of dasatinib and a methyl–urea by-product (Supplementary Fig. S4A4). As expected, this linker design exhibits a two-step release mechanism: a very fast initial protease-dependent event, and a much slower pH-dependent intramolecular cyclization event that results in the release of the native dasatinib drug conformation.

We sought to demonstrate that dasatinib released from the NDC retains its efficacy by performing Western blot analysis of glioma cells, derived from both RCAS/tv-a and human tumors, following treatment with cRGD-Das-NDCs (Supplementary Fig. S4B–S4D). Cells were exposed to increasing concentrations (10 nmol/L to 10 μmol/L) of cRGD-Das-NDC or the native dasatinib and subsequently stimulated with PDGF-BB. Investigating the ability of cRGD-Das-NDCs to inhibit PDGFR signaling in DXFM cells, we successfully demonstrated reductions in the levels of p-Akt and p-S6RP (Supplementary Fig. S4B), as well as p-Src and p-PRAS40 (Supplementary Fig. S4C). The reduction of phosphorylation levels occurred across all markers in a dose-dependent manner and was similar to those observed when using the native drug. In addition to effective inhibition of PDGFR signaling in cells derived from a murine origin, we also chose to evaluate the effectiveness of cRGD-Das-NDCs in a human glioma model by the use of TS543 neurosphere cultures (Supplementary Fig. S4D). Again, treatment of TS543 cells with cRGD-Das-NDCs demonstrated a dose-dependent reduction in the levels of p-PDGFRα that was equal to, or better than, results observed following treatment with free dasatinib.

Together, these results confirm the effectiveness of cRGD-Das-NDCs in their ability to inhibit PGFR signaling by reducing the level of active effectors in both PI3K (i.e., Akt, PRAS40, S6RP) and Src pathways. In addition, these effects were confirmed in both murine and human-derived models of glioma, thus demonstrating an effective delivery platform that performs irrespective of species or genetic background.

Dasatinib-functionalized NDCs, both targeted (cRGD-) and nontargeted (cRAD-), were first evaluated for their ability to replicate the tumor distribution profiles that had been observed when utilizing drug-free cRGD- or cRAD-C' dot platforms. Tumor distributions of cRGD- and cRAD-Das-NDC particles were again analyzed using fluorescence microscopy 96 hours after NDC administration. The data generated showed enhanced coverage of the tumor by cRGD-Das-NDCs (Fig. 6A1–A4), as against that observed with cRAD-Das-NDCs (Fig. 6A5–A8). These results corroborated our previous findings of targeting moieties' ability to enhance tumor penetration, accumulation, distribution, and retention of C' dots in comparison to nontargeting particle controls. Subsequent studies evaluating the biological properties of NDCs in vivo were therefore performed using cRGD-Das-NDC particles.

We next sought to evaluate the in vivo efficacy of cRGD-Das-NDCs to inhibit downstream pathway signaling in our murine model of GBM. Previously, we demonstrated that our drug-linker construct is capable of being cleaved from C' dots by enzymes such as cathepsin B (47). Interestingly, cathepsin B expression has been shown to increase in higher grade gliomas (64), is found in lysosomes, on cell surface membranes, involves cancer cells and endothelial cells comprising brain tumors, and may be secreted into the extracellular milieu (65, 66). We therefore administered a single dose of 45 μmol/L cRGD-Das-NDC to mice harboring RCAS/tv-a mGBM. Twenty-four hours after administration, whole-brain tissues were extracted and subjected to IHC staining for the detection of p-S6RP, a downstream regulator of the PI3K pathway (67). Visual observation showed an appreciable reduction in the p-S6RP staining intensity between untreated control mice (Fig. 6B, top row; Supplementary Fig. S5A) and those treated with 45 μmol/L cRGD-Das-NDC (Fig. 6B, bottom row; Supplementary Fig. S5B). ROI analysis across all treated and control mice confirmed that there was a significant reduction in staining intensity as a result of cRDG-Das-NDC treatment (Fig. 6C).

These results confirm that cRGD-Das-NDCs retain an increased ADR over the cRAD-functionalized NDCs (cRAD-Das-NDC). In vivo administration and successful delivery of dasatinib were further confirmed by p-S6RP IHC, where an appreciable inhibition of phosphorylation was observed across all treated mGBM in comparison with untreated controls. It should be noted that we do recognize the time discrepancy between the analysis of this in vivo study and our previous in vitro work to demonstrate release of effective drug. We attribute the more rapid appearance of functional dasatinib in brain tumors to be a result of tumor microenvironment components that may aid in cleavage and cyclization of the dasatinib precursor (68). Together, these results demonstrate that cRGD functionalization increases the accumulation, diffusion, and retention of C' dots in a model of mGBM that displays heterogenous BBB breakdown.

In this study, we have focused on the delivery of a drug carrier across a partially disrupted BBB rather than on the distinct challenges of crossing an intact barrier. A partially disrupted BBB reflects the majority of clinical cases of CNS tumor progression where contrast-enhancing tumors are detectable and growing. We used a tumor model that (i) is known to generate a heterogeneous BBB and (ii) is known to respond to the SMI dasatinib, providing a biomarker of target inhibition as an end effect of the NDC. We have aimed to characterize how functional surface modifications alter the behavior of the C' dot in the challenging microenvironment of a high-grade brain tumor. The glioma model provides a uniquely high contrast between tumor and normal brain tissue when assessing NDC delivery via fluorescence, autoradiography, or PET. Although our model is a CNS primary tumor, we expect the results to be relevant to other tumor types that disrupt the BBB. Brain metastases may be a better target for NDC development in the specific case where the primary tumor is sensitive to small-molecule drugs. At the moment, no effective SMI has emerged for gliomas. Although our mouse glioma model is sensitive to dasatinib, we do not suggest that improvements in delivery of this particular inhibitor will benefit patients with gliomas. Within the larger efforts to develop “magic bullet” therapeutics for cancer, our current study is focused on the magic rather than the bullet.

With obstacles such as the BBB limiting the delivery of therapeutics to malignant brain tumors, the enhanced tumor distribution and drug-carrying capabilities of our platform may provide a means by which drug accumulation can be preferentially enhanced at sites of disease without significant off-target accumulations, thereby improving treatment efficacy and therapeutic index. An initial phase 1 imaging study is now underway to investigate cRGD-C' dot accumulation in patients diagnosed with primary gliomas or CNS metastases (Supplementary Fig. S6).

This study aimed to elucidate the delivery of C' dots to a mGBM model, along with their intratumoral and intracellular distributions, and to further investigate C' dots as a tunable platform for precision drug delivery. Our findings demonstrated that targeting of α_v integrins with cRGD increased the accumulation, retention, and diffusion of C' dots throughout the tumor interstitium, leading to a more uniform distribution of this platform within malignant brain tumor tissues, in particular, beyond regions of BBB breakdown. The modular design of C' dots allowed for multimodal imaging of these particles as both optical probes and radiotracers through encapsulation of fluorescent dyes and surface attachment of various radiolabels such as ⁸⁹Zr and ¹²⁴I.

The ultimate goal of this study was to assess the potential of C' dots as a cancer-targeted ultrasmall vehicle for precision drug delivery, penetration, and ADR at the target site while minimizing off-target effects. We have shown that SMIs can be attached to C' dots via a cleavable linker, with retention of in vivo pharmacologic activity against a specific target. Diffusion and intratumoral retention of NDCs suggest that C' dots can be tailored to efficiently deliver small-molecule therapies with enhanced PK. Given the current challenges of systemic therapies in their ability to penetrate heterogenous tumors, these results suggest a mechanism whereby precision drug delivery can overcome technical hurdles associated with existing small-molecule therapies and provide new opportunities for drugs with otherwise unacceptable systemic toxicity, PK, and/or target tissue uptake. These findings, while not optimized for dose or drug loading, lay the foundation for further investigation of C' dots as potent drug delivery vehicles to treat both primary and metastatic diseases to the CNS.

K. Ma and U. Wiesner hold ownership interest (including patents) in Elucida Oncology. U. Wiesner holds ownership interest (including patents) in and is an advisory board member of Elucida Oncology. M. Overholtzer is listed as a coinventor on provisional patent applications on C' dot nanoparticles that are owned by Memorial Sloan Kettering Cancer Center and licensed to Elucida Oncology. C.M. Rudin is an employee/paid consultant for AbbVie, Amgen, Ascentage, Bicycle, Celgene, Daiichi Sankyo, Genentech Roche, Ipsen, Loxo, Pharmamar, Vavotek, Harpoon, Bridge Medicines, and Astra Zeneca; reports receiving commercial research grants from Daiichi Sankyo; and is an advisory board member/unpaid consultant for Elucida. M.S. Bradbury reports receiving other commercial research support from, holds ownership interest (including patents) in, and is an advisory board member/unpaid consultant for Elucida Oncology. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R. Juthani, B. Madajewski, B. Yoo, F. Chen, K. Ma, P. Zanzonico, U. Wiesner, M.S. Bradbury, C.W. Brennan

Development of methodology: R. Juthani, B. Madajewski, B. Yoo, L. Zhang, P.-M. Chen, F. Chen, M.Z. Turker, K. Ma, M. Overholtzer, S. Carlin, P. Zanzonico, U. Wiesner, M.S. Bradbury, C.W. Brennan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Juthani, B. Madajewski, B. Yoo, L. Zhang, F. Chen, M.Z. Turker, K. Ma, M. Overholtzer, V.A. Longo, S. Carlin, C.M. Rudin, U. Wiesner, M.S. Bradbury, C.W. Brennan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Juthani, B. Madajewski, L. Zhang, P.-M. Chen, F. Chen, M.Z. Turker, K. Ma, M. Overholtzer, V. Aragon-Sanabria, J. Huse, M. Gonen, P. Zanzonico, U. Wiesner, M.S. Bradbury, C.W. Brennan

Writing, review, and/or revision of the manuscript: R. Juthani, B. Madajewski, B. Yoo, F. Chen, K. Ma, V. Aragon-Sanabria, J. Huse, M. Gonen, C.M. Rudin, U. Wiesner, M.S. Bradbury, C.W. Brennan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Juthani, K. Ma, U. Wiesner, M.S. Bradbury

Study supervision: R. Juthani, B. Madajewski, U. Wiesner, M.S. Bradbury, C.W. Brennan

We acknowledge that this study was funded by grants from the National Institutes of Health (1U54 CA199081-01, to M.S. Bradbury and U. Wiesner) and Sloan Kettering Institute (Core Grant P30 CA008748CCSG, to M.S. Bradbury).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.