Ultrasmall Nanoparticle Delivery of Doxorubicin Improves Therapeutic Index for High-Grade Glioma Open Access

Glioblastoma multiforme (GBM) is a highly aggressive brain tumor with poor prognosis and few available treatment options (1, 2). Despite considerable efforts to develop new classes of small-molecule drugs, prospects for patients with such aggressive tumors have not improved for decades, with approximately 7% of patients surviving 5 years after diagnosis (3). In addition to intra- and intertumoral heterogeneity promoting resistance and eventual recurrence (4), a principal challenge leading to central nervous system (CNS) treatment failures has been the reduced transport of potentially effective small-molecule drugs (i.e., chemotherapies) across the blood–brain barrier (BBB) and blood–tumor barriers (5, 6) accompanied by marked reductions in tumor tissue penetration, diffusion, and accumulation of such drugs. These factors contribute to an insufficient therapeutic window between efficacy and dose-limiting toxicities (DLT). The anthracycline, doxorubicin (DOX), used in this work, exhibits anti-neoplastic properties against a broad spectrum of tumor types (e.g., breast, lung, sarcoma), but its clinical use is limited by cardiotoxicity, bone marrow depression, nephrotoxicity, and other adverse effects (7). Potent cytotoxicity of DOX in gliomas has been confirmed in experimental studies leading to significant interest in its clinical potential and efforts to overcome a combination of poor BBB penetrance and narrow therapeutic window (8, 9). In this regard, DOX is exemplary of a class of small-molecule drugs that would benefit from the design and advancement of newer-generation drug delivery systems that can be engineered to better navigate complex biological barriers and improve therapeutic efficacy in highly aggressive CNS cancers, which remains a critically important unmet need.

Although extensive efforts have led to the development of a diverse array of both clinical and preclinical nanocarrier-based drug delivery systems (10), the vast majority of these systems are relatively large in size (i.e., at least 60 nm), which limits their uniform delivery, penetration, diffusion, and accumulation within the tumor interstitium; this results in only local responses following their extravasation and concentration along vessel walls (11). For instance, DOX has been encapsulated in a variety of large organic nanocarriers (e.g., liposomes) for the treatment of patients with malignant brain tumor but, in many cases, has shown only modest increases in overall survival (11, 12). The composition of different DOX-encapsulated nanocarriers, as well as free DOX and other selected therapies, used to treat high-grade gliomas, brain metastases, and other solid tumors are summarized in Supplementary Table S1.

Successful clinical deployment of DOX-containing constructs for treating CNS malignancies may have also been compromised by the predominant use of models with limited clinical relevance—either using in vitro systems or human GBM xenografts (i.e., U87, U251; refs. 13–15). Although these models can be exploited for rapid and reproducible results, they do not recapitulate the heterogeneity and pathophysiology of human glioma biology (16). This has raised additional concerns regarding the translational potential of the published results (17, 18).

Herein, we report on the feasibility of a first-in-kind DOX-conjugated ultrasmall (<8 nm) fluorescent core-shell silica nanoparticle that improves upon the biological, therapeutic, and toxicological properties of the native drug in clinically relevant models of high-grade glioma. Specifically, a platelet-derived growth factor-B genetically engineered mouse model (i.e., RCAS/tv-a PDGFB GEMM; ref. 19) and an EGF (amplified/mutant, EGFRvIII) patient-derived xenograft model (i.e., EGFR PDX) were used for this work; these models are known to recapitulate variable BBB disruption in patients (20). EGFR amplification and its active mutant, EGFRvIII, occur in approximately 40% of GBM (21), leading to activation of oncogenic signaling pathways (19) and the promotion of angiogenesis and more aggressive tumor growth (22).

The Cy5 dye-encapsulating silica core and poly(ethylene glycol; PEG) shell (core-shell) base nanoparticles synthesized in water, Cornell prime dots (or C′ dots; refs. 23, 24), have previously been surface-adapted with an array of functional components (25–27) and radiolabels (28) for conducting image-guided surgical (29, 30) and other oncologic (31) applications. For early in vivo evaluations, the platform was previously conjugated with prototype tyrosine kinase inhibitors (TKI), initially dasatinib, via an enzymatically cleavable linker (28) with a focus on target inhibition. In a subsequent study, we attached a first-generation TKI, gefitinib, for evaluating growth inhibition in a human non–small cell lung cancer (NSCLC) model (32). Here, we describe the following advanced studies: (i) rather than an inhibitor, with DOX, we successfully test a different class of therapeutic agent; (ii) we construct an azido-functionalized acid-labile hydrazone linker exhibiting pH-sensitive controlled drug release and serum stability; (iii) we use clinically relevant high-grade glioma models; (iv) we show a clear survival benefit of DOX-C′ dot conjugates over the free drug in efficacy evaluations; and (v) we demonstrate suppression of early-stage (i.e., 24 hours) cardiotoxicity. In addition to these substantial advances, screening pharmacokinetic (PK) and metabolic cage studies were conducted in naïve mice by PET imaging following particle radiolabeling with the radiometal, zirconium-89 (⁸⁹Zr). On the basis of the favorable PK properties obtained, including off-target (i.e., liver) accumulations of approximately 5%ID/g or less over a 24-hour period post-injection (p.i.), an optimal drug-per-particle ratio (DPR) of approximately 30 was identified for subsequent in vivo imaging and therapeutic studies in glioma-bearing mice.

Azido-PEG4-Hydrazone linker–DOX (Supplementary Fig. S2) was prepared as follows. Commercially available azido-PEG4-hydrazide or azido-PEG4-hydrazide-t-Boc was obtained (Conju-Probe and BroadPharm, respectively). Both compounds were treated with trifluoroacetic acid before proceeding. For example, 250 mg (0.617 mmol) of the t-Boc analog was treated with TFA containing 5% water for 1 hour at room temperature. LC-MS was used to monitor deprotection. The TFA was removed under reduced pressure, water/acetonitrile was added, and the solution then frozen and lyophilized overnight. The oil was dissolved in 50-mL anhydrous methanol, and then 450 mg (0.775 mmol) of DOX HCl (Carbosynth, Ltd.) was added. The reaction was monitored by LC-MS and allowed to proceed at room temperature overnight. The methanol was removed under reduced pressure and a red solid was isolated. The crude material is a mixture of Azido-PEG4-Hydrazone linker–DOX and free DOX and was used without further purification. HRMS C38H50N6O15: [M+H]+ calc. 831.3407, obs. 831.3405, [M+Na]+ calc. 853.3226, obs. 853.3222.

Base DBCO-PEG-Cy5-C′ dots or deferoxamine (DFO)-conjugated DBCO-PEG-Cy5-C′ dots were synthesized and characterized according to previously published procedures (23, 24, 27). The conjugation of pH-sensitive DOX–drug-linker was achieved via the highly specific strain-promoted azide-alkyne cycloaddition reaction between DBCO groups on (DFO-)DBCO–PEG–Cy5-C′ dots surface and azide group on the Azido-PEG4-Hydrazone linker–DOX. For a typical synthesis of DOX-C′ dots having a DPR of approximately 30, Azido-PEG4-Hydrazone linker–DOX was mixed with DBCO-PEG-Cy5-C′ dots in PBS at a reaction ratio of 100:1 and reacted at room temperature under shaking (650 rpm) overnight. Free unreacted Azido-PEG4-Hydrazone linker–DOX or free DOX were removed by PD-10 column purification (PBS buffer as the mobile phase). As synthesized, DOX-C′ dots (or DFO–DOX-C′ dots) were then characterized in terms of their hydrodynamic diameter, surface chemical properties, and drug-to-particle ratios using a combination of TEM (transmission electron microscopy), FCS (fluorescence correlation spectroscopy), and UV/Vis spectroscopy.

To study their pH-responsive cleaving capability, 100 μL (7.5 μmol/L) of DOX-C′ dots (DPR ∼30) were suspended in sodium acetate buffer solution (pH 5.2) or PBS buffer (pH 7.4). The mixture was kept at 37°C under shaking (650 rpm) for 48 hours. To quantify the percentage (%) of free DOX cleaved from DOX-C′ dots (i.e., % cleaved) under different pH conditions, integrated UV-Vis spectrum was generated from the cleaved DOX-C′ dot samples peak at selected time point, following the modified HPLC (high-performance liquid chromatography) method (32). That is, only the AUC of retained DOX, i.e., DOX conjugated to C′ dots, is evaluated using this integrated technique, not the free cleaved DOX itself. The AUC ratio between DOX (AUC_DOX) and C′ dots (AUC_{C′ dot}) at t = 0 and other time points were calculated. The percentage of cleavage rate (% cleaved DOX) was calculated using the following equation (e.g., using t = 24 hours, pH 5.2):

For serum stability studies, 100 μL (15 μmol/L) of DOX-C′ dots (DPR ∼30) in PBS were mixed with 100 μL of human or mouse serum and kept at 37°C on a shaking platform (650 rpm) for 120 hours. The AUC ratio between DOX (AUC_DOX) and C′ dots (AUC_{C′ dot}) was acquired using the modified HPLC method (32). The percentage (%) of retained DOX in the DOX-C′ dots was calculated using the following equation (e.g., using t = 24 hours):

⁸⁹Zr was produced at Memorial Sloan Kettering Cancer Center (MSKCC) on a TR19/9 cyclotron (Ebco Industries Inc.) via the ⁸⁹Y(p,n)⁸⁹Zr reaction, and purified to yield ⁸⁹Zr with a specific activity of 5.28–13.43 mCi/μg (470–1,195 Ci/mmol) of zirconium (33). Activity measurements were performed using a CRC-15R Dose Calibrator (Capintec). For the quantification of activities, experimental samples were counted on an Automatic Wizard (33) γ-Counter (PerkinElmer). All in vivo experiments were performed according to protocols approved by the MSKCC Institutional Animal Care and Use Committee. A purity of greater than 95% was confirmed using radio-TLC for all of the ⁸⁹Zr-labeled DFO–Dox-C′ dots.

DFO–DOX-C′ dots were synthesized by reacting the Azido-PEG4-Hydrazone linker–DOX with DFO–DBCO-C′ dots via highly specific strain-promoted azide-alkyne cycloaddition reaction. For a typical ⁸⁹Zr labeling, 0.75 nmol of as synthesized DFO–DOX-C′ dots were mixed with 1 mCi of ⁸⁹Zr-oxalate in HEPES buffer (pH 8) at 37°C for 60 minutes; final labeling pH was kept at 7–7.5 (34, 35). An EDTA challenge process was introduced to remove any non-specifically bound ⁸⁹Zr by incubating the mixture at 37°C for an additional 60 minutes with EDTA. The final ⁸⁹Zr labeling yield was in the range of 70% to 80% (n > 5). As synthesized ⁸⁹Zr-DFO–DOX-C′ dots were then purified using a PD-10 column. The final radiochemical purity was estimated to be greater than 99% by radio-TLC (data not shown).

EGFRvIII-derived glioma cells, derived from PDX tumors prepared by the MSK Brain Tumor Center (BTC; vide infra), were cultured using NeuroCult NS-A media from StemCell Technologies, Inc. RCAS/tv-a (Nestin-tv-a Ink4a-Arf^−/−)-derived glioma cells (kind gift from Dr. Eric Holland, MSK, NY) were maintained in DMEM–high glucose media supplemented with 10% FBS, 1.8 g/L sodium bicarbonate, and 1% penicillin/streptomycin. The human NSCLC line, ECLC26 was received as a generous gift from Dr. Charles Rudin (MSK, NY). MDA-MB-231 (Cat. # CRM-HTB-26) and PC-3 (Cat. # CRL-1435) human breast and prostate cancer cell lines, respectively, were obtained from the ATCC. ECLC26, MDA-MB-231, and PC-3 cells were cultured in RPMI-1640 supplemented with 10% FBS and Plasmocin with 1% penicillin/streptomycin (Media Preparation Facility, MSK). Cultures were maintained in a humidified incubator with 5% CO₂ at 37°C. All cell lines underwent routine Mycoplasma testing (MycoAlert Mycoplasma Detection Kit, Lonza). Work was conducted from the original cell stock for all cell lines. Moreover, all experiments within a study were run from the same passage from this original stock to ensure reproducibility.

To visualize intracellular delivery of DOX and DOX-C′ dots, RCAS/tv-a, ECLC26 or MDA-MB-231 cells were plated in 8-well chamber slides (μ-slide, Ibidi) at a density of 1 × 10⁴ cells per well and allowed to attach overnight before incubation with 500 nmol/L DOX or DOX-NDC for 24 hours. Following treatment, cells were washed 3x with PBS, counterstained with Hoechst (5 μg/mL) in PBS for 20 minutes, then washed 2x with PBS before incubating in Live Cell Imaging Solution (Thermo Fisher Scientific) for microscopy. Samples were imaged using a Zeiss LSM880 point-scanning confocal microscope equipped with an Airyscan, super-resolution detector (Molecular Cytology Core, MSKCC). Images were processed using Zen software (Zeiss) and displayed using Imaris Image Analysis Software (Bitplane).

Median lethal dose, LD₅₀, values were determined by incubating increasing concentrations of each treatment across the examined tumor cell types in opaque 96-well plates. Briefly, cells were plated at a density of 5 × 10³ cells per well and allowed to attach overnight in full growth media. Next, cells were treated with free DOX or DOX-C′ dots containing media [10⁻⁵–10¹ μmol/L (DOX concentration)] for 7 days. At the conclusion of treatment, cell viability was assessed using the CellTiter-Glo assay (Promega). LD₅₀ values were calculated in a manner identical to previously reported methods (32) and displayed using Prism7 software (GraphPad).

All mouse experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) at MSK, and conformed to NIH guidelines for animal welfare. Using well-established protocols, 4 × 10⁴ DF-1 cells (ATCC) expressing the RCAS-PDGF-B retroviral vector were introduced via intracranial injection into neonatal Nestin-tv-a Ink4a-Arf^−/− mice 3 days after birth or into 6- to 8-week-old adults, as previously described (20). Mice were monitored for signs of tumor burden (head tilt, lethargy, and hydrocephalus). Tumor burden was confirmed by T2-weighted MRI using a 4.7T/40-cm bore Avance magnet (Bruker) before the initiation of studies.

For PDX generation, murine intracranial tumors were initiated by the BTC staff under an IRB-approved protocol using low-passage patient-derived orthotopic xenotransplants; these were procured from the MSK BTC collection and were EGFR amplified/mutated. Four BTC PDX models were generated from primary patient tumors that were resected before therapy. Each sample was passaged directly from mouse to mouse without culturing or extensive manipulation beyond tissue dissection and dissociation. Stereotactic administration of minced fresh (3 μL) patient-derived Grades 3 and 4 glioma tissue was performed by injection into the striatum of anesthetized female NSG mice brains 3–4 weeks of age at coordinates: −0.5 mm AP, 1.5 mm ML, and 2.5 mm DV with respect to the bregma (N = 3 or 4 mice, respectively). Subsequent passages of PDX in murine brain were carried out in NSG or nude mice (male or female) with the same protocol.

Human subjects’ research for tumor procurement was carried out under an MSK IRB-approved protocol. Regardless of age, race, gender, or diagnosis, all patients who presented for brain tumor surgery at MSKCC were approached to consent voluntarily for permission to use non-essential tissues for research purposes. Specimens were derived from de-bulking surgery and were in excess of that needed for diagnostic purposes.

Serial PET imaging of glioma (and naïve) mice was performed over a 72-hour time interval using a small-animal PET scanner (Focus 120 microPET; Concorde Microsystems) following administration of the particle tracer, according to well-established procedures (35). Adult brain tumor mice (n = 3; and naïve mice) were then intravenously injected with 200–300 μCi (7.4–11.1 MBq) ⁸⁹Zr-DFO–DOX-C′ dots approximately 2–3 weeks after intracranial injection of PDX lines. Before image acquisition, mice were anesthetized using a 2% isoflurane (Baxter Healthcare)/oxygen gas mixture and placed on the scanner bed; anesthesia was maintained using 1% isoflurane/oxygen gas mixture. An energy window of 350−700 keV and a coincidence timing window of 6 ns were used. Data were sorted into 2D histograms by Fourier rebinning, and images were reconstructed using filtered back-projection into a 128 × 128 × 63 (0.72 × 0.72 × 1.3 mm³) matrix. PET imaging data were normalized to correct for non-uniformity 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 of the injected dose per gram of tissue, or %ID/g) using a system calibration factor derived from the imaging of a mouse-sized water-equivalent phantom containing ⁸⁹Zr. Region-of-interest (ROI) analyses of the PET data were performed using IRW software. Mice were sacrificed after the last PET imaging session (i.e., 72 hours p.i.), and tumor and major organs harvested for ex vivo radioassay analysis and fluorescence microscopy. Mouse organs were wet-weighted, counted in a Wizard2 γ-Counter (PerkinElmer), and converted to %ID/g (mean ± SD). Whole-brain tissue samples were collected and frozen in Tissue-Tek O.C.T for sectioning.

For autoradiography, 10-μm-thick frozen brain tumor tissue sections were exposed to a phosphor-imaging plate (Fujifilm BAS-MS2325, Fuji Photo Film) at −20°C. Following exposure, the phosphor-imaging plate was read at a resolution of 25 μmol/L using a Typhoon 7000 IP (GE Life Sciences) plate reader. Images were saved as both .tiff and .gel files for future analysis and processed using ImageJ software (https://imagej.nih.gov/ij/). The presence of brain tumors was first confirmed via hematoxylin and eosin (H&E), followed by imaging of ⁸⁹Zr-DFO–DOX-C′ dot accumulations using fluorescence microscopy (Cy5).

Time-activity curves were derived for each tissue and fit to exponential functions, and the fitted functions analytically integrated, accounting for radioactive decay, to yield the corresponding cumulative activity. Mouse organ absorbed doses were then calculated by multiplying the cumulative activity by the ⁸⁹Zr equilibrium dose constant for non-penetrating radiations (positrons), assuming complete local absorption of such radiation and ignoring the contribution of penetrating radiations (i.e., γ-rays). Mouse normal organ cumulated activities were converted to human normal organ cumulated activities by taking into account differences in total-body and organ masses between mice and humans (assuming a 70-kg standard human). Calculated human normal-organ cumulated activities were entered into the OLINDA dosimetry program to compute standard human organ absorbed doses (36). This human dosimetry model is a “normal” (i.e., tumor-free) anatomic model.

Three cohorts of mice (>5 mice/cohort) with confirmed brain tumors by T2-weighted MRI were intravenously injected via tail vein with 3 doses of DOX-C′ dots (7 mg/kg × 3), DOX alone (7 mg/kg), or saline vehicle in a total volume of 200 μL every 3 days. To ensure therapeutically equivalent drug dosing (i.e., ∼7 mg/kg) for each cohort, a DPR of approximately 30 and a C′ dot concentration of 60 μmol/L was used to derive an effective DOX concentration of 1,800 μmol/L for DOX-C′ dot preparations. Before and after final treatments (i.e., 96 hours), T2-weighted MRI scans were acquired for assessing changes in tumor volumes. Specifically, ROI analyses were performed by contouring areas of signal abnormality on a slice-by-slice basis through the lesion. Mice were sacrificed by carbon dioxide (CO₂) asphyxiation, and whole-brain specimens extracted and frozen in Tissue-Tek O.C.T compound (Sakura Finetek) for future histologic analyses (vide infra). For survival studies, RCAS/tv-a glioma mice (n = 10/group) underwent an identical dose schedule to that described above. Mice were monitored daily until they reached one of the following endpoints: (i) moribund condition; (ii) tumor mass >10% body weight; (iii) body weight loss >10% of the original weight; or (iv) body condition score of 1/5. Once an endpoint was met, the animal was recorded as an event in a Kaplan–Meier survival curve and displayed using Prism 7 software (GraphPad).

For IHC and immunofluorescence (IF) assays below, representative formalin-fixed and paraffin-embedded whole-brain and cardiac specimens were sectioned at a thickness of 10 μm on a Discovery XT System (Ventana Medical Systems). IHC for Ki-67 was performed on paraffin brain sections (4 coronal sections per brain) using a Leica Bond RX automated stainer. After heat-induced epitope retrieval in a pH 9.0 buffer, the primary antibody, rabbit monoclonal antibody clone D3B5 (Cell Signaling Technology #12202), was applied at a concentration of 1:500, followed by a polymer detection system, according to the manufacturer's instructions (DS9800, Novocastra Bond Polymer Refine Detection, Leica Biosystems). The chromogen used was 3,3′-diaminobenzidine tetrahydrochloride (DAB), and sections were counterstained with hematoxylin. IF staining was performed for γH2AX, Iba1, and CD206 on paraffin brain and/or cardiac tissue sections as follows. After heat induced epitope retrieval in a pH 9.0 buffer, the primary antibodies (anti-γH2Ax 0.2 mg/mL; Abcam, #11174, anti-Iba1 0.1 mg/mL; Abcam, #178847 or anti-CD206 0.5 mg/mL; Cell Signaling, #64693) were applied. Secondary biotinylated goat anti-rabbit IgG secondary antibody was subsequently added (5.75 mg/mL; Vector Labs, #PK6101). Application of streptavidin-HRP D (DAB Map kit, Ventana Medical Systems) was followed by incubation with Tyramide Alexa Fluor 488 (Invitrogen, # T20922), prepared according to the manufacturer's instruction at a 1:150 dilution. Slides were counterstained with DAPI (5 mg /mL, Sigma Aldrich, #D9542) and mounted with Mowiol.

RCAS/tv-a glioma mice were treated with either 3 doses of 7 mg/kg DOX, 7 mg/kg DOX-C′ dots, or saline vehicle at 3-day intervals. Twenty-four hours after the final injection, mice were anesthetized, and a vertical incision was made along the ventral abdominal wall at the midline. Major organs and tissues were displaced to one side of the abdominal cavity to expose the caudal portion of the abdominal vena cava and aorta. A 25G x 3/4” needle attached to a syringe was inserted in the vena cava for blood collection until flow into the syringe ceased. Animals were euthanized using a combination of CO₂ inhalation, pneumothorax, and exsanguination. Acquired blood samples were centrifuged for 15 minutes (1,000 × g) at 4°C and analyzed in triplicate using a mouse Troponin I Type 3 (cardiac) ELISA kit (Novus Biologicals, #NBP3–00456), according to the manufacturer's instructions. Briefly, a sandwich-ELISA was performed using a 96-well microplate reader (SynergyTM HT, Biotek Instruments, Inc.) coated with an anti-mouse TNNI3/cTn-I primary antibody. Standard solution or samples (100 μL) were added to microELISA plates and incubated for 2 hours. A biotinylated anti-mouse TNNI3/cTn-I detection antibody and Avidin–horseradish peroxidase (HRP) conjugate were then added and washed 3x with Wash buffer. The substrate reagent was subsequently added and incubated for 15 minutes, followed by addition of stop solution. Measured sample optical densities (OD) at 450 nm wavelengths were converted to tissue concentrations of mouse TNNI3/cTn-I using a calibration curve, the latter generated from a serial dilution of a standard solution (OD versus concentration).

Following euthanization of animals by CO₂, blood was collected by puncture of the abdominal vena cava immediately after euthanasia as previously described (37). Gross examination was then performed, and organs fixed in 10% neutral-buffered formalin, followed by decalcification of bone in a formic acid solution (Surgipath Decalcifier I, Leica Biosystems). Tissues were then processed in ethanol and xylene and embedded in paraffin in a Leica ASP6025 tissue processor. Paraffin blocks were sectioned at 5 μm, stained with H&E, and examined by a board-certified veterinary pathologist (S. Monette). The following tissues were processed and examined: heart, lungs, submandibular and mesenteric lymph nodes, liver, kidneys, spleen, bone marrow (sternum, femur, tibia, and vertebrae), and brain (4 coronal sections).

Analysis of blood samples was performed after collection in tubes containing dipotassium EDTA anticoagulant (Catalog # 365974, BD Biosciences). Samples were subsequently processed on an IDEXX Procyte DX hematology analyzer for the following parameters: white blood cell count, red blood cell count, hemoglobin concentration, hematocrit, mean corpuscular volume, mean corpuscular hemoglobin, mean corpuscular hemoglobin concentration, red blood cell distribution width standard deviation and coefficient of variance, reticulocyte relative and absolute counts, platelet count, platelet distribution width, mean platelet volume, and relative and absolute differential counts of neutrophils, lymphocytes, monocytes, eosinophils, and basophils.

Analysis of blood samples was performed after collection in tubes containing a serum separator gel (Catalog # 365967, BD Biosciences). Following centrifugation, serum was collected and analyzed on a Beckman Coulter AU680 clinical chemistry analyzer for the following parameters: alkaline phosphatase, alanine aminotransferase, aspartate aminotransferase, creatine kinase, gamma-glutamyl transpeptidase, albumin, total protein, globulin, total bilirubin, blood urea nitrogen, creatinine, cholesterol, triglycerides, glucose, calcium, phosphorus, chloride, potassium (K), sodium (Na), Na/K ratio, and albumin/globulin ratio.

Group means and SEMs were calculated for time-dependent and/or concentration-dependent changes in cellular uptake, proliferation index, functional markers (Ki-67, γH2AX, troponin), TME markers (Iba1, CD206), and tumor volumes. Statistical comparisons between the experimental groups were performed using one-way ANOVA followed by post hoc Tukey's test, and considered statistically significant if P < 0.05. No less than three replicates were generated per group (*, P < 0.05; **, P < 0.01; ***, P < 0.001), unless otherwise noted. Kaplan–Meier analysis was used to visualize survival data; the log-rank (Mantel–Cox) test was used to compare survival curves. All graphs were constructed and analyzed using GraphPad Prism 7 (GraphPad Software).

The data generated in this study are available within the article and its Supplementary Data Files.

As illustrated in Fig. 1A and Supplementary Figs. S1 and S3, an ultrasmall C′ dot platform functionalized with DOX–hydrazone linkers and optionally surface modified with multiple (∼3) DFO chelators (termed either DOX–PEG–Cy5-C′ dots or DFO–DOX–PEG–Cy5-C′ dots, or simply DOX-C′ dots and DFO–DOX-C′ dots, respectively) was synthesized using existing protocols plus a highly specific strain-promoted azide-alkyne cycloaddition reaction (23, 24, 27). This involved reacting particle surface-bound DBCO groups (e.g., on DFO–DBCO-C′ dots) with azide functional groups on azido-PEG4-hydrazone–DOX-linker conjugates (Supplementary Fig. S2). On the basis of the distinct elution profiles of DFO–DBCO-C′ dots and azido-PEG4-hydrazone linker–DOX conjugates (Supplementary Fig. S4A), a simple PD-10 column purification method was selected for separating free DOX or unreacted azido-PEG4-hydrazone linker–DOX from DFO–DOX-C′ dots. As synthesized DFO–DOX-C′ dots showed >99% purity using HPLC (Supplementary Fig. S4B). Figure 1B shows the representative UV-Vis spectrum of DFO–DOX-C′ dots with a DPR approximately 30 in relation to DFO–DBCO-C′ dots without conjugated drugs (DPR = 0), with corresponding absorption peaks at 485 nm (DOX) and 651 nm (Cy5-C′ dots), respectively. A similar DPR was found for non–DFO-functionalized DBCO-C′ dots. The successful conjugation of DOX molecules was further confirmed by the clear solution color change from light blue to dark brown post-conjugation (Fig. 1C). We found that variations in average DPR from 6 to 30 can be precisely controlled by altering reaction ratios between the azido-PEG4-hydrazone–DOX andDFO–DBCO-C′ dots (Supplementary Fig. S5A); DPR-dependent changes in solution color were also observed (Supplementary Fig. S5B). Figure 1D shows a representative TEM image of DFO–DOX-C′ dots with uniform particle size and size distribution. FCS measurements of the same DFO–DOX-C′ dots and their quantitative analysis suggested a hydrodynamic diameter of 6.3 nm and no significant increase in particle size over unconjugated DFO–DBCO-C′ dots, likely due to backfolding of the hydrophobic drugs in-between PEG chains of the soft particle shell to reduce interactions with water (Fig. 1E). For complete characterization datasets for base DFO–DBCO-C′ dots and DBCO-C′ dots, please see Supplementary Fig. S1A and S1B.

Each DOX molecule was conjugated to C′ dots via an acid-labile (or pH-sensitive) hydrazone linker (Supplementary Fig. S2) designed to remain intact in the bloodstream (pH 7.3–7.5) and undergo hydrolysis and payload release in mildly acidic endosomal (pH 5.0–6.5) and lysosomal (pH 4.5–5.0) compartments upon cellular internalization. To demonstrate the pH-responsive drug release capabilities of DFO–DOX-C′ dots, conjugates were suspended in buffers having a pH of 5.2 and 7.4 and kept on a shaking platform at 37°C. The percentage of DOX (% DOX) released from DOX-C′ dots as a function of pH was measured over a 48-hour period using HPLC. We observed significant release (up to 80%) at pH 5.2, whereas no appreciable release was observed at pH 7.4, indicating a pH-sensitive process (Fig. 1F). We also evaluated the payload stability of DFO–DOX-C′ dots incubated in human and mouse sera at 37°C under shaking conditions. As shown in Fig. 1G, 95.0% ± 2.9% and 91.3% ± 2.7% (mean ± SD) of DOX molecules remained conjugated over a 48-hour period in human and mouse sera, respectively. The release of free DOX was about 7.7% in human serum following 120 hours of incubation, whereas 11.1% was found for mouse serum. Because most of the intravenously injected C′ dot dose will be cleared within 48 hours p.i., serum stability of DFO–DOX-C′ dots was considered acceptable for our studies.

In vitro studies were conducted with non–DFO-functionalized C′ dots or DOX-C′ dots, as these did not require radioisotopic labeling. The cytotoxicity of DOX-C′ dots (DPR ∼30) versus free DOX was initially evaluated in glioma cells. The DPR value of approximately 30 was selected on the basis of the results of our screening PK and biodistribution studies in non–tumor-bearing female nu/nu mice (vide infra). First, the potency (LD₅₀) of DOX-C′ dots was assessed relative to DOX in multiple tumor cell types: glioma cells harvested from the RCAS/tv-a model (Fig. 1H), as well as human primary lung cancer (ECLC26), breast cancer (MDA-MB-231), and prostate cancer (PC-3) cells (Supplementary Fig. S6A–S6C). For these studies, cells were exposed to a range of DOX-C′ dot and DOX concentrations (10⁻¹² – 10⁻⁵ mol/L) for up to 7 days. Viability was assessed as a percentage of the respective untreated controls and used to calculate a LD₅₀ of 1.67 nmol/L in glioma cells (Fig. 1H), versus 29.6 nmol/L for free DOX. This finding suggested that increases in drug lethality (i.e., factor of ∼18) could be achieved over that of the free drug following particle attachment. It should be noted that LD₅₀ values computed for DOX-C′ dots were based on the particle concentrations, as C′ dot–drug conjugates, maintained at sizes below 8 nm, are considered a single drug entity. In addition, although DOX is known to be potent and effective against cell lines derived from malignant gliomas, its penetration across the BBB is inadequate, severely limiting its utility in patients with GBM (38). The antiproliferative effects of DOX-C′ dots (DPR ∼30) in EGFRvIII glioma cells were also investigated and compared with those of the native drug using 250 nmol/L concentrations. For cells treated with DOX-C′ dots, statistically significant reductions in proliferative capacity were seen as early as 24 hours post-incubation, as against those observed with free DOX, reaching a maximum at 96 hours (P < 0.0001; Supplementary Fig. S6D).

DOX-induced antiproliferative activity and cytotoxicity are thought to be the result of DNA intercalation and inhibition of topoisomerase II–mediated DNA repair (39). Thus, loss of cell viability and proliferative capacity relies on adequate internalization and subsequent intranuclear localization of DOX. Internalization was confirmed via flow cytometry after incubating DOX-C′ dots with RCAS/tv-a glioma cells over a range of concentrations (0–250 nmol/L) for 4 hours. Results showed a proportional increase of particle signal intensity (Cy5 fluorescence) within cells as a function of increasing concentration; a 250 nmol/L dose of DOX-C′ dots led to the highest median fluorescence intensity among doses tested (MFI; 5,000.8 a.u.; Supplementary Fig. S6E) and total percentage of DOX-C′ dot positive cells (99%). Findings were further confirmed by confocal microscopy in glioma cells (Fig. 1I–K), as well as in breast and lung cancer cells (Supplementary Fig. S7), noting that Cy5 (red puncta; Supplementary Fig. S7C and S7F) and intranuclear DOX fluorescence signals per cell were visually higher with DOX-C′ dots than with DOX alone for all cell types.

Biodistribution and serial PET imaging studies were conducted in healthy and glioma-bearing (mGBM) mice following intravenous administration of ⁸⁹Zr-labeled DFO–DOX-C′ dots, shown schematically (Fig. 2A). To assess biodistribution, clearance, and brain tumor–targeting capabilities of DOX-C′ dots, DFO-modified particles were labeled with ⁸⁹Zr (t_1/2 = 78.4 hours) for quantitative PET imaging (Supplementary Fig. S3; steps 9–13 illustrate the procedural workflow). Using ⁸⁹Zr-labeled DFO–DOX-C′ dots, initial PET screening studies were conducted in normal, healthy mice for a range of DPR values (i.e., 4 to ∼50) to identify an optimal DPR value for subsequent studies (Supplementary Fig. S8). For DOX-C′ dots, a DPR approximately 30 or less showed no significant changes in tracer uptake (%ID/g) within major organs and tissues. However, for a DPR of approximately 50, liver accumulations were seen to rise to levels greater than 5%ID/g (Supplementary Figs. S8 and S9). On the basis of these findings, a DPR of approximately 30 was selected for additional clearance, biodistribution, safety, and efficacy studies.

Renal and fecal clearance values of the particles were quantified over a period of 7 days (Supplementary Fig. S10A; Supplementary Table S2). For this study, each normal, healthy mouse was injected systemically with approximately 50 μCi of ⁸⁹Zr-DFO–DOX-C′ dots (DPR ∼30) and housed in individual metabolic cages. Urine and fecal specimens were collected separately at 6, 24, 48, 72, 120, and 168 hours p.i., and their activities assayed using a CRC-55tR dose calibrator. As observed in previous C′ dot studies (28, 32), renal clearance of ⁸⁹Zr-DFO–DOX-C′ dots was found to dominate whole-body excretion profiles at all time points p.i., ranging from approximately 93% (of total excreted) at 6 hours p.i. to approximately 70% (of total excreted) at 168 hours p.i. Urinary and fecal clearance at 6 hours p.i. were found to be approximately 14%ID and approximately 1%ID, respectively, yielding a urinary-to-fecal clearance ratio close to 14, suggesting dominant renal clearance at early p.i. time points. Cumulative clearance in both urine and feces increased to 41.1 ± 2.7%ID and 20.5 ± 7.6%ID by 72 hours p.i., respectively. Total hepatic clearance of ⁸⁹Zr-DFO–DOX-C′ dots was 22.2 ± 8.2%ID on day 7. The final renal-to-hepatic-clearance ratio was approximately 2 (Supplementary Table S2). Organ biodistribution profiles showed liver uptake values as low as 3.7%ID following intravenous injection of ⁸⁹Zr-DFO–DOX-C′ dots (Supplementary Table S3); these values are significantly lower than those reported for other nanocarrier-based drug delivery systems, for example, ⁶⁴Cu-labeled liposomes, with reported liver uptake values nearly 4-fold higher reported (40).

We also conducted serial PET imaging in non–tumor-bearing female mice after intravenous injection of ⁸⁹Zr-DFO–DOX-C′ dots (DPR ∼30), each animal serving as its own control. We observed notably high activity in blood and urine at 1 hour p.i., that is, activity greater than whole-body background activity, in regions corresponding to the heart and bladder, respectively (Fig. 2B, inset). Clearance of ⁸⁹Zr-DFO–DOX-C′ dots from the blood pool and the body was observed over the course of the study, confirmed by ex vivo organ-specific uptake measurements at study termination (Fig. 2B). On the basis of the measured ⁸⁹Zr-DFO–DOX-C′ dot time-activity data, favorable radiation dosimetry (Supplementary Table S4) was estimated for the 70-kg Reference Adult using the OLINDA dosimetry program (36), with absorbed doses and effective dose coefficients (rad/mCi and rem/mCi) found to be quite comparable with those of other ⁸⁹Zr-labeled radiopharmaceuticals. For ⁸⁹Zr-trastuzumab, for example, the effective dose coefficient is 18 mSv (or 1.8 rem) per mCi (41) versus approximately 12 (or 1.2 rem) per mCi, a lower value, for our ⁸⁹Zr-DFO–DOX-C′ dots.

Next, to investigate PK and brain tumor targeting and accumulation of ⁸⁹Zr-DFO–DOX-C′ dots in RCAS/tv-a glioma mice (n = 3), PET images were acquired, and findings correlated with histopathology (Fig. 2C and D). Representative maximum intensity projection images, illustrated for a single mouse, revealed significant bladder and cardiac (∼18%ID/g) activity at 2 hours p.i. (Fig. 2C). Tumor uptake increased with time, to about 3.0 ± 0.4%ID/g at 24 hours p.i., and nearly 4.0 ± 0.15%ID/g at 72 hours (Fig. 2C and E). At study termination (∼72 hours p.i.), whole-brain tumor tissue specimens were harvested for assessing particle tracer uptake and distribution. Representative H&E staining of frozen tissue sections confirmed the presence of partially necrotic areas of tumor (Fig. 2D1 and D4), which corresponded to areas of particle localization seen in Cy5 fluorescence microscopy (Fig. 2D2 and 2D5) and autoradiography (Fig. 2D3), as well as to areas of signal abnormality on MR imaging (Fig. 2D6).

Additional tracer kinetic profiles acquired in these mice after ⁸⁹Zr-DFO–DOX-C′ dot injection were similar to those reported previously for other ⁸⁹Zr-labeled C′ dots (34), noting time-dependent increases in tumor %ID/g (Fig. 2E), as well as tumor-to-blood and tumor-to-muscle ratios (insets, Fig. 2E). Liver accumulations showed an overall mild decrease with time p.i., with tumor-to-liver ratios approaching 1 by 72 hours p.i. (Fig. 2E). Furthermore, ex vivo biodistribution findings at 48 and 72 hours p.i. were less than 5%ID/g for all major organs and tissues, and confirmed tumor uptake values (Fig. 2F; Supplementary Table S5). The favorable PK profiles determined for the particle tracer in RCAS/tv-a glioma mice—including their low reticuloendothelial system accumulation, bulk renal clearance, and passive tumor targeting—paved the way for in vivo therapeutic efficacy evaluations.

We next investigated therapeutic properties of DOX-conjugated C′ dots in both RCAS/tv-a glioma and PDX models harboring EGFR mutations and/or amplifications (n > 5 mice/cohort). Treatment efficacy of DOX-C′ dots was assessed in relation to free DOX (Fig. 3A) using bioequivalent dosing regimens (i.e., 7 mg/kg/dose; n = 3 doses), as well as to animals receiving vehicle (controls). All treatments were administered systemically via tail vein every 3 days (200 μL/dose). For both RCAS/tv-a and EGFR PDX cohorts, mice treated with DOX-C′ dots exhibited statistically significant reductions in tumor volume of 86% (P = 0.034) and 63% (P = 0.029), respectively, expressed as a percentage of the control group. By contrast, no significant percentage of volume reduction changes were seen in DOX-treated mice (Fig. 3B and D; Supplementary Fig. S11A and S11C). Correlative fluorescence imaging of a representative brain tumor specimen harvested from a particle-treated PDX glioma mouse 96 hours after the final dose confirmed particle fluorescence signal (Fig. 3C, right) throughout the area of infiltrative tumor demarcated on H&E staining (Fig. 3C, left), suggesting particle penetration and intratumoral distribution. Finally, mice treated with free DOX demonstrated significant (i.e., 15%–20%) loss of their initial body weight over the treatment interval of approximately 9–12 days, whereas mice treated with DOX-C′ dots exhibited 10% or less loss of their initial body weight (Supplementary Fig. S11B and S11D).

In addition to conducting growth inhibition studies, we also evaluated whether treatment with DOX-C′ dots prolonged survival over that found with DOX and saline vehicle. RCAS/tv-a mice were randomized to one of these three treatment groups (n = 10/cohort), and a dosing strategy identical to that used for the previous treatment study was used. Mice receiving vehicle demonstrated a median survival time of 13 days. By comparison, mice treated with DOX and DOX-C′ dots demonstrated median survival times of 12.5 and 23 days, respectively (Fig. 3E), the latter statistically significant relative to both DOX (P = 0.0004) and controls (P = 0.0005). Median survival times were not statistically significant when comparing DOX-treated mice to those administered saline vehicle. Together, these data demonstrated that the conjugation of DOX to C′ dots was efficacious and led to a survival benefit, in contrast with DOX itself, without adversely affecting mouse weights. Factors that may have contributed to these findings include improved PK, tissue penetration, and intratumoral particle distribution.

We selected proliferative (Ki-67) and DNA damage (γH2AX) markers as treatment response indicators based on the mechanisms of action exhibited by DOX; γH2AX also served as a marker of cardiac toxicity (vide infra). Before evaluation of these functional markers, significant tumor volume reductions were observed for DOX-C′ dot–treated RCAS/tv-a glioma mice over a 10-day interval on T2-weighted MRI (Fig. 3F), confirmed on H&E staining (Fig. 3G); these changes were also accompanied by marked intratumoral reductions in Ki-67 staining using IHC (Fig. 3H). By contrast, DOX- and vehicle-treated glioma mice showed no significant reductions in either tumor volume or Ki-67 staining. In addition, statistically significant drops in both the number of Ki-67 cells per unit area (P = 0.02) and the total number of Ki-67 cells (P = 0.02) were found for tumors treated with DOX-C′ dots, but not for those treated with DOX or vehicle (n = 3 mice/cohort; Fig. 3I and J).

Similarly, positive findings were obtained for γH2AX-stained tumor tissue specimens (Fig. 4A and B) derived from 3 additional cohorts of RCAS/tv-a glioma mice. Representative mice from each of these 3 cohorts were also assessed with T2-weighted MRI and H&E staining (Fig. 4A). A 53% increase in γH2AX staining was found in specimens treated with DOX-C′ dots, but not in DOX- (P = 0.04) or vehicle-treated specimens (P = 0.003); these findings are plotted in Fig. 4C and confirmed on MRI and adjacent H&E-stained tissues (Fig. 4A). Although considered an effective standard-of-care chemotherapeutic agent for treating a number of malignancies, including small cell lung and breast cancers, systemic administration of DOX is also associated with dose-limiting cardiotoxicity and myelosuppression (42). As DOX-C′ dots have a favorable PK profile, we aimed to reduce drug accumulations to non-target tissues as a means of mitigating these adverse effects. In cardiac tissues harvested from RCAS/tv-a glioma mice treated with DOX-C′ dots, we observed statistically significant decreases in the percentage of γH2AX-stained cardiac cells relative to that seen with DOX alone (i.e., P = 0.0001), in fact, restoring intensities to levels observed for control specimens (Fig. 4D). These findings suggested that DOX-C′ dots could abrogate early DNA damage responses typically seen in normal organs with DOX, thus effectively reducing its therapy-limiting toxicity and thereby increasing the doses that patients can receive during treatment, that is, increasing its therapeutic index (43).

We additionally monitored another highly sensitive and specific marker of early myocardial cell injury: elevation of serum cardiac troponin levels (i.e., cardiac troponin I; refs. 44, 45). To determine an optimal time point to assay serum troponin I enzyme levels, 4 cohorts of non–tumor-bearing mice were each intravenously injected with 3 doses of DOX (7 mg/kg), with each cohort harvested at a different time point (i.e., 1 day to 2 weeks p.i.). Serum enzyme levels were found to be maximal at 24 hours p.i., and this value was subsequently used for comparative studies with DOX-C′ dots and vehicle (Fig. 4E). Significantly reduced enzyme levels (P = 0.0376) were found in serum samples from mice treated with DOX-C′ dots versus those treated with DOX or saline vehicle (Fig. 4F). On histologic examination, enhanced apoptotic cell death (i.e., increased cleaved caspase-3) was also observed in a number of evaluated DOX-treated cardiac specimens (Supplementary Table S6), suggesting early cardiac injury. Corresponding to these toxicity data, mouse body and cardiac weights were significantly reduced for the majority of DOX-treated mice (n = 9; red, Supplementary Table S6) on limited gross examination, as compared with DOX-C′ dot– (n = 5) and vehicle-treated (n = 4) cohorts.

We further sought to examine changes in select tumor microenvironment (TME) response markers, namely ionized calcium–binding adaptor molecule 1 (Iba1) and the anti-inflammatory macrophage marker, CD206, by IHC (Fig. 4G). A prominent feature of high-grade gliomas is the presence of a large population of tumor-associated macrophages (TAM) within the TME; their interaction with glioma cells promotes disease progression, survival, and resistance (46). Prior work with particle-based platforms has shown that TAMs can be polarized toward a pro-inflammatory, anti-tumor phenotype (15). Following treatment of PDGF-B glioma mice with 3 doses of DOX-C′ dots, we observed significant reductions in the percentage of single-positive (CD206⁺) and double-positive (Iba1⁺CD206⁺) anti-inflammatory macrophages relative to DOX- (P = 0.0025) and vehicle-treated (P = 0.0008) tumors (Fig. 4H and I) 96 hours after the final treatment. No significant change in the %Iba1⁺ cells was noted (Fig. 4J) across treatment groups.

Finally, we performed additional limited toxicology studies using mice from each treatment arm of our growth inhibition studies. Analyses were conducted by an independent pathologist showing no abnormalities or obvious tissue damage in the treatment groups based on H&E staining (Supplementary Fig. S12) and gross histopathological examination. Complete blood counts (Supplementary Tables S7 and S8) and serum chemistries (Supplementary Tables S9 and S10) were also monitored in these mice in addition to histopathology (Supplementary Table S11). Although complete blood count results showed initial decreases in the number of reticulocytes over a 12-day post-treatment interval following administration of DOX-C′ dots, rapid recovery to normal levels was observed thereafter. By contrast, DOX-treated mice did not show a similar rate of recovery over the study interval. Hematologic results otherwise looked similar across all evaluated groups and were largely within normal limits. Evaluation of serum chemistry values revealed no significant abnormalities or differences between mice treated with DOX or DOX-C′ dots when compared with vehicle-treated mice. On the basis of the results of all histological, metabolic, and hematological parameters assayed, DOX-C′ dots were found to be safe and well-tolerated when administered to mice at dosages showing marked toxicity with DOX alone.

The development of newer-generation particle-based drug delivery platforms that can address potential hurdles limiting high-efficacy treatment of CNS malignancies remains a critical unmet need. A key determinant aiding successful treatment responses rests on the design of platforms exhibiting favorable PK and clearance that, in turn, can mitigate adverse off-target events. Tunability and control of particle surface chemistry are also crucial for achieving reproducible batch-to-batch performance. By leveraging the beneficial properties of these platforms, substantial reductions in tumor burden and DLTs might be achieved for a variety of toxic payloads that could be preferentially delivered to target sites (47).

In this work, we advanced a first-in-kind renally clearable nanoparticle–drug conjugate for effectively treating high-grade glioma, improving upon key biological properties that have limited the utility of the native drug. Rather than use of a small-molecule inhibitor as in our earlier work (28, 32), we used DOX as a chemotherapeutic payload conjugated via a molecularly engineered pH-sensitive azide-functionalized linker for enhanced drug release under conditions of acidic pH, typically found in endo/lysosomal compartments (48). We conducted first-time efficacy evaluations in clinically relevant GEMM and PDX high-grade glioma models with DOX-C′ dots, which overcame the lack of survival benefit found with the free drug using bioequivalent doses of DOX. This result was achievable given the favorable PK and clearance profiles of the C′ dot platform. The histopathology of harvested tumor specimens validated these findings, with significantly decreased cellular proliferation and enhanced DNA damage observed within a 24-hour p.i. period. A lack of acute cardiotoxicity was also found relative to the free drug, including analysis of a very short p.i. window using well-characterized molecular/biochemical assays and cardiac weights.

Serial PET imaging at early time points showed that ⁸⁹Zr-DFO–DOX-C′ dots were confined to the cardiac blood pool, followed by bulk renal excretion (Supplementary Fig. S8). It is important to note that quantitative PET imaging and metabolic cage results were essential to inform the translational developments of this platform: product safety, particle design (e.g., DPR), PK, tumor tissue penetration/localization, and off-target uptake. For instance, the influence of DPR on the resulting PK/biodistribution profiles for ⁸⁹Zr-DFO–DOX-C′ dots (Supplementary Fig. S8; Supplementary Table S5) was evaluated to maintain low liver %ID/g values of around 5 or less; this was achieved for a DPR value approximately 30. Even at the highest DPR value (i.e., 50), the %ID/g value for the liver only rose to approximately 8%, likely due to the slight increase in the size of the DOX-C′ dot platform.

The approach we present is aimed at treating the bulk tumor and does not necessarily address widespread infiltration of glioma cells through brain parenchyma protected by a fully intact BBB. Although a limitation, this may also be considered a desirable feature for classes of drugs, such as anthracyclines, that are neurotoxic. Importantly, in the brain tumor mice studied, the extent to which particle transport could be investigated was limited by their relatively short lifespans. In the RCAS/tv-a high-grade glioma model, for instance, we were only able to evaluate particle transport 96 hours after administering the final particle treatment using fluorescence microscopy. These studies did confirm, however, that DOX-C′ dots can effectively penetrate a variably permeable BBB in high-grade glioma, including those regions with a more intact barrier, which resulted in relatively uniform intratumoral distributions and contributed to the efficacy achieved in two different high-grade glioma models. These findings point to the potential utility of this particle therapy for treating human malignant brain tumors. In our prior work (28), particles were found to readily diffuse through gliomas beyond the initial distribution limited by the BBB given their size and surface chemical properties. These observations are usually not seen with larger-size agents (e.g., >60 nm), which can extravasate from blood vessels, but often cannot effectively penetrate or diffuse within the tumor interstitium (11). The extent to which DOX-C′ dots can more widely penetrate the BBB to treat cells infiltrating throughout the brain parenchyma, remains an area of active investigation, and would need to be assessed further in the clinical trial setting to draw definitive conclusions concerning utility.

In practice, targeting bulk glioma tumor with radiotherapy and surgery is clinically valuable for palliation and can improve prognosis, even if it fails to address infiltrating disease and is insufficient for long-term control. To that end, not all therapies studied in GBM demonstrate effectiveness beyond the site of the primary tumor, but have, nevertheless, led to a survival benefit. One such example is BCNU (1,3-bis-(2-chloroethyl)-1-nitrosourea) wafers (Gliadel wafers) implanted in the intratumoral cavity to treat residual disease. This agent has demonstrated prolonged local disease control and progression free survival (49, 50). Although current standard treatment includes the DNA alkylating agent temozolomide, an effective BBB penetrant (51), relapse or progression are typically seen within a year (52), and there are no highly effective second-line chemotherapy options to date. In this context, there has been extensive investigation of therapies, including DOX, for recurrent tumors resistant to temozolomide (53, 54).

At the preclinical level, significant gains have been made in the development of a diverse array of DOX-containing organic and inorganic-based nanoprobes for targeting CNS malignancies, which include newer-generation lipid-based products (17), polymersome assemblies (55), biological delivery vehicles (56), and mesoporous silica particles (57); these and many other classes of nanodelivery vehicles have been discussed in a number of excellent reviews (10, 58, 59). A property common to the majority of these agents, however, is their relatively large size (i.e., >60 nm), which can lead to significant accumulations in the liver and spleen in PK/biodistribution studies (40, 60, 61). Additional challenges include long circulation times, DLT, and premature or limited drug release, the latter often attributed to protein corona formation (62). Despite the scope of these extensive research efforts, the overall result has been a general lack of clinically translatable agents.

To address these technical hurdles, several investigator groups, including ours, have advanced ultrasmall inorganic or inorganic-organic hybrid drug delivery vehicles [i.e., surface-modified silica (28), gold (63), and iron oxide (64)]. Advantages of this ultrasmall class of agents include their bulk renal clearance, surface chemical control, low off-target accumulations (i.e., <5%ID/g), and improved target tissue penetration, diffusion, and uptake within solid tumors. Given the very different brain tumor models, techniques, and probes used in such investigations, a meaningful comparison of brain tumor uptake values is challenging. Largely driven by the EPR effects, overall brain tumor uptake among such platforms, to first order, is not substantially different. However, the actual time-dependent kinetics of larger platforms will substantially differ from our sub–8-nm particles. Moreover, the U87MG model used in many studies is not a clinically relevant model, and is known to have a leakier BBB than, say, GEMMs, which will limit interpretation and comparison of such results.

Furthermore, studies using this class of ultrasmall materials have principally focused on non–CNS-based applications without progressing to clinical trial testing (13). We previously developed and translated a sub–8-nm integrin-targeting C′ dot to the clinic (29–31) for image-guided surgical applications, including an active Phase 1 study for evaluating patients with malignant brain tumor. The targeted delivery, accumulation, diffusion, and retention of this intravenously injected tracer are currently being assessed in primary and metastatic CNS lesions using multimodal imaging (MRI/PET/optical). Importantly, the multimodal (PET/optical) imaging properties of the platform might be exploited for their use as a companion diagnostic tool to monitor treatment responses and select patients most likely to benefit from a specific particle-based therapy in future clinical trials (Supplementary Fig. S13).

Although conventional DOX remains the standard of care for curative treatment of multiple cancer types, such as breast cancer, a major barrier to its widespread use as a chemotherapeutic is its known cardiotoxic effects, which include reduction of left ventricular ejection fraction and elevated cardiac troponin levels indicative of myocardial cell death. These findings have accelerated the development of alternative nanoformulations that could reduce such effects and greatly improve cardiovascular outcomes for a large group of cancer survivors. Although PEGylated DOX HCl, a modified version of Doxil (65), and others, do, in fact, demonstrate less cardiotoxicity over the free drug in clinical practice, there continues to be an important unmet need to further reduce such adverse events. Moreover, such findings are usually reported at delayed intervals p.i. (∼1 week). To the best of our knowledge, assays that yield earlier assessments of cardiotoxicity, typically arising within a 24-hour treatment window, such as DNA damage markers (i.e., γH2AX), troponin levels (44, 66), or apoptotic cell death (cleaved caspases; ref. 67), as studied here, have not been integrated into preclinical studies investigating DOX treatment efficacy.

Finally, the development of ultrasmall drug delivery vehicles that can more readily penetrate and diffuse within solid tumors, even without a cancer-targeting moiety, is critically important, as demonstrated by our studies. This is particularly true for treating subsets of tumors that lack targetable surface markers, are considered undruggable, or that exhibit mutant or truncated cell surface receptors that can be difficult to target. Such probes are also important for treating tumors known to exhibit heterogeneous marker expression, so that even nearby tumor cells lacking target expression may still undergo cell death due to particle internalization. In turn, this may limit the emergence of drug resistance. Finally, as opposed to larger-sized platforms, the greater diffusivity of these essentially neutrally charged particles (24) within the TME means that therapeutic particles may also damage and kill tumor cells in areas with an intact barrier, areas that would ordinarily serve as impediments to attaining efficacious treatment responses.

F. Chen reports other support from Elucida Oncology during the conduct of the study. R. Lee reports grants from NIH during the conduct of the study. M.Z. Turker reports grants from National Cancer Institute, as well as grants and personal fees from Scholarship from Republic of Turkey during the conduct of the study; M.Z. Turker also reports personal fees from Elucida Oncology outside the submitted work. K. Ma reports other support from Elucida Oncology Inc. outside the submitted work; K. Ma also reports a patent for Nanoparticle Drug Conjugates issued and licensed to Elucida Oncology. P. Zanzonico reports personal fees from Novartis outside the submitted work. C.M. Rudin reports personal fees from AbbVie, Amgen, AstraZeneca, Daiichi Sankyo, Epizyme, Genentech/Roche, Ipsen, Jazz, Kowa, Merck, Syros, Bridge Medicines, Earli, and Harpoon Therapeutics outside the submitted work. C. Brennan reports other support from Elucida Oncology during the conduct of the study; C. Brennan also reports a patent for Nanoparticle imaging (WO2018009379A1) issued and licensed to Elucida Oncology and for Nanoparticle tumor treatment (EP3448436A1) issued and licensed to Elucida Oncology. U. Wiesner reports grants from NIH/NCI during the conduct of the study, as well as other support from Elucida Oncology, Inc. outside the submitted work; in addition, U. Wiesner has a range of patents and disclosures on Cornell dots pending, issued, licensed, and with royalties paid from Elucida Oncology, Inc.. U. Wiesner is a co-founder of Terapore Technologies, Inc., and sits on the scientific advisory board, which produces ultrafiltration membranes used in biotech and biopharmaceutical industries. M.S. Bradbury reports grants from National Institutes of Health, as well as other support from PSC-CUNY Research Award Program and Elucida Oncology during the conduct of the study; M.S. Bradbury also reports other support from Elucida Oncology outside the submitted work. In addition, M.S. Bradbury reports a patent for Nanoparticle Drug Conjugates licensed to Elucida Oncology and for Compositions and Methods for Targeted Particle Penetration, Distribution, and Response in Malignant Brain Tumors licensed to Elucida Oncology. No disclosures were reported by the other authors.

V. Aragon-Sanabria: Conceptualization, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. A. Aditya: Investigation, visualization, methodology, writing–original draft, writing–review and editing. L. Zhang: Investigation, visualization, methodology, writing–review and editing. F. Chen: Conceptualization, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. B. Yoo: Conceptualization, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. T. Cao: Investigation, visualization, methodology, writing–review and editing. B. Madajewski: Conceptualization, investigation, visualization, methodology, writing–review and editing. R. Lee: Investigation, visualization, methodology, writing–review and editing. M.Z. Turker: Investigation, visualization, methodology, writing–review and editing. K. Ma: Methodology, writing–review and editing. S. Monette: Investigation, visualization, methodology, writing–review and editing. P. Chen: Methodology, writing–review and editing. J. Wu: Methodology, writing–review and editing. S. Ruan: Writing–review and editing. M. Overholtzer: Investigation, writing–review and editing. P. Zanzonico: Methodology, writing–review and editing. C.M. Rudin: Methodology, writing–review and editing. C. Brennan: Conceptualization, writing–review and editing. U. Wiesner: Conceptualization, supervision, methodology, writing–original draft, writing–review and editing. M.S. Bradbury: Conceptualization, methodology, supervision, writing–original draft, writing–review and editing.

We thank Dr. L. Parada, Director of the Brain Tumor Center, and his staff for generating the PDX models used in this work. We also thank Hunter Mass Spectrometry for the use of their instrumentation. This study was funded by grants from the NIH (1U54 CA199081–01, to M.S. Bradbury and U. Wiesner) and Sloan Kettering Institute (core grant P30 CA008748CCSG), Elucida Oncology, Inc., and PSC-CUNY Research Award Program.

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.