Glioblastoma (GBM) has a dismal prognosis. Evidence from preclinical tumor models and human trials indicates the role of GBM-initiating cells (GIC) in GBM drug resistance. Here, we propose a new treatment option with tumor enzyme-activatable, combined therapeutic and diagnostic (theranostic) nanoparticles, which caused specific toxicity against GBM tumor cells and GICs. The theranostic cross-linked iron oxide nanoparticles (CLIO) were conjugated to a highly potent vascular disrupting agent (ICT) and secured with a matrix-metalloproteinase (MMP-14) cleavable peptide. Treatment with CLIO-ICT disrupted tumor vasculature of MMP-14–expressing GBM, induced GIC apoptosis, and significantly impaired tumor growth. In addition, the iron core of CLIO-ICT enabled in vivo drug tracking with MR imaging. Treatment with CLIO-ICT plus temozolomide achieved tumor remission and significantly increased survival of human GBM-bearing mice by more than 2-fold compared with treatment with temozolomide alone. Thus, we present a novel therapeutic strategy with significant impact on survival and great potential for clinical translation. Mol Cancer Ther; 16(9); 1909–21. ©2017 AACR.
Glioblastoma multiforme (GBM) is the most frequent primary malignant brain tumor in adults and the leading cause of cancer-related death in children (1). The mean survival time is 12 months in both adults and children (2). GBM contains GBM-initiating cells (GIC) that play a central role in GBM development and disease recurrence (3). GICs possess enhanced self-renewal and invasive properties, promote tumor angiogenesis, and are resistant to the limited number of current therapies, notably temozolomide (4, 5). Therefore, GICs represent the core problem of the dismal outcome of GBM. To achieve improved survival of GBM patients, novel therapeutic strategies are needed that target GICs.
GICs are preferentially found in the perivascular niche (6, 7) and depend on tumor vessels for nutrition supply and survival (8). The efficacy of current GBM therapies with oral or intravenous drugs is hindered by their limited transendothelial permeability to the GIC niche. Previous drug-loaded nanocarrier systems relied on the enhanced permeability and retention (EPR) effect in tumors (9, 10). However, the highly heterogeneous nature of the EPR effect can lead to poor delivery to the GIC niche and hence, poor therapeutic efficacy (10). A new, emerging strategy is to deliver vascular-disrupting agents (VDA), which do not rely on the EPR effect. VDAs target endothelial cells at the intraluminal surface of blood vessels, for example, by disrupting the colchicine-binding site of tubulin (11, 12). This leads to vascular collapse and starvation of tumor cells supplied by these vessels, a very effective therapeutic strategy (12, 13). In addition, VDAs selectively destabilize the tumor microvascular endothelial lining, causing a transient increase in vascular permeability and drug delivery to the perivascular tumor interstitium (14, 15), the location of the GIC niche. Thus, the highly vascularized nature of GBMs and perivascular location of the GIC-vascular niche have spurred a lot of interest in VDAs. Previous VDAs, including combretastatin and 5,6-dimethylxanthenone-4-acetic acid, have led to significant necrosis in gliomas (16, 17). However, the clinical efficacy of first-generation VDAs was limited by a high prevalence of cardiotoxicity (18–20).
To avoid concomitant toxic effects in normal organs, nontoxic VDA-prodrugs can be designed, which are activated by specific tumor enzymes (21). For example, matrix metalloproteinases (MMP) and specifically the membrane-type MMPs (MT-MMPs; MT1-MMP = MMP-14) subclass represent an ideal target for prodrug activation, because they are highly overexpressed in GBM (22–24) and can selectively cleave specific peptide sequences (21, 25–27). The azademethylcolchicine-peptide conjugate ICT2588 is metabolized by MMP-14 to release an active VDA, azademethylcolchicine, with efficacy against a range of solid tumors (26) and minimal systemic toxicity (28). We coupled ICT, a minor structural analogue of ICT2588 (modified to allow conjugation to nanoparticles), to cross-linked iron oxide nanoparticles (CLIO) to generate theranostic nanoparticles (CLIO-ICT). Initial feasibility studies showed MMP-14–specific cleavage, efficient drug delivery, and therapeutic efficacy in murine mammary adenocarcinomas (29). This new theranostic strategy should be particularly beneficial for GBM, where efficient drug treatment is limited by the blood—brain barrier and inability to reach the perivascular GIC niche. On the basis of the vascular disrupting properties of the ICT drug (26), CLIO-ICT should be able to permeate the tumor microvascular endothelium and reach tumor cells and GICs. Thus, we hypothesized that CLIO-ICT will induce significant apoptosis of GBM tumor cells and GIC, and thereby, prolong survival of GBM-bearing mice.
Materials and Methods
Chemicals and antibodies
The following antibodies were used: α-tubulin (Abcam), CD133 (Miltenyi Biotec), CD49F (Biolegend), CD15 (BD Biosciences), CD31 (Abcam), MMP-14 (Abcam), Nestin (Abcam), and cleaved caspase-3 (Cell Signaling Technology). The following chemicals were used: ferumoxytol (AMAG Pharmaceuticals), temozolomide (Sigma), and ilomastat (Selleckchem).
Synthesis and characterization of CLIO-ICTs
ICT and CLIO-ICT were synthesized according to a previously published protocol (29). To determine the relaxivities of ferumoxytol and CLIO-ICT, in vitro MRI studies were conducted on a 7T MR scanner (Bruker Biospin). The transverse relaxation times (T2) of ferumoxytol and CLIO-ICT in water with various of Fe concentrations (0, 2.5, 5, 10, 20, 40 μg/mL) were measured individually using a fast-spin echo sequence with a repetition time (TR) of approximately 3,000 ms, multiple echo times (TE) of 6.8, 13.6, 20.4, 27.3, 34, and 40.9 ms. The T2 relaxivity values (r2) was obtained from linear least-squares determination of the slope of 1/T2 relaxation rate (s−1) versus the Fe concentration plot. The concentration of iron (Fe) was determined using inductively coupled plasma-mass spectrometry (ICP-MS) and the iron (Fe) atoms in each iron oxide core was estimated to be approximately 6200 using Diamond crystal structure analysis software. The molar concentration of iron (Fe) in CLIO-ICT stock solution was then calculated. As ICT is linked with FITC, the average amount of ICT covalently immobilized to a single CLIO-ICT nanoparticle was determined by the emission intensity of FITC. In brief, the emission profile of FITC of a diluted CLIO-ICT solution was recorded upon excitation at 495 nm. The concentration of FITC was then estimated using a calibration plot obtained from a set of standard FITC solutions. The average number of ICT on each CLIO-ICT was then obtained. To determine the sizes of ferumoxytol and CLIO-ICT, the samples in deionized (DI) water at a Fe concentration of 100 μg/mL were analyzed using a Zetasizer Nano ZS equipment.
Two patient-derived GBM cell lines (pcGBM2, pcGBM39) and three commercially available GBM cell lines (U87-MG, U138, and A172 from ATCC) were used for in vitro studies. Both pcGBM2 and pcGBM39 were kindly provided by Dr. Sanjiv Sam Gambhir, Stanford University (Stanford, CA). HCN2, normal cortical neuron cells were purchased from ATCC. U87-MG, U138, A172 cells were grown in DMEM (Life Technologies) containing 10% FBS and 1% penicillin/streptomycin (Life Technologies). HCN2 cells were grown in DMEM with 4 mmol/L l-glutamine adjusted to contain 1.5 g/L sodium bicarbonate, 4.5 g/L glucose (90%), and FBS (10%). pcGBM2 and pcGBM39 were cultured as floating cellular spheres as described previously (30). All cell lines used were authentic and confirmed to be mycoplasma negative using the MycoAlert Mycoplasma Activity Kit (Lonza). To test for mycoplasma activity, MycoAlert substrate was added to the test media (cell supernatant) which catalyzes the conversion of ADP to ATP by mycoplasma enzymes in the cell supernatant. This was followed by addition of a luciferase reaction mix that cleaved ATP to produce a bioluminescent signal as an indication of mycoplasma activity. All cell lines used were from early passages.
Assessment of MMP-14 gene expression of different patient-derived GBM neurospheres (pcGBM2 and pcGBM39) and GBM cell lines (U87-MG, U138 and A172) as well as human neuronal cortical cells (HCN2) as controls was determined by qPCR as described previously (29).qPCR expression analysis for MMP-14 and the control marker GAPDH was done and the total cellular RNA was extracted from each sample with the QIAGEN RNeasy mini kit. cDNA was prepared from total RNA and quantitative real-time PCRs (qPCR) were carried out and analyzed on an Applied Biosystems StepOne Real-Time PCR System. The formation of double-stranded DNA product was monitored by TaqMan gene expression primers.
In vitro evaluation of the therapeutic efficacy of CLIO-ICT against GBM cancer cells and GICs
Patient-derived GBM neurospheres (pcGBM2 and pcGBM39) and GBM cell lines (U87-MG, U138 and A172) as well as human neuronal cortical cells (HCN2) cells were plated in 96-well plates at 2 × 104 cells/well and treated with CLIO-ICT (10 nmol/L), ICT (10 nmol/L), CLIO (0.01 mmol/L), and DMSO at 37°C, 5% CO2.
Cell viability after 96 hours was assessed using MTS assay kit (Promega) as per the manufacturer's instruction. At the end of the incubation period, the absorbance was measured at 490 nm in a microplate reader (Expert Plus V1.4 ASYS). Cell viability was also verified with cell counting following addition of Trypan blue (Sigma) at a final concentration of 0.1% (v/v), and routine examination of the cells under phase contrast microscopy.
For apoptotic assays control and treated pcGBM2, pcGBM39 U87-MG, U138 and A172 and HCN2 cells were analyzed for caspase-3 activity levels, a marker of cytotoxicity using the SensoLyte Homogeneous AMC Caspase-3/7 assay kit (AnaSpec, Inc.), according to the manufacturer's instructions. Release of the AMC fluorophore following cleavage of the specific fluorometric caspase substrate, DEVD-AMC was detected using a fluorometer (ex/em = 354 nm/442 nm). Data were normalized relative to the vehicle-treated controls. For flow cytometry–based apoptosis detection, control, and CLIO-ICT–treated pcGBM39, U87-MG and A172 cells were also analyzed by Annexin V-based methods using Annexin V Apoptosis Detection Kit (eBioscience, Inc.). In addition, control and CLIO-ICT pcGBM39 cells were scored for anti-active caspase-3 (BD Biosciences) with flow cytometry (BD FACS ARIA II, BD Biosciences).
For characterization of GICs, pcGBM39 and pcGBM2 were stained for cancer stem cell markers using anti-CD133-Biotin (Miltenyi Biotec), CD15-FITC (BD Biosciences), and CD49F-PE-Texas red (Biolegend) and analyzed with flow cytometry Immunofluorescence staining for α-tubulin (Cell Signaling Technology) at 1:25 dilution was conducted as described previously.
In vivo evaluation of the therapeutic efficacy of CLIO-ICT against GBM cancer cells and GICs
In vivo intracranial xenografts.
Cells (3 × 105) were injected stereotaxically into the striatum of anesthetized 6- to 8-week-old NOD scid gamma (NSG) mice, using the following coordinates: 2 mm posterior to the bregma, 2 mm lateral to the midline, and 3–4 mm deep with respect to the surface of the skull. Once tumor masses were detected with bioluminescence imaging, mice were randomized in four groups and CLIO-ICT, ICT, CLIO, or PBS were delivered intravenously twice a week for 14 days. Total cumulative doses for each drug regimen were: CLIO-ICT (80 mg/kg of ICT), ICT (80 mg/kg of ICT), and CLIO (0.5 mmol Fe/kg). For combination treatment with temozolomide (33 mg/kg), mice were treated with ICT and CLIO-ICT twice a week for 21 days. Total cumulative doses for temozolomide were 200 mg/kg, respectively. For determining in vivo tumorigenesis of GICs, 500 CD133+CD15+, CD133−CD15−, CD49F+CD15+, CD49F+CD15− pcGBM39 cells were injected stereotaxically into the striatum of anesthetized 6- to 8-week-old NOD scid gamma (NSG) mice as described above. Animals were scored for tumor formation with bioluminescence assays until 4 months. All animal maintenance, handling, surveillance, and experimentation were performed in accordance with and approval from the Stanford University Administrative Panel on Laboratory Animal Care (Protocol 24965).
Luminescent imaging was performed 30 days (for pcGBM39) and 60 days (for pcGBM2) after tumor injection on an IVIS Spectrum (Caliper Life Science) and quantified using Living Image 4.0 software. D-Luciferin (firefly) potassium salt solution (Biosynth) was prepared (15 mg/mL) and injected intraperitoneally (0.139 g luciferin per kg body weight). Total flux (photons per second) values were obtained by imaging mice until peak radiance was achieved and quantified with Living Image 4.0 software. Bioluminescence imaging was repeated at the end of the treatment and analyzed in a blind manner.
Mice were euthanized at the end of the treatment for further histologic examination. For histologic analysis, the brains were kept in 10% neutral buffered formalin for 24 hours, followed by 70% ethanol at room temperature for 24 hours. Brains were then embedded in paraffin for 3 hours at 67°C. Coronal sections (5-μm thick) were stained with hematoxylin and eosin, and images were acquired (Eclipse E800, Nikon). Brain tissues were also fixed in 4% paraformaldehyde at 4°C overnight and later immersed in 30% sucrose for 2 days. Brains were then embedded in optimal cutting temperature and stored in −80°C. Coronal sections (5-μm thick) were mounted on superfrost slides, rinsed with PBS, and permeabilized with 0.1% Triton X-100 made in PBS solution for 15 minutes. Subsequently, cells were blocked for 2 hours and stained with primary antibodies overnight to determine GIC populations or to evaluate apoptosis upon treatment with theranostic nanoparticles. The following dilutions were used: MMP-14 (1:200, Abcam), Nestin (1:200, Abcam), cleaved caspase-3 (1:300, Cell Signaling Technology), CD133 (1:10, Miltenyi Biotec), and CD15 (1:200, Abcam). Immunofluorescence images were acquired with a Leica SP8 confocal microscope using Leica AF software. Images were prepared using Adobe Photoshop (Adobe Systems) and analyzed using Velocity 64 software.
Coronal sections (5-μm thick) from PBS, CLIO, ICT, and CLIO-ICT–treated animals were mounted on superfrost slides, rinsed with PBS, and permeabilized with 0.1% Triton X-100 made in PBS solution for 15 minutes. Subsequently, cells were blocked for 2 hours and stained with CD31 (1:20, Abcam) antibody overnight to determine endothelial cells. Nuclei were counterstained with DAPI and imaged in Leica SP5 confocal microscope. Single endothelial cells or clusters of endothelial cells positive for CD-31 were considered as a vessel. At first, slides were examined at an original magnification of 10×. Three “hotspots” (areas with the highest microvessel concentration) from each slide were identified and these areas were photographed at 10×. The area of this field was recorded and the number of microvessels in this field was counted with ImageJ in a blindfolded manner. MVD (microvessel/mm2) were then assessed according to Weidner and colleagues (31). MVD of the specimen were estimated as a mean ± SD of MVD in three different fields from three independent experiments.
For detection of apoptosis in GICs and GBM cancer cells, mice were euthanized, and brain tumors were dissociated to single cells and stained with anti‐CD15-APC (BD Biosciences) or Annexin V-PE (BioLegend) and DAPI. Tumor cells were gated on the basis of GFP expression and mouse cells were gated out using a lineage mixture of Brilliant Violet 605–conjugated H2kb, H2kd Ter119, and CD45 antibodies. Flow cytometric analysis and cell sorting were performed on a BD FACS ARIA II (BD Biosciences) flow cytometer.
We evaluated tumor delivery of CLIO-ICT in GBM-bearing mice using MRI. CLIO-ICT comprised of ferumoxytol (CLIO-M) linked to ICT. Ferumoxytol (AMAG Pharmaceuticals) comprised of iron oxide nanoparticles that had a mean hydrodynamic diameter of 30 nm and provided superparamagnetic signal effects on T1- and T2-weighted MR images. pcGBM39-bearing mice underwent MRI before and 24 hours after intravenous injection of PBS, CLIO, ICT, and CLIO-ICT (n = 6) and at regular intervals during treatments. MRI studies of GBM-bearing mice were performed on a 7T MR scanner (Bruker Biospin), using a field of view of 2 cm × 2 cm and a slice thickness of 0.5 mm for the following acquisitions: T2-weighted fast-spin echo (FSE): repetition time (TR): 4,500 ms, echo time (TE): 42 ms, flip angle α: 90° and T2 multi-slice multi-echo (MSME): TR: 3,000 ms, TE: 8, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88 and 96 ms, α: 90°. T2 relaxation times of the tumors were calculated as a quantitative measure of tumor contrast enhancement.
Histopathologic correlations of MRI data
For Prussian blue iron stains, coronal brain tissue sections (5 μm) of formalin-fixed, paraffin-embedded tissue were deparaffinized with xylene, rehydrated, and stained according to the manufacturer's recommendation with the Sigma-Aldrich Accustain Iron Stain Kit. Sections were counterstained with nuclear fast red (Thermo Fisher Scientific). Representative images were captured using the Aperio ScanScope CS Slide Scanner with a 20× objective for whole-slide imaging.
Inductively coupled plasma mass spectrometry (ICP-MS)
ICP-MS analysis was conducted to study the biodistribution of theranostic CLIO-ICT nanoparticles. Briefly, 6‐to 8‐week‐old NOD scid gamma (NSG) mice (n = 3) were intravenously injected with PBS and CLIO-ICT (80 mg/kg of ICT, 0.5 mmol Fe/kg). After 24 hours, mice were euthanized and different organs including brain, heart, lung, liver, kidney, and spleen were harvested and processed for ICP-MS (inductively coupled plasma mass spectrometry) analysis. The tissues were weighed in preweighed glass tubes and treated with concentrated HNO3 (1 g tissue/mL HNO3) and 50 μL H2O2. The tubes were immersed in oil bath at 110°C overnight to completely digest the samples. Samples were considered completely digested when there was no liquid left in tubes. The samples were diluted with 5 mL of 2% HNO3, gently vortexed, and filtered through 0.44-micron Millipore filter and subjected to ICP-MS analysis. A standard solution of iron in the detection range was also included. Amount of iron per sample was obtained and expressed as iron per gram of tissue.
Determination of plasma vWF
We also evaluated the levels of von Willeband Factor (vWF) in plasma from animals treated with PBS, ICT, and CLIO-ICT. vWF is a marker of cardiotoxicity and was detected with kit based ELISA assays (MyBiosource.com) following manufacturer's protocol. Briefly, blood was collected from animals on day 5 after treatment with PBS, ICT, and CLIO-ICT into heparinized tubes to separate the plasma and stored in −80°C. The plasma was diluted 1:3 and 100 μL of samples or standards were added to murine vWF antibody precoated plates for 90 minutes at 37°C. The plate was washed with washing solution from the kit, followed by incubation with biotinylated vWF antibody for 60 minutes at 37°C. Then the plate was washed again and enzyme-working solution was added for 30 minutes at 37°C. After five washes, 100 μL of color reagent was added for not more than 30 minutes. The reaction was then stopped and OD values were measured at 450 nm on a microplate reader. A known concentration of vWF was used to generate a standard curve. The OD of samples was plotted over a standard curve and the concentration of plasma vWF was calculated after accounting for dilution factor. Our results indicated that CLIO-ICT and ICT did not induce any further increase in plasma vWF levels as compared with PBS. However, mice administered with doxorubicin showed slightly higher levels of vWF and served as our positive control.
Cell viability and apoptosis experiments were performed in biological triplicates, in three independent experiments. Magnetic resonance, immunofluorescence, and flow cytometry images were representative from three independent experiments. Results were presented as mean ± SD unless otherwise presented. Tumor and organ relaxation rates and fluorescence signals were compared within multiple experimental groups using one-way ANOVA. Results were analyzed using nonparametric two‐tailed Mann–Whitney test to compare two groups. Kaplan–Meier survival curves were compared using the log‐rank (Mantel–Cox) test. The level of significance was set at P < 0.05, as compared with the control group. Statistical analyses were carried out with Prism 6.0 software (GraphPad).
Physicochemical properties of CLIO-ICTs
The concept for MMP-14 activatable theranostic nanoparticles, CLIO-ICTs in GBMs is shown in Fig. 1A. ICT and CLIO-ICT were prepared and characterized as described previously (29). The hydrodynamic diameter of CLIO-ICT nanoparticles, as measured with a Zetasizer Nano ZS analyzer, was 22.10 ± 0.78 nm, slightly larger than ferumoxytol nanoparticles (20.27 ± 1.02 nm). The average number of ICT molecules per iron oxide nanoparticle was 4.7, based on the attached fluorescein absorption and known CLIO-ICT concentration. CLIO-ICTs had higher r2 relaxation rates (276.07 mmol/L−1 s−1) compared with original ferumoxytol nanoparticles (101.18 mmol/L−1 s−1, respectively; Fig. 1B). This is likely due to the modified surface coating.
In vitro antitumor effects of CLIO-ICTs in human GBMs
To evaluate MMP-14 levels, we analyzed the gene expression profile of MMP-14 by qPCR in five human GBM cell lines. HCN2, a cell line derived from normal neural cortex expressed very low levels of MMP-14 and served as our negative control (Fig. 1C). Expression data for MMP-14 was collected as Ct values and the gene expression levels were normalized to the reference control gene, GAPDH. While the cell lines U87 and U138 expressed increased levels of MMP-14 compared with HCN2 cells (P < 0.005), the A172 cell line showed no significant MMP-14 expression (P = 0.103; Fig. 1C). Neurospheres were derived from two GBM patients and cultured in neurosphere medium (30, 32). Compared with HCN2 cells, both pcGBM2 (P = 0.0027; Fig. 1C) and pcGBM39 (P = 0.001; Fig. 1C) showed significantly increased MMP-14 expression.
Next, we investigated the effect of theranostic CLIO-ICT nanoparticles on GBM cell viability. In all MMP-14–high GBM cell lines, treatment with both ICT (10 nmol/L) and CLIO-ICT (10 nmol/L ICT; 0.01 mmol/L CLIO) induced significant (P < 0.005) loss of cell viability when compared with CLIO (0.01 mmol/L) or DMSO (Fig. 1D). A172 cells were not responsive to treatment and MMP-14–negative normal neural cortex cells (HCN-2) did not show any significant cytotoxic effects after incubation with CLIO-ICT or ICT (Fig. 1D). The dose range of CLIO-ICT and ICT was chosen based on previous studies (29, 33, 34) and in the case of CLIO, was calculated on the basis of plasma concentrations of CLIO reached in patients after clinically applied doses (35–37). In apoptosis assays, GBM cells expressing high MMP-14 showed higher signal for both active caspase-3 fluorescence and Annexin-V/PI staining after incubation with CLIO-ICT and ICT (Fig. 1E and F; Supplementary Fig. S1A). A172 and HCN2 cells demonstrated minimal induction of caspase-3 upon treatment with CLIO-ICT (Fig. 1E). In addition, experiments with MMP inhibitor ilomastat rescued CLIO-ICT toxicity in GBMs, suggesting that activation of the CLIO-ICT was MMP-specific (Supplementary Fig. S1B). These results are consistent with MMP-14 selectivity shown for CLIO-ICT and ICT2588 as shown in previous studies (26, 29).
The apoptotic effects of most VDAs are a result of tubulin disruption in tumor cells and tumor endothelial cells. To evaluate whether our theranostic nanoparticles retained this therapeutic effect, we conducted α-tubulin staining in treated and control MMP-14–expressing GBM cells. Cells treated with both CLIO-ICT and ICT showed dramatic alteration in tubulin morphology (Fig. 1G). Control cells showed normal tubulin morphology as depicted by spindle shape of the microtubule structures, whereas treated cells showed retracted tubulin fibers that were aggregated around the nuclear periphery. The amount of tubulin was significantly reduced in presence of CLIO-ICT, which was quantified as total fluorescence in the red channel. Colchicine-treated cells were used as positive controls in these experiments (Supplementary Fig. S1C).
CLIO-ICTs induce apoptosis in GICs in vitro
To study the therapeutic effects of CLIO-ICTs on GICs, we incubated GBM cultures with CLIO-ICT (10 nmol/L) and examined GIC apoptosis by Annexin-V/DAPI staining.
We defined GICs (Supplementary Fig. S2) by a combination of established cell surface markers, CD133, CD15, and CD49F (38, 39). Considering the heterogeneity of tumors, we confirmed the reliability of these markers in identifying GICs in pcGBM39 and pcGBM2 (Supplementary Fig. S2), sorting CD133+CD15+, CD133−CD15−, CD49F+CD15+, and CD49F+CD15− cells from pcGBM39 and conducting in vivo tumorigenic assay (Supplementary Fig. S3). When injected into the right parietal lobe of NSG mice, only the CD15+/CD133+ and CD49F+/CD15+ population would generate tumors as early as 8 weeks, suggesting their tumorigenic potential. This confirmed the GIC like phenotype of these cells (Supplementary Fig. S3A and S3B).
Next, we investigated whether CLIO-ICTs could induce apoptosis of GICs. pcGBM39 cells treated with CLIO-ICT demonstrated a significant reduction in the percentage of CD133+, CD15+, and CD49F+ populations (P < 0.05; Fig. 1H and I). Furthermore, CD133+, CD15+, and CD49F+ cells from CLIO-ICT–treated pcGBM39 showed significant apoptosis as detected by Annexin-V/DAPI staining (P < 0.005; Fig. 1H and I).
Antitumor effects of CLIO-ICTs in orthotopic brain tumors initiated from patient-derived GBM neurospheres
As the VDA activity in tumors is documented (26), we hypothesized that CLIO-ICT will target tumor vasculature, GICs, and GBM cancer cells (Fig. 2A). To establish an orthotopic mouse model for GBM, tumor cells from patient-derived GBM neurospheres, pcGBM2 and pcGBM39, were injected into right parietal lobes of NSG mice and tested for efficacy of CLIO-ICT (Fig. 2B). The in vivo growth of these GBM neurospheres has been previously documented (30, 32). MMP-14 expression in these tumors was confirmed prior to drug treatment (Supplementary Fig. S4). Both pcGBM2 and pcGBM39 were engineered to express firefly-luciferase-GFP and firefly-luciferase–expressing reporter, respectively, for in vivo tumor detection with in vivo bioluminescence imaging.
At 3 to 4 weeks postintracranial cell injection, pcGBM39 tumors were detectable in the brain, whereas pcGBM2 tumors took longer (8 weeks). Mice were randomized into four treatment cohorts: each were intravenously injected with CLIO-ICT, ICT, CLIO, or PBS twice a week for 14 days. Total cumulative doses for each drug regimen were: CLIO-ICT (80 mg/kg of ICT), ICT (80 mg/kg of ICT), and CLIO (0.5 mmol Fe/kg). A longitudinal time course study was performed to study the antitumor properties of the theranostic nanoparticles. Both CLIO-ICT and ICT had significant antitumor effects on two different GBM types, whereas PBS-treated animals demonstrated increase in tumor size (Fig. 2C). No tumor response was observed in the CLIO-treated mice, suggesting that the therapeutic effect is derived directly from the released VDA entity (Fig. 2C). CLIO-ICT significantly inhibited pcGBM39 tumor growth and achieved remission of pcGBM2 tumors (Fig. 2C and D, P < 0.005). PBS- or CLIO-treated animals drastically lost body weight over time, while CLIO-ICT and ICT-treated sets demonstrated no significant weight loss (P > 0.05; Supplementary Fig. S5). CLIO-ICT improved survival by 100%, 200%, and 220% compared with ICT, CLIO, or PBS-treated mice (P < 0.05; Fig. 2E). Pathologic evaluation confirmed decreased tumor size and increased tumor apoptosis following CLIO-ICT treatment compared with CLIO or PBS treatment (Fig. 3A and B). We also investigated the biodistribution of CLIO-ICT in tumor free NSG mice using inductively coupled plasma mass spectrophotometry (ICP-MS). Results show that CLIO-ICT was accumulated in lungs, liver, spleen, and kidney (Supplementary Fig. S6A, P < 0.05). Brain and heart showed very little CLIO-ICT (Supplementary Fig. S6A, P > 0.05). Next, we measured CLIO-ICT toxicity in normal organs (brain, heart, lung, liver, spleen, and kidney) by evaluating the amount of apoptotic marker, caspase-3. Immunofluorescence data showed negligible expression of active caspase-3 in these organs from CLIO-ICT–treated animals when compared with PBS-treated group (Supplementary Fig. S6B). As previous VDAs are known to induce cardiotoxicity, we further confirmed that CLIO-ICT is minimally cardiotoxic. In addition to caspase-3 staining, we also measured plasma vWF levels in both PBS and CLIO-ICT–treated animals. vWF is conventionally used as a marker for cardiotoxicity (28). Our data show that no significant difference was observed for plasma vWF levels in PBS and CLIO-ICT–treated animals. In summary, CLIO-ICT was found to be nontoxic to normal organs (Supplementary Fig. S6C).
To further compare tumor retention of macromolecular CLIO-ICT and small molecular ICT, we measured drug quantities of FITC-labeled ICT and CLIO-ICT in pcGBM39 tumor specimen with immunofluorescence assays. CLIO-ICT–treated tumors demonstrated significantly higher FITC intensity as compared with ICT only (P < 0.05), suggesting improved tumor retention of ICT when linked to a nanocarrier (Fig. 3B).
To better understand the tumor retention of CLIO-ICT in vivo, we next imaged pcGBM39-bearing mice on a 7.0 T small-animal MRI system. We found that GBM showed a signal drop or negative (hypointense) enhancement on T2-weighted MR images after intravenous injection of CLIO and CLIO-ICT, whereas PBS- or ICT-administered mice showed little or no loss in tumor signal (Fig. 3C and D). When comparing pre- and postinjection images in each group, we found that tumor T2 relaxation times decreased significantly after CLIO and CLIO-ICT injection (P < 0.005) and did not change significantly after ICT or PBS injection (Fig. 3C and D). When comparing different treatment groups, tumor T2 relaxation times of CLIO and CLIO-ICT–treated tumors were significantly shorter compared with the ICT only and PBS groups (P < 0.05). This was consistent with histopathologic evaluations: Prussian blue staining demonstrated iron deposits within CLIO and CLIO-ICT–treated groups, but not PBS or ICT-treated groups (Fig. 3A).
Next, longitudinal MR scans were obtained on week 1 and 2 for each group to noninvasively monitor therapy response in tumors. For each time point, tumor volume was calculated and plotted against time. CLIO-ICT was most effective in inhibiting tumor growth (Fig. 3E). In addition, the degree of tumor T2 signal enhancement, quantified as tumor T2 relaxation time, correlated with tumor volume (Spearman rank correlation, r = 0.7; P < 0.001). Thus, CLIO-ICTs could be used for noninvasive monitoring of drug delivery and therapy response in orthotopic GBM models using MRI.
To understand the therapeutic mechanism, we evaluated the effect of CLIO-ICT on tumor vasculature and GICs. We used CD31 as a marker for GBM endothelial cells (7, 40). Both CLIO-ICT and ICT treatment significantly inhibited MVD (microvessel/mm2) in pcGBM39 tumors (P < 0.05; Fig. 4). In addition, CLIO-ICT significantly reduced the GICs in pcGBM39-bearing mice, as demonstrated by reduced staining for CD133, CD15, and Nestin (P < 0.005; Fig. 4) and significantly increased staining for Annexin-V/DAPI (P < 0.05; Fig. 5). We confirmed this observation using flow cytometry and firefly-luciferase-GFP–expressing pcGBM2 cells. We observed that CLIO-ICTs triggers GBM cell apoptosis in vivo, as GBM cells (GFP-positive) from CLIO-ICT-treated mice presented higher levels of Annexin-V and DAPI staining compared with glioblastoma cells from vehicle-treated mice (Fig. 5A and B). Of note, CLIO-ICT did not induce apoptosis in nontumor stromal (GFP-negative) cells (Fig. 5B and C), further confirming the absence of toxicity of this drug on normal cells. The percentage of CD15+ GICs was significantly reduced in presence of CLIO-ICTs, whereas apoptosis staining in CD15+ GICs from CLIO-ICT–treated mice was significantly increased (P < 0.05; Fig. 5B and D).
The above results suggest that the therapeutic efficacy of CLIO-ICT is due to disruption of tumor vasculature and GIC apoptosis.
CLIO-ICTs in combination with temozolomide impedes tumor growth and improves survival in GBM-bearing animals
Next, we assessed whether CLIO-ICT could be used in combination with standard clinical anti-GBM treatment, such as temozolomide. Before proceeding to in vivo studies, a panel of GBM cells was screened in vitro for temozolomide sensitivity with MTS-based viability assays, which demonstrated varying profiles of temozolomide toxicity (Fig. 6A). For our in vivo studies, we proceeded with temozolomide-responsive pcGBM39 and pcGBM2 cells to test whether CLIO-ICT would add therapeutic benefit. At three and eight weeks postintracranial injection, pcGBM39 and pcGBM2-bearing mice were randomized into treatment groups: temozolomide (200 mg/kg; cumulative dose) was orally administered either alone or in combination with ICT/CLIO-ICT (80 mg/kg; cumulative dose), twice per week for three consecutive weeks. At 3 weeks after initiation of treatment, MRI scans revealed 7-fold reduction in tumor size for temozolomide plus CLIO-ICT–treated pcGBM39-bearing mice relative to that of vehicle-treated mice, which showed 30-fold increase in tumor size after treatment (Fig. 6B and C). The combination of TMZ plus CLIO-ICT significantly inhibited pcGBM39 tumor growth (P < 0.005), induced complete remission of pcGBM2 tumors, and increased overall survival as compared with temozolomide plus ICT and single-drug regimens (Fig. 6D–I).
Data showed that novel MMP-14–activatable cross-linked iron oxide nanoparticles (CLIO-ICT), which where conjugated with the VDA ICT (a structural analogue of ICT2588), caused significant apoptosis of cancer cells and GICs in GBM. This led to significantly prolonged survival of pcGBM39-bearing mice and complete tumor remission of pcGBM2-bearing mice. CLIO-ICT synergized with temozolomide in vivo, thereby revealing a novel therapeutic combinatorial regimen for targeting GBMs.
Current GBM therapies include surgical resection followed by radiotherapy with concurrent temozolomide treatment (41). However 90% of GBM patients develop tumor recurrence (42). This is partly attributed to poor drug delivery and drug-resistant GICs (43). Nearly all large molecules and the majority of small molecular drugs including tyrosine kinase inhibitors are effluxed at the blood brain barrier by P-gp and Bcrp on the luminal side of endothelial cells, and never reach therapeutic levels in the brain (44). Another major challenge involves the multiple molecular pathways of GBMs. Identifying just one therapeutic target is insufficient, as this does not account for the considerable array of genetic and epigenetic heterogeneities found within and among GBM patients (45). Our approach addresses most of these challenges: CLIO-ICT does not rely solely on a passive delivery across the tumor microvessel wall via the EPR effect (9, 10). The VDA rather increases vascular permeability through active disruption of the tumor microvessel endothelial lining. This effect is specific to tumor vessels through tumor enzyme activation (28), and efficacious in tumors with different molecular codes. Both ICT-mediated vessel disruption and nanoparticle-mediated enhanced permeability and retention (EPR) support accumulation of CLIO-ICT in the tumor interstitium. CLIO-ICT causes tumor blood vessel collapse, cancer cell starvation, and jeopardizes the tumor microenvironment, making this a very effective therapeutic strategy (26). We further demonstrate for the first time that CLIO-ICT is capable of inducing apoptosis in GICs.
Agemy and colleagues reported that theranostic nanoparticles linked to a tumor vasculature homing peptide (CGKRK) showed exceptional efficacy in eliminating a lentivirus-induced GBM model. However, this drug could not completely inhibit the growth of orthotopically transplanted GBM (46). In contrast, our CLIO-ICT achieved almost complete remission of orthotopic pcGBM2 tumors and significantly inhibited pcGBM39 tumors, both of which express MMP-14. Literature shows that MMP-14 expression is positively correlated with tumor grade and disease progression of gliomas (23, 47). Therefore, our MMP-14–targeted approach seems specifically suited for the treatment of high-grade GBMs.
As our theranostic nanoparticles are based on two clinically applicable components, the combined CLIO-ICT has high potential for clinical translation: CLIO-ICT is composed of the FDA-approved nanoparticle compound ferumoxytol, linked to a VDA based on ICT2588. ICT2588 is on course to be evaluated in a phase I clinical trial in the United Kingdom for treatment of advanced tumors outside of the CNS in 2017. Further since CLIO-ICT are activated by tumor MMP-14 to release the potent VDA specifically in tumor tissue, potential side effects of cardiotoxicity, as noted with previous VDAs, are minimized (28).
VDA-based strategies sensitize GBMs to chemotherapy through three mechanisms: (i) VDAs jeopardize the perivascular GIC niche, which provides access to vascular nutritions but shields GIC from classical chemotherapy agents, such as temozolomide. (ii) VDAs disrupt tumor-promoting vascular endothelial cells. Borovski and colleagues, found that tumor microvascular endothelial cells (tMVEC) isolated from human GBM tumors promoted the proliferation of human CD133+ cells when GIC-tMVEC cocultures were exposed to irradiation and temozolomide (48, 49). In contrast, CLIO-ICT had inhibited both tumor endothelial cells and GIC, thereby eliminating endothelial cell–mediated GIC rescue. (iii) CLIO-ICT is activated by tumor-specific enzymes, that is, the activation is “built-in” and does not rely on external stimuli. In contrast, previous generations of theranostic nanoparticles had to be activated by near-infrared phototherapy (50). These approaches are more difficult to translate to the clinic and were accompanied with severe side effects, such as severe cerebral edema (50).
In summary CLIO-ICT holds significant clinical potential for improving targeted therapy and survival in GBM patients. Significant advantages of CLIO-ICT compared with other theranostic agents include selective and effective delivery to GBM, therapeutic efficacy independent of EPR effect, prolonged retention in the tumor tissue via VDA initiated vascular collapse, drug entrapment in the GIC niche, ability to track the drug with MRI, and eliminated toxicity-liability to normal organs. Further, our approach of combining CLIO-ICT with temozolomide offers refinements to current less optimal standards of care to achieve GBM remission.
Disclosure of Potential Conflicts of Interest
R.A. Falconer and P.M. Loadman declare that they are founding shareholders of Incanthera Ltd. No potential conflicts of interest were disclosed by the other authors.
Conception and design: S. Mohanty, S. Cheshier, J. Rao, P.M. Loadman, R.A. Falconer, H.E. Daldrup-Link
Development of methodology: S. Mohanty, Z. Chen, L. Pisani, F.T. Chin, S. Mitra, S.S. Gambhir, H.E. Daldrup-Link
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Mohanty, K. Li, G.R. Morais, J. Klockow, K. Yerneni, L. Pisani, F.T. Chin, S. Cheshier, E. Chang, S.S. Gambhir, P.M. Loadman, R.A. Falconer
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Mohanty, K. Li, G.R. Morais, K. Yerneni, L. Pisani, S. Mitra, S.S. Gambhir, J. Rao, H.E. Daldrup-Link
Writing, review, and/or revision of the manuscript: S. Mohanty, K. Li, F.T. Chin, S. Cheshier, E. Chang, S.S. Gambhir, J. Rao, P.M. Loadman, R.A. Falconer, H.E. Daldrup-Link
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Mohanty, L. Pisani, E. Chang, H.E. Daldrup-Link
Study supervision: S. Mohanty, S. Mitra, S.S. Gambhir, P.M. Loadman, R.A. Falconer, H.E. Daldrup-Link
Other (provided critical feedback on data shown in all versions of manuscripts): E. Chang
We thank the Small Animal Imaging Facility at Stanford for providing the equipment and infrastructure for this project. We thank Jamal Elbakay for technical assistance. We thank members of the Daldrup-Link, Gambhir, and Cheshier and ICT Bradford laboratories for valuable discussions.
This work was in part supported by grants from the National Cancer Institute NIH, R21CA176519 and R21CA190196 (to H.E. Daldrup-Link), the NIH 1U54CA199075 (to S.S. Gambhir), NCI training grant (T32CA118681; to J. Klockow), and University of Bradford, UoB-66031 (to R.A. Falconer).
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.