Targeted disruption of tumor vasculature via polyphenol nanoparticles to improve brain cancer treatment.

Despite being effective for many other solid tumors, traditional anti-angiogenic therapy has been shown to be insufficient for the treatment of malignant glioma. Here, we report the development of polyphenol nanoparticles (NPs), which not only inhibit the formation of new vessels but also enable targeted disruption of the existing tumor vasculature. The NPs are synthesized through a combinatory iron-coordination and polymer-stabilization approach, which allows for high drug loading and intrinsic tumor vessel targeting. We study a lead NP consisting of quercetin and find that the NP after intravenous administration preferentially binds to VEGFR2, which is overexpressed in tumor vasculature. We demonstrate that the binding is mediated by quercetin, and the interaction of NPs with VEGFR2 leads to disruption of the existing tumor vasculature and inhibition of new vessel development. As a result, systemic treatment with the NPs effectively inhibits tumor growth and increases drug delivery to tumors.

INTRODUCTION

Malignant gliomas are the most common primary brain tumors in adults. With the diagnosis comes a dismal prognosis, with a median survival of ~14 months.[1-3] The current standard of care involves surgical resection, followed by radiation and chemotherapy with temozolomide (TMZ). Despite the aggressive treatment, tumor recurrence is all but ensured.[1-3]

The lack of effective therapies for malignant gliomas could be largely attributed to the existence of the unique vasculature in those tumors, which not only supports tumor growth through the supply of oxygen and essential nutrients but also serves as the blood-brain barrier (BBB) in tumors, or blood-brain-tumor barrier (BBTB), to prevent drug delivery to the brain.[4,5] One promising approach to improve the treatment of malignant gliomas could be vessel-targeting therapy, such as anti-angiogenesis therapy through inhibition of the formation of new blood vessels.[6,7] Clinically, anti-angiogenesis therapy through intravenous administration of bevacizumab, a humanized monoclonal antibody targeting vascular endothelial growth factor (VEGF), has proved to be effective for the treatment of various solid tumors.[8] Unfortunately, the same therapeutic approach failed to provide overall survival benefit for patients with malignant gliomas.[6,7] As an alternative approach, vessel-targeting therapy through the disruption of the existing tumor vasculature has been explored. A few vascular-disrupting agents, such as ASA404 (DMXAA), have been tested in clinical trials and shown to be promising for some cancers,[9,10] while, similar to the anti-angiogenic therapy, the vascular-disrupting approach failed to effectively inhibit glioma development.[11] These findings suggest that vessel-targeting therapy through either inhibition of neovessel generation or disruption of the existing vasculature alone is insufficient for malignant gliomas, and effective treatment of the disease may require addressing both the existing tumor vasculature and neovessels simultaneously.

In this study, we report the development of polyphenol nanoparticles (NPs), which not only inhibit the growth of new vessels but also selectively disrupt the existing tumor vasculature. Polyphenols represent one of the largest categories of phytochemicals in plants, many of which are known to have excellent anti-angiogenic effects.[12-16] Nonetheless, most natural polyphenols have limited bioavailability because of poor solubility and stability as well as potential systemic toxicity, and thus cannot be translated for clinical applications.[12,13] To overcome this limitation, we formulated polyphenols into NPs through a combinatory Fe-coordination and polymer-stabilization approach and screened a collection of anti-angiogenic polyphenols. Through the screen, we identified NPs consisting of quercetin, which showed the greatest anti-angiogenic activities. We characterized quercetin NPs in mouse glioma models and unexpectedly found that quercetin NPs after intravenous administration selectively disrupted the existing tumor vasculature, leading to improved survival of tumor-bearing mice and enhanced drug delivery to brain tumors. The finding of dual-functional polyphenol NPs may suggest an effective vessel-targeting approach for the treatment of malignant gliomas.

RESULTS

Synthesis and screen of polyphenol NPs

Clinical translation of anti-angiogenic polyphenols is limited by their poor bioavailability.[12,13] To overcome this limitation, we developed an approach to assemble polyphenols into NPs through coordination via Fe, used as a powerful chemistry for the assembly of surface nanostructures.[17,18] The resulting nanostructures were further stabilized via amphiphilic polymers. By using catechin and curcumin, two hydrophobic polyphenols, and dopamine, a hydrophilic polyphenol, as examples, we showed that polyphenols were unable to form dispersed NPs without the presence of amphiphilic polymers, and, among all test polymers, F127 markedly improved the morphology and dispersion (Figure S1). Through this approach, we formulated 15 anti-angiogenic polyphenols into NPs, which are in a spherical shape and in diameters ranging from 2 to 150 nm (Figures 1A, 1B, and S2A; Table S1). The resulting NPs were evaluated for their anti-angiogenic activity based on the standard tube formation assay. HUVEC cells were engineered to express green fluorescent protein (GFP) and treated with the selected NPs at 200 μg/mL. Thirty minutes later, the cells were plated onto Matrigel. After 24 h of incubation, the formation of tubes was imaged. Analysis of the tube images found that compared to PBS control, treatment with many of the tested NPs reduced the number of junctions and tubular structures (Figures S2B and S2C). Among them, NPs consisting of ellagic acid, gallic acid, and quercetin exhibited the greatest inhibitory effects (Figures 1C and 1D). As we aimed to select NPs suitable for brain cancer treatment, we evaluated ellagic acid-, gallic acid-, and quercetin-NPs for their efficiency in brain cancer targeting in mice bearing GL261 glioma, which is one of the most widely used models for brain cancer research.[19-22] NPs were labeled with IR780, an infrared dye that allows for non-invasive imaging, and administered to mice 21 days after tumor inoculation through tail vein injection. After 12 h, the mice were euthanized. The brains were isolated and imaged using an in vivo imaging system (IVIS). We found that all three selected NPs penetrated brain tumors, and, among them, quercetin-NPs showed the greatest efficiency (Figure 1E). We found that the marked anti-angiogenic effect of quercetin NPs were not due to their cytotoxicity, as the NPs at a concentration of up to 200 μg/mL did not induce significant cell killing (Figure 1C). Due to their significant anti-angiogenesis activity and high brain tumor-targeting efficiency, quercetin-NPs were selected for further evaluation. To simplify the nomenclature, we designated quercetin-NPs as Q-NPs.

Figure 1.

Synthesis and characterization of polyphenol NPs

(A) Molecular structures of the selected polyphenols and diameters of NPs derived from the corresponding polyphenols. Blue: hydrophobic. Black: hydrophilic. Number represents the average diameter of NPs assembled from the specific polyphenol.

(B) Representative TEM images of gallic acid-, quercetin-, and ellagic acid-NPs. Scale bar: 50 nm.

(C) Cytotoxicity and anti-angiogenesis activities of NPs assembled from the selected polyphenols at the indicated concentrations; n = 5.

(D) Representative images of HUVEC-formed tubes after treatment with PBS or the indicated NPs. Scale bar: 1,000 μm; n = 3.

(E) Representative images (upper panel) and quantification (bottom panel) of IR780-labeled NPs accumulated in GL261 gliomas after intravenous administration.

Data represent means ± SEMs; n = 3.

Physical, chemical, and biological characteristics of Q-NPs

To obtain a formulation optimized for the maximized loading of quercetin, we synthesized an array of Q-NPs by varying the feeding concentration of free quercetin. Loading of quercetin was determined by the standard Folin-Ciocalteu method.[23] We found that the loading efficiency increased with the concentration initially, but reached a plateau at concentrations >7.5 mg/mL (Figure S3). Therefore, we selected the formulation synthesized with a feeding concentration of 7.5 mg/mL, which contain quercetin by 73% by weight.

We determined physical and chemical properties of Q-NPs in the selected formulation. Analysis by dynamic light scattering (DLS) showed that Q-NPs have an average hydrodynamic diameter of 16.3 nm and a negative surface charge of −20.2 mV (Figure S4). Analysis by energy dispersive X-ray spectroscopy (EDS) confirmed the existence of Fe, C, and O, among which Fe occupies 3.9% by weight (Figure 2A). Analysis by X-ray powder diffraction (XRD) found 2 peaks centered at 19.2° and 23.6°, suggesting that Q-NPs possess a high degree of crystallinity (Figure 2B). Analysis by Fourier transform infrared spectrum (FTIR) identified 3 peaks at 1,415, 1,594, and 1,644 cm−1 in the spectrum of Q-NPs but not in the spectrum of F127-Fe complexes, suggesting the coordination between polyphenols and Fe molecules. Further analysis of the FTIR spectra found 3 new peaks at 732, 784, and 1,214 cm−1 associated with F127-Fe complexes but not F127, indicating a possible interaction between Fe3+ and F127 (Figure 2C). Transmission electron microscopy (TEM) analysis found that, unlike Q-NPs, complexes consisting of either quercetin-F127 without Fe or quercetin-Fe without F127 had poor morphology (Figure S5). In addition to morphology, both Fe and F127 also contribute to stability of Q-NPs. We studied the interaction forces between the polyphenol-Fe complexes and F127 by determining the effects of NaCl, urea, and Triton X-100, which interfere with electrostatic interaction, hydrogen bonding, and hydrophobic interaction, respectively, on the self-assembly process.[24] We found that only Triton X-100 significantly altered the hydrodynamic size of Q-NPs, suggesting that hydrophobic interaction, but not other interactions, plays a role in NP self-assembly (Figure S6). We found that the critical micelle concentration (CMC) for Q-NPs was 4.5 × 10−6, which is significantly lower than that for F127 (3 × 10−5) (Figure 2D).

Figure 2.

Physical, chemical, and biological characteristics of Q-NPs

(A) EDS spectrum of Q-NPs.

(B) XRD spectrum of Q-NPs.

(C) FTIR spectra of Q-NPs, F127, and F127-Fe.

(D) Change in I372/I383 with concentrations in deionized water for Q-NPs and F127. Data represent means ± SEMs; n = 3.

(E and F) Characterization of cytotoxicity (E) and anti-angiogenesis activity (F) of Q-NPs at the indicated concentrations. Data represent means ± SEMs; n = 5.

(G) Representative microscopic images of the formation of HUVEC tubes after treatment with Q-NPs at the indicated concentrations. Scale bar: 1,000 μm; n = 5.

Preliminary studies showed that Q-NPs inhibited angiogenesis by 95% at 200 μg/mL, a concentration at which the NPs exhibited limited cytotoxicity (Figure 1C). We further characterized the anti-angiogenic activity of Q-NPs at serial concentrations. We found that while the half-maximal inhibitory concentration (IC50) of Q-NPs on cell proliferation is >1,000 μg/mL, the IC50 on tube formation is 39.2 μg/mL (Figures 2E–2G). In comparison, free quercetin has a limited solubility of 0.002 mg/mL in aqueous solution.[25] In a demonstration study, we dissolved 100 mg quercetin in 1 mL DMSO, and added the solution to 100 mL PBS to a final concentration of 1 mg/mL. We found that quercetin was immediately precipitated out (Figure S7A). Similar precipitation occurred when quercetin in DMSO was added to cell culture medium at final concentrations >125 μg/mL. As a result, the precipitated quercetin covered the top of the cells and killed cells through suffocation in addition to toxicity (Figures S7B and S7C). For this reason, free quercetin was not included in the cytotoxicity and angiogenesis studies.

It was shown that some polyphenols, such as tannic acid, react with Fe, resulting in Fenton reaction-mediated chemodynamic therapy.[26,27] To test whether the Fenton reaction plays a role in Q-NPs-mediated antitumor activities, we determined intracellular oxidative stress following Q-NP treatment through dichlorofluorescein diacetate (DCFH-DA) staining. We found that treatment with Q-NPs at the tested concentrations did not significantly increase intracellular oxidative stress (Figure S8), suggesting that the Fenton reaction does not play a role in the observed biological activities of Q-NPs. This could be explained by the fact that compared to tannic acid, quercetin has a lower density of phenolic hydroxyl groups and thus has limited efficiency in reaction with Fe molecules.[27] Collectively, these data suggest Q-NPs as a potent anti-angiogenesis agent without significant cytotoxicity.

Q-NPs inhibit angiogenesis and selectively disrupt the existing tumor vasculature

Q-NPs efficiently accumulated in GL261 gliomas in the brain after intravenous administration (Figure 1E). To further characterize the brain-targeting effect, we determined the kinetics of Q-NP accumulation in tumors. Three weeks after the inoculation of GL261 cells, mice were treated with IR780-labeled Q-NPs through tail vein injection. The accumulation of IR780 signal in the brain was monitored by IVIS. We found that the signal in the NP treatment group continued to increase in the first 12 h, after which it decreased (Figure 3A). In contrast, under the same imaging condition, the signal in the group treated with free dye was below the detection limit (Figure S9). Therefore, we chose 12 h as the time point for further characterization. Next, we characterized Q-NPs for drug delivery to brain tumors by using poly(lactic-co-glycolic acid) (PLGA) NPs, one of the most often used NPs for drug delivery, as the benchmark. Both NPs were synthesized with encapsulation of IR780 and intravenously administered to tumor-bearing mice. Control mice were treated with free dye. Doses for all groups were normalized to ensure that each mouse received the same amount of dye. At 12 h after treatment, the mice were euthanized. The brains were isolated and subjected to IVIS imaging. The results in Figure 3B show that compared to both free dye and PLGA NPs, Q-NPs accumulated in the brain at a significantly greater efficiency. We further compared Q-NPs with engineered dendrimer NPs, which have a size comparable to Q-NPs (10 nm for the dendrimer NPs versus 16 nm for Q-NPs) and were reported to cross the BBB in brain tumors.[28] Both NPs were synthesized with encapsulation of Cy 5.5 and intravenously administered to tumor-bearing mice. Control mice were treated with free dye. Doses for all of the groups were normalized to ensure that each mouse receives the same amount of dye. The mice were euthanized 12 h later. The brains were isolated and subjected to IVIS imaging. We found that, compared to both free dye and dendrimer NPs, Q-NPs accumulated in the brain at a significantly greater efficiency (Figure S10). IR780-based imaging by IVIS, although suitable for the detection of NPs, does not have the resolution to differentiate tumors from the normal brain nor to allow quantification of the delivery efficiency. To determine the difference in delivery efficiency between Q-NPs and PLGA NPs, we synthesized NPs with encapsulation of europium, a rare earth metal that does not naturally exist in the mouse brain. The resulting europium-loaded NPs, together with an equivalent dose of free europium as a control, were injected into tumor-bearing mice. After 12 h, the mice were euthanized. Tumors in the brain were isolated and subjected to quantification of europium by inductively coupled plasma-mass spectrometry (ICP-MS). We found that the average percentage injected dose per mouse brain (%ID/g of dry weight) for Q-NPs was 112.3%, which is 18.7 and 131.7 times greater than PLGA NPs and free agents, respectively (Figure 3C). The analysis also showed that the amount of Q-NPs accumulated in tumors was 15.1 times greater than that in the normal brain (Figure 3C). Using the same approach, we quantified the accumulation of Q-NPs in other organs and found that, compared to that in tumors, the amounts of Q-NPs in the heart, liver, spleen, lung, and kidney were significantly lower (Figure S11). To further confirm the findings, we quantified the accumulation of quercetin in the brain by liquid chromatography-MS (LC-MS) without the use of surrogate europium. Tumor-bearing mice were established and received intravenous administration of Q-NPs at 25 mg/kg. Twelve hours later, the mice were euthanized. Tumors in the brain and control normal brains were isolated and subjected to LC-MS analysis. We found that the concentration of quercetin in brain tumors was 87.9 μg/g (wet tissue), which is 18.3 times greater than that in normal brains (Figure S12), and the equivalent %ID/g of dry weight is 132.1%. Both findings were consistent with those determined by ICP-MS.

Figure 3.

Characterization of the interaction of Q-NPs with tumor vessels

(A) Change in the number of Q-NPs in brain tumors based on IR780 florescence quantification with time in mice receiving intravenous administration of IR780-labeled Q-NPs. Data represent means ± SEMs; n = 3.

(B) Representative images of free IR780, IR780-labeled Q-NPs, and PLGA NPs in the brain; n = 3.

(C) Quantification of %ID/g of free agent, Q-NPs, and PLGA NPs based on ICP-MS analysis. Data represent means ± SEMs; n = 3.

(D) Representative images of Q-NPs (red) in brain tumors (upper panel) or normal brain tissues (bottom panel). Blue: DAPI. Yellow: CD31. GFP: tumor cells. Scale bar: 100 μm.

(E) Representative images of brain tissues stained for Fe (blue) by Prussian Blue and red blood cells (red) by H&E. Scale bar: 100 μm. Free quercetin was given through intraperitoneal (i.p.) injection.

(F) Representative images of brain tissues immunostained for CD31 (green) and VEGFR2 (red). Blue: DAPI.

(G) Representative images and quantification of IR780-labeled Q-NPs accumulated in the brains isolated from mice with and without pre-treatment of anti-VEGFR2 antibody. Data represent means ± SEMs; n = 3.

(H) Western blot analysis of the expression of VEGFR2 and VEGFR2 (Tyr-1175) in cells with the indicated treatments. The lanes were run on the same gel but were noncontiguous.

(I) Representative images of bEnd.3 cells treated with Q-NPs and immunostained for VE-cadherin (green). Blue: DAPI. Scale bar: 10 μm.

(J) Representative images of tumor or normal brain tissues isolated from mice receiving Q-NP treatment and immunostained for VE-cadherin (green). Blue: DAPI. Red arrows denote disrupted homophilic interaction of VE-cadherin. Scale bars for the top and the bottom panels are 100 and 10 μm, respectively.

Statistical analyses were performed using unpaired, 2-tailed Student’s t test by Prism 8 (GraphPad). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

We examined the distribution of Q-NPs in brain regions with and without tumors through high-resolution confocal imaging. Tumor-bearing mice were established through the inoculation of GL261 cells engineered to express GFP. Three weeks later, the mice were treated with Q-NPs labeled with Cy5, a red dye that is suitable for detection by fluorescence microscopy. After 12 h, a cohort of mice were intravenously administrated with R-Phycoerythrin (PE)-conjugated anti-CD31 antibody, which labels blood vessels. One hour later, the mice were euthanized. The brains were isolated, sliced, subjected to DAPI staining, followed by imaging by confocal microscopy. We found that the Q-NPs bound to vessels preferably in tumors, but not in the regions without tumors (Figure 3D; Video S1). We examined the distribution of Q-NPs 24 h after treatment and found that most NPs remained bound to vessels without penetrating tumors (Figure S13; Video S2). Further analysis found that the binding of Q-NPs, as identified by Fe staining, led to inhibition of the growth of new vessels, disruption of the existing vessels, which were detected by CD31 immunostaining (Figure S14A), and extravasation of red blood cells (Figure 3E). In contrast, intraperitoneal injection with free quercetin did not cause a detectable reduction in vessel density nor induce vessel disruption (Figures 3E and S14A). We found that Q-NP treatment did not disrupt vessels in the normal brain without tumors (Figure S14B).

To determine the molecular mechanism accounting for the unprecedented degree of brain tumor targeting and vessel-specific inhibition, we studied major surface receptors that are involved in tumor angiogenesis, including VEGF receptor 2 (VEGFR2), TIE2, fibroblast growth factor receptor (FGFR), insulin-like growth factor 1 receptor (IGFR), and epidermal growth factor receptor (EGFR).[29,30] Cy5-labeled Q-NPs were synthesized and incubated with HUVEC and bEnd.3 cells that were engineered to overexpress the candidate molecules. Six hours later, the cells were collected and analyzed by flow cytometry. We found that overexpression of VEGFR2, but not the other receptors, significantly enhanced the binding/uptake of Q-NPs in both cells (Figure S15). We further characterized the role of VEGFR2 using an anti-VEGFR2 antibody and found that blocking VEGFR2 using the antibody significantly reduced the uptake of Q-NPs (Figure S15). VEGFR2 is known to upregulate in vessels in tumors but not normal tissues.[31] The preferential expression of VEGFR2 in the tumor vasculature in GL261 gliomas was confirmed by immunostaining (Figure 3F; Video S3). To validate that the interaction of Q-NPs with tumor vessels is mediated by VEGFR2, we treated tumor-bearing mice with anti-VEGFR2 antibody. Control mice were treated with PBS. Six hours later, the mice were treated with IR780-labeled Q-NPs. After an additional 12 h, the mice were euthanized. The brains were isolated and imaged. We found that pre-treatment with anti-VEGFR2 antibody significantly reduced the accumulation of NPs in tumors by 73%, while no significant changes were found in other organs (Figures 3G and S16). We studied the mechanism accounting for the interaction of Q-NPs with VEGFR2. Quercetin is known to specifically bind with VEGFR2 through various interactions involving significant negative binding energies.[32] We hypothesized that the interaction of Q-NPs with VEGFR2 is mediated by quercetin molecules that are anchored on the surface of the NPs. To test this hypothesis, we treated VEGFR2-expressing bEnd.3 with free quercetin. After 1 h, the cells were treated with Q-NPs labeled with Cy5. After an additional 6 h, confocal imaging was performed. We found that pre-treatment with quercetin significantly reduced Q-NP accumulation in cells in a dose-dependent manner (Figure S17), suggesting that the interaction of Q-NPs with VEGFR2 is mediated by quercetin.

Activation of VEGFR2 is a critical step in the initiation of tumor angiogenesis.[33,34] At the molecular level, phosphorylation of Tyr1175, a phosphorylation site in VEGFR2, triggers a downstream signaling cascade, which leads to increased endothelial cell proliferation and migration.[33,34] Inhibition of angiogenesis by blocking the VEGFR2 signaling has been suggested as a promising approach for cancer treatment.[31,35] Both quercetin and Fe, two major components of Q-NPs, are known to inhibit angiogenesis through suppression of the VEGFR2 signaling.[36,37] We speculate that the observed anti-angiogenesis effect of Q-NPs also results from inhibition of the VEGFR2 signaling. To test the hypothesis, we determined the impacts of various treatments on the level of VEGFR2 expression and phosphorylation in HUVEC cells. Briefly, HUVECs were treated with DMSO (0.1%), ferric chloride (50 μM), quercetin (50 μM), or Q-NPs (50 μg mL−1). After incubation for 120 min, the cells were washed with PBS and subjected to western blot analysis. We found that, similar to free quercetin and Fe molecules, Q-NPs effectively inhibited phosphorylation of VEGFR2 (Tyr-1175) (Figure 3H).

We investigated the mechanism accounting for the observed Q-NP-induced vessel-disruption effect (Figure 3E). Previous studies showed that certain NPs induce tumor vessel disruption through disturbance of the homophilic interaction of vascular endothelial (VE)-cadherin, a phenomenon called “nanomaterials-induced endothelial leakiness” (NanoEL).[38-40] We hypothesize that the observed vascular disruption effect of Q-NPs may share the same mechanism. We tested the hypothesis by treating bEnd.3 cells with Q-NPs, followed by immunostaining for VE-cadherin. We found that Q-NP treatment effectively disturbed the homophilic interaction of VE-cadherin in a concentration-dependent manner (Figure 3I). We further characterized the homophilic interaction of VE-cadherin in tumor vessels in vivo. Mice bearing GL261 tumors were established and treated with Q-NPs at 25 mg/kg. Twenty-four hours later, the mice were euthanized. The brains were isolated, sliced, and subjected to VE-cadherin immunostaining. Consistent with the finding in cell culture, we found that the treatment effectively disturbed the homophilic interaction of VE-cadherin in tumors, while no detectable disturbance was found in the normal brain (Figure 3J).

Our results suggest that Q-NPs inhibit angiogenesis through interaction with VEGFR2 and disrupt the existing tumor vasculature through disturbance of the homophilic interaction of VE-cadherin.

Q-NPs enhance drug delivery preferentially to brain tumors

One key feature of NanoEL is that treatment with the selected nanomaterials leads to increased vascular endothelium permeability and enhanced extravasation of circulating drugs to tumors.[38-40] We assessed whether Q-NP-mediated vessel disruption has similar effects to enhance BBB penetration and drug delivery to brain tumors. We evaluated treatment with Q-NPs on BBB permeability in an in vitro Transwell-based BBB model, which was established by culturing human brain microvascular endothelial cells (HBMECs) and normal human astrocyte (NHA) cells on the top and the bottom of the upper chamber membrane, respectively (Figure 4A).[41] When the transendothelial electrical resistance (TEER) value reached ~100 Ω, Q-NPs were added to the upper chamber to a final concentration of 50 μg/mL. After a 1-h incubation, tetramethylrhodamine isothiocyanate (TRITC)-dextran (4.4 kDa) was added to the same chamber. At various time points, the medium in the bottom chamber was sampled and subjected to quantification for TRITC-dextran. The results in Figure 4B showed that treatment with Q-NPs significantly compromised the integrity of the barrier in a time-dependent manner. Next, we assessed Q-NP treatment on drug penetration into tumors in tumor-bearing mice. Paclitaxel (PTX), a commonly used chemotherapy drug that is known to have a limited ability to penetrate brain tumors,[42] was used as a model. Three weeks after the inoculation of GL261 cells, mice were treated with Q-NPs at 50 mg/kg through intravenous administration. Twenty-four hours later, PTX was given through tail vein injection. After an additional 12 h, the mice were euthanized, the brains were harvested, and tumors in the brain were isolated and subjected to quantification of PTX by high-performance liquid chromatography (HPLC). The results in Figure 4C showed that, while treatment with Q-NPs did not alter the concentration of PTX in the normal brain, the treatment increased the penetration of PTX into tumors by 2.6-fold. We applied the same approach to determine the permeability using Evans Blue and found that the treatment enhanced leakage of the dye preferentially into brain tumors (Figures 4D, 4E, and S18). The results in Figure 4D showed that while treatment with Q-NPs did not alter the concentration of Evans Blue in the normal brain, the treatment increased the penetration of Evans Blue into tumors by 6.6-fold. Collectively, the in vitro and in vivo data suggest that treatment with Q-NPs disrupts the BBB and enhances drug delivery preferentially to tumors in the brain.

Figure 4.

Q-NPs treatment improved drug delivery to brain tumors

(A) Schematic diagram of the in vitro BBB permeability assay.

(B) Quantification of the permeability of TRITC-dextran in the in vitro BBB model with or without treatment of Q-NPs (50 μg/mL) at the indicated time points. Data represent means ± SEMs; n = 3.

(C) Quantification of ID%/g of PTX in normal brain tissues and brain tumors in mice with or without treatment of Q-NPs. Data represent means ± SEMs; n = 3.

(D) Quantification of Evans Blue in normal brain tissues and brain tumors in mice with or without treatment of Q-NPs. Data represent means ± SEMs; n = 3.

(E) Representative images of the brains isolated from mice receiving the indicated treatments followed by Evans Blue injection; n = 3.

Statistical analyses were performed using the unpaired, 2-tailed Student’s t test by Prism 8 (GraphPad). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Characterization of Q-NPs for brain cancer treatment in GL261 syngeneic glioma model

We characterized Q-NPs for brain cancer treatment in mice bearing GL261 gliomas. To enable non-invasively imaging tumor development in the brain, GL261 cells were engineered to express luciferase. One week after tumor inoculation, mice were randomly grouped and treated with PBS, Q-NPs, free PTX, or a mixture of Q-NPs with free PTX (Q-NPs + PTX). Q-NPs and PTX were given at 25 and 2 mg/kg, respectively. Due to its limited solubility, free quercetin at the dose equivalent to Q-NPs was quickly precipitated out in PBS before injection (Figure S7). As injection of such a suspension into mice led to immediate death, free quercetin was not included in both the efficacy and the following toxicity studies. Treatments were performed twice per week for 3 consecutive weeks. The mice were monitored for weight change, survival, and tumor growth. Luciferase imaging showed that treatment with Q-NPs or Q-NPs + PTX effectively inhibited tumor growth (Figures 5A and 5B). We found that treatment with Q-NPs significantly enhanced the survival of tumor-bearing animals by 14 days (p = 0.0025), and the efficacy of Q-NPs treatment was further improved through the co-administration of PTX (Figure 5C; p = 0.0414). By the end of the study, the brains were harvested and subjected to pathological analysis. H&E staining showed that treatment with Q-NPs disrupted the existing tumor vessels, accompanied by extravasation of red blood cells into tumors (Figures 5D and S19A). Immunostaining of blood vessels using anti-CD31 and anti-VEGFR2 antibodies confirmed that treatment with Q-NPs disrupted tumor vessels. Terminal deoxynucleotidyl transferase (TUNEL) staining identified massive cellular apoptosis in tumors isolated from mice treated with Q-NPs, but not PBS. Throughout the study, treatment with Q-NPs did not induce a significant loss of body weight (Figure 5E) nor detectable damage to major organs (Figure S19B), suggesting that Q-NPs are safe for intravenous administration. To further confirm that the NPs are safely used in vivo, we performed an alanine aminotransferase (ALT) assay and an aspartate aminotransferase (AST) assay and found that intravenous administration of Q-NPs did not result in significant hepatotoxicity (Figure S20).

Figure 5.

Characterization of Q-NPs for treatment of GL261 gliomas

(A and B) Representative images (A) and quantification (B) of tumors based on luciferase signal in the brains of mice receiving the indicated treatments. Data represent means ± SEMs; n = 10.

(C) Kaplan-Meier survival analysis of mice receiving the indicated treatments; n = 10.

(D) Representative immunohistochemical and fluorescence images of tumors isolated from mice receiving the indicated treatments. Scale bar: 50 μm.

(E) Changes in body weight with time in mice receiving the indicated treatments. Data represent means ± SEMs; n = 10.

Statistical analyses were performed using the unpaired, 2-tailed Student’s t test and 1-way ANOVA analysis by Prism 8 (GraphPad). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Characterization of Q-NPs for brain cancer treatment in human glioma xenografts

Recent evidence suggests that the histopathology of human malignant gliomas cannot be recapitulated by mouse models derived from existing cell lines. Establishment of a clinically relevant mouse model of human malignant gliomas requires the inoculation of brain cancer stem cells (BCSCs) derived from patient specimens.[43,44] For this purpose, we recently established a panel of human BCSCs and demonstrated that many of them are capable of mimicking human histopathology in mice.[45,46] Using PS30, one of our well-characterized lines, we investigated whether the tumor-targeting and inhibition effects of Q-NPs observed in the GL261 model could be repeated. The experiments were carried out according to the same designs as described above. We found that Q-NPs after intravenous administration efficiently penetrated into tumors but not the normal brain, and the number of Q-NPs accumulated in tumors was 38.5 times greater than that of control PLGA NPs (Figure S21). We found that treatment with Q-NPs enhanced the penetration of PTX selectively to tumors by 3.2-fold but not to the normal brain (Figure 6A). When used as a therapeutic agent, Q-NPs significantly inhibited tumor growth and enhanced the survival of tumor-bearing mice, and the efficacy was enhanced through combination with PTX (Figures 6B–6D). Of note, unlike that for mice bearing GL261 gliomas, the enhancement effect by PTX did not reach statistical significance. This is likely because PS30 as a BCSC line is highly resistant to conventional chemotherapy drugs, such as PTX.[45,46] More aggressive treatment regimens, such as the use of a higher dose and/or an increased treatment frequency, are needed for this model. Pathologically, treatment with Q-NPs disrupted tumor vasculature, leading to significant vessel loss and cellular apoptosis (Figures 6E and S22). Throughout the study, no significant body weight loss or damage to normal organs was found (Figure S23). All of the findings were consistent with those observed in the GL261 model, suggesting that the effects of Q-NPs for brain cancer drug delivery and treatment are not unique to the tested tumor models.

Figure 6.

Characterization of Q-NPs for brain cancer treatment in PS30 mouse xenografts

(A) Quantification of ID%/g of PTX in normal brain tissues and brain tumors in mice treated with PBS or Q-NPs. Data represent means ± SEMs; n = 3.

(B and C) Representative IVIS images (B) and quantification (C) of tumors in the brains of mice receiving the indicated treatments. Data represent means ± SEMs; n = 7.

(D) Kaplan-Meier survival analysis of tumor-bearing mice receiving the indicated treatments; n = 7.

(E) Representative immunohistochemical and fluorescence images of tumors isolated from mice receiving the indicated treatments. Scale bar: 50 μm.

Statistical analyses were performed using the unpaired, 2-tailed Student’s t test and 1-way ANOVA analysis by Prism 8 (GraphPad). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

DISCUSSION

Malignant glioma is a devastating disease without an effective treatment. While being effective for various solid tumors, the traditional anti-angiogenic therapy fails to extend the survival of patents with malignant glioma, suggesting that inhibition of the formation of new vessels alone is insufficient for this disease.[6-8] In this study, we screened a small collection of anti-angiogenic polyphenols that were formulated into NPs and identified Q-NPs, which exhibited significant anti-angiogenic activity and brain tumor-targeting efficiency (Figure 1). Unlike bevacizumab, a classical anti-angiogenic agent that inhibits new vessel formation,[8] Q-NPs have dual functions to not only limit the growth of new vessels but also induce targeted disruption of the existing tumor vasculature (Figures 3A–3E). Mechanistically, we showed that the tremendous vessel-targeting and anti-angiogenesis effects of Q-NPs are mediated by VEGFR2, which is preferentially expressed in brain tumor vessels (Figures 3F–3H). We found that Q-NPs induced NanoEL by selectively damaging the existing tumor vasculature through disruption of the homophilic interaction of VE-cadherin (Figures 3I and 3J). This disruption, in turn, leads to an increase in vascular endothelium permeability and enhanced extravasation of circulating drugs selectively into brain tumors (Figure 4). Through interaction with VEGFR2, Q-NPs identify tumor vasculature with a high degree of specificity, which is not often seen in traditional angiogenesis therapies. Despite the great benefit of enhancing drug delivery preferentially to brain tumors, Q-NP-induced NanoEL is not without potential limitations. A recent study in mouse breast cancer models showed that endothelial leakiness induced by TiO2 NPs may promote cancer cell extravasation, leading to increased tumor load and potential metastases.[47] Therefore, further characterization of the long-term consequences of NanoEL induced by Q-NPs in brain cancer models is warranted.

This study is also significant in that it develops an approach to formulate polyphenols into NPs. Clinical translation of polyphenols has been limited by their low bioavailability and systemic toxicity.[13,48] To address this translation hurdle, formulating polyphenols into NPs using existing nanomaterials, such as polymers or liposomes, have been explored.[49,50] Unfortunately, accumulating evidence suggests that this approach typically cannot encapsulate and deliver payloads by >10% by weight,[51] and, without further engineering, does not allow for disease targeting.[52,53] Both limitations can be overcome through the combinatory Fe-coordination and polymer-stabilization approach developed in this study. For instance, we showed that Q-NPs carry payload quercetin by up to 73% by weight (Figure S3), and without further surface modification recognize tumor vasculature with high efficiency and specificity (Figure 3).

In summary, we reported the development of polyphenol NPs, which have dual effects on the inhibition of the growth of new vessels and targeted disruption of the existing tumor vasculature. Due to their marked effects on reducing tumor development and enhancing drug delivery, the NPs have the potential to be used as promising anti-angiogenic agents for the treatment of brain cancer. In addition, we developed an approach to formulate polyphenols into NPs, which may have a broad application in improving the delivery of bioactive polyphenols and facilitating the translation of polyphenol drugs into clinical applications.

EXPERIMENTAL PROCEDURES

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jiangbing Zhou (jiangbing.zhou@yale.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

All data supporting the findings of this study are available within the article and are available from the lead contact upon request.

Synthesis of NPs

Q-NPs were synthesized at a mass ratio of F127:Fe:quercetin of 106:20:x (19.88–100). In a typical synthesis, 106 mg F127 was dissolved in 8.8 mL water at room temperature with stirring. Then, 0.2 mL FeCl3 aqueous solution at a concentration of 100 mg/mL was added to the above F127 solution. After 1 hour, various volumes of quercetin DMSO solution (from 0.994 to 5 mL at a concentration of 20 mg/mL) were added to the above mixture. After stirring overnight, the solutions were collected and dialyzed against deionized water using SnakeSkin Dialysis Tubing (10K MWCO, Thermo Fisher Scientific). The same procedures were used to synthesize other NPs, except that the mass of the polyphenol was changed accordingly.

Characterization of NPs

TEM micrographs were taken on a TECNAI G2 Spirit TWIN TEM (FEI). X-ray diffraction was performed on a Rigaku SmartLab X-ray Diffractometer using Cu Kα (0.15406, nm) radiation. DLS and zeta potential measurements were carried out on a Zetasizer Nano ZS (Malvern Instruments). Elemental analysis was performed on a PerkinElmer ICP-MS Elan DRC-e. Infrared spectra were captured with an FTIR/Raman Thermo Nicolet 6700 spectrometer (Thermo Fisher Scientific). EDS analysis was performed on a Hitachi SU8230 UHR cold field emission (CFE) scanning electron microscope equipped with a Bruker XFlash 5060FQ Annular EDS detector.

Tube formation assay

In a typical study, 200 μL Matrigel solution (Corning) was added to a 24-well culture plate and allowed to solidify for 45 min at 37°C. In the meantime, HUVECs were treated with selected NPs at various concentrations. Thirty minutes later, the treated cells were collected and seeded on Matrigel. After incubation for 24 h, the formation of tubes was imaged using an AMG EVOS fI inverted microscope. The length and width of the tube-like structures were calculated using NIH ImageJ software. The inhibition results are presented as the mean between five replicates.

In vitro BBB permeability assay

An in vitro BBB model was established according to our recent report.[41] Briefly, Transwell cell culture 8 μm inserts (Millipore) were coated with 5 μg/cm2 fibronectin in 150 μL PBS (1.68 μL/mL stock solution in 150 μL PBS per insert). HBMECs (33,000 cells/well) were seeded with 1% fetal bovine serum (FBS) containing EBM-2 in the upper chamber of the Transwell culture insert. In the lower chamber, NHA cells were seeded with 45,000 cells/well. After TEER values reached ~100 Ω (usually 2–3 days after adding HBMECs), NPs were added into the top chamber. TRITC-dextran (10 μL, 1 mg/mL, 4.4 kDa; Sigma-Aldrich) was added to the top chamber after 1 h. As various time points, including 1, 2, 3, 4, 6, and 24 h after treatment, medium in the bottom chamber was sampled and subjected to measurement for the amount of TRITC-dextran using a BioTek Synergy 2 Multi-Detection Microplate Reader with an excitation/emission wavelength of 550/575 nm.

VE-cadherin immunostaining

For the immunostaining study, bEnd.3 cells were treated with Q-NPs at various concentrations. After incubation for 24 h, the treated cells were collected and fixed by 4% paraformaldehyde (Sigma) for 20 min at room temperature. Primary antibodies (recombinant anti-VE-cadherin antibody [EPR18229] [ab205336], Abcam) were used at 1:1,000 and incubated overnight at 4°C. Secondary antibodies (chicken anti-rabbit immunoglobulin G [IgG] [H + L] cross-adsorbed secondary antibody, Alexa Fluor 488, A-21441, Thermo Fisher Scientific) were used at 1:1,000 and incubated for 1 h at room temperature. For the in vivo immunostaining study, at day 21 after tumor inoculation, Q-NPs were administrated through the tail vein. Twenty-four hours later, the mice were euthanized. The brain was isolated, sliced, and subjected to VE-cadherin staining according to the procedures described above.

Tumor models

Mice were purchased from Charles River Laboratories. For establishment of GL261 gliomas models, 5- to 6-week-old female C57BL/6 mice were used. For the establishment of PS30 models, nude mice (athymic NCr-nu/nu, 6 weeks old) were used. Mouse models were established according to our previously reported procedures.[22,54]

Characterization of drug delivery to the brain

Characterizations of drug delivery, including comparing gallic acid-, quercetin-, and ellagic acid- NPs, comparing Q-NPs with PLGA NPs and dendrimer NPs, were performed in mice 21 days after the inoculation of GL261 cells. NPs in each comparison group were normalized to ensure that each mouse received the same amount of dye or europium. Twelve hours after treatment, the accumulation of NPs in the brain and major organs was determined based on fluorescence, which was imaged using an animal optical imaging system (IVIS Lumina III, Caliper Life Sciences), or based on the amount of europium, which was quantified by ICP-MS. For the in vivo competition study, the same experimental design was used, except that each mouse in the competition was pre-treated with 200 μg anti-mouse VEGFR2 antibody before Q-NPs administration.

Microscopic analysis of NPs in the brain

Tumor-bearing mice were established through the intracranial injection of GL261 cells that were engineered to express GFP. After 21 days, the mice were treated with Cy5-labeled Q-NPs. At 12- and 24-h time points, a cohort of mice was intravenously administered 100 μL PE rat anti-mouse CD31 antibody (BD PharMingen #553373). One hour later, the mice were euthanized. The brains were isolated, sliced, subjected to DAPI staining, and imaged by confocal microscopy.

Characterization of Q-NPs for enhancing drug delivery

At day 21 after tumor inoculation, Q-NPs were administrated through the tail vein. Twenty-four hours later, PTX or Evans Blue was injected intravenously. After an additional 12 h, the mice were euthanized. The tumors and major organs were harvested, homogenized in 1 mL methanol, and centrifuged at 13,000 rpm for 30 min. The supernatant was collected and subjected to quantification of PTX by HPLC (Schimadzu Prominence LC-20AD). A reverse-phase C18 column (250 × 4.6 mm, 5.0 μm, XBridge) was used in this experiment. The mobile phase and the flow rate of the mobile phase were acetonitrile/water = 45/55 (v/v) and 1 mL/min. The injection volume was 30 μL. The column effluent was detected using a UV detector at λmax of 227 nm. The measurements were carried out in triplicate. The percentage of the injected dose per gram of tissue (%ID g−1 tissue) was calculated. For Evans Blue study, the tumors and brains were harvested and incubated in 1 mL 2,2-N-methylformamide overnight at 37°C with mild shaking. After centrifugation, the supernatant was collected and subjected to quantification of Evans Blue by measuring the optical density (OD)600.

Characterization of therapeutic evaluation and toxicity of Q-NP treatment

To assess the potential therapeutic effect of Q-NPs, tumor-bearing mice were prepared through intracranial injection of GL261 cells or PS30 cells. One week later, the mice were randomly grouped and treated with PBS, Q-NPs, free PTX, or a mixture of Q-NPs with free PTX (Q-NPs + PTX). Q-NPs and PTX were given at 25 mg/kg and 2 mg/kg, respectively. Treatments were performed twice per week for 3 consecutive weeks. Tumor size was monitored once per week for 3 weeks using IVIS. The weight, grooming, and general health of the animals were monitored on a daily basis. Mice were euthanized after either a 15% loss in body weight or when it was humanely necessary due to neurological symptoms. The brains were harvested and fixed for immunohistochemistry. To detect VEGFR2 expression in tumors, sectioned brains were stained with anti-VEGFR2 and anti-CD31 antibody. To analyze the therapeutic effects of NPs, slides of brain sections were prepared and stained with TUNEL and HE. To analyze the biodistribution of NPs, slides of brain sections were stained with HE and a Fe stain kit. For characterization of toxicity, healthy mice were used and received treatment of Q-NPs at a dose of 100 mg/kg. Control mice were treated with saline. After 10 days, blood was taken from the tail vein. After centrifugation at 3,000 rpm for 10 min, the serum was collected and analyzed using AST and ALT assay kits (MAK055, MAK052, respectively [Sigma-Aldrich]). One month later, the mice were euthanized. Major organs, including the heart, liver, spleen, kidney, and brain, were harvested and fixed for immunohistochemistry.

Statistical analysis

In vitro experiments were performed at least in triplicate. Data are presented as the means ± standard deviations (SDs). Statistical analyses were performed using the unpaired, two-tailed Student’s t test and one-way ANOVA analysis by Prism 8 (GraphPad) (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Study approval

The study is compliant with relevant ethical regulations regarding animal research and was approved by the Yale Institutional Animal Care & Use Committee (IACUC).