Gold Nanoparticle–Mediated Targeted Delivery of Recombinant Human Endostatin Normalizes Tumour Vasculature and Improves Cancer Therapy

Tumour vasculature is generally disordered because of the production of excessive angiogenic factors by tumour cells, which results in tumour progression and reduces the effectiveness of radiotherapy or chemotherapy. Transient anti-angiogenic therapies that regulate tumour vascular morphology and function and improve the efficiency of antitumour therapy are under investigation. Recombinant human endostatin (Endostar/rhES) is a vascular angiogenesis–disrupting agent that has been used to treat non-small cell lung cancer (NSCLC) in the clinical setting. In this study, we used gold nanoparticles (AuNPs) as a drug-delivery system (DDS) for targeted tumour delivery of rhES for short therapy, which resulted in transient tumour vascular normalization, reduced permeability and hypoxia, strengthened blood vessel integrity, and increased blood-flow perfusion. Moreover, combination therapy with 5-FU over this timeframe was substantially more effective than 5-FU monotherapy. In conclusion, our research demonstrates the potential use of AuNPs as a drug-delivery platform for transporting rhES into a tumour to induce transient tumour vascular normalization and enhance the antitumour efficacy of cytotoxic drugs.

Endostatin, a 20-kDa C-terminal non-collagenous fragment of type XVIII collagen, inhibits angiogenesis in different pathological conditions 11 . Moreover, endostatin specifically binds to nucleolin on the cell surface to inhibit tumour lymphangiogenesis and lymphatic metastasis 12 by down-regulating VEGF-C levels and VEGFR-3 gene expression 13 . Recombinant human endostatin (Endostar or rhES) is a modified human endostatin in which a nine-amino acid sequence (MGGSHHHHH) added to the N-terminus of endostatin confers a more stable structure while retaining endogenous endostatin function 14 .
To improve in vivo efficiency and reduce dosing frequency, a variety of drug-delivery systems (DDSs) have been developed. Gold nanoparticles (AuNPs), as a nanoparticle DDS, exhibit unique properties, including chemical stability, inherent biocompatibility, low cytotoxicity 15 , and the capability of delivering various biological molecules to their targets, including small drug molecules, peptides, proteins, and nucleic acids such as DNA or RNA 16,17 . AuNPs can become trapped in solid tumours and be retained at high concentrations for prolonged periods 18 due to the enhanced permeation and retention (EPR) effect, which is likely associated with the heterogeneity of tumour blood vessels 19 . The width between the endothelial cells in tumour vessels is ranging from 200 nm to 1.2 μ m 20 ; therefore, nanoscale drugs can extravasate and diffuse from tumour vessels but not normal vessels due to the leaky nature of tumour vessels 20,21 . Furthermore, vessels with excessive leakiness increase interstitial fluid pressure inside tumours and cause interstitial fluid to ooze from the tumour into the surrounding tissues, resulting in decreased drug concentrations and the spread of tumour cells 22 .
In this study, we used AuNPs as a drug carrier to deliver rhES into a tumour via passive targeting and facilitate transient vascular normalization to improve antitumour therapeutic efficacy.

Results
Characterization of rhES-AuNPs-PEG. The UV-visible spectrum of AuNPs revealed a characteristic surface plasmon resonance (SPR) absorption band at approximately 517 nm; after combination with rhES, the compound exhibited a peak at 550 nm, which was attributed to the introduction of groups on the surface of the AuNPs (Fig. 1a). Dynamic light scattering (DLS) indicated that the hydrodynamic diameters of AuNPs, PEG-, and rhES-PEG-functionalized AuNPs were 58.2 ± 7.1 nm, 70.6 ± 6.8 and 337.9 ± 13.0 nm, respectively (Fig. 1b), and the zeta potentials were −39.3 ± 1.1 mV, − 16.5 ± 0.5 mV and − 13.7 ± 1.0 mV, respectively. Field emission transmission electron microscopy (FETEM) images revealed that the AuNPs were well-dispersed spherical particles that were uniform in size (Fig. 1c). The rhES-AuNPs-PEG conjugates exhibited a slightly aggregated structure (Fig. 1d), consistent with the bathochromic shift in the UV-visible spectrum. Loading efficacy and release rate. The free rhES in the supernatant was measured as described previously, and the loading of rhES on AuNPs was determined to be 86%. The release of rhES at pH 6.8 and 7.4 was slow and sustained; however, the release ratio in acidic media (pH 5.0 and 6.0) continuously increased with time, and the release capability was increased gradually with the reduction of the pH value (pH 6.6 and 6.2). Thus, low pH resulted in more rapid release of rhES compared with release at pH 6.8 (P = 0.0091) and 7.4 (P = 0.0040) (Fig. 2a). The linear relationship between pH and the release ratio is presented in Fig. 2b; over different time periods, the ratio of rhES released from AuNPs was strongly dependent on pH. This pH-dependent release might enhance the selectivity of tumour targeting because the acidic conditions of the tumour microenvironment may trigger the release of rhES from the AuNPs. Furthermore, the lower release rate at pH 7.4 would protect the stability of the carrier delivery system in the circulation and normal tissues, as observed in previous research 23 .

Accumulation of rhES-AuNPs-PEG in tumours.
After intravenous injection, rhES-AuNPs-PEG accumulated rapidly and specifically in tumours at a higher level than rhES alone at 30 min (P = 0.0141) and 4 h (P = 0.0332) (Fig. 3a,b). The difference in localization demonstrates that rhES-AuNPs-PEG exhibited improved tumour-targeting efficiencies and longer retention. However, rhES localization to the lung, heart, and kidney was not significantly different. A significant increase in rhES was observed in the liver (within 30 min) and spleen, which may be due to phagocytosis by the reticuloendothelial system (RES).
In addition, optical in vivo imaging of Cy5-labeled regents revealed that after the i.v. injection, rhES-AuNPs-PEG rapidly accumulated in tumours as early as 30 min and increased in tumours in 4 h (Fig. S1). By contrast, rhES accumulated to lower levels in tumours than rhES-AuNPs-PEG, and we observed poor specific localization in vivo. This finding confirms that AuNPs are excellent nanoparticle drug carriers that increase tumour targeting.
Short-term rhES-AuNPs-PEG treatment facilitates transient vascular normalization. We next evaluated vascular normalization by measuring CD31 as a marker of the endothelium, α -smooth muscle actin (α -SMA) as a marker of pericytes, and FITC-labelled dextran as a marker of permeability. CD31 staining revealed that the vessels in the control treatment were chaotic and deformed (Fig. 4a). By contrast, after treatment with rhES-AuNPs-PEG for 4 days or 8 days, the tumour vessels appeared organized. The tumour vessels exhibited increased vessel diameter and improved structure, but similar effects were not observed after 12 days. Moreover, a significant difference in vascular density was not observed between day 0 and day 12. Our data indicate that rhES-AuNPs-PEG improved vascular morphology in tumours in a highly time-dependent manner.
The control tumours exhibited increased vascular permeability, which is associated with excessive deposition of dextran around the tumour vessels, as well as irregular CD31 staining. However, after treatment with rhES-AuNPs-PEG for 4 days and 8 days, leakage of dextran from vessels decreased (day 4, P = 0.0009; day 8, P = 0.0011; day 12, P = 0.2206; Fig. 4b). Additionally, double staining for CD31 and α -SMA demonstrated that the treatment increased the proportion of pericyte-covered vessels (day 4, P = 0.0078; day 8, P = 0.0082; day 12, P = 0.0815; Fig. 4c). These results indicate that rhES-AuNPs-PEG increased the integrity of the endothelium by increasing pericyte coverage, indicating that the tumour vessels were more mature.

rhES-AuNPs-PEG increase blood perfusion and reduce hypoxia in tumour sections.
Because rhES-AuNPs-PEG normalized the morphology of blood vessels, we next evaluated whether treatment could improve the function of the vessels. Vessel perfusion improved in tumours after 4 days of treatment, and perfusion gradually increased until day 8. However, an obvious decrease in perfusion was observed on day 12 (day 4, P = 0.0271; day 8, P = 0.0028; day 12, P = 0.7487; Fig. 5a,b). By contrast, vessels in the control tumours exhibited less perfusion during the entire course of treatment.
Because hypoxia triggers angiogenesis, we analysed tumour oxygenation by staining for the hypoxia marker hypoxia inducible factor-1α (HIF-1α ), which enables tumour cells to produce a variety of pro-angiogenic The rhES release profile was two-phase, with an initial quick release followed by continuous release. Release was higher and more rapid at low pH (pH 5.0 and 6.0) than at pH 6.8 and 7.4, indicating that the rhES release ratio from AuNPs was pH dependent. (b) The linear relationship between pH and release ratio. Different time points and pH values altered the release ratio; overall, low pH values promoted the release of the drug. factors 24 . Treatment with rhES-AuNPs-PEG relieved hypoxia in tumours (day 4, P = 0.0003; day 8, P < 0.0001; day 12, P = 0.4408; Fig. 5c,d), consistent with the increase in blood flow during the normalization window of the tumour vessels. Histological analysis revealed that the tumours in mice treated with rhES-AuNPs-PEG exhibited smaller necrotic areas in the tumoural regions (day 4, P = 0.0138; day 8, P = 0.0018; Fig. 5e,f). Together, these data indicate that treatment with rhES-AuNPs-PEG temporarily increases blood perfusion and alleviates hypoxia and necrosis in tumours. Pre-treatment with rhES-AuNPs-PEG increases intratumour drug delivery. To determine if vessel normalization increases chemotherapeutic efficacy, we administered 5-FU within and outside this normalization window. Compared with monotherapy, combination with rhES-AuNPs-PEG significantly increased 5-FU localization to the tumour sections (P = 0.0428; Fig. 6a) but had no effect on the kidneys (P = 0.8948; Fig. 6b) or liver (P = 0.1166; Fig. 6c), indicating that the normalization was tumour-specific.
Treatment with rhES-AuNPs-PEG improves cytotoxic therapeutic outcomes. Treatment with 5-FU monotherapy and 5-FU combined with rhES-AuNPs-PEG reduced growth significantly (Fig. 7a). However, tumours receiving combination therapy exhibited a more significant reduction in growth on day 12 compared to   plus 5-FU was 1.60-fold longer than that of the placebo group (Fig. 7b). These data suggest that tumour vascular normalization induced by short-term rhES-AuNPs-PEG treatment enhanced the efficacy of 5-FU.

Discussion
Solid tumours are always associated with pathological angiogenesis, which is initiated by pro-angiogenic factors that are produced by tumour cells under hypoxic conditions. These newly formed vessels are structurally abnormal and exhibit dysfunctions that impair perfusion and oxygenation. Moreover, the changes in the tumour microenvironment promote tumour invasion and metastasis 25 , and hypoperfusion limits the delivery of therapeutic agents into the tumour, which results in insensitivity to chemo/radiotherapy 26 . Most studies have only concentrated on anti-angiogenesis to "starve" tumours using high-dose angiogenesis inhibitors; however, emerging evidence has revealed a contradiction between anti-angiogenesis and traditional chemotherapy. Lower doses of anti-angiogenesis reagents can induce tumour vascular normalization, which results in temporarily increased blood flow and improves the effectiveness of subsequent chemo/radiotherapy. In fact, a variety of evidence supports the existence of a normalization window.
Endostatin is a broad-spectrum endogenous angiogenesis inhibitor 27,28 that suppresses angiogenesis via many pathways, such as bFGF/FGF-2, α 5β 1, E-selectin, metalloproteinases, and VEGF 27 . Furthermore, the recombinant form (Endostar/rhES) has been used in the clinical setting for the treatment of lung cancer 29 . In this study, we used AuNPs as a tool to specifically target tumour vessels for short therapy to induce tumour vascular normalization. AuNPs are excellent nano-drug carriers because of their large surface-area-to-volume ratio and biocompatibility and have been used successfully to deliver large biomacromolecules without compromising their activity 30 . Moreover, the grafting of methoxy polyethylene glycol-thiol (mPEG-Thiol) onto AuNPs after being combined with rhES to maintain colloidal stability 31 and prevent nonspecific serum protein binding to the surface 32,33 reduces clearance by macrophages of the RES 33 , consequently prolonging their circulation time in vivo and improving tumour uptake. Our data indicated that rhES was successfully directly bound to AuNPs (presumably via a noncovalent linkage), thus targeting delivery into tumours more efficiently. As previously mentioned, extravasation of nanoparticles, as a passive targeting method, is dependent on heterogeneous vessels 18 . The integrity of the vascular endothelium prevents macromolecules and nanoparticles from permeating into most normal tissues, and this selective extravasation effect favours a long circulation time and less toxicity.
Pericytes have been implicated as mediators of tumour angiogenesis and metastasis 34 . Pericytes normally envelope endothelial cells and coordinate intercellular signalling to establish direct cell-cell contact and support endothelial integrity, stabilization, and maturation 35,36 . The dissociation of pericytes and endothelial cells is the initial stage of tumour angiogenesis and facilitates endothelial cell migration and endothelial tubulogenesis 34,35 . Meanwhile, the detachment-mediated increase in vessel permeability leads to the leakage of plasma proteins 34 and cancer cell metastasis into the circulation 36 . Several molecules are involved in pericyte-mediated interaction signalling, including PDGFRβ s, VEGFRs, and Tie-2 [36][37][38] . Moreover, disruption of pericytes and endothelial cells activates pericytes to secrete more VEGF and bFGF, establishing a vicious cycle of continuous angiogenesis while promoting endothelial cell proliferation and migration into the surrounding tissue 34 . In this study, our observations suggested that treatment with rhES-AuNPs-PEG improved pericyte coverage, indicating that the tumour blood vessels were more mature.
Hypoxia is an indicative feature of solid tumours and a potent inducer of tumour angiogenesis 39,40 . HIF-1α is the primary mediator of the hypoxia response and induces overexpression of pro-angiogenic factors produced by tumour and host cells under hypoxic conditions 39 ; for example, VEGF is primarily regulated by HIF-1α at the transcriptional level 41 and causes tumour vascular angiogenesis. By contrast, increasing vascular perfusion is associated with reduced tumour hypoxia, and our data corroborate these findings. After treatment with rhES-AuNPs-PEG, HIF-1α expression was significantly reduced in the tumours. Our result is consistent with published studies demonstrating that endostatin down-regulates the expression of HIF-1α in the microvascular endothelium 42,43 . However, some data suggest that the anti-angiogenic activity of endostatin is independent of the HIF-1/VEGF pathway in vitro 44 .
Next, we examined the concentration of 5-FU in tumours by HPLC analysis. Treatment with rhES-AuNPs-PEG increased the concentration of intratumour 5-FU, which is closely related to increased blood flow in tumours, and this improvement was obvious after 4 and 8 days of treatment with rhES-AuNPs-PEG. Moreover, 5-FU combined with tumour vascular normalization therapy reduced disease progression.
In summary, we prepared a tumour-specific targeting drug-loaded nanosystem by surface-modifying rhES using AuNPs. Our data indicate that AuNPs enhanced the concentration of rhES in tumours. Furthermore, rhES promoted transient vascular normalization in H22 xenograft models, reduced hypoxia and leaking, and improved pericyte coverage and blood perfusion, thereby increasing 5-FU delivery into the tumour and improving therapeutic efficiency. However, further research is needed to fully elucidate the mechanism of tumour vascular normalization. Preparation and characterization of the rhES-AuNPs-PEG. The rhES-AuNPs-PEG was prepared via a self-assembly process. A calculated amount of rhES was added to the dispersion of AuNPs, and the reaction system was incubated in the dark at 4 °C for 12 hours. The preparation was oscillated gently every 2 hours. Subsequently, mPEG-Thiol was introduced and incubated for another 4 h for in vivo treatment. SPR in the UV-visible spectrum was observed on a model V-570 Jasco dual-beam spectrophotometer. The hydrodynamic size distribution and zeta potential were evaluated by DLS using a Malvern Zetasizer Nano ZS (Malvern, UK). The morphology and surface structure were visualized using field emission transmission electron microscopy (FETEM) (JEM2100, JEOL, Japan).

Reagents and cells. rhES was provided by
rhES-loading efficiency. The rhES-load efficiency was determined according to the previously described method 45 . The mixture was centrifuged at 13000 rpm for 20 min. Subsequently, the concentration of free rhES in the supernatant was determined using an ELISA method. Finally, the percentage loading of rhES on AuNPs was estimated using the following formula: rhES release in vitro. rhES release behaviour from the carrier was estimated using a centrifugal filter device (Millipore Amicon ® Ultra, 30-kD molecular weight cutoff). rhES-AuNPs-PEG and 10-mL PBS at different pH values were added to the filter device and centrifuge tube, respectively. The system was incubated at 37 °C for a predetermined length of time, and the filter device was centrifuged at 3000 rpm for 20 min. Next, 1 mL of PBS was withdrawn and replaced by an equal volume of fresh PBS. The rhES concentration was determined as described previously.
Tumour xenograft models and treatment regimens. Female Kunming mice (20-40 g weight, 6-8 weeks old) were purchased from the Medical Laboratory Animal Center (Guangdong, China) and maintained under specific pathogen-free conditions. All animal study protocols were approved and conducted in accordance with the guidelines of the Laboratory Animal Ethics Committee of Jinan University. Xenografted tumours were initiated by injection of 5 × 10 6 H22 cells into the subcutaneous tissue of the right flank. The tumour volume was measured using a Vernier calliper every other day and calculated as V = (length × width 2 ) × 0.523. Treatment with rhES-AuNPs-PEG or placebo was initiated when the tumour volume reached 175-200 mm 3 post-inoculation and was repeated daily from day 0 to day 12. From day 1, all animals bearing tumours underwent treatment with 5-FU (40 mg/kg) daily for 4 days, followed by treatment every 2 days for 1 week. Subsequently, the mice were euthanized, and the tumours and various organs were harvested for analysis.
In vivo tumour and tissue biodistribution. Mice bearing H22 tumours received rhES (10 mg/kg) or rhES-AuNPs-PEG containing the same dose of conjugated rhES for predetermined times (30 min and 4 h). The animals were sacrificed, and tissue samples from the liver, spleen, heart, lung, kidney, and tumours were collected and rapidly frozen for immunofluorescence detection using an anti-endostatin antibody (1:500; Abcam). In addition, we used optical in vivo imaging systems (Bruker, Germany) to detect the rhES-AuNPs-PEG biodistribution. The animals were anaesthetized with isoflurane and imaged 10 min, 30 min and 4 h after the injection of Cy5-labelled regents. The acquisition time was approximately 10 min, and image reconstruction was performed using InVivoScope 1.37 software (Bioscan).

Vascular leakage and perfusion. Extravascular diffusion visualization was assessed by intravenous injec-
tion of 0.25 mg/mouse FITC-labelled dextran (40 -kDa; Chondrex, USA). Tumours were harvested after 20 minutes, frozen in optimal cutting temperature compound (Sakura Finetek, Torrance, CA) and stored at − 80 °C. Cryosections with a thickness of 20 μ m were fixed in cold acetone and rehydrated in PBS. Microvessels were stained with an anti-CD31 antibody (1:500; Abcam), and the percentage of leakage was calculated as the ratio of the dextran + area to CD31 + area using Image J software.
Tumour blood perfusion was detected using FITC-conjugated lectin from Bandeiraea simplicifolia (Sigma-Aldrich, USA) and anti-CD31 staining. Subsequently, 10 mg/kg FITC-lectin was injected intravenously. After 1 minute, the tumours were excised, and sections were prepared as described previously. Tissue was visualized under a fluorescence microscope (Leica DM6000B), and the lectin + area was presented as a percentage of the CD31 + area.
Immunofluorescence. Tissues were fixed in 4% paraformaldehyde for 24 hours and paraffin-embedded, sectioned, dewaxed in xylene, and rehydrated through graded alcohols. Antigen retrieval was performed in citric acid buffer (pH 6.0). Sections were blocked in 2% normal goat serum for 1 hour and stained with primary antibodies as follows: anti-CD31 antibody (1:500; Abcam) for endothelium, α -SMA antibody (1:100; Proteintech) for pericytes, and HIF-1α antibody (1:50; Proteintech) for hypoxia. The sections were then washed and incubated with rhodamine-conjugated goat anti-rat IgG (H + L) (1:50; Proteintech) or goat anti-rabbit IgG-FITC (1:200; Santa Cruz) for 40 min at RT. The level of pericyte coverage was presented as a percent of the length along CD31 + vessels. The ratio of hypoxia in the tumour sections was presented as HIF-1α staining area/CD31 + area. HPLC quantification. The concentration of 5-FU in different organs was quantified using an Agilent 1100 series HPLC (Agilent ® Technologies, Santa Clara, CA, USA) on a COSMOSIL C18 column (250 mm × 4.5 mm, Scientific RepoRts | 6:30619 | DOI: 10.1038/srep30619 5 μ m) (Shimadzu, Tokyo, Japan). The eluents were methanol, 0.1% trifluoroacetic acid and deionized water (3:40:57, v/v/v), and the flow rate was 1.0 mL/min. 5-FU was detected at a wavelength of 265 nm. All responses obtained were analysed using Agilent ® ChemStation ® software. Statistical analysis. All data are presented as the mean ± standard error of the mean of independent triplicate samples. Statistics for different groups were compared by one-way analysis of variance or unpaired Student's t-test. Survival time was determined with the Kaplan-Meier test. All statistical analyses were performed using GraphPad Prism (version 5.0; GraphPad Software, La Jolla, CA). P values less than 0.05 were considered statistically significant (* P < 0.05, * * P < 0.01, and * * * P < 0.001).