Targeting transferrin receptor delivery of temozolomide for a potential glioma stem cell-mediated therapy

Glioma stem cells, which are sub-populations of tumor cells, are responsible for resistant responses to radiotherapy and chemotherapy after surgery. Targeting resistant glioma stem cell sub-populations might present a novel means to prevent tumor recurrence. Due to the high expression of transferrin receptors at the surface of brain capillary endothelial and tumor cells, especially glioma stem cells, targeting the transferrin receptor system provides an avenue for the entry of drug molecules into the brain. Nanoparticles that target glioma stem cell sub-populations, conjugate transferrin and encapsulate temozolomide, were developed as a potential therapeutic strategy to evaluate their effectiveness at damaging tumor cells. Nanoparticles were highly effective at penetrating the blood-brain barrier and delivering a high therapeutic dose of temozolomide. This effective means of delivery provoked enhanced cytotoxicity against glioma cells, and especially against glioma stem cells. The targeting transferrin receptor nanoparticles display an inherent capacity for a highly therapeutic approach in targeting glioma stem cells and non-stem cells tumors. In addition, transferrin nanoparticles encapsulating temozolomide have the potential of a promising anti-tumor strategy against glioma of the O6-methylguanine-DNA-methyltransferase gene promoter methylation.


INTRODUCTION
Glioma stem cells (GSCs) are a sub-population of stem cells that remain non-proliferative for extended periods that have the capacity to re-enter the cell cycle to reestablish a viable tumor under select microenvironmental conditions [1]. This sub-population of tumor cells is responsible for resistance to radio-and chemotherapy following surgery. Promising data has revealed that targeting resistant GSCs may present a novel approach at blocking tumor recurrence [2,3].
The grim prognosis of glioblastoma multiforme (GBM) is due in part to structural barriers including the blood-brain barrier (BBB), which prohibits entry of chemotherapeutic agents. Experimental methods aimed at achieving highly effective chemotherapeutic penetration to the site of the tumor, have been a major focus recently, and shown promise in the treatment of malignant diseases of the brain [4]. The BBB comprises brain capillary endothelial cells (BCECs) that predominantly restrict paracellular substrate flux and free exchange of molecules larger than 400 Da. In addition, due to the low permeability of the BBB and expression of transferrin (Tf) receptors (TfR) on the surface of BCECs, targeting the TfR system provides a route that allows the entry of drugs and nanoparticles to the brain [5][6]. Multiple experimental studies have assayed the expression of TfR1 and TfR2, and found that both were increased on both proliferating and malignant cells, including GBM, as compared to normal brain tissue. Clinical studies have revealed that TfR1 expression was correlated with poorer outcomes [7][8]. The expression of TfR was also increased in cancer stem cells (CSCs) as compared non-CSCs. The surface expression levels of TfR were noticeably elevated in CSCs, likely because of increased recycling or enhanced expression of TfR. In addition, TfR was essential in the maintenance of CSC and increased the frequency of CSCs by nearly 10-fold [9]. Thus, targeting of transferrin receptor therapeutics for GSCs has shown promising potential as a novel therapeutic to target GBM.
GBM with the O6-methylguanine-DNAmethyltransferase (MGMT) gene promoter methylation status is sensitive to temozolomide (TMZ) chemotherapy; however, recurrence is inevitable, and is perhaps due to the existence of GSCs, which when present display only partial uptake of TMZ. Clinical samples and GSC cell-lines expressing the methylated MGMT gene promoter were used to evaluate the effects of Tftargeted nanoparticles pre-loaded with TMZ with the aim of damaging GSCs and inhibiting regrowth of GBM orthotopic xenografts.

Identification of PAMAM-PEG-Tf bioconjugates
Bioconjugated products were purified by gel filtration to separate Tf and PAMAM-PEG-Tf, which were indicated by two peaks in the graphical spectra ( Figure 1A). PAMAM-PEG-Tf was obtained by SDS-PAGE, which showed a new band at about 110 kDa, which implied a Tf ligand that was covalently attached to PAMAM-PEG. No bands were seen at approximately 80 KDa in the PAMAM-PEG-Tf lane, which strongly supported the success of the purification procedure ( Figure 1B). Conjugation of Tf and PAMAM was aided by ultraviolet-visible spectra of all bands scanned. An apparent peak at 280nm from Tf was observed when detecting the biconjugated PAMAM-PEG-Tf, which indicated conjugation of PAMAM and Tf by PEG since no peak was seen in PAMAM ( Figure 1C). The results of SDS-PAGE and ultraviolet-visible spectra also showed that one Tf molecule was conjugated to one PAMAM-PEG molecule. The concentration of the nanoparticles PAMAM-PEG-Tf/ TMZ was calculated as 200 μM using a Tf standard curve.

Encapsulating efficiency (EE%) and drug loading capacity (LC%)
According to calculation by free TMZ in solution, the results of TMZ EE% was (74.5 ± 7.8%) in the PAMAM-PEG-Tf and (79.4 ± 2.9 %) in PAMAM. The LC% of TMZ was (14.71 ± 1.29 %) in PAMAM and (4.31 ± 0.35 %) in PAMAM-PEG-Tf. The solubility of TMZ in water was enhanced when it interacted with PAMAM or PAMAM-PEG-Tf. The TMZ concentration of 5.6 mM in PAMAM and 5.2 mM in PAMAM-PEG-Tf solution was calculated. There were no significant differences in terms of percent EE and TMZ concentration between PAMAM and PAMAM-PEG-Tf. However, the percent LC of TMZ in PAMAM-PEG-Tf was significantly lower than that found for PAMAM (P < 0.01) due to the presence of Tf.

Methylation status of the MGMT promoter of GBM
Methylation status of the MGMT promoter using a nested Methylation-Specific Polymerase Chain Reaction (MSP) assay in GBM resected tissues are shown in Figure  2A. Both the methylated and unmethylated status of the MGMT promoter were shown for samples 1 and 2, and the positive methylation status was shown in sample 3, as well as for cultured GSCs SU2 and 51A, which were all considered TMZ sensitive in the clinic.  Cryostat sections using TUNEL assay was used to determine the apoptotic effect of drugs and then analyzed by fluorescence microscopy after GBM samples were incubated with PAMAM-PEG-Tf/TMZ for 24 h ( Figure 2B). PAMAM-PEG-Tf/TMZ was indicated by red fluorescence staining, disrupted DNA was stained by green fluorescence and cell nuclei were visualized by DAPI counter-staining. As shown in Figure 2C, few TUNEL-positive cells appeared in the control group with no treatment, and an increased number of TUNEL-positive cells were observed when the sample was incubated with PAMAM-PEG-Tf/TMZ, PAMAM/TMZ or TMZ as compared the control (P < 0.01). There was no significant difference among the number of TUNEL-positive cells when the sample was incubated with PAMAM-PEG-Tf/ TMZ, PAMAM/TMZ or TMZ (P > 0.05).

Identification of sorted GSCs and non-GSCs
Isolated GSCs were identified by sub-sphere formation assay ( Figure 3A) and surface marker analysis ( Figure 3B). Tumor spheres were formed 3 to 4 weeks after primary culture of tumor cells from GBM patients. Subspheres were formed 4-5 days after primary spheres were dissociated. It was found that 100 ± 50 cells assembled in each subsphere, which was similar to that seen for primary spheres. GSCs showed strong expression of nestin and moderate expression of CD133 by immunofluoresence staining. Flow cytometry showed that (41.8 ± 1.
Fluorescence densities from PAMAM-PEG-Tf/TMZ in TfR+ GSCs as compared matched non-stem stem cells was markedly higher (P < 0.05 or P < 0.01; Figure 5D), and western blot analysis showed no expression of TfR on TfR-SU2 and 51A cells, which indicated successful TfR knockdown using siRNA ( Figure 5E).

BBB permeability in mouse brain with xenografts
Analysis on PAMAM-PEG-Tf/TMZ permeability to the BBB was performed at 2 h and 12h post-intravascular injection ( Figure 6). Images from mouse brain tumors were obtained, and microvessels were shown by a green signal from CD31 stained cells. Red signals from PAMAM-PEG-Tf/TMZ conjugated with the Alexa Fluor 555 unit from TfR, were seen near blood vessels 2 h postadministration. The red signal of the entire imaged area in the mouse brain tumor was shown on images 12 h postadministration.

The study of cytotoxicity and uptake in SOX2+ cells from intracranial xenografts
Cellular apoptosis from murine brain tumors was evaluated by TUNEL staining after 24 h of treatment with PAMAM-PEG-Tf/TMZ, in which, disrupted DNA showed a green signal ( Figure 7A). A number of apoptotic cells appeared in the area of the red signal from Tf with Alexa Fluor 555, which indicated active apoptosis in tumor cells absorbing to PAMAM-PEG-Tf/TMZ. The uptake by brain tumors and location in SOX2+ cells of PAMAM-PEG-Tf/ TMZ are shown in Figure 7B    xenografts was increased significantly than that in TfR-SU2 (29.1 ± 2.6)% and (24.5 ± 3.1)% (P < 0.01).
Microscopic examination of the brains revealed targeting features of PAMAM-PEG-Tf/TMZ in the tumor. PAMAM-PEG-Tf/TMZ was indicated by red fluorescence, and SOX2+ cells were shown by green fluorescence by immunostaining analysis. Xenografts from TfR+ cells showed a comprehensive red fluorescence, and SOX2+ cells similarly showed full patterns of red fluorescence as compared TfR-cell xenografts. In addition, minimal red fluorescence was seen in SOX2+ cells in spite of a complete moderate intensity of red fluorescence in SOXcells. As determined by fluorescence of individual mice, PAMAM-PEG-Tf/TMZ was absorbed by SOX2+/TfR+ cells, but not by SOX2+/TfR-cells. This observation indicated a specific location in TfR+ GSCs in vivo.

Therapeutic efficacy in vivo
The efficacy of oral TMZ and i.v. TMZ nanoparticles for tumor suppression was evaluated in the nude mouse intracranial xenograft models. The status of the brain tumor and suppression of tumor growth were monitored since mice were treated by chemotherapy. Figure 8 and Table 2 showed survival results using Kaplan-Meier curves, and multiple group comparisons were described using Cox survival plots. Significant anti-tumor efficacy was observed in all treated groups. For SU2 cells implanting mice ( Figure 8A), the median survival time (MST) of the mice administered oral TMZ was 43.5 ± 8.3 days (95% CI, 37.6 -49.4 days), and exhibited significant antitumor effects as compared the control group 32.6 ± 4.1 days (95% CI, 29.6 -35.6 days; P < 0.01) and i.v.  implanted with GSCs were used for the study of BBB permeability at periods of 2 h and 12 h post-intravascular adoptive transfer. Red signal patterns were derived from PAMAM-PEG-Tf/TMZ, and green signal patterns were derived from immunostaining anti-CD31 antibody. The fluorescent images at a magnification of ×100, were shown in the white pane, which was magnified to a ×400 field view, and shown on the next line. Scale bar =50 μm Mice bearing TfR+ gliomas and that had received i.v. PAMAM-PEG-TfR/TMZ, also showed an MST of 66.4 ± 11.9 days (95% CI, 57.9 -74.9 days), which was significantly longer as compared to that of TfR-cell implanted mice and PAMAM/TMZ administration (P < 0.01). PAMAM-PEG-Tf/TMZ administration retarded xenograft growth from TfR+ cells, and did so most significantly among the treated groups. Survival data of 51A cells ( Figure 8C) that had been implanted in mice was accordant with that of SU2 cells. When receiving the PAMAM-PEG-TfR/TMZ injection, MST of mice bearing TfR+ 51A cells was 69.6 ± 13.2 days, which was significantly increased as compared control (29.1 ± 4.1 days), PAMAM-PEG-Tf administration (31.0 ± 6.4 days), TMZ administration (41.5 ± 7.6 days), PAMAM/TMZ treatment (50.8 ± 10.3 days) and PAMAM-PEG-TfR/TMZ treatment for TfR-51A implanted mice (50.3 ± 10.2 days).
Results indicate that when administrated with TMZ, the survival time both in SU2 and 51A cells implanted mice was significantly different from that of counterpart control group (P < 0.01). The difference in MSTs between groups of mice administrated with PAMAM-PEG-TfR/ TMZ for TfR-cells and PAMAM/TMZ administration for TfR+ cells were not significant (P > 0.05). Non-targeting nanoparticles encapsulating TMZ extended survival time of GSCs implanted mice comparing to TMZ treatment alone (P < 0.05). Statistically significant differences were noted in TfR targeting and non-targeting groups of nanoparticles encapsulating TMZ (P < 0.01). Multivariate analysis and Cox proportional hazards model indicated that TMZ, nanoparticle and targeting factor correlated significantly with survival in mice bearing brain tumors ( Figure 8B for SU2 and Figure 8D for 51A cells implanted mice).

DISCUSSION
In this study we evaluated the validity of PAMAM-PEG-Tf/TMZ nanoparticles in damaging GSCs, as well as non-stem tumor cells. Nanoparticles that were targeted to TfR, with high expression in both populations, were prepared by conjugating Tf to PAMAM dendrimers, which could specifically target TMZ to brain tumors, and especially GSCs.
Overexpressed TfR on active proliferating surfaces of tumor cells is widely used to deliver drugs since iron is a basic element that is required during cellular metabolism. Complexes taking TfR as a target, is produced by Mebiopharm, SynerGene Therapeutics and Calando Pharmaceuticals, and has been used in clinical trials to deliver anti-cancer drugs [10].
TfR targeting is a sufficient anti-cancer therapy, so much so that a common marker is used to target both GSCs and non-stem tumor cell populations. However, taking TfR as a target to delete glioma is a more practical strategy due to over-expression on the surfaces of both populations. In previous studies, the prepared nanoparticles, which comprised a liposomal complex employing an anti-TfR single-chain variable fragment as a targeting ligand, carried the wtp53 gene, and showed anti-cancer activity by inducing death of both cancer stem cells and non-stem cancer cells [11]. This approach was used in Phase I trials in advanced solid cancer patients [12,13]. Combined use of nanoparticles and conditional TMZ chemotherapy increased its tumoricidal efficacy in TMZresistant glioma transplanted mice [14].
This study demonstrated that PAMAM-PEG-TfR/TMZ not only inhibited glioma growth, but also
accomplished tumor regression and delayed tumor recurrence, at least in orthotopic glioma nude mouse models. Such responses were likely produced by elimination of both GSCs and non-stem tumor cells, which indicated the broad applicability of targeting both populations. The specificity of PAMAM-PEG-Tf/TMZ also prevented significant side-effects of TMZ because of the TfR targeting effect. High uptake efficiency was observed by both populations in vitro with similar frequencies; however, a more significant fluorescence intensity was shown in GSCs as compared non-stem tumor cells at 12 h after 50μM PAMAM-PEG-Tf/TMZ treatment. The targeting ability of PAMAM-PEG-Tf/TMZ was further confirmed in vivo using intracranial transplanted tumor nude mice. Red fluorescence from PAMAM-PEG-Tf/TMZ was readily seen in the tumor area.
The fluorescent images showed clearly that the PAMAM-PEG-Tf/TMZ and the stem cell marker SOX2, were co-localized in intracranial xenografts of TfR+ GSCs, and evident apoptosis was present in the cells that fully exhibited PAMAM-PEG-Tf/TMZ. No absorption of PAMAM-PEG-Tf/TMZ was seen in SOX2+ cells from TfR-GSC xenografts. Therefore, PAMAM-PEG-Tf/TMZ suppressed tumor growth more effectively than TMZ and PAMAM/TMZ due to drug accumulation by TfR+ GSCs.
MSTs of mice with the TfR+ SU2 or 51A xenograft that was treated with PAMAM/TMZ, was similar to that as its TfR-counterpart of SU2 or the 51A xenograft, which received PAMAM-PEG-Tf/TMZ, due to no TfR targeting. However, the anti-tumor efficacy of nanoparticles comprised of PAMAM/TMZ and PAMAM-PEG-Tf/TMZ, was more potent than that of free TMZ. Due to non-targeting nanoparticle uptake by cells, PAMAM-PEG-Tf/TMZ was absorbed more easily by TfR-non-stem tumor cells due in part to the decreased dose, and longer period of therapy as compared with the targeting particles. Moreover, TfR-GSCs failed to absorb PAMAM-PEG-Tf/TMZ, but TfR-non-stem tumor cells could absorb abundant drugs, which indicated that the prepared nanoparticles were not absorbed by GSCs. Thus, targeting GSCs is an important option in preventing glioma recurrence.
Drug nanocarriers should need the characteristics of high drug loading and entrapment rate, biodegradable carrier material, low or no toxicity, appropriate particle size and long cycle period. The structure characteristics of common drug nanocarrier are showed in Table 3. Non-targeting uptake is obvious obstacle of small molecular chemical drugs for cancer therapy. A large amount of experimental researches focus on further development of novel and more efficient delivery systems. Nanoparticles linked with a ligand are widely used for the delivery of anticancer drugs. Application and synthesis of delivery vehicles targeted TfR played an important role because the TFR is overexpressed on the surface of various fast-growing malignant tumor cells. Receptor-mediated endocytosis of TF induced rapid tumor-cell-specific uptake of targeting nanoparticles, and the internalized nanoparticles could be effectively degraded to release functional drugs Nanoparticles Nanoparticles are composed of polymer material capsules and liquid (water or oil) inner cores. The drug is usually encapsulated by the polymer film in the inner core layer. Nanoparticles include polylactic acid (PLA), polyglycolic acid (PGA), polycaprolactone (PCL), polylactic acid glycolic acid (PLGA), natural polymer materials, and so on.
Nano-liposomes Nano-liposomes are multi-layer vesicle structure formed by phospholipids. They are. each layer are composed of lipid bimolecular membrane, interlayer and liposome inner core are the water phase, and the bimolecular membrane is oil phase.
Nano-micelles Nano-micelles are composed of hydrophilic shell and hydrophobic core. Hydrophilic segments include polyethylene glycol (PEG), polyethylene oxide (PEO), polyoxypropylene, and so on, and hydrophobic segments include PLA, PGA,PCL,PLGA, Chitosan, and so on.
Nano-magnetic particles Nano-magnetic particles are composed of drug magnet particle carrier and skeleton material.

Dendrimers
Dendrimers are symmetrically spherical polymer, showing a dendritic geometric appearance. Their molecular surfaces have functional groups with very high density, and wide cavities exist inside the molecules.
molecules in the cells. Kanwar et al., prepared ironsaturated bovine lactoferrin, and displayed a 60-80% similar sequence with Tf nanocapsules, an observation that validated the capacity of these nano-capsules to kill colon cancer stem cells and induce apoptosis by targeting survivin [15]. The transferrin-targeting nanoparticle delivery system that carries the tumor suppressor microRNA-1, efficiently delivered miR-1 and inhibited migration of GBM patient-derived GSCenriched spheres [16]. Nanocomplexes of cationic liposomes that were conjugated with TfR single-chain antibody fragments and which encapsulated TMZ were taken up by cancer stem cells [17]. Concordant with our results, these previous studies all showed ligands of TfR that had the ability to target cancer stem cells. Temozolomide (TMZ) was approved by the FDA as one of the most commonly used alkylating agents to target glioblastoma. MGMT promoter methylation as compared un-methylation is associated with longer survival when GBM patients are treated with TMZ [18]. DNA methylation signatures from primary cultured glioma cells were present in xenograft tumors, which indicated no tissue culture-related epigenetic changes [19]. This study selectively cultured GSCs from clinical samples or cell-lines that possessed the methylated MGMT promoter. These GSCs were sensitive to TMZ treatment, and could be used to validate the anti-tumor efficacy of PAMAM-PEG-Tf/TMZ. PAMAM-PEG-Tf/TMZ solution when administered intravenously to experimental mice with the aim of examining its anti-cancer efficacy. The results mimicked in vitro experiments and were concordant with observations made for clinical samples. Cytotoxicity was observed for in vitro cultured CSCs and non-stem tumor cells. Moreover, apoptosis was shown following in vivo administration and in studies of clinical samples.
Co-localization of CD31 positive cell staining and PAMAM-PEG-Tf/TMZ was seen to have accumulated in the xenograft. PAMAM-PEG-Tf/TMZ accumulated from nearby vessels two hours post-administration to a wide area including far from the vessels at 12 h postadministration. This observation indicated enhanced permeability of the BBB. TfR-targeting nanoparticles could efficiently deliver drugs through the BBB. In previous studies, and due to its ability to cross the BBB, Tf-targeted nanoparticles incorporating zoledronic acid increased the tumoricidal efficacy of this drug in intracranial U373 xenografts [20]. The delivery of a Tf-conjugated magnetic silica nanoparticle complex that was loaded with doxorubicin and paclitaxel, was enhanced in the intracranial U87 xenograft of BALB/c nude mice [21]. Our results were concordant with these previously published studies of nanoparticles that used Tf as the targeting ligand, which was shown to quite easily penetrate BBB.

CONCLUSION
Overall, in this study, the efficacy of a novel nanoparticle complex of PAMAM-PEG-Tf/TMZ was evaluated to GSCs. A high dose uptake and significant cytotoxicity of PAMAM-PEG-Tf/TMZ in GSCs was observed due to the targeting function of Tf, in which the ligand was highly expressed in GSCs. PAMAM-PEG-Tf/TMZ traversed the BBB and delivered TMZ to the avascular region of tumor, and delivered an effective dose of TMZ specifically to tumor cells. After surgery and radiotherapy, the chemotherapeutic protocol was commonly ineffective against drug-resistant GSCs. The achieved delivery mechanism, provoked a potent induction of glioma cell apoptosis, and especially GSCs. The targeting TfR nanoparticles could be used for an effective therapeutic strategy against GSCs, and against non-stem tumor cells, and it would provide a promising tumoricidal strategy for treating glioma displaying MGMT promoter methylation.

Synthesis of PAMAM-PEG-Tf bioconjugates
PAMAM and Tf were conjugated by bifunctional groups of maleimide-polyethylene glycol 2000-amino succinimidyl succinate (MAL-PEG2000-NHS) to synthesize PAMAM-PEG-Tf bioconjugates (Figure 9). PAMAM dendrimers were reacted with MAL-PEG2000-NHS at a ratio of 1:10 (mol/mol) in PBS (pH 8.0) for 2 h at room temperature. The surface NH 2 groups of PAMAM were specifically reacted with the NHS groups of MAL-PEG2000-NHS. The resulting conjugate, PAMAM-PEG, was purified by ultrafiltration through a molecular weight cut-off membrane of 5KDa, following which, the buffer was changed to PBS pH7.0. Simultaneously, Tf with an Alexa Fluor 555 (Molecular Probes) fluorochrome was thiolated using Traut's reagent, and then the thiolated Tf was coupled to the periphery of PAMAM-PEG at a ratio of 1:5 (PAMAM to peptide, mol/mol) in PBS pH 7.0 for 24 h at room temperature. The MAL groups of PAMAM-PEG were specifically reacted with the thiol groups of thiolated Tf. To purify the bioconjugated product, gel filtration was employed with a Sephacryl S-300 gel filtration column. To examine TfR conjugation and PAMAM, the absorbance was measured at a wavelength of 280 nm, and bioconjugated synthesis was determined by ultraviolet-visible spectra, and the purity of bioconjugate synthesis was characterized by SDS-PAGE that was stained with Coomassie Brilliant Blue. Protein concentrations of synthesized bioconjugates were determined by measuring absorbance at 280 nm using a multiskan spectrophotometer model 1510 (Thermo Fisher Scientific, Vantaa, Finland) according to a free Tf standard curve.

Preparation of PAMAM-PEG -Tf/TMZ nanoparticles
TMZ was encapsulated in the interior of the PAMAM-PEG-Tf by mixing TMZ with the carriers. Next, the free TMZ molecules were separated by gel filtration chromatography. The relative dose of TMZ in PAMAM and PAMAM-PEG-Tf was calculated using the TMZ standard curve with a maximum absorption of 329nm. The encapsulating efficiency (EE%) and the drug loading capacity (LC%) of TMZ in PAMAM and PAMAM-PEG-Tf was determined by measuring TMZ concentration. Both EE% and LC% were calculated as indicated below. EE% = (TMZ in PAMAM or PAMAM-PEG-Tf / total amount of TMZ in solution) ×100%, LC% = (TMZ in PAMAM or PAMAM-PEG-Tf / solutes total weight) ×100%. Preparation of PAMAM-PEG-Tf /TMZ was determined by ultraviolet-visible spectra.

Statement of ethics
Tumor tissues from three pathologically diagnosed human GBM (Grade IV) surgical resections were used for immunohistochemistry, MSP and primary cell culture. The patients had provided their written and informed consent, which were under Institutional Review Board approval of the First Affiliated Hospital of Soochow University, China.

Uptake and cytotoxic assay in surgical samples of glioblastoma
GBM tissues from surgical excision procedures were collected immediately for the measurement of uptake and cytotoxicity assay as previously described [22]. The samples were incubated with PAMAM-PEG-Tf/TMZ (50 μM TMZ) for 24 h, washed twice and frozen embedded in an optimal cutting temperature (OCT) compound. Then 10 μm cryostat sections were prepared using an ultracryotome, following which, apoptotic cells in the GBM samples were detected by terminal deoxynucleotidyl transferase (TdT)-mediated dUTP-biotin nick end labeling (TUNEL) assay, using an In Situ Cell Death Detection Kit (Promega, USA) and used according to the manufacturer's instructions. Sections were counter-stained with anti-fade sealant containing 4'6-diamidino-2-phenylindole (DAPI). Fluorescence images were visualized and captured using a fluorescence microscope (Olympus BX40, Japan).

Primary cell culture of human brain tumors and cell-lines
Freshly resected human glioblastoma tumor samples were dissociated for primary cell culture. Tumor tissues were washed with phosphate buffered saline (PBS), minced mechanically into small fragments and dissociated into single cells and/or small clumps by tripsin.
Cells were cultured on non-adherent plates in serumfree DMEM/F12 medium containing 2%B27 supplement BBB permeability, uptake and cytotoxicity of PAMAM-PEG-Tf/TMZ in murine brain xenografts PAMAM-PEG-Tf/TMZ solution (5mg/kg) was injected into the tail vein of mice that were implanted with intracranial glioma in a volume of 100 μl. Next, animals were anesthetized and sacrificed. Mouse brains were removed, and frozen in OCT embedding medium to prepare sections at a thickness of 10 μm. Sections were immunostained with anti-CD31 antibody to analyze BBB permeability of PAMAM-PEG-Tf/TMZ after 2 h and 12 h post-injection. The slides were stained by immunofluorescence to detect SOX2 protein expression in xenografts to assess uptake by GSCs, and TUNEL staining was carried out for analysis of cell apoptosis after 24 h post-injection.

Evaluation of therapeutic potential in vivo
Mice were randomly divided into six experimental groups, and there were 10 mice in each group: Group 1, control; group 2, PAMAM-PEG-Tf; Group 3, TMZ; group 4, PAMAM/TMZ; group 5, PAMAM-PEG-Tf/TMZ; and group 6, PAMAM-PEG-Tf/TMZ . TfR+ cells were implanted with intracranial xenografts in groups 1 to 5, and TfR-cells were used in group 6. Drugs that included PAMAM/TMZ and PAMAM-PEG-Tf/TMZ containing 5mg/kg TMZ were injected into the tail vein of mice following brain tumor formation. TMZ was administrated by intra-gastric injection using a suspension at a dose of 20 mg/kg. Mice received drugs for 5 days continuously and were monitored daily until severe neurological deficits appeared. Survival analysis was used to compare the differences of each group according to survival time.

Statistical analysis
Statistical analyses were carried out using SPSS version 19.0 (SPSS Inc., Chicago, IL, USA), and data was statistically determined by one-way ANOVA. The significance level was considered at an alpha value of P < 0.05. Each in vitro experiment was repeated at least three times. Overall mouse survivals that were implanted with intracranial glioma were estimated via Kaplan-Meier survival curves, and compared between groups via stratified log-rank tests. A Cox proportional hazards regression model was used to examine the validity of stratifying by groups and by factors.