Radiation sensitization of glioblastoma by cilengitide has unanticipated schedule-dependency


  • Conflict of Interest/Disclosure: Merck KGaA provided research support for some of the work described herein. Simon L. Goodman and Diane Hahn are employed by Merck-Serono, which is currently testing cilengitide in phase-3 clinical trials for cancer. Dr. Goodman is also a copatent holder on the cilengitide state-of-matter patent.


We investigated whether cilengitide could amplify the antitumor effects of radiotherapy in an orthotopic rat glioma xenograft model. Cilengitide is a specific inhibitor of αv series integrins, and acts as an antiangiogenic. U251 human glioma cells express αvβ3 and αvβ5 integrins. We used in vitro assays of adhesion and growth of tumor and endothelial cells to evaluate cytotoxicity and the potential for cilengitide to enhance radiation toxicity. Treatment was then evaluated in an orthotopic model to evaluate synergy with therapeutic radiation in vivo. In vitro, cilengitide blocked cell adhesion, but did not influence the effects of radiation on U251 cells; cilengitide strongly amplified radiation effects on endothelial cell survival. In vivo, radiotherapy prolonged the survival of U251 tumor-bearing rats from 50 to over 110 days. Cotreatment with cilengitide and radiation dramatically amplified the effects of radiation, producing survival over 200 days and triggering an enhanced apoptotic response and suppression of tumor growth by histology at necropsy. Signaling pathways activated in the tumor included NFκb, a documented mediator of cellular response to radiation. Because cilengitide has a short plasma half-life (t½ ∼ 20 min), antiangiogenic scheduling typically uses daily injections. We found that a single dose of cilengitide (4 mg/kg) given between 4 and 12 hr prior to radiation was sufficient to produce the same effect. Our results demonstrate that blockade of αv integrins mediates an unanticipated rapid potentiation of radiation, and suggests possible clinical translation for glioma therapy. © 2008 Wiley-Liss, Inc.

Integrins control cell attachment to extracellular matrices (ECMs) and participate in cellular defense against genotoxic assaults.1 These defense mechanisms are a major factor in the resistance of solid malignancies to radiotherapy. In this in vivo study, we examine the effects of an inhibitor of αv-integrins, cilengitide, on the growth of gliomas in response to external beam radiotherapy. Malignant gliomas, including anaplastic astrocytoma and glioblastoma multiforme (GBM), are the most common primary brain tumors, afflicting some 6/100,000 individuals annually within the United States.2 Current treatment options include surgery, radiation therapy (RT) and chemotherapy. But the efficacy of treatment is limited by the infiltrative nature of GBMs, by sustained tumor angiogenesis, and by a marked resistance to chemo and radiotherapies. Indeed, clinical prognosis is poor and the median survival from diagnosis of 12 months in GBM has not changed appreciably over a quarter century.3 Gliomas, and especially anaplastic gliomas, infiltrate and spread great distances in the brain from a peripheral zone of infiltrating cells in the highly vascularized cellular rim of tumor that surrounds a central necrotic core.4 The infiltrating tumor cells cause an almost inevitable local recurrence and clinical progression.5 Recurrence following surgery and radiation often occurs adjacent to the initial tumor in the margin which contains a leaky neovasculature.6 Angiogenesis in glioblastoma is dramatic compared with other malignancies, and clearly fosters tumor growth.7

It is well established that angiogenesis involves a substantial reprogramming of endothelial cells,8 including the increased expression and activity of integrins αvβ3 and αvβ5.9 We have focused on the αv series integrins, as αvβ3 has an interesting and restricted distribution (e.g., on gliomas and tumor invasive endothelia).10 Expression of αv-series integrins is modulated in response to stress factors including hypoxia and tumor irradiation.11 Integrin αvβ3 binds diverse ECM ligands with an exposed NH2-Arginine-Glycine-Aspartic acid (“RGD”) sequence.1, 12

αvβ3 and αvβ5 integrins and their ligands are also overexpressed in glioblastoma and interact to inhibit tumor and vascular apoptosis. Extracts of normal brain white matter contain the αvβ3 ligand osteopontin, which promotes cell migration via αvβ3 and αvβ5 integrins.13 In summary, αv integrins, especially αvβ3, appear to regulate the cellular behavior of gliomas and endothelial cells, and support growth factor-mediated cell survival.

Pharmacological antagonists of integrin αvβ3 have been used in glioma tumor models. Interference with integrin αvβ3 stimulates glioma and endothelial cell apoptosis and, more significantly, prolongs survival in orthotopic glioma models.14 αv integrin antagonists are in clinical trials, including Vitaxin/Medi52215, 16 and cilengitide (EMD121974), a cyclic RGD-containing peptide.12, 15, 22

Here, we examine in vivo the effect of cilengitide on pro-apoptotic RT in the U251 orthotopic model of human malignant glioma. Cilengitide dramatically amplifies the efficacy of RT in this model, and does so in a scheduling regimen which suggests an unexpected role for αv integrins in this system.

Material and methods

Materials and immunoreagents

The anti-human integrin antibodies used were against αv (17E6), β6 (5C4) (Calbiochem, San Diego, CA), αvβ3 (LM609), αvβ5 (P1F6), αvβ6 (10D5), α2 (P1E6), α3 (P1B5), α4 (P4G9), α5 (P1F6), α6 (GoH3), β1 (P4C10), and β4 (ASC3) were all from Chemicon (Temecula, CA). Antifactor VIII (vascular marker) polyclonal and anti-MIB1 (Ki67) were from Dako (Carpenteria, CA). Apoptag (TUNEL stain) was from Intergen (Danvers, MA). Alu II probe was from Biogenex (San Remon, CA). Cilengitide (EMD 121974: cyclo-(Arg-Gly-Asp-DPhe-(N-Me)-Val) was supplied as an apyrogenic sterile infusion solution in physiological saline at 15 mg/ml (25.5 mM) provided by CTEP NCI (Bethesda, MD) and Merck, KGaA (Darmstadt, Germany). All other reagents were of at least research grade. Gill II hematoxylin and eosin (E&E) Y were from SurgiPath (Richmond, IL).

Cell culture

Human glioma cell line U25117 was originally obtained from American Type Culture Collection (ATCC, Rockville, MD). Cells were cultured at 37°C in 5% CO2 and were maintained in DMEM containing 10% (v/v) fetal bovine serum, 4 mM glutamine, 100 IU/ml penicillin, 100 μg/ml streptomycin and 1% nonessential amino acid (Invitrogen Life Technologies, Inc., Rockville, MD). Human umbilical vein endothelial cells (HUVECs) were isolated18 and cultured in EGM-MV media (Promocell, Heidelberg, Germany), and used between passages 4 and 7.

Cell characterization

U251 glioma cells in log phase growth were analyzed for integrin expression profile using live-cell FACs, attachment dependency to ECM substrata, and the sensitivity of the attachment to cilengitide on vitronectin-coated substrata as previously described.19

Effect of cilengitide on the radiation sensitivity of U251 and endothelial cells in vitro

Nontissue culture treated 96-well plates were coated overnight with vitronectin (0.5 μg/ml) before platting freshly harvested HUVECs or U251 cells (5000 cells in 100 μl) in complete serum containing growth medium for 3 hr. Serially diluted cilengitide or the diluent alone (complete medium) was added (100 μl), and the cells returned to the incubator for 4 hr at 37°C, before being irradiated (1–25 Gy doses of X-rays; 122 kV; Faxitron 43855A) (Rhode & Schwarz, Cologne, Germany). Following irradiation, plates were returned to the incubator and cultured for a further 5 days, before being assessed for viable cells.20

Orthotopic U251 tumor growth and therapy studies in vivo

Following IACUC guidelines in an approved animal-use protocol, nude rats were inoculated intracerebrally as follows: animals were anesthetized with 100 mg/kg ketamine, 15 mg/kg xylazine and 0.05 ml atropine i.m. The surgical zone was swabbed with Betadine solution, the eyes coated with Lacri-lube and the head immobilized in a small animal stereotactic device (Kopf, Cayunga, CA). A #2701 10 μl Hamilton syringe with a 26s gauge needle containing U251MG tumor cells freshly harvested from log phase growth [5 × 105 in 10 μl of phosphate-buffered saline (PBS)] were injected at a rate of 0.5 μl/10 sec until the entire volume had been injected.

Treatment protocol: Drug administration schedule

Cilengitide was administered according to schedule by daily i.p. injection using a 26-gauge needle, beginning 2 weeks following tumor implant, at a dose of 4 mg/kg. Prior to RT, the drug was freshly prepared as a working concentration of 15 mg/ml, and administered in initial cohorts for 7 days before RT, or in subsequent cohorts as a single dose at various time points immediately preceding RT, as detailed in the text. Control animals had identical handling except that vehicle (control) was administered. Post-RT, cilengitide treatment was continued either daily in the initial cohorts, or only for 7 days in subsequent cohorts. Animals were then monitored until they became symptomatic from brain tumor. As detailed in the Results section, this de-escalation of treatment duration had no adverse consequence for outcomes.

Radiation treatment

For radiation treatment, rats were anesthetized as described in “Implantation” above, and restrained in a stereotactic head holder before being stereotactically irradiated once at 14–28 days post-tumor implantation following either 7 days i.p. dosing, or a single i.p. dose of cilengitide (see the earlier section). Radiation was delivered using a 6MV Varian Clinac (Varian Medical Systems, Inc., Palo Alto, CA) with a 10 cm × 10 cm field and a secondary collimator to generate a single collimated dorso-ventral beam of 6 MV X-rays with a placement of bolus material so that 95–100% of the central axis dose of 25 Gy was delivered to the tumor volume.21 The volume of brain irradiated was a cylinder of 3-mm radius at the 80% isodose boundary. The dose rate of 6 MV X-rays was 600 cGy per min using a source-to-skin surface distance of 75 cm.

Pharmacokinetic analysis

Plasma concentration-time profiles in Wistar rats were determined after single cilengitide i.p. administration of 2.5 or 15 mg/kg, or i.v. administration of 2.5 mg/kg (3 male and 3 female rats per dose). Plasma levels following retro-orbital sampling were measured using a validated HPLC/MS method, as described.22 The limit of quantification was 4 ng/ml. Pooled concentration profiles were analyzed by WinNonlin Version 5.0 (Pharsight Corporation, Mountain View, CA).

Tissue collection, processing and histologic staining

Conventional H&E staining was performed on sections from tumor and adjacent blocks using Gill II H&E Y (SurgiPath). Alu in situ hybridization and immunohistochemistry were performed on immediately adjacent sections, as previously described.23

Western blot

The expression of phospho-p65 RelA was examined by Western blotting using a phospho-specific antibody (S265) from Cell Signaling Technology (Danvers, MA). Pellets from freshly harvested cells in log phase growth (106 cells) were lysed by resuspension in 100 μl of lysis buffer (0.5% Na deoxycholate; 2% NP-40; 0.2% SDS; 1 mM PMSF; 50 μg/ml aprotinin; 50 μM leupeptin; 0.5 mM Na3VO4, 50 mM NaCl, 25 mM Tris-HCl, pH 7.4) on ice for 15 min, then an equal volume of 2× sample buffer was added and the samples boiled for 5 min. Lysates (30 μg protein) were resolved by SDS-PAGE and transferred to nitrocellulose membranes. The membranes were blocked with 5% dry milk in PBS and subsequently stained with the primary antibody (S265). Specific reactive bands were detected using goat-anti-rabbit or goat-anti-mouse IgG HRP conjugates (BioRad, Hercules, CA) and the immunoreactive bands visualized by ECL Western blotting detection kit (Amersham, Arlington Heights, IL).

Microarray methods

RNA Isolation: from untreated, 1 hr post cilengitide, and 8 hr post cilengitide-treated animals, total RNA was extracted and purified from triplicate xenograft specimens using Trizol (Invitrogen Life Technologies, San Diego, CA) followed by RNAeasy Purification (Qiagen, Mississauga, Ontario, Canada). The quality of RNA was assessed from the ratio of absorbance values at 260 and 280 nm; only those samples with 260/280 ratio of 1.8–2.2 were used. Visualization of intact 28S and 18S ribosomal RNA bands on a Bioanalyser 2100 (Agilent Technologies, Inc., Palo Alto, CA) confirmed the presence of high quality RNA.

Sample labeling and hybridization

Microarray experiments were performed using the human expression HG-U133 plus 2.0 gene chip arrays (Affymetrix, Santa Clara, CA) following the manufacturer's recommendations. Briefly, 5 μg total RNA was reverse-transcribed with a T7-(dT) 24 oligonucleotide as primer, labeled with biotin, and fragmented using Affymetrix reagents. Of the resulting cDNA, 10 μg was loaded on each chip. After washing and staining with streptavidin-phycoerythrin (Invitrogen, Carlsbad, CA), the chips were scanned using a Genechip Scanner 3000 workstation (Affymetrix).

Statistical analysis

Expression profiles from microarray experiments were analyzed using GeneSpring™ software version 7 (Silicon Genetics, CA). Gene expression data was normalized in 2 ways: “per chip” normalization and “per gene” normalization. For per chip normalization, all expression data on a chip was normalized to the 50th percentile of all values on that chip. For per gene normalization, the data for a given gene was normalized to the median expression level of that gene across all samples. Samples were classified to 1 of 3 groups: (i) No cilengitide treatment, (ii) 1 hr after cilengitide treatment and (iii) 8 hr after cilengitide treatment. Gene expression changes were identified as being statistically different when analyzed by “Group” using a parametric test where variances were not assumed equal (i.e., Welch ANOVA). For assignment of statistical significance, p-value cutoff was 0.001 and Benjamini and Hochberg False Discovery Rate multiple-testing-correction was used. Raw expression data are provided (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE12949) (Table I).

Table I. cDNA Microarray Studies
  1. Gene expression patterns were generated using cDNA microarrays, profiling tumor tissues harvested 1 and 8 hr after single dose cilengitide, given at 4 mg/kg. Ninety-four genes were lower at 8 hr compared with 1 hr and 104 genes were higher. Raw microarray data are on the NIH Gene Expression Omnibus (GEO) website http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE12949.

104 Genes with expression significantly higher in 8 hr samples
94 Genes with expression significantly lower in 8 hr samples

Pathway and gene network analysis

Genes identified as being significantly differentially expressed were analyzed with Ingenuity Pathways Analysis (Ingenuity Systems, Redwood City, CA; http://www.ingenuity.com), a web-delivered application. A data set containing gene identifiers and their expression values was uploaded as a text file. Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base. A fold-change-cutoff of 1.5-fold was set to identify genes whose expression was differentially regulated. These genes, called “focus genes,” were then used as the starting point for generating biological interaction networks with a network size of ∼35 elements. Ingenuity Pathways Analysis then computes a score for each network according to the fit of the user's set of significant genes.

Detection of apoptotic and autophagic cells

Apoptotic cells are identified by nuclear fragmentation into spherical structures of hyperdense chromatin, cell rounding and shrinkage. For autophagy detection, cells were stained with isoform B of human microtubule-associated protein 1 light chain 3 (Abgent AP1802a, San Diego, CA) as described.24 The paraffin sections were deparaffinized to water and heated in a citrate buffer, pH 6.0, for 20 min. Unless otherwise stated, all reagents are from Biocare Medical (Concord, CA). Blocking was with Rat Blocker for 22 min, and the antibody was diluted to 1:100 and incubated at room temperature for 90 min. Detection was with rabbit on rodent HRP polymer for 30 min followed by a purple chromogen and counterstain in hematoxylin. Autophagy was identified by intracytoplasmic vacuoles containing granular material. For apoptosis and autophagy cell counts, 15 high power fields were counted. Results are presented as mean values ± SEM.


Characterization by integrin expression and adhesion of U251 glioma cells

The U251 cell line used in this study expressed both αvβ3 and αvβ5 as seen from their FACS staining with LM609 and P1F6 antibodies, respectively (Fig. 1a). Expression of αvβ3 was uniform in the population, and judged by mean fluorescence intensity was about half that of αvβ5. U251 did not express ανβ6. Of the beta1 integrins, α3 and β1 subunits (which as heterodimers comprise the laminin-5 receptor) were strongly expressed; subunits that comprise the collagen (α2β1), fibronectin (α5β1, α4β1) and laminin receptors (α6 and β4) were also present (Supp. Info. Table I). This profile corroborates that previously reported for U251.25 U251 cells attached and spread on fibronectin, vitronectin and collagen I, but weakly to fibrinogen. U251 attachment to vitronectin was dependent on αv integrins, as shown by sensitivity to the blocking antibody 17E6, and was likely mediated entirely by αvβ3 and αvβ5 as revealed by sensitivity to cilengitide, but not to a RAD-control peptide. Cilengitide does not affect α5β1 or α4β1-mediated adhesion to fibronectin nor α2β1-mediated attachment to collagen (Figs. 1b and 1c).

Figure 1.

Integrin expression and attachment dependency characteristics of U251 cells. Expression, usage and sensitivity to cilengitide of integrins on U251 glioma cells: (a) U251 cells in log phase growth were harvested and analysed for integrin expression by FACS with complex or chain-specific primary mouse anti-integrin antibodies followed by Alexa488 labeled anti-mouse IgG and analysed by FACS, or with secondary antibody alone (Cntrl). Substantial amounts of αvβ3 and αvβ5, but not β6 integrin were present. (b) integrins on U251 mediated attachment to vitronectin (squares), fibronectin (triangles up), collagen type I (diamonds) but not fibrinogen (triangles down). (c) U251 adhesion to vitronectin was sensitive to the anti-αv integrin antibody 17E6 (squares open), and the RGD-peptide cilengitide (circles closed) but not to the corresponding RAD-peptide (circles open).

Cilengitide increases the sensitivity of U251MG and endothelial cells to radiation in vitro

Both U251 and HUVECs were sensitive to radiation, and cilengitide increased this sensitivity in vitro. U251 cells had a 50% survival at 5 days following 10-Gy irradiation (Fig. 2a), while HUVECs were more sensitive, and 50% survival was rendered by 5 Gy (Fig. 2c). The addition of cilengitide from 500 pM up to 5 μM had small but measurable affects on the radiation response of the U251 cells in vitro, as judged by their viability (Figs. 2a and 2b). However, cilengitide strongly amplified the response to irradiation in HUVECs. At the highest irradiation dosage, 12 Gy, some 30% HUVEC survival was seen after 5 days in the absence of cilengitide, but, with an EC50 of 0.5 μM added cilengitide reduced this survival to levels slightly above background. As a single agent (i.e., no radiation), cilengitide was only modestly suppressive of U251 cell growth (Fig. 2b), but killed HUVECs with an EC50 of some 4 μM (Fig. 2d). Thus, cilengitide at moderate concentrations dramatically amplified cell kill by radiation of endothelial cells in vitro, and worsened the survival of U251. The study was also performed on serum-coated tissue culture plastic with similar results (data not shown). Cell attachment to serum-coated plastic is mainly supported by αv integrin dependent interaction with vitronectin from serum.26 This context represents a more native form of vitronectin than the material routinely used in tissue culture experiments.27

Figure 2.

Cilengitide increases the cytotoxicity of the irradiation both on endothelia and glioblastoma cells. (a, b) U251 cells or (c, d) HUVECs treated with cilengitide, irradiated and assayed for viability after 5 days. To clarify effects of treatment, the same data set for each cell type is plotted both by radiation dosage and by cilengitide concentration. Cilengitide strongly amplified the toxicity of radiation on HUVECs (c, d), and less strongly affected U251 (a, b). Error bars indicate s.d. (n = 3) (a, c: Log survival on Y-axis). Each curve shows a single cilengitide concentration: in nanomolar: 0 (squares closed), 0.5 (squares open), 5 (circles), 50 (diamonds), 500 (triangles down) and 5000 (triangles). (b, d: linear survival Y-axis): Each curve shows a single radiation dose in Gy: 0 (squares closed), 2 (triangles), 6 (triangles down), 11.5 (diamonds) and 22.5 (circles).

Cilengitide pharmacokinetics in the rat

To assess the concentrations of cilengitide present during irradiation, and the exposure of target tissues, we performed a pharmacokinetic (PK) study in the Wistar rat, and the PK parameters derived are shown in Table II. A 2-compartment model with first order absorption was then applied to simulate the time course after single i.p. injection at 2.5, 4 and 15 mg/kg. The observed cilengitide plasma concentrations measured after i.p. administration and the PK profiles predicted by the model are shown (Fig. 3). Weighted correlations between observed and predicted i.p. profiles were 0.93 and 0.96 (2.5 mg/kg and 15 mg/kg profiles, respectively). Main PK parameters derived from the model, as well as observed AUCn and Cmax data are shown in Table II. Bioavailability for the i.p. route was close to unity. The plasma versus time profiles were characterized by an α half-life of 0.24 hr, a β half-life of 1.2 hr and an absorption half-life of 0.21 hr. Volumes of the central and peripheral compartments were 320 and 29 ml/kg and clearance was 924 ml/hr/kg. There was no gender-related difference in PK parameters. The results suggest that the cilengitide concentration in the plasma after 4 hr of 4 mg/kg i.p. administration was <10 ng/ml.

Figure 3.

Observed and simulated plasma concentration-time profiles of cilengitide following i.p. injection into Wistar rats. Mean and standard deviations (n = 6: 3m + 3f) are shown for observed measurements of cilengitide plasma concentration at 2.5 mg/kg i.p. (circles) simulated at 2.5 mg/kg i.p. (solid line), observed at 15 mg/kg i.p. (square), simulated at 15 mg/kg i.p. (hatched solid line) and simulated profile for 4 mg/kg i.p. dosing (broken line). The y axis is displayed as ng/ml.

Table II. Rat PK Data
Pharmacokinetic parameterValue
Vc (mL/kg)320
Vp (mL/kg)28.8
t½ α (hr)0.235
t½ β (hr)1.23
t½ a (hr)0.210
CL (ml/hr/kg)924
AUCn (h ng/ml); 2.5 mg/kg i.v.2380
AUCn (h ng/ml); 2.5 mg/kg i.p.2380
AUCn (h ng/ml); 15 mg/kg i.p.16500
*AUCn (h ng/ml); 4 mg/kg i.p.4330
Cmax (ng/ml); 2.5 mg/kg i.v.5660
Cmax (ng/ml); 2.5 mg/kg i.p.3530
Cmax (ng/ml); 15 mg/kg i.v.15200
*Cmax (ng/ml); 4 mg/kg i.v.4860

We measured the PK of cilengitide in rats at doses bracketing the therapeutic dosage used, and found that it had a short half-life (Supp. Info. Table II). Although a substantial Cmax nearing 10 μM was reached with 4 mg/kg dosing. By 4 hr, the time synergy with RT maximized, the systemic concentration was only 20 nM. By 8 hr, when synergy was still strong, the calculated systemic levels were below 1 nM, concentrations that have little effect on cell attachment in vitro.19 The IC50 of cilengitide for receptor inhibition in ligand binding assays is in the low nanomolar range,19 so it is conceivable that signal pathways initiated at these integrins might be modulated at such low concentrations of cilengitide. We used this PK data and similar measurements in other species to investigate the allometric scaling properties of cilengitide (Supp. Info. Fig. 1). We found that the measured human data from a phase I clinical trial using cilengitide was perfectly predicted by the linear regression of the allometric scaling curve (data not shown). The implication being that there is no anomalous concentration or distribution effect specific to the rat.

Effect of cilengitide and radiation on U251 in vivo

Orthotopic growth of U251 glioma cells is lethal to animals after∼30 days, is largely unaffected by the administration of cilengitide alone (daily i.p. injections continuously at 4 mg/kg), is impaired by treatment with radiation alone (25 Gy), but is effectively eliminated by combining radiation with cilengitide (Fig. 4). Significant enhancement of long-term survival is seen with combined cilengitide and radiation (Fig. 4a). The therapeutic benefit from radiation when combined with cilengitide is also manifest even if the cilengitide is delivered as a single dose prior to the radiation. Figure 4b shows a KM plot comparing U251-bearing animal survival in cohorts treated with single dose cilengitide prior to RT; therapeutic benefit from the combination treatments was only evident when the dosing interval between cilengitide and radiation was greater than 4 hr or less than 12 hr.

Figure 4.

Kaplan-Meier plot depicting U251-bearing nude rat survival under treatment regimens of control, cilengitide alone, RT alone and combined cilengitide-RT. (a) Kaplan-Meier plot comparing survival of U251 intracranial xenografts treated as control (vehicle), radiation alone (RT) (25 Gy), cilengitide alone (EMD) (4 mg/kg, delivered as daily i.p. injections for the duration of the study) and combined cilengitide and radiation (4 mg/kg delivered daily for the duration of the study starting 7 days preceding radiation [25 Gy] which was delivered once). Treatments were initiated 14 days following implant. Control vs. RT, Control vs. cilengitide-RT and RT vs. RT-cilengitide are all statistically significant at p < 0.0001 (b) Kaplan-Meier plot depicting survival of U251 intracranial xenograft animals treated with cilengitide as a single dose (4 mg/kg) administered prior to RT (25 Gy) with intervals between cilengitide and radiation as noted in the Figure. Outcomes of the RT-cilengitide group with 8 hr pretreatment were significantly different from all other treatment groups at p < 0.001.

Cilengitide: Detection of apoptotic and autophagic cells

Results of apoptosis and autophagy counts are presented in Table III. Representative images of the morphologic detection of each are shown in Figure 5. There appears to be a significant difference in the rate of apoptosis when Cilengitide is combined with radiation, but no difference with regards to the timing of administration before radiation at either 1 or 8 hr. Pairwise comparisons among the 5 treatment groups were done using generalized estimating equation methods which take into account 2 measurements for each animal. The difference between 8 hr + RT and cilengitide only was significant (p = 0.031). The differences between 1 hr + RT and cilengitide only, 1 hr + RT and RT only, and 8 hr + RT and RT only were of borderline statistical significance (p = 0.055, p = 0.072 and p = 0.05, respectively).

Figure 5.

Detection of apoptosis and autophagy in vivo morphologic detection of apoptotic cells by nuclear fragmentation into spherical structures of hyperdense chromatin, cell rounding and shrinkage is shown in panel (a). Autophagy is shown in cells stained with isoform B of human microtubule-associated protein 1 light chain 3 by intracytoplasmic vacuoles containing granular material (b).

Table III. Apoptosis and Autophagy Counts: Descriptive Statistics by Pairwise Comparisons
  • 1

    p values using generalized estimating equation (GEE) methods.

Treatment groupNApoptosisAutophagy
1 hr + RT310.
8 hr + RT510.
EMD only44.
RT only24.
Pairwise comparisonsp1p1
1 hr + RT vs. 8 hr + RT0.9790.837
1 hr + RT vs. EMD only0.0550.888
1 hr + RT vs. RT only0.0720.640
1 hr + RT vs. untreated0.2540.387
8 hr + RT vs. EMD only0.0310.702
8 hr + RT vs. RT only0.0500.740
8 hr + RT vs. untreated0.2080.441
EMD only vs. RT only0.8400.540
EMD only vs. untreated0.5470.305
RT only vs. untreated0.4880.708

Mechanism of treatment synergy—pAkt

Integrins act as survival factors. Our hypothesis predicts that integrin antagonism would be associated with reversal of PI3k/pAkt expression. Activation of the Akt family of genes is a poor prognostic factor in cancer, and may correlate with resistance to RT.28 Even allowing for the half-life of existing pAkt, cilengitide therapy triggered a decrease in expression of pAkt within 8 hr as assessed histologically in vivo (Supp. Info. Fig. 2). This observation leads us to suggest that blockade of survival signals induces a pattern of gene expression that makes the cells more prone to apoptosis, which is reflected in the survival studies we have described here.

cDNA microarray studies

To identify steps in signal transduction pathways that correlated with the induction of cell death and treatment synergy seen with combination therapy, we analyzed changes in gene expression patterns derived from intracerebrally implanted U251 tumors using human specific-cDNA microarrays. We profiled tumor tissues harvested in vivo 1 and 8 hr after a single dose of cilengitide, given at 4 mg/kg. Microarray data have been posted on the NIH Gene Expression Omnibus (GEO) website http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE12949. We found that the expression of 94 genes were significantly lower at 8 hr compared with 1 hr and 104 genes were higher (Table I). When we classified these genes, excluding unrecognized ESTs, p65 RelA, a major component of the NFκB complex, stood out by being differentially expressed in multiple replicates. Other genes encoding products implicated in apoptosis, including AIP1, were upregulated, and antiapoptotic genes, notably PI-3 kinase, were downregulated. The known positions of these genes are within the NFκB signaling network and are shown in Supporting Information Figure 3.

P65 RelA NFκB

The expression of one of the genes identified as regulated by cilengitide in the microarray, p65 RelA NFκb, was examined by Western blotting using a (S275) phosphospecific antibody (Supp. Info. Fig. 4). After exposure of U251 cells in vitro to cilengitide for 1 or 8 hr, phosphorylation at S275 was significantly downregulated (Supp. Info. Fig. 4, middle panel), consistent with reports that inhibition of RelA phosphorylation sensitizes to apoptosis in constitutive NF-κB-expressing and chemo-resistant cells.29


We have assessed the impact of a novel therapeutic combination that uses cilengitide, an inhibitor of α-v series integrins, in combination with external beam RT in a human orthotopic xenograft brain tumor model in the rat. Untreated tumors in this model cause mortality within 60 days, RT doubles the median survival to around 100 days, and as we describe here, cilengitide in combination with RT prolongs asymptomatic survival to over 300 days. Enhanced survival was accompanied by a rapid regression of the tumors following combined RT-cilengitide therapy, as observed in the living animals by both MRI and immunohistology. This effect was first observed on a schedule of daily treatment with cilengitide, and we presumed that it was due to inhibition of tumor angiogenesis. However, we subsequently discovered that cilengitide also had a similar synergistic effect even when given once within 4–12 hr before RT. It appears that very brief inhibition of αv integrins can greatly enhance the therapeutic effects of RT on orthotopic brain tumors, with enhanced apoptotic cell death, reduction of phospho-Akt, and short-term induction of genes that interact with the radiation damage and survival pathways, providing an initial suggestion of the possible mechanisms that cause the effect. This reveals a previously unexpected aspect of αv integrin inhibitors as amplifiers of radiotherapy.

Ligation of αv integrins, and especially αvβ3, can protect cells from apoptosis, and support growth-factor driven proliferation. Continuous inhibition of αv integrins in vivo can dramatically affect the response of heterotopic tumor xenografts to RT, involving the antiangiogenic activity well known for such inhibitors.11, 30 Our data support and extend this work and show that it also applies in an orthotopic intracerebral vascular bed. As ECM and vascular beds differ greatly between tissues and modulate tumor response to inhibitors,31 this is interesting. However, we were surprised by the unexpectedly short exposure to cilengitide that was needed for response in our model.

We measured the PKs of cilengitide in rats, and found that it had a short half-life and, although a substantial Cmax was reached. By the time synergy with RT maximized at 4 hr, the systemic concentration was some 20 nM, and by 8 hr, when synergy was still strong, we calculated the systemic levels to be below 1 nM. Such concentrations have little effect on cell attachment. Indeed, in in vitro models micromolar quantities are needed to produce effects on endothelial behavior.19 Because the IC50 of cilengitide for receptor inhibition in isolated αvβ3 and αvβ5 ligand binding assays is in the low nanomolar range,19 it is conceivable that signal pathways initiated at these integrins might be modulated at such low concentrations. Nevertheless, the synergistic effect disappears after 12 hr, in vivo, which suggests a model where cilengitide interaction with integrins triggers a response that sensitizes cells in the tumor environment to RT, and which decays within a time scale of 12 hr.

What possible mechanisms may be operating over this time scale? Both U251 cells and tumor invasive endothelia express αvβ3 and αvβ5 integrins. Their ligands, including osteopontin and vitronectin, are present in gliomas. As discussed, the ligation and inhibition of these integrins can regulate endothelial and glioblastoma attachment, migration and survival.19 However, in the histologically complex ECM of a tumor, it would seem that any therapeutic effect on αv-integrins should be negated by the multiple other integrins, including those insensitive to cilengitide, that might be engaged in this environment. Although such morphogenetic effects as blockade of migration and attachment may be relevant during continuous exposure to such drugs, the very short exposure that acts synergistically with RT suggests that other mechanisms must be operating in this system.

We gained some insight into such mechanisms by examining changes in gene expression that occurred on the time scale when cilengitide sensitized the tumor compartment to RT. These experiments suggested that genes affecting pro-apoptotic pathways involving RelA-NFκ-B were induced, whereas antiapoptotic genes like PI3K and pAkt were suppressed in the relevant time scale in situ by cilengitide. Analysis of protein phosphorylation supported this view. Post-translational modification of RelA regulates its ability to function as either an activator or repressor of gene expression. Of the 7 reported putative sites of RelA phosphorylation, 5 are thought to activate RelA transcriptional activity, either by enhancing binding to coactivator proteins or increasing RelA nuclear localization and stability.32 Further modulation of p65 RelA activity occurs via another gene, ING4, through physical interaction. Expression of ING4 is significantly reduced in gliomas as compared with normal human brain tissue, and the extent of reduction correlates with the progression from lower to higher grades of tumors. ING4 has been purported to regulate brain tumor angiogenesis through transcriptional repression of NFκ-B-responsive genes33via p65/RelA. This subunit of nuclear factor NFκ-B also inhibits p53 transcription activity. A novel NF-κB pathway involving IKKβ and p65/RelA Ser-536 phosphorylation results in p53 inhibition in the absence of NF-κB transcription activity.34 Integrin ligands modulate tumor and endothelial cell motility and protease secretion, which may involve NFκ-B-mediated signaling35 and a NFκ-B survival pathway.36

Ionizing radiation triggers free radical formation and induces single and double strand DNA breaks. These induce G2 cell cycle arrest and apoptosis.37 Cilengitide blocks the interaction between αv and the ECM, a block that in the long term can modulate the MAPK pathway, induce G1 cell cycle arrest, suppress Akt activity, as we have shown, and stimulate apoptosis. But in rat, cilengitide has a plasma half-life of only 20 min, and we have no reason to expect elevated drug concentrations in our model to account for these effects. These showed that cilengitide alone had little pro-apoptotic effect on the tumors alone, but that it greatly amplified the apoptosis induced by RT. This enhanced cell death might account for the corresponding lack of growth of the treated tumors, and extended asymptomatic survival of the animals.

Although the links are strong between the enhanced survival of glioma and endothelial cells and the production and activation of integrins of the α-v series, the PKs of cilengitide in rat suggest that cilengitide acts to amplify RT before RT occurs. In this, it differs from classical radiosensitizers, which need to be present during RT, and alter the biophysics of the radiation effects.38 As discussed, cilengitide likely triggers events modulating RT damage-response pathways. The optimal window for amplification occurs between 4 and 8 hr before RT, since 1 and 12 hr pre or 24 hr post-RT, cilengitide does not amplify RT efficacy. The systemic concentration of cilengitide when the efficacy window for RT opened, was below 20 nM (i.e., at the lower end of concentrations affecting overt cell attachment behavior). As yet, we cannot exclude that in the complex environment of the tumor, synergies between cellular compartments may amplify the effects of such low cilengitide concentrations, but, in summary, all available data suggest that the extent of αv integrin inhibition at and following synergistic RT is likely too low to have an effect at that moment.

We have also not formally excluded a selective enrichment of cilengitide within the tumor. As the tumor environment is enriched per αvβ3 and αvβ5, they might potentially act as an affinity sink for circulating material, and increase the concentration and distort drug clearance locally. The only relevant measurements in this context in the literature involve RGD-based imaging in heterotopic αvβ3-containing xenografts using reagents with similar inhibition kinetics to cilengitide.39, 40 They reported optimal enrichment of some 5-fold over plasma concentrations. If this were reflected in our model, it would still put the cilengitide concentrations below those that produce major cellular effects in vitro. During RT systemic concentrations in our experiments are near the IC50 for isolated αvβ319 and it is possible that partial receptor occupancy or selective enrichment in situ may indeed be generating active signals 4–8 hr after administration. Our current working hypothesis is that cilengitide induces a short-term signaling and/or transcription event, which interacts most strongly with DNA damage repair pathways between 4 and 8 hr of exposure, and decays to background within 12 hr, as supported by our gene array data. Other antiangiogenic agents enhance RT, however their scheduling and mechanism of action suggests that they are acting by a different route than cilengitide. Garkavtsev et al.41 found that anti-VEGF antibody DC101 given before RT, amplified RT by reducing tumor hypoxia. It had to be given 4 or 6 days before RT, and acted by “vascular normalization.” Oxygen enhances RT, when VEGF is blocked, vascular patency and perfusion recovers, and oxygenation improves. Might vascular normalization be responsible for the effects of cilengitide? The short-time interval between cilengitide and RT and its rapid clearance argue against this mechanism. Therefore, it appears unlikely that vascular normalization can be occurring in the brain after 4 hr.

In conclusion, we show here that the response of orthotopically implanted glioma to radiotherapy can be dramatically enhanced by cilengitide, a specific inhibitor of αv integrins. The ability of cilengitide to enhance subsequent tumor response and induction of death pathways when scheduled shortly before RT, shows an unexpected function of αv integrins and most likely of αvβ3 in this system—to support radioprotective intracellular responses—and suggests an interesting and, to date unexpected, therapeutic effect for such an inhibitor. Studies to further investigate the mechanisms involved are ongoing and will likely impact the clinical use of agents of this class.