18F-labeled RGD peptide: initial evaluation for imaging brain tumor angiogenesis

Nuclear Medicine and Biology 31 (2004) 179 –189

18

www.elsevier.com/locate/nucmedbio

F-labeled RGD peptide: initial evaluation for imaging brain tumor angiogenesis

Xiaoyuan Chena, Ryan Parka, Anthony H. Shahiniana, Michel Tohmea, Vazgen Khankaldyyanb, Mohammed H. Bozorgzadeha, James R. Badinga, Rex Moatsb, Walter E. Laugb, Peter S. Contia,* a

PET Imaging Science Center, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA 90033 b Departments of Pediatrics, Radiology, and Pathology, Childrens Hospital Los Angeles, Los Angeles CA 90027, USA

Abstract Brain tumors are highly angiogenesis dependent. The cell adhesion receptor integrin ␣v␤3 is overexpressed in glioma and activated endothelial cells and plays an important role in brain tumor growth, spread and angiogenesis. Suitably labeled ␣v␤3-integrin antagonists may therefore be useful for imaging brain tumor associated angiogenesis. Cyclic RGD peptide c(RGDyK) was labeled with 18F via N-succinimidyl-4-[18F]fluorobenzoate through the side-chain ⑀-amino group of the lysine residue. The radiotracer was evaluated in vivo for its tumor targeting efficacy and pharmacokinetics in subcutaneously implanted U87MG and orthotopically implanted U251T glioblastoma nude mouse models by means of microPET, quantitative autoradiography and direct tissue sampling. The N-4-[18F]fluorobenzoyl-RGD ([18F]FBRGD) was produced in less than 2 h with 20-25% decay-corrected yields and specific activity of 230 GBq/␮mol at end of synthesis. The tracer showed very rapid blood clearance and both hepatobiliary and renal excretion. Tumor-to-muscle uptake ratio at 30 min was approximately 5 in the subcutaneous U87MG tumor model. MicroPET imaging with the orthotopic U251T brain tumor model revealed very high tumor-to-brain ratio, with virtually no uptake in the normal brain. Successful blocking of tumor uptake of [18F]FB-RGD in the presence of excess amount of c(RGDyK) revealed receptor specific activity accumulation. Hence, N-4-[18F]fluorobenzoyl labeled cyclic RGD peptide [18F]FB-RGD is a potential tracer for imaging ␣v␤3-integrin positive tumors in brain and other anatomic locations. © 2004 Elsevier Inc. All rights reserved. Keywords: MicroPET; Quantitative Autoradiography; Angiogenesis; Glioblastoma; RGD

1. Introduction Brain tumors are the second leading cause of cancer death in children under age 15 and in young adults up to age 34 [16]. Each year more than 17,000 people in the United States are diagnosed with brain tumor [3,8]. Malignant gliomas, the most common primary brain tumors, remain largely incurable despite intensive multimodality treatment including surgical resection, irradiation, and chemotherapy. Due to the failure of traditional cytotoxic therapies to improve the prognosis of malignant gliomas, investigators are actively exploring the use of novel cytostatic therapies against these tumors [23]. Adjuvant anti-angiogenic therapy against nonmalignant

* Corresponding author. Tel.: ⫹1-323-442-5940; fax: ⫹1-323-4423253. E-mail address: [email protected] (P. S. Conti). 0969-8051/04/$ – see front matter © 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.nucmedbio.2003.10.002

endothelial cells that form tumor vasculature is potentially a significant breakthrough in the treatment of cancers, including glioma [10,24]. Glioma tends to be highly vascular, a feature that is critical to their continued growth. In fact, radiologists have long exploited the increased vascularity of gliomas to localize and diagnose such tumors by angiography and contrasted enhanced studies which take advantage of vessel permeability. Vascular proliferation is an important marker in the histological grading of gliomas [7]. The degree of vascularization has been shown to correlate well with tumor grade and aggressiveness, presumably because tumors with a faster growth rate require an increased supply of nutrients and oxygen [22]. Imaging of tumor angiogenesis could play a major role in development and application of anti-angiogenic therapies, since quantitative imaging of angiogenesis can be designed to evaluate delivery, pharmacology and efficacy of novel anti-angiogenesis drugs. The ␣v-integrins (␣v␤3, ␣v␤5) mediate the contact of

180

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

activated endothelial cells to proteins of the extracellular matrix during tumor angiogenesis, which is a prerequisite for survival of endothelial cells. The ␣v-integrins are overexpressed on brain tumor cells and sprouting endothelial cells, but not on normal brain cells and quiescent endothelial cells. Thus the ␣v-integrins are attractive targets for antiangiogenic therapy [25]. Antibody, peptide and peptidomimetic ␣v␤3 integrin antagonists have been shown to impair angiogenesis, growth and metastasis of solid tumors [2,4,5]. Recently, Laug et al. [19,27] have demonstrated that EMD 121974 (c[RGDf(NMe)V]) suppresses orthotopic brain tumor growth but not heterotopic brain tumor implanted subcutaneously in nude mice, suggesting that the brain environment is a critical determinant of brain tumor susceptibility to growth inhibition by this pentapeptide. Based on these data, phase I/II studies have been initiated in adult brain tumor patients with favorable results (www. ClinicalTrials.gov). The ability to non-invasively image brain tumor growth and spread, visualize and quantify integrin ␣v␤3 and ␣v␤5 expression at different tumor growth stages and to monitor the effect of angiogenesis inhibitors would create a robust platform for development and evaluation of anti-angiogenic cancer therapy as a standard in oncology practice. Several radiolabeled ligands of the ␣v␤3 integrin adhesion receptor have recently been developed based on the integrin’s recognition of the RGD sequence of adhesive proteins. Haubner et al. labeled c(RGDyV) with 125I for tumor targeting in melanoma M21-bearing nude mice and osteosarcoma-bearing BALB/c mice [11]. This compound exhibited high affinity and selectivity in vitro and receptorspecific tumor accumulation in vivo. However, [125I]c(RGDyV) also displayed high accumulation of radioactivity in the liver and predominant hepatobiliary excretion. A glycosylated, RGD-containing peptide c[RGDyK(SAA)] with increased water solubility was found to have improved pharmacokinetics relative to the original radiotracer [12]. The same glycopeptide was also labeled with 18F via a 2-[18F]fluoropropionate moiety [13] for microPET imaging of ␣v␤3 integrin positive tumors. Direct electrophilic radiofluorination of c[RGDf(NMe)V] with [18F]AcOF yielded modest tumor uptake, probably due to the low specific activity (ca. 30 GBq/mmol) [21]. A dimeric RGD peptide, E-[c(RGDfK)]2, was conjugated with DOTA and HYNIC chelators, which enabled efficient labeling with 111 In/90Y and 99mTc, respectively. These tracers showed specific tumor uptake in an OVCAR-3 model. Treatment with a single injection of 90Y labeled peptide caused a significant delay in growth of small subcutaneous tumors [14,15]. Cyclic peptide c(RGDyK) was also conjugated to DTPA and labeled with 111In for targeting of pancreatic cancer CA20948 [30]. Polypeptide RGD-4C conjugated to HYNIC for 99mTc labeling showed modest tumor uptake in human renal carcinoma and colon cancer models [26]. We recently used direct tissue sampling and quantitative autoradiography with 125I-labeled c(RGDyK) peptide in a sub-

cutaneous U87MG glioblastoma model and found very high tumor uptake at early time points. High and persistent kidney uptake in this study presumably resulted from the overall positive charge of the tracer. Modification of the pentapeptide with a monomethoxy poly(ethylene glycol) moiety (mPEG, M.W. ⫽ 2,000) produced prolonged tumor accumulation and retention of the radiolabel, drastically decreased kidney uptake and slightly increased liver retention compared to the 125I-RGD peptide [6]. In the current study, we labeled the cyclic RGD peptide c(RGDyK) with 18F (t1/2 ⫽ 110 min) by conjugation with N-succinimidyl-4-[18F]fluorobenzoate through the lysine side chain ⑀-amino group. The radiotracer was applied to both subcutaneous U87MG and orthopic, intracerebral U251T glioblastoma models using microPET, autoradiography and direct tissue sampling techniques to study in vivo biodistribution of the radiotracer.

2. Materials and methods 2.1. General All reagents, unless otherwise specified, were of analytical grade and commercially available. Dimethylsulfoxide (DMSO) was distilled from barium oxide, dispensed under N2 into vials containing activated molecular sieves (4 Å) and stored at 4°C. Anhydrous acetonitrile (MeCN) from a newly opened flask was stored over activated molecular sieves (4 Å) at room temperature (r.t.). All other solvents were used without further purification. 4,7,13,16,21,24Hexaoxa-1,10-diazabicyclo-[8.8.8]hexacosan (K222) and potassium carbonates (K2CO3), as well as sodium phosphate buffer, were obtained from Sigma. Ethyl 4-(dimethylamino)benzoate, methyl trifluoromethane-sulfonate, O-(N-succinimidyl)-N,N,N’,N’-tetramethyluronium tetrafluoroborate (TSTU), and tetrapropylammonium hydroxide in water (1.0 M) were from Fluka. Radio-TLC was performed using silica gel 60 F254 plates (Merck), a Bioscan system 200 and Winscan software, version 2.2. Cyclic RGD peptide c(RGDyK) was synthetically produced via solution cyclization of fully protected linear pentapeptide H-GlyAsp(OtBu)-D-Tyr(OtBu)-Lys(Boc)-Arg(Pbf)-OH, followed by trifluoroacetic acid deprotection in the presence of the free radical scavenger triisopropylsilane [6]. As a reference standard, 19F-RGD containing the 4-[19F]fluorobenzoyl moiety was synthesized by reaction of c(RGDyK) with N-succinimidyl 4-fluorobenzoate (SFB). SFB was prepared from 4-fluorobenzoic acid, N-hydroxysuccinimide and 1-(3-dimethylaminopropyl)3-ethylcarbodiimide (EDC), in 1:1 molar ratio in a mixture of Na2HPO4 buffer (0.1 M, pH 9.0) and MeCN (ratio 1:1) for 2 h, and then the reaction was quenched by adjusting pH to 3 with acetic acid. The FB-RGD peptide was isolated by repeated HPLC preparations, using a Vydac 218TP54 column (5 ␮m, 250 ⫻ 4.6 mm). Elution was performed at 1 mL/min with a gradient of

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

MeCN starting from 95% solvent A (0.1% TFA in water) and 5% solvent B (0.1% TFA in MeCN) (0-2 min) to 35% solvent A and 65% solvent B at 32 min. The absorbance was monitored at 218 nm. Fractions containing the FB-RGD were pooled, freeze-dried and stored in a freezer until use. The conversion, based on RGD peptide and determined using HPLC, was 83%. The identity of FB-RGD was confirmed by matrix-assisted laser desorption ionization MS (MALDI Q-TOF) m/z ⫽ 740.2 [M⫹1]⫹; calculated Mr for C34H42N9FO9 ⫽ 739.76. FB-RGD showed a single sharp peak from RP-HPLC at retention time of 18.4 min, which was separated cleanly from RGD (10.2 min). No-carrier-added [18F]F- was produced via the 18O(p,n) [18]F reaction by bombardment of an isotopically enriched [18O]water target (95% enrichment, Isonics) with 11 MeV protons using a Siemens RDS-112 negative ion cyclotron. The [18F]F- was trapped on an anion exchange resin (Dowex (Cl-) 1X-8, 200-400 mesh, converted to OH- form and then eluted with K2CO3 (0.6 mL 10 mM, 6 ␮mol) and combined with K222 (10 mg, 26.6 ␮mol). The solvent was evaporated under a stream of argon at 90°C. Azeotropic drying was repeated twice with 1 mL portions of MeCN. 2.2. Synthesis of

18

F-labeled RGD peptide

The dried K222/K⫹ complex ([KC222]⫹/18F-) was resolubilized with a solution of 5 mg (15.6 ␮mol) of ethyl 4-trimethlammoniumbenzoate trifluoromethane-sulfonate in 1 mL anhydrous DMSO and reacted with stirring at 120°C for 20 min to produce ethyl 4-[18F]fluorobenzoate. RadioTLC (eluent CH2Cl2/EtOAc ⫽ 4:1) showed ⬎90% conversion yield, with the radioactive peak eluting with the same Rf as ethyl 4-fluorobenzoate (Rf ⬇ 0.84). The ethyl ester was subsequently hydrolyzed using NaOH (0.1 M, 1 mL) at 120°C for 7 min. After acidification with 0.7 mL 1 M HCl, the solution was diluted with 8 mL H2O and loaded onto an activated C18 Sep-Pak® column (Waters). The cartridge was then washed with 5 mL H2O and dried with a stream of argon. The fixed 4-[18F]fluorobenzoic acid was then eluted with 2 mL of MeCN. Radio-TLC revealed radiochemical purity of ⬎98% (Rf ⬇ 0.33). The [18F]fluorobenzoic acid was azeotropically dried with addition of tetrapropylammonium hydroxide (15 ␮L, 1 M in H2O). Subsequently, a solution of 12 mg of TSTU in 1 mL of MeCN was added and heated for 5 min at 90°C, followed by acidification with 3 mL 5% HOAc, and diluted with 6 mL water. The reaction mixture was then passed through C18 Sep-Pak®, and the excess amount of activation reagent was removed by washing the cartridge with 10 mL 10% MeCN. Finally, the N-succinimidyl 4-[18F]fluorobenzoate ([18F]SFB) was eluted from the cartridge with 3 mL of CH2Cl2. The solvent was removed by a stream of argon at 50°C. [18F]SFB was redissolved in MeCN (500 ␮L). The RGD peptide (100 ␮g) was mixed at 50°C for 40 min in 1 mL of Na2HPO4 buffer (pH 8.5) until most of SFB had reacted according to radio-TLC. Final purification was accom-

181

plished by C18 reversed-phase chromatography (detection mode: radioactivity and UV at wavelength at 254 nm). The HPLC fractions containing the activity were combined and evaporated with a stream of argon to remove MeCN. The pH of the resultant aqueous solution was adjusted to 7.0 with 0.1 N NaOH. This solution was passed through an activated Waters C18 Sep-Pak® cartridge, washed with water and eluted with 200 ␮L portions of 80% ethanol. The ethanol fractions were pooled and evaporated to a small volume. The activity was reconstituted in phosphate-buffered saline and passed through a 0.22 ␮m Millipore filter into a sterile multidose vial for use in animal experiments. 2.3. Tumor models Animal experiments were conducted under a protocol approved by the USC Institutional Animal Care and Use Committee. Female athymic nude mice (nu/nu) obtained from Harlan (Indianapolis, IN) at 4-6 weeks of age were injected subcutaneously (s.c.) in the right hind leg with 107 U87MG glioblastoma cells suspended in 200 ␮L Eagle’s minimum essential medium (EMEM). When the tumors reached 0.4-0.6 cm in diameter (10-14 d after implant), the mice were used for biodistribution and microPET imaging experiments. Details of the orthotopic (intracranial) xenotransplant model in nu/nu mice have been described previously [18]. U251T tumor cells (106/10 ␮L RPMI) were injected over 10 min into the forebrain 1.5 mm lateral and 0.5 mm anterior to the bregma, at a depth of 2.5 mm. Mice were kept under ketamine/xylazine anesthesia during the procedure. This tumor cell number resulted in the growth of tumors in all experimental animals and a highly reproducible growth rate. Six to seven weeks after injection when the tumors reached 5 mm or more in diameter, the mice were used for microPET imaging studies. 2.4. Biodistribution Nude mice bearing subcutaneously xenografted human glioblastoma U87MG were injected intravenously (i.v.) with approximately 370 kBq (10 ␮Ci) of [18F]F-RGD. Animals were euthanized at 30 min, 1 h, and 2 h postinjection (p.i.). Blood, tumor and the major organs and tissues were collected, wet-weighed, and counted in a ␥-counter (Packard). The percent injected dose per gram (%ID/g) was determined for each sample. For each mouse, radioactivity of the tissue samples was calibrated against a known aliquot of the injectate. Values are quoted as mean ⫾ standard deviation (SD) of a group of 5 animals. The receptor-mediated localization of the 18F-labeled RGD peptide was investigated by injection of [18F]FB-RGD with different amounts (between 5 mg/kg and 15 mg/kg) of c(RGDyK) on s.c. U87MG tumor models. Biodistribution was determined as described above at 1 h p.i. in 5 mice/group.

182

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

2.5. MicroPET images Positron emission tomography (PET) was performed on a microPET R4 rodent model scanner (Concorde Microsystems, Knoxville, TN). The scanner has a computer-controlled bed, 10.8 cm transaxial and 8 cm axial field of view (FOV). The camera has no septa and operates exclusively in 3D list mode. All raw data were first sorted into 3D sinograms, followed by Fourier rebinning and 2D filtered backprojection (FBP) reconstruction using a ramp filter with the Nyquist limit (0.5 cycles/voxel) as the cut-off frequency. Thirty min after injection of approximately 7.4 MBq (200 ␮Ci) of [18F]F-RGD, animals were anesthetized with ketamine/xylazine and placed near the center of the FOV of the microPET, where the highest image resolution and sensitivity are available. Static imaging was performed for 10 min. Circular regions of interest were placed at the locations of maximum tracer uptake in the tumor and in the corresponding contralateral site (reference region). Relative tracer uptake was expressed as the ratio between mean counts in tumor and the reference region (tumor:background ratio). For the s.c. model, the mice were sacrificed immediately after PET imaging and then subjected to autoradiography and subsequent direct tissue sampling. For the orthotopic U251 model, the mouse was imaged twice: (a) control experiment with injection of 7.4 MBq (200 ␮Ci) of [18F]FRGD. Two 10 min static images were obtained at 30-40 min and 60-70 min postinjection under ketamine/xylazine anethesia; (b) blocking experiment with injection of 7.4 MBq of [18F]F-RGD and 10 mg/kg of c(RGDyK) two days after the control experiment. A 10 min static image was obtained 30 min postinjection. The brain tumor was excised post mortem, fixed in buffered formalin, and embedded in paraffin for H&E (Hexatoxylin and Eosin) staining. 2.6. Whole-body autoradiography Autoradiography was performed using a Packard Cyclone Storage Phosphor Screen system (Downers Grove, IL) and a Bright 5030/WD/MR cryomicrotome (Hacker Instruments, Fairfield, NJ). Immediately after microPET scanning, mice were killed by cervical dislocation and then frozen in a dry ice and isopropyl alcohol bath for two minutes. The bodies were then embedded in a 4% carboxymethyl cellulose (CMC) (Aldrich, Milwaukee, WI) in water mixture using a stainless steel mold. The mold was placed in the dry ice and isopropyl alcohol bath for five minutes and then into a –20°C freezer for one hour. The walls of the mold were then removed, and the frozen block was mounted in the cryomicrotome. The block was then cut into 50 ␮M sections, and desired sections were digitally photographed and captured for autoradiography. The sections were transferred into a chilled film cassette containing a Super Resolution screen (spatial resolution 0.1 mm; Packard, Meriden, CT) and kept there overnight at -20°C. Screens were laser-

scanned with the Packard Cyclone. Images were calibrated by an internal standard method.

3. Results 3.1. Radiolabeling N-succinimidyl 4-[18F]fluorobenzoate ([18F]SFB) was synthesized by modification of literature procedures [17,29,31]. The decay-corrected yields of [18F]N-succinimidyl 4-[18F]fluorobenzoate ([18F]SFB) were 60-65%, based on starting [18F]F- and up to 75-80%, based on resolubilized [18F]F- (in five preparations) The mean synthesis time, including C18 Sep-Pak cartridge purification, was 65 min from end of bombardment (EOB). The SFB precursor was sufficiently pure to proceed with peptide labeling without HPLC isolation. Radio-TLC (eluent CH2Cl2/EtOAc ⫽ 4:1) showed only one radioactive peak eluting with same Rf as reference standard N-succinimidyl 4-[18F]fluorobenzoate (SFB) (Rf ⬇ 0.72). In the conjugation labeling of the RGD peptide, the conversion of [18F]SFB into the lysine ⑀-amino group was completed within 40 min at 50°C. Dissolving [18F]SFB in either MeCN or DMF helped to reduce hydrolysis of the active succinimidyl ester and to augment the solubility of [18F]SFB and its contact with the peptide. Four radioactive peaks were detectable, namely 4-[18F]fluorobenzoic acid (tRet ⬇ 14.3 min), partially hydrolyzed [18F]SFB (tRet ⬇ 15.5 min), [18F]FB-RGD peptide (tRet ⬇ 17.6 min), and a trace amount of unreacted [18F]SFB (tRet ⬇ 21.4 min). The decay-corrected yield of [18F]FB-RGD, counted from [18F]SFB, ranged from 35 to 45% (in five preparations). After purification, the mobile phase was subsequently formulated in physiologically-buffered saline and sterile-filtered. The radiochemical purity of the labeled peptide was higher than 99% according to analytical HPLC analysis. The specific radioactivity of the labeled peptide was determined based on the labeling agent [18F]SFB, since the unlabeled RGD peptide (tRet ⬇ 10.2 min) is efficiently separated from the product. The specific radioactivity of [18F]SFB was estimated by radio-HPLC to be 230 GBq/ ␮mol at the end of synthesis (EOS) (or 350 GBq/␮mol at end of bombardment (EOB)), based upon 9.25 GBq (250 mCi) starting [18F]F-. For microPET studies, the injected dose of 7.4 MBq (200 ␮Ci) contained approximately 30 ng of peptide. For illustration purpose, the structure of [18F]FB-RGD is presented in Fig. 1. 3.2. Biodistribution Figure 2 shows the biodistribution of the 18F-labeled RGD peptide ([18F]FB-RGD) in female nude mice with s.c. U87MG glioblastoma tumors. Data were obtained at 30 min, 1 h and 2 h p.i. [18F]FB-RGD had very rapid blood clearance, resulting in very low activity concentration in

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

183

Fig. 1. Schematic structure of [18F]FB-RGD.

blood and muscle as early as 30 min p.i. (0.52 ⫾ 0.13 and 0.57 ⫾ 0.17 %ID/g, respectively). The initial activity accumulation in the ␣v-integrin expressing U87MG tumor was ca. 3% ID/g at 30 min, and decreased to about 1.5 %ID/g at the 2 h time point. The tracer had very fast renal clearance. At 1 h, the highest and second highest radioactive concentrations were found in the urine and gallbladder. Low radioactivity concentration in bone indicated that the tracer is stable against defluorination in vivo. Tumor-to-blood and muscle ratios at 2h were high (21 and 8.2, respectively). Similar to 125I-labeled RGD peptide [6], [18F]FB-RGD demonstrated rapid blood clearance and fast tumor accumulation. As shown in Figure 3, uptake of [18F]FB-RGD in the tumor was somewhat lower than that of [125I]RGD (P ⬍ 0.01), and the washout rate was also more rapid (P ⬍ 0.01). Remarkably, the kidney uptake of 18F-labeled tracer was

significantly lower than that of the 125I-labeled RGD peptide (P ⬍ 0.001). The same phenomenon has been observed in the case of 125I-labeled RGD-mPEG conjugate, where the positive charge on lysine amino group was neutralized [6]. The results of blocking studies of [18F]FB-RGD using the subcutaneous U87MG model are shown in Figure 4. Coinjection of [18F]FB-RGD with c(RGDyK) resulted in a significant decrease of radioactivity in all dissected tissues, except for the liver. Uptake in the tumor was reduced most pronouncedly from 2.6 (%ID/g) (control) to 0.35 (co-injection with 5 mg/kg RGD) and 0.26 (co-injection with 15 mg/kg RGD). Tumor-to-blood ratio decreased from 18.5 ⫾ 4.2 (control) to 5.8 ⫾2.5 (co-injection with 5 mg/kg RGD) and 4.3 ⫾2.0 (co-injection with 15 mg/kg RGD). This dose-dependent blockade of uptake in the tumor and several normal organs suggests integrin receptor-mediated uptake in these tissues.

Fig. 2. Biodistribution of [18F]FB-RGD in nude mice bearing subcutaneously xenotransplanted U87MG tumors. The data are reported as %ID/g ⫾ SD (n ⫽ 5).

184

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

Fig. 3. Comparison of biodistribution data of [18F]FB-RGD (䊊) and 125I-RGD (䊐) in U87MG glioblastoma-bearing nude mice. Error bars denote SD (n ⫽ 5).

3.3. MicroPET and QAR of subcutaneous tumor model Static microPET imaging (10 min scan) 30-40 min after i.v. administration of 18F-RGD is illustrated in a coronal image of a subcutaneously implanted mouse in Fig. 5. Regions of interest (ROIs) were placed on tumors as well as other organs of interest. The tumor/muscle ratio was 4.8 ⫾ 0.3 (n ⫽ 3), in good agreement with the biodistribution experiment in a separate group of mice (5.5 ⫾ 0.4; n ⫽ 5). Tumors were visualized with clear contrast. PET imaging was further validated by whole-body autoradiography in the

same animals. Tumor/muscle ratio measured with autoradiography was 5.1 ⫾ 0.3. As expected, we did not observe any tumor uptake in the brain, suggesting that either the tracer does not penetrate the intact blood-brain-barrier (BBB) or the normal brain tissue does not express significant amount of ␣v-integrins. 3.4. MicroPET imaging of orthotopic model To determine whether the radiotracer was able to detect orthotopically implanted brain tumors, a nude mouse inoc-

Fig. 4. Activity accumulation in tumor and tumor-to-blood ratios after coinjection of [18F]FB-RGD with different amounts of c(RGDyK) in U87MG glioma bearing mice (n ⫽ 3). Data were determined 60 min postinjection.

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

185

Fig. 5. (i) MicroPET scan of mouse administered 200 ␮Ci of [18F]FB-RGD and scanned 30 min after injection (10 min single frame). The mouse was anethetized with ketamine/xylazine and positioned prone in the scanner. The mouse carried a U87MG glioblastoma xenograft on the right thigh. (ii) digital autoradiograph of the section containing tumor after microPET scanning, and (iii) an anatomic photograph of the section. (A: U87MG tumor, B: small intestines, C: liver, and D: gallbladder.)

ulated 2 months previously with U251T cells in the forebrain was imaged with microPET at 30 min and 60 min postinjection (Figure 6). The left side of the figure shows a normal nude mouse for comparison. Virtually no uptake of 18 F-RGD was seen in the normal brain, illustrating impenetrability of the blood-brain-barrier for the RGD peptide radiotracer. With the orthotopic model, we observed heterogeneous uptake inside the tumor at 30 min postinjection. Regions of interest drawn on the outer (red color), intermediate (yellow color), and central (green color) portions of the tumor image indicated uptake of 1.3, 1.9, and 1.6

%ID/g, respectively. At 1 h time point, some tumor washout was observed, but the uptake inside was more homogeneous, with outer portion uptake of 0.9, and intermediate and central portions uptake of 1.1 %ID/g, respectively. The tumor-to-brain ratio in the microPET scan was about 8:1 at both 30 min and 60 min time points. Histological examination of the brain tumor after microPET (Fig. 7) agreed well with the anatomical and functional information obtained from non-invasive imaging. Transverse sectioning of the tumor revealed maximum diameter of 7 mm, which was in accordance with the size obtained from counting the cross-

Fig. 6. MicroPET imaging of [18F]FB-RGD (tail vein injection of 200 ␮Ci activity) in nude mouse brain (ketamine/xylazine anesthesia and positioned prone). Sagittal slices are shown for (A) a normal mouse with intact blood-brain barrier (30-40 min post injection); (B) a mouse with a 2 month old U251T gliobastoma originally inoculated into the forebrain (30-40 min post injection); and (C) same mouse as in (B) scanned 60-70 min post injection.

186

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

Fig. 7. Histopathology of orthotopically injected U251T brain tumor cells after microPET imaging. A: original magnification; B: ⫻ 10; C: ⫻ 20. Tumor size and density of tumor cells were in accordance with the volumetric and uptake information obtained from microPET scanning.

ing sagittal planes of the microPET image. Remarkably, no necrosis was found in this relatively large tumor. Histological sectioning also identified the outer portion (1 mm thickness) of the tumor to be less condensed than the ossified region. To determine whether the activity accumulation in orthotopic brain tumor xenografts is ␣v-integrin specific or simply due to non-specific diffusion, we performed a blocking experiment in which the mouse was pre-treated with 10 mg/kg of c(RGDyK) 10 min before injection of [18F]FBRGD. Averaged tumor-to-brain ratio decreased from 8 ⫾ 1 (control) to 2.5 ⫾ 0.8 at 30 min postinjection of the radiotracer, demonstrating receptor dependent tumor uptake in orthotopic brain tumor as well (Figure 8).

4. Discussion Recently we examined the effect of the cyclic peptide antagonist pentapeptide EMD 121974, an antiangiogenic agent, on both subcutaneous and orthotopic brain tumor

growth. Daily administration of the ␣v antagonist inhibited the growth of established intracerebral U87MG glioblastoma and DAOY medulloblastoma tumors. However, the heterotopic tumors grown under the skin showed little or no effect after treatment with EMD 121974 at the dose tested for orthotopic models. This specific tumor response rate observed with the ␣v antagonist is in accord with recent data showing that ␣v knockout mice develop an abnormal vascular system only in the central nervous system and intestine but not in the remaining organs, including skin [1]. It is thus of critical importance to elucidate the mechanism why subcutaneous tumors are not responsive to EMD 121974 treatment and only orthotopic tumors smaller than 1.5 mm at the initiation of therapy responded to daily treatment. These findings may have an influence on the design of clinical trials. It is anticipated that more appropriate selection of patients entering clinical trials for anti-integrin therapy, as well as improved treatment monitoring of patients with brain tumors, can be achieved by applying positron emission tomography (PET) and suitable radiotracers to visualize and quantify ␣v-integrin expression. In this study,

Fig. 8. Coronal microPET images of orthotopic U251T tumor bearing nude mouse after 30 min post-administration of 200 ␮Ci of [18F]FB-RGD (10 min single frame, under ketamine/xylazine anethsia and positioned prone), with or without co-injection of blocking dose (200 ␮g of c(RGDyK)). Tumor-to-brain ratio was 8 ⫾ 1 on the control experiment (left) and was decreased to 2.5 ⫾ 0.8 on the blocking experiment (right).

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

we labeled cyclic RGD peptide c(RGDyK) with 18F through 4-[18F]fluorobenzoyl moiety and demonstrated the suitability of this [18F]FB-RGD tracer for non-invasive imaging of ␣v-integrin expression in brain tumor models using PET. Brain tumors are highly angiogenic and are dependent on neovascularization for their continued growth and invasion. The cell adhesion receptors ␣v␤3 and ␣v␤5 integrins are reportedly overexpressed in brain tumor cells and human brain tumor specimens [28]. Previously, we introduced 125Ilabeled cyclic pentapeptide c(RGDyK) for imaging of the integrin status in subcutaneously xenoplanted U87MG tumor models. Very rapid and high tumor uptake was observed at early time points. However, high renal uptake could not be avoided due to the overall positive charge of the labeled radiotracer. Modification of the peptide by covalent binding of monomethoxy poly(ethylene glycol) (mPEG, M.W. ⫽ 2,000) resulted in prolonged tumor accumulation and retention of radiolabel as well as a large reduction in kidney uptake. Presumably, this occurred because PEGylation neutralized the positive charge on the lysine side chain amino group, and because the increase in molecular weight caused a decrease in glomerular filtration of the peptide into renal cells [6]. Haubner et al. [12] modified the same peptide with a sugar moiety to increase the hydrophilicity. The resultant 125I-c(RGDyK(SAA)) had increased tumor uptake, slightly longer circulatory retention, somewhat decreased liver uptake and more rapid renal clearance, compared with 125I-labeled c(RGDyV). Modification of the lysine side-chain amino group either by the sugar moiety modification or mPEG conjugation has minimum effect on the targeting ability of the modified tracers. Although not thoroughly investigated at this time, we believe that labeling of the same peptide with 18F by modification at the ⑀-amino functionality of the lysine residue will have limited effect on the receptor binding affinity and specificity of the resulting PET tracer. Haubner et al. [12] labeled c(RGDfK(SAA)) with 18F using 4-nitrophenyl 2-[18F]fluoropropionate ([18F]NPFP) in the presence of potassium salt of 1-hydroxy-benzotriazole as activating agent. This tracer was applied to ␣v␤3 positive melanoma M21 and osteosarcoma bearing nude mice. This tracer revealed fast blood and predominant renal clearance. Compared with the 125I-labeled analog, 18F-labeled c(RGDfK(SAA)) [13] indicated rapid blood and predominant renal clearance. Decreased tumor uptake and more rapid tumor washout rates were also observed. MicroPET experiments demonstrated the ability to clearly distinguish tumors from the adjacent thorax wall. While this electrophilic substitution method is suitable for 18F-labeling of c(RGDfK) derivatives, it may not be applied to c(RGDyK), since both Lys and Tyr will react with [18F]NPFP [31]. Very recently, Ogawa et al. [21] reported direct electrophilic radiofluorination of cyclic pentapeptide EMD 121974, which showed very high affinity for ␣v␤3 in vitro [9]. Due to the extreme reactivity of acetyl hypofluorite, [18F]AcOF [20], multiple radioactive species as mono-fluorinated iso-

187

mers and di-fluorinated congeners were identified. Since [18F]F2 was produced with carrier added, the specific activity of the resulting radiofluorinated peptide was as low as ca. 30 GBq/mmol, which presumably contributed to low tumor uptake (less than 1% ID/g) in all species tested. Application of radiotracer at such a low specific activity in vivo may have resulted in self-inhibition of tumor uptake through saturation of the available receptor binding sites. Although no microPET data are available for this tracer, we would expect that 200 ␮Ci of the radiotracer (equivalent to 150-200 ␮g of fluorinated RGD peptide) will likely serve as a blocking dose for a mouse. MicroPET imaging is therefore expected to reveal only non-specific binding of the tracer. The unfavorable high liver and intestine uptake would further prevent the use of such an electrophilic substituted 18 F-RGD tracer for potential clinical application. In the current study, we chose to modify the RGD peptide with N-succinimidyl-4-[18F]fluorobenzoate. The preparation of the prosthetic labeling moiety was straightforward and could easily be automated. SFB labeled macromolecules are usually more stable than NPFP labeled compounds [31]. The radiolabeling of RGD peptide with this succinimidyl ester did not require extensive heating and the labeled peptide was easily separated by HPLC chromatography employing a Vydac protein and peptide column. Compared to the [18F]F-RGD(SAA) reported by Haubner et al. [13], this tracer had modestly improved in vivo kinetics. Faster blood clearance was observed compared to RGDSAA, with tumor washout also slightly reduced. Similar to RGD-SAA, the tumor uptake of [18F]FB-RGD was decreased compared to the 125I-labeled analog, which can be explained by the fact that the addition of bulky prosthetic labeling moieties altered to some extent the tertiary structure of the peptide and thus lowered the receptor binding affinity and/or changed the in vivo kinetics of the resulting radiotracers. PET imaging of brain tumors is currently based on metabolic radiotracers such as FDG. However, FDG is relatively non-specific and has limited usefulness for assessing anti-angiogenic brain tumor treatment strategies that are not based on traditional suppression of tumor cell proliferation. For example, anti-integrin therapy in orthotopic brain tumor models has been very successful in inhibiting small tumor growth and spread [27], and subsequent phase I/II clinical trials in human patients have produced favorable effects. There is thus an urgent need to image ␣v-integrin expression level in brain tumors for appropriate selection of patients to enter clinical trials, dosage optimization of integrin inhibitors and subsequent monitoring of anti-integrin treatment efficacy. We have demonstrated herein the feasibility of imaging brain tumors with the ␣v␤3-integrin antagonist [18F]FB-RGD. As seen from Figure 6, normal mice had virtually no uptake of [18F]FB-RGD in the brain. This may be tentatively explained as being due to the tight blood-brain barrier BBB, which prohibits penetration by [18F]FB-RGD. On the

188

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

other hand, we observed significant glioma uptake in the orthotopic model, where the tumor cell line was xenoplanted into the forebrain of nu/nu mice, the tight intercellular junctions that constitutes BBB having been perturbed by the intracranial inoculation to allow the RGD peptide to cross the “leaky” BBB. The very high tumor-to-brain contrast (ca. 8:1) may indicate overexpression of ␣v-inetgrins during brain tumor growth and invasion. At 30 min p.i., the image indicated heterogeneous uptake, with highest accumulation an in intermediate sphere of tumor where the malignant cell mass was concentrated. At 1 h, the uptake difference between center of the tumor and the intermediate sphere disappeared. Although the microPET imaging demonstrated the ability of [18F]FB-RGD to visualize brain tumor with high contrast, the correlation between ␣v-integrin expression and magnitude of uptake of [18F]FB-RGD by tumor was not examined here. In addition, specific localization of tracer to endothelial vs. tumor receptors could not be directly addressed in our studies. However, the successful blockade of uptake of the radiotracer in brain tumors with co-injection of excess amount of non-radiolabeled RGD peptide demonstrates specific integrin receptor localization within the tumor bed. Semi-quantitative analysis of ␣v-integrin receptor density by this blocking experiment also will need to be validated by in vitro receptor autoradiography. Despite the additional work still required, we anticipate that quantitative PET imaging of ␣v-integrin expression will be useful for imaging brain tumor growth, spread and angiogenesis, as well as evaluating the effects of anti-integrin therapy on tumor angiogenesis, regression and necrosis.

5. Conclusion 18

F-labeled cyclic RGD peptide c(RGDyK) via a prosthetic labeling group 4-[18F]fluorobenzoyl moiety revealed very high tumor to background ratios in both subcutaneous and orthotopic brain tumor models. The results demonstrated both renal and hepatobiliary excretion pathways and rapid blood clearance. Activity accumulation in the tumors reached an early maximum, but tumor-to-normal tissue contrast increased at later time points. This study also demonstrated that optimization of the tracer to obtain prolonged tumor uptake and improve in vivo kinetics is needed. We anticipate that non-invasive serial studies of ␣v-inetgrin expression and functional activity using PET will become an important tool to evaluate the role of ␣v-intergins during brain tumor angiogenesis, growth and spread, and to monitor the efficacy of anti-integrin therapy.

Acknowledgments This work was carried out with partial contributions from NIH grants P20 CA86532 (to PSC), R01 CA82989 (to

WEL), ACS grant ACS-IRG-580007-42 (to XC), and the Robert E. and May R. Wright Foundation (to XC). The USC cyclotron team, particularly Joseph Cook, is acknowledged for radionuclide production. References [1] Bader BL, Rayburn H, Crowley D, Hynes RO. Extensive vasculogenesis, angiogenesis, and organogenesis precede lethality in mice lacking all alpha v integrins. Cell 1998;95:507–19. [2] Belvisi L, Bernardi A, Checchia A, Manzoni K, Potenza D, Scolastico D, Castorina M, Cupelli A, Giannini G, Carminati P, Pisano C. Potent integrin antagonists from a small library of RGD-including cyclic pseudopeptides, Org Lett 2001;3:1001– 4. [3] Black PM. Brain tumor. Part 2. N Engl J Med 1991;324:1555– 64. [4] Brooks PC, Stromblad S, Klemke R, Sarkar FH, Cheresh DA. Antiintegrin ␣v␤3 blocks human breast cancer growth and angiogenesis in human skin. J Clin Investig 1995;96:1815–22. [5] Buerkle MA, Pahernik SA, Sutter A, Jonczyk A, Messmer K, Dellian M. Inhibition of the alpha-␯ integrins with a cyclic RGD peptide impairs angiogenesis, growth and metastasis of solid tumors in vivo. Br J Cancer 2002;86:788 –95. [6] Chen X, Park R, Shahinian AH, Bading JR, Conti PS. Pharmacokinetics and tumor retention of 125I-labeled RGD peptide are improved by PEGylation. Nucl Med Biol 2003;30:in press. [7] Daumas-Duport C. Histological grading of gliomas. Curr Opin Neurol Neurosurg 1992;5:924 –31. [8] Davis FG, Freels S, Grutsch J, Barlas S, Brem S. Survival rates in patients with primary malignant brain tumors stratified by patient age and tumor histological type: an analysis based on surveillance. Epidemiology, and End Results (SEER) data, 1973-1991. J Neurosurg 1998;88:1–10. [9] Dechantsreiter MA, Planker E, Mathae B, Lohof E, Hoelzemann G, Jonczyk A, Goodman SL, Kessler H. N-methylated cyclic RGD peptide as highly active and selective ␣v␤3 integrin antagonists. J Med Chem 1999;42:3033–340. [10] Folkman J. Angiogenesis and angiogenesis inhibition: an overview. EXS 1997;79:1– 8. [11] Haubner R, Wester H-J, Reuning U, Senekowitsch-Schmidtke R, Diefenbach B, Kessler H, Stocklin G, Schwaiger M. Radiolabeled ␣v␤3 integrin antagonists: a new class of tracers for tumor targeting. J Nucl Med 1999;40:1061–71. [12] Haubner R, Wester H-J, Burkhart F, Senekowitsch-Schmidtke R, Weber W, Goodman SL, Kessler H, Schwaiger M. Glycosylated RGD-containing peptides: tracer for tumor targeting and angiogenesis imaging with improved biokinetics. J Nucl Med 2001;42:326 –36. [13] Haubner R, Wester H-J, Weber WA, Mang C, Ziegler SI, Goodman SL, Senekowitsch-Schmidtke R, Kessler H, Schwaiger M. Noninvasive imaging of ␣v␤3 integrin expression using 18F-labeled RGDcontaining glycopeptide and positron emission tomography. Cancer Res 2001;61:1781–5. [14] Janssen ML, Oyen WJ, Dijkgraaf I, Massuger LF, Frielink C, Edwards DS, Rajopadhye M, Boonstra H, Corstens FH, Boerman OC. Tumor targeting with radiolabeled ␣v␤3 integrin binding peptides in a nude mouse model. Cancer Res 2002;62:6146 –51. [15] Janssen M, Oyen WJG, Massuger LFAG, Frielink C, Dijkgraaf I, Edwards DS, Radjopadhye M, Corstens FHM, Boerman OC. Comparison of a monomeric and dimeric radiolabeled RGD-peptide for tumor targeting. Cancer Biother Radiopharm 2002;17:641– 6. [16] Jemal AJ, Murray T, Samuels A, Ghafoor A, Ward E, Thun MJ. Cancer statistics, 2003. CA Cancer J Clin 2003;52:5–26. [17] Kilbourn MR, Dence CS, Welch MJ, Mathias CJ. Fluorine-18 labeling of proteins. J Nucl Med 1987;28:462–70. [18] MacDonald TJ, Tabrizi P, Shimada H, Zlokovic BV, Laug WE. Detection of brain tumor invasion and micrometastasis in vivo by

X. Chen et al. / Nuclear Medicine and Biology 31 (2004) 179 –189

[19]

[20]

[21]

[22] [23] [24]

expression of enhanced green fluorescent protein. Neurosurgery 1998;43:1437– 43. MacDonald TJ, Taga T, Shimada H, Tabrizi P, Zlokovic BV, Cheresh DA, Laug WE. Preferential susceptibility of brain tumors to the antiangiogenic effects of an ␣v-integrin antagonist. Neurosurgery 2001;48:151–7. Namavari M, Bishop A, Satyamurthy N, Bida G, Barrio JR. Regioselective radiofluorodestannylation with [18F]F2 and [18F]CH3COOF: a high yield synthesis of 6-[18F]fluoro-L-DOPA. Appl Radiat Isot 1992;43:989 –96. Ogawa M, Hantaro K, Oishi S, Kawasumi Y, Fujii N, Kawaguchi M, Doi R, Imamura M, Yamamoto M, Ajito K, Mukai T, Saji H, Ito K. Direct electrophilic radiofluorination of a cyclic RGD peptide for in vivo ␣v␤3 integrin related tumor imaging. Nucl Med Biol 2003;30: 1–9. Plate KH, Mennel HD. Vascular morphology and angiogenesis in glial tumors. Exp Toxicol Pathol 1995;47:89 –94. Puduvalli VK, Yung AW. New frontiers in therapy of malignant gliomas. Forum 1998;8:261–9. Puduvalli VK, Sawaya R. Antiangiogenesis – therapeutic strategies and clinical implications for brain tumors. J Neurooncol 2000;50: 189 –200.

189

[25] Rooprai HK, Vanmeter T, Panou C, Schnull S, Trillopazos G, Davies D, Pilkington GJ. The role of integrin receptors in aspects of glioma invasion in vitro. Int J Dev Neurosci 1999;17:613–23. [26] Su Z-F, Liu G, Gupta S, Zhu Z, Rusckowski M, Hnatowich DJ. In vitro and in vivo evaluation of a technetium-99m-labeled cyclic RGD peptide as a specific marker of ␣v␤3 integrin for tumor targeting. Bioconjugate Chem 2002;13:561–70. [27] Taga T, Suzuki A, Gonzalez-Gomez I, Gilles FH, Stins M, Shimada H, Barsky L, Weinberg KJ, Laug WE. ␣v-Integrin antagonist EMD 121974 induces apoptosis in brain tumor cells growing on vitronectin and tenascin. Int J Cancer 2002;98:690 –7. [28] Tonn J-C, Wunderlich S, Kerkau S, Klein CE, Roosen K. Invasive behavior of human gliomas is mediated by interindividually different integrin patterns. Anticancer Res 1998;18:2599 – 605. [29] Vaidyanathan G, Zalutsky MR. Labeling proteins with fluorine-18 using N-succinimidyl 4-[18F]fluorobenzoate. Nucl Med Biol 1992;19:275– 81. [30] Van Hagen PM, Breeman WAP, Bernard HF, Schaar M, Mooij CM, Srinivasan A, Schmidt MA, Krenning EP, de Jong M. Evaluation of a radiolabeled cyclic DTPA-RGD analogue for tumor imaging and radionuclide therapy. Int J Cancer 2000;90:186 –98. [31] Wester H-J, Hamacher K, Sto¨ cklin G. A comparative study of n.c.a. fluorine-18 labeling of proteins via acylation and photochemical conjugation. Nucl Med Biol 1996;23:365–72.