64Cu-NODAGA-c(RGDyK) Is a Promising New Angiogenesis PET Tracer: Correlation between Tumor Uptake and Integrin α V β 3 Expression in Human Neuroendocrine Tumor Xenografts

Purpose. The purpose of this paper is to evaluate a new PET tracer 64Cu-NODAGA-c(RGDyK) for imaging of tumor angiogenesis using gene expression of angiogenesis markers as reference and to estimate radiation dosimetry for humans. Procedures. Nude mice with human neuroendocrine tumor xenografts (H727) were administered 64Cu-NODAGA-c(RGDyK) i.v. for study of biodistribution as well as for dynamic PET. Gene expression of angiogenesis markers integrin α V, integrin β 3, and VEGF-A were analyzed using QPCR and correlated to the tracer uptake in the tumors (%ID/g). From biodistribution data human radiation-absorbed doses were estimated using OLINDA/EXM. Results. Tumor uptake was 1.2%ID/g with strong correlations between gene expression and tracer uptake, for integrin α V  R = 0.76, integrin β 3  R = 0.75 and VEGF-A R = 0.81 (all P < 0.05). The whole body effective dose for humans was estimated to be 0.038 and 0.029 mSv/MBq for females and males, respectively, with highest absorbed dose in bladder wall. Conclusion. 64Cu-NODAGA-c(RGDyK) is a promising new angiogenesis PET tracer with potential for human use.


Introduction
Angiogenesis, the formation of new blood vessels, plays an important role in tumor growth, local invasiveness, and metastatic progression. Many preclinical studies and clinical trials confirm the importance of integrin α V β 3 in the process of tumor angiogenesis and metastasis [1,2]. Integrin α V β 3 is a cell adhesion molecule and is highly expressed on activated endothelial cells and tumor cells but is not expressed on resting endothelial cells and therefore specific for neoangiogenesis [3]. Several extracellular matrix (ECM) proteins like vitronectin, fibrinogen, and fibronectin interact with integrin α V β 3 via the amino acid sequence Arg-Gly-Asp (RGD) [4,5]. Vascular endothelial growth factor (VEGF-A) also plays an important role in both normal vascular tissue development and tumor neovascularization and is highly expressed in various human tumors [6,7].
In the last few years metal PET isotopes such as 68 Ga and 64 Cu have gained increased interest for labeling of peptides. Both 68 Ga and 64 Cu are excellent alternatives to 18 F. 68 Ga can be obtained from an in-house commercially available 68 Ge/ 68 Ga generator. The advantage of using cyclotron produced 64 Cu for PET imaging and radiotherapy is the longer halflife allowing imaging at late time points acquiring additional information [8]. A longer half-life also allows more nonspecific activity to be washed out. 64 Cu can now be produced in high yield and at high specific activity on a small biomedical cyclotron.
Neuroendocrine tumors (NET) express integrin α V β 3 at variable levels [19]. NET is one of the tumor types where antiangiogenesis treatment of patients seems promising [20] but patients need to be selected individually. To select the patients, PET scans with a tracer targeting tumor-angiogenesis may be extremely relevant. Therefore using a human NET xenograft model for preclinical studies to evaluate new PET tracers targeting tumor-angiogenesis is relevant.
The aim of the present study was therefore, as the first, to use and evaluate the PET tracer 64 Cu-NODAGA-c(RGDyK) using a human NET xenograft model. Biodistribution, PET/CT scans, correlation between gene expression of angiogenesis markers and tracer uptake, and radiation dosimetry extrapolation for humans were performed. Since investigation of tumor-angiogenesis using human xenografts in mice is challenged by the fact that the formation of new vessels in the tumors might be of either human or mice (host) origin, we designed QPCR assays to quantify both mice and human genes of Integrin α V , Integrin β 3 , and VEGF-A.

Stability Studies.
The stability of 64 Cu-NODAGAc(RGDyK) was investigated in either buffer or plasma from nude mice. For buffer stability, 160 MBq 64 Cu-NODAGAc(RGDyK) was incubated for 1, 2, 18, and 24 hours at room temperature in 450 μL ammonium acetate buffer. For the stability in plasma, 10 MBq 64 Cu-NODAGA-c(RGDyK) was incubated for 1, 2, 18, and 24 hours in 500 μL plasma. Plasma was mixed with equal amounts of acetonitrile after incubation. The resulting mixture was centrifugated and the supernatant was collected and filtered. The stability of 64 Cu-NODAGA-c(RGDyK) in plasma and buffer was analyzed and determined by the previously described HPLC conditions. Female NMRI (Naval Medical Research Institute) nude mice (6 weeks upon arrival) were acquired from Taconic Europe, Lille Skensved, Denmark. The mice were acclimated one week in the animal facilities before taken into experiments. At the age of 8 weeks (weight: 32.8 ± 4.2 g) the mice had H727 cells (5-7×10 6

Dosimetry Extrapolation for Humans.
For estimation of the human radiation dose using 64 Cu-NODAGA-c(RGDyK) the mouse biodistribution data were used. Assuming similar pharmacokinetics in mice and humans, the activity in organs at 1, 2, and 18 hours post injection (p.i.) was used to calculate single exponential effective half-lives and residence times for each organ. These data were then used for extrapolation to human dosimetry estimates using standard female (58 kg) and male (70 kg) human phantoms and the Organ Level INternal Dose Assessment/EXponential Modeling software (OLINDA/EXM; Vanderbilt University, Nashville, TN, USA) [21]. To ensure a conservative estimate, the urinary elimination fraction was set to 75% with voiding intervals of 5 hours. The effective dose for both adult male and female were then calculated using OLINDA/EXM.

Small Animal PET and CT
2.5.1. Imaging Experiments. A longitudinal dynamic study was performed scanning a mouse at 1, 2, and 18 hours after intravenous injection of 1.6 MBq 64 Cu-NODAGAc(RGDyK), allowing dynamic information of the tracer biodistribution in tumors as well as in the other organs of interest. The PET scans were performed using a small animal PET scanner (MicroPET Focus 120, Siemens Medical Solutions, Knoxville, TN, USA). The energy window for the emission PET scans was set to 350-650 keV and the time resolution was 6 ns. Scan time was 10 min.
The acquired data for emission scan was stored in listmode format and postprocessed to obtain 2 bytes 128 × 144 × 32 sinograms. Finally, the emission sinograms were reconstructed using MAP Algorithms and resulted into 4byte 256 × 256 × 95 image sets with a zoom factor of 1.443 and a voxel size of 0.87 × 0.87 × 0.79 mm 3 . Furthermore, the emission sinograms were corrected for dead time and decay time. Scatter and attenuation corrections were not applied to the emission data. The system was calibrated to provide activity concentrations as Bq/cc.
The small animal CT scans of the mouse parallel with the PET scans were acquired using a small animal computed tomography (microCAT II, Siemens Medical Solutions). The acquisition time of each CT scan was 6.5 minutes generating 360 projections at 360 • arc. The X-ray source settings were 75 kVp, 500 μA, and 270 ms for the voltage, the current, and the exposure time, respectively. The CT projections on the 3, 000 × 2, 970 CCD crystal were binned by 4 to increase sensitivity and reduce the dataset size. The projections were reconstructed by real-time reconstruction algorithm (the COBRA toolbox) using Shepp-Logan filter into 768 × 768 × 512 matrix with a voxel size of 0.092 × 0.092 × 0.092 mm 3 .

Image Analysis.
The fusion of the PET and CT images was performed manually using the ASIPro Toolbox (Siemens Medical Solutions). Also the analyses of PET images were performed in the ASIPro toolbox.
The measurement of the activity uptake in the tumors and the organs were estimated by manually outlining on CT images. The outlining of the organs on CT images in every slice resulted in 3D region of interest (ROI). The 3D ROIs of the organs were then applied on the corresponding PET images. The results of these ROIs acquired the concentration of the tracer in the organs expressed in percentage of the injected dose per gram tissue (%ID/g).

Quantitative Real-Time PCR.
Gene expression was quantified on the Mx3005P real-time PCR system from Stratagene. Human ITGAV, human ITGB3 and human VEGFA were tested in a triplex-, mice ITGAV and mice ITGB3 in a duplex, and mice VEGFA in a simplex-QPCR assay. The three human housekeeping genes HPRT, UBC, and RPLP were also measured in a triplex. The Brilliant QPCR Core Reagent Kit (Stratagene; cat# 600530) was used for all the multiplex assays. Assay optimization resulted in a 50% and 100% increase of Taq polymerase and dNTP in duplex and triplex assays, respectively. An MgCl 2 concentration of 5.5 mM was optimal in all designs. The Brilliant III Ultra-Fast QPCR Master Mix (Stratagene, Cat no. 600880) was used for simplex assay. NO-RT (no reverse transcription) for all the samples was tested using the housekeeping gene-triplex. The thermal profile for QPCR Core Reagent was: denaturation at 95 • C for 10 minutes, followed by 40 cycles with denaturation at 95 • C for 30 seconds and annealing/elongation at 60 • C for The samples were run in triplicates using 1 μL of cDNA in a total volume of 25 and 20 μL, respectively.
The optimal housekeeping genes for the H727 tumors were found by testing 12 human reference genes from TATAA Biocenter (Goteborg, SE) in 12 different H727 tumors from NMRI nude mice. With geNorm software [22] the genes and the number of genes giving the most stable endogenous normalization were found. The most stable housekeeping genes were HPRT, UBC, and RPLP. Normalization to these three housekeeping genes was executed for all assays.  Table 2.
QBasePlus software [23] based on the comparative method (2 −ΔΔCt ) [24] for relative quantification was used including normalization to three housekeeping genes. Furthermore an inter-run calibrator for correction between PCR runs within the same gene and an arbitrary overall calibrator for comparison of the gene levels in between mice and human genes were taking into account.

Statistics.
As shown in earlier studies [19] the PCR data were found not to be normal distributed when tested by a One-sample Kolmogorov-Smirnov test. All QPCR data were therefore log 10 transformed to obtain normal distribution and the log-transformed data were used for all the subsequent statistical analyses.
T-Test (two tailed) was used for comparison of mRNA levels of each gene-human versus mice-at the three different sample collecting times. Correlation between genes and %ID/g were evaluated using linear regression.
Data are presented as mean± SEM. All statistical analyses were performed using SPSS 17.0 statistical software (SPSS Inc., Chicago, Illinois, USA). P < 0.05 was considered significant.

QPCR Data.
In H727 tumors (removed 2 hours after tracer injection) integrin α V of both human and mice origin was expressed at the same level, whereas gene expression of h-VEGF-A was significantly higher than m-VEGF-A (P < 0.001). Integrin β 3 of both human and mice origin was expressed on a very low level ( Figure 2). Regression analyses for 64 Cu-NODAGA-c(RGDyK) tracer uptake, %ID/g (calculated from the biodistribution data) versus quantitative mRNA in the H727 tumors showed International Journal of Molecular Imaging 5   Table 2 shows the primers and probe sequences, the flourophores and quenchers used on the TaqMan probes and the amplicon length of the product for each gene.  significant correlation at 2 h p.i. with mice Integrin α V (R = 0.76, P < 0.05), mice Integrin β 3 (R = 0.75, P < 0.05), and mice VEGF-A (R = 0.81, P < 0.05), also for human Integrin α V at 2 h p.i. (R = 0.86, P < 0.01) (Figure 3).

Biodistribution Data.
Liver, kidneys, lung, spleen, heart, intestine, muscle, and blood were collected from 10 mice at the following time points: 1, 2, and 18 hours post injection. All tissues were γ-counted 120 sec. and %ID/g was calculated from the measured counts by taking counting efficiency of the γ-counter (F = 0.0943), half-life time for 64 Cu, decay from injection to counting time, and tissue weight into account. We found the highest %ID/g (mean ± SEM) at 1 h p.i. for the kidneys, intestines, liver and spleen (   times. Biodistribution data and tumor-to-organ ratios are shown in Figure 4.

Radiation Dosimetry.
Residence times based on biodistribution data are shown in Table 3 and were used for calculating estimates of human radiation-absorbed doses using OLINDA/EXM. Table 4 shows the OLINDA/EXM estimates for human adult females and males for each organ and the total effective dose. The highest radiation-absorbed doses were found to be the urinary bladder wall (0.014 and 0.010 mGy/MBq for females and males, resp.). However, this estimate is a worst-case scenario since we used a conservative model with a urinary elimination fraction of 75% with voiding intervals of 5 hours. The whole body effective dose was 0.038 and 0.029 mSv/MBq for females and males, respectively. Accordingly, an administered dose of 200 MBq 64 Cu-NODAGA-c(RGDyK) to humans would lead to a radiation burden of less than 8 mSv.

PET Scanning Data.
The tumors were clearly visible after injection of 64 Cu-NODAGA-c(RGDyK) at all measured time points: 1, 2 and 18 h pi ( Figure 5). 64 Cu-NODAGAc(RGDyK) showed relatively high uptake in H727 tumors; mean %ID/g of right and left tumors were 1.15, 0.99, and 0.76 respectively, at 1, 2, and 18 h pi. Tumor-to-muscle (T/M) ratios were 6.9 ± 0.8. Both %ID/g and T/M ratios were similar to the values found in the biodistribution study using the γ-counter. Tracer activity in organs did not disturb the tumor to background contrast from 2 h pi where most of the nonspecific uptake had been cleared.

Discussion
To the best of our knowledge this is the first study of the PET tracer 64 Cu-NODAGA-c(RGDyK). The major finding was a close correlation between tumor uptake of 64 Cu-NODAGAc(RGDyK) and the target Integrin α V β 3 validating its ability to image neo-angiogenesis. The cyclic RGD peptides c(RGDfK) and c(RDGyK) have frequently been used for radiolabeling. c(RGDfK), cyclo   is more hydrophilic than D-Phe and c(RGDyK) has therefore been reported as preferable due to faster renal excretion [10]. In our study we found a faster decrease of radioactivity in kidneys, intestine, and liver from 1 to 18 h (2.2 to 0.5, 1.6 to 0.4, and 1.1 to 0.4%ID/g) than that reported for 64 Cu-NODAGA-c(RGDfK) in the same time intervals (2.2 to 1.6, 2.1 to 1.3, and 1.7 to 1.0%ID/g) [12]. Taking together c(RGDyK) therefore seems superior to c(RGDfK). c(RGDyK), the peptide used in the present investigation, has previously been chelated to 1,4,7,10-tetraazacyclododecane-N,N ,N ,N -tetraacetic acid (DOTA) or 1,4,7triazacyclononane-1,4,7-triacetic acid (NOTA) and labeled with 64 Cu [11][12][13]25] or 68 Ga [14]. 64 Cu-DOTA-c(RGDyK) was first described in 2004 by Chen and coworkers [25]. The 64 Cu-DOTA complex seemed more unstable than 64 Cu-NOTA or 64 Cu-NODAGA, because Cu 2+ was transferred to copper-binding proteins, leading to relatively high blood and liver uptake [26]. NODAGA is, compared to the chelators DOTA and NOTA [8,11,[27][28][29], particularly useful due to high hydrophilicity leading to rapid renal excretion avoiding radioactivity accumulation in the gastrointestinal tract [30] and in the kidneys. Therefore, we chose NODAGA as chelator.
NODAGA-c(RGDy/fK) has been labeled with 68 Ga [9,12,15,18] and NODAGA-c(RGDfK) also with 64 Cu [12]. One paper compared 68 Ga-and 64 Cu-NODAGA-c(RGDfK) and found the highest tracer uptake using 68 Ga-NODAGAc(RGDfK) compared to 64 Cu-NODAGA-c(RGDfK) (5.2 versus 3.8%ID/g, 1 h p.i. in highly angiogenic U87MG xenograft tumors) but also 3 times higher tumor-to-blood ratio [12]. We found similar T/B ratio (12.7) after 1 hour using 64 Cu-NODAGA-c(RGDyK) as they did with 64 Cu-NODAGA-c(RGDfK). The advantage of 64 Cu labeling is the possibility of delayed imaging up to 12 or 18 h post injection with expected improved image contrast, and therefore improved visualization of low expressing Integrin α V β 3 tumors. Indeed, we found that delayed images after 2 and 18 hours showed higher tumor-to-background ratios compared to 1 hour images. This confirms the value of delayed images, possible with 64 Cu. In addition 64 Cu has a much better image quality than 68 Ga due to the shorter positron range. A comparison to 18 F-labeled RGD shows that tumor-to-liver and tumor-to-kidney ratios are similar [11].
For the neuroendocrine H727 xenograft tumors we found moderate levels of tracer uptake (1.2%ID/g 1 h p.i. as the highest) explained by the fact that neo angiogenic activity International Journal of Molecular Imaging in H727 tumors, reflecting low aggressiveness in general of NET tumors, is known to be lower than, for example, in glioblastoma U87MG xenograft tumors showing high expression of integrins [15].
We investigated the target specificity of the PET tracer by analyzing mRNA levels of both Integrin α V and Integrin β 3 using quantitative QPCR. Our results strongly support that 64 Cu-NODAGA-c(RGDyK) has high affinity to Integrin α V β 3 reflecting neo angiogenesis.
A challenge in preclinical studies using human xenografts in mice is, that the formation of new vessels in the tumors might be of both human and mouse origin. The interaction between mouse host stroma and human tumor cells and the role they play in the tumor-angiogenesis and tumor growth are of great importance in antiangiogenesis treatment. Therefore in preclinical studies using antiangiogenic drugs for treatment, it is of great importance to take into consideration species specificity. Accordingly, the antibody Bevacizumab only binds to human VEGF-A and not to murine VEGF-A [31], whereas the peptide Cilengitide is blocking Integrin α V β 3 of both human and mouse origin [32]. In the present study we therefore quantified both the gene expression level of mice and human Integrin α V , Integrin β 3 , and VEGF-A in our human H727 xenograft model. We found that Integrin α V and Integrin β 3 , of both mouse and human origin, are present at comparable levels. In contrast, VEGF-A is mainly of human origin. In general this seems to be true throughout tumor development as we previously (unpublished data) found dual origin from 1-3 weeks of tumor growth with a confirmed higher level of VEGF-A of human origin. Species specificity of our PCR designs were verified targeting human or mice gene sequences during blast tests, and as a further control QPCR primers targeting mice genes were tested in cDNA from a NET patient finding no gene expression.
The finding of correlation between tracer uptake of 64 Cu-NODAGA-c(RGDyK) and mRNA levels of mouse Integrin α V , Integrin β 3 , and VEGF-A two hours post injection indicates that 64 Cu-NODAGA-c(RGDyK) is indeed targeting the tumor-angiogenesis and therefore is a promising radiotracer for PET imaging and quantification of angiogenesis.
Taken together 64 Cu-NODAGA-c(RGDyK) seems promising for future use in humans. We therefore performed experiments to estimate dosimetry in humans based on projections from animal data. Here we found a favorable estimate of less than 8 mSv for a realistic clinical dose of 200 MBq. It should be noted that this estimate is conservative. Compared to recent dosimetry estimates for 68 Ga-NODAGA-c(RGDyK) [9] on a per MBq basis the radiation burden is comparable especially taking into consideration the more conservative urinary elimination model used in our calculations. In both studies the organ receiving the highest radiation-absorbed dose was the urinary bladder wall.

Conclusion
We, as the first, have described the use of 64 Cu-NODAGAc(RGDyK) as a PET tracer targeting angiogenesis. We found an excellent tumor-to-background ratio due to rapid elimination from nontumor tissue. Also an excellent image quality was obtained using 64 Cu. In addition, our compound seems stable as indicated by the low liver uptake. Using gene expression analyses we found a close correlation with key angiogenesis markers validating the target of the tracer. Finally the tracer had a favorable dosimetry. We therefore suggest that 64 Cu-NODAGA-c(RGDyK) should be further studied as a promising candidate for human angiogenesis PET.