VEGF-PET Imaging Is a Noninvasive Biomarker Showing Differential Changes in the Tumor during Sunitinib Treatment

Angiogenesis, the formation of new blood vessels, is one of the hallmarks of carcinogenesis. VEGF and its corresponding receptors (VEGFR) on endothelial cells are important players in the regulation of angiogenesis, providing targets for antiangiogenic agents (1). When used either as single agents or combined with chemotherapy, antiangiogenic drugs have improved disease outcome in several tumor types. However, this benefit is modest and often of limited duration. Further, unexplained paradoxical effects have been observed. For example, the VEGFR tyrosine kinase inhibitor (TKI) sunitinib is used for the clinical treatment of metastatic renal cell cancers and gastrointestinal stromal tumors as it blocks angiogenesis in primary tumors (2), but its use may also lead to increased invasiveness at the tumor boundary and promotion of metastases (3, 4). This complex and dynamic interaction between tumor cells and their microenvironment may be an important reason why previous investigation of potential biomarkers has failed to predict response to antiangiogenic therapy. The search for biomarkers has been especially directed towards circulating markers and visualizing anatomic tumor changes during antiangiogenic treatment (4). The biological responses that occur in response to antiangiogenic treatment are presumably dynamic over time and likely to be heterogeneous within the tumor. Therefore molecular imaging, enabling visualization of biological processes, might provide a better insight into how tumors respond to antiangiogenic treatment with agents such as sunitinib.

Well known techniques for molecular imaging include measurement of tissue glucose uptake with ¹⁸F-FDG positron emission tomography (PET) to assess metabolism and measurement of tissue perfusion with ¹⁵O-water PET. A new option is direct imaging the molecules involved in the promotion or inhibition of VEGF signaling using PET analysis, which may be achieved following molecular radiolabeling. Therapeutic inhibitors of the VEGF pathway, such as the monoclonal antibody bevacizumab and the antibody derivative ranibizumab, which bind and block VEGF-A, have proven clinical antiangiogenic effectiveness and are attractive for this purpose (5, 6).

Previously, we developed ⁸⁹zirconium-labeled bevacizumab (⁸⁹Zr-bevacizumab) as a biomarker for the PET analysis of VEGF levels, with the aim of providing insight into the available target for VEGF-dependent antiangiogenic therapy and thus assist in tumor response prediction (6). This approach proved promising, as tumor uptake of ⁸⁹Zr-bevacizumab mediated by VEGF-A binding significantly higher than control antibody was demonstrated, providing a potential new tracer for noninvasive imaging of VEGF signaling in the microenvironment of the tumor. Further, the ⁸⁹Zr label, with its half life of 78 hours, proved valuable for antibody imaging allowing high-resolution PET over at least 24 hours. However, maximum uptake did not occur until 4 to 7days post injection (pi; ref. 6), likely due to the 21days serum half-life of bevacizumab (7). To gain more dynamic insight into tumor response during antiangiogenic treatment, we have since developed the PET tracer ⁸⁹Zr-ranibizumab for potential use as noninvasive biomarker of VEGF signaling. Ranibizumab, a monoclonal antibody fragment (Fab) derivative of bevacizumab, is used to treat macular degeneration (8). It has a higher affinity for all soluble and matrix bound human VEGF-A isoforms than bevacizumab (7). In addition, it allows fast and sequential follow up PET scans, as its serum half-life is only 2 to 6 hours.

In this study, we investigated the biological effects of sunitinib treatment and subsequent withdrawal in human xenograft tumor models, using tumor growth assays, immunohistochemistry, and PET analyses (¹⁸F-FDG, ¹⁵O-water, and ⁸⁹Zr-ranibizumab). A secondary aim was to determine the utility of ⁸⁹Zr-ranibizumab-PET analysis as a biomarker for antiangiogenic treatment.

The human ovarian tumor cell lines A2780 (provided by Dr TC Hamilton, Fox Chase Cancer Center). The human ovarian cancer cell line SKOV-3 and the human colon cancer cell line Colo205 were obtained from the ATCC. Cell lines were quarantined until screening for microbial contamination and mycoplasm were performed and proven to be negative. Meanwhile, a reproducible supply of cells was established by cryopreservation. All experiments were performed within a predefined number of passages. Key features of the cell lines were routinely checked. Growth and morphology of both cell lines was observed and noted to be consistent with prior descriptions of the lines; no further genetic characterization was performed. A2780 and Colo205 were cultured in RPMI-1640 (Invitrogen) with 10% heat-inactivated fetal calf serum (FCS; Bodinco BV) and 2 mmol/L of l-glutamine (Invitrogen) at 37°C in a humidified atmosphere containing 5% CO_2, SKOV-3 in Dulbecco's modified Eagles medium with 4.5 g/mL of glucose and 10% FCS. Cells were subcultured 3 times per week.

For in vitro experiments sunitinib (LC Laboratories) was dissolved in dimethyl sulfoxide (DMSO) at 40 mg/mL and stored at −80 °C. The MTT assay was used to determine cytotoxicity of sunitinib in A2780 and Colo205 cells (9). Cells were seeded (A2780 3,750 cells per well, Colo205 3,000 cells per well) in a 96-well plate in quadruplicate for each sunitinib concentration (0–20,000 nmol/L) and cultured for 4 days. No cytotoxicity occurred at sunitinib levels up to 5,000 nmol/L, which is a relevant plasma level in mice (Supplement 1).

¹⁸F-FDG was produced using the coincidence ¹⁸F-FDG synthesis module (10). Carrier-added [¹⁵O]O₂ was produced by irradiation of a mixture of nitrogen and 1% oxygen gas with 7MeV deuterons from a Scanditronix MC-17 cyclotron. [¹⁵O]O₂ was reacted with hydrogen gas at 400°C to generate ¹⁵O-water. The ¹⁵O-water was trapped in a sterivial with 2.5mL of 0.9% NaCl and sterilized by filtration (22-μm Millex GP filter).

Conjugation and labeling of ranibizumab (Lucentis, Novartis Pharma) and Fab-IgG were executed as described for U36 (6, 11). The chelate desferrioxamine B (Df; Novartis Pharma) was succinylated (N-sucDf), and coupled to lysine residues of ranibizumab by means of a tetrafluorophenol-N-sucDf ester. Conjugation was performed at room temperature for 30 minutes at pH 9.5 to 9.7. Hereafter, the mixture was adjusted to pH 4.2 to 4.4 (0.1 mol/L of H₂SO₄) and 50 μL of 25 mg/mL of ethylenediaminetetraacetic acid (EDTA; Calbiochem) was added to remove Fe(III). The resultant solution was then purified by ultrafiltration, diluted in water for injection (5 mg/mL) and stored at −20°C. Labeling was performed with ⁸⁹Zr produced by Cyclotron BV. ⁸⁹Zr-oxalate was adjusted to pH 3.9 to 4.2 and mixed for 3 minutes, then adjusted to pH 6.7 to 6.9 using HEPES buffer. N-sucDf-ranibizumab was added and incubated for 45 minutes at 20°C. ⁸⁹Zr-ranibizumab labeling (specific activity [SA] 1,500 MBq/mg) resulted in yields of more than 95%. 24 hours storage in 37°C serum displayed no measurable decrease in protein-bound radioactivity and revealed adequate VEGF-A binding, comparable to unlabeled ranibizumab using a VEGF-coated enzyme-linked immunosorbent assay as described previously for bevacizumab (Supplement 2; 6). Control ⁸⁹Zr-Fab-IgG is a humanized Fab-fragment comparable in size to ranibizumab. It was similarly prepared and showed no binding affinity toward VEGF-A.

Animal experiments were performed with isofluran inhalation anesthesia (induction 3%, maintenance 1.5%). Tumor cells for xenografting were harvested by trypsinization and resuspended in culture medium and Matrigel (BD Bioscience). In vivo imaging and ex vivo biodistribution experiments were conducted using male nude HSD athymic mice (Harlan). 6 to 8weeks old mice were injected subcutaneously with 1.0 × 10⁵ SKOV3, 5 × 10⁶ A2780, 5 × 10⁶ A2780^luc+, or 5 × 10⁶ Colo205 cells mixed with 0.1mL Matrigel™. In vivo studies were commenced when resulting tumors measured 6 to 8mm in diameter. All animal experiments were approved by the animal experiments committee of the University of Groningen.

All tracers were injected intravenously into the penile vein. ¹⁸F-FDG (5.0 ± 1.0 Mbq) microPET images (Focus 220 rodent scanner (CTI Siemens)) were obtained 1 hour pi. Animals were fasted for 12 hours before ¹⁸F-FDG administration. Ten minutes dynamic PET imaging was taken after ¹⁵O-water (78 ± 8.9 Mbq) administration followed by microCT imaging using a MicroCAT II (CTI Siemens) for anatomic localization. ⁸⁹Zr-ranibizumab (3.5 ± 1.5 Mbq) images were taken 0, 3, 6, and 24 hours pi. Static images of 30 minutes acquisition time were obtained each time.

Following image reconstruction, quantification was performed with AMIDE Medical Image Data Examiner software (version 0.9.1, Stanford University; ref. 12). To quantify radioactivity within the tumor, 3D volumes of interest (VOIs) were drawn. ¹⁸F-FDG is presented as standardized uptake value (SUV), using mean tumor uptake per cm³ divided by mean body uptake. ¹⁵O-water is presented as percentage uptake per cm³ relative to tumor uptake in the first frame (30 seconds) of the scan. For ⁸⁹Zr-ranibizumab, the total injected dose was calculated by decay correction of total activity present at 0 hour after injection in the animal. The data were quantified as percentage injected dose per gram (%ID/g), assuming a tissue density of 1. Data are presented as percentage uptake compared with baseline and corrected for tumor volume. Following sacrifice, organs and tissues were excised, rinsed for residual blood, and weighed. Samples and primed standards were counted for radioactivity in a well-type LKB-1282-Compu-gamma system (LKB Wallac) and corrected for physical decay. Harvested tumors were divided, immediately frozen at −80°C, and paraffin embedded for further analysis.

On the basis of available ⁸⁹Zr-bevacizumab data for comparison (5), we firstly used the SKOV-3 xenograft model to evaluate ⁸⁹Zr-ranibizumab characteristics. In 4 groups of animals, 4 protein doses [3, 8, and 40 μg (all n = 4) and 350 μg (n = 1)] ranibizumab labeled with a fixed amount (MBq) of ⁸⁹Zr were administered and biodistribution was determined 24-hours pi. Subsequently, 4 groups (n = 4) of mice were injected with 5 μg of ⁸⁹Zr-ranibizumab and 1, 3, 6, and 24 hours thereafter a group was sacrificed and ex vivo biodistribution performed. One group of mice (n = 4) was injected with 5 μg of ⁸⁹Zr-Fab-IgG. Ex vivo biodistribution followed 24-hours pi.

Sunitinib malate was dissolved in DMSO at 75 mg/mL. Before administration, sunitinib was diluted in phosphate-buffered saline (PBS; 140 of mmol/L of NaCl, 9 mmol/L of Na₂HPO₄, 1.3 mmol/L of NaH₂PO₄; pH = 7.4) and administrated intraperitoneally (ip) once daily at 60 mg/kg (6.5 mL/kg) or placebo (vehicle). Sixty milligram per kilogram of sunitinib daily previously demonstrated antitumor efficacy in xenograft-bearing mice and changes in tumor-derived human VEGF plasma levels with minimal toxicity (13). Four different treatment schedules with once daily ip treatment were used. ¹⁸F-FDG imaging was performed in A2780-bearing animals at baseline (n = 4), day7 (n = 4) of sunitinib treatment and following a drug-free week (n = 5). Serial ¹⁵O-water PET was carried out in A2780-bearing animals (n = 5) at baseline and following 7days of sunitinib. Serial ⁸⁹Zr-ranibizumab imaging (5 ± 1 μg) was performed in A2780-bearing mice treated with sunitinib for 1week followed by a drug-free week (n = 4), sunitinib ip for 2weeks (n = 4), sunitinib for 1week (n = 4), or placebo treatment (n = 4) and Colo205 tumor bearing mice with sunitinib for 1week followed by a drug-free week (n = 5) at baseline, day7 and day 14. To assess relative attribution of non-VEGF driven uptake, serial ⁸⁹Zr-Fab-IgG imaging was performed in A2780-bearing animals (n = 5) at baseline, day7 of sunitinib and following a drug-free week. Tumor volumes were assessed by external calibration along the longest axis (x-axis) and the axis perpendicular to the longest axis (y-axis). Tumor volume was calculated by the formula: (x × y²)/2.

Plasma VEGF levels of sacrificed A2780-bearing mice were determined with the human VEGF ELISA kit (R&D Systems). Paraffin-embedded tumors (n = 4 per treatment group) were stained with hematoxylin and eosin and antibodies against von Willebrand factor (vWf), Ki67 anti-human and anti-mouse (Dako), VEGFR2 (Cell Signaling), GLUT-1 (Chemicon), HIF1α (BD Biosciences), and human VEGF-A (sc-152, Santa Cruz). The mean vascular density (MVD) and VEGFR2 were scored in 3-defined hot spot areas that contained the maximum number of microvessels, as previously described (6). To assess changes in tumor vessel diameter, the shortest diameter of the largest vessels (≥5 per slide) were measured using computerized Aperio Image Software (Aperio Technologies Inc). Blood vessel proliferation was determined by calculating the percentage blood vessels with Ki67-positive endothelial cells. The tumor proliferation index was calculated by percentage of Ki67-positive cells in at least 4 high-power fields (200×). MVD, tumor vessel diameter, VEGFR2, blood vessel proliferation, and tumor proliferation were obtained for tumor rim (area < 500 μm of tumor margin) and center (area > 500 μm of tumor margin). GLUT-1 and HIF1α staining was scored manually as percentage positive of total tumor tissue per slide.

Data are presented as means ± SEM. Statistical analysis was performed using the Mann–Whitney U test for unpaired data and the Wilcoxon Matched Pairs test for paired data (SPSS, version 14). Tumor growth curves (trends) were analyzed using linear regression analyses. All tests were 2 sided and the significance level was taken as 0.05 or less.

Sunitinib treatment was evaluated in the A2780 and Colo205 model, which were chosen for their angiogenic profile and constant growth rate (2, 14). In A2780-bearing mice, tumor growth diminished after 7days of daily sunitinib treatment. Tumors in treated animals increased 123.7% ± 16.0% from baseline, compared with 182.0% ± 18.1% in nontreated animals (Fig. 1A). Thereafter sunitinib was withdrawn for 7days and tumor growth accelerated to 190% ± 10.8% on day 14. Tumor growth tended to be slower when sunitinib was continued, resulting in a tumor volume of 140.0% ± 16.1% at day14 compared with day0 (P = 0.067). In the Colo205 model, sunitinib induced significant tumor growth stabilization at day7 after daily sunitinib treatment. When sunitinib was withdrawn, tumors showed regrowth, from an average volume of 86.6% ± 12.9% on day7 to 132.9% ± 26.8% on day14 (P = 0.096), with a significant tumor growth curve from zero (P = 0.0006).

At baseline, ¹⁸F-FDG PET showed a homogeneous tumor uptake in A2780 tumor-bearing animals (Fig. 1B). Seven days of sunitinib treatment resulted in a 55% homogeneous decrease in ¹⁸F-FDG uptake compared with nontreated animals. When sunitinib treatment was stopped, ¹⁸F-FDG uptake on day14 was slightly higher than on day 7.

¹⁵O-water PET in A2780 tumors showed a 3- to 8-fold lower uptake at baseline than well perfused organs such as heart and kidneys, resulting in suboptimal tumor visualization (Fig. 1C). Further, ¹⁵O-water PET tumor uptake at baseline was 15.8% higher in the center compared with the tumor rim. By day7 of sunitinib treatment, tumor uptake tended to decrease in the center compared with baseline (−16.8% ± 7.7%), though not significantly (P = 0.11) and no difference between tumor rim and center was observed.

⁸⁹Zr-ranibizumab was first evaluated in the SKOV-3 model. Within 3 hours pi of ⁸⁹Zr-ranibizumab clear tumor visualization was seen (Fig. 2A), with a plateau at 24 hours (Fig. 2B). At 24 hours, ⁸⁹Zr-ranibizumab tumor uptake was 2.54-fold higher than ⁸⁹Zr-Fab-IgG (3.96% ± 1.00% ID/g vs. 1.56% ± 0.38% ID/g, P = 0.034), signifying VEGF-A specificity of ⁸⁹Zr-ranibizumab uptake. Increasing doses of unlabeled ranibizumab blocked tumor uptake of ⁸⁹Zr-ranibizumab to the uptake level of ⁸⁹Zr-Fab-IgG (Fig. 2B). Furthermore, organ biodistribution revealed rapid blood clearance of ⁸⁹Zr-ranibizumab from 8.44% ± 2.19% ID/g 1 hour pi to 0.38% ± 0.38% ID/g 24 hours pi resulting in tumor/blood ratios higher than 10 (Supplement3).

In the next experiments ⁸⁹Zr-ranibizumab quantification was performed 24 hours pi unless otherwise stated. At baseline, ⁸⁹Zr-ranibizumab uptake was 8.75% lower in the tumor rim than in the center (P = 0.05), comparable to findings seen with ¹⁵O-water PET. In A2780 placebo-treated mice, ⁸⁹Zr-ranibizumab (% ID/g) tumor uptake remained constant between baseline and 1week (Δ 8%, P = 0.229). However after 7days of sunitinib treatment in A2780 and Colo205 xenografts, there was a pronounced reduction in ⁸⁹Zr-ranibizumab tumor uptake in the tumor center but minimal effect in the tumor rim, whereas ¹⁸F-FDG decreased homogeneously (Fig. 2D). In A2780 tumors, ⁸⁹Zr-ranibizumab uptake decreased by only 19.5% ± 5.1% in the rim, whereas a decrease of 45.4% ± 5.9% was found in the center compared with baseline (Fig. 3B). Control experiments with ⁸⁹Zr-Fab-IgG, showed only a minimal decrease in uptake at day7 following sunitinib treatment in A2780 tumors, indicating minimal changes in nonspecific uptake of Fab fragments following sunitinib treatment. In contrast, decreased ⁸⁹Zr-ranibizumab uptake was 3.2-fold greater than control ⁸⁹Zr-Fab-IgG experiments (Fig. 3D). Interestingly, in A2780 tumors treated for 14 days, ⁸⁹Zr-ranibizumab uptake in the rim increased 46% at day14 compared with day7 (Fig. 3D), whereas in the center the uptake remained low compared with baseline.

PET scans performed 7days after stopping sunitinib treatment showed higher ⁸⁹Zr-ranibizumab tumor uptake in bothmodels, which exceeded baseline values. In A2780, ⁸⁹Zr-ranibizumab tumor rim uptake increased 69.5% ± 18.3% versus 7days sunitinib, and 34.6% ± 11.4% (P = 0.056) versus baseline (Fig. 3B). Likewise, uptake in the tumorcenter increased andreturned to baseline. In Colo205 tumors, ⁸⁹Zr-ranibizumab uptake in the tumor center exceeded baseline values (31.7% ± 9.9%; P = 0.033) 7daysafter discontinuation (Supplement 4 and 5). Control experiments using ⁸⁹Zr-Fab-IgG showed some enhancement in passive tumor uptake, but were 2.83-fold lower than with ⁸⁹Zr-ranibizumab.

Differential tumor response on ⁸⁹Zr-ranibizumab-PET scan between tumor rim and center corresponds with microscopic changes. Histologic examination also revealed a differential effect of sunitinib on the tumor rim and center. After 7days of sunitinib treatment, nonaffected blood vessels were mainly present in the tumor rim. Tumor tissue surrounding these vessels retained a high proliferation rate, and low HIF1α and GLUT-1 expression (Fig. 4A). Tumor tissue surrounding affected vessels in the center showed a low tumor proliferation rate, high HIF1α, and GLUT-1 expression. The center contained predominantly vital areas with some areas of necrosis (Supplement 7). Seven days of sunitinib treatment resulted in a reduction in all vascular markers and decreased tumor cell proliferation, which was most pronounced in the tumor center (Supplement 8). Interestingly, at day14 the tumor blood vessel diameter returned to baseline and increased tumor proliferation rate was observed in the tumor rim, which coincided with increased VEGF-A staining. Interestingly, no changes in vascular markers or tumor and endothelial proliferation were observed in the tumor center compared with day 7.

After a 7day drug-free period, all vascular makers recovered and tumor vessel diameter even exceeded baseline. In addition, HIF1α and GLUT-1 expression decreased whereas anincreased VEGF-A expression and tumor proliferation were observed (Fig. 4 and Fig. 5). Furthermore, large areas of randomly distributed multiple blood-filled regions (pelioses) were present at baseline and after discontinuation in parallel with high endothelial proliferation (Fig. 4D). Together these findings indicate rapid tumor revascularization after stopping sunitinib, which closely matches in vivo PET findings.

After 7days of sunitinib treatment, tumor derived human VEGF plasma levels decreased by 60% (Fig. 5B), whereas after 14days of sunitinib treatment there was a 6.61-fold increase compared with day 7. After the 7day drug-free period a 4.79-fold increase was observed.

This study is the first to investigate the effect of sunitinib treatment and its withdrawal on tumor biology using the molecular imaging techniques ⁸⁹Zr-ranibizumab-PET, ¹⁸F-FDG PET, and ¹⁵O-water PET in human xenograft tumor models. ⁸⁹Zr-ranibizumab-PET corresponded best with findings observed by tumor proliferation and vascularization assays, histology, and immunohistochemistry. Strikingly, sunitinib treatment caused a clear decline in ⁸⁹Zr-ranibizumab-PET signal in the tumor center, with only minimal reduction of uptake in the tumor rim, with a pronounced rebound after discontinuation. These findings demonstrate that sunitinib treatment effectively inhibits proliferation, neo-vascularization and secondly that ⁸⁹Zr-ranibizumab-PET is a valuable molecular imaging technique with utility as a biomarker forthe determination of intratumoral VEGF status and/or angiogenesis.

Currently MRI and PET imaging are used to visualize antiangiogenic treatment effects mainly by measuring changes in blood flow and permeability (15–17). Preclinically, MRI demonstrated decreased transfer constant K^trans following anti-VEGF treatment in a xenografts model, especially in the tumor center (18). High variability in MRI results during follow up of antiangiogenic treatment hampers the predictive value for individual patients (19, 20). PET imaging has a higher sensitivity compared with MRI and allows whole body imaging (21, 22). In this study, we have demonstrated the utility of ⁸⁹Zr-ranibizumab-PET as a valuable imaging biomarker of the VEGF pathway. Interestingly, if used in patients this PET tracer could also provide insight in changes in VEGF levels in normal organs during sunitinib treatment, which might influence distant invasion after treatment with sunitinib (3).

Rapid blood clearance of ⁸⁹Zr-ranibizumab and VEGF-driven tissue uptake resulted in high tumor to background ratios that enable more rapid tumor follow up with maximal uptake of ⁸⁹Zr-ranibizumab in the tumor within 24 hours pi compared with 4days for ⁸⁹Zr-bevacizumab. Relatively high kidney uptake of ⁸⁹Zr-ranibizumab, as seen with other metal labeled proteins, is a consequence of glomerular filtration of the tracer followed by tubular reabsorption and subsequent lysosomal degradation (23, 24). ⁸⁹Zr-ranibizumab binds with 20–100 fold higher affinity than bevacizumab to all human VEGF-A isoforms, including matrix bound isoforms (6). Most likely, ⁸⁹Zr-ranibizumab binds like bevacizumab in vivo to VEGF located on the cell surface and the extra cellular matrix as shown by Stollman et al. using ¹¹¹In-bevacizumab in VEGF₁₆₅ and VEGF₁₈₉ over expressing melanoma xenografts (25). Moreover, changes in ⁸⁹Zr-ranibizumab tumor uptake were 3-fold higher compared with control ⁸⁹Zr-Fab-IgG, indicating VEGF-mediated tumor uptake. One of our most important findings however, is that ⁸⁹Zr-ranibizumab tumor uptake can be followed dynamically within the tumor at different time points during and after sunitinib treatment.

⁸⁹Zr-ranibizumab does not bind to murine VEGF and therefore gives no information about the contribution of mouse VEGF to tumor angiogenesis in our xenografts models. However, our results corresponded highly with the angiogenic state of the tumors shown by other assays, indicating human VEGF to be largely responsible for the angiogenesis effects observed. There was a striking difference between tumor rim and center as uptake demonstrated with VEGF-PET. In concordance with reduced tumor growth, sunitinib treatment resulted in a decrease of angiogenic markers like microvessel density (MVD), as seen in other tumor models (2), vessel diameter, VEGFR2 expression. Further, reduced endothelial proliferation and the absence of pelioses showed true inhibition of angiogenesis. The inhibition of angiogenesis was also associated with high HIF1α and GLUT-1 expression and decreased Ki67 staining, indicating increased hypoxia and cellular stress and reduced tumor cell proliferation. All effects were more pronounced in the center than in the tumor rim, as highlighted by ⁸⁹Zr-ranibizumab-PET. The ⁸⁹Zr-ranibizumab-PET signal is the sum of perfusion of the tracer into the tumor followed by binding to VEGF, it is therefore a resultant of changed perfusion, MVD and VEGF expression, reflecting VEGF biodistribution and bioavailability and allowing in vivo insight in overall tumor angiogenesis.

VEGF-A staining was performed with an antibody that binds to human all VEGF-A isoforms and to a lesser extent to murine VEGF. The VEGF-A expression, was less intense in the center although we had expected an increase as a result of hypoxia caused by VEGFR inhibition. When we analyzed the plasma level of tumor derived human VEGF in these mice, we observed an initial decrease after 7days of sunitinib treatment, which matches the results assessed with ⁸⁹Zr-ranibizumab-PET and underlines the striking finding of initial drop in VEGF expression by tumor cells during sunitinib treatment that has not been reported before. A potential explanation for the initial decrease of VEGF could be an off target effect of sunitinib. Apart from blocking VEGFR signaling, sunitinib binds more than 70 other kinases (26). For instance, it inhibits STAT3, an activator of VEGF transcription, which could explain the initial VEGF reduction (27, 28). Ebos et al. measured already elevated human plasma VEGF levels after 1week sunitinib treatment in a PC-3 xenograft model (13). We however observed, just after 2weeks of sunitinib, an enhanced VEGF-PET signal in the tumor rim, which corresponded microscopically with more intense VEGF-A staining, a raise in blood vessel diameter, tumor cell proliferation, and raise in plasma VEGF levels. These results suggest transient antitumor effects as well as dynamic changes in the VEGF pathway resulting in tumor adaptation to sunitinib treatment. These data are in concordance with vessel normalization already 8days following initiation of the pan-VEGFR-TKI cediranib in a glioblastoma model (19, 29). These results demonstrate the need for noninvasive follow up in continuous as well as intermittent dosing schedules in the clinic to understand the effect of antiangiogenic treatment.

An explanation for the differences observed between the tumor rim and center might be that tumor cells located at the tumor rim are likely not only dependent on tumor blood vessels, but also benefit from adjacent peritumoral blood vessels. Indeed, we did not observe changes in peritumoral blood vessels during sunitinib (Supplement 6), whereas others even observed a large increase in peritumoral blood vessel diameter following cediranib in their glioblastoma model (19). Taken together, our findings emphasize that differences exist in response to antiangiogenic therapy among tumor areas. The interest for biological processes occurring at the edge of the tumor is also attracted by the phenomenon of epithelial mesenchymal transition (EMT) in tumors. EMT contributes to tumor progression and enhances the metastatic potential of tumor cells. It has been proposed that induction of hypoxia is one of the instigators (3, 30, 31). Indeed, sunitinib has showed to increase invasiveness at the tumor boundary and enhanced the forming of metastases (3). VEGF-PET enables noninvasive follow up of this process in the primary tumor as well as in all metastatic lesions.

The strong rebound in ⁸⁹Zr-ranibizumab-PET signal and rapid tumor regrowth after sunitinib discontinuation reflects what is seen in the clinic. Indeed, comparably to our findings preclinical immunohistochemical analyses have demonstrated rapid revascularization after stopping the VEGFR-TKI AG-013736 (32). Moreover, regrowth of lymph node metastases and flare up of clinical symptoms have been reported during the sunitinib-free period or rapidly after discontinuation (33, 34).

In our study, the clinically most frequently used tracer ¹⁸F-FDG did not reflect intratumoral differences or rapid tumor revascularization and regrowth after stopping sunitinib. Furthermore, the homogeneous reduction in ¹⁸F-FDG uptake did not correspond with the vitality seen by histology and immunohistochemistry. In addition, no differences between tumor rim and center were observed in this study with ¹⁵O-water after sunitinib despite the clinical feasibility of ¹⁵O-water PET to assess baseline tumor perfusion status (35). ¹⁵O-water PET performance is partly hampered by a relative low tumor perfusion compared with well perfused organs. For example, in patients 2-fold lower perfusion rates were found in primary renal cell cancers versus normal kidneys (36).

The newly developed VEGF tracer can be clinically used and translate preclinical findings. Recently the first clinical trials have been initiated to visualize VEGF. ¹¹¹In-bevacizumab visualized all known melanoma lesions in a feasibility study, even very small (1 cm) lesions (37). At this moment serial ⁸⁹Zr-bevacizumab VEGF-PET scans are performed to study the role of these scans during antiangiogenic treatment in renal cell carcinoma patients (38). When more rapid insight in the dynamic changes in VEGF response in the tumors is required at shorter intervals, ⁸⁹Zr-ranibizumab offers an additional investigative opportunity.

In conclusion, ⁸⁹Zr-ranibizumab-PET allows noninvasive dynamic and spatial in vivo visualization and quantification of VEGF signaling. ⁸⁹Zr-ranibizumab-PET therefore has potential use for preclinical and clinical follow up, guidance of new treatment strategies, optimal dose finding, and exploration of drug combinations, to increase the benefit of antiangiogenic therapies.

The authors declare that no conflict of interest exists.

We thank T.H. Oude Munnink, T. Jones, E.M.E. van Straten, and M. Green for technical assistance and helpful comments on the manuscript.

A personal grant to W.B. Nagengast and grants RUG 2007-3739 and RUG 2009-4273 from the Dutch Cancer Society.