The study was performed with the approval of the institutional committee for animal research in accordance with the U.S. National Institutes of Health guidelines for the care and use of laboratory animals. A total of 20 Copenhagen rats (7–8 weeks old, Harlan Laboratories) were given subcutaneous injections of 6 × 10⁶ MAT-Ly-Lu-B2 rat prostate carcinoma cells into the left abdominal flank. When tumors reached a volume of approximately 800 mm³ based on 3D caliper measurements (a × b × c × 0.5), rats were randomly assigned to either the therapy (n = 10) or the control (n = 10) group. The therapy group received 10 mg/kg body weight sorafenib (Nexavar, Bayer Schering Pharma) daily by gastric gavage through a curved buttoned cannula. The control group received daily volume-equivalent doses of the solvent solution (polyethoxylated castor oil [Cremophor EL, BASF] and ethanol) only. For the therapeutic solution, 20 mg of sorafenib was dissolved in 2.5 mL of a 1:1 solution of polyethoxylated castor oil and ethanol followed by 7.5 mL of distilled water.

DCE-MRI was performed before therapy (day 0) and on day 7 after therapy. For MRI examinations, the rats were anesthetized by intraperitoneal injection (100 mg/kg body weight ketamine plus 10 mg/kg body weight xylazine). MRI preparation also included a 25-gauge butterfly tail vein catheter for contrast administration. After completion of MRI on day 7, the rats were sacrificed by IV injection of a 1-mL overdose of the ketamine-xylazine combination. Tumors were harvested and fixed with formaldehyde (4%) for immunohistochemical analysis.

DCE-MRI was performed with a clinical 3-T system (Magnetom Verio, Siemens Healthcare) with the rat in supine position. An eight-element phased-array wrist coil with inner coil dimensions of approximately 100 × 80 × 60 mm (length × width × height) was used for signal reception. A short time before contrast injection, a fast view-sharing 3D gradient-recalled echo time-resolved angiography with stochastic trajectories sequence (TWIST, Siemens Healthcare) was started for fast bolus tracking and acquisition of the baseline and contrast passage with high temporal resolution and a total of 300 acquired 3D datasets (TR/TE, 6.34/2.11; flip angle, 40°; matrix size, 128 × 90 × 24; FOV, 50 × 50 × 72 mm; spatial resolution, 0.39 × 0.56 × 3.0 mm; receiver bandwidth, 180 Hz/pixel; acquisition time, 10:13 minutes for all 300 datasets). For acceleration, the actual slice resolution was set to 50% of the nominal value. Partial Fourier acquisition was performed in both phase and 3D directions with a factor of 6/8. View sharing was used with a central k-space region diameter of 17% and a peripheral k-space sampling density of 20%, thus a temporal resolution of 2 seconds per 3D dataset was achieved. The contrast agent was administered in a manual fast bolus injection of 0.1 mmol/kg body weight of gadobutrol (Gadovist, Bayer Schering Pharma) followed by a saline bolus. Gadobutrol, a small-molecule macrocyclic contrast agent, has a molecular mass of 0.6 kDa and is clinically available for contrast-enhanced MRI.

Data were postprocessed at an external workstation with PMI version 0.4 software (Platform for Research in Medical Imaging) written in house in IDL version 6.4 (ITT Visual Information Solutions) [20]. A blood region of interest was drawn in the lumen of the intrahepatic inferior vena cava to extract an arterial input function, and a tissue region of interest was placed over the tumor periphery (Fig. 1). The tumor periphery was defined as the outer 2-mm rim of the tumor, a region representative of viable tumor tissue and less affected by the elevated interstitial pressure and necrosis that can be present in the tumor center [21].

Signal intensity versus time curves were extracted for both regions of interest, and tracer concentrations were approximated according to relative signal enhancement (S/S₀ – 1, where S is signal intensity). The two-compartment exchange model was fitted to the data, producing four independent model parameters: tissue blood flow, tissue blood volume, transendothelial permeability-surface area product, and extravascular extracellular volume. Tissue blood flow, in milliliters of blood per minute per 100 mL of tissue, was a measure of the perfusion of the tumor. Tissue blood volume, in milliliters of blood per 100 mL of tissue, was a measure of the vascularity of the tumor. The product of transendo thelial permeability and surface area, in milliliters per minute per 100 mL, was a measure of capillary wall permeability to gadobutrol multiplied by capillary surface area. Extravascular extracellular volume, in milliliters per 100 mL, is a measure of the volume of the interstitial space.

The formaldehyde-fixed and paraffin-embedded tissue was stained for antiangiogenic (rat endothelial cell antigen 1), antiproliferative (Ki-67), and proapoptotic (terminal deoxynucleotidyl transferase–mediated nick end labeling [TUNEL]) tumor drug effects. To analyze the tumor tissue morphologically and to measure the necrotic areas, H and E staining was performed. In the first step, the tissue samples were dewaxed and rehydrated according to standard procedures: preheating at 60°C, washing in xylene substitute (Neo-Clear, Merck), and rehydration in a graded series of ethanol strengths (100%, 96%, 90%, and 70% ethanol) and double-distilled water.

Ki-67 antigen staining—A Ki-67–specific monoclonal rabbit antirat antibody was used to quantify tumor cell proliferation. The tissue was demasked in 0.1 M citrate buffer (pH 6.0) with microwave irradiation at 600 W. Permeabilization was achieved by immersing the slides into 0.25% Triton X-100–Tris Cl solution (Qiagen). After administration of the primary antibody (ab16667, 1:100) and overnight incubation, the secondary antibody (goat antimouse secondary antibody, 1:100; Vector) was applied. Signal enhancement was achieved with avidin-biotinylated horseradish peroxidase complex (ABC Kit, Vector). After addition of 3,3-diaminobenzidine, the samples were rinsed in hematoxylin and mounted with Kaiser glycerogelatin. The results were expressed as the average number of Ki-67–positive cells in 10 random fields at 200× magnification (cells per high-power field).

Rat endothelial cell antigen 1 staining—For demasking of the tissue, the slides were covered with pepsin (S3002, Dako), one package dissolved in 500 mL of 0.2 N HCl (1:5) followed by microwave irradiation. Nonspecific binding sites were blocked with 3% bovine serum albumin in phosphate-buffered saline solution succeeded by an avidin-biotin blockade. The primary antibody (ab9774, a monoclonal mouse antirat antibody, 1:40) was applied and incubated overnight. After endogenous peroxidase treatment with 3% H₂O₂, the secondary antibody (rabbit antimouse secondary antibody, 1:150, Vector) was applied. Signal enhancement was achieved by incubation with horseradish peroxidase complex (ABC Kit, Vector) followed by 3,3-diaminobenzidine and hematoxylin. To quantify microvascular density, microvessels were quantified as described in the literature [22]. Results were expressed as the average number of rat endothelial cell antigen 1–positive cells in 10 random fields at 200× magnification (cells per high-power field).

Terminal deoxynucleotidyl transferase–mediated nick end labeling staining—TUNEL staining was performed with a commercially available apoptosis detection kit (In Situ Cell Death Detection Kit, Roche Diagnostics) according to the manufacturer’s instructions. Immediately after staining, samples were analyzed with fluorescence microscopy with a standard fluorescent filter set at 520 ± 20 nm. Results were expressed as the average number of apoptotic cells in 10 random fields at 200× magnification (cells per high-power field).

Continuous variables were presented as mean ± SD. DCE-MRI and immunohistochemical values in the treatment and control groups were compared by unpaired Student t test. A paired Student t test was used for intragroup comparisons of the DCE-MRI perfusion parameters between day 0 and day 7. Analyses were performed with SPSS for Microsoft Windows software (version 11.5, SPSS). Relations between DCE-MRI and immunohistochemical findings were evaluated with Pearson correlation coefficients. Values of p < 0.05 were considered statistically significant.

Tumor perfusion—In the sorafenib therapy group, tumor perfusion measured by blood flow decreased significantly (p < 0.01) over the course of the experiment from baseline to day 7 (47.9 ± 36.8 to 24.4 ± 18.6 mL/100 mL/min). In the control group, a significant increase in tumor perfusion was observed (37.6 ± 12.3 to 49.8 ± 15.0 mL/100 mL/min; p < 0.05). Individual values are displayed in Table 1 and Figure 2, which depicts the unidirectional decline in tumor perfusion in the therapy group.

Permeability–surface area product—No significant changes (p > 0.05) were observed in tumor endothelial permeability–surface area product over the course of the experiment in either group (6.1 ± 4.1 to 4.7 ± 2.4 mL/100 mL/min and 4.6 ± 3.0 to 5.1 ± 5.1 mL/100 mL/min; p > 0.05). Figure 3 depicts the omnidirectional development of endothelial permeability in the treatment group over the course of the experiment with individual values partly increasing and decreasing under therapy.

Tumor vascularity—Tumor vascularity measured as blood volume decreased significantly (p < 0.05) under sorafenib treatment from day 0 to day 7 (15.6% ± 11.4% to 5.4% ± 2.1%). No significant change in tumor blood volume was observed in the control group between baseline and day 7 (12.9% ± 3.3% to 10.3% ± 7.3%).

Tumor vascularity, measured as the number of rat endothelial cell antigen 1–positive endothelial cells in tumor sections, was significantly lower in the therapy group than in the control group (74.4 ± 16.9 vs 197 ± 75.8; p < 0.05) (Fig. 4). The number of Ki-67–stained cells, a marker of proliferation, was significantly higher in the control group than in the therapy group (10,198 ± 3064 vs 15,003 ± 3674; p < 0.02). TUNEL staining, a marker of apoptosis, revealed significantly more apoptotic cells in the therapy than the control group (6923 ± 3761 vs 3167 ± 1500; p < 0.05). Individual values are displayed in Table 2.

Modest but significant inverse correlations were observed between tumor blood flow and immunohistochemical tumor cell apoptosis (TUNEL) (r = –0.56; p < 0.05) and between tumor blood flow and immunohistochemical tumor vascularity (rat endothelial cell antigen 1) (r = 0.56). No significant correlations were observed between noninvasive DCE-MRI values and immunohistochemical measurements of tumor cell proliferation and tumor cell apoptosis (Table 3).

Use of the two-compartment exchange model [27] yielded measurements of tumor perfusion that declined significantly during the 1-week course of treatment with sorafenib. Our results are in accordance with those of Padhani and Dzik-Jurasz [28]. After a review of the literature, those authors concluded that findings at DCE-MRI with small-molecule contrast media are predictive of tumor response and can be used to monitor the effects of antivascular and antiangiogenic drugs in experimental and clinical settings with various parameters of tumor microcirculation, including tumor perfusion. Padhani and Dzik-Jurasz, however, emphasized that owing to the great variety of DCE-MRI data acquisition (e.g., MRI sequences, mode of contrast administration) and analysis (e.g., kinetic models, quantification techniques), meaningful comparisons and conclusions are difficult between imaging centers.

If DCE-MRI is to enter into widespread clinical practice, effective cross-site standardization of measurements and evaluation has to be implemented. Still, reports on the change in tumor perfusion during antiangiogenic therapy remain controversial. Studies have shown a possible dependence of tumor perfusion on the tumor entity, tumor size, and therapy applied [29]. Our data revealed unidirectional development of the individual values of tumor perfusion in the therapy group and in our model may support tumor perfusion as a reliable parameter of therapeutic response. After reviewing the literature, we find it difficult to draw clear conclusions about changes in perfusion during antiangiogenic therapies because most groups used tracer kinetic models that do not separate tissue perfusion from endothelial permeability [26]. Nevertheless, it has become clear that changes in endothelial permeability can be assessed more sensitively if macromolecular contrast agents are used [30], and these agents are not yet available for clinical use.

The finding that tumor permeability–surface area product was not significantly influenced by sorafenib treatment in our prostate carcinoma model may be explained in part by the unselective extracellular distribution profile of the small-molecule contrast medium gadobutrol [31] and the associated reduced sensitivity to the changes in endothelial permeability induced by multi–tyrosine kinase inhibitors such as sorafenib. The finding of omnidirectional development of individual values of endothelial permeability, however, support the hypothesis that permeability–surface area product assayed with small-molecule contrast medium may be of only limited importance in angiogenesis imaging. Raatschen et al. [32], however, successfully used DCE-MRI with the macromolecular contrast medium prototype albumin-(gadopentetate dimeglumine)₃₀ to monitor the effects of the anti–vascular endothelial growth factor antibody bevacizumab on experimental human melanoma xenografts in rats. Endothelial permeability, quantified as the endothelial transfer coefficient K^PS, was identified as a sensitive predictive biomarker of tumor growth response to bevacizumab. This successful application was mainly attributed to the selective distribution profile of macromolecular contrast medium, which remained intravascular in normal tissues but passed through the abnormal endothelial barrier into the interstitial extravascular space in almost all types of cancer. However, prototypes of macromolecular contrast media such as albumin-(gadopentetate dimeglumine)₃₀ are not suitable for human use because of concerns about bioelimination and immunogenicity [33]. Pending approval of a macromolecular MRI contrast medium for human use, the call for clinically available noninvasive surrogate parameters of response to angiogenic treatment may be answered in part with DCE-MRI and small-molecule contrast media with full awareness of the methodologic limitations.

In the treatment group, tumor vascularity, quantified at DCE-MRI as tumor blood volume, was significantly reduced between baseline and day 7. In this context, blood volume has proved to be a robust and reliable parameter. This finding holds true particularly for the early treatment period, when only the vessel fraction regresses. However, as soon as there is a strong decrease in tumor cell mass, data must be interpreted with caution [29].

Immunohistochemical analysis of the tumors revealed significant antiangiogenic, proapoptotic, and antiproliferative effects of sorafenib on prostate carcinoma allografts. The multi–tyrosine kinase inhibitor sorafenib has been found to suppress the receptor tyrosine kinases of vascular endothelial growth factor receptor and platelet-derived growth factor receptor and the RAF serine and threonine kinases along the RAF–mitogen-activated protein–extracellular signal-regulated kinase pathway [15] with effects on an array of tumor entities [16, 17, 19]. Our results are in accordance with the findings of Chang and coworkers [14], who reported a significant reduction in tumor vascularity and a significant increase in tumor cell apoptosis in experimental renal cell carcinoma xenografts in rats after as little as 3 days of daily sorafenib treatment (15 mg/kg body weight). In an in vitro experiment, Delgado et al. [34] found a significant proapoptotic effect of sorafenib on Seg-1 human esophageal adenocarcinoma cells as early as 48 hours after treatment. Our results confirm significant anti-tumor effects of sorafenib on the prostate carcinoma allografts investigated.

In our study, correlations between noninvasive DCE-MRI parameters and immunohistochemical measurements were performed for validation purposes and revealed a modest but significant inverse correlation between MRI-assayed tumor perfusion and immunohistochemical tumor cell apoptosis. This correlation may be explained by reduced tumor perfusion in prenecrotic tumor areas with increased apoptosis [35]. A modest correlation was also observed between tumor perfusion and immunohistochemical tumor vascularity. Moreover, no significant correlations between noninvasive DCE-MRI parameters and microscopic quantifications were observed. The lack of correlation between histologic and functional parameters raises the question of the biologic significance of functional parameters of tumor microcirculation quantified with dynamic imaging enhanced with small-molecule contrast medium. Our results are in accordance with those of Atkin et al. [36], who reported poor correlation between results with DCE-MRI methods for assessing tissue angiogenesis with immunohistochemical markers of angiogenesis in 15 patients with rectal cancer. Those authors concluded that DCE-MRI findings do not reflect static histologic vascular properties in patients with rectal cancer. The variable results reported in previous correlative studies of DCE-MRI and histologic angiogenic markers may be rooted in a lack of standardization of DCE-MRI and immunohistochemical techniques, there being variations in MRI protocols and immunoassaying quantification methods.

Despite the broad acknowledgment of the role of DCE imaging in the preclinical assessment of antiangiogenic therapeutic effect in animal models, most clinical studies remain preliminary [1, 3, 16, 17, 19]. The lack of standardized and validated MRI protocols is among the primary reasons for delayed establishment of DCE imaging in clinical routine. Multitudes of imaging protocols and pharmacokinetic models vary from site to site and complicate the comparison of study results and the performance of comprehensive multicenter studies. With the clinical introduction of novel and expensive molecular targeted therapies, use of sensitive imaging biomarkers of early therapeutic response will further increase and may favorably be combined with functional DCE-MRI. Clinically, perfusion MRI has been found [37] to have potential in a multiparametric MRI approach to the diagnosis of prostate cancer by including not only anatomic but also functional imaging information, such as perfusion and diffusion, in diagnostic consideration. Our results support the hypothesis that with novel molecular targeted therapies available for prostate cancer treatment [38], findings at perfusion MRI performed with a multiparametric approach may be used as noninvasive imaging biomarkers of therapeutic response in early and sensitive monitoring of therapy.

The results of our study are limited in several respects. We investigated only one tumor-and-therapy combination in a rat model with possibly limited translational relevance to humans. For validation of the noninvasive MRI parameters, only a narrow immunohistochemical scope was investigated and correlated with limited success. Other immunohistochemical parameters may produce better correlations. However, the selected correlation parameters are accepted as representing central aspects of tumor metabolism. To investigate the reproducibility of kinetic parameters such as tumor perfusion, additional studies featuring repeated measurements will have to be performed.