Eighteen athymic male nude mice received subcutaneous injections of human cancer cells PC-3 on both flanks of the back in order to initiate tumors in each mouse (2 × 18 = 36 tumors). The PC-3 cancer cell line (American Type Culture Collection, Manassas, VA) was maintained as monolayers in E-MEM medium supplemented with 15% fetal bovine serum (FBS) at 37℃. Cells were harvested by trypsinization in trypsin/ethylenediaminetetraacetic acid (EDTA), washed in Hank's Balanced Salt Solution (HBSS) without Ca²⁺ and Mg²⁺, and pelleted by centrifugation at 150 × g for 5 minutes. Subsequently, re-suspended cells were counted in a hemocytometer using 0.4% trypan blue, and brought to a final concentration of 1 × 10⁶ cells/ml. Male, athymic nude mice (4-8 weeks old) were obtained from an animal facility (Orient, Seoul, Korea), housed under pathogen-free conditions, and maintained under controlled conditions (12-hour dark-light cycles; temperature, 20-24.8℃) with free access to sterilized mouse chow. Tumors were initiated by injection of approximately 1 × 10⁶ cells per site (50 µl). Injections on each back flank were at least 20 mm apart and as far from the lung and heart as possible in order to minimize motion effects on MR imaging (Fig. 1).

Initially, 36 tumors in 18 mice were imaged when the tumors reached 5-10 mm (shortest diameter), which was typically 2-4 weeks after implantation. In order to compare MR images and pathological specimens, 10 mice (20 tumors) were sacrificed for specimen preparation immediately after the initial MR imaging. The remaining eight mice (16 tumors) were sacrificed after the second MR imaging, eight days later. Next, tumors were excised and histologic analyses were performed.

MR images were acquired using a high field (7.0 Tesla, 20 cm horizontal bore magnet) small-animal MR scanner (Bruker BioSpin, Rheinstetten, Germany). Animals were anesthetized with a continuous supply of 2% isoflurane (EZ Anesthesia, Palmer, PA) in air, placed in the prone position in the animal bed, and inserted into the radiofrequency coil (38 mm i.d.) inside the magnet. Scout transverse images were acquired for correct positioning of the pelvic region of the mouse. Different MR sequences were obtained in order to optimize tumor visualization and signal contrast. Fast T1-weighted (Bruker RARE sequence; field of view [FOV] = 3 × 2.5 cm, matrix = 256 × 256, spatial resolution = 117 × 98 µm/pixel, slice thickness = 1.0 mm, repetition time [TR] = 1300 ms, TEeff = 9.0 ms) and T2-weighted sequences (Bruker RARE sequence; FOV = 3 × 2.5 cm, matrix = 256 × 256, spatial resolution = 117 × 98 µm/pixel, slice thickness = 1.0 mm, TR = 2500 ms, TEeff = 36.0 ms) provided the best contrast, and were used for morphological examinations. The total acquisition time, including induction of anesthesia, positioning, and set-up, was approximately 12 minutes per animal. After image acquisition, DW images were obtained using the Bruker SE-DWI protocol (FOV = 3 × 2.5 cm, matrix = 128 × 96, spatial resolution = 234 × 260 µm/pixel, slice thickness = 1.0 mm, TR = 2500 ms, TEeff = 26.0 ms, flip angle = 90°, excitations = 3). This protocol is based on the pulsed gradient spin echo (PGSE) sequence of Stejskal Tanner, with a series of four diffusion experiments/images. The diffusion gradient duration was 7 ms, while the diffusion separation time, D (big delta), was set to 14 ms. Three orthogonal gradient directions were applied in the DW MR. Corresponding b-values for the diffusion-sensitizing gradient were 0, 200, 500, 1000, and 2000 seconds/mm², respectively. ADC maps were constructed automatically on a pixel-by-pixel basis from the DW images for quantitative analysis. The ADC values were extracted from the experimental data from the indicated b-value ranging from 0 to 2000 s/mm² using a least squares mono exponential fitting. During each imaging session, animal respiration rates and core-body temperatures were monitored continuously, with the temperature maintained via a feedback system that provided warm air to the bore of the magnet. Typically, the respiration rate was maintained at 40/min, and the core-body temperature was maintained between 35℃ and 37℃.

The symbol '오' was used as an orientation marker in order to maintain proper orientation after tumor excision, where the upper round-shaped component 'ㅇ' of the symbol represents the cephalad direction, the mid vertical component 'ㅣ' of the symbol represents the mid vertical line of the tumor, and the lower horizontal line component 'ㅡ' represents the axial MR plane in the same orientation as the parallel line of the MR section at the most center level of the tumor. The orientation symbol was marked on skin overlying the tumor right after placement of the animal in the prone position in the animal bed inside the magnet. Although the skin marking is not visible on MRI, the horizontal line of the marking was considered closest to being parallel to the axial MR plane because the marking was made while holding the same MR position on the animal table inside the magnet. Animals were euthanized by CO₂ immediately after MR imaging. The gross findings were evaluated and recorded. Thereafter, we excised the tumor, including the overlying skin, where the orientation was marked. A total of 24 tumors were harvested within 24 hours after the last MRI. Histological slides were made at the NCC Histology Core Facility. The excised tissues were fixed in a large volume of 10% formalin for a minimum of one day to allow for complete tissue fixation. Histological examinations of all tissues were performed on paraffin-embedded sections. First, the paraffin block was bisected along the lower horizontal line component (ㅡ) on the overlying skin. Next, a serial section was made parallel to the sectional face. Subsequently, the block was sectioned along approximately the same plane as the axial MR images to permit the correlation calculation between the histological findings and MR images. Consecutive 5-µm sections were cut and mounted on glass slides. Sections were stained with Hematoxylin and Eosin (H&E). Expression of vascular endothelial growth factor (VEGF) was examined immunohistochemically with an anti-VEGF mouse monoclonal antibody (1:80, sc-7269; Santa Cruz Biotechnology, Fremont, CA) using the streptavidin-peroxidase technique with a SPTM kit (Zymed Laboratories, South San Francisco, CA). Polyclonal antibody (sc-1506r; Santa Cruz Biotechnology) against PECAM1 (CD31) was used for microvessel staining using the LSAB method (LSAB kit; Dako, Glostrup, Denmark). Negative controls were prepared by substitution of phosphate buffered saline for the primary antibody, and known positive controls were included for each staining procedure.

Tissue sections of the entire specimen were examined using a microscope at magnifications ranging from × 40 to × 400 by a board-certificated pathologist specially trained in genitourinary pathology. Features including tumor cellularity, angiogenesis, and degree of tumor necrosis, were analyzed for comparison with ADC values. For assessment of cellularity, four different areas of the slides (counterstained with H&E) were evaluated on × 400 high power fields (HPFs); in each case, the selection was made by a pathologist using an Olympus BX50 microscope (Olympus Co., Tokyo, Japan) and sections were photographed using a Spot Insight 2MP Firewire Color Mosaic Digital Microscope Camera (Diagnostic Instruments Inc., Sterling Heights, MI). Tumor cells included in each digital photograph were independently counted by two trained investigators and mean cellular number per one HPF was calculated in each case (8). The average score of the four areas was recorded. Microvessel density (MVD) was evaluated according to a previously described method (9). MVD within a designated hot spot (the most intensely neovascularized area at × 40 low power field) was scored on HPFs (× 200). Microvessels stained with CD-31 were counted on HPFs and the average score of the four areas was recorded. VEGF expression was semiquantitatively graded into four levels according to the percentage of positive tumor cells in the corresponding area. VEGF-positive tumor cells were not present in the examined tumor region for grade 0; VEGF-positive tumor cells were present in less than 25% for grade 1, between 25% and 50% for grade 2, and greater than 50% for grade 3 (10). For analysis of the degree of tumor necrosis, the pathologist marked necrotic areas on slides using a marking pen, according to a subjective decision of the pathologist. Thereafter, images of the tumors were photographed at a magnification of × 10 and imported into a Windows program that was developed with visual C++ using open source library CxImage, which performed calculation of percentage (%) of necrosis within the tumor (Fig. 1).

Apparent diffusion coefficient values were computed and analyzed for each region, as compared with corresponding regions on pathological specimens. A region-of-interest (ROI) that was minimally smaller than the actual solid portion of the lesion was carefully placed to ensure that cystic or necrotic areas were not included. Mean ADC values were then obtained after repeated measurement (3 times per tumor) (Fig. 1). Quantitative analysis of DW MR imaging, such as ADC measurement and drawing ROI, was performed by two experienced radiologists in consensus, who were unaware of the detailed pathologic findings. Three-dimensional volumetric measurements were used for assessment of tumor volume. In all contiguous axial images, tumor mass was outlined on the computer monitor using software (Rapidia 2.8, Infinitt, Seoul, Korea). The area of tumor in each slice was multiplied by the slice profile, and total tumor volume was automatically calculated by summation of the adjacent volumes.

Computed mean ADC values within solid portions of a mass on DW MR images were compared with pathological parameters, including tumor volume (measured on MR images), cellularity, MVD, VEGF expression, and tumor necrosis on corresponding sections of pathological specimens. Tumor necrosis on histological examination was defined as intratumoral necrosis that represents the fractional necrotic area in the entire section. For each tumor, the mean ADC value was plotted against the pathological parameters. The Pearson's correlation test was used for identification of correlation between pathological parameters and ADC values. A p value less than 0.05 was considered significant. All analyses were performed using Windows-based statistical software (STATA 10.0, Stata, College Station, TX; S-Plus 6.2, Insightful, Seattle, WA).