The study protocol was approved by the local ethics committee for animal care and use. Experiments were performed in 10 adult (5-week-old) (SLC, Kotoh-cho, Japan) female nude mice, each weighing 17-20g. A human DLD-1 colon cancer cell suspension (1 × 10⁶ cells in 100 µL phosphate buffered saline) was injected subcutaneously into the right hind limbs of the mice. Tumors were allowed to grow for approximately 7 to 14 days, until reaching a predetermined size, the longest diameter of which ranged from 0.6 cm to 2 cm.

All MRI studies were performed using a 3T MR scanner (MAGNETOM Tim Trio, Siemens Healthcare, Erlangen, Germany) with a four-phased array wrist coil. Mice were placed prone in a plastic holder for ease of fixation and injection of contrast agent, and an inhaled anesthesia was used to avoid movement during imaging. Anesthesia was maintained with 2% isoflurane in a mixture of 100% 1 L/min oxygen. The tail vein was cannulated for intravenous injection of gadolinium contrast agent.

All images were obtained in the transverse plane using the following sequences: transverse T2-weighted turbo spin-echo (TSE) sequence (repetition time [TR] msec/echo time [TE] msec) 4930/128, echo train length of 25, one signal acquired, matrix of 114 × 192) applied with a section thickness of 1.0 mm, an intersection gap of 1.1 mm, and a flip angle of 160°. The field of view was 35 × 60 mm², which included the tumor in its entirety (20 sections) with a resultant voxel size of 0.3 × 0.3 × 1.0 mm³.

Dynamic contrast-enhanced MR imaging included two precontrast T1-weighted measurements (3D VIBE, TR/TE 8.3/2.3, 20 axial slices [slice thickness, 1.04 mm], field of view 50 × 50 mm, matrix 90 × 128) with different flip angles (2°, 15°) to determine the T1 relaxation time in the blood and tissue before the arrival of contrast agent on a pixel-by-pixel basis. This was followed by T1-measured dynamic contrast-enhanced series using a TWIST sequence (TR/TE 5.8/2.4, flip angle of 12°, and all other parameters the same as for the precontrast image).

Sixty dynamic measurements were performed on the entire volume of 20 sections. The acquisition time was 6.2 seconds with a single dynamic measurement. After the first five measurements, an intravenous bolus of gadopentetate dimeglumine (Magnevist; Schering, Erlangen, Germany) at a concentration of 0.05 mmol/mL was administered by manual injection at a dose of 0.3 mmol/kg over a maximum period of five seconds. The results of a previous pilot study showed that this dose of contrast agent provides maximal signal enhancement with the same sequence used in this study (data not shown).

The scanned data were transferred onto an image processing workstation (Leonardo, Siemens Healthcare, Erlangen, Germany) and analyzed using Tissue4D® software (Siemens Healthcare, Erlangen, Germany). A value for the arterial input function was automatically calculated by the software without an arterial region of interest (ROI) measurement procedure. To verify the applicability of the programmed arterial input function, the enhancement pattern of the normal thigh muscles opposite to the tumor was also analyzed. All mice showed similar patterns of enhancement in muscle ROI, with minimal variation between animals.

Tumor ROI was drawn at the midpoint along the z-axis of the mass by a third-year resident under the supervision of a senior radiologist. Two ROIs were drawn and measured for each mouse. ROI 1 covered the whole transection area of the tumor. ROI 2 covered only the hotspot area in the tumor. Hotspots in the MR images were selected to be matched with a hotspot in immunohistochemical staining. Dynamic data from ROIs 1 and 2 were fitted pixel by pixel to a pharmacokinetic model described by Tofts et al. (6), of which the quantitative parameters K^trans, K_ep, and V_e, were calculated.

After DCE-MRI is performed, the tumor was dissected from the surrounding tissue under ether anesthesia and sterile conditions. The mice were euthanized at the end of the surgical procedure in a CO₂ chamber. The dissected tumor was fixed with 10% formaldehyde solution and preserved at 4℃ for 24 hours. Then, 6 µm thick slide sections were made to correspond with the MR images. Immunohistochemical staining was also performed as follows: Deparaffinized slide sections were incubated with an anti-CD31 primary monoclonal antibody (Abcam, Cambridge, MA) followed by incubation with a biotinylated secondary antibody. The sections were then incubated with avidin-biotinylated peroxidase complex reagent and binding reactions were visualized using 3,3-diaminobenzidine (DAB). The stained microvessels were heterogeneous and the areas of highest neovascularization were mostly found at the periphery of the tumor. When scanned in a low-power field (100 ×), highly vascularized areas were considered to be hotspots. Three hotspots were examined respectively and vessels were manually counted in a high-power field (200 ×) with ocular grids. The grading and counting method described by Weidner et al. (14) was adapted for evaluation of MVD. Any brown-stained endothelial cell or endothelial cell cluster that was clearly separate from the adjacent microvessels, tumor cells, and other connective tissue elements was considered as a single countable microvessel. The sum of counts per field (total area = 0.578 mm²) represented a single hotspot. The highest vessel count among the three hotspots was used as the MVD. Hematoxylin and Eosin staining was performed in a consecutive slide to confirm to tumor necrosis. The area including the hot spots was scanned with low-power fields (2 ×, 40 ×). The outer tumor border and tumor necrosis was manually circumscribed by using Image J software (NIH, Bethesda, MD). Then, the total pixel area of the tumor and necrotic region was automatically calculated. Histological necrotic fraction (NF) was defined as the division of total pixel area of necrosis by the total pixel area within the tumor border.

The mice were divided into two subgroups according to their K^trans, K_ep, and V_e values. If a parameter was less than the median value, the mouse was assigned to the "low" group. And, if a parameter was above the median value, then the mouse was assigned to the "high" group. The Mann-Whitney test was performed to validate the differences in MVD between the low and high groups according to each parameter. The relationship between MVD, perfusion parameters and necrotic fraction of the tumor was examined by Spearman's correlation analysis. The correlation coefficient (τ) expresses the degree of change in the two corresponding variables. A coefficient closer to 1 indicates a positive correlation and a coefficient closer to -1 indicates a negative correlation. Statistical significance was accepted with a p value of less than 0.05. Statistical analyses were performed using computer software (PASW Statistics 18; SPSS Inc., Chicago, IL).