Clear cell renal cell carcinoma (ccRCC), the most common renal malignancy, is a highly heterogeneous neoplasm characterized by a robust tumor vasculature. The clinical importance of tumor heterogeneity in the classification and risk stratification of ccRCC has been recently recognized (1). Arterial Spin Labeled (ASL) MRI was used to explore the association between heterogeneous in vivo perfusion and the underlying histology and genomic profile to identify key genes linked to tumor angiogenesis.

Eight patients with renal masses (confirmed ccRCC) were imaged in the supine position with a 1.5T MR scanner with an 8-channel Torso coil (Excite HDx, GE Medical Systems, Waukesha, WI). All patients signed informed consent prior to enrollment in this prospective HIPAA compliant study. Axial and coronal T2-weighted images were acquired for anatomic reference. A single-slice coronal 2D ASL through the center of the mass was acquired with pseudo-continuous labeling in the upper abdominal aorta, optimized with background and vascular signal suppression (2). Offline reconstruction of ASL perfusion maps was completed using IDL (ITT Visual Information Systems). ASL images were analyzed on the open-source DICOM viewer Osirix X. A radiologist drew regions of interest (ROIs) of approximately 1 cm² within regions of the tumor that subjectively demonstrated the highest (high flow) and lowest (low flow) signal intensity, respectively, avoiding areas demonstrating signal intensity similar to that of the background (i.e. no perfusion). The mean ROI values represent blood flow levels in milliliters per minute per 100 g of tissue (mL/min/100g). Immediately after nephrectomy, the surgical specimen was positioned spatially to match its anatomic orientation on the ASL images with the help of fiducial markers placed during surgery. Fresh tumor samples collected from areas of high ASL measured perfusion were compared with samples with low ASL perfusion in the same tumor and samples from uninvolved renal parenchyma (URP). Microvessel density was assessed using CD34 immunohistochemistry (ICH) and ccRCC histology confirmed with CAIX ICH (Fig. 1). Genome-wide microarray profiling was performed on high and low perfusion regions of tumor as well as URP to characterize the underlying genomic profile in each region of sample. Survival analysis was performed for genes exhibiting expression correlation to ASL measures using over 400 ccRCC tumors data from The Cancer Genome Atlas (TCGA)(3).MRI was performed 19 +/-12 days prior to nephrectomy. All patients had ccRCC Fuhrman grade 2 or 3. There was no correlation between ASL perfusion and tumor grade (p>0.05). Average tumor size was 7.7±2.1 cm (range 3.9-11.7 cm). Average tumor blood flow in the high and low flow areas were 593±283 ml/100g/min and 191±108 ml/100g/min, respectively. Average blood flow in the renal cortex of the kidney harvesting the tumor was 421±164 ml/100g/min. The expression of a set of 52 genes was significantly correlated with ASL perfusion in different tumor regions. The expression correlation of top perfusion associated genes was validated using Q-RTPCR. Further pathway-based analysis depicted significant upregulation of multiple immune and inflammatory response pathways as well as hypoxia pathways in low perfusion compared to high perfusion regions of tumors. Ephrin-A5 (EFNA5) expression correlated with perfusion using both microarray (R2 = 0.504 and P = 0.002) and RT-PCR (R2 = 0.374 & P = 0.01) (Fig. 2) and exhibited highest significant differences between low and high perfusion regions across all samples (Fold Change = 2.88 and P < 0.02). Interestingly, EFNA5 has not been associated with ccRCC in previous genomics studies based on random tumor samples probably due to its heterogeneous perfusion associated expression. Higher expression of EFNA5 is associated with poor 3 and 5 years survival (P = 0.0009) (Fig. 3). EFNA5 depicted significant amplifications in about 10% ccRCCs and a pattern of low survival association (P < 0.06) on the basis of TCGA data.The integrated use of ASL MRI and genomic profiling offers a platform for correlating tumor pathophysiology in vivo with genomic heterogeneity. Using this novel methodology we report for the first time overexpression of EFNA5 as a signature of intratumoral vascular heterogeneity in ccRCC and confirm that EFNA5 overexpression correlates with patient survival.We propose MRI-based targeted tissue sampling to characterize the heterogeneous neoplastic states and associated complex genetic alterations driving angiogenesis in ccRCC.