Endothelial gene-expression analysis. (A) Interrelationship between ECs of different sources and their gene-expression signatures. ECs from a malignant and proangiogenic environment (TEC) are compared with ECs from organ-matched and patient-matched nonmalignant sources (NEC) and with nonmalignant proangiogenic microenvironment (placenta)–derived ECs (PLEC) to identify the subset of genes that showed expression induced by the tumor microenvironment specifically (tumor EC signature markers). (B) Venn diagram representation of subtraction repertoire screening. Four pairwise comparisons were performed by cDNA array screening of SSH repertoires: tumor-conditioned (HUVEC+) compared with quiescent HUVECs (HUVEC–), colorectal carcinoma ECs compared with normal colon ECs (TEC compared with NEC), colorectal carcinoma ECs compared with placental ECs (TEC compared with PLEC), and placental ECs compared with normal colon ECs (PLEC compared with NEC). Included are clones that showed at least a 2-fold difference in expression. The shaded area represents the TAGs (overexpressed in TECs compared with NECs and in TECs compared with PLECs). (C) TAG markers are strongly biased toward genes associated with extracellular matrix remodeling. (D) GAG/A markers (overexpressed in TECs and PLECs compared with NECs) show a diverse functional profile. (E) GAG/B markers (overexpressed in TECs and in activated HUVECs) are strongly biased to protein regulation. Functional distributions among the 3 different classes were significantly different (P < .001), as determined by χ2 test. TAG functional classification differed from that of GAG/A (P < .001) and GAG/B (P = .006), though GAG/A functional distribution did not differ significantly from that of GAG/B (P = .630). Classifications were obtained using the Gene Ontology Database and were graphically represented according to the data presented in Tables 2 and 3.