An angiogenic response originating from peritumoral sinusoids and portal tracts that leads to the formation of metastases with sinusoidal- and portal-type angiogenic patterns, respectively, occurs during the course of liver colonization by murine 51b colon carcinoma (51b-CC) cells. We found a 5-fold increase in endogenous endostatin levels from hepatic blood over baseline (25 +/- 6 ng/mL) when micrometastatic foci had a detectable size and a 14-fold increase when macrometastases were developed. Despite this endogenous endostatin production, subcutaneous administration of recombinant human endostatin (rh-E; 50 mg/kg) decreased metastasis number by 60% when dosed from days 1 to 20 after 51b-CC cell injection, by 40% when given from days 10 to 20, and by 30% when administered as a single dose 30 minutes before 51b-CC cell injection compared with controls. In addition, administration of rh-E from days 10 to 20 decreased overall metastasis volume by 90% compared with controls. rh-E increased the number of necrotic sinusoidal-type metastases by 7-fold and decreased their intrametastatic CD31(+)-microvessel density by 80% without affecting portal-type metastases. Flow cytometry showed rh-E binding to mouse liver sinusoidal cells but not to CD45(+) cells (leukocytes and Kupffer cells) or 51b-CC cells. Furthermore, rh-E induced sinusoidal endothelium cell apoptosis. In conclusion, despite the direct correlation between metastasis development and endogenous endostatin generation in the liver, administration of rh-E inhibited micrometastasis generation and macrometastasis growth very efficiently. The antiangiogenic mechanism was selective for sinusoidal-type metastases, in which the neovasculature originating from sinusoidal endothelium cells was targeted by rh-E.