The antiangiogenic protein endostatin showed considerable preclinical antitumor activity, but limited efficacy in phase I/II trials. Prior studies using an anti-HER2 antibody-murine endostatin fusion showed enhanced antitumor activity compared to anti-HER2 antibody or endostatin given alone, or in combination. We have generated two anti-HER2 human endostatin fusion proteins by fusing either wild-type or a mutant human endostatin (huEndo-P125A) to the 3' end of a humanized anti-HER2 IgG3 antibody. Antitumor efficacy was examined in murine and human breast tumor models. HuEndo-P125A antibody fusion protein (αHER2-huEndo-P125A) inhibited VEGF and bFGF induced endothelial cell proliferation, and tube formation in vitro, more efficiently than endostatin alone, wild-type endostatin fusion protein (αHER2-huEndo), or parental anti-HER2 antibody (αHER2 IgG3). Wild-type and mutant human endostatin was rapidly cleared from serum in mice (T½(2) = 2.0-2.1 hours), whereas αHER2-huEndo fusion proteins had a significantly prolonged half-life (T½(2) = 40.7-57.5 hours). Treatment of SK-BR-3 breast cancer xenografts with anti-HER2 IgG3-huEndo-P125A fusion resulted in greater inhibition of tumor growth and improved survival, compared to treatment with either αHER2 IgG3 (P = 0.025), human endostatin (P = 0.034), or anti-HER2 IgG3-huEndo (P = 0.016). αHER2-huEndo-P125A specifically inhibited tumors expressing HER2 in mice simultaneously implanted with murine mammary tumor EMT6 cells and with EMT6 engineered to express HER2 antigen (EMT6-HER2). Targeting of endostatin using antibody fusion proteins could improve antitumor activity of either anti-HER2 antibody and/or endostatin and provides a versatile approach that could be applied to other tumor targets with alternative antibody specificities.