Because its expression appears to be largely restricted to human melanomas, 9-O-acetyl-GD3 is a candidate antigen for vaccine construction. Searching for potential sources, we compared chemically O-acetylated calf brain GD3 and 9-O-acetyl-GD3 extracted from bovine buttermilk with 9-O-acetyl-GD3 from human melanoma. Three fractions (F1-F3) of chemically O-acetylated GD3 differed in the number and position of O-acetyl groups. O-Acetylation sites were the lactose portion in F1 and lactose as well as sialic acid in F2 and F3. Natural (melanoma- or buttermilk-derived) 9-O-acetyl-GD3 was O-acetylated solely on the sialic acid moiety. While F1 was not reactive with monoclonal antibodies against 9-O-acetyl-GD3, F2 and F3 were as reactive as the natural products. Immunization with the natural products induced high-titer antibodies against natural 9-O-acetyl-GD3 as well as F2 and F3. In contrast, mice immunized with the synthetic fractions produced antibodies only against the immunogen but not against natural 9-O-acetyl-GD3. Only immunization with the natural products induced production of antibodies reactive with surface antigens of melanoma cells expressing 9-O-acetyl-GD3. The findings suggest (a) that C-9 of the subterminal sialic acid is the site of chemical O-acetylation in F2 and F3, as opposed to C-9 of the terminal sialic acid in the natural products; (b) that O-acetylation of both the terminal and subterminal sialic acid moieties of GD3 results in recognition by three murine monoclonal antibodies (D1.1, ME 311, and Jones) reactive with human melanoma cells; (c) that O-acetylation of the terminal sialic acid is critical, on the other hand, for inducing an immune response against melanoma 9-O-acetyl-GD3; and (d) that O-acetyl GD3 from bovine buttermilk can substitute as immunogen for inducing an immune response against human melanoma cell surface antigens in the mouse.