F1-20/AP-3 is a synapse-specific phosphoprotein. In this study we characterize the ability of bacterially expressed F1-20/AP-3 to bind and assemble clathrin cages. We find that both of two bacterially expressed alternatively spliced isoforms of F1-20/AP-3 can bind and assemble clathrin as efficiently as preparations of F1-20/AP-3 from bovine brain. This establishes that the clathrin assembly activity found in F1-20/AP-3 preparations from brain extracts is indeed encoded by the cloned gene for F1-20/AP-3. It also demonstrates that post-translation modification is not required for activation of the clathrin binding or assembly function of F1-20/AP-3. Ultrastructural analyses of the clathrin cages assembled by bacterially expressed F1-20/AP-3 reveals a strikingly narrow size distribution. This may be important for the regulation of quantal size during neurotransmission. We also express the 33 kD NH2-terminus of F1-20/AP-3 in E. coli, and measure its ability to bind to clathrin triskelia, to bind to clathrin cages, and to assemble clathrin triskelia into clathrin cages. It has been suggested that the 33 kD NH2-terminus of F1-20/AP-3 constitutes a clathrin binding domain. We find that the bacterially expressed 33 kD NH2-terminus of F1-20/AP-3 binds to clathrin triskelia, fails to bind to preassembled clathrin cages, and is not sufficient for clathrin assembly. The finding that the 33 kD NH2-terminus of F1-20/AP-3 binds to clathrin triskelia but fails to assemble clathrin triskelia into clathrin cages is consistent with the published proteolysis studies. The finding that the 33 kD NH2-terminus of F1-20/AP-3 fails to bind to clathrin cages is novel and potentially important. It is clear from these experiments that the 33 kD NH2-terminus of F1-20/AP-3 is sufficient to carry out some aspects of clathrin binding; however it appears that defining the regions of the protein involved in clathrin binding and assembly may be more complex than originally anticipated.