The mechanism of viral capsid assembly is an intriguing problem because of its fundamental importance to research on synthetic viral particle vaccines, gene delivery systems, antiviral drugs, chimeric viruses displaying antigens or ligands, and the study of macromolecular interactions. The genes coding for the scaffolding (gp7), capsid (gp8), and portal vertex (gp10) proteins of the procapsid of bacteriophage phi 29 of Bacillus subtilis were expressed in Escherichia coli individually or in combination to study the mechanism of phi 29 procapsid assembly. When expressed alone, gp7 existed as a soluble monomer, gp8 aggregated into inclusion bodies, and gp10 formed the portal vertex. Circular dichroisin spectrum analysis indicated that gp7 is mainly composed of alpha helices. When two of the proteins were coexpressed, gp7 and gp8 assembled into procapsid-like particles with variable sizes and shapes, gp7 and gp10 formed unstable complexes, and gp8 and gp10 did not interact. These results suggested that gp7 served as a bridge for gp8 and gp10. When gp7, gp8, and gp10 were coexpressed, active procapsids were produced. Complementation of extracts containing one or two structural components could not produce active procapsids, indicating that no stable intermediates were formed. A dimeric gp7 concatemer promoted the solubility of gp8 but was inactive in the assembly of procapsid or procapsid-like particles. Mutation at the C terminus of gp7 prevented it from interacting with gp8, indicating that this part of gp7 may be important for interaction with gp8. Coexpression of the portal protein (gp20) of phage T4 with phi 29 gp7 and gp8 revealed the lack of interaction between T4 gp20 and phi 29 gp7 and/or gp8. Perturbing the ratio of the three structural proteins by duplicating one or another gene did not reduce the yield of potentially infectious particles. Changing of the order of gene arrangement in plasmids did not affect the formation of active procapsids significantly. These results indicate that phi 29 procapsid assembly deviated from the single-assembly pathway and that coexistence of all three components with a threshold concentration was required for procapsid assembly. The trimolecular interaction was so rapid that no true intermediates could be isolated. This finding is in accord with the result of capsid assembly obtained by the equilibrium model proposed by A. Zlotnick (J. Mol. Biol. 241:59-67, 1994).