We have shown recently that an average of three or more C9 molecules must bind to C5b-8 on Escherichia coli strain J5 to cause direct complement killing in the absence of serum lysozyme. We initially confirmed and extended this observation by showing that deposition of a large number of C5b-9 complexes bearing 1C9 per C5b-8 was not bactericidal for J5. To identify the target site for bactericidal C5b-9 deposition, we measured release of periplasmic and cytoplasmic markers of different size from J5 as the C9:C5b-8 ratio was changed, because the diameter of the C5b-9 channel is known to increase as the C9:C5b-8 ratio increases. To facilitate measurement of release of the periplasmic marker beta-lactamase (BLA), J5 was transformed for high level constitutive TEM-1 BLA production (J5-Amp). Multimeric C9 within C5b-9 (C9:C5b-8 greater than 3) was required to release BLA (m.w. 28,900) from J5-Amp regardless of whether cells bore 310, 560, or 890 C5b-9/organism. Curves of both BLA release and killing vs C9:C5b-8 ratio were sigmoidal and nearly superimposable. Release of the small cytoplasmic marker 86Rb, a potassium analog, also required a minimum C9:C5b-8 ratio of 3:1; specific 86Rb release did not occur in the absence of killing. Release of the large cytoplasmic marker beta-galactosidase (m.w. 505,000) did not occur even at the highest achievable C9:C5b-8 ratio of 11:1, despite greater than 99.9% killing, indicating that there was no dissolution of the peptidoglycan layer due to incomplete removal of serum lysozyme. Complement-mediated killing of J5 requires sufficient damage to the outer membrane or formation of a sufficiently large C5b-9 channel to release the large periplasmic marker BLA. The requirement of multimeric C9 for 86Rb release suggests that at low C9:C5b-8 ratios, either C5b-9 does not have access to the cytoplasmic space or that the J5 K+ transport systems are able to compensate for putative C5b-9 channels.