Our previous experiments showed that immune IgG and F(ab')2, but not Fab', mediated serum killing of Escherichia coli 0111B4, strain 12015 (12015), without significantly increasing the extent of terminal complement (C) component attachment to the bacterial surface. We concluded that bactericidal antibody must change either the site or the nature of C5b-9 bacterial attachment. To pursue this possibility, conditions necessary for elution of C5b-9 from the bacterial surface were examined. Forty-two to 44% of 125I-C9 was released from the serum-resistant nonpresensitized 12015 by 1 M NaCl or 0.1% trypsin, compared with the 21 to 24% release from the serum-sensitive presensitized isolate under the same condition. When strain 12015 bearing 125I-C9 was lysed in a French pressure cell, 73.1% of 125I-C9 was released with the capsular fraction if the organisms had not been presensitized. In contrast, on presensitized 12015, 70.2% of 125I-C9 remained associated with the outer membrane after such lysis. These results suggested that C5b-9 was trapped within or underneath the capsule of 12015 in the absence of bactericidal antibody, but that addition of antibody led to C5b-9 insertion into the outer membrane with bacterial killing. The requirement of C components preceding C5 for bacterial killing was next examined. Minimal killing of presensitized 12015 occurred when a terminal C complex was formed by acid activation from purified C5, C6, C7, C8, and C9 in the absence of C3 or earlier components. In contrast, between 1.2 and 3 log killing of nonpresensitized rough Salmonella minnesota and rough E. coli was observed in the same system. Killing of 12015 was examined with bacteria incubated in C5-deficient serum (C5D), followed by washing and the addition of purified C5, C6, C7, C8, and C9 to permit C5b-9 formation. Antibody was added before or after incubation in C5D serum, or after the addition of purified C5-C9. Under conditions of equivalent C3 and C9 binding, significant killing occurred only when antibody was added before incubation in C5D serum. These results show that antibody must be present at or before the time of C5 convertase formation to mediate killing of 12015 by C5b-9. Therefore, antibody is unlikely to be functioning primarily to alter the bacterial surface to expose sites for C5b-9 insertion, nor is the effect of antibody simply to increase C3 and terminal component binding. We postulate that antibody mediates killing of 12015 by localizing C5b-9 around antibody-clustered sites of C3 and C5 convertase formation.