We have shown previously that multimeric C9 within C5b-9 (C9:C5b-8 greater than 3:1) is needed for killing of a rough strain of Escherichia coli. We now extend these studies using serum sensitive, rough (R) and serum resistant, wild type (WT) strains of Salmonella typhimurium as well as a mutant S. typhimurium strain (TS) with a temperature sensitive mutation in synthesis of keto-deoxy-octulosonate, a constituent within the deep core structure of Salmonella LPS. Both R and TS required multimeric C9 within C5b-9 to be killed. Addition at 37 degrees C of increasing inputs of C9 to TS or R bearing C5b-9 led to a dose-related increase in C9 binding and killing. In contrast, addition of high inputs of C9 to the same strains at 4 degrees C, a procedure that limits the C9:C5b-8 ratio to 1:1, resulted in low C9 binding and minimal killing. Bactericidal C5b-9 formed at 37 degrees C on R and TS with high inputs of C9 co-sedimented with the bacterial outer membrane on sucrose density gradient analysis. Non-bactericidal C5b-9 on R, WT, and TS co-sedimented near the inner membrane, despite the presumed lack of association between these constituents. Whereas 125I C9 within the non-bactericidal pools immunoprecipitate with anti-C5, 125I C9 within bactericidal pools did not immunoprecipitate with anti-C5, anti-C7, or anti-C9. These findings suggest that bactericidal C5b-9 may be deposited in a unique location or configuration within the bacterial cell wall.