In contrast, in the crystal structure of Tae4 from S. typhimurium no disulfide bond could be observed. These cysteine MLN4924 residues are conserved among homologous Tae4 family members and were shown to be important in Tae4 from E. cloacae. Therefore, we wondered whether mutations of these two cysteine residues in Tae4 impair growth of E. coli cells after TWS119 periplasmic localization of Tae4. Indeed, variants with single mutation of either one of this cysteine to a serine residue could no longer evoke the growth phenotype. This effect on the toxicity of Tae4 after mutation of these cysteine residues has also been observed for the E. cloacae Tae4/Tai4 system and disulfide bond formation was suggested to be important for correct positioning of the third catalytic aspartate residue. However, although we do not observe any disulfide bond between Cys135 and Cys139 in our crystal structure, the position of Asp137 is virtually identical when compared with the E. cloacae homologue. It remains to be shown whether the observed reduced toxicity is indeed due to this disulfide bond as a requirement for enzyme catalysis or whether it serves as protection mechanism against proteolytic degradation. Whereas the effector molecules are potentially secreted through the injection needle of the type-6-secretion system, immunity proteins are shuttled into the periplasmic region to confer immunity against attacks from siblings and to avoid selfintoxication by self-encoded effector molecules. Thus, we expected Tai4 to be exported and truncated by its periplasmic leader sequence. Indeed, when building the Tai4 polypeptide chain we could model the residues Gln27 to Lys127 in the electron density but could not assign the N-terminal region. We could verify truncation of the N-terminal leader sequence by mass spectrometry experiments showing that the purified Tai4 had an effective mass of 11,695 Da, which is in excellent agreement with the predicted mass of Tai4 truncated after Ala26. Cleavage at this position is most likely, since the upstream region from Gln27 contains a classical periplasmic localization sequence and residues that are essential for efficient export are conserved. In support of a periplasmic localization of Tai4 we showed that only full-length Tai4 could completely rescue the growth phenotype after export of Tae4 in the periplasmic compartment. In contrast, the Tai4DN26 variant could only mildly counteract Tae4 toxicity. Notably cells still grow slightly better when Tai4DN26 is present in the cytoplasm when compared with cells that express periplasmic Tae4 alone. This partial rescue is most likely due to strayed inhibitor molecules that enter the periplasm upon partial cell lysis by periplasmic Tae4. Thus, our structural model consists of a fully processed Tai4 molecule that had been exported into the periplasm. The overall structure of the Tai4 immunity protein showed a purely a-helical fold consisting of six helices stacked onto each other. Strikingly, a closer inspection of the electron density map revealed that one side formed by the a-helices b, c and e of Tai4 makes extensive contact with the identical surface of 
a crystal-symmetry related molecule. The two molecules are in a head-to-tail arrangement and related by a crystallographic twofold symmetry. Each molecule establishes a contact surface of 1410 A ? 2 and the oligomeric state within the biological assembly of Tai4 is therefore most likely dimeric. Whereas the center of this dimer interface is mainly hydrophobic, the boundary area is formed by hydrogen bonds between the two molecules. Additional support for the biological relevance of this dimer interface comes from our bioinformatic studies, showing that the hydrophobic nature of the interface as well as residues that form hydrogen bonds are conserved.