Without being exposed to the external environment, suggesting that it remains embedded in the mycobacterial outer membrane. Surprisingly, when the same construct was expressed in M. marinum it was efficiently secreted in the culture supernatant in an ESX-5-dependent manner, while it remained associated with the cell wall when expressed in M. tuberculosis or M. bovis BCG. Moreover, in M. marinum the secreted form of the PE domain had a lower apparent molecular weight than predicted, indicative of a maturation process, as was previously demonstrated for PE_PGRS33 wt and for LipY. In the cell fractionation experiments involving M. tuberculosis extracts, Rv1698 was not only found in the cell wall fraction, but also in the cytoplasmic fraction. This imperfect fractionation could be due to the fact that, for biosafety reasons, M. tuberculosis was lysed by bead beating instead of sonication. Since our data suggested that the first portion of the PE domain contains functions required for protein export, we constructed three chimeric proteins in which the first 30, 43 or 61 residues of the PE domain of PE_PGRS33 were fused to the coding sequence of GFP. All these proteins were able to localize in the cell wall, although in M. smegmatis only the construct including the first 61 residues of the PE domain was translocated with a fair efficiency. These result clearly indicated that PE_PGRS33-mediated translocation is more efficient in M. bovis BCG than in M. smegmatis. It should be noted that, while M. bovis BCG encodes an ESX-5 secretion system, M. smegmatis does not. Moreover, this species encodes only few PE and no PE_PGRS proteins. However, M. smegmatis chromosome encodes other type VII secretion systems that might, with low efficiency, complement the absence of ESX-5 and have a role in the secretion of PE_PGRS33 and its derivatives in this species. Finally, the MK-2206 chimera including only the first 30 amino acids of the PE domain fused to GFP was partially released in M. bovis BCG culture supernatant, but not in that of M. smegmatis. This interesting finding suggests that the first 30 amino acids of the PE domain contain sufficient information to allow protein translocation. The only structural information available for a PE protein derives from PE25, a protein including only the PE domain, whose structural gene is followed by the gene encoding a protein of the PPE family. These two proteins were shown to interact to form a heterodimer, for which the crystal structure was solved. The PE structure included two antiparallel ahelices connected by a loop. If the different PE domains have a similar folding, the first chimera would include most of the first a-helix, while the second would include both the first a-helix and the loop, and the last chimera would include the first ahelix, the loop and part of the second a-helix. Our data suggest an involvement of the first a-helix in directing the protein to the secretion system.