The availability of complete genome sequences has allowed the prediction of all exported proteins of the corresponding organisms with dedicated algorithms. Even though numerous studies report on genome-based predictions of signal peptides and cell retention signals, they lack a proteomic verification. For example, 180 secretory and 114 lipoprotein signal peptides were predicted recently for the Gram-positive eubacterium Bacillus subtilis. In the present studies, proteomic approaches were used to define the extracellular complement of the B. subtilis secretome. Using different growth conditions and a hyper-secreting mutant, approximately 200 extracellular proteins were visualized by two-dimensional (2D) gel electrophoresis, of which 82 were identified by mass spectrometry. These include 41 proteins that have a potential signal peptide with a type I signal peptidase (SPase) cleavage site, and lack a retention signal. Strikingly, the remaining 41 proteins were predicted previously to be cell associated because of the apparent absence of a signal peptide (22), or the presence of specific cell retention signals in addition to an export signal (19). To test the importance of the five type I SPases and the unique lipoprotein-specific SPase of B. subtilis, the extracellular proteome of (multiple) SPase mutants was analyzed. Surprisingly, only the processing of the polytopic membrane protein YfnI was strongly inhibited in Spase I mutants, showing for the first time that a native eubacterial membrane protein is a genuine Spase I substrate. Furthermore, a mutation affecting lipoprotein modification and processing resulted in the shedding of at least 23 (lipo-)proteins into the medium. In conclusion, our observations show that genome-based predictions reflect the actual composition of the extracellular proteome for approximately 50%. Major problems are currently encountered with the prediction of extracellular proteins lacking signal peptides (including cytoplasmic proteins) and lipoproteins.