||Secretory production of recombinant proteins by E. coli has many advantages over intracellular production. Induced expression of a secretory exoglucanase, Exg, engineered in the Tac-cassette excretion vector was lethal to E. coli. An exponentially growing culture harboring the recombinant construct suffered slow growth and 99.9% of its cells died within 60-100 min after induction. Analysis of subcellular fractions revealed the presence of Pre-Exg in the inner membrane of cultures expressing high levels but not low levels of Pre-Exg. As only Pre-Exg but not Mat-Exg was detectable in the cytoplasm, and Exg was shown by cross-linking experiments to be physically associated with the Sec proteins, it was concluded that secretion and processing of Pre-Exg took place in the SecYEG translocation machinery. The results supported the idea that cell death was caused by some unusual tie-up of Pre-Exg with the SecYEG translocation machinery, thus imposing an inhibitory effect on the secretion of endogenous secretory proteins. A new model, designated 'Saturated Translocation', was proposed to explain the interchangeable lethal and non-lethal properties of Pre-Exg, and to address the possible scenarios that might occur in the course of cell death triggered by secretion of Pre-Exg. Furthermore, we propose here that the cell lysate ratio (Pre/Mat RQ) of the unprocessed precursor Exg protein (Pre-Exg) and its processed mature product (Mat-Exg) reflects the capacity of E. coli to secrete Exg. A Pre/Mat RQ of 20/80, designated the 'Critical Value (CV)', was an important threshold measurement. A rise in the Pre/Mat RQ triggered a mass killing effect. The use of various secretion signal peptides did not improve the viability of cells expressing high levels of Pre-Exg under strong tac promoter control. However, use of the weaker vegG promoter in conjunction with a change in start codon of the spa leader sequence from ATG to TTG in a pM1vegGcexL plasmid construct resulted in a high level (0.9 u ml-1) of excreted Exg in shake-flask cultures. This was 50% higher than the best result obtained from plasmid construct lacUV5par8cex, using the lacUV5 promoter and the ompA leader sequence. Variations in the excreted Exg activities were attributable to differences in the Pre/Mat RQ values of the induced cultures harboring pM1vegGcexL and lacUV5par8cex. These values were 18/82 and 10/90, respectively. Employing fed-batch cultivation in two-liter fermentors, an induced JM101(pM1vegGcexL) culture yielded 4.5 u ml-1 of excreted Exg, which was over sevenfold greater than previously reported. Our results illustrate the successful application of the Pre/Mat RQ ratio as a guide to the attainment of a maximum level of secreted/excreted Exg.