Somatic non-endocrine cells, engineered by gene transfer for glucose-regulated secretion of mature human insulin, might represent a future tool for the cure of insulin-dependent diabetes mellitus. Non-endocrine cells lack the β-cell specific proteases for proinsulin cleavage, but are able to process proteins via the constitutive pathway. To exploit this property, we introduced two mutations in the human preproinsulin cDNA (HPI) by overlap extension PCR. Such a mutated cDNA (HPIm) should encode a proinsulin molecule containing the cleavage sequences for the ubiquitous protease furin. The HPIm cDNA was then inserted into a retroviral vector carrying the cDNA coding for the truncated receptor of the human nerve growth factor (ΔNGFr), for cell marking and positive selection. By this retroviral vector the HPIm cDNA was stably transduced into murine fibroblasts (NIH3T3) and human hepatoma cells (HepG2), resulting in mature insulin release (assessed by specific IRMA, cross reactivity with proinsulin <0,001%), ranging from 13.2 to 37.2 pM/24h/10 6 cells. Biological activity of this mutated insulin was assessed in diabetic athymic mice (blood glucose levels >22.4 mM, plasma IRMA levels <1.5 pM). I.p. injection of 6-10x10 7 transduced murine fibroblasts produced hypoglycemia in 60% of the recipients, and plasma IRMA levels of 26.4-73.2 pM. Primary cultures of rat hepatocytes were subsequently transduced, with a 40-50% transduction efficiency (measured by NGFr expression) resulting in mature insulin accumulation in the culture medium (414+/-18 pM/24h/10 6 cells). Such engineered cell lines and primary hepatocytes represent an useful model for further genetic manipulation aimed at developing the secretion of mature insulin in response to extra-cellular glucose variations.