Myoblasts transfected with HB10D insulin secrete more hormone than those transfected with wild-type insulin (Simonson, G.D., Groskreutz, J.D., Garman, C.M., and MacDonald, M.J. (1996) Hum. Gene. Ther. 7: 71-78), indicating that production of wild-type insulin is not efficient in these cells. The ability of non- cells to produce insulin was examined in several cell lines. In clones of neuroendocrine GH₄C₁ cells stably transfected with proinsulin, two thirds of ³⁵S-proinsulin was degraded within 3 h of synthesis, while ³⁵S-prolactin was stable. In transiently transfected neuroendocrine AtT20 cells, half of ³⁵S-proinsulin was degraded within 3 h after synthesis, while ³⁵S-growth hormone was stable. In transiently transfected fibroblast COS cells, ³⁵S-proinsulin was stable for longer, but less than 10% was secreted 8 h after synthesis. Proinsulin formed a concentrated patch detected by immunofluorescence in transfected cells that did not co-localize with calreticulin or BiP, markers for the endoplasmic reticulum, but did co-localize with membrin, a marker for the cis-medial Golgi complex. Proinsulin formed a Lubrol-insoluble aggregate within 30 min after synthesis in non- cells, but not in INS-1E cells, a cell line that normally produces insulin. Over 45% of ³⁵S-HB10D proinsulin was secreted from COS cells 3 h after synthesis, and this mutant formed less Lubrol-insoluble aggregate in the cells than did wild-type hormone. These results indicate that proinsulin production from these non- cells is not efficient and that proinsulin aggregates in their secretory pathways. Factors in the environment of the secretory pathway of cells may prevent aggregation of proinsulin to allow efficient production.