Although mammalian cell factories are now widely used in production of biomedically important proteins they are still relatively ineffective compared with those from prokaryotic organisms. In an attempt to improve the CHO cell line we have used a biotechological approach to investigate whether or not the choice of signal peptide may have a major impact on the synthesis/ secretion of a recombinant protein (Gaussia luciferase) in transfected cells. In experiments where the native luciferase signal peptide was replaced with that derived from a series of sources the amount of luciferase produced varied by a factor of more than 100. The most effective signal peptide being that derived from Gaussia luciferase. Surprisingly two signal peptides of mammalian origin, human interleukin-2 and albumin, proved to be far inferior to the signal peptide derived from the marine organism. A reduction in level of mRNA available for translation proved not to be responsible for the major decrease observed in luciferase production. When tested in a hepatic cell line (HepG2) the marine luciferase signal peptide again proved to be superior to that from albumin. The Gaussia luciferase signal peptide was also much more effective than that derived from Vargula luciferase in production of Vargula luciferase. This demonstrates that a native signal peptide is not necessarily always the most effective. The signal peptide from Metridia, seven oikosin signal peptides derived from Oikopleura and two prokaryotic signal peptides all proved to be far inferior to the Gaussia luciferase signal peptide with respect to recombinant luciferase production in CHO cells. Overall the results described here indicate the vital importance of selection of the signal peptide when aiming to produce maximal amounts of recombinant protein in a mammalian expression system.