N-myristoylation is a cotranslational modification involved in protein-protein interactions as well as in anchoring polypeptides to phospholipid bilayers; however, its role in targeting proteins to specific subcellular compartments has not been clearly defined. The mammalian myristoylated flavoenzyme NADH-cytochrome b5 reductase is integrated into ER and mitochondrial outer membranes via an anchor containing a stretch of 14 uncharged amino acids downstream to the NH2-terminal myristoylate glycine. Since previous studies suggested that the anchoring function could be adequately carried out by the 14 uncharged residues, we investigated a possible role for myristic acid in reductase targeting. The wild type (wt) and a nonmyristoylatable reductase mutant (gly2-->ala) were stably expressed in MDCK cells, and their localization was investigated by immunofluorescence, immuno-EM, and cell fractionation. By all three techniques, the wt protein localized to ER and mitochondria, while the nonmyristoylated mutant was found only on ER membranes. Pulse-chase experiments indicated that this altered steady state distribution was due to the mutant's inability to target to mitochondria, and not to its enhanced instability in that location. Both wt and mutant reductase were resistant to Na2CO3 extraction and partitioned into the detergent phase after treatment of a membrane fraction with Triton X-114, demonstrating that myristic acid is not required for tight anchoring of reductase to membranes. Our results indicate that myristoylated reductase localizes to ER and mitochondria by different mechanisms, and reveal a novel role for myristic acid in protein targeting.

Two forms of NADH-cytochrome b5 reductase are produced from one gene: a myristylated membrane-bound enzyme, expressed in all tissues, and a soluble, erythrocyte-specific, isoform. The two forms are identical in a large cytoplasmic domain (Mr approximately 30,000) and differ at the NH2-terminus, which, in the membrane form, is responsible for binding to the bilayer, and which contains the myristylation consensus sequence and an additional 14 uncharged amino acids. To investigate how the two differently targeted forms of the reductase are produced, we cloned a reductase transcript from reticulocytes, and studied its relationship to the previously cloned liver cDNA. The reticulocyte transcript differs from the liver transcript in the 5' non-coding portion and at the beginning of the coding portion, where the seven codons specifying the myristoylation consensus are replaced by a reticulocyte-specific sequence which codes for 13 non-charged amino acids. Analysis of genomic reductase clones indicated that the ubiquitous transcript is generated from an upstream "housekeeping" type promoter, while the reticulocyte transcript originates from a downstream, erythroid-specific, promoter. In vitro translation of the reticulocyte-specific mRNA generated two products: a minor one originating from the first AUG, and a major one starting from a downstream AUG, as indicated by mutational analysis. Both the AUGs used as initiation codons were in an unfavorable sequence context. The major, lower relative molecular mass product behaved as a soluble protein, while the NH2-terminally extended minor product interacted with microsomes in vitro. The generation of soluble reductase from a downstream AUG was confirmed in vivo, in Xenopus oocytes. Thus, differently localized products, with respect both to tissues and to subcellular compartments, are generated from the same gene by a combination of transcriptional and translational mechanisms.