We previously showed that translation from the rat BACE1 5' leader is cap-dependent and that four AUG codons (AUG1-4) in the 5' leader were bypassed, partially or completely, depending on the cell line. Two other groups reported comparable results with human BACE1 sequences in different cell lines, although different mechanisms were postulated. In contrast, a third group working with the human sequence reported that most translation events are initiated at AUG2. Using reporter constructs with the rat BACE1 5' leader in rat cells, we now show that this apparent discrepancy between studies can be explained by the use of different expression systems and differences in interpretation. When reporter constructs were transcribed in the nucleus, the upstream AUG codons did not affect translation, but when mRNAs were transcribed in the cytoplasm or when in vitro transcripts were transfected into cells, the upstream AUG codons inhibited translation. These findings suggest that when transcription occurs in the nucleus, the BACE1 mRNA initiates translation by a shunting mechanism. The results are less consistent with either leaky scanning or reinitiation and provide a caveat against the use of cytoplasmic expression systems or RNA transfection for analyses of translation initiation.