The degradation of extended protein isoforms points to a misfiring translation initiation process.

There is ever increasing evidence for significant amounts of translation upstream of known AUG start codons in protein coding genes. Some of this translation is from upstream open reading frames (ORFs) that are unconnected to the main coding exons, but upstream initiation codons that are in-frame with coding exons can produce N-terminally extended protein isoforms. N-terminal extensions have much more proteomics support than the shorter proteins predicted to be produced from upstream ORFs. The upstream regions that produce N-terminal extensions have certain characteristics in common. They are highly GC-rich, most of the predicted start codons are non-AUG, and most do not conserve their reading frames beyond simians. The extended isoforms themselves are found significantly more frequently in dysregulated cells than in normal tissues. Approximately one in seven of these N-terminal extensions are upstream of signal peptides and would almost certainly block their recognition by the signal recognition particle. As a result, N-terminally extended isoforms containing exposed, hydrophobic signal peptides would be expected to accumulate in the cytoplasm. However, this analysis finds that those N-terminal extensions that would block signal recognition are practically not detected at the protein level even though the transcripts that would produce these extensions are found as expected in ribosome profiling experiments. This is a clear indication that these mislocated proteins are degraded after translation. Theprobable degradation of these extended proteins strongly suggests that their translation is a side effect of inefficient translation initiation.