A fully mature mRNA is usually associated to a reference open reading frame encoding a single protein. Yet, mature mRNAs contain unconventional alternative open reading frames (AltORFs) located in untranslated regions (UTRs) or overlapping the reference ORFs (RefORFs) in non-canonical +2 and +3 reading frames. Although recent ribosome profiling and footprinting approaches have suggested the significant use of unconventional translation initiation sites in mammals, direct evidence of large-scale alternative protein expression at the proteome level is still lacking. To determine the contribution of alternative proteins to the human proteome, we generated a database of predicted human AltORFs revealing a new proteome mainly composed of small proteins with a median length of 57 amino acids, compared to 344 amino acids for the reference proteome. We experimentally detected a total of 1,259 alternative proteins by mass spectrometry analyses of human cell lines, tissues and fluids. In plasma and serum, alternative proteins represent up to 55% of the proteome and may be a potential unsuspected new source for biomarkers. We observed constitutive co-expression of RefORFs and AltORFs from endogenous genes and from transfected cDNAs, including tumor suppressor p53, and provide evidence that out-of-frame clones representing AltORFs are mistakenly rejected as false positive in cDNAs screening assays. Functional importance of alternative proteins is strongly supported by significant evolutionary conservation in vertebrates, invertebrates, and yeast. Our results imply that coding of multiple proteins in a single gene by the use of AltORFs may be a common feature in eukaryotes, and confirm that translation of unconventional ORFs generates an as yet unexplored proteome.

Four cationic peptides were purified from mussel (Mytilus galloprovincialis) hemocytes. A combination of Edman degradation and mass spectrometry of plasma revealed (i) a previously characterized molecule, mytilin B (Charlet, M., Chernysh, S., Philippe, H., Hetrut, C., Hoffmann, J., and Bulet, P. (1996)J. Biol. Chem. 271, 21808–21813) and (ii) three new isoforms, mytilin C, D, and G1. The four molecules exhibited complementary antimicrobial properties. The cDNA sequence coding for the mytilin B precursor was obtained from a hemocyte cDNA library. This precursor contains a putative signal peptide of 22 residues, a processing peptide sequence of 34 amino acids, and a C-terminal extension of 48 residues rich in acidic residues. Distribution of mytilin B mRNA and of the corresponding peptide in various mussel tissues revealed that mytilins are synthesized and stored in a specific hemocyte subtype. Furthermore, in an experimental model of infection, we showed (i) a recruitment of hemocytes containing mytilins toward the injection site within hours following bacterial challenge, (ii) that mytilins probably play a prominent role in killing intracellular bacteria after phagocytosis, and (ii) later an increase of mytilin-like material occurred in the plasma suggesting a secondary systemic role.