The results presented here demonstrate that translation that begins downstream of the ER signal peptide drives nuclear localization of Bmp2, Bmp4, and Gdf5. In each case, elimination of the signal peptide prevents translation of the nascent polypeptide into the ER and thereby prevents transit through the secretory pathway, which in turn prevents contact with the Golgi-localized proprotein convertases that would otherwise cleave the proprotein and destroy its NLS. These findings indicate that nuclear Bmp2, Bmp4, and Gdf5 proteins translated from alternative downstream start codons cannot function in their traditional role as ligands binding to cell surface receptors, because they cannot enter the secretory pathway. Likewise, processed and secreted Bmp2, Bmp4, and Gdf5 would not be likely to enter the nucleus after binding to cell surface receptors and being internalized, as their NLSs are destroyed by proprotein processing. This mechanism for producing nuclear variants of BMP family proteins stands in contrast to several other growth factors including EGF family members, IFNγ, FGF family members, prolactin, and growth hormone [36
]. In these examples, growth factors are secreted from the cell, bind to plasma membrane receptors, and are internalized prior to translocation of the ligand and/or its receptor to the nucleus.
The nuclear localization of growth factor variants that avoid the secretory pathway is not unprecedented. A nuclear form of parathyroid hormone-related peptide (PTHrP) can be generated by translation from an alternative start site downstream of the conventional initiator ATG, producing a protein with a truncated signal peptide much like nBmp2, nBmp4, and nGdf5. Loss of the signal peptide enables PTHrP to bypass the ER and secretory pathway, and an embedded NLS then interacts with importin β1 to direct PTHrP to the nucleus [38
]. Another example of nuclear localization due to utilization of an alternative start codon is found in the basic fibroblast growth factor (bFGF). In one form of this protein, nuclear localization is determined by translational initiation at an upstream alternative start site. The utilization of an upstream CUG start codon produces a variant protein with an extended amino terminus containing an NLS that directs nuclear localization [40
]. In the case of fibroblast growth factor 3 (FGF3), translation can initiate at a CUG codon that is 87 nucleotides upstream of the first AUG in the protein-coding frame. These 87 nucleotides code for two NLS signals and a hydrophobic secretory signal, and the balance between nuclear localization and secretion of this protein variant is determined by the competing signals [41
]. These examples demonstrated a precedent for altering the subcellular localization of proteins by initiating translation from alternative start sites, and they show that nuclear localization of growth factors can occur without prior secretion of the protein from the cell [43
The novel location of the bipartite NLS in Bmp2, overlapping the site of proprotein processing, initially led us to consider a different mechanism of nuclear localization. We examined whether inhibition of proprotein processing might leave the NLS intact and thus lead to nuclear localization of Bmp2. This hypothesis was considered because regulation of furin activity has been shown to affect the activity of furin substrates in other cases. For example, the pro-β-NGF protein is a neurotropin that has opposing activities depending on whether or not it is cleaved by furin. Cleaved β-NGF promotes cell survival, whereas uncleaved β-NGF promotes apoptosis of neurons [44
]. Furin is also responsible for cleavage of the transmembrane receptor Notch. Cleavage results in the release of the Notch intracellular domain, which goes to the nucleus and activates genes involved in development and differentiation [45
]. Uncleaved Notch, in contrast, inhibits cell differentiation [46
]. The experiments presented here, however, do not support a role for furin modulation in regulating the localization of Bmp2.
Instead, this work has demonstrated that nBmp2, nBmp4, and nGdf5 are all produced by initiating translation from a location downstream of the signal peptide. In each protein, an alternative start codon was identified, the mutation of which reduced nuclear localization by approximately 50%. These codons were located at amino acid positions 58, 83, and 173 in Bmp2, Bmp4 and Gdf5, respectively. Interestingly, these different start sites produce nuclear variants that are 336, 326, and 329 amino acids long for nBmp2, nBmp4, and nGdf5, respectively. The differences in the lengths of the three proproteins, therefore, are almost entirely accounted for by differences upstream of the alternative start codons, suggesting that selective pressure has played a role in maintained the lengths of the nuclear variants.
The observation that mutation of the alternative start codons reduced but did not eliminate synthesis of the nuclear variants of each protein suggests that, at least in ectopically expressed fusion constructs, other sites can serve as start codons for synthesis of the nuclear variants when the primary alternative start sites are eliminated. Indeed, the NetStart 1.0 program predicted several weaker alternative start codons in each propeptide coding region. It is not clear, however, whether usage of any other alternative start codons ever occurs in vivo.
The observation that only 20-30% of cells that expressed the three wild type BMP/GFP fusion constructs showed nuclear localization of the fusion proteins suggested that translational start site selection and/or nuclear translocation might be regulated, perhaps in association with the cell cycle. Indeed, immunofluorescence staining of endogenous Bmp2 in cells undergoing mitosis demonstrated that cells entering M-phase of the cell cycle display the most intense staining, which might reflect increased utilization of the alternative start codon during the G2
/M phase of the cell cycle. A similar pattern of subcellular localization throughout the cell cycle has been reported for PTHrP, and it has been suggested that this pattern "supports a role for PTHrP in cell division" [47
]. Dissolution of the nuclear envelope seemed to allow nBmp2 to spread throughout the cell, either by diffusion or active transport, and when the nuclear envelope reassembled during cytokinesis, nBmp2 was no longer preferentially localized to the nucleus. These observations suggest that new nuclear translocation of nBmp2 is required to re-establish nuclear localization every time a cell completes M-phase of the cell cycle. Additional experiments will be required to determine whether the same nBmp2 molecules can be re-transported to the nucleus, or whether de novo protein synthesis is required after each cell division. Likewise, additional experiments are needed to explore whether nBmp2 plays any role in cell division.