Cell type specific profiling of alternative translation identifies novel protein isoforms in the mouse brain

Translation canonically begins at a single AUG and terminates at the stop codon, generating one protein species per transcript. However, some transcripts may use alternative initiation sites or sustain translation past their stop codon, generating multiple protein isoforms. Through other mechanisms such as alternative splicing, both neurons and glia exhibit remarkable transcriptional diversity, and these other forms of post-transcriptional regulation are impacted by neural activity and disease. Here, using ribosome footprinting, we demonstrate that alternative translation is likewise abundant in the central nervous system and modulated by stimulation and disease. First, in neuron/glia mixed cultures we identify hundreds of transcripts with alternative initiation sites and confirm the protein isoforms corresponding to a subset of these sites by mass spectrometry. Many of them modulate their alternative initiation in response to KCl stimulation, indicating activity-dependent regulation of this phenomenon. Next, we detect several transcripts undergoing stop codon readthrough thus generating novel C-terminally-extended protein isoforms in vitro. Further, by coupling Translating Ribosome Affinity Purification to ribosome footprinting to enable cell-type specific analysis in vivo, we find that several of both neuronal and astrocytic transcripts undergo readthrough in the mouse brain. Functional analyses of one of these transcripts, Aqp4, reveals readthrough confers perivascular localization, indicating readthrough can be a conserved mechanism to modulate protein function. Finally, we show that AQP4 readthrough is disrupted in multiple gliotic disease models. Our study demonstrates the extensive and regulated use of alternative translational events in the brain and indicates that some of these events alter key protein properties. [NOTE: this GEO series only contains in vivo data] For in vivo experiments (4 samples in total): 2 replicates were prepared from P21 Snap25::eGFP-RpL10a mouse brains for profiling neurons, 2 replicates from Aldh1l1:: eGFP-RpL10a mouse brains for profiling astrocytes. All samples were uniquely indexed and pooled across two lanes. For each sample, the uploaded fastq files are union of both lanes.