High-content imaging reveals the ability of microexons to shape protein localisation

Alternative splicing massively expands the human proteome by generating multiple protein isoforms per gene. However, the functional properties of most protein isoforms are unknown, and it remains unclear whether most have distinct or redundant functions. The development of methods to address this question and characterise proteins at scale are needed. We hypothesized that high-content imaging combined with classifier-based analysis could provide a powerful and broadly applicable approach for characterizing subcellular localization and behavioural differences between protein isoforms. As a proof-of-concept, we examine 42 isoform pairs in HeLa cells, each pair consisting of two isoforms from the same gene that differ in the inclusion/exclusion of a tissue-regulated cassette exon, many differing by 10 or fewer amino acids. We identify localisation differences for ~38% (16/42) of isoform pairs, and features describing these differences. We perform follow-up examination of two sorting nexin 2 (SNX2) isoforms that differ by four amino acids inserted into an α-helix, which drastically altered subcellular localisation. We find the localisation differences are caused by the presence of additional residues rather than their composition, which reduces vesicle association while enhancing interactions with the insulin receptor. Lastly, we generalise these findings by showing that microexons frequently overlap α-helices, are enriched in helix-stabilizing residues, have lengths that preserve helical register, and often modulate surface charge. Together, our findings support a broad role for alternative splicing in modulating protein subcellular localization and increasing functional proteomic diversity.