The fidelity of splice site selection is critical for proper gene expression. In particular, proper recognition of 3-splice site (3SS) sequences by the spliceosome is challenging considering the low complexity of the 3SS consensus sequence YAG. Here, we show that absence of the Prp18p splicing factor results in genome-wide activation of alternative 3SS in S. cerevisiae, including highly unusual non-YAG sequences. Usage of these non-canonical 3SS in the absence of Prp18p is enhanced by upstream poly(U) tracts and by their potential to interact with the first intronic nucleoside, allowing them to dock in the spliceosome active site instead of the normal 3SS. The role of Prp18p in 3SS fidelity is facilitated by interactions with Slu7p and Prp8p, but cannot be fulfilled by Slu7p, identifying a unique role for Prp18p in 3SS fidelity. This fidelity function is synergized by the downstream proofreading activity of the Prp22p helicase, but is independent from another late splicing helicase, Prp43p. Our results show that spliceosomes exhibit remarkably relaxed 3SS sequence usage in the absence of Prp18p and identify a network of spliceosomal interactions centered on Prp18p which are required to promote the fidelity of the recognition of consensus 3SS sequences.

We describe an RT-PCR protocol that allows high-resolution mapping of splicing products and isoforms using fluorescently labeled primers. Each species contains one fluorescent group allowing a direct comparison of the different isoforms despite size differences. A custom-size ladder enables the precise determination of cDNA lengths and discrimination of isoforms differing by less than five nucleotides on polyacrylamide gels. This protocol also allows the detection of products from in vitro splicing reactions, circumventing the need to use radiolabeled transcripts. For complete details on the use and execution of this protocol, please refer to Gabunilas and Chanfreau (2016).

Multiplexed assays allow functional testing of large synthetic libraries of genetic elements, but are limited by the designability, length, fidelity and scale of the input DNA. Here, we improve DropSynth, a low-cost, multiplexed method that builds gene libraries by compartmentalizing and assembling microarray-derived oligonucleotides in vortexed emulsions. By optimizing enzyme choice, adding enzymatic error correction and increasing scale, we show that DropSynth can build thousands of gene-length fragments at >20% fidelity.