High-throughput interrogation of programmed ribosomal frameshifting in human cells

Martin Mikl; Yitzhak Pilpel; Eran Segal

to detect splicing events and obtained an RNA readout (at least 100 reads mapped) for 10666 variants 115 (80% of the library, 94% of the variants with readout in the -1 frame and 96% of the variants with 116 readout in the +1 frame). Even without any filtering, only 2.95% of the variants showed gaps in the 117 mapping of at least 20 nt which could represent events of cryptic splicing (Fig S2EF; 0 .95% introducing 118 a shift to frame +1 and 0.64% to frame -1). Missing stretches can also come from synthesis or cloning 119 errors (and are therefore present and filtered out on the DNA level. Considering only those cases that 120 had a potential (degenerate, cryptic) donor or acceptor splice site in the area of the gap further 121 reduced the fraction of variants for which we cannot rule out a relevant splicing event to 36 (0.55%; 122 Fig S2EF, red; 0.22% introducing a shift to frame +1 and 0.14% to frame -1). Importantly, even for 123 these cases there was no correlation between the fraction of potentially spliced reads and GFP 124 expression (Pearson r=0.01, p=0.21 for -1 PRF and r=0.02, p=0.09 for +1 PRF), indicating that cryptic 125 splicing events do not constitute a common source for false positive signal in our assay. 126