Fig 1.
Transcriptome-wide de novo alternative splicing upon depletion of functional Exon Junction Complex.
(A) Overview of upregulated de novo splice junctions in EJC-depleted cells. Top: schematic of cryptic 5’ and 3’ SS. In this toy gene model, canonical pre-mRNA exons and introns are depicted as blue boxes and black lines. The ends of these introns are marked by splice signatures (GU: donor and AG: acceptor, shown in black). Cryptic splice sites identified in the EJC LOF datasets can be found within sequences that are normally exonic or intronic. These sites and the putative de novo intron are shown as red text and dashed lines. Bottom: Pie chart indicating the distribution of different splice junction classes. (B) Sashimi plot depicting HISAT2-mapped sequencing coverage along a portion of straw, which has defective splicing under core-EJC LOF. The gene model depicts the location of the cryptic 5’ SS relative to the annotated 5’ SS. Junction spanning read counts mapping to the canonical junction are circled, whereas cryptic junction read counts are squared. Note that spliced reads mapping to the cryptic junction are found in eIF4AIII-, mago- and tsu-KD but not the control comparison. EJC-CLIP [29] shows recruitment of EJC to exon-exon junctions. Region containing the cryptic 5’ SS has been zoomed on the right. (C) Sashimi plot depicting RNA-seq coverage at the mask gene, where depletion of the EJC (shown here, eIF4AIII-RNAi) results in utilization of a cryptic exonic 3’ SS. (D) Schematic for rt-PCR validation of exonic cryptic splicing, which yields shorter, internally deleted products. (E) Validation of de novo splicing events in core-EJC depleted cells. EJC core components (eIF4AIII, mago, tsu and btz) were depleted from Drosophila S2 cells using dsRNA. After knockdown, eight targets identified in (A) were evaluated using rt-PCR and demonstrated splicing defects (asterisk). Importantly, only knockdown of core-EJC factors yielded cryptic bands, but not btz or control conditions. unkempt (unk) generates several products due to multiple alternative exons included within its rt-PCR amplicon (S1D Fig). # indicates products of unknown identity.
Fig 2.
EJC-depletion leads to activation of cryptic 3’ splice sites.
(A) Depiction of 3’ SS position of spurious junctions relative to exon-exon boundaries as density and dot plot. The dot plot indicates splice site scores as calculated via NNSPLICE. Horizontal dashed line depicts threshold for strong 3’ SS, and vertical dashed lines specify 50 nt flanking exon-exon junctions. (B) Sashimi plot depicting HISAT2-mapped sequencing coverage along a portion of CG7408, which has a cryptic 3’ SS that is activated under core-EJC LOF. Junction spanning read counts mapping to the canonical junction are circled, whereas cryptic junction read counts are squared. Note that spliced reads mapping to the cryptic junction are found in eIF4AIII, mago and tsu KD but not the control comparison. EJC-CLIP [29] shows recruitment of EJC to exon-exon junctions. (C) Validation of CG7408 cryptic 3’ SS activation in core-EJC, but not btz or lacZ KD conditions. (D) Schematic of CG7408 splicing reporters. Exons 1–4 (introns included) were cloned and subjected to further manipulation. Locations of deleted introns (Δ), as well as a construct lacking all introns (mRNA) are included. For reference, the position of the cryptic 3’ SS is marked on exon 2. genomic+spacer represents modified versions of the genomic splicing reporter with insertions of 36–48 nt spacer sequences on exon 2. (E) rt-PCR of reporter (D) constructs ectopically expressed in S2 cells demonstrates that intron 2 is required for accurate processing of the minigene. Canonical and cryptic products are indicated. (F) Cryptic splicing is detected with the inclusion of 36–48 nt spacer sequences. (G) Schematic of out-of-order splicing and positional requirement of the core-EJC for accurate 3’ SS definition.
Fig 3.
EJC-depletion leads to activation of cryptic 5’ splice sites.
(A) Metagene of cryptic 5’ SS position relative to exon-exon boundaries as density and dot plot. The dot plot indicates splice site scores as calculated via NNSPLICE (see Materials and Methods). Horizontal dashed line depicts threshold for strong 3’ SS, and vertical dashed lines specify 50 nt flanking exon-exon junctions. (B) Sashimi plot depicting HISAT2-mapped sequencing coverage along a portion of CG3632, which has a cryptic 5’ SS that is activated under core-EJC LOF. Junction spanning read counts mapping to the canonical junction are circled, whereas cryptic junction read counts are squared. Note that spliced reads mapping to the cryptic junction are found in eIF4AIII, mago and tsu KD but not the control comparison. (C) Validation of CG3632 cryptic 5’ SS activation (asterisk) in core-EJC, but not btz or lacZ KD conditions. (D) Schematic of CG3632 splicing reporters. Exons 13–15 (introns included) were cloned and subjected to further manipulation. Locations of deleted introns (Δ), as well as a construct lacking all introns (mRNA) are included. The position of the cryptic 5’ SS is marked on exon 14, and was mutated in Δi13+SD mut. (E) rt-PCR of reporter (D) constructs ectopically expressed in S2 cells demonstrates that intron 13 is required for accurate processing of the minigene. Canonical products are indicated by the line and cryptic products by an asterisk.
Fig 4.
EJC-depletion leads to activation of dual cryptic splice sites and resplicing of mRNA. (A) Above: Schematic of resplicing splicing versus alternative resplicing, both of which would yield the same aberrant mRNA product. Below: Sequence of CkIIβ transcript lost due to cryptic splicing. Cryptic 3’ SS activated is highlighted in red, as well as a potential regenerated 5’ SS. Scores listed are generated by NNSPLICE. Conservation across Drosophilid family is shown. (B) Schematic of CkIIβ splicing reporters. Exons 2–4 (introns included) were cloned and subjected to further manipulation. Locations of deleted introns (Δ), as well as a construct lacking all introns (mRNA) are included. For reference, the position of the cryptic 3’ SS and potential 5’ recursive splice sites is marked on exon 3. (C) rt-PCR of CkIIβ reporter constructs in S2 cells demonstrates that introns are required for accurate processing of the minigene. Canonical and cryptic products are indicated. (D) Validation of CG31156 cryptic 5’ SS activation in core-EJC, but not btz or lacZ KD conditions. (E) Schematic of CG31156 splicing reporters with and without introns. Location of potential 3’ recursive splice site on exon 2 is indicated along with conservation scores. (F) rt-PCR of reporter constructs in S2 cells demonstrates that introns are required for accurate processing of the minigene. Canonical and cryptic products are indicated.
Fig 5.
Inherent features of splice sites predispose mRNA resplicing.
(A) Model for mRNA re-splicing. Top, Binding sites of U1 snRNA and U2AF35 define the 5’ SS and 3’ SS, respectively, but also impose constraints on flanking exonic sequences that intrinsically regenerate splice site mimics in a recursive fashion. (Bottom) When located in proximity to another cryptic splice site, these can lead to mRNA resplicing in the absence of the EJC. An example of dual cryptic splice sites with a regenerated 3’ SS is shown, but this can also occur with a regenerated 5’ SS. (B) Comparison of splice site strengths for cases of dual cryptic splice site activation. Cases that contain regenerated 3’ and 5’ splice sites at exon junctions and their structures are schematized and distinguished by red and blue dot. Dashed lines mark thresholds for reasonably strong splice sites. (C) Re-splicing on cDNAs. Constructs bearing cDNA segments of baboon, eIF4G1 and straw were expressed in S2 cells and yielded re-spliced amplicons. Gene specific primers that can amplify both endogenous and ectopic products only show re-splicing from cells expressing cDNA reporters. To verify this directly reflects expression from the cDNA reporters, we tested amplicons that include a vector-specific primer, which yields larger bands relative to the gene-specific primers. These also demonstrate mostly re-spliced products.