Optimal antisense target reducing INS intron 1 retention is adjacent to a parallel G quadruplex

Abstract

Splice-switching oligonucleotides (SSOs) have been widely used to inhibit exon usage but antisense strategies that promote removal of entire introns to increase splicing-mediated gene expression have not been developed. Here we show reduction of INS intron 1 retention by SSOs that bind transcripts derived from a human haplotype expressing low levels of proinsulin. This haplotype is tagged by a polypyrimidine tract variant rs689 that decreases the efficiency of intron 1 splicing and increases the relative abundance of mRNAs with extended 5' untranslated region (5' UTR), which curtails translation. Co-expression of haplotype-specific reporter constructs with SSOs bound to splicing regulatory motifs and decoy splice sites in primary transcripts revealed a motif that significantly reduced intron 1-containing mRNAs. Using an antisense microwalk at a single nucleotide resolution, the optimal target was mapped to a splicing silencer containing two pseudoacceptor sites sandwiched between predicted RNA guanine (G) quadruplex structures. Circular dichroism spectroscopy and nuclear magnetic resonance of synthetic G-rich oligoribonucleotide tracts derived from this region showed formation of a stable parallel 2-quartet G-quadruplex on the 3' side of the antisense retention target and an equilibrium between quadruplexes and stable hairpin-loop structures bound by optimal SSOs. This region interacts with heterogeneous nuclear ribonucleoproteins F and H that may interfere with conformational transitions involving the antisense target. The SSO-assisted promotion of weak intron removal from the 5' UTR through competing noncanonical and canonical RNA structures may facilitate development of novel strategies to enhance gene expression.

Figure 1.

Location of SSOs in the human proinsulin gene. (A) Schematics of the INS reporter and its mRNA products. SSOs are shown as black horizontal bars below exons (numbered boxes) and below intron 1 (line); their sequences are in Supplementary Table S1. Start and stop codons are denoted by arrowheads. Canonical (solid lines) and cryptic (dotted lines) splicing is shown above the primary transcript; designation of cryptic splice sites is in grey. SSOs targeting intron 1 segments del4-del7 are shown in the lower panel. (B) mRNA isoforms (numbered 1–6) generated by the INS reporter construct. Description of isoforms that do not produce proinsulin is in grey.

Figure 2.

SSO-induced inhibition of INS intron 1 retention. (A) Cotransfection of the INS reporter construct (IC D-F) with the indicated SSOs into HEK293 cells. Spliced products described in Figure 1B are shown to the right. Bars represent percentage of intron 1-containing isoforms relative to natural transcripts (upper panel) or percentage of splicing to the cryptic 3' splice site of intron 2 relative to the total (lower panel). Error bars denote SD; sc, scrambled control; SSO-, ‘no SSO’ control. Final concentration of SSOs was 1, 3, 10 and 30 nM, except for SSO6 and SSO8 (10 and 30 nM). (B) SSO21-mediated promotion of intron 1 splicing in clones lacking the cryptic 3' ss of intron 2. RNA products are to the right. (C) A fold change in SSO21-induced intron 1 retention in transcripts containing and lacking the cryptic 3' ss of intron 2. The final concentration of SSO21 was 30 nM in duplicate transfection. Designation of the reporter constructs is at the bottom.

Figure 3.

INS SSOs targeting cryptic 3' splice sites. (A) Activation of cryptic 3' ss of intron 2 (cr3' ss+126; Figure 1A) by SSO6 and promotion of exon 2 skipping by SSO8. Concentration of each SSO was 2, 10, 50 and 250 nM. SSOs are shown at the top, spliced products to the right, reporter at the bottom. (B) A predicted stable hairpin between the authentic and cryptic 3' ss of INS intron 2. Bases targeted by SSO6 are denoted by asterisks and predicted splicing enhancer hexamers (listed to the right) are denoted by a dotted line. (C) SSO4 does not prevent activation of cryptic 3' ss 81 base pairs downstream of its authentic counterpart (cr3' ss+81) in cells depleted of U2AF35 but induces exon skipping. The final concentration of each SSO in COS7 cells was 5, 20 and 80 nM. The final concentration of the siRNA duplex U2AF35ab (29) was 70 nM. The reporter was the same as in panel A.

Figure 4.

Optimization of the intron retention target by antisense microwalk at a single-nucleotide resolution. (A) Location of oligoribonucletoides. Microwalk SSOs and oligos used for CD/NMR are represented by horizontal black bars below and above the primary transcript, respectively. Intron 1 sequences predicted to form RNA G-quadruplexes are highlighted in grey. Microwalk direction is shown by grey arrows; winner oligos are highlighted in black. A box denotes a single nucleotide polymorphism reported previously (20). (B) Intron retention levels of each microwalk SSO in two cell lines. Error bars denote SDs obtained from two independent cotransfections with reporter IC D-F.

Figure 5.

Biophysical characterization of RNA secondary structure formation. (A) Far-UV CD spectrum at 25°C for CD1 (19-mer) and CD2 (20-mer) RNAs, revealing ellipticity maximum at 265 and 270 nm, respectively. (B) ¹H NMR spectra of CD1 and CD2 recorded at 800 MHz and 298 K showing characteristic groups of resonances from H-bonded G bases. (C) Sigmoidal CD melting curves for the two RNAs showing a transition mid-point at 56.8 ± 0.2°C and 69.0 ± 0.45°C, respectively. The two curves have been displaced slightly from each other for clarity. (D) The proposed parallel quadruplex structure with two stacked G-tetrads connected by short loop sequences for CD1 (top panel). Predicted hairpin structures for CD2 are shown at the bottom panel. G→C mutations are in red.

Figure 6.

Conformational quadruplex/hairpin transitions involving the antisense target. (A) Schematic equilibrium between hairpin (black) and quadruplex (dark blue) structures proposed to form within the G-rich motif encompassing oligoribonucleotide CD3. CD4 contains a CC→UU mutation (in red). (B) The NMR spectrum in the 9–15 ppm region reveals imino proton signals corresponding to hydrogen bonded bases. The signals between 10 and 12 ppm are characteristic of Hoogsteen hydrogen bonded Gs within a G-tetrad (red box), while signals > 12 ppm are indicative of Watson–Crick A-U and G-C base pairs within hairpin structures (black box). In CD3, hairpin H1 is significantly populated, but mutations in CD4 destabilize H1 making H2 the major species, with both in equilibrium with the quadruplex structure. (C) Mfold predictions of two possible hairpins, consistent with the NMR data. (D) Reduction of intron retention upon destabilization of the hairpin structure by the CC→UU mutation. Error bars denote SD of a duplicate experiment with reporter IC D-C. Del5, the IC D-C reporter lacking segment del5 (Figure 1A); M1, a reporter containing two substitutions (Supplementary Table S2) to destabilize both the G-quadruplex and the stem-loop.

Figure 7.

Identification of proteins that interact with pre-mRNAs encompassing the antisense target for intron retention. (A) Intron retention levels for wild type and mutated reporter constructs (IC D-C) following transient transfections into HEK293T cells. Mutations are shown in Supplementary Table S2. RNA products are to the right. The presence of predicted RNA quadruplexes, hairpins H1/H2 and the upstream and downstream C₄ run are indicated below the gel figure. Error bars denote SDs obtained from two replicate experiments. (B) Intron retention levels of tested RNAs correlate with their predicted stabilities across the antisense target. (C) Western blot analysis of a pull-down assay with antibodies indicated to the right. NE, nuclear extracts; B, beads-only control; AV3, control RNA oligo containing a cytosine run and a 3' ss AG (7). The sequence of CD5 RNA is shown in Figure 4A.

Figure 8.

Splicing pattern of quadruplex-rich and -poor minigenes upon DHX36 depletion. (A) Schematics of reporter constructs. Predicted quadruplexes are denoted by black rectangles; their densities are shown in Table 1. Exons (boxes) are numbered; forward slash denotes shortening of F9 intron 3 (24). The F9 and TSC2 minigenes contain branch point substitutions c.253–25C and c.5069–18C, respectively, that impair splicing (24). Cr5' ss-104; cryptic 5' ss 104 upstream of authentic 5' ss of intron 2. (B) Immunoblot with antibodies against DHX36. sc, scrambled siRNA; c, untreated cells. Error bars are SDs of two transfection experiments. (C–E) Intron retention and exon skipping of the indicated reporters. The final concentration of DHX36 siRNA was 50 nM. RNA products are shown schematically to the right. Error bars are SDs of two transfection experiments.