Pre-mRNA secondary structures influence exon recognition

Abstract

The secondary structure of a pre-mRNA influences a number of processing steps including alternative splicing. Since most splicing regulatory proteins bind to single-stranded RNA, the sequestration of RNA into double strands could prevent their binding. Here, we analyzed the secondary structure context of experimentally determined splicing enhancer and silencer motifs in their natural pre-mRNA context. We found that these splicing motifs are significantly more single-stranded than controls. These findings were validated by transfection experiments, where the effect of enhancer or silencer motifs on exon skipping was much more pronounced in single-stranded conformation. We also found that the structural context of predicted splicing motifs is under selection, suggesting a general importance of secondary structures on splicing and adding another level of evolutionary constraints on pre-mRNAs. Our results explain the action of mutations that affect splicing and indicate that the structural context of splicing motifs is part of the mRNA splicing code.

Figure 1. Scheme Illustrating the Computation of PU Values and the Control Datasets

(A) The figure shows a silencer site (red) bound by the polypyrimidine tract binding protein in the intron upstream of the mouse c-src exon N1 [53] together with 30 nt of its up- and downstream pre-mRNA sequence context. We computed PU values of the splicing motif for all context lengths from 11 nt up to 30 nt. The optimal secondary structure in dot-parenthesis notation is shown below the sequence. To get a single value that measures the single-strandedness of this motif, we averaged these 20 values. (B) Controls: Given an experimentally verified splicing motif (red) and its natural pre-mRNA context, we randomly selected a substring (blue) with the same length from the flanking regions in control 1. In control 2, we copied the verified motif to a randomly selected position. In control 3, we modified the flanking regions by dinucleotide shuffling [28]. The dataset in controls 4 and 5 consists of 10,000 randomly selected motifs (blue) from human exons (shown as open boxes) and introns (lines), respectively.

Figure 2. Influence of mRNA Conformation on Splice Site Selection

(A) Experimental strategy. The structure of the SXN-derived minigenes is schematically indicated. White boxes represent constitutive globin exons, the black box represents the alternative exon where the motifs are introduced (shaded part). The motifs are either single-stranded in a loop (L) or double-stranded in a stem structure (S). Nucleotides that are part of the motif are indicated by blue (L) and red (S) crosses, flanking nucleotides by black dots. The thick arrow indicates the RSV promoter, small arrows the location of the PCR primers. (B) Analysis of the reporter constructs in vivo. L, motif is in a loop structure (single-stranded); S, motif is in a stem structure (double-stranded). A total of 1 μg of each construct was expressed in HEK293 cells and its RNA was analyzed after 24 h of transfection. An ethidium bromide stained agarose gel of representative experiments is shown. The structure of the products is schematically shown on the right. The sequences containing the motifs (underlined) and their PU values were: CD44 ESE: loop (ATCCATGGGGCTGGATGTGACGTACAACCACAATACGTCACATACTTCCTCTCATGA, PU = 0.998), stem (ATGATGGGTATGTGCGTTGCTTCGGCAACCACAACTCATCGCATACTTCCTCTCATGA, PU < 0.001), predicted ESE: loop (ATCCATGGGGCTGGATGTGACGTAACAAGGCATACGTCACATAGCTTCCTCTCATGA, PU = 0.994), stem (CTACCTTGCGCATGATACGCATGCGCAAGGTAGCACTGCATGAGCTTCCTCACGTTT, PU = 0.164), hnRNP A1 ESS: loop (ATCCATGGGGCTGGATGTGACGTAGTAGGGTATACGTCACATAGCTTCCTCTCATGA, PU = 0.976), stem (CTACCCTACGCATGATACGCATGCGTAGGGTAGCACTGCATGAGCTTCCTCACGTTT, PU = 0.126), predicted ESS: loop (ATCCATGGGGCTGGATGTGACGTAGTAAGTGAATACGTCATATCTTACCTCTCATGA, PU = 0.857), stem (ATCCAGTAAGCTACGCTCCGATGCGTAAGTGAGTCCGCTCACTTACGCATCTCATGA, PU < 0.001). (C) Statistical analysis. The average percent exon inclusion of three independent transfection experiments for S and L constructs is: CD44 ESE, 70% versus 7%; predicted ESE, 75% versus 34%; hnRNP A1 ESS, 15% versus 37%; predicted ESS, 2% versus 81%. Error bars indicate the standard deviation from at least three independent experiments.

Figure 3. Selection on the Structural Context of Predicted Exonic Enhancer, Silencer, and Splicing-Neutral Motifs

The differential fraction of enhancers and silencers that are selected for single-stranded contexts is shown at the y-axis. The differential fraction is the difference between the fraction of enhancers/silencers and the fraction of neutral motifs (background); positive values indicate a preference for single-strandedness. To infer selection for single-strandedness, we compared exons with decoy regions, pseudo exons, and intron flanks (x-axis). The left side shows the differential fraction of enhancers and silencers that are selected to be single-stranded in exons (more GC rich flanks and higher PU values in exons), the right side the differential fraction in decoy regions and intron flanks (more GC rich flanks and higher PU values in decoy regions and intron flanks). Selection for single-strandedness in pseudo exons cannot be inferred since no enhancer and no silencer motif has more GC rich flanks in pseudo exons. Asterisks above or below bars indicate the significance in a Fisher's exact test comparing enhancers/silencers with neutral motifs; asterisks above brackets indicated the significance comparing enhancers with silencers (***, p-value < 0.0001; **: p < 0.01).

Figure 4. mRNA Secondary Structures Can Explain Experimental Observations

(A) A structural change explains exon skipping by the 25G->A mutation in the human CFTR exon 12. NI scores [36] for the wild-type (wt) and the 25G->A mutation (highlighted gray) indicate the strength of ESE (blue, positive score) and ESS (red, negative score) motifs. Hexamers with an NI score >0.8 and <−0.8 are considered to be strong ESEs and ESSs, respectively; the remaining hexamers are “splicing-neutral” (open boxes). A positive PU value difference (PU_wild-type minus PU_mutant) indicates a higher single-strandedness for the respective hexamer in the wild-type exon and vice versa. Bars give the NI score or the PU value change for the hexamer starting at this position. Compared to the wild-type sequence, the 25G->A mutation does not change ESE or ESS motifs. However, the ESE at position 16 becomes more double-stranded in the 25G->A mutant, and the ESSs at position 23–25 become more single-stranded, explaining the decrease in exon inclusion from 80%–25% [43]. (B) Due to their structural conformation, a predicted ESE is not active in the exon 8 of the rat beta-tropomyosin gene. The three purine-rich ESE candidates with NI scores of 1 characterized in [54] are shown in blue letters. Bars for hexamers that overlap these three investigated 9-mers are highlighted gray. While the first motif is highly double-stranded, the second and the third motif have at least one hexamer in a more single-stranded context. This explains experimental results as the mutation of the second and to a lesser extent of the third motif affects splicing, while mutating the first motif shows no effect [54]. Thus, observed splicing effects correlate with the single-strandedness of these three motifs. Average PU values using the context lengths 11–30 nt are shown in (A) and (B).