Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Dec 6;72(5):849-861.e6.
doi: 10.1016/j.molcel.2018.08.044. Epub 2018 Oct 11.

Extensive Structural Differences of Closely Related 3' mRNA Isoforms: Links to Pab1 Binding and mRNA Stability

Free PMC article

Extensive Structural Differences of Closely Related 3' mRNA Isoforms: Links to Pab1 Binding and mRNA Stability

Zarmik Moqtaderi et al. Mol Cell. .
Free PMC article


Alternative polyadenylation generates numerous 3' mRNA isoforms that can vary in biological properties, such as stability and localization. We developed methods to obtain transcriptome-scale structural information and protein binding on individual 3' mRNA isoforms in vivo. Strikingly, near-identical mRNA isoforms can possess dramatically different structures throughout the 3' UTR. Analyses of identical mRNAs in different species or refolded in vitro indicate that structural differences in vivo are often due to trans-acting factors. The level of Pab1 binding to poly(A)-containing isoforms is surprisingly variable, and differences in Pab1 binding correlate with the extent of structural variation for closely spaced isoforms. A pattern encompassing single-strandedness near the 3' terminus, double-strandedness of the poly(A) tail, and low Pab1 binding is associated with mRNA stability. Thus, individual 3' mRNA isoforms can be remarkably different physical entities in vivo. Sequences responsible for isoform-specific structures, differential Pab1 binding, and mRNA stability are evolutionarily conserved, indicating biological function.

Keywords: CLIP-READS; DMS; DREADS; Pab1; RNA folding; RNA structure; alternative polyadenylation; mRNA half-life; mRNA isoform; yeast.


Figure 1.
Figure 1.. DREADS schematic
(A) Exponentially-growing yeast cells are treated with DMS, which methylates exposed A and C residues in cellular RNAs. During library construction, reverse transcriptase (RT) stalls at DMS-modified positions (blue asterisks) as well as at naturally occurring stall sites on mRNA molecules. Each resulting library fragment contains information on both the specific poly(A) isoform (downstream end) and the endpoint of the RT reaction (upstream end). Paired-end deep sequencing thus makes it possible to link multiple RT stops with individual 3’ isoforms. Subtraction of the naturally occurring (non-DMS) RT stops in the untreated control allows for precise mapping of DMS-reactive, accessible A and C residues in each 3’ isoform. (B) Example of a library fragment subjected to paired-end sequencing. Gray segments at the fragment ends represent adapter sequences added during library construction. The R1 sequence read encompasses the junction between the poly(A) tail remnant and the last nt of genomically-encoded 3’UTR sequence; this allows mapping of the specific 3’ isoform. The R2 sequence read supplies information about where the reverse transcription reaction stopped (naturally or at potential DMS-modified site) on the same mRNA molecule during library construction.
Figure 2.
Figure 2.. Closely-related mRNA isoforms can possess radically different DMS reactivity profiles
(A)Examples of DMS reactivity profiles (limited to 200nt upstream of the poly(A) tail for display clarity) of biological replicates and closely related (2 nt apart) neighboring isoforms. Reactivities are scaled relative to the isoform’s highest observed reactivity value (100). Top: A and C reactivities for two biological replicates of the RPS31 mRNA “+48” isoform arising from polyadenylation 48 nt downstream of the ORF. Bottom: A and C reactivities (average of two replicates) for the same +48 isoform compared to those of the isoform ending 2 nt away, at 50 nt downstream of the ORF. The indicated coordinates for both isoforms are defined with respect to the same +48 position and hence represent the identical nucleotides in the two isoforms. (B)Genome-wide percentile distribution of Pearson correlation coefficients for DREADS isoform reactivity profiles. Gray bars: correlations for the same isoform’s reactivity profiles in two biological replicates. Black bars: correlations of every isoform’s reactivity profile with every other same-gene isoform’s reactivity profile. (C)Percentage of poorly correlating DREADS isoform reactivity profiles increases as a function of inter-isoform spacing. (D)Genome-wide percentile distribution of Pearson correlation coefficients for isoform reactivity profiles (SHREADS). Gray bars: correlations for the same isoform’s reactivity profiles in two biological replicates. Black bars: correlations of every isoform’s reactivity profile with every other same-gene isoform’s reactivity profile (E) DREADS and SHREADS identify a highly overlapping subset of structurally distinct isoform pairs. Same gene isoform pairs exhibiting substantial structural differences (ΔR ≥ 0.3) are far more likely to be identified by both methods than by chance alone (P = 10−51).
Figure 3.
Figure 3.. Single point mutations can alter an isoform’s structural profile
DMS reactivity profile comparison of the wild-type SOD1 +71 (SOD +71wt) isoform to two mutants possessing an identical cleavage/polyadenylation site (SOD1 +71A66U, SOD1 +71UGG). Reactivities for each isoform are scaled relative to that isoform’s highest observed reactivity value (100). Top: A comparison of A and C reactivities for SOD +71wt and the single substitution mutant SOD1 +71A66U reveal modest differences in reactivity profiles. Bottom: Extensive differences in reactivity profiles between SOD +71wt and the triple substitution mutant SOD1 +71UGG.
Figure 4.
Figure 4.. DMS reactivity profiles are influenced by cellular factors
(A) Percentage of poorly correlating mRNA isoform reactivity profiles (R < 0.3) in in vivo (from Figure 2C, black bars) and in vitro (gray bars) refolded samples as a function of the distance between isoform endpoints. (B) Left panel: DMS reactivity profile comparison of identical mRNA isoforms from native D. hansenii and from a D. hansenii chromosome segment in an S. cerevisiae host strain. ΔR is the difference by which an isoform’s reactivity profiles correlate in biological replicates compared to the correlation of that isoform’s reactivity profiles in the native species (D. hansenii or K. Lactis) versus when transplanted into S. cerevisiae (Rreplicates – Rcross-species). Blue bars represent isoforms for which the interspecies Pearson correlation is worse (three different ΔR thresholds shown) than the same-species biological replicate correlation. Red bars represent isoforms for which the interspecies correlation is better than the biological replicate; these bars represent the experimental error. Right panel: the same analysis with identical K. lactis isoforms from native and S. cerevisiae-YAC strains.
Figure 5.
Figure 5.. Relationship of isoform-specific protein binding and structure
(A) Structural similarity of same-gene isoform pairs (compared to structural similarity of same-isoform biological replicates) when protein binding sites are present in common sequences (gray bars) or in the extra sequence unique to the longer of the two isoforms (black bars). ΔR is a measure of the difference in reactivity profiles of two same-gene isoforms compared to the biological replicates of those isoforms (see methods); small ΔR values (≤ 0.1) indicate that the two same-gene isoforms are structurally similar, while larger ΔR values (> 0.3) are indicative of widespread structural differences. (B) The sequence unique to the longer of two 3’ isoforms is more likely to be evolutionarily conserved when the isoforms are structurally different. For pairs of isoforms with similar folding (left, ΔR < 0.1), the extra sequence unique to the longer isoform is more frequently non-conserved (median PhastCons score in the sequence element < 0.33, black bar) than conserved (median PhastCons score in element > 0.67, gray bar). For structurally dissimilar isoform pairs (right, ΔR > 0.3), this sequence is more frequently conserved than non-conserved. (C) CLIP-READS schematic. Exponentially growing cells are irradiated with UV light (254 nm) to crosslink RNA-protein complexes, after which the protein of interest is immunoprecipitated from the extract. Immunopurified protein:mRNA complexes are treated with proteinase K to digest bound proteins and mRNAs are captured on oligo(dT) beads. This mRNA population is subjected to isoform-specific deep sequencing by the 3’ READS method (Jin et al., 2015), which allows identification and quantification of the individual mRNA 3’ isoforms bound to the protein of interest (in this example, Pab1) in vivo. An input sample consisting of a portion of the same extract is also subjected to deep sequencing in parallel to the CLIP-READS sample. Isoform frequencies obtained for the CLIP-READS sample are then divided by frequencies obtained in the input sample to arrive at isoform-specific protein binding levels. (D) Isoform-specific Pab1 binding exhibits considerable variation. Expression-corrected relative Pab1 binding levels for isoforms in two representative genes, with the value of 1 representing the median expression-corrected Pab1 binding for the entire dataset of > 25,000 isoforms. (E) Correlation of differential Pab1 binding with extensive structural differences in same-gene isoforms ≤ 20 nt apart.
Figure 6.
Figure 6.. Relationship of isoform properties with stability
(A) A two-part structural motif consisting of a predicted unstructured region (−24 to −18) and a double-stranded region in the poly(A) tail (+8 to +16) correlates with increased isoform stability. (B) Relationship between half-life, evolutionary conservation and predicted structural character in sequence windows near the poly(A) site. +1 is the first A of the poly(A) tail; −1 is the last nt before the poly(A) tail. Blue bars denote the subset of isoforms with unstructured sequences (≥ 6 predicted unstructured residues) in the indicated sequence window, while red bars denote isoforms in which this sequence is structured (≥ 6 predicted structured residues).
Figure 7.
Figure 7.. Potential models for differential structures of 3’ mRNA isoforms in vivo
The model depicts a subset of all same-gene isoforms that exhibit extensive structural differences in vivo; for simplicity, many common structural parts are not shown. Identically positioned residues with varying DMS reactivities are circled and numbered. For each panel, a representation of the DMS reactivity pattern across the five numbered positions is depicted in the upper right-hand corner, with DMS-reactive positions in green and unreactive positions in black. Left panel: In vitro, extra sequences (red N’s) or mutations (blue asterisks) near the poly(A) site have no effect on RNA folding or isoform reactivity profiles. In an environment devoid of RNA-binding proteins, the three isoforms possess identical structures. Right panel: In vivo, the presence of mutations or additional sequences near poly(A) sites results in differential association of Pab1 and/or other RNA-binding proteins that alter RNA-RNA interactions within the mRNA that in turn can influence protein binding, leading to a cascade of differential structural and binding effects over an extensive region upstream of the poly(A) tail. For isoforms that are not closely spaced, the cascade of structural effects may also stem from intrinsic differences in RNA-RNA interactions.

Similar articles

See all similar articles

Cited by 4 articles

Publication types

MeSH terms

LinkOut - more resources