Distal Alternative Last Exons Localize mRNAs to Neural Projections

Abstract

Spatial restriction of mRNA to distinct subcellular locations enables local regulation and synthesis of proteins. However, the organizing principles of mRNA localization remain poorly understood. Here we analyzed subcellular transcriptomes of neural projections and soma of primary mouse cortical neurons and two neuronal cell lines and found that alternative last exons (ALEs) often confer isoform-specific localization. Surprisingly, gene-distal ALE isoforms were four times more often localized to neurites than gene-proximal isoforms. Localized isoforms were induced during neuronal differentiation and enriched for motifs associated with muscleblind-like (Mbnl) family RNA-binding proteins. Depletion of Mbnl1 and/or Mbnl2 reduced localization of hundreds of transcripts, implicating Mbnls in localization of mRNAs to neurites. We provide evidence supporting a model in which the linkage between genomic position of ALEs and subcellular localization enables coordinated induction of localization-competent mRNA isoforms through a post-transcriptional regulatory program that is induced during differentiation and reversed in cellular reprogramming and cancer.

Figure 1. Cellular fractionation and sequencing reveals mRNA isoforms associated with neurite localization

A) Cells are grown on top of porous membranes, allowing growth of neurites through the pores, enabling fractionation. B) Soma and neurite lysates from primary cortical neurons were immunoblotted for beta-actin, a marker of both soma and neurite, and histone H3, a marker of soma. C) LRs in two cell lines. Differentially enriched genes in both cell lines are shown in blue. D) Schematic showing differential isoform enrichment. E) The fraction of the expressed alternative isoform pairs that were significantly differentially enriched between soma and neurite fractions for different classes of alternative isoforms. At left, the inclusion isoform is pictured in blue, and the exclusion isoform is pictured in red. F) Distribution of Δψ values of different isoform classes. Boxes indicate 25th and 75th percentiles, lines indicate 5^th and 95^th percentiles. See also Figure S1, Table S1.

Figure 2. 3′ UTRs of neurite distal ALE isoforms confer neurite localization

A) The subcellular localization of RNA from a reporter gene (Fig. S2E) containing the proximal (left) or distal (right) alternative last exon from the indicated gene was monitored using RNA FISH. The fluorescent protein product of the reporter is colored in green while probes against the RNA are shown in red. B) Quantification of FISH results. Values are the mean intensity across the projection in the red channel divided by the mean intensity in the green channel. * p < 0.05; ** p < 0.01. C) qRT-PCR analysis of neurite versus soma expression of proximal and distal reporter genes (mean and SD of 6 replicates). See also Figure S2.

Figure 3. Distinctive properties of 3′ UTRs of neurite-localized distal ALEs

A) Left: Lengths of UTRs of neurite-localized distal ALEs identified in N2A and CAD cells, proximal ALEs of the same genes, and distal and proximal ALEs not associated with localization. Middle: UTRs from the indicated regions were aligned with homologous regions from human, rat, dog, and cow. RNA secondary structure minimum free energies (MFE) were then calculated for successive 100 nt windows of the alignment using RNAalifold. For each alignment, the median MFE was recorded. Right: PhastCons scores of 30-way alignments of UTRs from the indicated classes of ALEs. The score for each UTR was defined as the mean PhastCons score for all basepairs within the UTR. B) Increased PSI values following differentiation of CAD cells indicate preferential accumulation of neurite distal ALE isoforms but not of nonlocalized distal ALE isoforms. See also Figure S3, Table S2.

Figure 4. Mbnl motifs are enriched and conserved within localized distal ALEs, and Mbnl promotes RNA localization to projections

A) Enrichment of 6mers (hexanucleotides) between neurite distal and soma proximal UTRs and conservation of 6mers between mouse and human. Conservation is measured by a z-score representing the number of SD above the mean conservation of 50 control 6mers matched for CpG and C+G% content, in neurite distal UTRs. B) Metagene analysis of Mbnl motif frequency across UTRs from indicated classes (excluding the last 50 nt to exclude PAS motifs). These classes correspond to those defined in Figure 2. C) Relative CLIP-seq cluster densities in the UTRs of distal and proximal ALEs. Control UTRs consist of randomly sampled UTRs from all ALE events that were not differentially localized. Error bars are the standard error of random samplings of controls. D) Change in LR upon Mbnl knockdown for genes that were (blue) or were not (pink) localized in the control sample. E) Mbnl motif frequency across 3′ UTRs of ALEs as a function of the change in localization of that ALE in cortical neurons from Mbnl1 / Mbnl2 DKO mice. ALEs were classified by their ΔΔψ values as described in Supplemental Methods. Error bars represent +/− SEM. See also Figure S4, Table S3.

Figure 5. ALE and tandem UTR isoforms are generally coordinately regulated

A) For each row, the fraction of alternative isoform events that displayed an increase in relative abundance of the inclusion isoform in the differentiated or reprogrammed sample relative to its corresponding control was calculated. For each class of isoforms red indicates a shift in the differentiated/reprogrammed sample towards the proximal AFE, ALE or tandem UTR isoform or toward exon skipping, while blue corresponds to a shift toward the distal AFE, ALE or tandem UTR isoform or exon inclusion. The number inside the boxes corresponds to the number of significantly changing alternative isoforms in each sample. B) As in A, but comparing cancer samples to matched non-tumor controls. All samples significantly biased toward distal or proximal by chi-square test (P < 0.05) except those marked NS. C) The fraction of tandem UTR and ALE events displaying shifts towards distal PAS following the knockdown of RNA binding proteins in K562 cells. Gene names with significant shifts toward distal or proximal shown in bold. D) Correlation and clustering of isoform types indicated in A and B. E) Generally, development and differentiation result in a shift toward the inclusion of more distal ALEs. Conversely, becoming cancerous results in a shift toward more proximal ALEs.