The cellular EJC interactome reveals higher-order mRNP structure and an EJC-SR protein nexus

Abstract

In addition to sculpting eukaryotic transcripts by removing introns, pre-mRNA splicing greatly impacts protein composition of the emerging mRNP. The exon junction complex (EJC), deposited upstream of exon-exon junctions after splicing, is a major constituent of spliced mRNPs. Here, we report comprehensive analysis of the endogenous human EJC protein and RNA interactomes. We confirm that the major "canonical" EJC occupancy site in vivo lies 24 nucleotides upstream of exon junctions and that the majority of exon junctions carry an EJC. Unexpectedly, we find that endogenous EJCs multimerize with one another and with numerous SR proteins to form megadalton sized complexes in which SR proteins are super-stoichiometric to EJC core factors. This tight physical association may explain known functional parallels between EJCs and SR proteins. Further, their protection of long mRNA stretches from nuclease digestion suggests that endogenous EJCs and SR proteins cooperate to promote mRNA packaging and compaction.

Figure 1. Purification of endogenous EJCs and RNA footprints

A. RIPiT schematic. EJC core factors, colored; non-EJC proteins, gray; RNA, black line. B. Co-purification of EJC factors requires their co-expression. Western blot of total extract (TE, lanes 1–4) or FLAG IPs (lanes 5–8) from extracts in which FLAG-eIF4AIII and Myc-Magoh were either co-expressed (+) or expressed in separate cells (−) and mixed prior to IP. C. Confirmation of EJC purification. Western blots of EJC or non-EJC RNA-binding proteins in total extracts (lanes 1–3), control RIPiT (lane 4) or EJC RIPiTs (lanes 5 and 6). For each lane, indicated are the FLAG-fusion protein (box above) and antibodies used for the 1^st and 2^nd IPs (table above lanes 4–6). D. EJC purification under denaturing conditions after formaldehyde crosslinking. Western blots as in C showing protein levels in total extracts (lanes 1 and 2) or RIPiTs (lanes 3 and 4) from formaldehyde-crosslinked cells. SDS and deoxycholate were included during the α-FLAG IP. * indicates IgG_H. E. Length distribution of RNase I-resistant EJC footprints. Base-hydrolyzed synthetic polyU₃₀ RNA (lane 1) or immunoprecipitated RNA fragments from RIPiTs indicated at top (lanes 2–5) were 5′-end ³²P-labeled and separated by denaturing PAGE. Autoradiograph pixel intensity profiles of lanes 2–5 are at right. F. Same as E except samples in lanes 3 and 4 were additionally incubated with RNases A+T1 after α-eIF4AIII IP. See also Figure S1.

Figure 2. The EJC interactome

A. The EJC proteome. Left: Protein names organized into indicated classes. Center: Heat map showing fold-enrichment of each protein in FLAG-eIF4AIII or FLAG-Magoh IPs over its level in a FLAG-only (control) IP (not shown). Black rectangles: protein not detected. All proteins shown were at least 10-fold enriched in FLAG-EJC protein IPs except in the “Other proteins detected” class, which includes <10-fold enriched RNA binding proteins, previously reported EJC-interacting proteins similarly enriched in our EJC and control IPs, and select proteins depleted in the EJC IPs compared to the control IP. Bottom: Heat map log₂ color scale. Right: Bar graph showing protein stoichiometries relative to RBM10 in FLAG-eIF4AIII (blue), FLAG-Magoh (red) or FLAG-only (tan) IPs. Dashed blue line: level of co-purifying Y14 and Magoh in FLAG-eIF4AIII IP. Dashed red line: level of co-purifying eIF4AIII in FLAG-Magoh IP. These dashed lines assembled EJC core stoichiometries in each IP. B. Confirmation of EJC proteomic analyses. Western blots of total extracts or RIPiTs as in Figure 1B showing levels of EJC-interacting factors either detected or not in mass spec analysis. C. SRSF1 is tightly associated with but does not crosslink to EJC core factors. Western blots of indicated proteins in total extracts (lanes 1 and 5) or RIPiTs (lanes 2–4 and lanes 6–8). Native and denaturing α-FLAG IPs were as in Figure 1C and 1D, respectively. D. Two distinct size complexes containing EJC core proteins. Western blots of indicated proteins (right) in total cell extract (TE), input FLAG-eIF4AIII IP sample (IP) and gel filtration fractions of RNase A-treated FLAG-eIF4AIII complexes (lanes 1–17). Peak elution fractions of known MW size markers are indicated at top. * indicates IgG_H (RNPS1 panel) or a cross-reacting unknown protein (SKAR panel). E. RNA footprints of EJC-core containing complexes. Autoradiogram as in Figure 1D showing RNA profiles from FLAG-eIF4AIII:Y14 RIPiT (lane 1) or from indicated gel filtration fractions in C (lane 2: fractions 2 and 3; lane 3: fractions 11, 12 and 13). Pixel intensity profile is on the left. See also Figure S2 and Table S1.

Figure 3. EJC occupancy on endogenous mRNAs

A. Distribution of short EJC footprints (FLAG-eIF4AIII:Y14 and FLAG-Magoh:eIF4AIII), HMW EJC footprints (FLAG-eIF4AIII-HMW), nuclear mRNP footprints or RNA-Seq reads on a representative gene (top half), ENO1, and its spliced mRNA (bottom half). Scale is indicated at top; reads per million at the highest position within each library window are indicated at left. B. Meta-exon analysis of read distribution. Composite plots of distances from the centers of all exon-mapping reads to the 5′ (left panel) or 3′ end (right panel) of each read's parent exon for the indicated libraries. C. Meta-analysis of exons upstream of major or minor spliceosome introns. Composite plots for FLAG-Magoh:eIF4AIII library reads mapping to exons upstream of introns excised by the “major” (~1.7×10⁵ introns) or “minor” (427 introns) spliceosome. In both plots, the fraction of reads at each nucleotide position out of all reads within 100 nt from exon 3′ end was plotted. D-G. Profiles of short EJC footprints, nuclear mRNP footprints and RNA-Seq reads (as in A) on individual exons. Distinguishing features of each exon are indicated at bottom. See also Figure S3.

Figure 4. Variability in cEJC occupancy and contributing factors

A. Signal variability among cEJC sites on individual mRNAs. Smoothed histograms of coefficient of variation (C_v) in the number of reads at each cEJC site (−15 to −31 nts from exon junctions) within a single transcript in a set of 4366 highly expressed transcripts. Inset: parameters used to calculate C_v. B. Fractions of occupied cEJC sites per transcript. Included were all mRNAs with >10 introns and representative transcript RPKM >1. All cEJC sites with a significant peak (probability-value < 0.01) overlapping the −24 position in the FLAG-Magoh:eIF4AIII set were considered occupied. The box-plots show the interquartile range (IQR) of fraction occupied cEJC sites in each RPKM bin. The whiskers are drawn at 1.5 times the IQR and outliers are shown as open circles. Median (black horizontal line) and its confidence interval (notch) are also indicated within each box-plot. C. Range of mappability scores at cEJC-free (white) and cEJC-occupied (gray) sites from all spliced transcripts with RPKM >10. Occupied sites were binned by peak significance (probability-values). Box-plots are as in B. D. Nucleotide frequency plot at cEJC-free and -occupied sites. All sites with mappability score ≥8 from all spliced transcripts with RPKM >10 were used. E. Interquartile range of the minimal free energy of folding for two different 40 nt exonic windows (−50 to −90 nt from exon junctions, left; −10 to −50 nt from exon junctions, right) for cEJC-free (white) and cEJC-occupied (gray) sites from D except that included cEJC-occupied sites had a peak probability-value <10⁻¹². The total number of sites in each dataset is indicated. Box-plots are as in B except only the IQR is shown. The dashed line accentuates statistically significant (indicated by non-overlapping notches) difference in median minimal free energy of cEJC-free and -occupied sites. See also Figure S4.

Figure 5. Functional classes enriched in cEJCs

A. Average cEJC peak-height versus expression level (RPKM) for transcripts of 7536 mRNAs with highly reproducible cEJC peaks. Each dot represents an mRNA with 4 or more exons. Black line: locally weighted non-linear fit (LOESS). The top and bottom 5% deviating furthest from the fit are indicated. B. Functional classes enriched in the higher cEJC occupancy set. The p-values shown were adjusted for multiple hypotheses testing. C. Average cEJC occupancy up- and downstream of alternative spliced exons. AS-NMD (PTC+): Red line: fraction of cEJC occupied sites on successive exons flanking PTC-inducing alternative exon inclusion/exclusion events; Black line: total number of exons at each position for AS-NMD transcript set (RPKM > 1). AS only (PTC-): Same as above, except for a dataset where cassette exon inclusion/exclusion does not generate a PTC.

Figure 6. Non-canonical EJC (ncEJC) sites

A-B. First exon ncEJCs. A. ANKRD11 (top) and OGT exon 1 showing read distributions at canonical (gray) and non-canonical (colored) sites from indicated short footprint libraries. B. Top: Two related sequence motifs enriched under first exon ncEJC peaks. Bottom: Frequency of occurrence of inset motif among three equal size bins of first exon ncEJC peaks (black dots connected by dashed line) or in scrambled peak sequences (box-plots: IQR and median (horizontal line with box-plot) of the frequencies from 1000 iterations; whiskers are at 1.5 times IQR and outliers are as open circles). Peaks were binned according to their probability-values. C-D. Internal exon ncEJCs. C. cEJC and ncEJC sites in KPNB1 exon 10 (top) and BRD2 exon 11 (bottom, color scheme as in A). D. Left: Dendrogram showing related motifs detected in internal exon ncEJC peaks. Right: Same as B, except internal ncEJC peaks were divided into five equal size bins. E. Functional classes enriched in ncEJC read densities. See also Figure S6.

Figure 7. A new view of mRNP structure incorporating EJC-EJC and EJC-SR protein collaboration

A. Co-occupancy of adjacent cEJC sites. Bar-plots showing expected (light gray) and observed (dark gray) frequencies of co-occupied cEJC sites (mappability ≥8) at indicated transcript RPKMs. The two distributions are significantly different (p-value < 1×10⁻¹⁵, chi-squared test). B. Occurrence of ncEJC peaks on cEJC-occupied (red line) and cEJC-free exons (blue line). C. Knockdown efficiency of eIF4AIII quantified by Odyssey imaging. Bar plot: Lane 2:1 ratio for each protein. D. mRNA binding efficiency of proteins upon eIF4AIII knockdown. Western blots showing levels of indicated proteins in total extracts (lanes 1 and 2) and in oligo-dT selected RNAs (lanes 3 and 4) from UV-crosslinked HEK293 cells. Bar plot: Lane 4:3 ratio for each protein. E. The EJC interactome and a new view of mRNP structure: Solid black line: exonic RNA; Dashed black line: a generic intron; Color ovals: proteins enriched more >10-fold in the EJC proteome (Figure 2) listed in descending order of stoichiometry. Black ovals: undetected proteins known to bind to mRNA ends; Green spheres: bridging protein-protein interactions. See also Figure S7.