Rbfox Proteins Regulate Splicing as Part of a Large Multiprotein Complex LASR

Abstract

Rbfox proteins control alternative splicing and posttranscriptional regulation in mammalian brain and are implicated in neurological disease. These proteins recognize the RNA sequence (U)GCAUG, but their structures and diverse roles imply a variety of protein-protein interactions. We find that nuclear Rbfox proteins are bound within a large assembly of splicing regulators (LASR), a multimeric complex containing the proteins hnRNP M, hnRNP H, hnRNP C, Matrin3, NF110/NFAR-2, NF45, and DDX5, all approximately equimolar to Rbfox. We show that splicing repression mediated by hnRNP M is stimulated by Rbfox. Virtually all the intron-bound Rbfox is associated with LASR, and hnRNP M motifs are enriched adjacent to Rbfox crosslinking sites in vivo. These findings demonstrate that Rbfox proteins bind RNA with a defined set of cofactors and affect a broader set of exons than previously recognized. The function of this multimeric LASR complex has implications for deciphering the regulatory codes controlling splicing networks.

Fig. 1

Rbfox proteins are found in the high molecular weight nuclear fraction. (A) Preparation of the soluble and HMW nuclear fractions. (B) Immunoblot analysis of soluble and HMW fractions from mouse brain. Rbfox1, Rbfox2, and Rbfox3 were detected with an antibody recognizing their nearly identical RRMs.

Fig. 2

Rbfox recruitment to introns in the HMW fraction and to 3′ UTRs in the soluble nuclear fraction. (A) Distribution of Rbfox iCLIP tags in 5′ UTRs, CDS, 3′ UTRs, and introns. Data for mouse brain iCLIP clusters of Rbfox1, 2, and 3 with width ≥2 nt are shown. See Table S1 and Fig. S2 for additional detail. (B) Genome browser view of iCLIP reads mapped to the 3′ portion of the Snap25 gene. iCLIP tracks from the soluble nuclear fraction of whole brain and the cerebellar HMW fraction are aligned as indicated. Significant iCLIP reads from Rbfox1, 2, and 3 are colored as indicated. GCAUG motifs are shown below. See Data File S1 for additional examples.

Fig. 3

Rbfox proteins coprecipitate from the HMW fraction with a distinct set of other RNA binding proteins. (A) Immunopurification of Rbfox proteins from HEK293 nuclear fractions. Soluble and HMW nuclear extracts were prepared from cells stably expressing Flag-tagged Rbfox1, 2, or 3. Proteins were immunoprecipitated with anti-Flag, eluted with Flag peptide, resolved on SDS-PAGE, and stained with SimplyBlue^™. The major interacting proteins are indicated on the right. Flag-Rbfox bands are indicated by asterisks. (−) denotes nuclear fractions from the parental HEK293 cell line that does not express Flag-tagged protein. (B) Sedimentation of protein complexes from mouse brain HMW extract through 10-50% glycerol gradients. Gradient fractions from top to bottom run from left to right. 40S and 60S markers from a parallel gradient are indicated below. Rbfox1, Rbfox2, Rbfox3 and their binding partners are indicated on the right. MyEF2 and hnRNP M are detected with common antibody, as are the ILF3 gene products NF110 and NF90. See Fig. S1 for additional data.

Fig. 4

The 55S Rbfox complex is present in HEK293 cells and is heterogeneous in size. (A) Gradient sedimentation of HMW extract from HEK293T cells transiently expressing Flag-tagged Rbfox1. Proteins detected by immunoblot are indicated as in Fig. 3B. (B) Native gel analysis of complexes separated on glycerol gradients. HMW extracts from HEK293 cells stably expressing Flag-tagged Rbfox3 were fractionated and protein complexes were resolved by native PAGE and probed by immunoblot. Flag-Rbfox3 (top) and hnRNP M (bottom) are indicated on the right. See Fig. S3 for further analyses.

Fig. 5

Rbfox3 can regulate alternative splicing through an hnRNP M binding site. (A) Diagram of the minigene DUP-51M1 and its mutant DUP-51ΔMsite. The M binding site is in Bold. Arrows indicate primers used to detect DUP-51 premRNA. (B) DUP-51M1 or DUP-51ΔMsite were transfected into HEK293T cells with control vector (−) or Flag-Rbfox3 expression vector. Exon 2 splicing was measured by RT-PCR with primers in the flanking exons. The spliced products are indicated (top). Average exon inclusion with standard deviation from four experiments is quantified below. Rbfox3 expression caused a 4.6-fold decrease in DUP-51M1 exon 2 splicing. Statistical significance (red) was measured by unpaired, two-tailed, unequal variance Student's t-test. HnRNP M, Flag-tagged Rbfox3 and U1-70K, as a loading control are indicated (bottom). (C) As in B, cells expressing DUP-51 minigenes and Flag-Rbfox3 were UV-irradiated in vivo and lysed under denaturing conditions to prevent copurification of hnRNP M. RNA:protein crosslinks were immunoprecipitated with anti-Flag. HnRNP M, Flag-Rbfox3, and GAPDH in the lysates (lanes: input) and immunoprecipitates (lanes: Flag IP) were measured by immunoblot (top). DUP-51 pre-mRNA and GAPDH mRNA were detected by RT-PCR (bottom). Amounts of coprecipitated RNA normalized to the Rbfox3 protein over three experiments are graphed, with means, standard deviation, and p-value as in B. See Fig. S4 and S6 for additional analyses.

Fig. 6

Rbfox1 stimulates hnRNP M splicing activity across many exons. (A) Immunoblot of hnRNP M and Rbfox1 in Rbfox2 null HEK293 cells. Rbfox2-knockout cells and derivative cells with Flag-Rbfox1 at the Flp-in locus were grown in doxycycline, transfected with control or hnRNP M-targeted shRNA plasmids and harvested 84 hours post transfection. Relative protein expression over three experiments is graphed below with standard deviation. (B) Comparison of hnRNP M splicing activity in cells expressing Flag-Rbfox1 to that in cells not expressing Rbfox proteins. hnRNP M-regulated exons on this chart were defined in Rbfox1 expressing cells as showing a ∣ΔPSI∣ > 10 (PSI in hnRNP M-expressing cells minus PSI in hnRNP M-knockdown cells) and FDR < 0.5. X-axis: ΔPSI values for these exons in Rbfox1 expressing cells. Y axis: corresponding ΔPSI values in Rbfox1-lacking cells. (C) Rbfox1 splicing activity in hnRNP M expressing cells and hnRNP M-depleted cells is compared as in B. See Table S4 for the rMATS analysis.

Fig. 7

Enrichment of sequence motifs near sites of Rbfox binding. Histogram of pentamer enrichment z-scores within 40 nucleotides of the crosslink sites. iCLIP clusters showing overlap for all three Rbfox paralogs in forebrain or hindbrain HMW fractions were analyzed. Motif enrichments were calculated for crosslink sites less than 40 nucleotides from the nearest GCAUG motif (A) or for sites more than 40 nt from this motif (B). The top 10 percent of z-scores is shaded darker gray. The Rbfox binding GCAUG, UGCAU (red) and similar pentamers (orange) are indicated as dots below and sorted by z-score above. Motifs recognized by hnRNP M (blue), hnRNP C (green), and hnRNP H (gray) are similarly shown. (C) Motif distribution near the crosslink sites. The fraction of sequences with an individual motif aligning at each nucleotide relative to the crosslink site is plotted. Smaller groups of the cluster subsets A and B were analyzed: 1) those containing a GCAUG sequence within 5 nucleotides of the crosslink site (orange lines), and 2) those with a crosslink >100 nt away from the nearest GCAUG (red lines). (D) WebLOGO plots of the sequence adjacent to (U)GCAUG motifs (Crooks et al., 2004). Top: Intronic sequences containing a GCAUG motif within 5 nt of an Rbfox crosslink site. Middle: Sequences containing UGCAUG from introns with Rbfox iCLIP clusters, but >100 nt away from the nearest crosslink site. Bottom: Mean PhyloP placental conservation scores of the Weblogo sequences (orange line, crosslinked to Rbfox; blue line, no Rbfox crosslinking). See Fig. S5 for additional information.