MBNL1 and RBFOX1 co-regulate alternative splicing events transcriptome-wide through a conserved buffering mechanism

  1. Melissa Hale
  2. Joseph Ellis
  3. Ryan Meng
  4. Sunny McDaniel
  5. Amy Mahady
  6. Stacey Wagner
  7. Jared Richardson
  8. Ona McConnell
  9. Eric Wang
  10. J. Andrew Berglund

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    A major question in the field of alternative splicing regulation is how information from multiple, independently-acting splicing factors is integrated into a single signal that informs the spliceosome to include or skip an exon. This study addresses the extent of co-regulation of alternative splicing events by the two RNA binding proteins RBFOX1 and MBNL1. The work has implications for how the effects of perturbing one splicing factor can be buffered by the action of a second RNA binding protein, with broad relevance in development and disease. Additional work would provide more support for a generalized mechanism underlying regulation alternative splicing events.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Alternative splicing (AS) is controlled by cis -regulatory elements recognized by networks of trans -acting factors. Here we investigate modes and mechanisms of AS co-regulation by MBNL1 and RBFOX1, two RNA binding proteins (RBPs) critical for developmental AS transitions. We generated two cell models that express each RBP under separate inducible promoters. Transcriptome-wide categorization of the impacts of RBFOX1 expression on MBNL1 splicing revealed a common co-regulatory mode through which RBFOX1 buffers MBNL1 dose-dependent splicing regulation by reducing the total range of exon inclusion or exclusion. Minigene mutational analysis and in vitro binding experiments suggest that this buffering mechanism occurs through a shared cis -regulatory element previously unidentified as critical for MBNL1-dependent activity. Overall, our studies define a conserved co-regulatory mechanism through which RBFOX1 and MBNL1 can fine-tune and provide redundancy for AS outcomes. These studies indicate overlapping use of RNA motifs with potential implications for when activity of RBPs is disrupted.

Article activity feed

  1. eLife

    Evaluation Summary:

    A major question in the field of alternative splicing regulation is how information from multiple, independently-acting splicing factors is integrated into a single signal that informs the spliceosome to include or skip an exon. This study addresses the extent of co-regulation of alternative splicing events by the two RNA binding proteins RBFOX1 and MBNL1. The work has implications for how the effects of perturbing one splicing factor can be buffered by the action of a second RNA binding protein, with broad relevance in development and disease. Additional work would provide more support for a generalized mechanism underlying regulation alternative splicing events.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

  2. eLife

    Reviewer #1 (Public Review):

    In this study, Hale and colleagues study the extent of co-regulation of alternative splicing events by the two RNA binding proteins RBFOX1 and MBNL1. The authors use inducible expression to look at the relationship between the dose of each factor and their impact on exon inclusion. The results support a model where for a subset of splicing events, RBFOX1 can act to buffer MBNL1 dose, and in one case the authors provide more detailed mechanistic evidence that this likely occurs through a consensus RBFOX binding site (UGCAUG) that also serves as a low-affinity binding site (YGCA) for MBNL1.

    In general, I found the current study interesting and timely, highlighting a key challenge in our ability to understand and predict alternative splicing outcomes. Specifically, as the current study demonstrates, differing concentrations of RNA binding proteins that co-regulate alternative splicing events can often lead to diverse and complex regulatory outcomes, which we still do not fully understand.

    A strength of this study is that it highlights the importance of understanding alternative splicing regulation as a product of interactions between multiple RNA binding proteins, where distinct concentrations of these factors can lead to diverse outcomes on splicing patterns. The work demonstrates this concept through a combination of more detailed mechanistic work on a reporter gene and through the use of in vitro binding assays. It also begins to extend these ideas through transcriptome-wide analysis of splicing patterns in cells expressing differing concentrations of two RNA binding proteins.

    I do feel that in its current form, the study would benefit from additional genome-wide analysis to further strengthen the generality of the model put forward by the authors to explain buffering of MBNL1 dosage via the co-expression of RBFOX1.

    Overall however, this study has increased our appreciation for the importance of moving away from studying RNA binding proteins in isolation and without consideration of other cis elements and trans factors.

  3. eLife

    Reviewer #3 (Public Review):

    In this manuscript, the authors address an important problem in understanding how alternative splicing outcomes are determined. In many cases, the influence of multiple factors accounts for the relative abundance of different splice variants and these splicing factors and their target transcripts frequently change in expression over the course of a developmental program. But how are the contributions from these multiple trans-acting regulators combined to produce a binary outcome, such as inclusion or skipping of an exon? The splicing factors examined in this manuscript are the well-studied Muscleblind (MBNL) and RBFOX families, each of which has three separately encoded paralogous loci, and which change in expression during developmental processes to control splicing of large sets of exons. These factors are particularly important in the context of muscle development, and MBNL proteins are sequestered within nuclear foci driven by expanded-repeat encoded transcripts containing canonical MBNL binding sites which is a causal factor in myotonic dystrophies. Thus it is important to understand how these factors interact to control splicing in normal and disease contexts. The interactions between RBFOX and MBNL, as well as another family of splicing regulators known as CUGBP or CELF proteins in co-regulating splicing events have all been extensively documented in the literature, including in some cases examples of overlapping binding sites.

    The novel regulatory interaction that the authors set out to characterize here began with the identification of a non-canonical, overlapping binding site for MBNL embedded within an RBFOX motif controlling the splicing of exon 11 in the insulin receptor (INSR) transcript. RBFOX showed a 'buffering' response curve on minigene splicing, where at low concentrations of MBNL protein RBFOX significantly enhanced splicing, while the effect of MBNL was dominant when it was present at increased concentrations. The effects of each protein could be dissected by mutations pinpointing each motif, and recombinant MBNL exhibited a higher binding affinity for the hybrid site compared to RBFOX RRM, consistent with the dynamics seen in the minigene splicing assay. Through use of an innovative dual-inducible transgene system to independently manipulate expression of MBNL and RBFOX1, alternative splicing was measured first by performing a multipoint dose-response curve on individual endogenously expressed genes by RT-PCR, then examining the extreme end-points of the curve in a genome wide manner using RNA-seq.

    A strength of the paper is in the use of these innovative cell lines with independently-titratable MBNL and RBFOX protein expression to enable dose-response testing. This system is used in a two-step process in which the coincidence of altered splicing events detected in a simplified end-point RNA-seq experiment with the dose-dependent validation of endogenous splicing responses allows the inference of complex dose-response curves from a much simpler RNA-seq experiment. However, the main weakness is the generalization of the proposed mechanism, which affects the significance of the conclusion. Simply put, although the 'buffering' type interaction on which they focus could was found among a larger group of exons in the genomewide data, these accounted for only about 1/3 of the affected splicing events. More importantly, there was little evidence provided for the role of hybrid binding sites among these buffered splicing events, with only ~30% of the exons having an identifiable motif like the one found in INSR. While the INSR minigene experiments nicely define a mechanism responsible for the splicing of that exon, and the genomic data hint at the possibility of other such interactions, the conclusions of this manuscript in its current form do not represent a major, generalizable advancement in our understanding of splicing regulation.

  4. eLife

    Reviewer #2 (Public Review):

    In this work, Hale et al. performed systematic analysis how dosage dependent splicing regulation by MBNL1 is affected by different levels of RBFOX1 expression. Starting with analysis of INSR exon 11 using minigene splicing reporters, the authors observed that the magnitude of MBNL dependent splicing is large when RBFOX1 is low, and it is reduced when RBFOX expression is high, a phenomenon denoted "buffering" co-regulatory mechanism. By mutagenesis and in vitro binding assays, the authors proposed that the RBFOX1 binding motif UGCAUG can directly bind MBNL1 through the imperfect UGCA sequence, and an increasing level of MBNL1 can outcompete RBFOX1 binding. To generalize this observation, this study elegantly generated two cell lines. In each cell line, MBNL1 and RBFOX1 are controlled by inducible promoters, which can be independent controlled by DOX/ponA titration. Exons co-regulated by MBNL1 and RBFOX1 were identified by RNA-seq analysis and were binned based on how RBFOX1 level affects the magnitude of MBNL1-dependent splicing. The authors observed that exons subject to buffering co-regulation is most abundant. The competitive binding of RBFOX1 and MBNL1 through YGCAUG appears to account for a subset of these exons.

    Overall, systematic investigations of combinational splicing regulation by multiple RBPs are lacking and this study provided a very nice experimental system with several interesting observations. I hope the comments below can be helpful to improve the manuscript:

    1. Previous studies reported that MBNL1 and RBFOX1 frequently regulate exon inclusion or skipping in the same direction. The authors found the magnitude of MBNL1-dependent splicing changes is smaller when RBFOX1 level is high, which was denoted "buffering" co-regulation. This is certainly correct at sementic level, but this observation does not directly imply whether there is mechanistic coordination between the two RBPs or if there is, what is the nature of such coordination.

    a) This is because the study measured the magnitude of splicing regulation by delta_PSI, which is bounded by the baseline inclusion level. For example, when the baseline exon inclusion is 0.1 and the inclusion can readily increase to 0.4 when MBNL1 is induced if the exon is strongly regulated by MBNL1. However, if RBFOX1 increases the baseline exon inclusion level to 0.7, then the maximal possible delta_PSI upon MBNL1 expression will be 0.3, even when there is no direct coordination between the two RBPs. Therefore, it is not precise to claim the regulation is non-additive based on this observation.

    b) The authors might refer to Baeza-Centurion et al. 2019 Cell 176:549, which addressed this issue. There might be more rigorous theoretical framework if the authors want to argue whether the regulation is additive or coordinated.

    1. Through mutagenesis analysis of the UGCAUG elements into UCGAUG (MUT1, which disrupted the RBFOX binding site) or into CGCUUG (MUT2, which disrupted the RBFOX binding sites while creating a MBNL binding site), as well as gel shift experiment, the authors proposed that RBFOX1 and MBNL1 can compete in binding to the UGCAUG element. These data are consistent with the theory, but do not exclude other possibilities. For example, the juxtaposition of RBFOX1 and MBNL1 can stabilize the binding of the two proteins to RNA. The reduction in MBNL1-dependent splicing in MUT1 can be explained by destabilization of RNA-mediated protein-protein interactions. The restoration of splicing by MUT2 could be due to the fact that the newly created CGCU increases the affinity of MBNL1 binding to clustered YGCY motif sites, so that it is less dependent on stabilization by interacting with RBFOX1.

    2. The observation of MBNL1 binding to UGCAUG containing sequences in vitro, or the observation that it can outcompete RBFOX1 binding at high concentration is not very surprising. This result does not prove such binding/competition occurs in cells at endogenous protein level. However, it is somewhat surprising that MBNL1 can outcompete RBFOX1 RRM in binding to UGCAUG at equivalent concentration. Is it because the substrate sequence used in gel shift has three YGCY like sites (GGCU, UGCA, and UGCG)? If this is the case, how this situation can be generalized? It might be helpful to determine Kd from the gel shift and compare the values with those reported in the literature.

    3. Comparison of RBFOX1 motifs in different sets of exons. Before the analysis presented in Figure 3 E,F, it will be helpful to examine motif enrichment in RBFOX-dependent vs. independent exons in different regions (upstream intron, exon, and downstream intron), which will provide a positive control, as the expected patterns are very well established in the literature. In their analysis, the authors should distinguish upstream/downstream intron depending on RBFOX-dependent inclusion or skipping.

    Some of the requests above might go beyond the immediate scope of the study, but nevertheless the authors should probably at least discuss how these issues affect the interpretation of the data.

  5. Version published to 10.1101/2021.04.20.440659v1 on bioRxiv