Our focus this fall is on RNA : Protein interaction methods and analysis. We are reading papers to learn about classic and novel techniques for identifying which RNA binding proteins interact with an RNA and the position they interact with. We are also interested in learning about bioinformatics methods that predict RNA:protein interactions or recognize conserved sequence or structural motifs.
We are reading:
Sequence, Structure, and Context Preferences of Human RNA Binding Proteins by Dominguez, D., et al… Yeo, G. W., Graveley, B. R., & Burge, C. B. (2018). Molecular Cell, 70(5), 854-867.e9. https://doi.org/10.1016/j.molcel.2018.05.001
RBPmap: A web server for mapping binding sites of RNA-binding proteins. Paz, I., Kosti, I., Ares, M., Cline, M., & Mandel-Gutfreund, Y. (2014). Nucleic Acids Research, 42(W1), 1–7. https://doi.org/10.1093/nar/gku406
Analysis of RNA–protein networks with RNP-MaP defines functional hubs on RNA. Weidmann, C. A., Mustoe, A. M., Jariwala, P. B., Calabrese, J. M., & Weeks, K. M. (2020). Nature Biotechnology. https://doi.org/10.1038/s41587-020-0709-7
Clusters of mammalian conserved RNA structures in UTRs associate with RBP binding sites. Gadekar VP, Munk AW, Miladi M, Junge A, Backofen R, Seemann SE, Gorodkin J. NAR Genom Bioinform. 2024 Aug 9;6(3):lqae089. doi: 10.1093/nargab/lqae089
We attended the 2024 RNA Society Meeting in Edinburgh and had a great time talking about RNA biology and meeting people. Austin Herbert, Allie Randazza and Debarati Majumdar presented posters on their graduate research projects. Congratulations to Debarati for receiving an RNA Society Travel Award!
Lackey lab in at the closing dinner in the National Museum of Scotland.
Posters
Exploring alternative polyadenylation isoforms of DNMT3A
Debarati Majumdar, Austin Herbert, Lela Lackey
The DNMT3A (DNA methyl transferase) gene is known for its pivotal role in de novo DNA methylation and gene regulation. DNMT3A’s role in DNA methylation is well‐documented. However, DNMT3A produces various mRNA isoforms through alternative polyadenylation (APA) and alternative splicing. Each of these mRNA isoforms may be regulated differently in ways that could significantly impact the origins and progression of various DNMT3A‐associated conditions, such as acute myeloid leukemia (AML) and neurodevelopmental disorders. Our hypothesis is that specific mRNA isoforms of DNMT3A, resulting from alternative polyadenylation (APA) events, may display variable stability, unique subcellular localizations, and differential impacts on downstream gene expression, ultimately influencing the pathogenesis of disorders. As a preliminary dataset, we used GTEx, 1000 genomes and UCSC Genome browser to identify mRNA isoforms and check their levels in various tissues. Once we confirmed DNMT3A mRNA isoform variants, we constructed plasmids containing the 3’UTR of the DNMT3A gene linked to the nano luciferase reporter gene to quantitatively assess the 3’UTR’s effect on gene expression by monitoring nanoluciferase activity. Subsequently we will employ various molecular biology assays to investigate RNA stability including (Actinomycin D inhibition followed by qPCR) RNA localization (Fluorescence In Situ Hybridization; Cellular fractionation followed by qPCR) and translational efficiency (Polysome profiling) of these isoforms. Understanding the distinct roles of DNMT3A mRNA isoforms will not only advance our knowledge of the regulatory mechanisms underlying epigenetic control, but also lay the groundwork for isoform‐specific therapeutic interventions in DNMT3A‐associated diseases.
High‐throughput workflow to study the impact of mutations on RNA structure in the adenine riboswitch Alexandra Randazza, FNU Jiamutai, Vijay Shankar, Lela Lackey
Directly linking individual RNA structures with function is difficult. We aim to create a workflow to determine the impact of specific RNA structures on RNA function. To do so, we incorporate mutations at specific positions in an RNA construct to alter individual structural elements and analyze the impact of altering those structures on function. We are developing a high‐throughput method to analyze hundreds to thousands of variants at once. For our initial tests, we are using the adenine riboswitch. The bacterial adenine riboswitch alters its structure upon binding adenine to facilitate the initiation of translation of an adenine deaminase protein. Purely computational structure analysis of mutant adenine riboswitch was unable to detect restoration of structure with rescue mutations. Our initial tests included six sequence variants that are expected to either destroy or maintain the aptamer’s structure. The ability of these variants to respond to adenine was first analyzed individually. In the future, we will pool them together. We have created a computational pipeline to separate each variant in the pool and generate structural data to measure the impact of the sequence variation on structure. We plan to expand this method to larger pools of mutants. To link mutations and structural changes to function, we will quantify the appropriate response of the adenine riboswitch to adenine. This will allow us to highlight structures that are essential in the adenine riboswitch and use this method to understand the role of RNA structure in other systems.
SF3B1, a core component of the spliceosome involved in branch point recognition and 3’ splice site selection is frequently mutated in human hematopoietic malignancies. Mice and zebra fish with conditional SF3B1 knock‐in mutations develop macrocytic anemia. A hallmark of SF3B1 mutation is an increased use of upstream cryptic 3’ splice sites (C3SS) in a broad number of genes, a finding that is recapitulated across multiple isogenic and patient cell types. Current studies suggest the common SF3B1 K700E mutation results in gain of function, allowing mutant SF3B1 spliceosomes to recognize cryptic 3’ splice sites normally inaccessible to the wild‐type protein. We asked whether precursor RNAs with SF3B1 K700E sensitive splice sites are structurally different from control cryptic 3’ splice sites. Utilizing publicly available bulk RNA sequencing data, we define a core set of 73 cryptic 3’ splice sites shared between isogenic SF3B1 K700E and myelodysplastic syndrome patient cell lines. Through a combination of experimental and in‐silico methods, we assess precursor RNA secondary structure in this subset of 3’ splice sites sensitive to cryptic mis‐splicing in SF3B1 mutant backgrounds. Occurring at a mean distance of 28 base pairs upstream, SF3B1 sensitive C3SS are significantly closer to their paired canonical splice sites than what is observed among control C3SS (mean 78 bp). Both SF3B1 sensitive and control C3SS contain a distinct polypyrimidine tract and a strong consensus AG splice site motif. Experimentally based structure models of intron‐exon junctions around SF3B1 sensitive C3SS reveals that cryptic sites structurally mimic their canonical counterparts. Despite structural mimicry at the immediate splice sites, cryptic 3’ splice sites lack downstream structural signatures observed in the exons of their paired canonical splice sites. We anticipate that SF3B1 mutation sensitive C3SS contain distinct structural patterns compared to control C3SS. These observations will yield insights on the modulation of cryptic 3’ splice site choice by precursor RNA secondary structures susceptible to mis‐splicing in SF3B1 mutants. Finally, we will use these findings to investigate a generalized mechanistic role of precursor RNA structures in contributing to regulation of cryptic 3’ splicing.
Our focus this summer is on RNA structure analysis methods. We are reading papers to understand analysis methods that quantify how similar two structures are to one another. Specifically, we are focusing on reading papers that describe methods that allow us to identify conserved or enrichment substructures across different RNAs, such as within precursor RNAs with alternatively spliced intron-exon junctions.
By creating an Ago2 construct containing a neural specific promoter and GFP tag, researchers were able to selectively create a GFP-Ago2 protein only in mouse neural cells that could also be the target of RNA immunoprecipitation (RIP). Enrichment of neural specific miRNAs by qPCR confirmed the specificity of this construct, and sequencing of bound miRNAs found some well known transcripts to be incorporated by Ago2, including Snca and Itgb1. The relative enrichment of bound transcripts was analyzed, with 2177 gene targets confirmed by bioinformatic analysis. Of these enriched genes, three (ex. miR-124, miR-125, and let-7) displayed a high degree of binding to Ago2 and conserved binding sites, thus confirming the efficacy of their RIP-seq and construct.
Specifically focusing on miR-124 (which is highly expressed in hippocampal cells), researchers created a new construct, one with what’s called a miR-sponge, theoretically capable of “absorbing” miRNAs and inhibiting their functions. All within the same construct, they performed RIP on GFP-Ago2 and confirmed the specificity of their sponge construct in neural cells and specificity for miR-124 via qPCR. Taking it a step further, they sequenced the sponge RIP samples and analyzed the bound mRNA targets. What they found was a substantial drop in the number of bound transcripts, about 497 less, and that reduced enrichment in Ago2 binding likely accounted for such a drop off in bound transcripts and an increase in mRNA expression level. Among the top gene networks controlled by 300 of the lost 497 transcripts – nerve development, metabolism, and transcription. Finally, to confirm their construct could more broadly be applied to other mRNA networks, researchers inhibited miR-125 (expressed in glial and neural cells) with the same sponge construct, finding it too lost bound transcripts, about 384 out of 2177. Altogether, Malmevik et al. provided key insight into the mouse hippocampal miRNA targetome with an innovative approach that has broad applications for the characterization of miRNA functions and targets.
11/19/21 – In our pre-Thanksgiving journal club we discussed a bioRxiv paper from the Li and Staley laboratories at the University of Chicago. In this work, Zeng and co-authors describe CoLa-Seq (co- transcriptional lariat sequencing). CoLa-Seq is a new technique that detects lariat containing precursor RNAs and splicing by-products to identify branchpoints. The authors call the two lariat containing species they detect with CoLa-Seq NLIs (nascent lariat intermediates) and ELIs (excised lariat introns). To obtain reads corresponding to NLIs and ELIs, the authors enriched for precursor RNAs by isolating chromatin. They further selected RNA in the process of splicing by decapping and degrading linear RNA, leaving only RNA protected by the 2’-5′ lariat linkage. Using CoLa-Seq, Zeng, et al. identified the largest number of branchpoints. In addition, CoLa-Seq provides a technique to continue branchpoint identification in other cell lines and under additional conditions. A reasonable protocol for branchpoint identification is important as, even with this study, many branchpoints remain unmapped. Branchpoint selection is important for recognition of the 3’ splice site and understanding of alternative splicing.
In addition to describing a new technique and documenting an extensive number of branchpoints, Zeng, et al., analyzed their results to make several novel biological insights. They analyze the timing of splicing by measuring the number of nucleotides past the 3’ splice site in NLI reads. One thing that surprised me from their timing data is the variability at the same intron. In addition, they found that splicing can happen in-order, out-of-order and concurrently. These three different splicing modes occur in different ratios in most transcripts. As spliced intermediates are rare and collected from a large population of cells, I wonder how the state of the cell and the level of transcript affect splicing timing and the order of splicing. Interestingly, splicing did not seem to depend on transcription of the downstream exon, even for long introns, as would be expected for the exon-definition model of splicing. Zeng, et al., also used extensive modeling to try and understand what elements control splicing timing and order. As an RNA structure lab, we were most intrigued by the role of GC content in splicing timing! However, as GC content captures both structural and sequence motifs, it is still too early to say what role RNA structure has in regulating splicing. We look forward to seeing the final version of this manuscript in press.
Using miRNA-seq and enrichment analysis, the authors pinpointed six miRNA targets that were dysregulated in both their MM cohort and R-ISS cohort. Survival analysis via Kaplan-Meier survival curves and Cox regression analysis discovered the only candidate with a significant association with worse overall and progression-free survivability: miR-181a. GO Ontology analysis documented this miRNA to be enriched in several key biological processes, including B cell apoptosis and response to glucocorticoids. Next, survival analyses were performed to evaluate the prognostic utility of miR-181a via Kaplan-Meier curves and Cox univariate and multivariate regression analysis. The result: high miR-181a levels were associated with a shorter life expectancy in both overall survivability (OS) and cancer specific survivability, confirmed by an independent cohort. Multivariate Cox regression analysis, incorporating variables such as age, gender, LDH, and cytogenetics, further confirmed miR-181a overexpression and its association with post-treatment progression.
Next, researchers evaluated if miR-181a could serve as a co-prognostic indicator of MM along with R-ISS, high-risk cytogenetics, and 1st line therapy response. Indeed, miR-181a proved invaluable in assisting prognosis with all three existing indicators: in combination with R-ISS, it provided better stratification for patients’ survivability; with high-risk cytogenetics, an increase in inferior disease outcome; and with therapy response, it could predict the risk for disease progression and a worse survival outcome. Finally, decision curve analysis (DCA) compared the net benefit of treating patients evaluated for high miR-181a levels/R-ISS, and response to 1st line therapy to those only evaluated for the latter two. Not surprisingly, there was a significant net benefit in treating those with elevated miR-181a levels compared to those not evaluated for it.
To summarize, Papadimitriou, et al. indicated a key miRNA component of MM prognosis that, along with existing markers and indicators, demonstrates value in significantly improving patients’ prognosis, predictability of disease progression, and survivability outcome.
11/5/21– Our ability to solve problems and treat diseases are limited only by our knowledge of the problem, our imagination, and our tools. While previous studies have found ribozymes to alter mRNA, and therefore protein expression, via small molecules or short nucleotide oligomers, they suffer in the ability to detect longer RNA sequences and/or providing significant fold changes in reporter expression. Our lab read a paper on “RNA-responsive elements for eukaryotic translational control”. In this paper, Zhao et al. developed eToeholds as modular regulators for eukaryotic mRNA translational control which can be altered by the presence of a specific trigger RNA. This is a rather astounding tool, as the module itself could be altered, in theory, to any RNA you wish to detect with relatively high specificity (~40-50 base pairs). The eToeholds themselves can be altered not only for detecting specific RNAs, but also at the IRES for optimization in different systems.
One of their most sensitive eToehold designs utilized the CrPV IRES and had pairing between the 8-6 regions on the secondary structure which had an mKate reporter. After additional optimization to reduce basal expression of the modulated RNA, this eToehold design would produce a 16-fold change in protein expression when triggered with the GFP mRNA in HEK293. Test showed that the eToeholds made with a different target in the same IRES design would not mispair to the different targets and another design developed to allow 5’ caps on these modulated RNAs to allow similar levels of sensitivity. However, when IRES systems were changed to human ones such as from hepatitis C, polio virus, or enterovirus 71 with the same eToehold procedure, the sensitives were found to be much weaker to compromise with higher modulated mRNA magnitudes. But, when testing the claims that the eToehold tool could strongly detect infections, cell states, and cell types, the results were significant and supported the value of this tool. One point to add before the conclusion is that this tool specifically works in the cytosol, so nuclear retained RNAs would not be able act as a trigger; however, that still could be used for test to determine nuclear retention. In conclusion, the eToehold switches produced show strong potential to detect and target cells that express specific mRNAs, whether they be viral, endogenous, or exogenous, and have relatively high specificity in doing so such that it could be used either for diagnostics, therapeutics, or research.
This week in journal club we discussed the paper “Nonsense-Mediated RNA Decay is a Unique Vulnerability of Cancer Cells Harboring SF3B1 or U2AF1 Mutations” by Cheruiyot et al. Here, the authors conduct a crispr/cas9 knockout based forward genetic screen to identify previously unknown promoters of nonsense-mediated decay. Top hits from this screen included known NMD-associated factors UPF1, SMG6, and RUVBL1, and previously reported splicing factors not typically associated with NMD, SF3B1 and U2AF1. Both SF3B1 and U2AF1 contain reoccurring hotspot mutations involved in different types of cancers and thus were targeted for their unconventional roles in NMD. Cell lines created with these splicing factor hotspot mutations showed a distinct sensitivity to NMD inhibition, as cells exhibited various cancer related phenotypes such as reduced survivability, increased chromosomal aberrations and R-loop formations, and reduction in DNA replication and transcription fork progression speed. This paper sets the ground work for determining the specific interactions between splicing factors and the NMD pathway.
10-22-21: For this week’s journal club, I elected to look at an AGO1 paper that was recently published in J Cell Biol by Acuña, et al. and relevant to many of our lab’s interests in ongoing experiments within our lab. This paper checked a lot of the boxes: timely, discussing argonaute, proposing functional roles for argonaute, utilizing immunocytochemistry, and incorporating large-scale experimental and historical ChIP-Seq.
The experiments performed by Acuña, et al. were interesting first and foremost because argonaute has been widely studied for its role in small RNA-mediated post transcriptional processing and in transcriptional gene silencing. However, they propose an alternate function for argonaute-1: estradiol triggered transcription in human cells. They used ChIP and ChIP-Seq on transfected MCF-7 cells to locate and identify ERα binding sites within the cells relative to argonaute-1 (AGO1), which they found in significance. ERα binding motifs were present throughout the cells expressing AGO1, indicating that they are indeed prominently linked in some capacity.
To further interrogate this connection, they treated these AGO1 transfected MCF-7 cells with estradiol. Upon enrichment, signals of AGO1 expression increased greater than MCF-7 cells that were untreated. Additionally, they detected no change in subcellular localization of AGO1 during E2 treatment. To complete their follow-up, they proceeded to knockdown the function of AGO1 to react to E2, showing a marked decrease in ERα activity within the knockdown cells. They were also able to rescue the functionality of these proteins and restore levels of production to near those of endogenous levels. More interestingly, when they conducted their experiments showing that estrogen increased in response to E2’s presence in relation to AGO1, they also found that when they silenced the AGO1’s ability to incorporate RNA, that the relation continued to exist. This implies that AGO1 and ERα are binding together on the chromatin and that it is activating co-transcriptionally. They were also able to demonstrate that AGO1 is preferentially enriched at active transcriptional enhancers. All of this points to the possibility that AGO1 does indeed act in an estrogen-dependent manner as a transcription coactivator, at least as the authors suggest. The most marked criticism of this paper is simply a wish this lab would like to have granted: investigating this same system in the rest of the family of argonaute proteins.
Equally important as this conclusion, however, is the implication that proteins may harbor multiple and various functions. In the case of this AGO1, miRNA’s were shown to not even be necessary for its binding on chromatin with ERα. Previous studies indicating the importance of AGO1’s interactions with small RNA’s would have and did entirely miss this potential function. This gives rise to a compelling question: how to identify multiple functions of a protein and, more importantly, what determines the significance and impact of any given functions it may have out of multiple? Is AGO1 more important to transcriptional gene silencing or coactivating with ERα? In what ways would you determine the significance and relevance of AGO1’s influence in both of those mechanisms relative to each other? These are questions we may hope to answer soon, and wrestle with our principles of microbiology and genetics in doing so.
9-17-21: RNA silencing is a fascinating mechanism in eukaryotes by which the cell’s own RNA regulates the expression of genes at the transcriptional, post-transcriptional, and translational level. These RNA molecules belong to the subclass of small RNAs (sRNAs) and include small-interfering RNAs (siRNAs), micro RNAs (miRNAs), and piwi-interacting RNAs (piRNAs). These sRNA, along with an Argonaute protein, join to form the RNA-induced silencing complex (RISC).
The process by which miRNAs are incorporated into the RNA-induced silencing complex (RISC) was of great interest in our review in the publication “Argonaute-3 activates the let-7a passenger strand microRNA”, published in RNA Biology in 2013. The authors, J. Winters and S. Diederichs, detailed the unusual affinity one of the Argonaute family proteins, Argonaute 3 (Ago3), had for the passenger strand of Let-7a. When miRNAs are synthesized, they are processed into a duplex, consisting of a guide strand and passenger strand. It has been known that the guide strand is the one incorporated into RISC to assist with cleaving target RNA, while the passenger strand is sent for degradation. In addition, Ago3 was thought to be catalytically inactive, overshadowed by its cleavage-inducing family member Argonaute 2 (Ago2). Nevertheless, Winters and Diedrichs found that Ago3 not only incorporated the passenger strand of let-7a, called let-7a-3p, but it was also catalytically activated by this strand.
Analysis by northern blotting, with confirmation from qRT-PCR, confirmed that overexpression of Ago3 led to an increase in the expression of let-7a-3p and an increase in the ratio between passenger strand (let-7a-3p) and guide strand (let-7a-5p). Luciferase assays detailed the overexpression of Ago3 in another crucial way: an increase in let-7a-3p’s silencing activity, as evidenced by a decrease in luciferase expression upon let-7a-3p binding to Ras-related GTP binding protein RAB10. Winters and Diedrichs then created recombinant let-7a by switching its terminal loop with another miRNA, miR-193a, to see if any structural component affected Ago3’s incorporation of the miRNA. The recombinant hairpin loop did not affect incorporation, suggesting the let-7a duplex was responsible for it instead. Next, the authors questioned whether Ago3 would be affected by base-pairing and thermodynamic stability at the 5’ end of let-7a, a property by which the less stable strand is usually incorporated into RISC. However, Ago3 did not significantly increase incorporation of let-7a-3p when mutated strands were co-transfected with it, suggesting a departure from previous knowledge. Finally, the Winters and Diedrichs questioned the domains in Ago3 responsible for incorporating let-7a-3p into RISC, and after co-transfecting constructs of Ago3, chimeric Ago3 (domain swapped with Ago1), and let-7a, they determined by qRT-PCR the binding of let-7a-3p by the PAZ and MID domains, which bind the 5’ and 3’ end of miRNA, respectively.
To conclude, Winters and Diedrichs discovered a novel mechanism by which Argonaute 3, previously thought to be catalytically dead, could be activated by and incorporate a passenger strand of miRNA, let-7a-3p, increasing its expression and silencing activity, while detailing the significance of miRNA structure, stability, and sequence on incorporation into RISC. The implications of this research certainly lead to the possibility of other miRNA passenger strands being utilized to some capacity, though there remains little current research. Another possible candidate for further research is the role Ago3 has in regulating the tumor suppressor abilities of let-7a, which targets RAS and HMGA2. Such findings would shed insight into how RNA silencing can regulate oncogenes or what miRNA dysregulation may implicate.
