Category: Journal Club

Written by Luke Hatfield

8/20/21 – The world of RNA structure is bubbling with potential and a necessary desire to find accurate and reliable methods of interpreting not only two dimensional RNA structure, but three dimensional as well. Some studies have already been completed using complex in silico simulations, kinetics based physics models to interpret 3D structure conformation using nucleotide based SHAPE-MaP probing, and now direct crosslinking of folding conformations using SHAPE-JuMP from the Weeks lab at UNC, Chapel Hill. This method is accomplished using a SHAPE compound developed specifically for this method, called trans-bis-isatoic anhydride (TBIA). This TBIA acts like a typical SHAPE compound in that it reacts with a nucleotide’s 2’-OH group. However, unlike other SHAPE compounds, it is a double ended molecule. This means that when the primary oxygen bonds the molecule can then use another available oxygen on the other side of the compound to attach to a nearby nucleotide, which can be paired or unpaired. This then links the two strands of RNA that are in close proximity to each other. The RNA is then crosslinked so that the specific conformation the structure is taking is tethered using TBIA. Reverse transcription is then performed, utilizing a uniquely mutated RT-C8 reverse transcriptase, that is capable of “jumping” the TBIA, inducing mutations where the nucleotides were linked. These mutations are then read out (to an impressive depth of 500000 reads) and compared to standard sequencing reads and SHAPE-MaP reads in order to identify the locations of the mutations from SHAPE-JuMP. SHAPE-JuMP provides important details about which nucleotides were crosslinked and skipped, showing interaction sites of the TBIA, while the other standard reads serve to help fill in the gaps and build the 3D structures. This provides unique structural information about what sections of the RNA in question is in proximity to each other, giving insight into the 3D conformations the RNA are taking when in vitro.

This method is still fairly new and there are a fair few improvements to be made. However, this introductory experiment is promising and gives a fair amount of credence to the ability of this technique and related compounds to improve upon the current understanding and methodology in uncovering RNA three dimensional structure. As our lab currently has two significant projects underway regarding RNA structure and form, observing and learning from techniques and experiments like these is crucial in moving forward with innovative and creative approaches to the questions that are at the forefront of RNA research. The primary drawback to utilizing this technique for our research projects is in the limited scope of RNA’s that it can probe. The P546 domain, VS riobzyme, RNase P, and Group II intron interrogated in this paper are no longer than 412 nucleotides at the longest (Group II intron) and 158 nucleotides long at the shortest (P546 domain). The current focus of our research involves RNA lengths averaging around three kilobase pairs, with some alternative structures reaching up to seven or even eight kilobase pairs. It is unclear how well the TBIA and SHAPE-JuMP would be able to handle forming its bonds it relies on to JuMP at strands of this length – but the ideas and concepts presented here will help us with our experimental setup and planning in the future.

Week 5 (2/12/2021) – We discussed a computational paper that connects theory with experimental SHAPE data by T. Hurst in the Chen lab. Single-nucleotide resolution, chemical probing experiments (like SHAPE) provide information on the accessibility of a nucleotide. Clearly the 3D structure of the RNA will impact how accessible each nucleotide is to the reagent, yet how extensive this information is and how to use it to create actual 3D models is a difficult problem that the field has been struggling with for a long time. One problem is the lack of known 3D structures to compare against. Hurst and Chen look at 17 crystal structures of small, highly structured RNAs and then analyze SHAPE reactivities from the same RNAs for what 3D information is encoded in the data. In a comparison approach to distinguish structures folded “randomly” from those informed by reactivity, the guided structures perform better. This is a clever approach that demonstrates that SHAPE reactivities do contain information about tertiary structure. However, practical limitations on the number and types of structures they have to work with prevent this technique from truly taking SHAPE data to 3D just yet. I would have liked to see whether only SHAPE data generated with the reagent 1M7 are suitable or other common reagents like 5NIA and NAI also have 3D information. Can you employ different reagents to collect different 3D information and combine that to get a better picture? What about reagents like DMS that probe base accessibility, but can provide 3D interactions in a different manner (PAIR-MaP)? The second half of this paper concerned the skew in reactivity by 1M7 toward certain base contexts. Hurst and Chen did a great job explaining how reactivity is influences by the 5′ and 3′ neighboring bases and how 1M7 does not react equally in all contexts. I think this is a reminder that experimental structure data has its own biases and both computational and experimental approaches are needed to understand both 3D and 2D RNA structure.

Upcoming (2/19/21) – We will discuss Gawroński, et al.’s paper on light sensitive RNA structures!

Week 4 (2/5/2021) – We discussed the Iserman, et al. paper from my alma mater, UNC Chapel Hill, and the Weeks and Gladfelter labs. This publication describes potential phase transition by the SARS coronavirus nucleocapsid protein and SARS-CoV-2 genomic RNA. We also read and considered the McSwiggen, et al., 2019 publication from the Tjian and Darzacq group that provides a critical look at the evidence for phase transitions, including how we should evaluate phase transition experiments and whether phase transition has reached the standard of biological relevance. Austin Herbert, Mitzy Garner and Emily Alonzo, graduate applicants to the Clemson Center for Human Genetics, joined us for our discussion. Iserman, et al., clearly show that the nucleocapsid (N) protein forms droplets that are concentration and temperature dependent. Addition of specific RNAs, including long RNAs and the 5′ leader region, to N-protein increases droplet formation. The authors then address the tricky problem of how primarily non-specific interactions between N-protein and RNA govern phase transition. Some evidence they include is that the N-protein interacts different with the 5′ leader region than it does with other RNAs that do not promote phase transition. However, the corona virus genome is very large and this study only investigates a fraction of the complete sequence. Much of the claims are speculation that will be interesting to follow-up on, but do not just form a complete picture yet. For example, the authors bring up the concept of therapeutics to disrupt phase transition, however, their evidence for biologically important phase transition on corona virus packaging is thin. In addition, they only test three chemicals that are known to have other effects on the cell (like kanamycin). Impressively, this substantial work on RNA structure, protein binding and phase transition was published within the same year that the novel SAR-CoV-2 was identified. It will be interesting to see how some of these introductory concepts are carried out in the future.

Upcoming (2/12/2021) – We will discuss a computational paper that connects theory with experimental SHAPE data by Hurst in the Chen lab.

1/22/2021 – We discussed Yu et al.‘s recently published paper suggesting that RNA structural intermediates may be recognized by RNA processing machinery and that the transitional landscape of folding should be considered for regulation and function from the Lucks lab. Realistic molecular dynamic simulation of RNA molecules remains difficult, however, simulations can provide insight into the biology of RNA structure. Yu, et al.’s paper addresses some very tricky questions on co-transcriptional folding and how structures evolve during transcript to result in their final mature form. The R2D2 (reconstructing RNA dynamics from data) method uses experimental SHAPE data during impaired transcription to capture structures of progressively longer transcripts. The authors took this to the next step and performed molecular dynamic simulations on these structural stages. They delve deep into the maturation of the SRP RNA. The highlight I took from this paper is that key nucleotides that influence the final RNA structure may not be those that form a long hairpin, but the nucleotides that stabilize or destabilize intermediary structures. More work is needed to understand how the lessons from SRP apply to other RNAs and how to identify key nucleotides. In particular, I’d be interested in learning more about the conservation of nucleotides that participate in intermediary structures and whether variants in intermediary positions are associated with human disease.

Upcoming (1/29/2021) – First data presentation in the lab! Luke Hatfield will present an overview of his projects since he joined in early December.

1/15/2021 – We read Marinus et al.‘s publication on developing novel SHAPE reagents to probe cellular RNA structure and improve signal-to-noise ratio in structural data from the Incarnato lab. There are a variety of RNA secondary structure probes, but all have caveats. For example, dimethyl sulfate (DMS) is highly reactive and easily penetrates cell membranes, but it only reacts well with A and C nucleotides (and it is highly toxic!). Most sensitive chemical probes like 1M7 and NIA result in better structure information from all four bases, but do not easily penetrate cell membranes and require extensive effort to develop good signal from background noise. Marinus et al. address this problem by synthesizing additional reagents based on NIA. The best of the new reagents, 2-aminopyrimidine-3-carboxylic acid imidazolide (2A3), only differs from NIA by replacing a methyl group with an amino group. However, this subtle difference makes this molecule more reactive with RNA, dramatically improving the signal-to-noise ratio, and allows the molecule to penetrate cell membranes more effectively. I would have liked to see the comparison with 5NIA, which I have used in lab and can permeate some types of mammalian cells. However, I think an additional reagent that improves RNA secondary structure experimental data is extremely useful and hope to test it out soon!

Upcoming (1/22/2021) – We will discuss Yu et al.‘s publication suggesting that RNA structural intermediates may be recognized by RNA processing machinery and that the transitional landscape of folding should be considered for regulation and function. This paper was recently published from the Lucks lab.

1/8/2021 – For our first meeting of 2021, we discussed Liu, et al.’s 2021 paper on the nuclear structurome from the Ding lab. The authors used Arabidopsis seedlings (dunked in RNA secondary structure chemical probing reagents) to develop RNA structures on cytoplasmic and nuclear RNAs. At the core of this paper is the idea that in vivo and in vitro RNA structures differ significantly and, by extension, that precursor mRNA and mature mRNA also have dramatically different structures. This would imply that the structures of mature mRNAs are not accurate representations for understanding RNA binding protein recognition, particularly in the context of splicing and alternative polyadenylation. The authors also support the ability of SHAPE reagents (NAI) to detect protein binding by comparing the reactivity of “deproteinized” RNAs to RNA protein complexes. By comparing the reactivity of the nuclear and cytoplasmic RNAs, Liu et al. detect a 2-nucleotide “structural motif” of unstructured RNA at the -1 and -2 positions around the 5′ splice site (within the exon). This “motif” corresponds with better recognition of the splice site and, in their experiments, are the only nucleotides required to be unpaired within the entire U1 snRNP recognition motif (spans -3 to +6). They did not detect a clear structural signal at the 3′ splice site, but did see an increase in single-stranded signal around the branch point. Likewise, strong polyadenylation signals have an overall high reactivity, suggested unpaired nucleotides. Is the difference between nuclear and cytoplasmic RNA structure due to protein binding differences between the two compartments? What the mechanism underlying dramatic structural changes? Is this an active process or inherent in the splicing transition where two exons form newly conjoined sequences that rearrange? I think the idea of RNAs with difference nuclear and cytoplasmic structures is intriguing. I’d like to see whether this concept and the 5′ splice site “structural motif” hold up humans.

Upcoming (1/15/2021) – We will discuss Marinus et al.‘s publication on novel SHAPE reagents with new abilities from the Incarnato lab!

12/18/2020 – We discussed Mi Seul, et al.’s paper on AGO3 structure and activity (2017) from the Nakanishi lab, as well as a secondary paper on AGO3 activity (2020) with tyRNAs (tiny RNAs that are 14-17 nts long). My favorite part of the 2017 paper is the beautiful crystal structure of AGO3 + RNA the authors present. It is available on the pdb under the id 5VM9. In addition to their structural studies the authors also show compelling biochemical experiments demonstrating that the AGO3/microRNA complex is capable of cleaving target mRNAs under specific circumstances. Although AGO2 easily binds a wide variety of microRNAs and directs mRNA cleavage based on the microRNA/mRNA interaction, AGO3 was previously thought to be inactive. Any residual activity in AGO3 preparations in biochemical assays was thought to come from contaminating AGO2. By preparing and purifying AGO3 from insect cells the authors clearly show that AGO3 in complex with specific microRNAs can cleave mRNAs. In their most recent publication, Mi Seul, et al., expand their analysis and show that AGO3’s cleavage ability is influenced by the size of the small guide RNA. In comparison to AGO2, AGO3 poorly cleaves mRNAs when it is complexed with a typical microRNA (21-23nt long). However, AGO3 outperforms AGO2 when it is complexed with small guide RNAs (14 nt long). Interestingly the sequence of these cleavage inducing tiny RNAs (cityRNAs) is more important for cleavage than traditional microRNA guided cleavage by AGO2. It remains difficult to determine which mRNAs AGO3 activity might be relevant for, as these papers still only test a handful of microRNAs and similar cityRNAs. We look forward to learning more about the functions of other, less known members of the argonaute family.

Upcoming (12/31/2020): We are reading Bubenik, et al.’s recent paper on deriving RNA structure from low abundance transcripts (i.e. pre-cursor mRNAs) from the Berglund lab.

There is so much amazing RNA research going on! We have a weekly journal club to talk about the latest publications. Here are the papers that we have been reading:

Upcoming 12/11/20: We are reading Rafael de Cesaris Araujo Tavares, et al.’s paper from Anna Marie Pyle’s lab about RNA structure across SAR-CoV2. This paper was recently published in the Journal of Virology – The global and local distribution of RNA structure throughout the SARS-CoV-2 genome.

12/4/20: We read Meredith Corley, et al.’s paper from Gene Yeo’s lab describing the fSHAPE protocol and its relevance. This paper was recently published in Molecular Cell – Footprinting SHAPE-eCLIP Reveals Transcriptome-wide Hydrogen Bonds at RNA-Protein Interfaces.

One interesting result we noticed in this paper was that RNA binding proteins generally interact with well structured regions (low Shannon entropy), but are more likely to directly contact accessible nucleotides within those structured regions. A question we would have liked to see more about in the discussion is how transcriptome-wide SHAPE protocols performed in the presence of proteins are influenced by protein protections and whether in-cell structure data is inaccurate or if eCLIP is necessary to see protein protection in standard SHAPE protocols.