Deep Learning for Elucidating Modifications to RNA-Status and Challenges Ahead.

RNA-binding proteins and chemical modifications to RNA play vital roles in the co- and post-transcriptional regulation of genes. In order to fully decipher their biological roles, it is an essential task to catalogue their precise target locations along with their preferred contexts and sequence-based determinants. Recently, deep learning approaches have significantly advanced in this field. These methods can predict the presence or absence of modification at specific genomic regions based on diverse features, particularly sequence and secondary structure, allowing us to decipher the highly non-linear sequence patterns and structures that underlie site preferences. This article provides an overview of how deep learning is being applied to this area, with a particular focus on the problem of mRNA-RBP binding, while also considering other types of chemical modification to RNA. It discusses how different types of model can handle sequence-based and/or secondary-structure-based inputs, the process of model training, including choice of negative regions and separating sets for testing and training, and offers recommendations for developing biologically relevant models. Finally, it highlights four key areas that are crucial for advancing the field.

A new mRNA structure prediction based approach to identifying improved signal peptides for bone morphogenetic protein 2.

BACKGROUND: Signal peptide (SP) engineering has proven able to improve production of many proteins yet is a laborious process that still relies on trial and error. mRNA structure around the translational start site is important in translation initiation and has rarely been considered in this context, with recent improvements in in silico mRNA structure potentially rendering it a useful predictive tool for SP selection. Here we attempt to create a method to systematically screen candidate signal peptide sequences in silico based on both their nucleotide and amino acid sequences. Several recently released computational tools were used to predict signal peptide activity (SignalP), localization target (DeepLoc) and predicted mRNA structure (MXFold2). The method was tested with Bone Morphogenetic Protein 2 (BMP2), an osteogenic growth factor used clinically for bone regeneration. It was hoped more effective BMP2 SPs could improve BMP2-based gene therapies and reduce the cost of recombinant BMP2 production. RESULTS: Amino acid sequence analysis indicated 2,611 SPs from the TGF-ß superfamily were predicted to function when attached to BMP2. mRNA structure prediction indicated structures at the translational start site were likely highly variable. The five sequences with the most accessible translational start sites, a codon optimized BMP2 SP variant and the well-established hIL2 SP sequence were taken forward to in vitro testing. The top five candidates showed non-significant improvements in BMP2 secretion in HEK293T cells. All showed reductions in secretion versus the native sequence in C2C12 cells, with several showing large and significant decreases. None of the tested sequences were able to increase alkaline phosphatase activity above background in C2C12s. The codon optimized control sequence and hIL2 SP showed reasonable activity in HEK293T but very poor activity in C2C12. CONCLUSIONS: These results support the use of peptide sequence based in silico tools for basic predictions around signal peptide activity in a synthetic biology context. However, mRNA structure prediction requires improvement before it can produce reliable predictions for this application. The poor activity of the codon optimized BMP2 SP variant in C2C12 emphasizes the importance of codon choice, mRNA structure, and cellular context for SP activity.

Organ- and Cell-Selective Delivery of mRNA In Vivo Using Guanidinylated Serinol Charge-Altering Releasable Transporters.

Selective RNA delivery is required for the broad implementation of RNA clinical applications, including prophylactic and therapeutic vaccinations, immunotherapies for cancer, and genome editing. Current polyanion delivery relies heavily on cationic amines, while cationic guanidinium systems have received limited attention due in part to their strong polyanion association, which impedes intracellular polyanion release. Here, we disclose a general solution to this problem in which cationic guanidinium groups are used to form stable RNA complexes upon formulation but at physiological pH undergo a novel charge-neutralization process, resulting in RNA release. This new delivery system consists of guanidinylated serinol moieties incorporated into a charge-altering releasable transporter (GSer-CARTs). Significantly, systematic variations in structure and formulation resulted in GSer-CARTs that exhibit highly selective mRNA delivery to the lung (∼97%) and spleen (∼98%) without targeting ligands. Illustrative of their breadth and translational potential, GSer-CARTs deliver circRNA, providing the basis for a cancer vaccination strategy, which in a murine model resulted in antigen-specific immune responses and effective suppression of established tumors.

The mechanical properties of lipid nanoparticles depend on the type of biomacromolecule they are loaded with.

For drug delivery systems, the mechanical properties of drug carriers are suspected to play a crucial role in the delivery process. However, there is a lack of reliable methods available to measure the mechanical properties of drug carriers, which hampers the establishment of a link between delivery efficiency and the mechanical properties of carriers. Lipid nanoparticles (LNPs) are advanced systems for delivering nucleic acids to target cell populations for vaccination purposes (mRNA) or the development of new drugs. Hence, it is crucial to develop reliable techniques to measure the mechanical properties of LNPs. In this article, we used AFM to image and probe the mechanical properties of LNPs which are loaded with two different biopolymers either pDNA or mRNA. Imaging the LNPs before and after indentation, as well as recording the retraction curve, enables us to obtain more insight into how the AFM tip penetrates into the particle and to determine whether the deformation of the LNPs is reversible. For pDNA, the indentation by the tip leads to irreversible rupture of the LNPs, while the deformation is reversible for the mRNA-loaded LNPs. Moreover, the forces reached for pDNA are higher than for mRNA. These results pave the way toward the establishment of the link between the LNP formulation and the delivery efficiency.

Silicon Ether Ionizable Lipids Enable Potent mRNA Lipid Nanoparticles with Rapid Tissue Clearance.

The advent of mRNA for nucleic acid (NA) therapeutics has unlocked many diverse areas of research and clinical investigation. However, the shorter intracellular half-life of mRNA compared with other NAs may necessitate more frequent dosing regimens. Because lipid nanoparticles (LNPs) are the principal delivery system used for mRNA, this could lead to tolerability challenges associated with an accumulated lipid burden. This can be addressed by introducing enzymatically cleaved carboxylic esters into the hydrophobic domains of lipid components, notably, the ionizable lipid. However, enzymatic activity can vary significantly with age, disease state, and species, potentially limiting the application in humans. Here we report an alternative approach to ionizable lipid degradability that relies on nonenzymatic hydrolysis, leading to a controlled and highly efficient lipid clearance profile. We identify highly potent examples and demonstrate their exceptional tolerability in multiple preclinical species, including multidosing in nonhuman primates (NHP).

Data-balanced transformer for accelerated ionizable lipid nanoparticles screening in mRNA delivery.

Despite the widespread use of ionizable lipid nanoparticles (LNPs) in clinical applications for messenger RNA (mRNA) delivery, the mRNA drug delivery system faces an efficient challenge in the screening of LNPs. Traditional screening methods often require a substantial amount of experimental time and incur high research and development costs. To accelerate the early development stage of LNPs, we propose TransLNP, a transformer-based transfection prediction model designed to aid in the selection of LNPs for mRNA drug delivery systems. TransLNP uses two types of molecular information to perceive the relationship between structure and transfection efficiency: coarse-grained atomic sequence information and fine-grained atomic spatial relationship information. Due to the scarcity of existing LNPs experimental data, we find that pretraining the molecular model is crucial for better understanding the task of predicting LNPs properties, which is achieved through reconstructing atomic 3D coordinates and masking atom predictions. In addition, the issue of data imbalance is particularly prominent in the real-world exploration of LNPs. We introduce the BalMol block to solve this problem by smoothing the distribution of labels and molecular features. Our approach outperforms state-of-the-art works in transfection property prediction under both random and scaffold data splitting. Additionally, we establish a relationship between molecular structural similarity and transfection differences, selecting 4267 pairs of molecular transfection cliffs, which are pairs of molecules that exhibit high structural similarity but significant differences in transfection efficiency, thereby revealing the primary source of prediction errors. The code, model and data are made publicly available at https://github.com/wklix/TransLNP.

Elucidating Structural Configuration of Lipid Assemblies for mRNA Delivery Systems.

The development of mRNA delivery systems utilizing lipid-based assemblies holds immense potential for precise control of gene expression and targeted therapeutic interventions. Despite advancements in lipid-based gene delivery systems, a critical knowledge gap remains in understanding how the biophysical characteristics of lipid assemblies and mRNA complexes influence these systems. Herein, we investigate the biophysical properties of cationic liposomes and their role in shaping mRNA lipoplexes by comparing various fabrication methods. Notably, an innovative fabrication technique called the liposome under cryo-assembly (LUCA) cycle, involving a precisely controlled freeze-thaw-vortex process, produces distinctive onion-like concentric multilamellar structures in cationic DOTAP/DOPE liposomes, in contrast to a conventional extrusion method that yields unilamellar liposomes. The inclusion of short-chain DHPC lipids further modulates the structure of cationic liposomes, transforming them from multilamellar to unilamellar structures during the LUCA cycle. Furthermore, the biophysical and biological evaluations of mRNA lipoplexes unveil that the optimal N/P charge ratio in the lipoplex can vary depending on the structure of initial cationic liposomes. Cryo-EM structural analysis demonstrates that multilamellar cationic liposomes induce two distinct interlamellar spacings in cationic lipoplexes, emphasizing the significant impact of the liposome structures on the final structure of mRNA lipoplexes. Taken together, our results provide an intriguing insight into the relationship between lipid assembly structures and the biophysical characteristics of the resulting lipoplexes. These relationships may open the door for advancing lipid-based mRNA delivery systems through more streamlined manufacturing processes.

Programmable RNA Loading of Extracellular Vesicles with Toehold-Release Purification.

Synthetic nanoparticles as lipid nanoparticles (LNPs) are widely used as drug delivery vesicles. However, they hold several drawbacks, including low biocompatibility and unfavorable immune responses. Naturally occurring extracellular vesicles (EVs) hold the potential as native, safe, and multifunctional nanovesicle carriers. However, loading of EVs with large biomolecules remains a challenge. Here, we present a controlled loading methodology using DNA-mediated and programmed fusion between EVs and messenger RNA (mRNA)-loaded liposomes. The fusion efficiency is characterized at the single-particle level by real-time microscopy through EV surface immobilization via lipidated biotin-DNA handles. Subsequently, fused EV-liposome particles (EVLs) can be collected by employing a DNA strand-replacement reaction. Transferring the fusion reaction to magnetic beads enables us to scale up the production of EVLs one million times. Finally, we demonstrated encapsulation of mCherry mRNA, transfection, and improved translation using the EVLs compared to liposomes or LNPs in HEK293-H cells. We envision this as an important tool for the EV-mediated delivery of RNA therapeutics.

G-Clamp Heterocycle Modification Containing Interstrand Photo-Cross-Linker to Capture Intracellular MicroRNA Targets.

MicroRNAs (miRNAs) play indispensable roles in post-transcriptional gene regulation. The identification of target mRNAs is essential for dissecting the recognition basis, dynamics, and regulatory mechanism of miRNA-mRNA interactions. However, the lack of an unbiased method for detecting weak miRNA-mRNA interactions remains a long-standing obstacle for miRNA research. Here, we develop and provide proof-of-concept evidence demonstrating a chemical G-clamp-enhanced photo-cross-linking strategy for covalent capture of intracellular miRNA targets in different cell lines. This approach relies on an aryl-diazirine-G-clamp-modified-nucleoside (ARAGON) miRNA probe containing an alkynyl group that improves the thermal stability of miRNA-target mRNA duplex molecules and can rapidly cross-link with the complementary strand upon UV 365 nm activation, enhancing the transient capture of mRNA targets. After validating the accuracy and binding properties of ARAGON-based miRNA probes through the successful enrichment for the known targets of miR-106a, miR-21, and miR-101, we then extend ARAGON's application to screen for previously unknown targets of different miRNAs in various cell lines. Ultimately, results in this study uncover GAB1 as a target of miR-101 in H1299 lung cancer cells and show that miR-101 silencing of GAB1 can promote apoptosis in H1299 cells, suggesting an oncogenic mechanism of GAB1. This study thus provides a powerful and versatile tool for enhanced screening of global miRNA targets in cells to facilitate investigations of miRNA functions in fundamental cellular processes and disease pathogenesis.

Unearthing a novel function of SRSF1 in binding and unfolding of RNA G-quadruplexes.

SRSF1 governs splicing of over 1500 mRNA transcripts. SRSF1 contains two RNA-recognition motifs (RRMs) and a C-terminal Arg/Ser-rich region (RS). It has been thought that SRSF1 RRMs exclusively recognize single-stranded exonic splicing enhancers, while RS lacks RNA-binding specificity. With our success in solving the insolubility problem of SRSF1, we can explore the unknown RNA-binding landscape of SRSF1. We find that SRSF1 RS prefers purine over pyrimidine. Moreover, SRSF1 binds to the G-quadruplex (GQ) from the ARPC2 mRNA, with both RRMs and RS being crucial. Our binding assays show that the traditional RNA-binding sites on the RRM tandem and the Arg in RS are responsible for GQ binding. Interestingly, our FRET and circular dichroism data reveal that SRSF1 unfolds the ARPC2 GQ, with RS leading unfolding and RRMs aiding. Our saturation transfer difference NMR results discover that Arg residues in SRSF1 RS interact with the guanine base but not other nucleobases, underscoring the uniqueness of the Arg/guanine interaction. Our luciferase assays confirm that SRSF1 can alleviate the inhibitory effect of GQ on gene expression in the cell. Given the prevalence of RNA GQ and SR proteins, our findings unveil unexplored SR protein functions with broad implications in RNA splicing and translation.

Protein G-quadruplex interactions and their effects on phase transitions and protein aggregation.

The SERF family of proteins were originally discovered for their ability to accelerate amyloid formation. Znf706 is an uncharacterized protein whose N-terminus is homologous to SERF proteins. We show here that human Znf706 can promote protein aggregation and amyloid formation. Unexpectedly, Znf706 specifically interacts with stable, non-canonical nucleic acid structures known as G-quadruplexes. G-quadruplexes can affect gene regulation and suppress protein aggregation; however, it is unknown if and how these two activities are linked. We find Znf706 binds preferentially to parallel G-quadruplexes with low micromolar affinity, primarily using its N-terminus, and upon interaction, its dynamics are constrained. G-quadruplex binding suppresses Znf706's ability to promote protein aggregation. Znf706 in conjunction with G-quadruplexes therefore may play a role in regulating protein folding. RNAseq analysis shows that Znf706 depletion specifically impacts the mRNA abundance of genes that are predicted to contain high G-quadruplex density. Our studies give insight into how proteins and G-quadruplexes interact, and how these interactions affect both partners and lead to the modulation of protein aggregation and cellular mRNA levels. These observations suggest that the SERF family of proteins, in conjunction with G-quadruplexes, may have a broader role in regulating protein folding and gene expression than previously appreciated.

Recent developments, opportunities, and challenges in the study of mRNA pseudouridylation.

Pseudouridine is an abundant mRNA modification found in diverse organisms ranging from bacteria and viruses to multicellular plants and humans. New developments in pseudouridine profiling provide quantitative tools to map mRNA pseudouridylation sites. Sparse biochemical studies establish the potential for mRNA pseudouridylation to affect most stages of the mRNA life cycle from birth to death. This recent progress sets the stage for deeper investigations into the molecular and cellular functions of specific mRNA pseudouridines, including in disease.

Structural basis of exoribonuclease-mediated mRNA transcription termination.

Efficient termination is required for robust gene transcription. Eukaryotic organisms use a conserved exoribonuclease-mediated mechanism to terminate the mRNA transcription by RNA polymerase II (Pol II)1-5. Here we report two cryogenic electron microscopy structures of Saccharomyces cerevisiae Pol II pre-termination transcription complexes bound to the 5'-to-3' exoribonuclease Rat1 and its partner Rai1. Our structures show that Rat1 displaces the elongation factor Spt5 to dock at the Pol II stalk domain. Rat1 shields the RNA exit channel of Pol II, guides the nascent RNA towards its active centre and stacks three nucleotides at the 5' terminus of the nascent RNA. The structures further show that Rat1 rotates towards Pol II as it shortens RNA. Our results provide the structural mechanism for the Rat1-mediated termination of mRNA transcription by Pol II in yeast and the exoribonuclease-mediated termination of mRNA transcription in other eukaryotes.

Structures of the ribosome bound to EF-Tu-isoleucine tRNA elucidate the mechanism of AUG avoidance.

The frequency of errors upon decoding of messenger RNA by the bacterial ribosome is low, with one misreading event per 1 × 104 codons. In the universal genetic code, the AUN codon box specifies two amino acids, isoleucine and methionine. In bacteria and archaea, decoding specificity of the AUA and AUG codons relies on the wobble avoidance strategy that requires modification of C34 in the anticodon loop of isoleucine transfer RNAIleCAU (tRNAIleCAU). Bacterial tRNAIleCAU with 2-lysylcytidine (lysidine) at the wobble position deciphers AUA while avoiding AUG. Here we report cryo-electron microscopy structures of the Escherichia coli 70S ribosome complexed with elongation factor thermo unstable (EF-Tu) and isoleucine-tRNAIleLAU in the process of decoding AUA and AUG. Lysidine in tRNAIleLAU excludes AUG by promoting the formation of an unusual Hoogsteen purine-pyrimidine nucleobase geometry at the third position of the codon, weakening the interactions with the mRNA and destabilizing the EF-Tu ternary complex. Our findings elucidate the molecular mechanism by which tRNAIleLAU specifically decodes AUA over AUG.

Supramolecular Lipid Nanoparticles Based on Host-Guest Recognition: A New Generation Delivery System of mRNA Vaccines For Cancer Immunotherapy.

Dendritic cell (DC) maturation is a crucial process for antigen presentation and the initiation of T cell-mediated immune responses. Toll-like receptors play pivotal roles in stimulating DC maturation and promoting antigen presentation. Here, a novel message RNA (mRNA) cancer vaccine is reported that boosts antitumor efficacy by codelivering an mRNA encoding tumor antigen and a TLR7/8 agonist (R848) to DC using supramolecular lipid nanoparticles (SMLNP) as a delivery platform, in which a new ionizable lipid (N2-3L) remarkably enhances the translation efficiency of mRNA and a ß-cyclodextrin (ß-CD)-modified ionizable lipid (Lip-CD) encapsulates R848. The incorporation of R848 adjuvant into the mRNA vaccine through noncovalent host-guest complexation significantly promotes DC maturation and antigen presentation after vaccination, thus resulting in superior antitumor efficacy in vivo. Moreover, the antitumor efficacy is further boosted synergized with immune checkpoint blockade by potentiating the anticancer capability of cytotoxic T lymphocytes infiltrated in tumor sites. This work indicates that SMLNP shows brilliant potential as next-generation delivery system in the development of mRNA vaccines with high efficacy.

Flipping the script: Understanding riboswitches from an alternative perspective.

Riboswitches are broadly distributed regulatory elements most frequently found in the 5'-leader sequence of bacterial mRNAs that regulate gene expression in response to the binding of a small molecule effector. The occupancy status of the ligand-binding aptamer domain manipulates downstream information in the message that instructs the expression machinery. Currently, there are over 55 validated riboswitch classes, where each class is defined based on the identity of the ligand it binds and/or sequence and structure conservation patterns within the aptamer domain. This classification reflects an "aptamer-centric" perspective that dominates our understanding of riboswitches. In this review, we propose a conceptual framework that groups riboswitches based on the mechanism by which RNA manipulates information directly instructing the expression machinery. This scheme does not replace the established aptamer domain-based classification of riboswitches but rather serves to facilitate hypothesis-driven investigation of riboswitch regulatory mechanisms. Based on current bioinformatic, structural, and biochemical studies of a broad spectrum of riboswitches, we propose three major mechanistic groups: (1) "direct occlusion", (2) "interdomain docking", and (3) "strand exchange". We discuss the defining features of each group, present representative examples of riboswitches from each group, and illustrate how these RNAs couple small molecule binding to gene regulation. While mechanistic studies of the occlusion and docking groups have yielded compelling models for how these riboswitches function, much less is known about strand exchange processes. To conclude, we outline the limitations of our mechanism-based conceptual framework and discuss how critical information within riboswitch expression platforms can inform gene regulation.

Deadenylation kinetics of mixed poly(A) tails at single-nucleotide resolution.

Shortening of messenger RNA poly(A) tails, or deadenylation, is a rate-limiting step in mRNA decay and is highly regulated during gene expression. The incorporation of non-adenosines in poly(A) tails, or 'mixed tailing', has been observed in vertebrates and viruses. Here, to quantitate the effect of mixed tails, we mathematically modeled deadenylation reactions at single-nucleotide resolution using an in vitro deadenylation system reconstituted with the complete human CCR4-NOT complex. Applying this model, we assessed the disrupting impact of single guanosine, uridine or cytosine to be equivalent to approximately 6, 8 or 11 adenosines, respectively. CCR4-NOT stalls at the 0, -1 and -2 positions relative to the non-adenosine residue. CAF1 and CCR4 enzyme subunits commonly prefer adenosine but exhibit distinct sequence selectivities and stalling positions. Our study provides an analytical framework to monitor deadenylation and reveals the molecular basis of tail sequence-dependent regulation of mRNA stability.

Rational design of oligonucleotides for enhanced in vitro transcription of small RNA.

All kinds of RNA molecules can be produced by in vitro transcription using T7 RNA polymerase using DNA templates obtained by solid-phase chemical synthesis, primer extension, PCR, or DNA cloning. The oligonucleotide design, however, is a challenge to nonexperts as this relies on a set of rules that have been established empirically over time. Here, we describe a Python program to facilitate the rational design of oligonucleotides, calculated with kinetic parameters for enhanced in vitro transcription (ROCKET). The Python tool uses thermodynamic parameters, performs folding-energy calculations, and selects oligonucleotides suitable for the polymerase extension reaction. These oligonucleotides improve yields of template DNA. With the oligonucleotides selected by the program, the tRNA transcripts can be prepared by a one-pot reaction of the DNA polymerase extension reaction and the transcription reaction. Also, the ROCKET-selected oligonucleotides provide greater transcription yields than that from oligonucleotides selected by Primerize, a leading software for designing oligonucleotides for in vitro transcription, due to the enhancement of template DNA synthesis. Apart from over 50 tRNA genes tested, an in vitro transcribed self-cleaving ribozyme was found to have catalytic activity. In addition, the program can be applied to the synthesis of mRNA, demonstrating the wide applicability of the ROCKET software.

The repertoire and structure of adhesion GPCR transcript variants assembled from publicly available deep-sequenced human samples.

Alternative splicing and multiple transcription start and termination sites can produce a diverse repertoire of mRNA transcript variants from a given gene. While the full picture of the human transcriptome is still incomplete, publicly available RNA datasets have enabled the assembly of transcripts. Using publicly available deep sequencing data from 927 human samples across 48 tissues, we quantified known and new transcript variants, provide an interactive, browser-based application Splice-O-Mat and demonstrate its relevance using adhesion G protein-coupled receptors (aGPCRs) as an example. On average, 24 different transcript variants were detected for each of the 33 human aGPCR genes, and several dominant transcript variants were not yet annotated. Variable transcription starts and complex exon-intron structures encode a flexible protein domain architecture of the N- and C termini and the seven-transmembrane helix domain (7TMD). Notably, we discovered the first GPCR (ADGRG7/GPR128) with eight transmembrane helices. Both the N- and C terminus of this aGPCR were intracellularly oriented, anchoring the N terminus in the plasma membrane. Moreover, the assessment of tissue-specific transcript variants, also for other gene classes, in our application may change the evaluation of disease-causing mutations, as their position in different transcript variants may explain tissue-specific phenotypes.

Pliability in the m6A-Binding Region Extends Druggability of YTH Domains.

Epitranscriptomic mRNA modifications affect gene expression, with their altered balance detected in various cancers. YTHDF proteins contain the YTH reader domain recognizing the m6A mark on mRNA and represent valuable drug targets. Crystallographic structures have been determined for all three family members; however, discrepancies are present in the organization of the m6A-binding pocket. Here, we present new crystallographic structures of the YTH domain of YTHDF1, accompanied by computational studies, showing that this domain can exist in different stable conformations separated by a significant energetic barrier. During the transition, additional conformations are explored, with peculiar druggable pockets appearing and offering new opportunities for the design of YTH-interfering small molecules.