A Deep Neural Network for Predicting and Engineering Alternative Polyadenylation.

Alternative polyadenylation (APA) is a major driver of transcriptome diversity in human cells. Here, we use deep learning to predict APA from DNA sequence alone. We trained our model (APARENT, APA REgression NeT) on isoform expression data from over 3 million APA reporters. APARENT's predictions are highly accurate when tasked with inferring APA in synthetic and human 3'UTRs. Visualizing features learned across all network layers reveals that APARENT recognizes sequence motifs known to recruit APA regulators, discovers previously unknown sequence determinants of 3' end processing, and integrates these features into a comprehensive, interpretable, cis-regulatory code. We apply APARENT to forward engineer functional polyadenylation signals with precisely defined cleavage position and isoform usage and validate predictions experimentally. Finally, we use APARENT to quantify the impact of genetic variants on APA. Our approach detects pathogenic variants in a wide range of disease contexts, expanding our understanding of the genetic origins of disease.

Cryo-EM Structure of Human Dicer and Its Complexes with a Pre-miRNA Substrate.

Human Dicer (hDicer) is a multi-domain protein belonging to the RNase III family. It plays pivotal roles in small RNA biogenesis during the RNA interference (RNAi) pathway by processing a diverse range of double-stranded RNA (dsRNA) precursors to generate ∼22 nt microRNA (miRNA) or small interfering RNA (siRNA) products for sequence-directed gene silencing. In this work, we solved the cryoelectron microscopy (cryo-EM) structure of hDicer in complex with its cofactor protein TRBP and revealed the precise spatial arrangement of hDicer's multiple domains. We further solved structures of the hDicer-TRBP complex bound with pre-let-7 RNA in two distinct conformations. In combination with biochemical analysis, these structures reveal a property of the hDicer-TRBP complex to promote the stability of pre-miRNA's stem duplex in a pre-dicing state. These results provide insights into the mechanism of RNA processing by hDicer and illustrate the regulatory role of hDicer's N-terminal helicase domain.

Dicer-like Enzymes with Sequence Cleavage Preferences.

Dicer proteins are known to produce small RNAs (sRNAs) from long double-stranded RNA (dsRNA) templates. These sRNAs are bound by Argonaute proteins, which select the guide strand, often with a 5' end sequence bias. However, Dicer proteins have never been shown to have sequence cleavage preferences. In Paramecium development, two classes of sRNAs that are required for DNA elimination are produced by three Dicer-like enzymes: Dcl2, Dcl3, and Dcl5. Through in vitro cleavage assays, we demonstrate that Dcl2 has a strict size preference for 25 nt and a sequence preference for 5' U and 5' AGA, while Dcl3 has a sequence preference for 5' UNG. Dcl5, however, has cleavage preferences for 5' UAG and 3' CUAC/UN, which leads to the production of RNAs precisely matching short excised DNA elements with corresponding end base preferences. Thus, we characterize three Dicer-like enzymes that are involved in Paramecium development and propose a biological role for their sequence-biased cleavage products.

Assembly mechanism of Integrator's RNA cleavage module.

The modular Integrator complex is a transcription regulator that is essential for embryonic development. It attenuates coding gene expression via premature transcription termination and performs 3'-processing of non-coding RNAs. For both activities, Integrator requires endonuclease activity that is harbored by an RNA cleavage module consisting of INTS4-9-11. How correct assembly of Integrator modules is achieved remains unknown. Here, we show that BRAT1 and WDR73 are critical biogenesis factors for the human cleavage module. They maintain INTS9-11 inactive during maturation by physically blocking the endonuclease active site and prevent premature INTS4 association. Furthermore, BRAT1 facilitates import of INTS9-11 into the nucleus, where it is joined by INTS4. Final BRAT1 release requires locking of the mature cleavage module conformation by inositol hexaphosphate (IP6). Our data explain several neurodevelopmental disorders caused by BRAT1, WDR73, and INTS11 mutations as Integrator assembly defects and reveal that IP6 is an essential co-factor for cleavage module maturation.

Structural mechanism of angiogenin activation by the ribosome.

Angiogenin, an RNase-A-family protein, promotes angiogenesis and has been implicated in cancer, neurodegenerative diseases and epigenetic inheritance1-10. After activation during cellular stress, angiogenin cleaves tRNAs at the anticodon loop, resulting in translation repression11-15. However, the catalytic activity of isolated angiogenin is very low, and the mechanisms of the enzyme activation and tRNA specificity have remained a puzzle3,16-23. Here we identify these mechanisms using biochemical assays and cryogenic electron microscopy (cryo-EM). Our study reveals that the cytosolic ribosome is the activator of angiogenin. A cryo-EM structure features angiogenin bound in the A site of the 80S ribosome. The C-terminal tail of angiogenin is rearranged by interactions with the ribosome to activate the RNase catalytic centre, making the enzyme several orders of magnitude more efficient in tRNA cleavage. Additional 80S-angiogenin structures capture how tRNA substrate is directed by the ribosome into angiogenin's active site, demonstrating that the ribosome acts as the specificity factor. Our findings therefore suggest that angiogenin is activated by ribosomes with a vacant A site, the abundance of which increases during cellular stress24-27. These results may facilitate the development of therapeutics to treat cancer and neurodegenerative diseases.

Structural basis for the activity of the type VII CRISPR-Cas system.

The newly identified type VII CRISPR-Cas candidate system uses a CRISPR RNA-guided ribonucleoprotein complex formed by Cas5 and Cas7 proteins to target RNA1. However, the RNA cleavage is executed by a dedicated Cas14 nuclease, which is distinct from the effector nucleases of the other CRISPR-Cas systems. Here we report seven cryo-electron microscopy structures of the Cas14-bound interference complex at different functional states. Cas14, a tetrameric protein in solution, is recruited to the Cas5-Cas7 complex in a target RNA-dependent manner. The N-terminal catalytic domain of Cas14 binds a stretch of the substrate RNA for cleavage, whereas the C-terminal domain is primarily responsible for tethering Cas14 to the Cas5-Cas7 complex. The biochemical cleavage assays corroborate the captured functional conformations, revealing that Cas14 binds to different sites on the Cas5-Cas7 complex to execute individual cleavage events. Notably, a plugged-in arginine of Cas7 sandwiched by a C-shaped clamp of C-terminal domain precisely modulates Cas14 binding. More interestingly, target RNA cleavage is altered by a complementary protospacer flanking sequence at the 5' end, but not at the 3' end. Altogether, our study elucidates critical molecular details underlying the assembly of the interference complex and substrate cleavage in the type VII CRISPR-Cas system, which may help rational engineering of the type VII CRISPR-Cas system for biotechnological applications.

New Type III CRISPR variant and programmable RNA targeting tool: Oh, thank heaven for Cas7-11.

Özcan et al. (2021) and van Beljouw et al. (2021) characterize a novel Type III-E CRISPR-Cas subtype, composed of a single polypeptide with crRNA processing and sequence-specific RNA cleavage activities, that provides a new RNA knockdown tool for mammalian cells with fewer off-target effects than current technologies.

Co-transcriptional splicing regulates 3' end cleavage during mammalian erythropoiesis.

Pre-mRNA processing steps are tightly coordinated with transcription in many organisms. To determine how co-transcriptional splicing is integrated with transcription elongation and 3' end formation in mammalian cells, we performed long-read sequencing of individual nascent RNAs and precision run-on sequencing (PRO-seq) during mouse erythropoiesis. Splicing was not accompanied by transcriptional pausing and was detected when RNA polymerase II (Pol II) was within 75-300 nucleotides of 3' splice sites (3'SSs), often during transcription of the downstream exon. Interestingly, several hundred introns displayed abundant splicing intermediates, suggesting that splicing delays can take place between the two catalytic steps. Overall, splicing efficiencies were correlated among introns within the same transcript, and intron retention was associated with inefficient 3' end cleavage. Remarkably, a thalassemia patient-derived mutation introducing a cryptic 3'SS improved both splicing and 3' end cleavage of individual ß-globin transcripts, demonstrating functional coupling between the two co-transcriptional processes as a determinant of productive gene output.

Structural basis for self-cleavage prevention by tag:anti-tag pairing complementarity in type VI Cas13 CRISPR systems.

Bacteria and archaea apply CRISPR-Cas surveillance complexes to defend against foreign invaders. These invading genetic elements are captured and integrated into the CRISPR array as spacer elements, guiding sequence-specific DNA/RNA targeting and cleavage. Recently, in vivo studies have shown that target RNAs with extended complementarity with repeat sequences flanking the target element (tag:anti-tag pairing) can dramatically reduce RNA cleavage by the type VI-A Cas13a system. Here, we report the cryo-EM structure of Leptotrichia shahii LshCas13acrRNA in complex with target RNA harboring tag:anti-tag pairing complementarity, with the observed conformational changes providing a molecular explanation for inactivation of the composite HEPN domain cleavage activity. These structural insights, together with in vitro biochemical and in vivo cell-based assays on key mutants, define the molecular principles underlying Cas13a's capacity to target and discriminate between self and non-self RNA targets. Our studies illuminate approaches to regulate Cas13a's cleavage activity, thereby influencing Cas13a-mediated biotechnological applications.

GTSF1 accelerates target RNA cleavage by PIWI-clade Argonaute proteins.

Argonaute proteins use nucleic acid guides to find and bind specific DNA or RNA target sequences. Argonaute proteins have diverse biological functions and many retain their ancestral endoribonuclease activity, cleaving the phosphodiester bond between target nucleotides t10 and t11. In animals, the PIWI proteins-a specialized class of Argonaute proteins-use 21-35 nucleotide PIWI-interacting RNAs (piRNAs) to direct transposon silencing, protect the germline genome, and regulate gene expression during gametogenesis1. The piRNA pathway is required for fertility in one or both sexes of nearly all animals. Both piRNA production and function require RNA cleavage catalysed by PIWI proteins. Spermatogenesis in mice and other placental mammals requires three distinct, developmentally regulated PIWI proteins: MIWI (PIWIL1), MILI (PIWIL2) and MIWI22-4 (PIWIL4). The piRNA-guided endoribonuclease activities of MIWI and MILI are essential for the production of functional sperm5,6. piRNA-directed silencing in mice and insects also requires GTSF1, a PIWI-associated protein of unknown function7-12. Here we report that GTSF1 potentiates the weak, intrinsic, piRNA-directed RNA cleavage activities of PIWI proteins, transforming them into efficient endoribonucleases. GTSF1 is thus an example of an auxiliary protein that potentiates the catalytic activity of an Argonaute protein.

The kinase Rio1 and a ribosome collision-dependent decay pathway survey the integrity of 18S rRNA cleavage.

The 18S rRNA sequence is highly conserved, particularly at its 3'-end, which is formed by the endonuclease Nob1. How Nob1 identifies its target sequence is not known, and in vitro experiments have shown Nob1 to be error-prone. Moreover, the sequence around the 3'-end is degenerate with similar sites nearby. Here, we used yeast genetics, biochemistry, and next-generation sequencing to investigate a role for the ATPase Rio1 in monitoring the accuracy of the 18S rRNA 3'-end. We demonstrate that Nob1 can miscleave its rRNA substrate and that miscleaved rRNA accumulates upon bypassing the Rio1-mediated quality control (QC) step, but not in healthy cells with intact QC mechanisms. Mechanistically, we show that Rio1 binding to miscleaved rRNA is weaker than its binding to accurately processed 18S rRNA. Accordingly, excess Rio1 results in accumulation of miscleaved rRNA. Ribosomes containing miscleaved rRNA can translate, albeit more slowly, thereby inviting collisions with trailing ribosomes. These collisions result in degradation of the defective ribosomes utilizing parts of the machinery for mRNA QC. Altogether, the data support a model in which Rio1 inspects the 3'-end of the nascent 18S rRNA to prevent miscleaved 18S rRNA-containing ribosomes from erroneously engaging in translation, where they induce ribosome collisions. The data also demonstrate how ribosome collisions purify cells of altered ribosomes with different functionalities, with important implications for the concept of ribosome heterogeneity.

Widespread Backtracking by RNA Pol II Is a Major Effector of Gene Activation, 5' Pause Release, Termination, and Transcription Elongation Rate.

In addition to phosphodiester bond formation, RNA polymerase II has an RNA endonuclease activity, stimulated by TFIIS, which rescues complexes that have arrested and backtracked. How TFIIS affects transcription under normal conditions is poorly understood. We identified backtracking sites in human cells using a dominant-negative TFIIS (TFIISDN) that inhibits RNA cleavage and stabilizes backtracked complexes. Backtracking is most frequent within 2 kb of start sites, consistent with slow elongation early in transcription, and in 3' flanking regions where termination is enhanced by TFIISDN, suggesting that backtracked pol II is a favorable substrate for termination. Rescue from backtracking by RNA cleavage also promotes escape from 5' pause sites, prevents premature termination of long transcripts, and enhances activation of stress-inducible genes. TFIISDN slowed elongation rates genome-wide by half, suggesting that rescue of backtracked pol II by TFIIS is a major stimulus of elongation under normal conditions.

Selective Roles of Vertebrate PCF11 in Premature and Full-Length Transcript Termination.

The pervasive nature of RNA polymerase II (Pol II) transcription requires efficient termination. A key player in this process is the cleavage and polyadenylation (CPA) factor PCF11, which directly binds to the Pol II C-terminal domain and dismantles elongating Pol II from DNA in vitro. We demonstrate that PCF11-mediated termination is essential for vertebrate development. A range of genomic analyses, including mNET-seq, 3' mRNA-seq, chromatin RNA-seq, and ChIP-seq, reveals that PCF11 enhances transcription termination and stimulates early polyadenylation genome-wide. PCF11 binds preferentially between closely spaced genes, where it prevents transcriptional interference and consequent gene downregulation. Notably, PCF11 is sub-stoichiometric to the CPA complex. Low levels of PCF11 are maintained by an auto-regulatory mechanism involving premature termination of its own transcript and are important for normal development. Both in human cell culture and during zebrafish development, PCF11 selectively attenuates the expression of other transcriptional regulators by premature CPA and termination.

Structural Basis of Transcription: RNA Polymerase Backtracking and Its Reactivation.

Regulatory sequences or erroneous incorporations during DNA transcription cause RNA polymerase backtracking and inactivation in all kingdoms of life. Reactivation requires RNA transcript cleavage. Essential transcription factors (GreA and GreB, or TFIIS) accelerate this reaction. We report four cryo-EM reconstructions of Escherichia coli RNA polymerase representing the entire reaction pathway: (1) a backtracked complex; a backtracked complex with GreB (2) before and (3) after RNA cleavage; and (4) a reactivated, substrate-bound complex with GreB before RNA extension. Compared with eukaryotes, the backtracked RNA adopts a different conformation. RNA polymerase conformational changes cause distinct GreB states: a fully engaged GreB before cleavage; a disengaged GreB after cleavage; and a dislodged, loosely bound GreB removed from the active site to allow RNA extension. These reconstructions provide insight into the catalytic mechanism and dynamics of RNA cleavage and extension and suggest how GreB targets backtracked complexes without interfering with canonical transcription.

A Complex of U1 snRNP with Cleavage and Polyadenylation Factors Controls Telescripting, Regulating mRNA Transcription in Human Cells.

Full-length transcription in the majority of human genes depends on U1 snRNP (U1) to co-transcriptionally suppress transcription-terminating premature 3' end cleavage and polyadenylation (PCPA) from cryptic polyadenylation signals (PASs) in introns. However, the mechanism of this U1 activity, termed telescripting, is unknown. Here, we captured a complex, comprising U1 and CPA factors (U1-CPAFs), that binds intronic PASs and suppresses PCPA. U1-CPAFs are distinct from U1-spliceosomal complexes; they include CPA's three main subunits, CFIm, CPSF, and CstF; lack essential splicing factors; and associate with transcription elongation and mRNA export complexes. Telescripting requires U1:pre-mRNA base-pairing, which can be disrupted by U1 antisense oligonucleotide (U1 AMO), triggering PCPA. U1 AMO remodels U1-CPAFs, revealing changes, including recruitment of CPA-stimulating factors, that explain U1-CPAFs' switch from repressive to activated states. Our findings outline this U1 telescripting mechanism and demonstrate U1's unique role as central regulator of pre-mRNA processing and transcription.

Efficient Pre-mRNA Cleavage Prevents Replication-Stress-Associated Genome Instability.

Cellular mechanisms that safeguard genome integrity are often subverted in cancer. To identify cancer-related genome caretakers, we employed a convergent multi-screening strategy coupled to quantitative image-based cytometry and ranked candidate genes according to multivariate readouts reflecting viability, proliferative capacity, replisome integrity, and DNA damage signaling. This unveiled regulators of replication stress resilience, including components of the pre-mRNA cleavage and polyadenylation complex. We show that deregulation of pre-mRNA cleavage impairs replication fork speed and leads to excessive origin activity, rendering cells highly dependent on ATR function. While excessive formation of RNA:DNA hybrids under these conditions was tightly associated with replication-stress-induced DNA damage, inhibition of transcription rescued fork speed, origin activation, and alleviated replication catastrophe. Uncoupling of pre-mRNA cleavage from co-transcriptional processing and export also protected cells from replication-stress-associated DNA damage, suggesting that pre-mRNA cleavage provides a mechanism to efficiently release nascent transcripts and thereby prevent gene gating-associated genomic instability.

Critical factors for precise and efficient RNA cleavage by RNase Y in Staphylococcus aureus.

Cellular processes require precise and specific gene regulation, in which continuous mRNA degradation is a major element. The mRNA degradation mechanisms should be able to degrade a wide range of different RNA substrates with high efficiency, but should at the same time be limited, to avoid killing the cell by elimination of all cellular RNA. RNase Y is a major endoribonuclease found in most Firmicutes, including Bacillus subtilis and Staphylococcus aureus. However, the molecular interactions that direct RNase Y to cleave the correct RNA molecules at the correct position remain unknown. In this work we have identified transcripts that are homologs in S. aureus and B. subtilis, and are RNase Y targets in both bacteria. Two such transcript pairs were used as models to show a functional overlap between the S. aureus and the B. subtilis RNase Y, which highlighted the importance of the nucleotide sequence of the RNA molecule itself in the RNase Y targeting process. Cleavage efficiency is driven by the primary nucleotide sequence immediately downstream of the cleavage site and base-pairing in a secondary structure a few nucleotides downstream. Cleavage positioning is roughly localised by the downstream secondary structure and fine-tuned by the nucleotide immediately upstream of the cleavage. The identified elements were sufficient for RNase Y-dependent cleavage, since the sequence elements from one of the model transcripts were able to convert an exogenous non-target transcript into a target for RNase Y.

The Integrator complex cleaves nascent mRNAs to attenuate transcription.

Cellular homeostasis requires transcriptional outputs to be coordinated, and many events post-transcription initiation can dictate the levels and functions of mature transcripts. To systematically identify regulators of inducible gene expression, we performed high-throughput RNAi screening of the Drosophila Metallothionein A (MtnA) promoter. This revealed that the Integrator complex, which has a well-established role in 3' end processing of small nuclear RNAs (snRNAs), attenuates MtnA transcription during copper stress. Integrator complex subunit 11 (IntS11) endonucleolytically cleaves MtnA transcripts, resulting in premature transcription termination and degradation of the nascent RNAs by the RNA exosome, a complex also identified in the screen. Using RNA-seq, we then identified >400 additional Drosophila protein-coding genes whose expression increases upon Integrator depletion. We focused on a subset of these genes and confirmed that Integrator is bound to their 5' ends and negatively regulates their transcription via IntS11 endonuclease activity. Many noncatalytic Integrator subunits, which are largely dispensable for snRNA processing, also have regulatory roles at these protein-coding genes, possibly by controlling Integrator recruitment or RNA polymerase II dynamics. Altogether, our results suggest that attenuation via Integrator cleavage limits production of many full-length mRNAs, allowing precise control of transcription outputs.

Oncogenic RAS Signaling Promotes Tumor Immunoresistance by Stabilizing PD-L1 mRNA.

The immunosuppressive protein PD-L1 is upregulated in many cancers and contributes to evasion of the host immune system. The relative importance of the tumor microenvironment and cancer cell-intrinsic signaling in the regulation of PD-L1 expression remains unclear. We report that oncogenic RAS signaling can upregulate tumor cell PD-L1 expression through a mechanism involving increases in PD-L1 mRNA stability via modulation of the AU-rich element-binding protein tristetraprolin (TTP). TTP negatively regulates PD-L1 expression through AU-rich elements in the 3' UTR of PD-L1 mRNA. MEK signaling downstream of RAS leads to phosphorylation and inhibition of TTP by the kinase MK2. In human lung and colorectal tumors, RAS pathway activation is associated with elevated PD-L1 expression. In vivo, restoration of TTP expression enhances anti-tumor immunity dependent on degradation of PD-L1 mRNA. We demonstrate that RAS can drive cell-intrinsic PD-L1 expression, thus presenting therapeutic opportunities to reverse the innately immunoresistant phenotype of RAS mutant cancers.

Structural basis of the unfolded protein response.

The unfolded protein response (UPR) is a network of intracellular signaling pathways that maintain the protein-folding capacity of the endoplasmic reticulum (ER) in eukaryotic cells. Dedicated molecular sensors embedded in the ER membrane detect incompletely folded or unfolded proteins in the ER lumen and activate a transcriptional program that increases the abundance of the ER according to need. In metazoans the UPR additionally regulates translation and thus relieves unfolded protein load by globally reducing protein synthesis. If homeostasis in the ER cannot be reestablished, the metazoan UPR switches from the prosurvival to the apoptotic mode. The UPR involves a complex, coordinated action of many genes that is controlled by one ER-embedded sensor, Ire1, in yeasts, and three sensors, Ire1, PERK, and ATF6, in higher eukaryotes, including human. We discuss the emerging molecular understanding of the UPR and focus on the structural biology of Ire1 and PERK, the two recently crystallized UPR sensors.