Bipartite Binding Research Articles

LPS binding caspase activation and recruitment domains (CARDs) are bipartite lipid binding modules.

Caspase-11 is an innate immune pattern recognition receptor (PRR) that detects cytosolic bacterial lipopolysaccharides (LPS) through its caspase activation and recruitment domain (CARD), triggering inflammatory cell death known as pyroptosis. Caspase-11 also detects eukaryotic (i.e. self) lipids. This observation raises the question of whether common or distinct mechanisms govern the interactions with self and nonself lipids. In this study, using biochemical, computational, and cell-based assays, we report that the caspase-11 CARD functions as a bipartite lipid-binding module. Distinct regions within the CARD bind to phosphate groups and long acyl chains of self and nonself lipids. Self-lipid binding capability is conserved across numerous caspase-11 homologs and orthologs. The symmetry in self and nonself lipid detection mechanisms enabled us to engineer an LPS-binding domain de novo, using an ancestral CARD-like domain present in the fish Amphilophus citrinellus. These findings offer critical insights into the molecular basis of LPS recognition by caspase-11 and highlight the fundamental and likely inseparable relationship between self and nonself discrimination.

Bipartite binding interface recruiting HP1 to chromosomal passenger complex at inner centromeres.

Maintenance of ploidy depends on the mitotic kinase Aurora B, the catalytic subunit of the chromosomal passenger complex (CPC) whose proficient activity is supported by HP1 enriched at inner centromeres. HP1 is known to associate with INCENP of the CPC in a manner that depends on the PVI motif conserved across HP1 interactors. Here, we found that the interaction of INCENP with HP1 requires not only the PVI motif but also its C-terminally juxtaposed domain. Remarkably, these domains conditionally fold the β-strand (PVI motif) and the α-helix from a disordered sequence upon HP1 binding and render INCENP with high affinity to HP1. This bipartite binding domain termed SSH domain (Structure composed of Strand and Helix) is necessary and sufficient to attain a predominant interaction of HP1 with INCENP. These results identify a unique HP1-binding module in INCENP that ensures enrichment of HP1 at inner centromeres, Aurora B activity, and thereby mitotic fidelity.

Yeast Rad52 is a homodecamer and possesses BRCA2-like bipartite Rad51 binding modes

Homologous recombination (HR) is an essential double-stranded DNA break repair pathway. In HR, Rad52 facilitates the formation of Rad51 nucleoprotein filaments on RPA-coated ssDNA. Here, we decipher how Rad52 functions using single-particle cryo-electron microscopy and biophysical approaches. We report that Rad52 is a homodecameric ring and each subunit possesses an ordered N-terminal and disordered C-terminal half. An intrinsic structural asymmetry is observed where a few of the C-terminal halves interact with the ordered ring. We describe two conserved charged patches in the C-terminal half that harbor Rad51 and RPA interacting motifs. Interactions between these patches regulate ssDNA binding. Surprisingly, Rad51 interacts with Rad52 at two different bindings sites: one within the positive patch in the disordered C-terminus and the other in the ordered ring. We propose that these features drive Rad51 nucleation onto a single position on the DNA to promote formation of uniform pre-synaptic Rad51 filaments in HR.

Modulation of Multidrug Resistance Protein 1-mediated Transport Processes by the Antiviral Drug Ritonavir in Cultured Primary Astrocytes

The Multidrug Resistance Protein 1 (Mrp1) is an ATP-dependent efflux transporter and a major facilitator of drug resistance in mammalian cells during cancer and HIV therapy. In brain, Mrp1-mediated GSH export from astrocytes is the first step in the supply of GSH precursors to neurons. To reveal potential mechanisms underlying the drug-induced modulation of Mrp1-mediated transport processes, we investigated the effects of the antiviral drug ritonavir on cultured rat primary astrocytes. Ritonavir strongly stimulated the Mrp1-mediated export of glutathione (GSH) by decreasing the Km value from 200 nmol/mg to 28 nmol/mg. In contrast, ritonavir decreased the export of the other Mrp1 substrates glutathione disulfide (GSSG) and bimane-glutathione. To give explanation for these apparently contradictory observations, we performed in silico docking analysis and molecular dynamics simulations using a homology model of rat Mrp1 to predict the binding modes of ritonavir, GSH and GSSG to Mrp1. The results suggest that ritonavir binds to the hydrophilic part of the bipartite binding site of Mrp1 and thereby differently affects the binding and transport of the Mrp1 substrates. These new insights into the modulation of Mrp1-mediated export processes by ritonavir provide a new model to better understand GSH-dependent detoxification processes in brain cells.

Zika virus RNA structure controls its unique neurotropism by bipartite binding to Musashi-1

Human RNA binding protein Musashi-1 (MSI1) plays a critical role in neural progenitor cells (NPCs) by binding to various host RNA transcripts. The canonical MSI1 binding site (MBS), A/GU(1-3)AG single-strand motif, is present in many RNA virus genomes, but only Zika virus (ZIKV) genome has been demonstrated to bind MSI1. Herein, we identified the AUAG motif and the AGAA tetraloop in the Xrn1-resistant RNA 2 (xrRNA2) as the canonical and non-canonical MBS, respectively, and both are crucial for ZIKV neurotropism. More importantly, the unique AGNN-type tetraloop is evolutionally conserved, and distinguishes ZIKV from other known viruses with putative MBSs. Integrated structural analysis showed that MSI1 binds to the AUAG motif and AGAA tetraloop of ZIKV in a bipartite fashion. Thus, our results not only identified an unusual viral RNA structure responsible for MSI recognition, but also revealed a role for the highly structured xrRNA in controlling viral neurotropism.

Structural mechanism of CRL4‐instructed STAT2 degradation via a novel cytomegaloviral DCAF receptor

Human cytomegalovirus (CMV) is a ubiquitously distributed pathogen whose rodent counterparts such as mouse and rat CMV serve as common infection models. Here, we conducted global proteome profiling of rat CMV‐infected cells and uncovered a pronounced loss of the transcription factor STAT2, which is crucial for antiviral interferon signalling. Via deletion mutagenesis, we found that the viral protein E27 is required for CMV‐induced STAT2 depletion. Cellular and in vitro analyses showed that E27 exploits host‐cell Cullin4‐RING ubiquitin ligase (CRL4) complexes to induce poly‐ubiquitylation and proteasomal degradation of STAT2. Cryo‐electron microscopy revealed how E27 mimics molecular surface properties of cellular CRL4 substrate receptors called DCAFs (DDB1‐ and Cullin4‐associated factors), thereby displacing them from the catalytic core of CRL4. Moreover, structural analyses showed that E27 recruits STAT2 through a bipartite binding interface, which partially overlaps with the IRF9 binding site. Structure‐based mutations in M27, the murine CMV homologue of E27, impair the interferon‐suppressing capacity and virus replication in mouse models, supporting the conserved importance of DCAF mimicry for CMV immune evasion.

Structural mechanism of BRD4-NUT and p300 bipartite interaction in propagating aberrant gene transcription in chromatin in NUT carcinoma

BRD4-NUT, a driver fusion mutant in rare and highly aggressive NUT carcinoma, acts in aberrant transcription of anti-differentiation genes by recruiting histone acetyltransferase (HAT) p300 and promoting p300-driven histone hyperacetylation and nuclear condensation in chromatin. However, the molecular basis of how BRD4-NUT recruits and activates p300 remains elusive. Here, we report that BRD4-NUT contains two transactivation domains (TADs) in NUT that bind to the TAZ2 domain in p300. Our NMR structures reveal that NUT TADs adopt amphipathic helices when bound to the four-helical bundle TAZ2 domain. The NUT protein forms liquid-like droplets in-vitro that are enhanced by TAZ2 binding in 1:2 stoichiometry. The TAD/TAZ2 bipartite binding in BRD4-NUT/p300 triggers allosteric activation of p300 and acetylation-driven liquid-like condensation on chromatin that comprise histone H3 lysine 27 and 18 acetylation and transcription proteins BRD4L/S, CDK9, MED1, and RNA polymerase II. The BRD4-NUT/p300 chromatin condensation is key for activating transcription of pro-proliferation genes such as ALX1, resulting ALX1/Snail signaling and epithelial-to-mesenchymal transition. Our study provides a previously underappreciated structural mechanism illuminating BRD4-NUT’s bipartite p300 recruitment and activation in NUT carcinoma that nucleates a feed-forward loop for propagating histone hyperacetylation and chromatin condensation to sustain aberrant anti-differentiation gene transcription and perpetual tumor cell growth.

A splicing silencer in SMN2 intron 6 is critical in spinal muscular atrophy.

Spinal muscular atrophy (SMA) is a fatal neuromuscular disease caused by homozygous deletions or mutations of the SMN1 gene. SMN2 is a paralogous gene of SMN1 and a modifying gene of SMA. A better understanding of how SMN2 exon 7 splicing is regulated helps discover new therapeutic targets for SMA therapy. Based on an antisense walk method to map exonic and intronic splicing silencers (ESSs and ISSs) in SMN2 exon 7 and the proximal regions of its flanking introns, we identified one ISS (ISS6-KH) at upstream of the branch point site in intron 6. By using mutagenesis-coupled RT-PCR with SMN1/2 minigenes, immunochromatography, overexpression and siRNA-knockdown, we found this ISS consists of a bipartite hnRNP A1 binding cis-element and a poly-U sequence located between the proximal hnRNP A1 binding site (UAGCUA) and the branch site. Both HuR and hnRNP C1 proteins promote exon 7 skipping through the poly-U stretch. Mutations or deletions of these motifs lead to efficient SMN2 exon 7 inclusion comparable to SMN1 gene. Furthermore, we identified an optimal antisense oligonucleotide that binds the intron six ISS and causes striking exon 7 inclusion in the SMN2 gene in patient fibroblasts and SMA mouse model. Our findings demonstrate that this novel ISS plays an important role in SMN2 exon 7 skipping and highlight a new therapeutic target for SMA therapy.

Bipartite binding and partial inhibition links DEPTOR and mTOR in a mutually antagonistic embrace.

The mTORC1 kinase complex regulates cell growth, proliferation, and survival. Because mis-regulation of DEPTOR, an endogenous mTORC1 inhibitor, is associated with some cancers, we reconstituted mTORC1 with DEPTOR to understand its function. We find that DEPTOR is a unique partial mTORC1 inhibitor that may have evolved to preserve feedback inhibition of PI3K. Counterintuitively, mTORC1 activated by RHEB or oncogenic mutation is much more potently inhibited by DEPTOR. Although DEPTOR partially inhibits mTORC1, mTORC1 prevents this inhibition by phosphorylating DEPTOR, a mutual antagonism that requires no exogenous factors. Structural analyses of the mTORC1/DEPTOR complex showed DEPTOR's PDZ domain interacting with the mTOR FAT region, and the unstructured linker preceding the PDZ binding to the mTOR FRB domain. The linker and PDZ form the minimal inhibitory unit, but the N-terminal tandem DEP domains also significantly contribute to inhibition.

BusR senses bipartite DNA binding motifs by a unique molecular ruler architecture.

The cyclic dinucleotide second messenger c-di-AMP is a major player in regulation of potassium homeostasis and osmolyte transport in a variety of bacteria. Along with various direct interactions with proteins such as potassium channels, the second messenger also specifically binds to transcription factors, thereby altering the processes in the cell on the transcriptional level. We here describe the structural and biochemical characterization of BusR from the human pathogen Streptococcus agalactiae. BusR is a member of a yet structurally uncharacterized subfamily of the GntR family of transcription factors that downregulates transcription of the genes for the BusA (OpuA) glycine-betaine transporter upon c-di-AMP binding. We report crystal structures of full-length BusR, its apo and c-di-AMP bound effector domain, as well as cryo-EM structures of BusR bound to its operator DNA. Our structural data, supported by biochemical and biophysical data, reveal that BusR utilizes a unique domain assembly with a tetrameric coiled-coil in between the binding platforms, serving as a molecular ruler to specifically recognize a 22 bp separated bipartite binding motif. Binding of c-di-AMP to BusR induces a shift in equilibrium from an inactivated towards an activated state that allows BusR to bind the target DNA, leading to transcriptional repression.

Control of the Serine Integrase Reaction: Roles of the Coiled-Coil and Helix E Regions in DNA Site Synapsis and Recombination.

Bacteriophage serine integrases catalyze highly specific recombination reactions between defined DNA segments called att sites. These reactions are reversible depending upon the presence of a second phage-encoded directionality factor. The bipartite C-terminal DNA-binding region of integrases includes a recombinase domain (RD) connected to a zinc-binding domain (ZD), which contains a long flexible coiled-coil (CC) motif that extends away from the bound DNA. We directly show that the identities of the phage A118 integrase att sites are specified by the DNA spacing between the RD and ZD DNA recognition determinants, which in turn directs the relative trajectories of the CC motifs on each subunit of the att-bound integrase dimer. Recombination between compatible dimer-bound att sites requires minimal-length CC motifs and 14 residues surrounding the tip where the pairing of CC motifs between synapsing dimers occurs. Our alanine-scanning data suggest that molecular interactions between CC motif tips may differ in integrative (attP × attB) and excisive (attL × attR) recombination reactions. We identify mutations in 5 residues within the integrase oligomerization helix that control the remodeling of dimers into tetramers during synaptic complex formation. Whereas most of these gain-of-function mutants still require the CC motifs for synapsis, one mutant efficiently, but indiscriminately, forms synaptic complexes without the CC motifs. However, the CC motifs are still required for recombination, suggesting a function for the CC motifs after the initial assembly of the integrase synaptic tetramer. IMPORTANCE The robust and exquisitely regulated site-specific recombination reactions promoted by serine integrases are integral to the life cycle of temperate bacteriophage and, in the case of the A118 prophage, are an important virulence factor of Listeria monocytogenes. The properties of these recombinases have led to their repurposing into tools for genetic engineering and synthetic biology. In this report, we identify determinants regulating synaptic complex formation between correct DNA sites, including the DNA architecture responsible for specifying the identity of recombination sites, features of the unique coiled-coil structure on the integrase that are required to initiate synapsis, and amino acid residues on the integrase oligomerization helix that control the remodeling of synapsing dimers into a tetramer active for DNA strand exchange.

Bipartite binding of the N terminus of Skp2 to cyclin A

Skp2 and cyclin A are cell-cycle regulators that control the activity of CDK2. Cyclin A acts as an activator and substrate recruitment factor of CDK2, while Skp2 mediates the ubiquitination and subsequent destruction of the CDK inhibitor protein p27. The N terminus of Skp2 can interact directly with cyclin A but is not required for p27 ubiquitination. To gain insight into this poorly understood interaction, we have solved the 3.2Å X-ray crystal structure of the N terminus of Skp2 bound to cyclin A. The structure reveals a bipartite mode of interaction with two motifs in Skp2 recognizing two discrete surfaces on cyclin A. The uncovered binding mechanism allows for a rationalization of the inhibitory effect of Skp2 on CDK2-cyclin A kinase activity toward the RxL motif containing substrates and raises the possibility that other intermolecular regulators and substrates may use similar non-canonical modes of interaction for cyclin targeting.

The Myeloid Zinc Finger Gene, MZF-1, Regulates the CD34 Promoter In Vitro

MZF-1 is a C2H2 zinc finger gene encoding a putative transcriptional regulator of myeloid differentiation. The MZF-1 protein contains 13 C2H2 zinc fingers arranged in bipartite DNA binding domains containing zinc fingers through 4 and, in the carboxy-terminus, 5 through 13. We previously identified the DNA consensus binding site recognized by the two DNA binding domains. To assess the transcription regulatory function of MZF-1, the full-length MZF-1 coding region was fused to the DNA binding domain of the yeast transactivator GAL4. The expression vector was cotransfected with the chloramphenicol acetyl transferase (CAT) reporter gene regulated by the thymidine kinase promoter containing GAL4 DNA binding sites into NIH 3T3, 293, K562, and Jurkat cell lines. MZF-1 represses CAT reporter gene expression via GAL4 binding sites in the nonhematopoietic cell lines NIH 3T3 and 293. In contrast, MZF-1 activates CAT reporter gene expression in the hematopoietic cell lines K562 and Jurkat. The MZF-1 binding sites are present in the promoters of several genes expressed during myeloid differentiation, including the CD34 promoter. MZF-1 transcriptional regulation of this physiologically relevant promoter was assessed in both hematopoietic and nonhematopoietic cell lines. Recombinant MZF-1 protein specifically binds to the consensus binding sites in the CD34 promoter in mobility shift assays. MZF-1 expression vectors were cotransfected with the luciferase reporter plasmids regulated by the CD34 promoter into both nonhematopoietic and hematopoietic cell lines. As with the heterologous DNA binding domain, MZF-1 represses reporter gene expression in nonhematopoietic cell lines and activates expression in hematopoietic cell lines. Activation of CD34 expression in hematopoietic cell lines is dependent on the presence of intact MZF-1 binding sites. The cell type-specific regulation of the CD34 promoter by MZF-1 suggests the presence of tissue-specific regulators/adapters or differential MZF-1 modifications that determine MZF-1 transcriptional regulatory function.

DNA Polymerase ι Interacts with Both the TRAF-like and UBL1-2 Domains of USP7

Reversible protein ubiquitination is an essential signaling mechanism within eukaryotes. Deubiquitinating enzymes are critical to this process, as they mediate removal of ubiquitin from substrate proteins. Ubiquitin-specific protease 7 (USP7) is a prominent deubiquitinating enzyme, with an extensive network of interacting partners and established roles in cell cycle activation, immune responses and DNA replication. Characterized USP7 substrates primarily interact with one of two major binding sites outside the catalytic domain. These are located on the USP7 N-terminal TRAF-like (TRAF) domain and the first and second UBL domains (UBL1-2) within the C-terminal tail. Here, we report that DNA polymerase iota (Pol ι) is a novel USP7 substrate that interacts with both TRAF and UBL1-2. Through the use of biophysical approaches and mutational analysis, we characterize both interfaces and demonstrate that bipartite binding to both USP7 domains is required for efficient Pol ι deubiquitination. Together, these data establish a new bipartite mode of USP7 substrate binding.

The fungal CCAAT-binding complex and HapX display highly variable but evolutionary conserved synergetic promoter-specific DNA recognition.

To sustain iron homeostasis, microorganisms have evolved fine-tuned mechanisms for uptake, storage and detoxification of the essential metal iron. In the human pathogen Aspergillus fumigatus, the fungal-specific bZIP-type transcription factor HapX coordinates adaption to both iron starvation and iron excess and is thereby crucial for virulence. Previous studies indicated that a HapX homodimer interacts with the CCAAT-binding complex (CBC) to cooperatively bind bipartite DNA motifs; however, the mode of HapX-DNA recognition had not been resolved. Here, combination of in vivo (genetics and ChIP-seq), in vitro (surface plasmon resonance) and phylogenetic analyses identified an astonishing plasticity of CBC:HapX:DNA interaction. DNA motifs recognized by the CBC:HapX protein complex comprise a bipartite DNA binding site 5′-CSAATN12RWT-3′ and an additional 5′-TKAN-3′ motif positioned 11–23 bp downstream of the CCAAT motif, i.e. occasionally overlapping the 3′-end of the bipartite binding site. Phylogenetic comparison taking advantage of 20 resolved Aspergillus species genomes revealed that DNA recognition by the CBC:HapX complex shows promoter-specific cross-species conservation rather than regulon-specific conservation. Moreover, we show that CBC:HapX interaction is absolutely required for all known functions of HapX. The plasticity of the CBC:HapX:DNA interaction permits fine tuning of CBC:HapX binding specificities that could support adaptation of pathogens to their host niches.

Bipartite mechanism for laminin-integrin interactions: Identification of the integrin-binding site in LG domains of the laminin α chain

Laminins are major cell-adhesive proteins consisting of α, β, and γ chains, in which the three C-terminal globular domains of the α chain (LMα/LG1–3) and the C-terminal tail region of the γ1 chain (LMγ1-tail) are required for binding to integrin. Despite the recent progress on the role of LMγ1-tail in coordinating the metal ion-dependent adhesion site of the integrin β subunit, the mechanism by which LMα/LG1–3 interacts with integrin remains to be elucidated. We found that basic residues on the bottom face of LMα5/LG2 are required for binding laminin-511 to α6β1 integrin. Intermolecular cysteine scanning assays demonstrated that the basic residues in LMα5/LG2 were in contact with the acidic residues within the laminin-binding X1 region of the integrin α subunit in the laminin-integrin complex. These results indicate that LMα5/LG2 interacts directly with the integrin α subunit and comprises a bipartite integrin binding site of laminin-511 with the LMγ1-tail.

Bipartite anchoring of SCREAM enforces stomatal initiation by coupling MAP kinases to SPEECHLESS.

Cell-fate in eukaryotes is controlled by MAP Kinases (MAPKs) that translate external cues to cellular responses. In plants, two MAPKs, MPK3/6, regulate diverse processes of development, environmental response, and immunity. Yet, the mechanism bridging these shared signaling components with a specific target remains unresolved. Focusing on the development of stomata, epidermal valves for gas exchange and transpiration, we report here that the bHLH protein SCREAM functions as a scaffold recruiting MPK3/6 to downregulate SPEECHLESS, a transcription factor initiating stomatal cell lineages. SCREAM directly binds with MPK3/6 through an evolutionarily-conserved yet unconventional bipartite motif. Mutations in this motif abrogate association, phosphorylation and degradation of SCREAM, unmask hidden non-redundancies between MPK3 and MPK6, and result in uncontrolled stomatal differentiation. Structural analyses of MPK6 at the 2.75Å resolution unraveled bipartite binding of SCREAM with MPK6 that is distinct from an upstream MAPKK. Our findings elucidate, at the atomic resolution, the mechanism directly linking extrinsic signals to transcriptional reprogramming during the establishment of stomatal cell-fate, and highlight a unique substrate-binding mode adopted by plant MAPKs.

ATP-binding Cassette Exporters: Structure and Mechanism with a Focus on P-glycoprotein and MRP1.

Proteins that belong to the ATP-binding cassette superfamily include transporters that mediate the efflux of substrates from cells. Among these exporters, P-glycoprotein and MRP1 are involved in cancer multidrug resistance, protection from endo and xenobiotics, determination of drug pharmacokinetics, and the pathophysiology of a variety of disorders. To review the information available on ATP-binding cassette exporters, with a focus on Pglycoprotein, MRP1 and related proteins. We describe tissue localization and function of these transporters in health and disease, and discuss the mechanisms of substrate transport. We also correlate recent structural information with the function of the exporters, and discuss details of their molecular mechanism with a focus on the nucleotide-binding domains. Evaluation of selected publications on the structure and function of ATP-binding cassette proteins. Conformational changes on the nucleotide-binding domains side of the exporters switch the accessibility of the substrate-binding pocket between the inside and outside, which is coupled to substrate efflux. However, there is no agreement on the magnitude and nature of the changes at the nucleotide- binding domains side that drive the alternate-accessibility. Comparison of the structures of Pglycoprotein and MRP1 helps explain differences in substrate selectivity and the bases for polyspecificity. P-glycoprotein substrates are hydrophobic and/or weak bases, and polyspecificity is explained by a flexible hydrophobic multi-binding site that has a few acidic patches. MRP1 substrates are mostly organic acids, and its polyspecificity is due to a single bipartite binding site that is flexible and displays positive charge.

Slx5/Slx8‐dependent ubiquitin hotspots on chromatin contribute to stress tolerance

Chromatin is a highly regulated environment, and protein association with chromatin is often controlled by post‐translational modifications and the corresponding enzymatic machinery. Specifically, SUMO‐targeted ubiquitin ligases (STUbLs) have emerged as key players in nuclear quality control, genome maintenance, and transcription. However, how STUbLs select specific substrates among myriads of SUMOylated proteins on chromatin remains unclear. Here, we reveal a remarkable co‐localization of the budding yeast STUbL Slx5/Slx8 and ubiquitin at seven genomic loci that we term “ubiquitin hotspots”. Ubiquitylation at these sites depends on Slx5/Slx8 and protein turnover on the Cdc48 segregase. We identify the transcription factor‐like Ymr111c/Euc1 to associate with these sites and to be a critical determinant of ubiquitylation. Euc1 specifically targets Slx5/Slx8 to ubiquitin hotspots via bipartite binding of Slx5 that involves the Slx5 SUMO‐interacting motifs and an additional, novel substrate recognition domain. Interestingly, the Euc1‐ubiquitin hotspot pathway acts redundantly with chromatin modifiers of the H2A.Z and Rpd3L pathways in specific stress responses. Thus, our data suggest that STUbL‐dependent ubiquitin hotspots shape chromatin during stress adaptation.

The Solution Structure of FUS Bound to RNA Reveals a Bipartite Mode of RNA Recognition with Both Sequence and Shape Specificity

Fused in sarcoma (FUS) is an RNA binding protein involved in regulating many aspects of RNA processing and linked to several neurodegenerative diseases. Transcriptomics studies indicate that FUS binds a large variety of RNA motifs, suggesting that FUS RNA binding might be quite complex. Here, we present solution structures of FUS zinc finger (ZnF) and RNA recognition motif (RRM) domains bound to RNA. These structures show a bipartite binding mode of FUS comprising of sequence-specific recognition of a NGGU motif via the ZnF and an unusual shape recognition of a stem-loop RNA via the RRM. In addition, sequence-independent interactions via the RGG repeats significantly increase binding affinity and promote destabilization of structured RNA conformation, enabling additional binding. We further show that disruption of the RRM and ZnF domains abolishes FUS function in splicing. Altogether, our results rationalize why deciphering the RNA binding mode of FUS has been so challenging.