Nature Structural & Molecular Biology

Linker histones are essential for chromatin compaction, yet how they contribute to higher-order fiber assembly remains poorly understood. Here, we determined cryo-electron microscopy structures of Arabidopsis dodeca-nucleosome fibers containing distinct H2A/H3 variants and linker histone H1.3, revealing a noncanonical binding mode that a laterally positioned H1.3 connects the acidic patch of one nucleosome and the DNA of the neighboring nucleosome, thereby scaffolding dinucleosomes into two-start chromatin fibers. This noncanonical binding mode is structurally conserved when H1.3 is replaced by Gallus gallus H5. Furthermore, incorporation of H2A.W and H3.3 further induces back-to-back fiber dimerization. Cryo-electron tomography and in vivo cross-linking mass spectrometry analyses support the physiological relevance of H1 lateral engagement. Our findings establish that linker histones act as active architectural scaffolds in higher-order chromatin organization.

Pre-mRNA splicing determines the expressed proteome and is frequently dysregulated in cancer. The tumour-suppressor RBM5 controls an exon network regulating apoptosis, yet its molecular mechanism is elusive. Using in vivo spliceosome capture and cryogenic electron microscopy, we determined structures of precatalytic spliceosomes arrested by RBM5 immediately after U2 snRNP branchpoint recognition. Despite intron diversity, the U2-pre-mRNA duplex, branchpoint adenine, and downstream polypyrimidine tract are well-resolved. RBM5 binds the outer SF3B1 HEAT surface and performs dual functions: First, its helix-loop-helix motif and upstream zinc-finger domain sterically block tri-snRNP and Prp8 docking and prevent progression to pre-B and Bact complexes; Second, its G-patch activates DHX15 and places this DExH-box helicase on the pre-mRNA as it exits SF3B1, poised for branch helix unwinding. DHX15 binding to SF3B1 is facilitated by U2SURP/SR140, which engages SF3B1 near RBM5s helix-loop-helix. Functional assays confirm that disruption of the RBM5 interfaces with either DHX15 or SF3B1 inhibit exon repression. Mutations at these regulatory interfaces are common in cancer genomes and predicted to disrupt its regulation of apoptotic isoforms. Thus, RBM5 acts as a dual-action spliceosome gatekeeper that couples helicase activation with physical stalling to enforce tumour-suppressive alternative splicing programmes.

The cyclin-dependent kinase CDK11 functions in transcription, mitotic progression, and mRNA splicing. Specifically, spliceosome activation during the B to Bact transition depends on phosphorylation of the U2 snRNP component SF3B1 by the CDK11-cyclin L-SAP30BP complex. Here, we present the structure of this spliceosome-activating CDK-cyclin complex, determined by cryogenic electron microscopy at 2.3 [A] resolution. Our structure and biochemical experiments show that SAP30BP forms extensive interactions with cyclin L2, thereby stabilising it, and forms critical interactions with the C-terminal kinase lobe of CDK11 that promote complex assembly. Destabilisation of cyclin L2 in the absence of SAP30BP suggests that these principles are applicable to all CDK11-cyclin L complexes. Furthermore, we identify a pseudo-substrate sequence near the CDK11 C-terminus and provide evidence for a role of this segment in CDK11 auto-regulation. Finally, the structure of the CDK11-cyclin L2-SAP30BP complex bound to the clinical high-affinity CDK11 inhibitor OTS964 and a comparison to OTS964-bound off-target complexes provide insight into the mechanism of OTS964 selectivity and specificity.

Upon DNA damage, chromatin remodeling is rapidly initiated to promote chromatin accessibility, thereby facilitating the recruitment and assembly of repair factors. Although this enhanced accessibility has been linked to poly(ADP-ribose) polymerase (PARP) activity, the mechanism by which cells overcome the nucleosome barrier remains unclear. Using our designer chromatin system, we uncovered a previously uncharacterized activity of PARP1, whereby it directly and asymmetrically evicts histone dimers proximal to DNA strand breaks from nucleosomes to generate oriented hexasomes. In the presence of HPF1, PARP1 generates stable PARylated hexasomes, an open chromatin intermediate that can serve as a bifunctional hub for recruitment of DNA- and PAR-dependent factors. Using cellular assays, we demonstrated that PARP activity is both required and sufficient to drive chromatin accessibility and the recruitment of repair factors, with direct involvement of subnucleosomal species. Unexpectedly, we identified the C-terminal tail of histone H2A, a motif harboring recurrent cancer-associated mutations, as a critical determinant of efficient PARP1-mediated nucleosome disassembly. Deletion of the H2A tail sensitizes cells to DNA-damaging agents and PARP inhibitors, implicating a functional role of PARP1-mediated nucleosome disassembly in DNA repair. Together, our findings support a model in which PARP1 directly drives histone eviction, leading to the formation of subnucleosomes that facilitate efficient DNA repair.

RNA 5-methylcytidine (m5C) is a prevalent modification that drives RNA stability and function. In humans, m5C is deposited on distinct RNA substrates by DNMT2/TRDMT1 and the NSUN family, to regulate diverse cellular processes, but how m5C writers recognise their substrates remains unclear. NSUN2 is a major m5C methyltransferase with broad roles in cell physiology and strong links to cancer and neurodevelopmental disorders 1. Here, we reconstitute an active human NSUN2-tRNA complex and capture its post-catalytic, tRNA-bound structure at 3.1 [A] resolution. Using an integrated approach combining biochemistry, cryo-electron microscopy, crosslinking mass spectrometry and molecular dynamics simulations, we show that NSUN2 remodels the tRNA to access the variable-loop target cytidine. Recognition is driven by RNA architecture, with NSUN2 exploiting the L-shaped tRNA scaffold to position the target base in the catalytic centre. We further show that Gly679 at the NSUN2-tRNA interface is important for the stability of the complex, providing a mechanistic basis for how the disease-associated Gly679Arg substitution can impair tRNA binding. Together, these findings establish an RNA-structure-guided mechanism for NSUN2 substrate recognition and methylation and provide general principles for m5C deposition on cellular RNAs and their fundamental role in disease.

We report the discovery and structure of a previously unknown ~1 MDa hollow protein assembly, identified during a survey of ciliary complexes from the ciliate Tetrahymena thermophila. By combining mass spectrometry, structure prediction, and cryo-electron microscopy, we define a homotetrameric cage-like complex with a distinctive elliptical architecture and a large internal cavity. A sequence survey revealed several thousand homologs spanning diverse unicellular eukaryotes--including green algae, fungi, amoebozoans, choanoflagellates, and SAR lineages--as well as predominantly gram-negative bacteria, indicating an ancient evolutionary origin and arguing against a eukaryote-specific function. We determined a near-atomic resolution structure of the complex from the slime mold Dictyostelium discoideum, demonstrating conservation of overall architecture and cavity despite low sequence identity. Together, these results establish the CAGE complex (Conserved Assembly in Gram-negative bacteria and Eukaryotes) as a new class of large protein cage broadly distributed across the tree of life. While its biological function remains unknown, its size, architecture, and conservation suggest possible roles in transport or protein/RNA homeostasis.

At the endoplasmic reticulum (ER), membrane protein quality control is tightly regulated to ensure excess subunits are recognized and degraded to protect cellular homeostasis. Using genome wide CRISPR screens, we identified a factor of unknown function, thioredoxin domain containing protein 15 (TXNDC15), and showed that it regulates membrane protein stability by tuning the activity of the E3-ubiquitin ligase, MARCHF6. TXNDC15 modulates MARCHF6 in two opposing ways: first, it enhances the binding, ubiquitination, and degradation of membrane protein subunits with soluble cytosolic domains; and second, it prevents the inappropriate recruitment and ubiquitination of subunits with globular lumenal domains. Patient mutations to TXNDC15 that cause the ciliopathy Meckel-Gruber syndrome, disrupted its binding to MARCHF6, allowing degradation of critical ciliary proteins as they transit through the ER leading to defects in ciliogenesis. The regulatory function of TXNDC15 therefore exemplifies how protein quality control maintains the integrity of the proteome to prevent disease.

P2X receptors are trimeric ATP-gated ion channels that assemble as homo- or heterotrimers, with heteromeric forms exhibiting intrinsic asymmetry that influences function. Here, we report four high-resolution cryo-EM structures of human P2X2/3 heterotrimers representing distinct functional states, including ATP-bound assemblies (P2X332 and P2X223), the apo form, and a ligand/ATP-bound closed conformation. The three ATP-binding sites show asymmetric recognition of MgATP{superscript 2}- and ATP-, and channel activation requires occupancy of only two MgATP{superscript 2}- molecules. Gefapixant binds a single allosteric site and selectively inhibits MgATP{superscript 2}-, but not ATP-, binding, indicating orthosteric-allosteric coupling within the heterotrimer. Structural features of the transmembrane domain define ion permeation, particularly for Ca{superscript 2}. Despite asymmetric ligand interactions, gating remains largely symmetric, with minor differences in desensitization. These findings provide a structural framework linking asymmetry to coordinated channel function and open avenues for subtype-selective therapeutic intervention.

Proper polyadenylation site (PAS) selection is critical for RNA isoform determination. Core spliceosomal components, including U1 snRNP, regulate PAS choice, but whether they work with other splicing factors in this role remains unclear. Here, we establish that the splicing factor SRSF1 regulates PAS selection independently of and through interactions with U1 snRNP. Independent of U1 snRNP, SRSF1 binds RNA near proximal PASs within 3 UTRs to promote their usage, and, in line with this observation, breast cancer tumors with altered SRSF1 levels display shifted 3'-end selection. In conjunction with U1 snRNP, SRSF1 acts on PASs through U1 snRNP-mediated SRSF1-Pol II interactions. Consistent with co-transcriptional regulation, SRSF1 reduces the Pol II elongation index and limits transcription readthrough. Together, our results reveal that SRSF1 shapes RNA isoform determination beyond its canonical role in splicing, through a combination of direct RNA binding and U1 snRNP-dependent coordination with Pol II.

RAF kinases activate MEK in the RAS-MAPK signaling pathway, and changes in RAF kinase signaling have been linked to tumor formation. RAF1 requires the HSP90-CDC37 chaperone system for proper activation, but how the HSP90-CDC37 chaperone system regulates RAF kinase maturation remains enigmatic. We present novel cryo-EM structures of previously uncharacterized RAF1 chaperone complexes, including a 2:2:2 RAF1-HSP90-CDC37 complex (RRHCC), intermediate assemblies (RHCC), and a RAF1-HSP90-CDC37-p23 complex (RHCp23). These reveal an asymmetric stepwise folding mechanism unique among HSP90 kinase clients in which one RAF1 threads through the HSP90 lumen while another is captured in a "casting mold" formed by CDC37 and HSP90 that stabilizes the partially folded C helix of RAF1. The RHCp23 structure shows how p23 cooperates with CDC37 to regulate ATP hydrolysis and client release. The HSP90-CDC37 system supports pre-dimerization of RAF1 and BRAFV600E homodimers and RAF1 heterodimers, a mechanism unique to RAF among kinase clients of HSP90. Phosphoproteomics reveals selective activating phosphorylations within RRHCC. These RAF isoform complexes differentially activate MEK signaling and cell proliferation, establishing HSP90-CDC37 as not just a passive stabilizer but an active regulator of RAF signaling with therapeutic implications.

Endogenous retroviruses (ERVs) compromise genome integrity when expressed, and cells have evolved chromatin-based pathways to silence their transcription. The histone H3.3 chaperone DAXX localizes to a subset of ERVs and enforces their silencing through incompletely defined mechanisms. Using complementary biochemical and genetic approaches, we identify a conserved basic patch within the DAXX histone-binding domain that engages DNA, promotes H3.3 nucleosome assembly in vitro, and is required for H3.3 enrichment at DAXX-bound ERVs in cells. Despite failure to deposit H3.3, DAXX with substitutions in this basic patch retains localization to ERVs and preserves silencing, indicating that histone H3.3 is dispensable for DAXX-mediated repression of ERVs. By contrast, ERV silencing requires the DAXX C-terminal SUMO-interacting motif, which mediates recruitment of SUMOylated repressors, including MORC3. These findings define modular outputs downstream of DAXX recruitment that uncouple nucleosome assembly from ERV silencing and highlight SUMO-dependent effector recruitment as the primary mechanism of silencing.

Peroxisomes import all matrix proteins post-translationally from the cytosol, a process that requires recycling of cargo receptors across the peroxisomal membrane. The membrane-embedded ubiquitin ligase, composed of Pex2, Pex10, and Pex12, is central to this process, but its mechanism remains unclear. Here we determined cryo-electron microscopy structures of the Saccharomyces cerevisiae Pex2-10-12 complex in closed and open states bound to Pex8, an essential factor of previously undefined function. The structures reveal how Pex2-10-12 gates its retro-translocation pore to control receptor entry and how the closed-to-open transition repositions the Pex10 RING domain to enable receptor mono-ubiquitination. Pex8 docks onto Pex2-10-12 from the matrix and guides receptors into the pore. Functional analyses show that the receptors N-terminal segment downstream of its mono-ubiquitination site initiates a loop insertion into the pore. These findings establish how Pex2-10-12 coordinates receptor recognition, retro-translocation, and ubiquitination, providing the molecular basis for receptor recycling in peroxisomal protein import.

Gene architecture in higher eukaryotes exhibits substantial heterogeneity. While most genes follow a canonical pattern of GC-rich exons and AT-rich introns, a subset displays GC-leveled architecture, characterized by uniformly high GC content across both exons and introns. These genes are often associated with nuclear speckles, membraneless compartments enriched in RNA-processing factors, yet the mechanistic basis of this spatial and functional relationship remains unclear. Here, we identify the nuclear speckle protein SON as a critical factor that safeguards the splicing of GC-rich genes. Acute depletion of SON in mouse embryonic stem cells selectively impairs the splicing of short, GC-rich introns. These SON-dependent introns are enriched in highly expressed and functionally essential genes, whose GC-rich architecture contributes to efficient RNA processing and expression. Mechanistically, these introns harbor atypical C-rich, U-poor polypyrimidine tracts at their 3 splice sites, which exhibit reduced affinity for core splicing factors. SON is recruited to these sites via U2 snRNP and further interacts with SR proteins to stabilize the association of U2 snRNP and U2AFs at these C-rich weak splice sites. Notably, the evolutionary expansion of SONs intrinsically disordered region is required to promote efficient splicing of GC-rich genes that emerged during evolution. Together, our study suggests that the evolutionary transition toward GC-rich gene architecture enhances gene expression efficiency, with SON acting to safeguard the splicing of this gene class.

Human translation initiation requires single-nucleotide precision to establish the reading frame, yet initiation at non-AUG codons plays key roles in gene expression. How the initiation machinery balances precision with this regulated flexibility remains unclear. Here, we define a conformational branchpoint governed by the human initiation factor eIF5 that gates commitment to start codons. Using single-molecule and structural approaches, we demonstrate that eIF5 reversibly occupies two conformations, which depends on a strictly conserved loop in the protein that monitors start codon identity. AUG codons favor the conformation that is stabilized by an eIF5-stimulated GTP hydrolysis step, which commits the complex to the start site. Non-AUG codons favor a standby conformation that destabilizes eIF5 and likely overlaps the binding site of an ancient structural homolog. This branchpoint complements enforcement of start codon fidelity by upstream steps and intrinsically controls the efficiency of non-AUG initiation.

Summary The canonical B-form DNA helix and its protein interactions are well-characterized, yet the behavior of torsionally constrained DNA, ubiquitous in cells, remains underexplored. Using cryo-electron tomography (cryo-ET), we 3D-reconstructed entire negatively supercoiled DNA substrates with bound RNA polymerase (RNAP), revealing DNA supercoiling conformational diversity and its interplay with molecular motors. RNAP and DNA-binding proteins like dCas9 preferentially localize at plectoneme apices during transcription, acting as torsional blocks. Together, dCas9 and RNAP on opposing plasmid apices can segregate "twin-supercoiling domains" without the need for external DNA end-tethering. The generation of twin domains reveals as regions of reduced supercoiling and the presence of multiple transcribing RNAP complexes. Negative supercoiling and apex localization of RNAP favor initiation but disfavor elongation, leading to slow-moving RNAP clusters. Topoisomerase I relieves RNAP pauses by removing them from apical constraints; the resulting RNAP load-and-release process from the apex provides a molecular mechanism for the "transcriptional bursting" phenomenon.

Metabolism, gene expression, and signaling all require the adaptation of protein activity to the mixture of reactants and products in a cell. This trait to adapt, called allostery, is hardwired in the structure of proteins. Binding a ligand at one location in a protein can change distant locations, thus tuning protein activity. How allostery works has been subject of intense research since its discovery sixty years ago. The challenge is to understand the order of events that follow ligand-binding in the three-dimensional architecture of proteins. Here we simplify this task by studying allostery in DNA, a nearly one-dimensional system. DNA can transmit allosteric signals over many nanometers to generate cooperativity in the binding of transcription factors, an archetype of the long-range action of allostery. We found that binding of the transcription factor ComK amplifies intrinsic microsecond structural fluctuations in DNA many nanometers distant from the binding site. Yet, it is not protein binding per se, but the intrinsically disordered region (IDR) of the protein that amplifies these fluctuations. IDR removal does not only rigidify DNA, but it also abolishes allostery. The result is a structurally distorted protein-DNA complex that lost its function. These findings have important implications for our understanding of transcription activation and suggest a new functional role for IDRs in transcription factors.

Genome-wide DNA supercoiling is closely linked to chromatin organization and gene expression, yet the mechanisms establishing genome-scale supercoiling in living cells and its functional consequences remain unclear. Here, we show that genome-wide supercoiling arises from transcription-driven asymmetric topological relaxation together with contributions from SMC complexes in human cells. During RNA polymerase elongation, human topoisomerases preferentially relax positive over negative supercoils, leading to the accumulation of negative supercoiling around genes. This imbalance enriches supercoiling at transcriptionally active regions including TAD boundaries, and promotes the emergence of large-scale topology. In parallel, SMC complexes independently shape genome-wide supercoiling, with cohesin contributing to interphase topology and condensin establishing an overall positively supercoiled mitotic genome. Functionally, the accumulation of transcription-driven negative supercoiling represses local transcription, revealing a supercoiling-mediated negative feedback mechanism. Together, these findings define the mechanistic basis of genome-scale supercoiling in human cells and establish DNA topology as an integral regulatory layer of transcription.

Transcription factors often act within defined developmental windows, yet how naive pluripotent cells acquire competence to execute specific transcription factor-driven fate programmes remains unclear. Pioneer transcription factors that engage target sites in closed chromatin to initiate gene expression programmes often act at the top of hierarchies in cell identity transitions. However, we show that the ability of ASCL1 to induce a coherent neuronal programme emerges only after exit from pluripotency, coincident with progressive chromatin remodelling and accumulation of permissive histone marks at neuronal ASCL1 target sites. Binding analysis reveals that although ASCL1 can access a subset of neuronal loci in mESCs and EpiLCs, ASCL1 is preferentially diverted to non-neuronal sites, resulting in divergent transcriptional responses. Increasing global histone acetylation enhances activation of individual neuronal genes but is insufficient to drive full neuronal differentiation. In contrast, co-expression of the homeodomain transcription factor PHOX2B redirects ASCL1 towards neuronal targets while suppressing inappropriate programmes in mESCs. These findings demonstrate that ASCL1 pioneer activity is highly context-dependent and that developmental priming of chromatin is essential for appropriate lineage specification. HIGHLIGHTSO_LIEctopic ASCL1 drives non-neuronal transcriptional responses in naive and formative pluripotent cells C_LIO_LIASCL1 occupies distinct, predominantly non-neuronal genomic targets in pluripotent cells due to differential chromatin accessibility C_LIO_LIASCL1 pioneer activity is locus- and cell type-specific and predicted by histone acetylation status C_LIO_LICo-expression of ASCL1 with Phox2 homeodomain cofactors potentiates neuronal lineage acquisition in pluripotent cells C_LI

The emergence of SARS-CoV-2 has caused millions of deaths and excess morbidity in the worldwide population. In addition to its respiratory symptoms, SARS-CoV-2 has become known for its neurotropism and long-term neurological sequelae, with a post-acute infection syndrome commonly referred to as long-COVID. Next to the host receptor angiotensin-converting enzyme 2 (ACE2) additional interactions of the SARS-CoV-2 spike (S) protein have been described for neuronal co-receptors specific to the nervous system including cell adhesion protein contactin 1 (CNTN1). Details of the spike-CNTN1 interaction have remained elusive. Here, we quantified the spike-CNTN1 interaction by surface plasmon resonance and resolved the structure of the complex by single particle cryo-electron microscopy (cryo-EM). Spike and CNTN1 interact with nanomolar affinity, driven by an avidity effect and mediated by the horseshoe moiety of CNTN1. The cryo-EM structure reveals that the CNTN1 Ig1-4 horseshoe is wedged in between two receptor binding domains (RBDs) and interacts, through Ig3, with a unique receptor interface at the base of the RBD in the up-conformation. This receptor interface is not previously described for other spike receptors but overlaps with the epitopes of several neutralizing monoclonal antibodies. Comparison of our data with available spike structures suggests one spike trimer can bind three CNTN1 molecules, or alternatively, different co-receptors such as ACE2 and CNTN1, simultaneously. These findings shed new light on the molecular determinants of SARS-CoV-2 neurotropism.

ADP-ribosylation is known as a protein modification, yet recent studies have expanded the range of ADP-ribosyltransferase (ART) substrates to include nucleic acids. tRNA 2'-phosphotransferase 1 (TRPT1) and several PARP family members can modify the 5'-phosphate of single-stranded RNA. Here, we show that PARP10 and PARP15 extend this activity beyond the 5'-phosphate terminus and generate N3-ADP-ribosyl uracil and N1-ADP-ribosyl guanine bases. The base-linked ADP-ribosylation is reversed selectively by the macrodomain-containing hydrolase TARG1. In TARG1 knockout cells, N1-ADP-ribosyl guanine can be detected. Together, these findings establish guanine and uracil ADP-ribose as two novel nucleotide modifications and reveal PARP15 and TARG1 as an enzyme pair which can dynamically regulate guanine ADP-ribosylation in living cells.