Gerhard Hummer

The nuclear pore complex (NPC) regulates molecular traffic between nucleus and cytoplasm ^1,2 . Its permeability barrier is formed by intrinsically disordered phenylalanine-glycine-rich nucleoporins (FG-NUPs), whose physical state has long been debated ^3–13 . Deciphering how FG-NUPs behave inside living cells is crucial for understanding how NPCs achieve both selective and rapid transport. By combining site-specific labelling with time-resolved fluorescence anisotropy, here we show that FG domains exhibit nanosecond-scale, liquid-like segmental motion in live cells, whereas reconstituted FG condensates undergo a liquid-to-solid transition in vitro. Experiments and molecular-dynamics simulations further show that importin-β, a principal nuclear transport receptor, and O-linked β- N -acetylglucosamine, a key post-translational modification of FG-NUPs, together stabilize the dynamic FG network. Finely balanced FG–FG and FG–importin-β interactions maintain FG-NUP liquidity and generate a radial organization in which importin-β is enriched near the NPC periphery and extended FG domains remain in the central channel, forming the transport barrier. This organization explains recent MINFLUX observations ¹⁴ of rim-biased transport trajectories and is supported here by direct MINFLUX imaging of importin-β in mammalian cells. Together, our results reconcile conflicting models of FG-NUP organization ^4,5,7,9,15 and establish a framework for understanding how cellular context tunes intrinsically disordered proteins in health and disease.

The endoplasmic reticulum (ER) forms specialized microdomains that organize autophagy, lipid droplet formation, and the repair of damaged lysosomes. The PI3P-binding protein DFCP1/ZFYVE1 marks these microdomains, raising the question as to whether and how one protein could define such a subcompartment. Here we determine the structure of full-length human DFCP1 bound to membranes by cryo-electron tomography and single-particle cryo-electron microscopy. DFCP1 assembles into a helical coat, built from tetramers that stack around a constricted membrane tube. Its ATPase domains form the organizing core of the helix, while its ER-targeting domains insert deeply into the outer leaflet, and its FYVE domains grip PI3P. Coarse-grained molecular dynamics simulations show that the coat does not simply follow the curvature of the membrane but acts as a barrier to lipid diffusion, almost completely blocking the movement of its cognate lipid PI3P past the filament and concentrating it beneath the coat. Point mutations that break the assembly, membrane-insertion or PI3P-binding interfaces each impair lysosomal repair in cells, and the assembly is also required for the focal recruitment of the lipid transporter VPS13C. We propose that DFCP1 collars fence off a specialized ER microdomain, so that locally made lipids can reach sites of repair rather than escaping into bulk ER.

Phosphoinositide 3-kinases (PI3Ks) are major regulators of cell growth, proliferation and signalling, constituting key therapeutic targets in cancer, inflammation, and other diseases. Individual class I PI3K isoforms control key cellular functions, imposing the need to generate isoform-specific inhibition for therapeutic intervention. Roginolisib is a selective PI3Kδ inhibitor that shows promise for the treatment of cancer. Using a combination of X-ray crystallography, molecular dynamics simulations, and hydrogen-deuterium exchange mass spectrometry, we describe the mechanism driving roginolisib’s potent and isoform-selective inhibition of PI3Kδ. Roginolisib uniquely stabilises the catalytic C-terminal helix kα12, locking the enzyme in an inactive conformation. This binding mode also results in more sustained inhibition of phosphatidylinositol 3,4,5-trisphosphate formation in tumour samples of CLL patients. Inhibition of PI3Ks by stabilization of helix kα12 into an inactive conformation has not, to the best of our knowledge, been described before and may provide the basis for novel, more selective and effective pharmacological strategies.

Mitochondria import hundreds of proteins into the matrix and the inner membrane (IM). The translocase of the IM of mitochondria (TIM23 complex) sorts these proteins using topogenic signals in their structure and directs them into the matrix or inserts them into the IM. How these signals are deciphered is unknown. Here we show that Tim17, the protein-conducting subunit of the TIM23 complex, contains a highly conserved selectivity filter for incoming proteins. Using a fluorescence-based sorting reporter, we identified the mitochondrial oxidoreductase Dmo2 (akin to DMAC1 in humans) as factor for intramitochondrial protein sorting. Molecular dynamics simulations show that a Dmo2-dependent disulfide bond into Tim17 constitutes a negatively charged selectivity filter at the vestibule of the TIM23 complex that arrests membrane proteins and facilitates the lateral insertion. Our work provides a molecular rationale as to how the TIM23 translocase differentiates between soluble and membrane proteins.

Root-mean-square deviation (RMSD) is the standard metric of structural comparison in molecular dynamics (MD) simulations. In its conventional form, RMSD assigns equal weight to all atoms regardless of mobility. Hence, flexible loops and disordered regions can dominate a global RMSD, while the rigid functional core contributes negligibly to the overall metric. To address this issue, we introduce the Bayes-optimal RMSD (BRMSD), which optimizes per-atom weights jointly with structural averages by maximizing a Bayesian posterior. In a trade-off between low RMSD and weight uniformity, a position-fluctuation parameter σ controls the transition from classical RMSD (σ → ∞) to a progressive focus on a rigid core (σ → 0). The BRMSD framework supports analysis modules for structural alignment, focused alignment onto a user-specified domain, trajectory smoothing, soft K-means conformational clustering, and rigid-domain identification. These modules are implemented in the open-source Python package BRMSD and benchmarked on two MD systems, the endoplasmic reticulum translocon-associated protein SND3 and the phosphotransferase adenylate kinase.

Disordered proteins are challenging targets for drug design because they lack well-defined binding pockets. Small molecules can bind them, but when the bound state remains conformationally heterogeneous, these interactions are not readily characterized in terms of conventional binding pockets. Here, we use the concept of binding paths to describe how small molecules interact with disordered proteins. A binding path is a stochastic sequence of transient residue clusters that a ligand engages as it diffuses across a disordered protein, with each cluster forming and breaking up on time scales comparable to the overall residence time. We formalize this approach using a Markov state model (MSM), in which each state corresponds to a set of residues simultaneously contacting the ligand and the transition probabilities capture how the ligand moves between these sets. We apply this framework to the interactions of 10074-G5 with Aβ42 and of fasudil with α-synuclein, two disordered proteins associated with Alzheimer’s disease and Parkinson’s disease, respectively. The MSM enables both a dynamic mapping of transient binding sites and a calculation of dissociation constants consistent with experimental measurements. By generalizing the concept of static binding pockets to dynamic binding paths, this approach provides a quantitative framework for characterizing small-molecule binding by disordered proteins and for identifying druggable regions on these challenging targets.

The human Unc51-like kinase 1 (ULK1) autophagy-initiating complex consists of ULK1, FIP200, and the Hop/Rev7/Mad2 (HORMA) domain heterodimer ATG13:ATG101. Phosphatidylinositol 3-phosphate (PI3P) is essential to recruit ULK1 complex (ULK1C) to membranes for ULK1, but ULK1C subunits do not contain PI3P-binding domains. Here, we show that the ATG13:ATG101 dimer forms a complex with the PI3P-binding protein WD40 interacting with phosphoinositide protein 3 (WIPI3), as well as WIPI2. Bound to WIPI2 and WIPI3, ATG13:ATG101 inserts its Trp-Phe (WF) finger into the membrane. Molecular dynamics simulations show that WIPIs and the WF finger cooperatively stabilize the complex on membranes. Biochemical reconstitution and cell-based assays show that WIPI3:ATG13 engagement promotes ATG16L1 phosphorylation, autophagy, and mitophagy. A kinase domain (KD)–proximal Pro-Val-Pro (PVP) motif in the ULK1 intrinsically disordered region docks onto the ATG13:ATG101 HORMA dimer brings the ULK1 KD close to the membrane. The PVP motif is essential for in vitro ULK1 phosphorylation of ATG16L1 and important for autophagy and mitophagy. These data establish a stepwise pathway for recruitment of the ULK1 KD to the vicinity of the membrane surface.

In open mitosis, re-establishing nucleocytoplasmic compartmentalization requires the LEM2-ESCRT machinery to coordinate spindle clearance with sealing of the remaining nuclear envelope pores. The structural basis of this topologically unique and fundamental membrane-remodeling process is poorly understood. Here, we combine biochemical reconstitution, cryo-electron tomography, subtomogram averaging and large-scale molecular dynamics simulations to define the structural mechanism of nuclear membrane sealing. We structurally resolve that LEM2s winged-helix domain (WH) co-polymerizes with the ESCRT-II/III protein CHMP7 to form a membrane-bound scaffold whose geometry is progressively remodeled by downstream ESCRT-III proteins as it transitions from the flat membrane surrounding the pore towards the negatively curved membrane neck. In parallel, LEM2 positions its intrinsically disordered low-complexity domain within the pore, where condensation around spindle microtubules mechanically couples the membrane-ESCRT-LEM2 scaffold to the spindle and narrows the remaining diffusion path, restoring compartmentalization before membrane closure is complete. Remarkably, the LEM2-WH domain alone forms tightly constricted membrane tubes, coating the negatively curved inner surface, revealing an intrinsic membrane-remodeling activity of the receptor itself. Together, our work establishes a structural framework for how receptor-ESCRT co-polymerization, low complexity domain-mediated sealing and receptor-driven membrane remodeling guide nuclear-envelope pores from spindle-containing openings to terminal constriction and fusion.

Chemotherapy-induced peripheral neuropathy (CIPN) is a common and dose-limiting side effect of first-line chemotherapeutics such as paclitaxel, for which effective treatments are currently unavailable. We found that plasma collected 24 h after paclitaxel administration enhanced sensory neuron activation. To identify potential mediators of this effect, we applied a broad LC-HRMS-based lipidomics approach to quantify plasma lipids in breast cancer patients before and 24 h after treatment. We observed significant elevations in lysophospholipids with 18:1 fatty acid chain, especially lysophosphatidylcholine (LPC) and lysophosphatidylglycerol (LPG) after paclitaxel treatment. Assessing their activity on sensory neurons revealed that these unsaturated lysophospholipids, but not their saturated counterparts, activated the transient receptor potential vanilloid 1 (TRPV1) channel, a key mediator of nociception. To investigate TRPV1–LPL interactions, we combined molecular dynamics simulations with targeted mutagenesis. Unlike LPC, which is zwitterionic due to its choline headgroup, LPG carries a negative charge and contains a glycerol headgroup instead. As a result of lacking a positive countercharge in its headgroup, LPG adopts a more tilted orientation within the vanilloid binding pocket (VBP) and forms stronger interactions with residues S510 and S512. Overall, the agonist’s potency appears to depend on its interaction with and proximity to Y511, which was confirmed through electrophysiological recordings and calcium imaging using mutated hTRPV1 channels. These findings reveal lipid–channel interactions and provide mechanistic insight into endogenous lipid modulation of pain signaling in the context of paclitaxel-induced CIPN.

Organelle membrane identity is encoded by protein and lipid composition. Human cells produce thousands of distinct lipid species for this purpose. Yet, the precise molecular functions of most lipids remain unknown, owing to limited methods for in situ investigations of individual lipid species. Here, we use minimally modified bifunctional lipid probes to detect and functionally characterize lipid-protein interactions across time and subcellular compartments. We leverage this approach to map time-resolved protein interactomes of individual lipids representing major membrane lipid classes as they are transported through the organelle system of human cells. We identify hundreds of unique lipid-protein interaction candidates, measure lipid-induced protein abundance changes, and distinguish between directly and indirectly interacting subunits of associated protein complexes. Informed by this multimodal dataset, we interrogate the lipid specificity of membrane organizing complexes, in particular the selective co-association of phosphatidylethanolamine with the ancestral subunits of the nuclear pore (NDC1) and MICOS (MIC60) complexes. Using super-resolution microscopy and molecular dynamics simulations, we demonstrate that phosphatidylethanolamine stabilizes both NDC1 and MIC60 in highly curved membrane nanodomains, indicating that major membrane organizing complexes are positioned by selective lipid-protein interactions. Altogether, our experimental strategy provides a blueprint for discovering and characterizing cellular functions of individual lipids.