Shenzhen Bay Laboratory

We report a chiral bifunctional N-heterocyclic carbene (NHC) catalyst that enables enantio- and diastereoselective alkylation of homoenolates. Quantitative steric maps (%Vbur) of the catalytic pocket explain stereocontrol and guide catalyst optimization for metal-free SN2 alkylations with unactivated and bulky secondary halides (>20:1 dr, up to 99% ee)─a significant advance in homoenolate reactivity. The method’s usefulness is demonstrated through short syntheses of bioactive targets, including (+)-coerulescine, pyrroloindolines, and CRTH2 receptor antagonists, using a modular, transient acyl-trapping strategy. Mechanistic studies show a dynamic kinetic resolution (DKR) pathway and confirm the importance of noncovalent interactions in stereoselectivity. Additionally, the chiral NHC allows the isolation and X-ray crystallographic analysis of a stable triazolium-derived homoenolate intermediate, providing direct structural insight into a longstanding challenge in NHC catalysis. By linking homoenolate structure to reactivity, this work creates a framework for rational NHC catalyst design in stereoselective transformations.

Myocardial ischemia is the common etiology of heart failure (HF). However, the precise molecular mechanisms that govern the ischemic myocardium into HF remain poorly defined. Selective accumulation of succinate is a hallmark of ischemia. Succinate is a predominant regulator of protein succinylation modification. The present study unravels the novel role of succinate in ischemia-induced HF via succinylation. A clinical cohort study that was performed on 1554 Chinese patients with coronary artery disease (CAD) indicated that serum succinate levels were positively correlated with the increase of HF biomarker, decrease of left ventricular ejection fraction (LVEF), incidence of HF, and risk of death. In cardiomyocytes deprived of oxygen and glucose (OGD) and in mice subjected to ligation of the left anterior descending coronary artery (LAD), succinate levels and global protein succinylation were elevated. Succinylation proteomic analysis identified the α-subunit of mitochondrial ATP synthase (ATP5A1) as an important succinylated protein, and K531 was identified as the functional succinylation site. Ablation of succinylation by the ATP5A1-K531R mutant ameliorated OGD-induced cardiomyocyte death, mitochondrial dysfunction and energy metabolic dysfunction. K531R delivered by cardiac-specific adeno-associated virus (AAV9) achieved short-term cardioprotective effects against ischemic injury, and exerted prolonged protective effects against HF development. Sirtuin 5 was confirmed as a desuccinylase of ATP5A1, whereas carnitine palmitoyltransferase 1A (CPT1A) was recognized as a trans-succinylase. Mechanistically, succinylation of ATP5A1-K531 impeded the assembly of ATP synthase and impaired its activity. Elevated succinate potentially serves as a risk predictor of HF. Targeting desuccinylation of ATP5A1-K531 might be a promising therapeutic strategy.

Molecular graph neural networks have achieved remarkable success in property prediction, yet their architectural components remain largely domain-agnostic-designed for generic graphs rather than molecules specifically.We argue that this represents a missed opportunity, particularly at the graph pooling stage, where all atomic information is irreversibly compressed into a single molecular vector.Current pooling methods, from simple sum/mean aggregation to learned hierarchical schemes, ignore the rich substructure vocabulary-functional groups, pharmacophores, ring systems-that chemists have curated over decades and encoded in molecular fingerprints.To test whether this domain knowledge can be productively integrated into neural architecture design, we develop FPPOOL, a pooling framework that uses molecular fingerprints as chemistry-aware guides to decompose molecular graphs into biochemically meaningful groups before hierarchical aggregation.Across 40 benchmarks spanning diverse molecular properties and activity-cliff challenges, FPPOOL consistently outperforms conventional pooling approaches while providing built-in interpretability that traces predictions to specific functional units without post-hoc explanation methods.More broadly, our results suggest a design principle for molecular machine learning: rather than relying solely on end-to-end learning, systematically encoding established chemical knowledge into the architecture-at the pooling stage and potentially beyond-can yield models that are simultaneously more accurate, more interpretable, and more aligned with how chemists reason about molecular structure.

Living cells undergo dynamic biological processes. For example, motor proteins transport cargos by taking nanometer-sized steps. However, it is challenging to measure nanometer-sized steps in living cells. Using cell-permeable, extremely bright, and photostable deuterium congeners of tetramethyl(silicon)rhodamine (SiR-d12) connected chloroalkane linker to label single HaloTag-fused myosin VI in living cells and total internal reflection fluorescence microscopy (TIRFM), we measured nanometer-sized steps of single myosin VI in living cells. The measured step size of wild-type myosin VI was larger than that predicted from its short-lever arms. Furthermore, myosin VI harboring a mutation in the ATP-binding pocket exhibited longer dwell times between steps, reduced velocity, and shorter run lengths than wild-type myosin VI, underscoring the critical role of the ATP-binding pocket in motility. Therefore, our direct measurements of nanometer-sized steps of single motor proteins in living cells provide mechanistic insights into the dynamics and biological processes of motor proteins in living cells.

Topological structures in condensed matter systems unlock new possibilities for the development of nanoelectronic devices. However, the potential of antiferroelectrics to host topological features remains largely unexplored, constrained by significant energy barriers from antiparallel dipole coupling that suppress polarization rotation and challenges in high-quality film fabrication. Here for the first time, we find that, dislocations, the most common one-dimensional topological structures in crystals, exhibit unexpectedly strong couplings with polar topologies and induce ordered polar antihedgehog lattices in antiferroelectric PbZrO3 driven by the interplay of electrostrictive effect and the flexoelectric field. Combined atomic-resolution transmission electron microscopy and phase-field simulations, it is revealed that the polarizations converging at dislocation cores and diverging between dislocations define lattices characterized by checkerboard-like antihedgehogs, respectively. Unexpected interplay between polar and structural topologies establishes a new paradigm for topology design.

The Epstein-Barr virus (EBV) is associated with multiple lymphoid malignancies and autoimmune diseases. The glycoprotein gp350, the most abundant envelope protein on the EBV virion surface, has been the primary target for the vaccine design. However, all the attempts have failed to prevent asymptomatic infection in clinical trials. This setback highlights an incomplete understanding of the immune response to gp350 and emphasizes the need for more comprehensive antibody studies. Here, we identified three human-derived anti-gp350 monoclonal antibodies from a nonimmune human scFv library. Of these, the S54 antibody exhibited a potent neutralizing activity against virus infection of Raji B cells. Using cryo-EM, we resolved the structure of the S54-gp350 complex, revealing that the S54 epitope substantially overlaps with the CR2 recognition site. Our findings elucidate the molecular mechanism by which anti-gp350 antibodies block EBV infection of B cells in vitro, providing valuable insights for antibody-based therapy and vaccine development.