Design of modular, transferable protein assemblies hasbroad applicabilityand in structural biology could help with the ever-troublesome crystallizationbottleneck, including finding robustly behaved protein crystals forrapidly characterizing ligands or drug candidates or generating multiplepolymorphs to illuminate diverse conformations. Nanobodies as crystallizationchaperones are well-established but still unreliable, as we show here.Instead, we show an exemplar of how robust crystallization behaviorcan be engineered by exploring many combinations (>200) of nanobodysurface mutations over several iterations. Critically, what neededtesting was crystallization and diffraction quality, since target–nanobodybinding affinity is decoupled from crystallizability enhancement.Our study yielded multiple polymorphs, all mediated by the same interface,with dramatically improved resolution and diffraction reliabilityfor some mutants; we thus name them ‘Gluebodies’ (Gbs).We further demonstrate that these Gb mutations do transfer to someother targets, both for achieving robust crystallization in alternativepacking forms and for establishing the ability to crystallize a keyearly stage readout. Since the Gb interface is evidently a favoredinteraction, it may be broadly applicable for modular assembly; morespecifically, this work suggests that Gbs should be routinely attemptedfor crystallization whenever nanobodies are available.

This study investigates combined gadolinium and boronneutron capturetherapy (GdBNCT) with carbonic anhydrase IX (CA IX) inhibition formesothelioma treatment using a multifunctional platform (Gd-B-CA-SF).This platform incorporates a carborane cage for BNCT, a Gd complexfor GdNCT/MRI, and a sulfamido group for CA IX inhibition. In vitrostudies confirmed CA IX inhibition and selective binding of the compoundto AB22 mesothelioma cells, with a 10-fold higher boron uptake comparedto that in healthy Met-5a mesothelium. The combination of B and Gdin one molecule allows the exploration of potential additional effectscoming from the nuclides 10B and 157Gd. Irradiationof AB22 cells showed complete tumor regrowth inhibition after treatmentwith 157Gd-10B-CA-SF, an effect strictly relatedto the presence of 157Gd and its localization on the cytosolicmembrane. In vivo MRI studies confirmed higher accumulation of Gd-B-CA-SFin AB22 tumors in mice, demonstrating effective targeting. In vivoNCT experiments showed reduced tumor growth in 157Gd-10B-CA-SF treated mice, confirming the effectiveness of both 157Gd and 10B, even at relatively low concentrations.These results demonstrate the translational potential of Gd-B-CA-SFas a theranostic agent for an innovative approach to cancer treatment.

Nature provides access to biological catalysts that can expand the chemical transformations accessible to synthetic chemists. Among these, α-ketoglutarate, non-heme iron-dependent (NHI) enzymes stand out as scalable biocatalysts for catalyzing selective oxidation reactions. Many NHI enzymes require protein engineering to improve their activity, selectivity, or stability. However, the reliance of this strategy on the innate stability of the enzyme can thwart the success of the engineering campaign. Harnessing innately stable enzymes can overcome these challenges and accelerate biocatalyst engineering. Herein, we highlight the use of ancestral sequence reconstruction (ASR) to mine for thermostable enzymes that can serve as superior starting points for protein engineering. In our effort to develop a biocatalytic route to tropolones, we identified an NHI enzyme that demonstrated poor stability, diminished activity at high substrate concentrations, and a limited substrate scope. We compared the in-lab evolution of the modern NHI enzyme and its ancestor, demonstrating the improved evolvability profile of the latter. By engineering the ancestral protein, we accessed variants with enhanced thermostability and expression, increased rates, and a substrate scope broader than those of their modern counterparts. Altogether, this work provides a strategy to rapidly access enzyme backbones that can accelerate engineering of more robust and synthetically useful NHI enzymes.

Proteins achieve diverse biological functions throughprecise sequence-structurerelationships, yet they can also function through statistical ensemblesrather than as individual, static entities. Inspired by this paradigm,recent work has explored random heteropolymers (RHPs) as synthetic,scalable, and versatile protein mimetics. RHPs have been found tofunction as polymer ensembles capable of folding, binding, catalyzing,and stabilizing biomolecules with control over the monomer sequence.In this Outlook, we highlight recent advances in the discovery andmechanistic understanding of functional RHPs, emphasizing their emergentbehaviors and utility across sustainability, human health, and pharmaceuticals.We discuss how autonomous experimentation, machine learning, and multiscalemodeling are converging to accelerate design and discovery in thisvast chemical space. By embracing statistical design principles, wepropose a new framework for creating functional polymers that mirrorbiological systems.

Generative machinelearning models for exploring chemical spacehave shown immense promise, but many molecules that they generateare too difficult to synthesize, making them impractical for furtherinvestigation or development. In this work, we present a novel approachby fine-tuning Meta’s Llama3 Large Language Models (LLMs) tocreate SynLlama, which generates full synthetic pathways made of commonlyaccessible building blocks and robust organic reaction templates.SynLlama explores a large synthesizable space using significantlyless data and offers strong performance in both forward and bottom-upsynthesis planning compared to other state-of-the-art methods. Wefind that SynLlama, even without training on external building blocks,can effectively generalize to unseen yet purchasable building blocks,meaning that its reconstruction capabilities extend to a broader synthesizablechemical space than those of the training data. We also demonstratethe use of SynLlama in a pharmaceutical context for synthesis planningof analog molecules and hit expansion leads for proposed inhibitorsof target proteins, offering medicinal chemists a valuable tool fordiscovery.

Antimicrobial resistance remains a formidable challengeto modernmedicine, with bacterial resistance mechanisms increasingly erodingthe utility of clinically important antibiotics. While recent effortshave expanded the antibacterial pipeline, the development of resistancein priority pathogens continues to exceed the pace of new drug development.One emerging strategy to overcome resistance is the rational designof hybrid antibiotics that engage multiple binding sites. Here wedescribe the design, synthesis, and microbiological and structuralcharacterization of hybrid antibiotics of azithromycin, tedizolid,and chloramphenicol that span the peptidyltransferase center (PTC)and nascent peptide exit tunnel (NPET) in the bacterial ribosome.We characterize the binding of four such hybrids by cryo-electronmicroscopy, granting insight into their molecular mechanisms of action.We identify a hybrid of azithromycin and tedizolid that is activeagainst a diverse panel of multidrug-resistant Gram-positive bacteriaand is minimally affected by ribosomal protection (ABC-F) resistancemechanisms. These results extend our understanding of ribosome inhibitionand provide a pipeline for the rational design of dual-action antibioticsthat target the ribosome. In a broader context, this work offers aframework for developing bifunctional inhibitors that engage adjacentbinding sites by means of a rational cycle of synthetic optimization,biological evaluation, and structural characterization.

Pancreatic ductaladenocarcinoma (PDAC) remains refractory to currentimmune checkpoint blockade (ICB) therapies, necessitating innovativetherapeutic strategies. Emerging evidence implicates aberrant sialoglycanupregulation as a key mediator of immune evasion in PDAC. Herein,we report Y-320, a highly potent oral sialylation inhibitor discoveredthrough high-throughput screening. Y-320 suppresses α-2,3/2,6-sialylationin PDAC cells (IC50 ≈ 200 nM) with >300-foldhigheractivity than the known pan-inhibitor P-3Fax-Neu5Ac. Structuralanalyses reveal competitive occupation of multiple sialyltransferases’substrate-binding pockets as Y-320’s action mechanism. In vivo, Y-320 significantly inhibits tumor growth and remodelsthe tumor immune microenvironment. Mechanistic studies establish thatthe therapeutic efficacy of Y-320 depends on the coordinated engagementbetween CD8+ T cell and macrophage. Importantly, Y-320synergizes with anti-PD-1 therapy to overcome ICB resistance in PDAC,demonstrating superior tumor suppression compared to monotherapies.Our findings demonstrate that Y-320 shows promise for use as a therapeuticagent for cancer and validates sialylation inhibition as a novel glycoimmunecheckpoint strategy for PDAC and other immunotherapy-resistant malignancies.

Multivariate (MTV) porous materials exhibit unique structuralcomplexitiesbased on their diverse spatial arrangements of multiple building blockcombinations. These materials possess potential synergistic functionalitiesthat exceed the sum of their individual components. However, the exponentiallyincreasing design complexity of these materials poses significantchallenges for accurate ground-state configuration prediction anddesign. To address this, we propose a Hamiltonian model for quantumcomputing that integrates compositional, structural, and balance constraintsdirectly into the Hamiltonian, enabling efficient optimization ofthe MTV configurations. The model employs a graph-based representationto encode linker types as qubits. Our framework enables quantum encodingof a vast linker design space, allowing representation of exponentiallymany configurations with linearly scaling qubit resources, and facilitatingefficient search for optimal structures based on predefined designvariables. To validate our model, a variational quantum circuit wasconstructed and executed using the Sampling Variational Quantum Eigensolver(VQE) algorithm in the IBM Qiskit. Simulations on experimentally knownMTV porous materials (e.g., Cu-THQ-HHTP, Py-MV-DBA-COF, MUF-7, andSIOC-COF2) successfully reproduced their ground-state configurations,demonstrating the validity of our model. Furthermore, VQE calculationswere performed on a real IBM 127-qubit quantum hardware for validationpurposes signaling a first step toward a practical quantum algorithmfor the rational design of porous materials.

Sodium-ion batteries (SIBs) are considered potentialalternativesto lithium-ion batteries (LIBs) due to the abundant resources andlow sodium cost. The rational nanostructural design for anode materialsplays a crucial role in SIBs. TiO2, as a common electrodematerial, suffers from the drawbacks of low specific surface areaand poor conductivity. To overcome these limitations, we propose astrategy combining solvent evaporation-induced self-assembly and chemicaloxidative polymerization to construct an ultrathin polypyrrole (PPy)-coatedmesoporous TiO2 microsphere (meso-TiO2@PPy)core–shell structure. The combination of the mesoporous structureand the conductive coating endows the micrometer-sized TiO2 spheres with high specific surface area, excellent conductivity,and abundant sodium-ion diffusion pathways, leading to a dominantpseudocapacitance (94%) of total charge storage. Remarkably, suchintegration allows for a high reversible capacity of 160.6 mAh g–1 at 1 A g–1, good rate performance,and stable cycling performance (capacity retention of 80.8% after2000 cycles). Our research provides a pathway for the design of compositiveanode materials for high-performance SIBs.