While there is an abundance of static data for the structure of biological macromolecules, the data regarding their folding mechanisms and dynamics is scarce, posing a challenge to the training of AI models. The World Model approach, which has been very successful in robotics, can come in handy in this regard. From limited data, it can build a latent, approximated representation of the spatio-temporal folding environment that can be used to train downstream AI models. Here, we developed DreamFold, a World Model-based generative framework to perform biomolecular simulations. DreamFold encodes protein structures and dihedral angle moves into latent vectors using variational autoencoders. Then, it uses another neural network to rapidly predict the next latent structure after a latent move, allowing the model to develop its own understanding of protein dynamics. Finally, in a "hallucinated" latent environment, an agent learns, through evolutionary algorithms, a policy to drive folding simulations toward a target. Latent configurations can then be decoded back into atomistic structures. DreamFold can compute protein folding pathways four orders of magnitude faster than molecular dynamics (MD), up to ~30,000×. Furthermore, for the monomeric proteins tested here (up to 449 aa), the computational cost scales linearly with the number of atoms (N) as O(N) (in comparison, MD scales as O(NlogN)). We have validated our results against established MD benchmarks and other available experimental data. Finally, we show how DreamFold allows us to identify folding intermediates with cryptic binding sites that are therapeutically valuable. Overall, DreamFold facilitates the study of protein dynamics by generative AI with applications in structure-based drug discovery.

Eukaryotic proteomes are regulated by the ubiquitin-proteasome system. Unwanted proteins are tagged with ubiquitin, which is recognized by the 26S proteasome for degradation. To enter the proteasome's 20S core particle and be hydrolyzed into peptides, the substrate must be unfolded and threaded through the 19S regulatory particle. Tyrosine "aromatic paddles" of the regulatory particle's Rpt motor subunits unfold and translocate the substrate by pulling it toward the core particle. We previously showed that for substrates degraded from their N-termini, polyglycine tracts inserted before a stable domain impair the proteasome's ability to grip or unfold the substrate at multiple points along the translocation pathway. Herein we compare unfolding from the N- and C-termini of model substrates. We find that a combination of local substrate structural elements, grip sequence, and possibly fundamental asymmetry of the proteasomal unfolding machinery that favors N-terminal unfolding affect the proteasome's ability to unfold substrates. Although the basic unfolding and translocation mechanism is conserved regardless of substrate orientation, we find differences in the ability of polyglycine tracts to disrupt unfolding from the N versus the C terminus. Finally, comparison of the rates of unfolding versus substrate release shows that different regions along the translocation pathway employ different mechanisms to facilitate unfolding and translocation. Interactions with aromatic paddles primarily increase the rate of substrate unfolding, while interactions predicted to take place with the core particle α-ring N-termini primarily decrease the rate of premature release of partially degraded substrates.

De novo protein design often produces thermostable proteins that denature above 100°C, which complicates the analysis of their stability. Thermostable proteins can be unfolded by combined chemical and thermal denaturation followed by global analysis of multiple melting curves. Here, we have developed CheMelt, a new online tool for global analysis of unfolding data via an intuitive graphical user interface. We use nanoscale differential scanning fluorimetry followed by CheMelt data analysis to dissect the combined thermal and chemical denaturation of 35 de novo designed protein binders. Thirteen present sufficient fluorescence changes to extract thermodynamic parameters of unfolding. These de novo designed proteins have systematically lower ΔCp and m-values than comparable natural proteins. We show that a high thermostability of a designed protein does not necessarily imply a high equilibrium stability, and demonstrate the potential of CheMelt in dissecting thermodynamic properties for protein design and engineering. CheMelt can be accessed at https://spc.embl-hamburg.de/app/chemelt.

HIV-1 Vpu supports viral adaptation through host-protein interactions. Although mainly membrane-associated, we recently identified a soluble Vpu form that forms a stable complex with Ca[2+]-bound calmodulin (Ca[2+]-CaM), potentially influencing Vpu trafficking. Here, to determine the binding affinity and identify regions of soluble Vpu involved in CaM binding, we used ensemble Förster resonance energy transfer (eFRET). We tested Cy3-labeled full-length (FL) Vpu, a C-terminal fragment (helices 2 and 3), and a Cy3-labeled FL Vpu V22A/W23Y/I33N mutant with substitutions of key residues in Vpu's helix 1 and helices 1-to-2 loop having a role in the interaction with Ca[2+]-CaM. All Vpu variants were labeled at residue L42C. Ca[2+]-CaM was tagged with Cy5 at residue S39C. eFRET analysis of 100 nM Cy3-Vpu variants mixed with Cy5-Ca[2+]-CaM (in the range 100 nM-2.5 μM) revealed heterocomplexes formation with characteristic dissociation constants (Kd) and binding free energies (∆G). FL Vpu-Ca[2+]-CaM showed highest stability (Kd ~74 nM, ∆G ~-9.7 kcal/mol), while the truncated C-terminal region and V22A/W23Y/I33N mutant formed weaker complexes with Ca[2+]-CaM (Kd~182 nM and 800 nM, ∆G ~-9.2 kcal/mol and ~-8.3 kcal/mol). The reduced Vpu-Ca[2+]-CaM stability after disruption of binding sites in and near Vpu helix 1 may explain how this interaction is regulated, including through lipid competition that promotes Vpu membrane insertion. We propose that, at the membrane, hydrophobic helix 1 dissociates from Ca[2+]-CaM and inserts into the lipid bilayer, weakening the complex and releasing CaM. These findings clarify HIV-1 Vpu interactions with cellular components and may inform antiviral development.

Pyridoxal-5'-phosphate (PLP)-dependent enzymes of fold type IV exhibit remarkable catalytic diversity, catalyzing either S- and R-selective transamination or β-elimination within a highly conserved structural scaffold. Here, we report the identification, biochemical characterization, and structural analysis of PLP fold type IV 4-amino-4-deoxychorismate lyase (ADCL) from Micromonospora aurantiaca (Micau5708). Sequence-structural analysis revealed a hybrid architecture of the active site of Micau5708. It combines a D-amino acid transaminase (DATA) conserved phenylalanine (F30) and arginine (R87) residues with an ADCL-specific threonine residue (T32) and a (R)-selective amine transaminase-specific histidine residue (H25). While Micau5708 shows only ADCL activity, substitution H25R introduces transamination activity, preserving ADCL activity. The 1.85 Å resolution crystal structure of Micau5708 in complex with substrate analog chorismate provides the first direct visualization of substrate binding in an ADCL active site. Site-directed mutagenesis, together with comparative analysis of DATAs of group II demonstrates that ADCL/DATA catalytic promiscuity is enabled by a specific amino acid combination in the active site: phenylalanine, threonine, and arginine residues (F30, T32, and R87), and arginine at position 25 with a flexible side chain. Furthermore, the Micau5708-chorismite complex structure has helped revise the catalytic mechanism of ADCLs, demonstrating the T32 assistance in protonation of the oxygen atom rather than the β-methylene carbon of the leaving enol-pyruvate. Together, these findings provide new insights into the structural basis of reaction specificity, catalytic promiscuity, and evolutionary plasticity within the PLP fold type IV enzyme superfamily. They also highlight the limitations of sequence-based functional prediction.

As an emergent field stimulated by insights from the biological, chemical, physical, and quantum sciences, quantum biology explores how quantum phenomena such as superradiance, tunneling, and entanglement can be leveraged to enhance, influence, and probe biological processes across a diversity of scales and physical degrees of freedom. This review surveys key theoretical, computational, and experimental developments over the last decade, providing relevant historical background and scientific context going back as far as Bohr's 1932 lecture on "Light and Life" in Copenhagen. Reflecting major new areas of investigation, biophysical intuitions, and advancements in critical areas of study from earlier decades, we highlight superradiance and quantum cooperativity; exciton-phonon interactions and long-range electrodynamic organization; quantum tunneling, entanglement, and dispersion effects in protein processes; ultraweak metabolic photon emissions; photosynthesis; magnetosensing and chirality-induced spin dynamics; and emerging quantum technologies employed to delicately observe biosystems or to harness life's complexity and hierarchical order. In each section of the review, we provide foundational information or primers to augment understanding of the underlying physical processes. We highlight various theoretical and experimental milestones while looking forward to new discoveries on these "supremely interesting [and] unsurveyably intricate" protein architectures, which Schrödinger envisioned in What is Life? would serve a far-reaching and consummate purpose of communication and information processing.

F1-ATPase achieves unidirectional rotation of its γ shaft through coordinated conformational cycling of the α3β3 ring, yet how shaft-ring interactions contribute to the directionality of rotation remains unclear. Here we analyzed an axle-less TF1, in which the lower γ region is truncated, by combining single-molecule rotation assays with cryo-EM structural analysis under catalysis conditions. Under ATP-saturated conditions, wild-type TF1 exhibited three pauses per turn corresponding to the catalytic dwells, whereas axle-less TF1 exhibited six, revealing an additional intermediate dwell that may be stabilized by γ truncation. This dwell occurred at 40° between the binding and catalytic dwells. High-speed recordings revealed frequent backsteps during the 40° transition between the intermediate and catalytic dwells. Dwell-time analysis indicated an approximately zero free-energy bias between these states, consistent with the absence of a detectable directional bias. Cryo-EM resolved the corresponding binding, intermediate, and catalytic dwell structures, showing a major β conformational change from 0° to 40°, but minimal β rearrangement between 40° and 80°, suggesting that the rotation is not coupled with a large β conformational transition. Thus, the 0-80° step in axle-less TF1 consists of two regimes: a directional 0-40° motion accompanied by a major β conformational change, followed by a 40-80° interval with little detectable β rearrangement and markedly reduced directional bias. These findings suggest that, in axle-less TF1, the latter half of the 80° substep proceeds through thermal diffusion without directional bias, and that the lower γ region may normally help bias this interval toward forward rotation.

Autoantibodies against tumor-associated autoantigens are clinically valuable biomarkers for cancer diagnosis; however, the structural determinants governing epitope selectivity remain unknown. Here, we investigated whether the intrinsic conformational stability of target autoantigens regulates the epitope propensity of tumor-associated autoantibodies in non-small cell lung cancer. Using a dual-antigen Luminex bead-based assay that presents each autoantigen in both native and S-cationized denatured forms, we directly compared IgG autoantibody reactivity toward conformational and linear epitopes across 11 autoantigens. Intrinsically disordered aggregation-prone autoantigens, including cancer/testis antigens, NY-ESO-1, and XAGE-1b, predominantly elicit linear epitope-directed responses, whereas thermodynamically stable soluble autoantigens generally drive conformational epitope recognition. Intriguingly, for autoantigens harboring both ordered and disordered segments, the immune response is strictly guided by domain-specific biophysics: while p53-specific antibodies preferentially target their flanking disordered regions, the Wilms' tumor protein 1 exceptionally drives conformational recognition directed toward its structured zinc-finger domains. Recombinant solubility in Escherichia coli broadly correlated with epitope class across all 11 autoantigens, confirming that prokaryotic folding efficiency generally reflects intrinsic conformational stability in vivo for autonomously folding monomeric cytosolic proteins. Independent computational validation was provided by the concordance between the experimental solubility ratios and AlphaFold3-derived Rosetta energy unit/solvent-accessible surface area values, demonstrating the convergence of thermodynamic estimates and patient-derived immune data within a unifying biophysical framework. These findings establish that the autoantibody epitope propensity is closely associated with the thermodynamic stability of the target autoantigen and provide a rational basis for tailoring antigen preparation strategies for autoantibody-based cancer diagnosis.

Drosophila behavior/human splicing (DBHS) proteins are involved at nearly every stage of the RNA lifecycle. There are three paralogues in mammals-SFPQ, NONO and PSPC1-which form both homodimer and heterodimers. It is likely that the non-redundant and overlapping roles of the different DBHS paralogues emerge from their dynamic and combinatorial nature; that is, their ability to mix and match dimer partners. A strong preference for heterodimerisation over homodimerisation has been demonstrated. However, the underlying molecular determinants of DBHS partner preference are poorly understood. This study describes the design, and biochemical and structural characterization of a chimeric DBHS protein designed to be a homodimer with an interface that emulates a DBHS heterodimer. In vitro experiments show that this chimera indeed remains as a homodimer even when in the presence of a native DBHS protein that prefers heterodimerisation. The ease of purification, production of high-quality crystals for X-ray diffraction, and perfect crystallographic symmetry of this engineered homodimer presents this as an attractive approach for future structural studies on DBHS proteins. This ability to engineer partner preference also highlights the potential for DBHS proteins to serve as versatile building blocks in synthetic biology. Furthermore, comparison of the dimer interface of this engineered homodimer with native DBHS proteins highlights subtle conformational differences in residues at the core of the dimer interface, indicating that maximized packing of a core tryptophan residue combined with increased hydrophobic or polar complementarity at the interface is likely to determine DBHS partner preference.

A central challenge in molecular biology is understanding how amino acid substitutions modulate various features of protein function and stability. To illuminate the complexities of this relationship, high-throughput (HTP) assays are increasingly used to assess site-saturating mutagenesis libraries. A common downstream analysis is to average the set of 20 outcomes at each amino acid position for comparison with structural and evolutionary features. Average values clearly identify positions that tolerate most substitutions (neutral positions) and positions where most substitutions abolish activity (toggle positions). However, average values conceal the existence of rheostat positions, where different amino acid substitutions sample a wide range of outcomes. To quantitatively identify rheostat positions, we previously developed a histogram-based analysis that we here expand by: (i) incorporating new position classes observed in experimental studies of rheostat positions; (ii) formalizing a hierarchy of class assignments; (iii) refining error-based identification of neutral positions; and (iv) statistically assessing the robustness of class assignments to changes in experimental and computational parameters. RheoScale 2.0 is implemented in Excel and newly implemented in Python for facile integration with existing HTP pipelines; all parameters are customizable. Example analyses are shown for three HTP datasets of the SARS-CoV-2 papain-like protease. Results illustrate two aspects that influence interpretation of HTP data: First, position assignments (and substitution outcomes) depend highly on the measured feature. Second, many protein positions play multiple roles in the sequence-structure-function relationship. The recognition of varied position roles will advance understanding of pathogen evolution, protein engineering, and variant interpretation for personalized medicine.

The integration of coarse-grained (CG) approaches into docking workflows offers a powerful strategy for modeling large biomolecular assemblies with reduced computational costs. We present here the implementation of the MARTINI2 CG force field into the HADDOCK3 integrative modeling platform. This development enables the use of the CG representations and parameters within HADDOCK3 for efficient sampling and scoring of large macromolecular complexes, including protein-protein and protein-nucleic acid complexes. The implementation takes advantage of the modular and flexible architecture of HADDOCK3, allowing a seamless combination of MARTINI2 representation with the various modules. Conversion from and to all-atom models is integrated into the CG modeling workflow. The performance of the protocol is first assessed on protein-protein and protein-DNA benchmarks and then illustrated on a few representative large-scale systems, demonstrating a significant reduction in computational costs while maintaining biologically relevant accuracy. HADDOCK3 is freely available from https://github.com/haddocking/haddock3.

Hyaluronic acid (HA) is a biologically versatile polysaccharide synthesized by vertebrates and several microbial pathogens. To date, Cryptococcus neoformans CPS1p is the only reported bona fide hyaluronic acid synthase (HAS) in fungi, which is functionally related to bacterial HASs. Considering the phylogenetic and biochemical connection between HASs and chitin synthases (CHSs), essential for fungal cell wall synthesis, it is reasonable to hypothesize the former might be more common in fungi than expected. In this work, a comprehensive in silico survey of putative HASs in the Fungal Tree of Life was carried out. 68 putative HASs, mainly in Basidiomycota, were found, although other artificial intelligence-inferred putative HASs were found among Ascomycota. Global fold and arrangement of essential amino acids were shared by all kingdoms' HASs; however, C. neoformas CPS1p and additional putative fungal HASs (fungal HASs) showed exclusive conserved sequence signatures. Moreover, fungal HASs bore an only 3-helices transmembrane pore and their gating loop, which regulates the entrance of substrates to the catalytic site, was directly connected to an also exclusive intrinsically disordered C-terminus. Phylogenetically, fungal HASs were found in a clade different to that of bacterial, animal and viral HASs, and might share a common ancestor with Class I, III or VI CHSs. The atypical features of fungal HASs could influence the size and biological role of the HA they potentially synthesize and also highlight regulatory differences among HASs at the level of the gating loop configuration.

Multispecific, multivalent antibodies (Abs) are a burgeoning class of drugs that dramatically expand the pharmacological repertoire beyond traditional therapeutic Abs. Here, we present a simple, modular approach to developing multispecific, multivalent Abs based on a Fab-phage library with a single light chain. Using this library, we created three Abs targeting unique sites on programmed death-ligand 1 (PD-L1) and another antibody targeting CD30. Biophysical and cellular characterization of these Abs demonstrated their functional equivalence to clinically relevant Abs targeting PD-L1 or CD30. We then combined these paratopes into a series of bispecific, tetravalent, triparatopic Abs that retained the functionality of the parental Abs. Structural analysis of each of the Abs in complex with their cognate antigens demonstrated the adaptability of the common light chain to form diverse paratopes with an array of distinct heavy chains.

Lactate dehydrogenases catalyze the reduction of pyruvate to lactate, the generation of which is of importance for the energetic metabolism of malignant cells. In particular, human cancer cells overexpress lactate dehydrogenase A (hLDH-A), whose catalytic activity depends on its assembly into the corresponding homotetramer (denoted as hLDH-5), suggesting that compounds interfering with the protein-protein interactions responsible for the generation of hLDH-5 should represent inhibitors of high selectivity. Not surprisingly, quite a number of competitive inhibitors were designed, synthesized, and tested against hLDH-5. However, only a few representatives of the repertoire accordingly obtained feature an appropriate selectivity. Here we show how the ionic strength and β-NADH binding affect the monomers-to-tetramer hLDH-A transition. Moreover, by taking advantage of the isolation of monomeric hLDH-A at neutral pH, we tested the effectiveness of cyclic peptides and long-chain dicarboxylates in hampering the assembly of hLDH-5. Interestingly, the cyclic peptide LCO15 and the long-chain dicarboxylate crocetin were found to be very effective inhibitors of hLDH-A: the catalytic activity of the enzyme was inhibited by 97% and 86% when these compounds were administered at 40 and 20 μM, respectively. Furthermore, the amount of lactate produced by MCF7 human cancer cells was found to decrease by 30% in the presence of 80 μM crocin (a diester of crocetin). Overall, our observations indicate that the assembly of hLDH-A into catalytically competent hLDH-5 can be appropriately inhibited by means of cyclic peptides and long-chain dicarboxylates.

Rigid enzymes catalyze chemical reactions by stabilizing the transition state through a specific conformation. Catalytic residues are precisely positioned, and enzyme dynamics are kept to a minimum. It is thought that conserved residues around the active site (second-shell residues) play a crucial role in positioning catalytic residues. Asp233 and Asp246 are two highly conserved second-shell residues in class A β-lactamases, rigid enzymes that inactivate β-lactam antibiotics. The two aspartates share a short hydrogen bond, linking β-strands 3 and 4. The role of this interaction in Mycobacterium tuberculosis β-lactamase BlaC was studied by mutating the Asp residues to Ala. Disruption of the hydrogen bond subtly affects the activity and stability of the enzyme, suggesting the interaction helps to fine-tune the active site. The effects are larger for BlaC D246A than for D233A, indicating that effects cannot solely be attributed to the loss of the hydrogen bond. Molecular dynamics calculations indicate a shift in the conformational landscape due to the mutations, altering the conformational equilibria of the catalytic residues toward less active states. The results illustrate that second-shell residues act as a complex network that supports the efficient positioning of the catalytic residues.