Turning high-throughput structural biology into predictive inhibitor design.

A common challenge in drug design pertains to finding chemical modifications to a ligand that increases its affinity to the target protein. An underutilized advance is the increase in structural biology throughput, which has progressed from an artisanal endeavor to a monthly throughput of hundreds of different ligands against a protein in modern synchrotrons. However, the missing piece is a framework that turns high-throughput crystallography data into predictive models for ligand design. Here, we designed a simple machine learning approach that predicts protein-ligand affinity from experimental structures of diverse ligands against a single protein paired with biochemical measurements. Our key insight is using physics-based energy descriptors to represent protein-ligand complexes and a learning-to-rank approach that infers the relevant differences between binding modes. We ran a high-throughput crystallography campaign against the SARS-CoV-2 main protease (MPro), obtaining parallel measurements of over 200 protein-ligand complexes and their binding activities. This allows us to design one-step library syntheses which improved the potency of two distinct micromolar hits by over 10-fold, arriving at a noncovalent and nonpeptidomimetic inhibitor with 120 nM antiviral efficacy. Crucially, our approach successfully extends ligands to unexplored regions of the binding pocket, executing large and fruitful moves in chemical space with simple chemistry.

Liquid-liquid phase separation underpins the formation of replication factories in rotaviruses.

RNA viruses induce the formation of subcellular organelles that provide microenvironments conducive to their replication. Here we show that replication factories of rotaviruses represent protein-RNA condensates that are formed via liquid-liquid phase separation of the viroplasm-forming proteins NSP5 and rotavirus RNA chaperone NSP2. Upon mixing, these proteins readily form condensates at physiologically relevant low micromolar concentrations achieved in the cytoplasm of virus-infected cells. Early infection stage condensates could be reversibly dissolved by 1,6-hexanediol, as well as propylene glycol that released rotavirus transcripts from these condensates. During the early stages of infection, propylene glycol treatments reduced viral replication and phosphorylation of the condensate-forming protein NSP5. During late infection, these condensates exhibited altered material properties and became resistant to propylene glycol, coinciding with hyperphosphorylation of NSP5. Some aspects of the assembly of cytoplasmic rotavirus replication factories mirror the formation of other ribonucleoprotein granules. Such viral RNA-rich condensates that support replication of multi-segmented genomes represent an attractive target for developing novel therapeutic approaches.

Theoretical and Data-Driven Approaches for Biomolecular Condensates.

Biomolecular condensation processes are increasingly recognized as a fundamental mechanism that living cells use to organize biomolecules in time and space. These processes can lead to the formation of membraneless organelles that enable cells to perform distinct biochemical processes in controlled local environments, thereby supplying them with an additional degree of spatial control relative to that achieved by membrane-bound organelles. This fundamental importance of biomolecular condensation has motivated a quest to discover and understand the molecular mechanisms and determinants that drive and control this process. Within this molecular viewpoint, computational methods can provide a unique angle to studying biomolecular condensation processes by contributing the resolution and scale that are challenging to reach with experimental techniques alone. In this Review, we focus on three types of dry-lab approaches: theoretical methods, physics-driven simulations and data-driven machine learning methods. We review recent progress in using these tools for probing biomolecular condensation across all three fields and outline the key advantages and limitations of each of the approaches. We further discuss some of the key outstanding challenges that we foresee the community addressing next in order to develop a more complete picture of the molecular driving forces behind biomolecular condensation processes and their biological roles in health and disease.

Learning the molecular grammar of protein condensates from sequence determinants and embeddings.

Intracellular phase separation of proteins into biomolecular condensates is increasingly recognized as a process with a key role in cellular compartmentalization and regulation. Different hypotheses about the parameters that determine the tendency of proteins to form condensates have been proposed, with some of them probed experimentally through the use of constructs generated by sequence alterations. To broaden the scope of these observations, we established an in silico strategy for understanding on a global level the associations between protein sequence and phase behavior and further constructed machine-learning models for predicting protein liquid-liquid phase separation (LLPS). Our analysis highlighted that LLPS-prone proteins are more disordered, less hydrophobic, and of lower Shannon entropy than sequences in the Protein Data Bank or the Swiss-Prot database and that they show a fine balance in their relative content of polar and hydrophobic residues. To further learn in a hypothesis-free manner the sequence features underpinning LLPS, we trained a neural network-based language model and found that a classifier constructed on such embeddings learned the underlying principles of phase behavior at a comparable accuracy to a classifier that used knowledge-based features. By combining knowledge-based features with unsupervised embeddings, we generated an integrated model that distinguished LLPS-prone sequences both from structured proteins and from unstructured proteins with a lower LLPS propensity and further identified such sequences from the human proteome at a high accuracy. These results provide a platform rooted in molecular principles for understanding protein phase behavior. The predictor, termed DeePhase, is accessible from https://deephase.ch.cam.ac.uk/.

The Chromatin Regulator HMGA1a Undergoes Phase Separation in the Nucleus.

The protein high mobility group A1 (HMGA1) is an important regulator of chromatin organization and function. However, the mechanisms by which it exerts its biological function are not fully understood. Here, we report that the HMGA isoform, HMGA1a, nucleates into foci that display liquid-like properties in the nucleus, and that the protein readily undergoes phase separation to form liquid condensates in vitro. By bringing together machine-leaning modelling, cellular and biophysical experiments and multiscale simulations, we demonstrate that phase separation of HMGA1a is promoted by protein-DNA interactions, and has the potential to be modulated by post-transcriptional effects such as phosphorylation. We further show that the intrinsically disordered C-terminal tail of HMGA1a significantly contributes to its phase separation through electrostatic interactions via AT hooks 2 and 3. Our work sheds light on HMGA1 phase separation as an emergent biophysical factor in regulating chromatin structure.

Surface Electrostatics Govern the Emulsion Stability of Biomolecular Condensates.

Liquid-liquid phase separation underlies the formation of biological condensates. Physically, such systems are microemulsions that in general have a propensity to fuse and coalesce; however, many condensates persist as independent droplets in the test tube and inside cells. This stability is crucial for their function, but the physicochemical mechanisms that control the emulsion stability of condensates remain poorly understood. Here, by combining single-condensate zeta potential measurements, optical microscopy, tweezer experiments, and multiscale molecular modeling, we investigate how the nanoscale forces that sustain condensates impact their stability against fusion. By comparing peptide-RNA (PR25:PolyU) and proteinaceous (FUS) condensates, we show that a higher condensate surface charge correlates with a lower fusion propensity. Moreover, measurements of single condensate zeta potentials reveal that such systems can constitute classically stable emulsions. Taken together, these results highlight the role of passive stabilization mechanisms in protecting biomolecular condensates against coalescence.

Rapid Structural, Kinetic, and Immunochemical Analysis of Alpha-Synuclein Oligomers in Solution.

Oligomers comprised of misfolded proteins are implicated as neurotoxins in the pathogenesis of protein misfolding conditions such as Parkinson's and Alzheimer's diseases. Structural, biophysical, and biochemical characterization of these nanoscale protein assemblies is key to understanding their pathology and the design of therapeutic interventions, yet it is challenging due to their heterogeneous, transient nature and low relative abundance in complex mixtures. Here, we demonstrate separation of heterogeneous populations of oligomeric α-synuclein, a protein central to the pathology of Parkinson's disease, in solution using microfluidic free-flow electrophoresis. We characterize nanoscale structural heterogeneity of transient oligomers on a time scale of seconds, at least 2 orders of magnitude faster than conventional techniques. Furthermore, we utilize our platform to analyze oligomer ζ-potential and probe the immunochemistry of wild-type α-synuclein oligomers. Our findings contribute to an improved characterization of α-synuclein oligomers and demonstrate the application of microchip electrophoresis for the free-solution analysis of biological nanoparticle analytes.

Analysis of αB-crystallin polydispersity in solution through native microfluidic electrophoresis.

In recent years, significant advancements have been made in the understanding of the population distributions and dynamic oligomeric states of the molecular chaperone αB-crystallin and its core domain variants. In this work, we provide solution-phase evidence of the polydispersity of αB-crystallin using microfluidic methods, used for separating the oligomeric species present in solution according to their different electrophoretic mobilities on-chip in a matter of seconds. We in particular demonstrate that microfluidic high-field electrophoresis and diffusion can detect the oligomerisation of these highly dynamic molecular chaperones and characterise the dominant oligomeric species present. We thereby provide a robust microfluidic method for characterising the individual species within complex protein mixtures of biological relevance.

Quaternization of Vinyl/Alkynyl Pyridine Enables Ultrafast Cysteine-Selective Protein Modification and Charge Modulation.

Quaternized vinyl- and alkynyl-pyridine reagents were shown to react in an ultrafast and selective manner with several cysteine-tagged proteins at near-stoichiometric quantities. We have demonstrated that this method can effectively create a homogenous antibody-drug conjugate that features a precise drug-to-antibody ratio of 2, which was stable in human plasma and retained its specificity towards Her2+ cells. Finally, the developed warhead introduces a +1 charge to the overall net charge of the protein, which enabled us to show that the electrophoretic mobility of the protein may be tuned through the simple attachment of a quaternized vinyl pyridinium reagent at the cysteine residues. We anticipate the generalized use of quaternized vinyl- and alkynyl-pyridine reagents not only for bioconjugation, but also as warheads for covalent inhibition and as tools to profile cysteine reactivity.

Enhancing the Resolution of Micro Free Flow Electrophoresis through Spatially Controlled Sample Injection.

Free flow electrophoresis is a versatile technique for the continuous separation of mixtures with both preparative and analytical applications. Microscale versions of free flow electrophoresis are particularly attractive strategies because of their fast separation times, ability to work with small sample volumes, and large surface area to volume ratios facilitating rapid heat transfer, thus minimizing the detrimental effects of Joule heating even at high voltages. The resolution of microscale free flow electrophoresis, however, is limited by the broadening of the analyte beam in the microfluidic channel, an effect that becomes especially pronounced when the analyte is deflected significantly away from its original position. Here, we describe and demonstrate how restricting spatially the sample injection and collection to the regions where the gradients in the velocity distribution of the carrier medium are the smallest allows this broadening effect to be substantially suppressed and hence the resolution of microscale free flow electrophoresis devices to be increased. To demonstrate this concept, we fabricated microfluidic free flow electrophoresis devices with spatially restricted injection nozzles implemented through the use of multilayer soft-photolithography and further integrated quartz based observation areas for fluorescent detection and imaging. With these devices, we demonstrated a 5-fold reduction in the extent of beam broadening relative to conventional free flow electrophoresis approaches with nonrestricted sample introduction. The manifold enhancement in the achievable resolution of microscale free flow electrophoresis devices opens up the possibility of rapid separation and analysis of complex mixtures.

Real-Time Intrinsic Fluorescence Visualization and Sizing of Proteins and Protein Complexes in Microfluidic Devices.

Optical detection has become a convenient and scalable approach to read out information from microfluidic systems. For the study of many key biomolecules, however, including peptides and proteins, which have low fluorescence emission efficiencies at visible wavelengths, this approach typically requires labeling of the species of interest with extrinsic fluorophores to enhance the optical signal obtained - a process which can be time-consuming, requires purification steps, and has the propensity to perturb the behavior of the systems under study due to interactions between the labels and the analyte molecules. As such, the exploitation of the intrinsic fluorescence of protein molecules in the UV range of the electromagnetic spectrum is an attractive path to allow the study of unlabeled proteins. However, direct visualization using 280 nm excitation in microfluidic devices has to date commonly required the use of coherent sources with frequency multipliers and devices fabricated out of materials that are incompatible with soft lithography techniques. Here, we have developed a simple, robust, and cost-effective 280 nm LED platform that allows real-time visualization of intrinsic fluorescence from both unlabeled proteins and protein complexes in polydimethylsiloxane microfluidic channels fabricated through soft lithography. Using this platform, we demonstrate intrinsic fluorescence visualization of proteins at nanomolar concentrations on chip and combine visualization with micron-scale diffusional sizing to measure the hydrodynamic radii of individual proteins and protein complexes under their native conditions in solution in a label-free manner.

Combining Affinity Selection and Specific Ion Mobility for Microchip Protein Sensing.

The sensitive detection of proteins is a key objective in many areas of biomolecular science, ranging from biophysics to diagnostics. However, sensing in complex biological fluids is hindered by nonspecific interactions with off-target species. Here, we describe and demonstrate an assay that utilizes both the chemical and physical properties of the target species to achieve high selectivity in a manner not possible by chemical complementarity alone, in complex media. We achieve this objective through a combinatorial strategy, by simultaneously exploiting free-flow electrophoresis for target selection, on the basis of electrophoretic mobility, and conventional affinity-based selection. In addition, we demonstrate amplification of the resultant signal by a catalytic DNA nanocircuit. This approach brings together the inherent solution-phase advantages of microfluidic sample handling with isothermal, enzyme-free signal amplification. With this method, no surface immobilization or washing steps are required, and our assay is well suited to monoepitopic targets, presenting advantages over conventional ELISA techniques.

Gradient-free determination of isoelectric points of proteins on chip.

The isoelectric point (pI) of a protein is a key characteristic that influences its overall electrostatic behaviour. The majority of conventional methods for the determination of the isoelectric point of a molecule rely on the use of spatial gradients in pH, although significant practical challenges are associated with such techniques, notably the difficulty in generating a stable and well controlled pH gradient. Here, we introduce a gradient-free approach, exploiting a microfluidic platform which allows us to perform rapid pH change on chip and probe the electrophoretic mobility of species in a controlled field. In particular, in this approach, the pH of the electrolyte solution is modulated in time rather than in space, as in the case for conventional determinations of the isoelectric point. To demonstrate the general approachability of this platform, we have measured the isoelectric points of representative set of seven proteins, bovine serum albumin, ß-lactoglobulin, ribonuclease A, ovalbumin, human transferrin, ubiquitin and myoglobin in microlitre sample volumes. The ability to conduct measurements in free solution thus provides the basis for the rapid determination of isoelectric points of proteins under a wide variety of solution conditions and in small volumes.

Fluctuations in the Kinetics of Linear Protein Self-Assembly.

Biological systems are characterized by compartmentalization from the subcellular to the tissue level, and thus reactions in small volumes are ubiquitous in living systems. Under such conditions, statistical number fluctuations, which are commonly negligible in bulk reactions, can become dominant and lead to stochastic behavior. We present here a stochastic model of protein filament formation in small volumes. We show that two principal regimes emerge for the system behavior, a small fluctuation regime close to bulk behavior and a large fluctuation regime characterized by single rare events. Our analysis shows that in both regimes the reaction lag-time scales inversely with the system volume, unlike in bulk. Finally, we use our stochastic model to connect data from small-volume microdroplet experiments of amyloid formation to bulk aggregation rates, and show that digital analysis of an ensemble of protein aggregation reactions taking place under microconfinement provides an accurate measure of the rate of primary nucleation of protein aggregates, a process that has been challenging to quantify from conventional bulk experiments.

Protein Condensate Atlas from predictive models of heteromolecular condensate composition.

Biomolecular condensates help cells organise their content in space and time. Cells harbour a variety of condensate types with diverse composition and many are likely yet to be discovered. Here, we develop a methodology to predict the composition of biomolecular condensates. We first analyse available proteomics data of cellular condensates and find that the biophysical features that determine protein localisation into condensates differ from known drivers of homotypic phase separation processes, with charge mediated protein-RNA and hydrophobicity mediated protein-protein interactions playing a key role in the former process. We then develop a machine learning model that links protein sequence to its propensity to localise into heteromolecular condensates. We apply the model across the proteome and find many of the top-ranked targets outside the original training data to localise into condensates as confirmed by orthogonal immunohistochemical staining imaging. Finally, we segment the condensation-prone proteome into condensate types based on an overlap with biomolecular interaction profiles to generate a Protein Condensate Atlas. Several condensate clusters within the Atlas closely match the composition of experimentally characterised condensates or regions within them, suggesting that the Atlas can be valuable for identifying additional components within known condensate systems and discovering previously uncharacterised condensates.

Targeting nucleic acid phase transitions as a mechanism of action for antimicrobial peptides.

Antimicrobial peptides (AMPs), which combat bacterial infections by disrupting the bacterial cell membrane or interacting with intracellular targets, are naturally produced by a number of different organisms, and are increasingly also explored as therapeutics. However, the mechanisms by which AMPs act on intracellular targets are not well understood. Using machine learning-based sequence analysis, we identified a significant number of AMPs that have a strong tendency to form liquid-like condensates in the presence of nucleic acids through phase separation. We demonstrate that this phase separation propensity is linked to the effectiveness of the AMPs in inhibiting transcription and translation in vitro, as well as their ability to compact nucleic acids and form clusters with bacterial nucleic acids in bacterial cells. These results suggest that the AMP-driven compaction of nucleic acids and modulation of their phase transitions constitute a previously unrecognised mechanism by which AMPs exert their antibacterial effects. The development of antimicrobials that target nucleic acid phase transitions may become an attractive route to finding effective and long-lasting antibiotics.

Direct digital sensing of protein biomarkers in solution.

The detection of proteins is of central importance to biomolecular analysis and diagnostics. Typical immunosensing assays rely on surface-capture of target molecules, but this constraint can limit specificity, sensitivity, and the ability to obtain information beyond simple concentration measurements. Here we present a surface-free, single-molecule microfluidic sensing platform for direct digital protein biomarker detection in solution, termed digital immunosensor assay (DigitISA). DigitISA is based on microchip electrophoretic separation combined with single-molecule detection and enables absolute number/concentration quantification of proteins in a single, solution-phase step. Applying DigitISA to a range of targets including amyloid aggregates, exosomes, and biomolecular condensates, we demonstrate that the assay provides information beyond stoichiometric interactions, and enables characterization of immunochemistry, binding affinity, and protein biomarker abundance. Taken together, our results suggest a experimental paradigm for the sensing of protein biomarkers, which enables analyses of targets that are challenging to address using conventional immunosensing approaches.

Microchip Free-Flow Electrophoresis for Bioanalysis, Sensing, and Purification.

The separation of complex mixtures is ubiquitous throughout molecular biology, and techniques such as gel-based electrophoresis are common laboratory practice. Such methods are not without their drawbacks, however, which include non-specific interactions between analyte and the separation matrix, poor yields in purification and non-continuous analyte throughput. Microfluidic techniques, which exploit physical phenomena unique to the microscale, promise to improve many aspects of traditional laboratory procedures. These methods offer a quantitative, solution-based alternative to traditional gel electrophoresis, with rapid measurement times enabling the analysis of transient or weak biomolecular interactions that would be challenging to observe with traditional methods. Here, we present a protocol for the lithographic fabrication and operation of microfluidic chips capable of free-flow electrophoretic (FFE) fractionation and analysis of biological analytes. We demonstrate the efficacy of our approach through a protein-sensing methodology based on FFE fractionation of DNA-protein mixtures. In addition, the FFE technique described here can be readily adapted to suit a variety of preparative and analytical applications, providing information on the charge, zeta-potential, and interactions of analytes.

New Frontiers for Machine Learning in Protein Science.

Protein function is fundamentally reliant on inter-molecular interactions that underpin the ability of proteins to form complexes driving biological processes in living cells. Increasingly, such interactions are recognised as being formed between proteins that exist on a broad spectrum of dynamic conformational states and levels of intrinsic disorder. Additionally, the sizes of the structures formed can range from simple binary complexes to large dynamic biomolecular condensates measuring 100 nm or more. Understanding the parameters that govern such interactions, how they form, how they lead to function and what happens when they take place in unintended manners and lead to disease, represent some of the core questions for molecular biosciences. In light of recent advances made in solving the protein folding problem by machine learning methods, we discuss here the challenges and opportunities brought by these new data-driven approaches for the next frontiers of biomolecular science.

Rapid highly sensitive general protein quantification through on-chip chemiluminescence.

Protein detection and quantification is a routinely performed procedure in research laboratories, predominantly executed either by spectroscopy-based measurements, such as NanoDrop, or by colorimetric assays. The detection limits of such assays, however, are limited to µ M concentrations. To establish an approach that achieves general protein detection at an enhanced sensitivity and without necessitating the requirement for signal amplification steps or a multicomponent detection system, here, we established a chemiluminescence-based protein detection assay. Our assay specifically targeted primary amines in proteins, which permitted characterization of any protein sample and, moreover, its latent nature eliminated the requirement for washing steps providing a simple route to implementation. Additionally, the use of a chemiluminescence-based readout ensured that the assay could be operated in an excitation source-free manner, which did not only permit an enhanced sensitivity due to a reduced background signal but also allowed for the use of a very simple optical setup comprising only an objective and a detection element. Using this assay, we demonstrated quantitative protein detection over a concentration range of five orders of magnitude and down to a high sensitivity of 10 pg mL - 1 , corresponding to pM concentrations. The capability of the platform presented here to achieve a high detection sensitivity without the requirement for a multistep operation or a multicomponent optical system sets the basis for a simple yet universal and sensitive protein detection strategy.