A Phage-Assisted Continuous Selection Approach for Deep Mutational Scanning of Protein-Protein Interactions

ProteinGym: Large-Scale Benchmarks for Protein Design and Fitness Prediction

Predicting the effects of mutations in proteins is critical to many applications, from understanding genetic disease to designing novel proteins that can address our most pressing challenges in climate, agriculture and healthcare. Despite a surge in machine learning-based protein models to tackle these questions, an assessment of their respective benefits is challenging due to the use of distinct, often contrived, experimental datasets, and the variable performance of models across different protein families. Addressing these challenges requires scale. To that end we introduce ProteinGym, a large-scale and holistic set of benchmarks specifically designed for protein fitness prediction and design. It encompasses both a broad collection of over 250 standardized deep mutational scanning assays, spanning millions of mutated sequences, as well as curated clinical datasets providing high-quality expert annotations about mutation effects. We devise a robust evaluation framework that combines metrics for both fitness prediction and design, factors in known limitations of the underlying experimental methods, and covers both zero-shot and supervised settings. We report the performance of a diverse set of over 70 high-performing models from various subfields (eg., alignment-based, inverse folding) into a unified benchmark suite. We open source the corresponding codebase, datasets, MSAs, structures, model predictions and develop a user-friendly website that facilitates data access and analysis.

Virus-assisted directed evolution of biomolecules

Directed evolution is a powerful technique that uses principles of natural evolution to enable the development of biomolecules with novel functions. However, the slow pace of natural evolution does not support the demand for rapidly generating new biomolecular functions in the laboratory. Viruses offer a unique path to design fast laboratory evolution experiments, owing to their innate ability to evolve much more rapidly than most living organisms, facilitated by a smaller genome size that tolerate a high frequency of mutations, as well as a fast rate of replication. These attributes offer a great opportunity to evolve various biomolecules by linking their activity to the replication of a suitable virus. This review highlights the recent advances in the application of virus-assisted directed evolution of designer biomolecules in both prokaryotic and eukaryotic cells.

Accurate proteome-wide missense variant effect prediction with AlphaMissense

The vast majority of missense variants observed in the human genome are of unknown clinical significance. We present AlphaMissense, an adaptation of AlphaFold fine-tuned on human and primate variant population frequency databases to predict missense variant pathogenicity. By combining structural context and evolutionary conservation, our model achieves state-of-the-art results across a wide range of genetic and experimental benchmarks, all without explicitly training on such data. The average pathogenicity score of genes is also predictive for their cell essentiality, capable of identifying short essential genes that existing statistical approaches are underpowered to detect. As a resource to the community, we provide a database of predictions for all possible human single amino acid substitutions and classify 89% of missense variants as either likely benign or likely pathogenic.

Deep mutational scan of a drug efflux pump reveals its structure-function landscape

Drug efflux is a common resistance mechanism found in bacteria and cancer cells, but studies providing comprehensive functional insights are scarce. In this study, we performed deep mutational scanning (DMS) on the bacterial ABC transporter EfrCD to determine the drug efflux activity profile of more than 1,430 single variants. These systematic measurements revealed that the introduction of negative charges at different locations within the large substrate binding pocket results in strongly increased efflux activity toward positively charged ethidium, whereas additional aromatic residues did not display the same effect. Data analysis in the context of an inward-facing cryogenic electron microscopy structure of EfrCD uncovered a high-affinity binding site, which releases bound drugs through a peristaltic transport mechanism as the transporter transits to its outward-facing conformation. Finally, we identified substitutions resulting in rapid Hoechst influx without affecting the efflux activity for ethidium and daunorubicin. Hence, single mutations can convert EfrCD into a drug-specific ABC importer.

Fragment-based drug discovery supports drugging 'undruggable' protein-protein interactions

Protein-protein interactions (PPIs) have important roles in various cellular processes, but are commonly described as 'undruggable' therapeutic targets due to their large, flat, featureless interfaces. Fragment-based drug discovery (FBDD) has achieved great success in modulating PPIs, with more than ten compounds in clinical trials. Here, we highlight the progress of FBDD in modulating PPIs for therapeutic development. Targeting hot spots that have essential roles in both fragment binding and PPIs provides a shortcut for the development of PPI modulators via FBDD. We highlight successful cases of cracking the 'undruggable' problems of PPIs using fragment-based approaches. We also introduce new technologies and future trends. Thus, we hope that this review will provide useful guidance for drug discovery targeting PPIs.

Integrating deep mutational scanning and low-throughput mutagenesis data to predict the impact of amino acid variants

BACKGROUND

Evaluating the impact of amino acid variants has been a critical challenge for studying protein function and interpreting genomic data. High-throughput experimental methods like deep mutational scanning (DMS) can measure the effect of large numbers of variants in a target protein, but because DMS studies have not been performed on all proteins, researchers also model DMS data computationally to estimate variant impacts by predictors.

RESULTS

In this study, we extended a linear regression-based predictor to explore whether incorporating data from alanine scanning (AS), a widely used low-throughput mutagenesis method, would improve prediction results. To evaluate our model, we collected 146 AS datasets, mapping to 54 DMS datasets across 22 distinct proteins.

CONCLUSIONS

We show that improved model performance depends on the compatibility of the DMS and AS assays, and the scale of improvement is closely related to the correlation between DMS and AS results.

Scalable Functional Assays for the Interpretation of Human Genetic Variation

Scalable sequence-function studies have enabled the systematic analysis and cataloging of hundreds of thousands of coding and noncoding genetic variants in the human genome. This has improved clinical variant interpretation and provided insights into the molecular, biophysical, and cellular effects of genetic variants at an astonishing scale and resolution across the spectrum of allele frequencies. In this review, we explore current applications and prospects for the field and outline the principles underlying scalable functional assay design, with a focus on the study of single-nucleotide coding and noncoding variants.

Agrochemical control of gene expression using evolved split RNA polymerase

Chemically-inducible gene expression systems are valuable tools for rational control of gene expression both for basic research and biotechnology. However, most chemical inducers are confined to certain groups of organisms. Therefore, dissecting interactions between different organisms could be challenging using existing chemically-inducible systems. We engineered a mandipropamid-induced gene expression system (Mandi-T7) based on evolved split T7 RNAP system. As a proof-of-principle, we induced GFP expression in E. coli cells grown inside plant tissue.

Methods for the directed evolution of biomolecular interactions

Noncovalent interactions between biomolecules such as proteins and nucleic acids coordinate all cellular processes through changes in proximity. Tools that perturb these interactions are and will continue to be highly valuable for basic and translational scientific endeavors. By taking cues from natural systems, such as the adaptive immune system, we can design directed evolution platforms that can generate proteins that bind to biomolecules of interest. In recent years, the platforms used to direct the evolution of biomolecular binders have greatly expanded the range of types of interactions one can evolve. Herein, we review recent advances in methods to evolve protein-protein, protein-RNA, and protein-DNA interactions.

PPI-HotspotDB: Database of Protein-Protein Interaction Hot Spots

Single-point mutations of certain residues (so-called hot spots) impair/disrupt protein-protein interactions (PPIs), leading to pathogenesis and drug resistance. Conventionally, a PPI-hot spot is identified when its replacement decreased the binding free energy significantly, generally by ≥2 kcal/mol. The relatively few mutations with such a significant binding free energy drop limited the number of distinct PPI-hot spots. By defining PPI-hot spots based on mutations that have been manually curated in UniProtKB to significantly impair/disrupt PPIs in addition to binding free energy changes, we have greatly expanded the number of distinct PPI-hot spots by an order of magnitude. These experimentally determined PPI-hot spots along with available structures have been collected in a database called PPI-Hotspot^DB. We have applied the PPI-Hotspot^DB to create a nonredundant benchmark, PPI-Hotspot+PDB^BM, for assessing methods to predict PPI-hot spots using the free structure as input. PPI-Hotspot^DB will benefit the design of mutagenesis experiments and development of PPI-hot spot prediction methods. The database and benchmark are freely available at https://ppihotspot.limlab.dnsalias.org.

Systematic molecular evolution enables robust biomolecule discovery

Evolution occurs when selective pressures from the environment shape inherited variation over time. Within the laboratory, evolution is commonly used to engineer proteins and RNA, but experimental constraints have limited the ability to reproducibly and reliably explore factors such as population diversity, the timing of environmental changes and chance on outcomes. We developed a robotic system termed phage- and robotics-assisted near-continuous evolution (PRANCE) to comprehensively explore biomolecular evolution by performing phage-assisted continuous evolution in high-throughput. PRANCE implements an automated feedback control system that adjusts the stringency of selection in response to real-time measurements of each molecular activity. In evolving three distinct types of biomolecule, we find that evolution is reproducibly altered by both random chance and the historical pattern of environmental changes. This work improves the reliability of protein engineering and enables the systematic analysis of the historical, environmental and random factors governing biomolecular evolution.

Phage-Assisted Continuous Evolution and Selection of Enzymes for Chemical Synthesis

Ligand-dependent biosensors are valuable tools for coupling the intracellular concentrations of small molecules to easily detectable readouts such as absorbance, fluorescence, or cell growth. While ligand-dependent biosensors are widely used for monitoring the production of small molecules in engineered cells and for controlling or optimizing biosynthetic pathways, their application to directed evolution for biocatalysts remains underexplored. As a consequence, emerging continuous evolution technologies are rarely applied to biocatalyst evolution. Here, we develop a panel of ligand-dependent biosensors that can detect a range of small molecules. We demonstrate that these biosensors can link enzymatic activity to the production of an essential phage protein to enable biocatalyst-dependent phage-assisted continuous evolution (PACE) and phage-assisted continuous selection (PACS). By combining these phage-based evolution and library selection technologies, we demonstrate that we can evolve enzyme variants with improved and expanded catalytic properties. Finally, we show that the genetic diversity resulting from a highly mutated PACS library is enriched for active enzyme variants with altered substrate scope. These results lay the foundation for using phage-based continuous evolution and selection technologies to engineer biocatalysts with novel substrate scope and reactivity.

A System for the Evolution of Protein-Protein Interaction Inducers

Molecules that induce interactions between proteins, often referred to as "molecular glues", are increasingly recognized as important therapeutic modalities and as entry points for rewiring cellular signaling networks. Here, we report a new PACE-based method to rapidly select and evolve molecules that mediate interactions between otherwise noninteracting proteins: rapid evolution of protein-protein interaction glues (rePPI-G). By leveraging proximity-dependent split RNA polymerase-based biosensors, we developed E. coli-based detection and selection systems that drive gene expression outputs only when interactions between target proteins are induced. We then validated the system using engineered bivalent molecular glues, showing that rePPI-G robustly selects for molecules that induce the target interaction. Proof-of-concept evolutions demonstrated that rePPI-G reduces the "hook effect" of the engineered molecular glues, due at least in part to tuning the interaction affinities of each individual component of the bifunctional molecule. Altogether, this work validates rePPI-G as a continuous, phage-based evolutionary technology for optimizing molecular glues, providing a strategy for developing molecules that reprogram protein-protein interactions.

Contingency and chance erase necessity in the experimental evolution of ancestral proteins

The roles of chance, contingency, and necessity in evolution are unresolved because they have never been assessed in a single system or on timescales relevant to historical evolution. We combined ancestral protein reconstruction and a new continuous evolution technology to mutate and select proteins in the B-cell lymphoma-2 (BCL-2) family to acquire protein–protein interaction specificities that occurred during animal evolution. By replicating evolutionary trajectories from multiple ancestral proteins, we found that contingency generated over long historical timescales steadily erased necessity and overwhelmed chance as the primary cause of acquired sequence variation; trajectories launched from phylogenetically distant proteins yielded virtually no common mutations, even under strong and identical selection pressures. Chance arose because many sets of mutations could alter specificity at any timepoint; contingency arose because historical substitutions changed these sets. Our results suggest that patterns of variation in BCL-2 sequences – and likely other proteins, too – are idiosyncratic products of a particular and unpredictable course of historical events.

One of the most fundamental and unresolved questions in evolutionary biology is whether the outcomes of evolution are predictable. Is the diversity of life we see today the expected result of organisms adapting to their environment throughout history (also known as natural selection) or the product of random chance? Or did chance events early in history shape the paths that evolution could take next, determining the biological forms that emerged under natural selection much later?

These questions are hard to study because evolution happened only once, long ago. To overcome this barrier, Xie, Pu, Metzger et al. developed an experimental approach that can evolve reconstructed ancestral proteins that existed deep in the past. Using this method, it is possible to replay evolution multiple times, from various historical starting points, under conditions similar to those that existed long ago. The end products of the evolutionary trajectories can then be compared to determine how predictable evolution actually is.

Xie, Pu, Metzger et al. studied proteins belonging to the BCL-2 family, which originated some 800 million years ago. These proteins have diversified greatly over time in both their genetic sequences and their ability to bind to specific partner proteins called co-regulators. Xie, Pu, Metzger et al. synthesized BCL-2 proteins that existed at various times in the past. Each ancestral protein was then allowed to evolve repeatedly under natural selection to acquire the same co-regulator binding functions that evolved during history.

At the end of each evolutionary trajectory, the genetic sequence of the resulting BCL-2 proteins was recorded. This revealed that the outcomes of evolution were almost completely unpredictable: trajectories initiated from the same ancestral protein produced proteins with very different sequences, and proteins launched from different ancestral starting points were even more dissimilar.

Further experiments identified the mutations in each trajectory that caused changes in coregulator binding. When these mutations were introduced into other ancestral proteins, they did not yield the same change in function. This suggests that early chance events influenced each protein’s evolution in an unpredictable way by opening and closing the paths available to it in the future.

This research expands our understanding of evolution on a molecular level whilst providing a new experimental approach for studying evolutionary drivers in more detail. The results suggest that BCL-2 proteins, in all their various forms, are unique products of a particular, unpredictable course of history set in motion by ancient chance events.

Systems for in vivo hypermutation: a quest for scale and depth in directed evolution

Traditional approaches to the directed evolution of genes of interest (GOIs) place constraints on the scale of experimentation and depth of evolutionary search reasonably achieved. Engineered genetic systems that dramatically elevate the mutation of target GOIs in vivo relieve these constraints by enabling continuous evolution, affording new strategies in the exploration of sequence space and fitness landscapes for GOIs. We describe various in vivo hypermutation systems for continuous evolution, discuss how different architectures for in vivo hypermutation facilitate evolutionary search scale and depth in their application to problems in protein evolution and engineering, and outline future opportunities for the field.

Phage-assisted continuous and non-continuous evolution

Directed evolution, which applies the principles of Darwinian evolution to a laboratory setting, is a powerful strategy for generating biomolecules with diverse and tailored properties. This technique can be implemented in a highly efficient manner using continuous evolution, which enables the steps of directed evolution to proceed seamlessly over many successive generations with minimal researcher intervention. Phage-assisted continuous evolution (PACE) enables continuous directed evolution in bacteria by mapping the steps of Darwinian evolution onto the bacteriophage life cycle and allows directed evolution to occur on much faster timescales compared to conventional methods. This protocol provides detailed instructions on evolving proteins using PACE and phage-assisted non-continuous evolution (PANCE) and includes information on the preparation of selection phage and host cells, the assembly of a continuous flow apparatus and the performance and analysis of evolution experiments. This protocol can be performed in as little as 2 weeks to complete more than 100 rounds of evolution (complete cycles of mutation, selection and replication) in a single PACE experiment.

Phage-DMS: A Comprehensive Method for Fine Mapping of Antibody Epitopes

Understanding the antibody response is critical to developing vaccine and antibody-based therapies and has inspired the recent development of new methods to isolate antibodies. Methods to define the antibody-antigen interactions that determine specificity or allow escape have not kept pace. We developed Phage-DMS, a method that combines two powerful approaches-immunoprecipitation of phage peptide libraries and deep mutational scanning (DMS)-to enable high-throughput fine mapping of antibody epitopes. As an example, we designed sequences encoding all possible amino acid variants of HIV Envelope to create phage libraries. Using Phage-DMS, we identified sites of escape predicted using other approaches for four well-characterized HIV monoclonal antibodies with known linear epitopes. In some cases, the results of Phage-DMS refined the epitope beyond what was determined in previous studies. This method has the potential to rapidly and comprehensively screen many antibodies in a single experiment to define sites essential for binding interactions.

Split T7 RNA polymerase biosensors to study multiprotein interaction dynamics

Protein-protein interactions (PPIs) are involved in nearly all cellular processes. PPIs are particularly crucial for mediating selectivity along signaling pathways. Thus, measuring the competitive interplay between PPIs in a cell is important for both understanding fundamental cellular regulation and developing therapeutics targeting those whose dysregulation is associated with disease. A variety of split protein reporter-based tools are available to measure if two proteins interact within a cell and thereby characterize the general determinants of their interactions. PPIs, however, occur within complex networks facilitated by dynamic biophysical nuances that determine activity and selectivity. Evolved, proximity-dependent split T7 RNA polymerase (RNAP) biosensors have recently been used to perform deep mutational scanning of PPI interfaces, and to create synthetic gene circuits. In this chapter, we present the application of proximity-dependent split RNAP biosensors as a method to measure multidimensional PPIs in live cells. Orthogonal split RNAP "tags" encode each interaction in a unique RNA signal, thereby enabling the study of multiple competitive PPIs in live cells. Each unique RNA signal can be quantified via established RNA analysis methods. Herein, we provide advice and protocols to aid other researchers in using the split RNAP biosensor, focusing primarily on how to detect multiple PPIs in mammalian cells, including their dynamic interplay in the presence of small molecule inhibitors.