Explore the vast range of titles available.

Protein-based therapeutics have led to new paradigms in disease treatment. Projected to be half of the top ten selling drugs in 2023, proteins have emerged as rivaling and, in some cases, superior alternatives to historically used small molecule-based medicines. This review chronicles both well-established and emerging design strategies that have enabled this paradigm shift by transforming protein-based structures that are often prone to denaturation, degradation, and aggregation in vitro and in vivo into highly effective therapeutics. In particular, we discuss strategies for creating structures with increased affinity and targetability, enhanced in vivo stability and pharmacokinetics, improved cell permeability, and reduced amounts of undesired immunogenicity. Ebrahimi and Samanta review the key advances in the chemical and structural modification of proteins that have enabled their rise as indispensable tools in medicine and outline emerging protein engineering strategies that can potentially unlock structures with improved therapeutic properties.

Protein–protein interactions in cells are rapidly identified with improved proximity labeling methods. Protein interaction networks and protein compartmentalization underlie all signaling and regulatory processes in cells. Enzyme-catalyzed proximity labeling (PL) has emerged as a new approach to study the spatial and interaction characteristics of proteins in living cells. However, current PL methods require over 18 h of labeling time or utilize chemicals with limited cell permeability or high toxicity. We used yeast display-based directed evolution to engineer two promiscuous mutants of biotin ligase, TurboID and miniTurbo, which catalyze PL with much greater efficiency than BioID or BioID2, and enable 10-min PL in cells with non-toxic and easily deliverable biotin. Furthermore, TurboID extends biotin-based PL to flies and worms.

Targeted degradation of cell surface and extracellular proteins via lysosomal delivery is an important means to modulate extracellular biology. However, these approaches have limitations due to lack of modularity, ease of development, restricted tissue targeting and applicability to both cell surface and extracellular proteins. We describe a lysosomal degradation strategy, termed cytokine receptor-targeting chimeras (KineTACs), that addresses these limitations. KineTACs are fully genetically encoded bispecific antibodies consisting of a cytokine arm, which binds its cognate cytokine receptor, and a target-binding arm for the protein of interest. We show that KineTACs containing the cytokine CXCL12 can use the decoy recycling receptor, CXCR7, to target a variety of target proteins to the lysosome for degradation. Additional KineTACs were designed to harness other CXCR7-targeting cytokines, CXCL11 and vMIPII, and the interleukin-2 (IL-2) receptor-targeting cytokine IL-2. Thus, KineTACs represent a general, modular, selective and simple genetically encoded strategy for inducing lysosomal delivery of extracellular and cell surface targets with broad or tissue-specific distribution. KineTACs are modular bispecific antibodies for degradation of extracellular and cell-surface proteins.

Adenine base editors (ABEs) catalyze A-to-G transitions showing broad applications, but their bystander mutations and off-target editing effects raise safety concerns. Through structure-guided engineering, we found ABE8e with an N108Q mutation reduced both adenine and cytosine bystander editing, and introduction of an additional L145T mutation (ABE9), further refined the editing window to 1–2 nucleotides with eliminated cytosine editing. Importantly, ABE9 induced very minimal RNA and undetectable Cas9-independent DNA off-target effects, which mainly installed desired single A-to-G conversion in mouse and rat embryos to efficiently generate disease models. Moreover, ABE9 accurately edited the A 5 position of the protospacer sequence in pathogenic homopolymeric adenosine sites (up to 342.5-fold precision over ABE8e) and was further confirmed through a library of guide RNA–target sequence pairs. Owing to the minimized editing window, ABE9 could further broaden the targeting scope for precise correction of pathogenic single-nucleotide variants when fused to Cas9 variants with expanded protospacer adjacent motif compatibility. bpNLS, bipartite nuclear localization signals. A precise adenine base editor variant, ABE9, was developed to generate single adenine transition at pathogenic homopolymeric adenine sites with minimal DNA/RNA off-target effects, suggesting promising potential for gene therapeutics.

The recent discovery of Is PETase, a hydrolytic enzyme that can deconstruct poly(ethylene terephthalate) (PET), has sparked great interest in biocatalytic approaches to recycle plastics. Realization of commercial use will require the development of robust engineered enzymes that meet the demands of industrial processes. Although rationally engineered PETases have been described, enzymes that have been experimentally optimized via directed evolution have not previously been reported. Here, we describe an automated, high-throughput directed evolution platform for engineering polymer degrading enzymes. Applying catalytic activity at elevated temperatures as a primary selection pressure, a thermostable Is PETase variant (HotPETase, T m  = 82.5 °C) was engineered that can operate at the glass transition temperature of PET. HotPETase can depolymerize semicrystalline PET more rapidly than previously reported PETases and can selectively deconstruct the PET component of a laminated multimaterial. Structural analysis of HotPETase reveals interesting features that have emerged to improve thermotolerance and catalytic performance. Our study establishes laboratory evolution as a platform for engineering useful plastic degrading enzymes. Enzymes for poly(ethylene terephthalate) (PET) deconstruction are of interest for plastics recycling, but reports on their directed evolution are missing. Now, an automated, high-throughput directed evolution platform is described, affording HotPETase that effectively achieves depolymerization above the glass transition temperature of PET.

The incorporation of light-responsive domains into engineered proteins has enabled control of protein localization, interactions and function with light. We integrated optogenetic control into proximity labeling, a cornerstone technique for high-resolution proteomic mapping of organelles and interactomes in living cells. Through structure-guided screening and directed evolution, we installed the light-sensitive LOV domain into the proximity labeling enzyme TurboID to rapidly and reversibly control its labeling activity with low-power blue light. ‘LOV-Turbo’ works in multiple contexts and dramatically reduces background in biotin-rich environments such as neurons. We used LOV-Turbo for pulse-chase labeling to discover proteins that traffic between endoplasmic reticulum, nuclear and mitochondrial compartments under cellular stress. We also showed that instead of external light, LOV-Turbo can be activated by bioluminescence resonance energy transfer from luciferase, enabling interaction-dependent proximity labeling. Overall, LOV-Turbo increases the spatial and temporal precision of proximity labeling, expanding the scope of experimental questions that can be addressed with proximity labeling. The light-sensitive LOV domain was engineered into the TurboID enzyme, creating ‘LOV-Turbo’. LOV-Turbo enables optogenetic control over proximity labeling, increasing the spatiotemporal precision of this technique.

Fluorescent biosensors enable the study of cell physiology with spatiotemporal resolution; yet, most biosensors suffer from relatively low dynamic ranges. Here, we introduce a family of designed Förster resonance energy transfer (FRET) pairs with near-quantitative FRET efficiencies based on the reversible interaction of fluorescent proteins with a fluorescently labeled HaloTag. These FRET pairs enabled the straightforward design of biosensors for calcium, ATP and NAD + with unprecedented dynamic ranges. The color of each of these biosensors can be readily tuned by changing either the fluorescent protein or the synthetic fluorophore, which enables simultaneous monitoring of free NAD + in different subcellular compartments following genotoxic stress. Minimal modifications of these biosensors furthermore allow their readout to be switched to fluorescence intensity, fluorescence lifetime or bioluminescence. These FRET pairs thus establish a new concept for the development of highly sensitive and tunable biosensors. Fluorescent proteins and HaloTag allow the flexible design of FRET-based biosensors with adjustable color using different fluorescent proteins or fluorophores and readout can be modified to fluorescence intensity, lifetime or bioluminescence.

Machine learning has been increasingly used for protein engineering. However, because the general sequence contexts they capture are not specific to the protein being engineered, the accuracy of existing machine learning algorithms is rather limited. Here, we report ECNet (evolutionary context-integrated neural network), a deep-learning algorithm that exploits evolutionary contexts to predict functional fitness for protein engineering. This algorithm integrates local evolutionary context from homologous sequences that explicitly model residue-residue epistasis for the protein of interest with the global evolutionary context that encodes rich semantic and structural features from the enormous protein sequence universe. As such, it enables accurate mapping from sequence to function and provides generalization from low-order mutants to higher-order mutants. We show that ECNet predicts the sequence-function relationship more accurately as compared to existing machine learning algorithms by using ~50 deep mutational scanning and random mutagenesis datasets. Moreover, we used ECNet to guide the engineering of TEM-1 β-lactamase and identified variants with improved ampicillin resistance with high success rates. Protein engineering is an active area of research in which machine learning has proven quite powerful. Here, the authors present a deep learning method that integrates both general and protein-specific sequence representations to improve the engineering of one’s protein of interest.

The organelles and subcellular compartments of yeast provide distinct environments and physical separation from the cytosol, enabling opportunities to target biosynthetic pathways to these compartments and enhance production of desirable compounds. Each subcellular compartment in yeast offers a unique physiochemical environment and metabolite, enzyme, and cofactor composition. While yeast metabolic engineering has focused on assembling pathways in the cell cytosol, there is growing interest in embracing subcellular compartmentalization. Beyond harnessing distinct organelle properties, physical separation of organelles from the cytosol has the potential to eliminate metabolic crosstalk and enhance compartmentalized pathway efficiency. In this Perspective we review the state of the art in yeast subcellular engineering, highlighting the benefits of targeting biosynthetic pathways to subcellular compartments, including mitochondria, peroxisomes, the ER and/or Golgi, vacuoles, and the cell wall, in different yeast species. We compare the performances of strains developed with subcellular engineering to those of native producers or yeast strains previously engineered with cytosolic pathways. We also identify important challenges that lie ahead, which need to be addressed for organelle engineering to become as mainstream as cytosolic engineering in academia and industry.

A key bottleneck in the microbial production of therapeutic plant metabolites is identifying enzymes that can improve yield. The facile identification of genetically encoded biosensors can overcome this limitation and become part of a general method for engineering scaled production. We have developed a combined screening and selection approach that quickly refines the affinities and specificities of generalist transcription factors; using RamR as a starting point, we evolve highly specific (>100-fold preference) and sensitive (half-maximum effective concentration (EC50) < 30 μM) biosensors for the alkaloids tetrahydropapaverine, papaverine, glaucine, rotundine and noscapine. High-resolution structures reveal multiple evolutionary avenues for the malleable effector-binding site and the creation of new pockets for different chemical moieties. These sensors further enabled the evolution of a streamlined pathway for tetrahydropapaverine, a precursor to four modern pharmaceuticals, collapsing multiple methylation steps into a single evolved enzyme. Our methods for evolving biosensors enable the rapid engineering of pathways for therapeutic alkaloids.A combined screening and selection approach enables the evolution of the generalist transcription factor RamR into specific and sensitive biosensors for various alkaloids and in turn a streamlined pathway for tetrahydropapaverine biosynthesis.