Protein Engineering Research Articles

A systematic analysis of regression models for protein engineering.

To optimize proteins for particular traits holds great promise for industrial and pharmaceutical purposes. Machine Learning is increasingly applied in this field to predict properties of proteins, thereby guiding the experimental optimization process. A natural question is: How much progress are we making with such predictions, and how important is the choice of regressor and representation? In this paper, we demonstrate that different assessment criteria for regressor performance can lead to dramatically different conclusions, depending on the choice of metric, and how one defines generalization. We highlight the fundamental issues of sample bias in typical regression scenarios and how this can lead to misleading conclusions about regressor performance. Finally, we make the case for the importance of calibrated uncertainty in this domain.

Protein Grafting Techniques: From Peptide Epitopes to Lasso-Grafted Neobiologics.

Protein engineering techniques have vastly expanded their domain of impact, notably following the success of antibodies. Likewise, smaller peptide therapeutics have carved an increasingly significant niche for themselves in the pharmaceutical landscape. The concept of grafting such peptide onto larger protein scaffolds, thus harvesting the advantages of both, has given rise to a variety of protein engineering strategies that are reviewed herein. We also describe our own "Lasso-Grafting" approach, which combines traditional grafting concepts with mRNA display to streamline the production of multiple grafted drug candidates for virtually any target.

Diverse crystalline protein scaffolds through metal-dependent polymorphism.

As protein crystals are increasingly finding diverse applications as scaffolds, controlled crystal polymorphism presents a facile strategy to form crystalline assemblies with controllable porosity with minimal to no protein engineering. Polymorphs of consensus tetratricopeptide repeat proteins with varying porosity were obtained through co-crystallization with metal salts, exploiting the innate metal ion geometric requirements. A single structurally exposed negative amino acid cluster was responsible for metal coordination, despite the abundance of negatively charged residues. Density functional theory calculations showed that while most of the crystals were the most thermodynamically stable assemblies, some were kinetically trapped states. Thus, crystalline porosity diversity is achieved and controlled with metal coordination, opening a new scope in the application of proteins as biocompatible protein-metal-organic frameworks (POFs). In addition, metal-dependent polymorphic crystals allow direct comparison of metal coordination preferences.

Structural elucidation of the mesothelin-mucin-16/CA125 interaction.

Mesothelin (MSLN) is a cell-surface glycoprotein expressed at low levels on normal mesothelium but overexpressed in many cancers. Mesothelin has been implicated to play role/s in cell adhesion and multiple signaling pathways. Mucin-16/CA125 is an enormous cell-surface glycoprotein, also normally expressed on mesothelium and implicated in the progression and metastasis of several cancers, and directly binds mesothelin. However, the precise biological function/s of mesothelin and mucin-16/CA125 remain mysterious. We report protein engineering and recombinant production, qualitative and quantitative binding studies, and a crystal structure determination elucidating the molecular-level details governing recognition of mesothelin by mucin-16/CA125. The interface is small, consistent with the ∼micromolar binding constant and is free of glycan-mediated interactions. Sequence comparisons and modeling suggest that multiple mucin-16/CA125 modules can interact with mesothelin through comparable interactions, potentially generating a high degree of avidity at the cell surface to overcome the weak affinity, with implications for functioning and therapeutic interventions.

Cold Spot SCANNER: Colab Notebook for predicting cold spots in protein–protein interfaces

BackgroundProtein–protein interactions (PPIs) are conveyed through binding interfaces or surface patches on proteins that become buried upon binding. Structural and biophysical analysis of many protein–protein interfaces revealed certain unique features of these surfaces that determine the energetics of interactions and play a critical role in protein evolution. One of the significant aspects of binding interfaces is the presence of binding hot spots, where mutations are highly deleterious for binding. Conversely, binding cold spots are positions occupied by suboptimal amino acids and several mutations in such positions could lead to affinity enhancement. While there are many software programs for identification of hot spot positions, there is currently a lack of software for cold spot detection.ResultsIn this paper, we present Cold Spot SCANNER, a Colab Notebook, which scans a PPI binding interface and identifies cold spots resulting from cavities, unfavorable charge-charge, and unfavorable charge-hydrophobic interactions. The software offers a Py3DMOL-based interface that allows users to visualize cold spots in the context of the protein structure and generates a zip file containing the results for easy download.ConclusionsCold spot identification is of great importance to protein engineering studies and provides a useful insight into protein evolution. Cold Spot SCANNER is open to all users without login requirements and can be accessible at: https://colab.research.google.com/github/sagagugit/Cold-Spot-Scanner/blob/main/Cold_Spot_Scanner.ipynb.

Protein engineering using circular permutation - structure, function, stability, and applications.

Protein engineering is important for creating novel variants from natural proteins, enabling a wide range of applications. Approaches such as rational design and directed evolution are routinely used to make new protein variants. Computational tools like de novo design can introduce new protein folds. Expanding the amino acid repertoire to include unnatural amino acids with non-canonical side chains in vitro by native chemical ligation and in vivo via codon expansion methods broadens sequence and structural possibilities. Circular permutation (CP) is an invaluable approach to redesigning a protein by rearranging the amino acid sequence, where the connectivity of the secondary structural elements is altered without changing the overall structure of the protein. Artificial CP proteins (CPs) are employed in various applications such as biocatalysis, sensing of small molecules by fluorescence, genome editing, ligand-binding protein switches, and optogenetic engineering. Many studies have shown that CP can lead to either reduced or enhanced stability or catalytic efficiency. The effects of CP on a protein's energy landscape cannot be predicted a priori. Thus, it is important to understand how CP can affect the thermodynamic and kinetic stability of a protein. In this review, we discuss the discovery and advancement of techniques to create protein CP, and existing reviews on CP. We delve into the plethora of biological applications for designed CP proteins. We subsequently discuss the experimental and computational reports on the effects of CP on the thermodynamic and kinetic stabilities of proteins of various topologies. An understanding of the various aspects of CP will allow the reader to design robust CP proteins for their specific purposes.

Green energy breakthroughs: Harnessing nano-catalysts and enzymatic catalysts for bioenergy generation

Microbial fuel cells and anaerobic digestion are promising bioenergy methods but inefficient and scalable. With its exceptional surface area-to-volume ratio, nano-catalysts are revolutionizing this field. Researchers have improved energy conversion efficiency by immobilizing or embedding nanoparticles on electrodes or catalytic supports to promote electron transport. Nanomaterials like graphene-based catalysts and metal nanoparticles speed up organic matter breakdown and make bioenergy systems more profitable. Researchers are working harder to find new bioenergy-generating methods to address environmental concerns and the demand for sustainable energy. The study discusses how nano-catalysts and enzymatic catalysts are revolutionizing bioenergy production in green energy. Additionally, enzymatic catalysts are the best bioenergy catalysts. Highly specialized, biocompatible, and renewable enzymes provide unrivaled biofuel production potential. Protein engineering and immobilization have improved enzymes for bioenergy systems. Enzymatic catalysts enable biofuel generation from more feedstocks without harsh chemical procedures by controlling biological reactions quickly and precisely. Combining nano- and enzymatic catalysts is promising. Nano-catalysts improve enzymatic processes by creating a favorable microenvironment, enhancing surface binding sites, and enabling substrate transport. This combination may maximize bioenergy systems' biofuel output and wastewater treatment capabilities. According to this study, integrating nano- and enzymatic catalysts is a green energy breakthrough. These advances will improve bioenergy generation by overcoming current techniques, opening new sustainable energy production prospects, and expediting our transition to a greener, more sustainable future.

A Structural In Silico Analysis of the Immunogenicity of L-Asparaginase from Penicillium cerradense

L-asparaginase is an essential drug used to treat acute lymphoid leukemia (ALL), a cancer of high prevalence in children. Several adverse reactions associated with L-asparaginase have been observed, mainly caused by immunogenicity and allergenicity. Some strategies have been adopted, such as searching for new microorganisms that produce the enzyme and applying protein engineering. Therefore, this work aimed to elucidate the molecular structure and predict the immunogenic profile of L-asparaginase from Penicillium cerradense, recently revealed as a new fungus of the genus Penicillium and producer of the enzyme, as a motivation to search for alternatives to bacterial L-asparaginase. In the evolutionary relationship, L-asparaginase from P. cerradense closely matches Aspergillus species. Using in silico tools, we characterized the enzyme as a protein fragment of 378 amino acids (39 kDa), including a signal peptide containing 17 amino acids, and the isoelectric point at 5.13. The oligomeric state was predicted to be a homotetramer. Also, this L-asparaginase presented a similar immunogenicity response (T- and B-cell epitopes) compared to Escherichia coli and Dickeya chrysanthemi enzymes. These results suggest a potentially useful L-asparaginase, with insights that can drive strategies to improve enzyme production.

Protein engineering of lipase A from Candida antarctica to improve esterification of tertiary alcohols.

Chiral tertiary alcohols are important organic compounds in science as well as in industry. However, their preparation in enantiomerically pure form is still a challenge due to their complex structure and steric hindrances compared with primary and secondary alcohols, so kinetic resolution could be an attractive approach. Lipase A from Candida antarctica (CAL-A) has been shown to catalyze the enantioselective esterification of various tertiary alcohols with excellent enantioselectivity but low activity. Here we report a mutagenesis study by rational design to improve CAL-A activity against tertiary alcohols. Single mutants of CAL-A were selected, expressed, immobilized and screened for esterification of the tertiary alcohol 1,2,3,4-tetrahydronaphthalene-1-ol. A double mutant V278S+S429G showed a 1.5-fold higher reaction rate than that of the wild type CAL-A, while maintaining excellent enantioselectivity.

Unraveling the Properties of Phage Display Fab Libraries and Their Use in the Selection of Gliadin-Specific Probes by Applying High-Throughput Nanopore Sequencing

Directed evolution is a pivotal strategy for new antibody discovery, which allowed the generation of high-affinity Fabs against gliadin from two antibody libraries in our previous studies. One of the libraries was exclusively derived from celiac patients’ mRNA (immune library) while the other was obtained through a protein engineering approach (semi-immune library). Recent advances in high-throughput DNA sequencing techniques are revolutionizing research across genomics, epigenomics, and transcriptomics. In the present work, an Oxford Nanopore in-lab sequencing device was used to comprehensively characterize the composition of the constructed libraries, both at the beginning and throughout the phage-mediated selection processes against gliadin. A customized analysis pipeline was used to select high-quality reads, annotate chain distribution, perform sequence analysis, and conduct statistical comparisons between the different selection rounds. Some immunological attributes of the most representative phage variants after the selection process were also determined. Sequencing results revealed the successful transfer of the celiac immune response features to the immune library and the antibodies derived from it, suggesting the crucial role of these features in guiding the selection of high-affinity recombinant Fabs against gliadin. In summary, high-throughput DNA sequencing has improved our understanding of the selection processes aimed at generating molecular binders against gliadin.

Improvement of Chalcone Synthase Activity and High-Efficiency Fermentative Production of (2S)-Naringenin via In Vivo Biosensor-Guided Directed Evolution.

Chalcone synthase (CHS) catalyzes the rate-limiting step of (2S)-naringenin (the essential flavonoid skeleton) biosynthesis. Improving the activity of the CHS by protein engineering enhances (2S)-naringenin production by microbial fermentation and can facilitate the production of valuable flavonoids. A (2S)-naringenin biosensor based on the TtgR operon was constructed in Escherichia coli and its detection range was expanded by promoter optimization to 0-300 mg/L, the widest range for (2S)-naringenin reported. The high-throughput screening scheme for CHS was established based on this biosensor. A mutant, SjCHS1S208N with a 2.34-fold increase in catalytic activity, was discovered by directed evolution and saturation mutagenesis. A pathway for de novo biosynthesis of (2S)-naringenin by SjCHS1S208N was constructed in Saccharomyces cerevisiae, combined with CHS precursor pathway optimization, increasing the (2S)-naringenin titer by 65.34% compared with the original strain. Fed-batch fermentation increased the titer of (2S)-naringenin to 2513 ± 105 mg/L, the highest reported so far. These findings will facilitate efficient flavonoid biosynthesis and further modification of the CHS in the future.

SpyTag Peptide with Alkoxyl Aspartic Acids for pH-Dependent Activation of the SpyCatcher/Tag System.

The SpyCatcher/SpyTag system is a protein pair that forms a covalent isopeptide bond without an additional energy supply. The ability to connect fused proteins makes this system an attractive tool for several protein engineering applications. Conditional activation of the SpyCatcher/SpyTag complex formation further expands the use of this system. Here, we evaluated the pH activation of SpyTag using alkoxyaspartic acids in the isopeptide-forming residue. We found that a peptide with an ethoxy group can be activated by hydrolysis under high pH conditions. However, the hydrolysis induces isoaspartate (isoAsp) formation, which is confirmed by an isoAsp-inserted short peptide. We overcame this problem by changing the C-terminal side of the aspartic acid position to Pro, which does not form isoAsp under high pH conditions. The findings of this study provide fundamental knowledge of the synthetic construction of the modified SpyTag peptide.

Thermodynamics of a hyperthermostable carboxylesterase from Anoxybacillus geothermalis D9

Hyperthermostable enzymes are highly desirable biocatalysts due to their exceptional stability at extreme temperatures. Recently, a hyperthermostable carboxylesterase EstD9 from Anoxybacillus geothermalis D9 was biochemically characterized. The enzyme exhibited remarkable stability at high temperature. In this study, we attempted to probe the conformational adaptability of EstD9 under extreme conditions via in silico approaches. Circular dichroism revealed that EstD9 generated new β-sheets at 80 °C, making the core of the hydrolase fold more stable. Interestingly, the profiles of molecular dynamics simulation showed the lowest scores of radius of gyration and solvent accessible surface area (SASA) at 80 °C. Three loops were responsible for protecting the catalytic site, which resided at the interface between the large and cap domains. To further investigate the structural adaptation in extreme conditions, the intramolecular interactions of the native structure were investigated. EstD9 revealed 18 hydrogen bond networks, 7 salt bridges, and 9 hydrophobic clusters, which is higher than the previously reported thermostable Est30. Collectively, the analysis indicates that intramolecular interactions and structural dynamics play distinct roles in preserving the overall EstD9 structure at elevated temperatures. This work is relevant to both fundamental and applied research involving protein engineering of industrial thermostable enzymes.

Computational scoring and experimental evaluation of enzymes generated by neural networks.

In recent years, generative protein sequence models have been developed to sample novel sequences. However, predicting whether generated proteins will fold and function remains challenging. We evaluate a set of 20 diverse computational metrics to assess the quality of enzyme sequences produced by three contrasting generative models: ancestral sequence reconstruction, a generative adversarial network and a protein language model. Focusing on two enzyme families, we expressed and purified over 500 natural and generated sequences with 70-90% identity to the most similar natural sequences to benchmark computational metrics for predicting in vitro enzyme activity. Over three rounds of experiments, we developed a computational filter that improved the rate of experimental success by 50-150%. The proposed metrics and models will drive protein engineering research by serving as a benchmark for generative protein sequence models and helping to select active variants for experimental testing.

For antibody sequence generative modeling, mixture models may be all you need.

Antibody therapeutic candidates must exhibit not only tight binding to their target but also good developability properties, especially low risk of immunogenicity. In this work, we fit a simple generative model, SAM, to sixty million human heavy and seventy million human light chains. We show that the probability of a sequence calculated by the model distinguishes human sequences from other species with the same or better accuracy on a variety of benchmark datasets containing >400 million sequences than any other model in the literature, outperforming large language models (LLMs) by large margins. SAM can humanize sequences, generate new sequences, and score sequences for humanness. It is both fast and fully interpretable. Our results highlight the importance of using simple models as baselines for protein engineering tasks. We additionally introduce a new tool for numbering antibody sequences which is orders of magnitude faster than existing tools in the literature. All tools developed in this study are available at https://github.com/Wang-lab-UCSD/AntPack. Supplementary data are available at Bioinformatics online.

Differentiating stable and unstable protein using convolution neural network and molecular dynamics simulations

Protein stability is a critical aspect of molecular biology and biochemistry, hinges on an intricate balance of thermodynamic and structural factors. Determining protein stability is crucial for understanding and manipulating biological machineries, as it directly correlated with the protein function. Thus, this study delves into the intricacies of protein stability, highlighting its dependence on various factors, including thermodynamics, thermal conditions, and structural properties. Moreover, a notable focus is placed on the free energy change of unfolding (ΔGunfolding), change in heat capacity (ΔCp) with protein structural transition, melting temperature (Tm) and number of disulfide bonds, which are critical parameters in understanding protein stability. In this study, a machine learning (ML) predictive model was developed to estimate these four parameters using the primary sequence of the protein. The shortfall of available tools for protein stability prediction based on multiple parameters propelled the completion of this study. Convolutional Neural Network (CNN) with multiple layers was adopted to develop a more reliable ML model. Individual predictive models were prepared for each property, and all the prepared models showed results with high accuracy. The R2 (coefficient of determination) of these models were 0.79, 0.78, 0.92 and 0.92, respectively, for ΔG, ΔCp, Tm and disulfide bonds. A case study on stability analysis of two homologous proteins was presented to validate the results predicted through the developed model. The case study included in silico analysis of protein stability using molecular docking and molecular dynamic simulations. This validation study assured the accuracy of each model in predicting the stability associated properties. The alignment of physics-based principles with ML models has provided an opportunity to develop a fast machine learning solution to replace the computationally demanding physics-based calculations used to determine protein stability. Furthermore, this work provided valuable insights into the impact of mutation on protein stability, which has implications for the field of protein engineering. The source codes are available at https://github.com/Growdeatechnology.

Enhancing the activity and succinyl-CoA specificity of 3-ketoacyl-CoA thiolase Tfu_0875 through rational binding pocket engineering

The 3-ketoacyl-CoA thiolase is the rate-limiting enzyme for linear dicarboxylic acids production. However, the promiscuous substrate specificity and suboptimal catalytic performance have restricted its application. Here we present both biochemical and structural analyses of a high-efficiency 3-ketoacyl-CoA thiolase Tfu_0875. Notably, Tfu_0875 displayed heightened activity and substrate specificity for succinyl-CoA, a key precursor in adipic acid production. To enhance its performance, a deep learning approach (DLKcat) was employed to identify effective mutants, and a computational strategy, known as the greedy accumulated strategy for protein engineering (GRAPE), was used to accumulate these effective mutants. Among the mutants, Tfu_0875N249W/L163H/E217L exhibited the highest specific activity (320% of wild-type Tfu_0875), the greatest catalytic efficiency (kcat/KM = 1.00 min−1mM−1), the highest succinyl-CoA specificity (KM = 0.59 mM, 28.1% of Tfu_0875) and dramatically reduced substrate binding energy (−30.25 kcal mol−1v.s. −15.94 kcal mol−1). A structural comparison between Tfu_0875N249W/L163H/E217L and the wild type Tfu_0875 revealed that the increased interaction between the enzyme and succinyl-CoA was the primary reason for the enhanced enzyme activity. This interaction facilitated rapid substrate anchoring and stabilization. Furthermore, a reduced binding pocket volume improved substrate specificity by enhancing the complementarity between the binding pocket and the substrate in stereo conformation. Finally, our rationally designed mutant, Tfu_0875N249W/L163H/E217L, increased the adipic acid titer by 1.35-fold compared to the wild type Tfu_0875 in shake flask. The demonstrated enzymatic methods provide a promising enzyme variant for the adipic acid production. The above effective substrate binding pocket engineering strategy can be beneficial for the production of other industrially competitive biobased chemicals when be applied to other thiolases.

Engineering of the Lrp/AsnC-type transcriptional regulator DecR as a genetically encoded biosensor for multilevel optimization of L-cysteine biosynthesis pathway in Escherichia coli.

L-cysteine is an important sulfur-containing amino acid being difficult to produce by microbial fermentation. Due to the lack of high-throughput screening methods, existing genetically engineered bacteria have been developed by simply optimizing the expression of L-cysteine-related genes one by one. To overcome this limitation, in this study, a biosensor-based approach for multilevel biosynthetic pathway optimization of L-cysteine from the DecR regulator variant of Escherichia coli was applied. Through protein engineering, we obtained the DecRN29Y/C81E/M90Q/M99E variant-based biosensor with improved specificity and an 8.71-fold increase in dynamic range. Using the developed biosensor, we performed high-throughput screening of the constructed promoter and RBS combination library, and successfully obtained the optimized strain, which resulted in a 6.29-fold increase in L-cysteine production. Molecular dynamics (MD)simulations and electrophoretic mobility shift analysis (EMSA)showed that the N29Y/C81E/M90Q/M99E variant had enhanced induction activity. This enhancement may be due to the increased binding of the variant to DNA in the presence of L-cysteine, which enhances transcriptional activation. Overall, our biosensor-based strategy provides a promising approach for optimizing biosynthetic pathways at multiple levels. The successful implementation of this strategy demonstrates its potential for screening improved recombinant strains.

Design of a recombinant asparaginyl ligase for site-specific modification using efficient recognition and nucleophile motifs

Asparaginyl ligases have been extensively utilized as valuable tools for site-specific bioconjugation or surface-modification. However, the application is hindered by the laborious and poorly reproducible preparation processes, unstable activity and ambiguous substrate requirements. To address these limitations, this study employed a structure-based rational approach to obtain a high-yield and high-activity protein ligase called OaAEP1-C247A-aa55-351. It was observed that OaAEP1-C247A-aa55-351 exhibits appreciable catalytic activities across a wide pH range, and the addition of the Fe3+ metal ion effectively enhances the catalytic power. Importantly, this study provides insight into the recognition and nucleophile peptide profiles of OaAEP1-C247A-aa55-351. The ligase demonstrates a higher recognition ability for the “Asn-Ala-Leu” motif and an N-terminus “Arg-Leu” as nucleophiles, which significantly increases the reaction yield. Consequently, the catalytic activity of OaAEP1-C247A-aa55-351 with highly efficient recognition and nucleophile motif, “Asn-Ala-Leu” and “Arg-Leu” under the buffer containing Fe3+ is 70-fold and 2-fold higher than previously reported OaAEP1-C247A and the most efficient butelase-1, respectively. Thus, the designed OaAEP1-C247A-aa55-351, with its highly efficient recognition and alternative nucleophile options, holds promising potential for applications in protein engineering, chemo-enzymatic modification, and the development of drugs.

Protein Engineering with Lightweight Graph Denoising Neural Networks.

Protein engineering faces challenges in finding optimal mutants from a massive pool of candidate mutants. In this study, we introduce a deep-learning-based data-efficient fitness prediction tool to steer protein engineering. Our methodology establishes a lightweight graph neural network scheme for protein structures, which efficiently analyzes the microenvironment of amino acids in wild-type proteins and reconstructs the distribution of the amino acid sequences that are more likely to pass natural selection. This distribution serves as a general guidance for scoring proteins toward arbitrary properties on any order of mutations. Our proposed solution undergoes extensive wet-lab experimental validation spanning diverse physicochemical properties of various proteins, including fluorescence intensity, antigen-antibody affinity, thermostability, and DNA cleavage activity. More than 40% of ProtLGN-designed single-site mutants outperform their wild-type counterparts across all studied proteins and targeted properties. More importantly, our model can bypass the negative epistatic effect to combine single mutation sites and form deep mutants with up to seven mutation sites in a single round, whose physicochemical properties are significantly improved. This observation provides compelling evidence of the structure-based model's potential to guide deep mutations in protein engineering. Overall, our approach emerges as a versatile tool for protein engineering, benefiting both the computational and bioengineering communities.