Accurate Broadcasting of Substrate Fitness for Lactazole Biosynthetic Pathway from Reactivity-Profiling mRNA Display

Substrate prediction for RiPP biosynthetic enzymes via masked language modeling and transfer learning

Ribosomally synthesized and post-translationally modified peptide (RiPP) biosynthetic enzymes often exhibit promiscuous substrate preferences that cannot be reduced to simple rules. Large language models are promising tools for predicting the specificity of RiPP biosynthetic enzymes. However, state-of-the-art protein language models are trained on relatively few peptide sequences. A previous study comprehensively profiled the peptide substrate preferences of LazBF (a two-component serine dehydratase) and LazDEF (a three-component azole synthetase) from the lactazole biosynthetic pathway. We demonstrated that masked language modeling of LazBF substrate preferences produced language model embeddings that improved downstream prediction of both LazBF and LazDEF substrates. Similarly, masked language modeling of LazDEF substrate preferences produced embeddings that improved prediction of both LazBF and LazDEF substrates. Our results suggest that the models learned functional forms that are transferable between distinct enzymatic transformations that act within the same biosynthetic pathway. We found that a single high-quality data set of substrates and non-substrates for a RiPP biosynthetic enzyme improved substrate prediction for distinct enzymes in data-scarce scenarios. We then fine-tuned models on each data set and showed that the fine-tuned models provided interpretable insight that we anticipate will facilitate the design of substrate libraries that are compatible with desired RiPP biosynthetic pathways.

Illuminating Substrate Preferences of Promiscuous F420H2-Dependent Dehydroamino Acid Reductases with 4-Track mRNA Display

Stereoselective reduction of dehydroamino acids is a common biosynthetic strategy to introduce d-amino acids into peptidic natural products. The reduction, often observed during the biosynthesis of lanthipeptides, is performed by dedicated dehydroamino acid reductases (dhAARs). Enzymes from the three known dhAAR families utilize nicotinamide, flavin, or F420H2 coenzymes as hydride donors, and little is known about the catalysis performed by the latter family proteins. Here, we perform a bioinformatics-guided identification and large-scale in vitro characterization of five F420H2-dependent dhAARs. We construct an mRNA display-based pipeline for ultrahigh throughput substrate specificity profiling of the enzymes. The pipeline relies on a 4-track selection strategy to deliver large quantities of clean data, which were leveraged to build accurate substrate fitness models. Our results identify a remarkably promiscuous enzyme, referred to as MaeJC, that is capable of installing d-Ala residues into arbitrary substrates with minimal recognition requirements. We integrate MaeJC into a thiopeptide biosynthetic pathway to produce d-amino acids-containing thiopeptides, demonstrating the utility of MaeJC for the programmable installation of d-amino acids in ribosomal peptides.

Assessing substrate scope of the cyclodehydratase LynD by mRNA display-enabled machine learning models

Many of the biosynthetic pathways for ribosomal synthesized and post-translationally modified peptide (RiPP) natural products make use of multi-domain enzymes with separate recruitment and catalysis domains that separately bind and modify peptide substrates. This "division of labor" allows RiPP enzymes to use relatively open and promiscuous active sites to perform chemistry at multiple residues within a peptide substrate seemingly regardless of the surrounding context. Defining, measuring, and predicting the seemingly broad substrate promiscuity of RiPPs necessitates high throughput assays, capable of assessing activity against very large libraries of peptides. Using mRNA display, a high throughput peptide display technology, we examine the substrate promiscuity of the RiPP cyclodehydratase, LynD. The vast substrate profiling that can be done with mRNA display enables the construction of deep learning models for accurate prediction of substrate processing by LynD. These models further inform on epistatic interactions involved in enzymatic processing. This work will facilitate the further elucidation of other RiPP enzymes and enable their use in the modification of mRNA display libraries for selection of modified peptide-based inhibitors and therapeutics.

Cell-Free Systems: Ideal Platforms for Accelerating the Discovery and Production of Peptide-Based Antibiotics

Peptide-based antibiotics (PBAs), including antimicrobial peptides (AMPs) and their synthetic mimics, have received significant interest due to their diverse and unique bioactivities. The integration of high-throughput sequencing and bioinformatics tools has dramatically enhanced the discovery of enzymes, allowing researchers to identify specific genes and metabolic pathways responsible for producing novel PBAs more precisely. Cell-free systems (CFSs) that allow precise control over transcription and translation in vitro are being adapted, which accelerate the identification, characterization, selection, and production of novel PBAs. Furthermore, these platforms offer an ideal solution for overcoming the limitations of small-molecule antibiotics, which often lack efficacy against a broad spectrum of pathogens and contribute to the development of antibiotic resistance. In this review, we highlight recent examples of how CFSs streamline these processes while expanding our ability to access new antimicrobial agents that are effective against antibiotic-resistant infections.

Engineering tRNAs for the Ribosomal Translation of Non-proteinogenic Monomers

Ribosome-dependent protein biosynthesis is an essential cellular process mediated by transfer RNAs (tRNAs). Generally, ribosomally synthesized proteins are limited to the 22 proteinogenic amino acids (pAAs: 20 l-α-amino acids present in the standard genetic code, selenocysteine, and pyrrolysine). However, engineering tRNAs for the ribosomal incorporation of non-proteinogenic monomers (npMs) as building blocks has led to the creation of unique polypeptides with broad applications in cellular biology, material science, spectroscopy, and pharmaceuticals. Ribosomal polymerization of these engineered polypeptides presents a variety of challenges for biochemists, as translation efficiency and fidelity is often insufficient when employing npMs. In this Review, we will focus on the methodologies for engineering tRNAs to overcome these issues and explore recent advances both in vitro and in vivo. These efforts include increasing orthogonality, recruiting essential translation factors, and creation of expanded genetic codes. After our review on the biochemical optimizations of tRNAs, we provide examples of their use in genetic code manipulation, with a focus on the in vitro discovery of bioactive macrocyclic peptides containing npMs. Finally, an analysis of the current state of tRNA engineering is presented, along with existing challenges and future perspectives for the field.

A Compact Reprogrammed Genetic Code for De Novo Discovery of Proteolytically Stable Thiopeptides

Thiopeptides make up a group of structurally complex peptidic natural products holding promise in bioengineering applications. The previously established thiopeptide/mRNA display platform enables de novo discovery of natural product-like thiopeptides with designed bioactivities. However, in contrast to natural thiopeptides, the discovered structures are composed predominantly of proteinogenic amino acids, which results in low metabolic stability in many cases. Here, we redevelop the platform and demonstrate that the utilization of compact reprogrammed genetic codes in mRNA display libraries can lead to the discovery of thiopeptides predominantly composed of nonproteinogenic structural elements. We demonstrate the feasibility of our designs by conducting affinity selections against Traf2- and NCK-interacting kinase (TNIK). The experiment identified a series of thiopeptides with high affinity to the target protein (the best KD = 2.1 nM) and kinase inhibitory activity (the best IC50 = 0.15 μM). The discovered compounds, which bore as many as 15 nonproteinogenic amino acids in an 18-residue macrocycle, demonstrated high metabolic stability in human serum with a half-life of up to 99 h. An X-ray cocrystal structure of TNIK in complex with a discovered thiopeptide revealed how nonproteinogenic building blocks facilitate the target engagement and orchestrate the folding of the thiopeptide into a noncanonical conformation. Altogether, the established platform takes a step toward the discovery of thiopeptides with high metabolic stability for early drug discovery applications.

Cheminformatics-Guided Cell-Free Exploration of Peptide Natural Products

There have been significant advances in the flexibility and power of in vitro cell-free translation systems. The increasing ability to incorporate noncanonical amino acids and complement translation with recombinant enzymes has enabled cell-free production of peptide-based natural products (NPs) and NP-like molecules. We anticipate that many more such compounds and analogs might be accessed in this way. To assess the peptide NP space that is directly accessible to current cell-free technologies, we developed a peptide parsing algorithm that breaks down peptide NPs into building blocks based on ribosomal translation logic. Using the resultant data set, we broadly analyze the biophysical properties of these privileged compounds and perform a retrobiosynthetic analysis to predict which peptide NPs could be directly synthesized in augmented cell-free translation reactions. We then tested these predictions by preparing a library of highly modified peptide NPs. Two macrocyclases, PatG and PCY1, were used to effect the head-to-tail macrocyclization of candidate NPs. This retrobiosynthetic analysis identified a collection of high-priority building blocks that are enriched throughout peptide NPs, yet they had not previously been tested in cell-free translation. To expand the cell-free toolbox into this space, we established, optimized, and characterized the flexizyme-enabled ribosomal incorporation of piperazic acids. Overall, these results demonstrate the feasibility of cell-free translation for peptide NP total synthesis while expanding the limits of the technology. This work provides a novel computational tool for exploration of peptide NP chemical space, that could be expanded in the future to allow design of ribosomal biosynthetic pathways for NPs and NP-like molecules.

Deep Learning-Driven Library Design for the De Novo Discovery of Bioactive Thiopeptides

Broad substrate tolerance of ribosomally synthesized and post-translationally modified peptide (RiPP) biosynthetic enzymes has allowed numerous strategies for RiPP engineering. However, despite relaxed specificities, exact substrate preferences of RiPP enzymes are often difficult to pinpoint. Thus, when designing combinatorial libraries of RiPP precursors, balancing the compound diversity with the substrate fitness can be challenging. Here, we employed a deep learning model to streamline the design of mRNA display libraries. Using an in vitro reconstituted thiopeptide biosynthesis platform, we performed mRNA display-based profiling of substrate fitness for the biosynthetic pathway involving five enzymes to train an accurate deep learning model. We then utilized the model to design optimal mRNA libraries and demonstrated their utility in affinity selections against IRAK4 kinase and the TLR10 cell surface receptor. The selections led to the discovery of potent thiopeptide ligands against both target proteins (KD up to 1.3 nM for the best compound against IRAK4 and 300 nM for TLR10). The IRAK4-targeting compounds also inhibited the kinase at single-digit μM concentrations in vitro, exhibited efficient internalization into HEK293H cells, and suppressed NF-kB-mediated signaling in cells. Altogether, the developed approach streamlines the discovery of pseudonatural RiPPs with de novo designed biological activities and favorable pharmacological properties.

Tyrosinase-Catalyzed Peptide Macrocyclization for mRNA Display

mRNA display of macrocyclic peptides has proven itself to be a powerful technique to discover high-affinity ligands for a protein target. However, only a limited number of cyclization chemistries are known to be compatible with mRNA display. Tyrosinase is a copper-dependent oxidase that oxidizes tyrosine phenol to an electrophilic o-quinone, which is readily attacked by cysteine thiol. Here we show that peptides containing tyrosine and cysteine are rapidly cyclized upon tyrosinase treatment. Characterization of the cyclization reveals it to be widely applicable to multiple macrocycle sizes and scaffolds. We combine tyrosinase-mediated cyclization with mRNA display to discover new macrocyclic ligands targeting melanoma-associated antigen A4 (MAGE-A4). These macrocycles potently inhibit the MAGE-A4 binding axis with nanomolar IC50 values. Importantly, macrocyclic ligands show clear advantage over noncyclized analogues with ∼40-fold or greater decrease in IC50 values.

Enzymatic Macrolactamization of mRNA Display Libraries for Inhibitor Selection

mRNA display is a powerful, high-throughput technology for discovering novel, peptide ligands for protein targets. A number of methods have been used to expand the chemical diversity of mRNA display libraries beyond the 20 canonical amino acids, including genetic code reprogramming and biorthogonal chemistries. To date, however, there have been few reports using enzymes as biocompatible reagents for diversifying mRNA display libraries. Here, we report the evaluation and implementation of the common industrial enzyme, microbial transglutaminase (mTG), as a versatile biocatalyst for cyclization of mRNA display peptide libraries via lysine-to-glutamine isopeptide bonds. We establish two separate display-based assays to validate the compatibility of mTG with mRNA-linked peptide substrates. These assays indicate that mTG has a high degree of substrate tolerance and low single round bias. To demonstrate the potential benefits of mTG-mediated cyclization in ligand discovery, high diversity mTG-modified libraries were employed in two separate affinity selections: (1) one against the calcium and integrin binding protein, CIB1, and (2) the second against the immune checkpoint protein and emerging therapeutic target, B7-H3. Both selections resulted in the identification of potent, cyclic, low nanomolar binders, and subsequent structure-activity studies demonstrate the importance of the cyclization to the observed activity. Notably, cyclization in the CIB1 binder stabilizes an α-helical conformation, while the B7-H3 inhibitor employs two bridges, one mTG-derived lactam and a second disulfide to achieve its potency. Together, these results demonstrate potential benefits of enzyme-based biocatalysts in mRNA display ligand selections and establish a framework for employing mTG in mRNA display.

Emulating nonribosomal peptides with ribosomal biosynthetic strategies

Peptide natural products are important lead structures for human drugs and many nonribosomal peptides possess antibiotic activity. This makes them interesting targets for engineering approaches to generate peptide analogues with, for example, increased bioactivities. Nonribosomal peptides are produced by huge mega-enzyme complexes in an assembly-line like manner, and hence, these biosynthetic pathways are challenging to engineer. In the past decade, more and more structural features thought to be unique to nonribosomal peptides were found in ribosomally synthesised and posttranslationally modified peptides as well. These streamlined ribosomal pathways with modifying enzymes that are often promiscuous and with gene-encoded precursor proteins that can be modified easily, offer several advantages to produce designer peptides. This review aims to provide an overview of recent progress in this emerging research area by comparing structural features common to both nonribosomal and ribosomally synthesised and posttranslationally modified peptides in the first part and highlighting synthetic biology strategies for emulating nonribosomal peptides by ribosomal pathway engineering in the second part.

Accurate Models of Substrate Preferences of Post-Translational Modification Enzymes from a Combination of mRNA Display and Deep Learning

Promiscuous post-translational modification (PTM) enzymes often display nonobvious substrate preferences by acting on diverse yet well-defined sets of peptides and/or proteins. Understanding of substrate fitness landscapes for PTM enzymes is important in many areas of contemporary science, including natural product biosynthesis, molecular biology, and biotechnology. Here, we report an integrated platform for accurate profiling of substrate preferences for PTM enzymes. The platform features (i) a combination of mRNA display with next-generation sequencing as an ultrahigh throughput technique for data acquisition and (ii) deep learning for data analysis. The high accuracy (>0.99 in each of two studies) of the resulting deep learning models enables comprehensive analysis of enzymatic substrate preferences. The models can quantify fitness across sequence space, map modification sites, and identify important amino acids in the substrate. To benchmark the platform, we performed profiling of a Ser dehydratase (LazBF) and a Cys/Ser cyclodehydratase (LazDEF), two enzymes from the lactazole biosynthesis pathway. In both studies, our results point to complex enzymatic preferences, which, particularly for LazBF, cannot be reduced to a set of simple rules. The ability of the constructed models to dissect such complexity suggests that the developed platform can facilitate a wider study of PTM enzymes.

The RaPID Platform for the Discovery of Pseudo-Natural Macrocyclic Peptides

Although macrocyclic peptides bearing exotic building blocks have proven their utility as pharmaceuticals, the sources of macrocyclic peptide drugs have been largely limited to mimetics of native peptides or natural product peptides. However, the recent emergence of technologies for discovering de novo bioactive peptides has led to their reconceptualization as a promising therapeutic modality. For the construction and screening of libraries of such macrocyclic peptides, our group has devised a platform to conduct affinity-based selection of massive libraries (>10¹² unique sequences) of in vitro expressed macrocyclic peptides, which is referred to as the random nonstandard peptides integrated discovery (RaPID) system. The RaPID system integrates genetic code reprogramming using the FIT (flexible in vitro translation) system, which is largely facilitated by flexizymes (flexible tRNA-aminoacylating ribozymes), with mRNA display technology.We have demonstrated that the RaPID system enables rapid discovery of various de novo pseudo-natural peptide ligands for protein targets of interest. Many examples discussed in this Account prove that thioether-closed macrocyclic peptides (teMPs) obtained by the RaPID system generally exhibit remarkably high affinity and specificity, thereby potently inhibiting or activating a specific function(s) of the target. Moreover, such teMPs are used for a wide range of biochemical applications, for example, as crystallization chaperones for intractable transmembrane proteins and for in vivo recognition of specific cell types. Furthermore, recent studies demonstrate that some teMPs exhibit pharmacological activities in animal models and that even intracellular proteins can be inhibited by teMPs, illustrating the potential of this class of peptides as drug leads.Besides the ring-closing thioether linkage in the teMPs, genetic code reprogramming by the FIT system allows for incorporation of a variety of other exotic building blocks. For instance, diverse nonproteinogenic amino acids, hydroxy acids (ester linkage), amino carbothioic acid (thioamide linkage), and abiotic foldamer units have been successfully incorporated into ribosomally synthesized peptides. Despite such enormous successes in the conventional FIT system, multiple or consecutive incorporation of highly exotic amino acids, such as d- and β-amino acids, is yet challenging, and particularly the synthesis of peptides bearing non-carbonyl backbone structures remains a demanding task. To upgrade the RaPID system to the next generation, we have engaged in intensive manipulation of the FIT system to expand the structural diversity of peptides accessible by our in vitro biosynthesis strategy. Semilogical engineering of tRNA body sequences led to a new suppressor tRNA (tRNA^Pro1E2) capable of effectively recruiting translation factors, particularly EF-Tu and EF-P. The use of tRNA^Pro1E2 in the FIT system allows for not only single but also consecutive and multiple elongation of exotic amino acids, such as d-, β-, and γ-amino acids as well as aminobenzoic acids. Moreover, the integration of the FIT system with various chemical or enzymatic posttranslational modifications enables us to expand the range of accessible backbone structures to non-carbonyl moieties prominent in natural products and peptidomimetics. In such systems, FIT-expressed peptides undergo multistep backbone conversions in a one-pot manner to yield designer peptides composed of modified backbones such as azolines, azoles, and ring-closing pyridines. Our current research endeavors focus on applying such in vitro biosynthesis systems for the discovery of bioactive de novo pseudo-natural products.

Site-Specific Nonenzymatic Peptide S/O-Glutamylation Reveals the Extent of Substrate Promiscuity in Glutamate Elimination Domains

Formation of dehydroalanine and dehydrobutyrine residues via tRNA-dependent dehydration of serine and threonine is a key post-translational modification in the biosynthesis of lanthipeptide and thiopeptide RiPPs. The dehydration process involves two reactions, wherein the O-glutamyl Ser/Thr intermediate, accessed by a dedicated enzyme utilizing Glu-tRNA^Glu as the acyl donor, is recognized by the second enzyme, referred to as the glutamate elimination domain (ED), which catalyzes the eponymous reaction yielding a dehydroamino acid. Many details of ED catalysis remain unexplored because the scope of available substrates for testing is limited to those that the upstream enzymes can furnish. Here, we report two complementary strategies for direct, nonenzymatic access to diverse ED substrates. We establish that a thiol-thioester exchange reaction between a Cys-containing peptide and an ^α thioester of glutamic acid leads an S-glutamylated intermediate which can act as a substrate for EDs. Furthermore, we show that the native O-glutamylated substrates can be accessible from S-glutamylated peptides upon a site-specific S-to-O acyl transfer reaction. Combined with flexible in vitro translation utilized for rapid peptide production, these chemistries enabled us to dissect the substrate recognition requirements of three known EDs. Our results establish that EDs are uniquely promiscuous enzymes capable of acting on substrates with arbitrary amino acid sequences and performing retro-Michael reaction beyond the canonical glutamate elimination. To facilitate substrate recruitment, EDs apparently engage in nonspecific hydrophobic interactions with their substrates. Altogether, our results establish the substrate scope of EDs and provide clues to their catalysis.

Directing evolution of novel ligands by mRNA display

mRNA display is a powerful biological display platform for the directed evolution of proteins and peptides. mRNA display libraries covalently link the displayed peptide or protein (phenotype) with the encoding genetic information (genotype) through the biochemical activity of the small molecule puromycin. Selection for peptide/protein function is followed by amplification of the linked genetic material and generation of a library enriched in functional sequences. Iterative selection cycles are then performed until the desired level of function is achieved, at which time the identity of candidate peptides can be obtained by sequencing the genetic material. The purpose of this review is to discuss the development of mRNA display technology since its inception in 1997 and to comprehensively review its use in the selection of novel peptides and proteins. We begin with an overview of the biochemical mechanism of mRNA display and its variants with a particular focus on its advantages and disadvantages relative to other biological display technologies. We then discuss the importance of scaffold choice in mRNA display selections and review the results of selection experiments with biological (e.g., fibronectin) and linear peptide library architectures. We then explore recent progress in the development of "drug-like" peptides by mRNA display through the post-translational covalent macrocyclization and incorporation of non-proteogenic functionalities. We conclude with an examination of enabling technologies that increase the speed of selection experiments, enhance the information obtained in post-selection sequence analysis, and facilitate high-throughput characterization of lead compounds. We hope to provide the reader with a comprehensive view of current state and future trajectory of mRNA display and its broad utility as a peptide and protein design tool.