Advances, opportunities, and challenges in methods for interrogating the structure activity relationships of natural products.

Time span in literature: 1985-early 2024Natural products play a key role in drug discovery, both as a direct source of drugs and as a starting point for the development of synthetic compounds. Most natural products are not suitable to be used as drugs without further modification due to insufficient activity or poor pharmacokinetic properties. Choosing what modifications to make requires an understanding of the compound's structure-activity relationships. Use of structure-activity relationships is commonplace and essential in medicinal chemistry campaigns applied to human-designed synthetic compounds. Structure-activity relationships have also been used to improve the properties of natural products, but several challenges still limit these efforts. Here, we review methods for studying the structure-activity relationships of natural products and their limitations. Specifically, we will discuss how synthesis, including total synthesis, late-stage derivatization, chemoenzymatic synthetic pathways, and engineering and genome mining of biosynthetic pathways can be used to produce natural product analogs and discuss the challenges of each of these approaches. Finally, we will discuss computational methods including machine learning methods for analyzing the relationship between biosynthetic genes and product activity, computer aided drug design techniques, and interpretable artificial intelligence approaches towards elucidating structure-activity relationships from models trained to predict bioactivity from chemical structure. Our focus will be on these latter topics as their applications for natural products have not been extensively reviewed. We suggest that these methods are all complementary to each other, and that only collaborative efforts using a combination of these techniques will result in a full understanding of the structure-activity relationships of natural products.

Diversity and taxonomic distribution of bacterial biosynthetic gene clusters predicted to produce compounds with therapeutically relevant bioactivities.

Bacteria have long been a source of natural products with diverse bioactivities that have been developed into therapeutics to treat human disease. Historically, researchers have focused on a few taxa of bacteria, mainly Streptomyces and other actinomycetes. This strategy was initially highly successful and resulted in the golden era of antibiotic discovery. The golden era ended when the most common antibiotics from Streptomyces had been discovered. Rediscovery of known compounds has plagued natural product discovery ever since. Recently, there has been increasing interest in identifying other taxa that produce bioactive natural products. Several bioinformatics studies have identified promising taxa with high biosynthetic capacity. However, these studies do not address the question of whether any of the products produced by these taxa are likely to have activities that will make them useful as human therapeutics. We address this gap by applying a recently developed machine learning tool that predicts natural product activity from biosynthetic gene cluster (BGC) sequences to determine which taxa are likely to produce compounds that are not only novel but also bioactive. This machine learning tool is trained on a dataset of BGC-natural product activity pairs and relies on counts of different protein domains and resistance genes in the BGC to make its predictions. We find that rare and understudied actinomycetes are the most promising sources for novel active compounds. There are also several taxa outside of actinomycetes that are likely to produce novel active compounds. We also find that most strains of Streptomyces likely produce both characterized and uncharacterized bioactive natural products. The results of this study provide guidelines to increase the efficiency of future bioprospecting efforts. ONE-SENTENCE SUMMARY: This paper combines several bioinformatics workflows to identify which genera of bacteria are most likely to produce novel natural products with useful bioactivities such as antibacterial, antitumor, or antifungal activity.

RNA sectors and allosteric function within the ribosome.

The ribosome translates the genetic code into proteins in all domains of life. Its size and complexity demand long-range interactions that regulate ribosome function. These interactions are largely unknown. Here, we apply a global coevolution method, statistical coupling analysis (SCA), to identify coevolving residue networks (sectors) within the 23S ribosomal RNA (rRNA) of the large ribosomal subunit. As in proteins, SCA reveals a hierarchical organization of evolutionary constraints with near-independent groups of nucleotides forming physically contiguous networks within the three-dimensional structure. Using a quantitative, continuous-culture-with-deep-sequencing assay, we confirm that the top two SCA-predicted sectors contribute to ribosome function. These sectors map to distinct ribosome activities, and their origins trace to phylogenetic divergences across all domains of life. These findings provide a foundation to map ribosome allostery, explore ribosome biogenesis, and engineer ribosomes for new functions. Despite differences in chemical structure, protein and RNA enzymes appear to share a common internal logic of interaction and assembly.

A Machine Learning Bioinformatics Method to Predict Biological Activity from Biosynthetic Gene Clusters.

Research in natural products, the genetically encoded small molecules produced by organisms in an idiosyncratic fashion, deals with molecular structure, biosynthesis, and biological activity. Bioinformatics analyses of microbial genomes can successfully reveal the genetic instructions, biosynthetic gene clusters, that produce many natural products. Genes to molecule predictions made on biosynthetic gene clusters have revealed many important new structures. There is no comparable method for genes to biological activity predictions. To address this missing pathway, we developed a machine learning bioinformatics method for predicting a natural product's antibiotic activity directly from the sequence of its biosynthetic gene cluster. We trained commonly used machine learning classifiers to predict antibacterial or antifungal activity based on features of known natural product biosynthetic gene clusters. We have identified classifiers that can attain accuracies as high as 80% and that have enabled the identification of biosynthetic enzymes and their corresponding molecular features that are associated with antibiotic activity.

Rotamer-Restricted Fluorogenicity of the Bis-Arsenical ReAsH.

Fluorogenic dyes such as FlAsH and ReAsH are used widely to localize, monitor, and characterize proteins and their assemblies in live cells. These bis-arsenical dyes can become fluorescent when bound to a protein containing four proximal Cys thiols-a tetracysteine (Cys4) motif. Yet the mechanism by which bis-arsenicals become fluorescent upon binding a Cys4 motif is unknown, and this nescience limits more widespread application of this tool. Here we probe the origins of ReAsH fluorogenicity using both computation and experiment. Our results support a model in which ReAsH fluorescence depends on the relative orientation of the aryl chromophore and the appended arsenic chelate: the fluorescence is rotamer-restricted. Our results do not support a model in which fluorogenicity arises from the relief of ring strain. The calculations identify those As-aryl rotamers that support fluorescence and those that do not and correlate well with prior experiments. The rotamer-restricted model we propose is supported further by biophysical studies: the excited-state fluorescence lifetime of a complex between ReAsH and a protein bearing a high-affinity Cys4 motif is longer than that of ReAsH-EDT2, and the fluorescence intensity of ReAsH-EDT2 increases in solvents of increasing viscosity. By providing a higher resolution view of the structural basis for fluorogenicity, these results provide a clear strategy for the design of more selective bis-arsenicals and better-optimized protein targets, with a concomitant improvement in the ability to characterize previously invisible protein conformational changes and assemblies in live cells.

In Vivo Biosynthesis of a ß-Amino Acid-Containing Protein.

It has recently been reported that ribosomes from erythromycin-resistant Escherichia coli strains, when isolated in S30 extracts and incubated with chemically mis-acylated tRNA, can incorporate certain ß-amino acids into full length DHFR in vitro. Here we report that wild-type E. coli EF-Tu and phenylalanyl-tRNA synthetase collaborate with these mutant ribosomes and others to incorporate ß(3)-Phe analogs into full length DHFR in vivo. E. coli harboring the most active mutant ribosomes are robust, with a doubling time only 14% longer than wild-type. These results reveal the unexpected tolerance of E. coli and its translation machinery to the ß(3)-amino acid backbone and should embolden in vivo selections for orthogonal translational machinery components that incorporate diverse ß-amino acids into proteins and peptides. E. coli harboring mutant ribosomes may possess the capacity to incorporate many non-natural, non-α-amino acids into proteins and other sequence-programmed polymeric materials.

Predicting fungal secondary metabolite activity from biosynthetic gene cluster data using machine learning.

Fungal secondary metabolites (SMs) contribute to the diversity of fungal ecological communities, niches, and lifestyles. Many fungal SMs have one or more medically and industrially important activities (e.g., antifungal, antibacterial, and antitumor). The genes necessary for fungal SM biosynthesis are typically located right next to each other in the genome and are known as biosynthetic gene clusters (BGCs). However, whether fungal SM bioactivity can be predicted from specific attributes of genes in BGCs remains an open question. We adapted machine learning models that predicted SM bioactivity from bacterial BGC data with accuracies as high as 80% to fungal BGC data. We trained our models to predict the antibacterial, antifungal, and cytotoxic/antitumor bioactivity of fungal SMs on two data sets: (i) fungal BGCs (data set comprised of 314 BGCs) and (ii) fungal (314 BGCs) and bacterial BGCs (1,003 BGCs). We found that models trained on fungal BGCs had balanced accuracies between 51% and 68%, whereas training on bacterial and fungal BGCs had balanced accuracies between 56% and 68%. The low prediction accuracy of fungal SM bioactivities likely stems from the small size of the data set; this lack of data, coupled with our finding that including bacterial BGC data in the training data did not substantially change accuracies currently limits the application of machine learning approaches to fungal SM studies. With >15,000 characterized fungal SMs, millions of putative BGCs in fungal genomes, and increased demand for novel drugs, efforts that systematically link fungal SM bioactivity to BGCs are urgently needed.IMPORTANCEFungi are key sources of natural products and iconic drugs, including penicillin and statins. DNA sequencing has revealed that there are likely millions of biosynthetic pathways in fungal genomes, but the chemical structures and bioactivities of >99% of natural products produced by these pathways remain unknown. We used artificial intelligence to predict the bioactivities of diverse fungal biosynthetic pathways. We found that the accuracies of our predictions were generally low, between 51% and 68%, likely because the natural products and bioactivities of only very few fungal pathways are known. With >15,000 characterized fungal natural products, millions of putative biosynthetic pathways present in fungal genomes, and increased demand for novel drugs, our study suggests that there is an urgent need for efforts that systematically identify fungal biosynthetic pathways, their natural products, and their bioactivities.

Statistical Coupling Analysis Predicts Correlated Motions in Dihydrofolate Reductase.

The role of dynamics in enzymatic function is a highly debated topic. Dihydrofolate reductase (DHFR), due to its universality and the depth with which it has been studied, is a model system in this debate. Myriad previous works have identified networks of residues in positions near to and remote from the active site that are involved in dynamics and others that are important for catalysis. For example, specific mutations on the Met20 loop in E. coli DHFR (N23PP/S148A) are known to disrupt millisecond-timescale motions and reduce catalytic activity. However, how and if networks of dynamically coupled residues influence the evolution of DHFR is still an unanswered question. In this study, we first identify, by statistical coupling analysis and molecular dynamic simulations, a network of coevolving residues, which possess increased correlated motions. We then go on to show that allosteric communication in this network is selectively knocked down in N23PP/S148A mutant E. coli DHFR. Finally, we identify two sites in the human DHFR sector which may accommodate the Met20 loop double proline mutation while preserving dynamics. These findings strongly implicate protein dynamics as a driving force for evolution.

Evolution-guided engineering of trans-acyltransferase polyketide synthases.

Bacterial multimodular polyketide synthases (PKSs) are giant enzymes that generate a wide range of therapeutically important but synthetically challenging natural products. Diversification of polyketide structures can be achieved by engineering these enzymes. However, notwithstanding successes made with textbook cis-acyltransferase (cis-AT) PKSs, tailoring such large assembly lines remains challenging. Unlike textbook PKSs, trans-AT PKSs feature an extraordinary diversity of PKS modules and commonly evolve to form hybrid PKSs. In this study, we analyzed amino acid coevolution to identify a common module site that yields functional PKSs. We used this site to insert and delete diverse PKS parts and create 22 engineered trans-AT PKSs from various pathways and in two bacterial producers. The high success rates of our engineering approach highlight the broader applicability to generate complex designer polyketides.

Effective molarity redux: Proximity as a guiding force in chemistry and biology.

The cell interior is a complex and demanding environment. An incredible variety of molecules jockey to identify the correct position-the specific interactions that promote biology that are hidden among countless unproductive options. Ensuring that the business of the cell is successful requires sophisticated mechanisms to impose temporal and spatial specificity-both on transient interactions and their eventual outcomes. Two strategies employed to regulate macromolecular interactions in a cellular context are co-localization and compartmentalization. Macromolecular interactions can be promoted and specified by localizing the partners within the same subcellular compartment, or by holding them in proximity through covalent or non-covalent interactions with proteins, lipids, or DNA- themes that are familiar to any biologist. The net result of these strategies is an increase in effective molarity: the local concentration of a reactive molecule near its reaction partners. We will focus on this general mechanism, employed by Nature and adapted in the lab, which allows delicate control in complex environments: the power of proximity to accelerate, guide, or otherwise influence the reactivity of signaling proteins and the information that they encode.

Predicting fungal secondary metabolite activity from biosynthetic gene cluster data using machine learning.

Fungal secondary metabolites (SMs) play a significant role in the diversity of ecological communities, niches, and lifestyles in the fungal kingdom. Many fungal SMs have medically and industrially important properties including antifungal, antibacterial, and antitumor activity, and a single metabolite can display multiple types of bioactivities. The genes necessary for fungal SM biosynthesis are typically found in a single genomic region forming biosynthetic gene clusters (BGCs). However, whether fungal SM bioactivity can be predicted from specific attributes of genes in BGCs remains an open question. We adapted previously used machine learning models for predicting SM bioactivity from bacterial BGC data to fungal BGC data. We trained our models to predict antibacterial, antifungal, and cytotoxic/antitumor bioactivity on two datasets: 1) fungal BGCs (dataset comprised of 314 BGCs), and 2) fungal (314 BGCs) and bacterial BGCs (1,003 BGCs); the second dataset was our control since a previous study using just the bacterial BGC data yielded prediction accuracies as high as 80%. We found that the models trained only on fungal BGCs had balanced accuracies between 51-68%, whereas training on bacterial and fungal BGCs yielded balanced accuracies between 61-74%. The lower accuracy of the predictions from fungal data likely stems from the small number of BGCs and SMs with known bioactivity; this lack of data currently limits the application of machine learning approaches in studying fungal secondary metabolism. However, our data also suggest that machine learning approaches trained on bacterial and fungal data can predict SM bioactivity with good accuracy. With more than 15,000 characterized fungal SMs, millions of putative BGCs present in fungal genomes, and increased demand for novel drugs, efforts that systematically link fungal SM bioactivity to BGCs are urgently needed.

Inonotus obliquus upregulates muscle regeneration and augments function through muscle oxidative metabolism.

Skeletal muscle wasting related to aging or pathological conditions is critically associated with the increased incidence and prevalence of secondary diseases including cardiovascular diseases, metabolic syndromes, and chronic inflammations. Much effort is made to develop agents to enhance muscle metabolism and function. Inonotus obliquus (I. obliquus; IO) is a mushroom popularly called chaga and has been widely employed as a folk medicine for inflammation, cardiovascular diseases, diabetes, and cancer in Eastern Europe and Asia. However, its effect on muscle health has not been explored. Here, we aimed to investigate the beneficial effect of IO extract in muscle regeneration and metabolism. The treatment of IO in C2C12 myoblasts led to increased myogenic differentiation and alleviation of dexamethasone-induced myotube atrophy. Network pharmacological analysis using the identified specific chemical constituents of IO extracts predicted protein kinase B (AKT)-dependent mechanisms to promote myogenesis and muscle regeneration. Consistently, IO treatment resulted in the activation of AKT, which suppressed muscle-specific ubiquitin E3 ligases induced by dexamethasone. IO treatment in mice improved the regeneration of cardiotoxin-injured muscles accompanied by elevated proliferation and differentiation of muscle stem cells. Furthermore, it elevated the mitochondrial content and muscle oxidative metabolism accompanied by the induction of peroxisome proliferator-activated receptor γ coactivator α (PGC-1α). Our current data suggest that IO is a promising natural agent in enhancing muscle regenerative capacity and oxidative metabolism thereby preventing muscle wasting.

Artificial intelligence for natural product drug discovery.

Developments in computational omics technologies have provided new means to access the hidden diversity of natural products, unearthing new potential for drug discovery. In parallel, artificial intelligence approaches such as machine learning have led to exciting developments in the computational drug design field, facilitating biological activity prediction and de novo drug design for molecular targets of interest. Here, we describe current and future synergies between these developments to effectively identify drug candidates from the plethora of molecules produced by nature. We also discuss how to address key challenges in realizing the potential of these synergies, such as the need for high-quality datasets to train deep learning algorithms and appropriate strategies for algorithm validation.

Total synthesis reveals atypical atropisomerism in a small-molecule natural product, tryptorubin A.

Molecular shape defines function in both biological and material settings, and chemists have developed an ever-increasing vernacular to describe these shapes. Noncanonical atropisomers-shape-defined molecules that are formally topologically trivial but are interconvertible only by complex, nonphysical multibond torsions-form a unique subset of atropisomers that differ from both canonical atropisomers (e.g., binaphthyls) and topoisomers (i.e., molecules that have identical connectivity but nonidentical molecular graphs). Small molecules, in contrast to biomacromolecules, are not expected to exhibit such ambiguous shapes. Using total synthesis, we found that the peptidic alkaloid tryptorubin A can be one of two noncanonical atropisomers. We then devised a synthetic strategy that drives the atropospecific synthesis of a noncanonical atrop-defined small molecule.

Mechanism of Allosteric Coupling into and through the Plasma Membrane by EGFR.

Epidermal growth factor receptor (EGFR) interacts through its extracellular domain with seven different growth factors. These factors induce different structures within the cytoplasmic juxtamembrane (JM) segment of the dimeric receptor and propagate different growth factor-dependent signals to the cell interior. How this process occurs is unknown. Here we apply diverse experimental and computational tools to show that growth factor identity is encoded by the EGFR transmembrane (TM) helix into discrete helix dimer populations that differ in both cross-location and cross-angle. Helix dimers with smaller cross-angles at multiple cross locations are decoded to induce an EGF-type coiled coil in the adjacent JM, whereas helix dimers with larger cross-angles at fewer cross locations induce the TGF-α-type coiled coil. We propose an updated model for how conformational coupling across multiple EGFR domains results in growth factor-specific information transfer, and demonstrate that this model applies to both EGFR and the related receptor ErbB2.

Improved Carbohydrate Recognition in Water with an Electrostatically Enhanced ß-Peptide Bundle.

The selective recruitment of oligosaccharides, or even simple sugars, in water solvent is an unsolved molecular recognition problem. Structure-guided, electrostatic redesign led to a significant increase in the affinity of a ß-peptide "borono-bundle" for simple sugars in neutral aqueous solution. The affinity for fructose (663 M(-1)) in water should allow its recruitment to the bundle surface for selective catalysis, and future work will focus in this direction.

A ß-boronopeptide bundle of known structure as a vehicle for polyol recognition.

Despite significant progress in the design of receptors and sensors for simple polyols and monosaccharides, few synthetic receptors discriminate among multiple saccharide units simultaneously, especially under physiological conditions. Described here is the three-dimensional structure of a supramolecular complex-a ß-peptide bundle-designed for the potential to interact simultaneously with as many as eight discrete monosaccharide units. The preliminary evaluation of this construct as a vehicle for polyol binding is also presented.