TL;DR
This paper tackles the vast untapped potential of Actinomycetes for discovering new natural products by generating and sharing a massive dataset of untargeted metabolomics data from nearly a thousand microbial strains. By combining this data with advanced computational tools, they demonstrate that a significant portion of microbial biosynthetic potential remains uncharacterized, highlighting the critical need for such public resources to bridge the gap between genomic information and actual metabolite production.
What Problem Does This Paper Address?
Actinomycetes, particularly the Streptomyces genus, are renowned "natural product factories," responsible for a huge number of bioactive compounds used in medicine and biotechnology. While genomic sequencing has revealed thousands of "biosynthetic gene clusters" (BGCs) that could produce these molecules, only a tiny fraction (estimated at 3%) of this potential has been experimentally confirmed. This disconnect between genetic blueprints and actual chemical output is a major bottleneck. Existing databases of natural products often lack crucial spectral data (like MS/MS spectra) needed to identify molecules produced by these organisms. Therefore, there's a critical need for comprehensive, publicly available metabolomic datasets that can be used to link BGCs to actual produced molecules and to discover novel compounds.
Methods
The researchers employed an untargeted liquid chromatography-tandem mass spectrometry (LC-MS/MS) approach to analyze microbial extracts. This method separates molecules based on their chemical properties and then fragments them to generate unique spectral fingerprints (MS/MS spectra).
Key methodological aspects include:
- OSMAC Strategy (One Strain Many Compounds): To maximize the diversity of detected molecules, they cultured 440 Actinomycetes strains under five different media conditions. This approach aims to trigger the production of different sets of compounds by varying the environmental cues.
- Molecular Networking: This computational technique visualizes the relationships between MS/MS spectra. Spectra with similar fragmentation patterns are clustered together, forming "molecular families." This allows for the identification of related compounds and the detection of novel analogues.
- Spectral Matching to Reference Libraries: The generated MS/MS spectra were compared against existing public spectral libraries (e.g., from GNPS - Global Natural Products Social Molecular Networking) to identify known molecules.
- Network Annotation Propagation (NAP): This advanced method uses existing annotations and spectral similarity to predict the identity of unannotated spectra, effectively "propagating" knowledge from known to unknown molecules. They also utilized a custom database of reported Streptomyces products.
- MS-Cluster: Used for merging similar spectra and creating consensus spectra, reducing redundancy.
- ModiFinder: A tool used to predict the location of chemical modifications in unknown molecules by comparing their spectra to known compounds, as demonstrated with the detection of edapochelins.
Datasets
The primary dataset introduced in this paper consists of untargeted LC-MS/MS data from 948 bacterial strains, predominantly Actinomycetes. This generated nearly two million MS/MS spectra.
- Newly Introduced Benchmark: The paper provides a large, curated metabolomics dataset as a public resource. This differs from existing datasets by its scale and the inclusion of diverse culture conditions, aiming to capture a broader spectrum of microbial metabolites.
- Publicly Available: The data has been deposited in the Mass Spectrometry Interactive Virtual Environment (MassIVE) repository and is accessible via GNPS.
Key Insights and Ideas
- Vast Undiscovered Potential: Despite Actinomycetes being heavily studied, the data reveals that a significant majority of detected MS/MS spectra (and thus potential molecules) remain unannotated, underscoring the immense untapped natural product reservoir.
- Bridging the Genomics-Metabolomics Gap: The study strongly emphasizes that genomic information (BGCs) alone is insufficient; metabolomic data is crucial for confirming the actual production of molecules and understanding biosynthetic potential.
- Power of Molecular Networking: Molecular networking proved invaluable for visualizing chemical space, grouping related compounds, and identifying potential novel molecules and analogues, even when direct annotation was challenging.
- Culture Conditions Matter: The OSMAC strategy highlighted that different media can elicit the production of distinct sets of molecules, emphasizing the importance of diverse cultivation approaches for comprehensive discovery.
- Value of Public Data Sharing: The paper champions the creation and sharing of large, public metabolomics datasets as a fundamental resource for the scientific community, enabling advanced computational analyses and accelerating discovery.
Key Results
- Dataset Size: Nearly two million MS/MS spectra were generated from 948 microbial strains.
- Annotation Rate: An overall annotation rate of 13.3% was achieved, corresponding to 2352 annotated molecules.
- Unannotated Molecules: The majority of detected MS/MS spectra (over 86%) remained unannotated, indicating a vast number of potentially novel compounds.
- Molecular Families: Molecular networking revealed over 8000 unannotated molecular families, suggesting a rich chemical diversity yet to be explored.
- Detection of Novel Compounds: The study successfully demonstrated the detection of recently discovered non-ribosomal peptides, such as edapochelins, by leveraging their metabolomics data and computational tools.
Limitations and Future Work
- Annotation Limitations: The authors acknowledge that annotation rates are still limited by the availability of comprehensive spectral libraries. Not every annotation is guaranteed to be correct, with an estimated ~1% false discovery rate.
- Redundancy in Spectra: Many unannotated spectra might represent different ion forms (adducts, in-source fragments, multimers) of the same molecule, contributing to redundancy.
- Culture Condition Specificity: The detected molecules are inherently limited to the specific culture conditions, extraction methods, and LC-MS/MS settings used in the study.
- Future Work: The authors envision this dataset serving as a valuable resource for:
- Investigating the ecological roles of thousands of detected molecules.
- Linking detected molecules to their corresponding BGCs, especially for "silent" or "cryptic" BGCs.
- Dereplication (identifying known compounds) and identifying unknown producers of known bioactive molecules.
- Benchmarking and improving machine learning algorithms for metabolite identification and gene-cluster prediction.
- Facilitating the discovery of new antibiotics and other bioactive compounds by expanding the available reference spectra.
Summary updated