9:37 AM - 10:00 AM
[BCG03-03] Metabolite analysis and machine learning-assisted discrimination of archaea and bacteria using minimally invasive single-cell Raman microspectroscopy
★Invited Papers
Keywords:Raman spectroscopy, Single-cell analysis, Archaea, Machine learning
The vast diversity of microorganisms inhabiting various earth environments is now rapidly being unveiled by virtue of microbiome analysis using next-generation sequencing. To understand the types and functions of specific microorganisms among them, nondestructive single-cell analysis is required that enables us to gain access to potential nonculturable microorganisms (microbial dark matter). Here we report that Raman microspectroscopy can be used for minimally invasive metabolic analysis and species discrimination of bacteria and archaea at the single-cell level. Raman spectra are recorded by irradiating laser-trapped, label-free single cells with laser light and encode the "fingerprints" of biomolecules such as proteins, DNAs, lipids, sugars, and metabolites.
We explore the capability of single-cell Raman microspectroscopy to detect characteristic metabolites in methanogenic archaea Methanosarcina mazei and Methanopyrus kandleri. In both archaea, we observe the Raman spectroscopic features of corrinoids such as vitamin B12 analogues. These features allow for investigating cell-to-cell variation in the corrinoid concentration in M. kandleri and for imaging the spatial distributions of the corrinoid in relatively larger M. mazei cells, both of which could provide useful insight into increased microbial corrinoid production.
In addition to the detection of particular metabolites, the molecular fingerprints encoded in Raman spectra can represent cell's phenotypic information. We develop an accurate discrimination method of bacterial and archaeal species based on machine learning of single-cell Raman data. As a proof-of-concept, we measure Raman spectra of six microorganisms (three bacterial and three archaeal species) and construct a discrimination model using random forest. The model distinguishes among the tested species with >99% accuracies. Owing to the minimally invasive nature of our approach, specific discriminated cells can easily be subjected to further analysis such as single-cell genomics and ecophysiological analysis.
We explore the capability of single-cell Raman microspectroscopy to detect characteristic metabolites in methanogenic archaea Methanosarcina mazei and Methanopyrus kandleri. In both archaea, we observe the Raman spectroscopic features of corrinoids such as vitamin B12 analogues. These features allow for investigating cell-to-cell variation in the corrinoid concentration in M. kandleri and for imaging the spatial distributions of the corrinoid in relatively larger M. mazei cells, both of which could provide useful insight into increased microbial corrinoid production.
In addition to the detection of particular metabolites, the molecular fingerprints encoded in Raman spectra can represent cell's phenotypic information. We develop an accurate discrimination method of bacterial and archaeal species based on machine learning of single-cell Raman data. As a proof-of-concept, we measure Raman spectra of six microorganisms (three bacterial and three archaeal species) and construct a discrimination model using random forest. The model distinguishes among the tested species with >99% accuracies. Owing to the minimally invasive nature of our approach, specific discriminated cells can easily be subjected to further analysis such as single-cell genomics and ecophysiological analysis.