Development of a method for quantifying metabolites in Escherichia coli colonies using hyperspectral imaging

Fermentation by microorganisms has attracted attention for the synthesis of bulk and fine chemicals with high added value, including pharmaceutical intermediates. To accelerate the development of high-producing microbial strains, a rapid screening method is warranted. This study aimed to develop a novel, nondestructive approach to quantify metabolite production in microbial colonies using hyperspectral imaging (HSI). As a model, we examined the heterologous production of 1,3,5-trihydroxyanthraquinone (AQ256), an anthraquinone with antimicrobial and anticancer activities, using Escherichia coli. Fluorescence spectral data from HSI, along with AQ256 concentrations measured via high-performance liquid chromatography, were used to construct regression models. In addition, red-green-blue (RGB)-based models were developed, as AQ256 exhibits a characteristic reddish-brown color. Four regression models were compared: multiple linear regression, partial least squares regression (PLSR), support vector regression, and random forest regression. Among them, the PLSR model based on HSI data showed the highest prediction accuracy (R² = 0.75 ± 0.23, root mean square error = 0.08 ± 0.02, mean absolute error = 0.07 ± 0.02). In particular, it outperformed the RGB-based model in extrapolation beyond the training data. These findings demonstrate that the HSI-based method enables accurate, nondestructive quantification of metabolites and has strong potential for high-throughput screening of microbial strains that produce various valuable compounds at elevated yields.