Development of a Clustering-Based Screening Using Hyperspectral Imaging for High-Producing Strains in Escherichia coli

Abstract

Efficient screening of high-producing strains from large combinatorial libraries remains an important challenge in synthetic biology. Hyperspectral imaging (HSI) enables the non-destructive evaluation of microbial colonies through the simultaneous acquisition of spectral and spatial information. In this study, we developed a novel screening strategy based on the clustering analysis of HSI-derived spectral data without the need to construct training strain sets with different production levels. As a proof of concept, we evaluated this approach using Escherichia coli with differential expression of biosynthetic genes for the production of a type II polyketide secondary metabolite, 1,3,8-trihydroxyanthraquinone (AQ256). A combinatorial library was constructed in which the transcription of AQ256 biosynthetic genes (antA-I and Pl-mcat) from Photorhabdus laumondii TTO1 was controlled by the T7 promoter or its variants. Clustering analysis revealed that AQ256 accumulation was divided into statistically distinct clusters, including those enriched in high-producing strains. The highest AQ256-producing strain, AQ-6′, identified through this screening, produced 49.0 mg L−1 in LB medium, which increased to 89.2 mg L−1 after optimizing cultivation conditions. Furthermore, proteomic analysis indicated that the increased expression of AntD/AntE, responsible for polyketide chain elongation, was associated with enhanced production. These results demonstrate that clustering based on HSI-derived spectral data enables the screening of high-producing strains without the need to prepare training data labeled according to production levels in advance and provides a useful strategy for enhancing type-II polyketide synthase-derived metabolite production.