Analysis of the carotenoid cycle during microbial growth by combining fluorescence imaging and deep learning

Detection and real-time tracking of changes in the relative carotenoid content during microbial growth can be extremely challenging. Additionally, analyzing its content cycling time can also be highly difficult. In this study, Raman tweezers were employed to identify the carotenoid in two strains of marine Bacillus spores (designated #2430 and #2966). Further, live-cell dynamic imaging and fluorescence microscopy were utilized to observe the growth morphology and monitor the variations in the relative carotenoid content within the spores. Additionally, a UNet-VGG16-based deep learning algorithm was incorporated to augment the precision of the analysis. The results of this study revealed the presence of varying levels of carotenoids between the two strains of spores across different growth stages. The dividing spores showed a progressive elongation in the cycle of relative carotenoid content fluctuation. In contrast, the non-dividing spores demonstrated more stable cycles, likely due to the lack of carotenoid accumulation, as they required carotenoids for their normal growth rather than cell division. To the best of our knowledge, this study is the first to detect dynamic and real-time variations in the relative carotenoid content during the growth of the two marine Bacillus spore strains, thereby providing novel perspectives on their metabolic characteristics and potential applications.