Peaks to Pixels: Multivariate Unmixing of Hyperspectral Resonance Raman Image for Mapping Newly Synthesized Carotenoids in Micro-factories

Carotenoids are naturally occurring biomolecules with potent antioxidant activity that humans must obtain through their diet, as they cannot synthesize them endogenously. These pigments are naturally produced by plants and microbes, including red yeast cells that act as micro-factories, particularly genera such as Rhodosporidium and Rhodotorula. However, selectively visualizing newly synthesized carotenoids at the single-cell level using fluorescence staining or mass spectrometry remains a major challenge. Taking advantage of the strong resonance Raman enhancement of carotenoids with excitation at 532 nm laser, we developed a single-cell nascent carotenoid detection, monitoring, and imaging framework by coupling resonance Raman spectroscopy with reverse carbon stable isotope probing (rRrSIP). This approach enables rapid detection and high-resolution imaging of intracellular carotenoids within individual red yeast cells. Furthermore, we implemented both univariate imaging to map carotenoid distribution and multivariate unmixing algorithms, specifically, vertex component analysis (VCA) coupled with non-negative least squares to extract pure carotenoid Raman signatures from hyperspectral datasets and map the spatial distribution across multiple cells, without requiring extraction procedures. This combined strategy not only captures nascent carotenoid distribution but also validates turnover dynamics in pure carotenoids in cellulo. Additionally, employing another unmixing algorithm, the multivariate curve resolution–alternating least squares (MCR-ALS) framework enabled us to visualize the localization of newly synthesized carotenoids within the lipid-rich regions of cells. Overall, this integrative strategy enhances high-resolution, rapid tracking and visualization of both global and nascent carotenoids directly within individual cells.