Hyperspectral Confocal Fluorescence & Raman Microscopy for Characterizing Plant Response at the Cellular and Subcellular Level: Current Progress and Future Opportunities

Le résumé fourni par la source

Hyperspectral confocal fluorescence and Raman microscopy have both demonstrated the ability to provide detailed images with sub-micron spatial resolution of biomolecules in plants based on signals from the interaction of excitation light with the endogenous molecules within the plant cells and tissues. Hyperspectral confocal fluorescence microscopy (HCFM) is based on the principles of fluorescence which occurs when tissue absorbs light of a shorter wavelength and partially re-emits it at a longer wavelength. Typical applications of HCFM in plant science have leveraged autofluorescence emitted by naturally occurring photosynthetic pigments in plant tissue such as chlorophylls and bilins [1-3]. In contrast, hyperspectral confocal Raman microscopy (HCRM) is based on spontaneous Raman spectroscopy, which probes the vibrational excitations of molecules through the inelastic scattering of photons. Typical applications of HCRM in plant science have studied molecules found in the cell walls such as lignin and cellulose, as well as carotenoids [4]. Though the two methods arise from distinct optical phenomena they fall under the umbrella field of spectroscopic imaging and are often similar in their microscope hardware design, the most common of which is a raster scanning confocal microscope coupled to a spectrometer. They also share the advantages of being label-free and nondestructive. This presentation will summarize the current state-of-the-art in the field of hyperspectral confocal fluorescence and Raman microscopy for plant science, using recent work by my lab to identify pigment changes in tall fescue in response to heavy metal stress [5] and carotenoid localization in pepper plants [6] as example applications. The discussion will detail the methods at all the key steps in the experimental process including sample preparation, hardware considerations, spectral image acquisition techniques, pre-processing methods [7], spectral deconvolution and modeling, and quantitative image analysis. Emphasis will be placed on linking each step in the experimental process to its potential effect on the end results. Despite the success of these spectroscopic imaging methods, they are still limited to applications with static or slow temporal behavior, in the tens of seconds for fluorescence and minutes to tens of minutes for Raman methods. Faster acquisition times can be achieved but the trade-off is lower spatial or spectral resolution and often both. Additionally, the spectral signatures, while information rich, are often spectrally congested and contaminated by signal from background molecules, presenting a significant challenge for automated analysis and data interpretation. In recent years, advances in imaging sensors and artificial intelligence and machine learning (AI/ML) have emerged that have the potential to address these challenges. The future promise of these technologies for 1) increasing the speed of spectroscopic imaging data acquisition, 2) increasing the efficiency of spectral analysis, and 3) improving feature selection and analysis automation will be presented [8].

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.