(664e) Excitation-Scanning Hyperspectral Imaging Technologies for Multilabel Cellular Imaging

Hyperspectral imaging technologies were originally developed by NASA and the Department of Defense for satellite imagery and remote sensing applications. More recently, these technologies have found a wide range of potential uses in the biomedical field, from basic research studies to clinical translation. One area in which hyperspectral imaging technologies have been highly utilized is fluorescence microscopy. Typical hyperspectral imaging fluorescence microscopy approaches involve optical filtering or dispersion of the emitted fluorescence in order to measure the fluorescence emission spectrum (emission-scanning). However, an alternative technological approach can be used to allow measurement of the fluorescence excitation spectrum (excitation-scanning). When implemented using broad-band or long-pass fluorescence emission filters, we have previously shown that the technique of excitation-scanning hyperspectral imaging may allow for significantly improved signal strength and/or imaging speed. Here, we report on the capabilities of excitation-scanning hyperspectral imaging microscopy for detection of many (5+) fluorescent signals in cell and tissue preparations for two application areas: live cell imaging (cell signaling) and tissue diagnostics.

Cell signaling studies were performed utilizing both pulmonary microvascular endothelial cells and human airway smooth muscle cells. Cells were labeled to allow detection of either cAMP (using the H188 Turquoise-Epac-Venus FRET reporter) or Ca²⁺ (using the calcium indicator Cal-520 AM). Time-lapse cell signaling spectral image data were analyzed using custom spectral unmixing and image post-processing algorithms implemented in MATLAB (MathWorks), as well as a custom software package for automated ROI-based cell signal quantitation called S8. Tissue imaging studies were performed using de-identified pairs of resected cancerous and non-involved colorectal tissues, obtained in accordance with the University of South Alabama Institutional Review Board (IRB exempt). Tissue spectral image data were analyzed using convolutional neural network-based approaches to differentiate cancerous and non-involved tissues. For both live cell imaging studies and tissue imaging studies, images were acquired using a custom excitation-scanning spectral imaging microscope that allowed for sequential excitation of fluorescence across a range of wavelengths, using either a 300 W Xe arc lamp (Titan X300, Sunoptic Surgical) and tunable filter assembly (VF-5, Sutter Instruments) or a custom LED-based spectral illumination module, with detection provided by a high-speed sCMOS camera (Prime 95B, Photometrics).

Results from both cell and tissue imaging studies indicate that sufficient information is present in excitation-scanning spectral image data to allow separation of 5+ fluorescent labels with sufficient signal strength so as to visualize and automatically quantify fluorescent molecule levels within subcellular regions. Additionally, results from tissue imaging studies indicate that excitation-scanning spectral image data can be used for automatic classification and differentiation of cancerous and non-involved tissue types. Finally, our results indicate that excitation-scanning spectral imaging can be implemented using a range of technological approaches, including broad-band illumination sources (arc lamps or supercontinuum laser) with tunable filters or through combination of many discrete wavelength sources (LEDs). These results suggest that excitation-scanning spectral imaging is a versatile alternative to emission-scanning spectral imaging that may be implemented on a range of microscope platforms. This work was supported by NIH awards P01HL066299, R01HL58506, and R01HL137030, and NSF award MRI1725937. Drs. Leavesley and Rich disclose financial interest in a start-up company, SpectraCyte LLC, that was formed to commercialize spectral imaging technologies.