A major challenge in phenotype discovery projects is the inconsistency in the measurements of rare markers across subsets of varying AF. Full spectral cytometers incorporate autofluorescence (AF) extraction to the unmixing workflows, improving the resolution of AF-affected markers. The recommended procedure to isolate pure AF references in Aurora, called “Discover, Distinguish and Designate” (DDD), involves the isolation of unique AF subsets in unstained samples via their progressive evaluation of NxN plots based on raw FCS fluorescence parameters. Although this manual approach easily enables the isolation of references for populations with unique AF profiles such as AMΦ, it becomes increasingly laborious if applied to the isolation of all potentially relevant AF subsets in a study.
As an alternative approach, we made use of the inherent power of dimensionality reduction algorithms to resolve AF complexity across Aurora 64 fluorescence detectors into a two-dimensional map of defined subsets. We demonstrate that opt-SNE consistently resolves complex unstained cellular mixes into subsets with a high level of spectral purity and unique scatter profiles, reveals differences in the AF fingerprint of tissues-resident cells, demonstrates replica consistency in fasting models and highlights minor differences as between female and male mice tissues. When applied to the evaluation of viral infection, drastic changes in AF makeup per tissue were revealed. Our analysis of opt-SNE projections based on virtual Aurora configurations, revealed the critical requirement of UV and violet lasers in combination for the resolution of mouse AF subsets.
This is the first reported instance of the application of high-dimensionality reduction algorithms to the label free dissection of unique AF signatures in complex mammalian cell mixtures. The workflow here described can also be used to unveil differences in the autofluorescence fingerprints of tissues in homeostasis and after immunological challenges.