Pooled CRISPR screening of high-content cellular phenotypes by ghost cytometry

Abstract

Fast enrichment of cells based on morphological information remains a challenge, limiting genome-wide perturbation screening for diverse high-content phenotypes of cells. Here we show that multi-modal ghost cytometry-based cell sorting is applicable to fast pooled CRISPR screening for both fluorescence and label-free high-content phenotypes of millions of cells. By employing the high-content cell sorter in the fluorescence mode, we enabled the genome-wide CRISPR screening of genes that affect NF-κB nuclear translocation. Furthermore, by employing the multi-parametric, label-free mode, we performed the large-scale screening to identify a gene involved in macrophage polarization. Especially the label-free platform can enrich target phenotypes without invasive staining, preserving untouched cells for downstream assays and unlocking the potential to screen for the cellular phenotypes even when suitable markers are lacking.

Label-free ghost cytometry for manufacturing of cell therapy products

“…However, we foresee the importance of evaluating the application of LF-GC for the assessment of more complex phenotypes related to the therapeutic potential of cellular therapies. For instance, we have recently shown that LF-GC can accurately classify the exhaustion states of T cells, 17 of which were reported as a key indicator of CAR-T cell functionality. 28,29 LF-GC performed 5-part cell type classification from peripheral WBC and classification of polarized M1/M2 macrophages.…”

“…28,29 LF-GC performed 5-part cell type classification from peripheral WBC and classification of polarized M1/M2 macrophages. 16,17 Lastly, future investigations exploring the use of LF-GC to assess non-T cell therapies such as CAR-macrophages or CAR-NK cells may offer additional insights into any shared or unique mechanisms driving the therapeutic potential of cell-based immunotherapies.…”

“…A possible approach to overcome this issue is the employment of an unsupervised machine learning approach, where training populations are defined solely by the morphological differences present in a mixed population. 17 Future studies that assess a variety of machine learning approaches, including supervised and unsupervised methods as well as their combinations, to define 'ground truth' populations will clarify the degree of achievable classification accuracy.…”

“…LF-GC has been shown to classify many complex phenotypes that are critical for the manufacturing of cellular therapies including; live-dead state, differentiations state, and functional subtyping. [15][16][17] In this work, we investigate the use of LF-GC for QC in cell therapy manufacturing. We show that LF-GC is a minimally disruptive, label-free cell analytical method that can be used to characterize important features of cell therapy products including cell number, viability, and phenotypic state.…”