Table 1.

Evaluation of methods for the high-throughput screening of cell–cell fusion

Methods	Experimental setup needed	Quantification methods	Throughput for genetic screens	Comparison among variants/perturbations
High-content imaging (used in refs. 19,20)	High-content screening microscope	Less precise; using the area of syncytium/fraction of fused cells under the microscope	Limited; >3,000 drugs were screened in ref. 19 and ~6,000 drugs and >30 spike protein variants were screened in ref. 20. Screening of genetic variants and perturbations requires the individual library constructs to be generated and delivered into microwell arrays for imaging. With an imaging throughput of ~30 s per well, screening a genome-wide CRISPR library of 37,722 sgRNAs is estimated to take ~13 d.	Potentially higher variation; There may be a lag time in cell fixation/image acquisition over a large number of samples to evaluate the fast cellular process. The non-pooled experimental setup may be more subject to technical variations such as cell density.
Droplet microfluidics-based method (established in this study)	Droplet microfluidics system and FACS sorter	More quantitative; using NGS	Moderate; 760 spike protein variants were screened in this study. Genetic variant libraries were pooled assembled and delivered into cells. Based on the current system, obtaining droplets with the paired cells for screening a DMS library of ~380 protein variants took ~40 min, followed by ~2 h of FACS sorting. At this rate, screening a genome-wide CRISPR library of 37,722 sgRNAs is estimated to take ~42 h for droplet encapsulation and ~126 h for FACS sorting (total: ~7 d).	Less variation; Pooled assay allows head-to-head comparison of the variants/perturbations under the same experimental setting. Less interference by the neighbouring cells/syncytia due to compartmentalization. Fusion-incompetent cells are greatly depleted, thus offering a wider assay range in defining the enriched fusion-competent variants.
Size-exclusion selection-based method (established in this study)	Cell strainer and FACS sorter	More quantitative; using NGS	High; 760 spike protein variants were screened in this study. Genetic variant and perturbation libraries are pooled assembled and delivered into cells. Only a filtration step using a cell strainer is needed to collect the fused cells, which takes seconds. With a reverse selection approach to collect unfused cells, a genome-wide CRISPR library with 37,722 sgRNAs was screened in this study and it took ~8 h to sort the unfused cells (total: ~8 h).	Less/moderate degree of variation; Pooled assay allows head-to-head comparison of the variants/perturbations under the same experimental setting. There are chances that some fusion-incompetent cells are trapped by neighbouring syncytia as bystanders and retained on the cell strainer, thus resulting in more variation and a narrower assay range.