Supplementary Fig. 1: ECCITE-seq enables CRISPR screens with single-cell multimodal readout.

a, Species-mixing proof-of-principle experiment: protein tag reads associated with each cell barcode. Points are colored based on species classification using transcripts as shown in Fig. 1b. About 1.2% of either human or mouse cells show cross-reactivity with mouse or human antibodies respectively. b, Mixed-species ECCITE-seq experiment demonstrating measurement of six cellular modalities. Cells of different origin were stained with ECCITE-seq and hashing antibodies, washed and combined to a single 10x run. All cells were stained with a mix of anti-human CD29 and anti-mouse CD29 antibodies. NIH-3T3 cells were split into 7 tubes and stained with 7 barcoded hashing antibodies (Hashtag-A to Hashtag-G), followed by washing and pooling. MyLa, Sez4 and PBMCs were stained with Hashtag_1, Hashtag_2 and Hashtag_3 respectively. Eventually, a cell mix comprising 65% NIH-3T3 cells, 25% PBMCs, and 5% each MyLa and Sez4 cells was loaded into the 10x Chromium for droplet formation and reverse transcription. After emulsion breakage and cDNA amplification, the distinctly sized products were separated with size selection capture beads and amplified separately. A pool consisting 88% cDNA, 7% guide-tag, 3% hashtag and 2% protein-tag library was submitted for NGS sequencing. c,d, sgRNA representation as measured by direct sgRNA capture or genomic DNA amplification of the guide variable region in two different cell lines. To assess the direct capture, we assigned a guide to each cell with a single sgRNA and quantified their proportion in the total cell population (orange). This ratio is compared to the ratio of genomic reads matching each sgRNA to the total genomic sgRNA reads. e, scRNA-seq saturation curves for the ECCITE-seq versus standard 10x V(D)J run using the K562 cells expressing the targeting sgRNA library. The same cell suspension was used into two parallel experiments; one underwent a standard 10x V(D)J run while the other was preceded by staining with CITE-seq and hashing antibodies, as well as spike-in of guide RT primer in the reverse transcription reaction. For this analysis, data were downsampled to the same total reads per cell after correcting for the different fraction of reads in cell. f. Levels of target mRNA or protein (moving median smoothing window = 151) in 1,000 cells ranked on decreasing sgRNA counts for CD46 sgRNA 2 or non-targeting sgRNA 4. g, Adjusted p-values of detecting the indicated gene expression or protein abundance change as a function of cell number. Different-size, randomized cell samples assigned to unique targeting sgRNAs were tested against equal size, randomized cell samples from the non-targeting sgRNAs groups, using the Wilcoxon rank sum test.