Massively parallel phenotyping of coding variants in cancer with Perturb-seq

Genome sequencing studies have identified millions of somatic variants in cancer, but it remains challenging to predict the phenotypic impact of most. Experimental approaches to distinguish impactful variants often use phenotypic assays that report on predefined gene-specific functional effects in bulk cell populations. Here, we develop an approach to functionally assess variant impact in single cells by pooled Perturb-seq. We measured the impact of 200 TP53 and KRAS variants on RNA profiles in over 300,000 single lung cancer cells, and used the profiles to categorize variants into phenotypic subsets to distinguish gain-of-function, loss-of-function and dominant negative variants, which we validated by comparison with orthogonal assays. We discovered that KRAS variants did not merely fit into discrete functional categories, but spanned a continuum of gain-of-function phenotypes, and that their functional impact could not have been predicted solely by their frequency in patient cohorts. Our work provides a scalable, gene-agnostic method for coding variant impact phenotyping, with potential applications in multiple disease settings. Overall design: To map variant impact, we profiled A549 cells (ATCC CCL-185) with 64 channels of 10X Chromium Single Cell 3 RNA-seq v2 (10X Genomics #120237), expressing a pool of barcoded variant sequences The 64 channels were equally split into two experiments: 32 channels for TP53 variants and 32 channels for KRAS variants. We loaded 7,000 cells per channel for each cDNA library to obtain a total of 224,000 cells per experiment (448,000 cells total across the TP53 and KRAS experiments). For each experiment, paired-end libraries were sequenced over 32 lanes on an Illumina Hiseq 2500 per sequencing parameters recommended by 10X Genomics: cell barcode read length 26 bp, index read length 8 bp and transcript read length 98 bp. No dial-out PCR was done, in contrast to typical Perturb-seq (Dixit et al. 2016; Jaitin et al. 2016; Adamson et al. 2016). Variants were assigned to cells based on reads from the RNA-seq data. Detailed information about the supplementary file formats in this work (raw and processed data) is provided in the supplementary file SCEVIP.README.pdf.