Single-cell CRISPR screens enable the exploration of mammalian gene function and genetic regulatory networks. However, use of this technology has been limited by reliance on indirect indexing of single-guide RNAs (sgRNAs). Here we present direct-capture Perturb-seq, a versatile screening approach in which expressed sgRNAs are sequenced alongside single-cell transcriptomes. Direct-capture Perturb-seq enables detection of multiple distinct sgRNA sequences from individual cells and thus allows pooled single-cell CRISPR screens to be easily paired with combinatorial perturbation libraries that contain dual-guide expression vectors. We demonstrate the utility of this approach for high-throughput investigations of genetic interactions and, leveraging this ability, dissect epistatic interactions between cholesterol biogenesis and DNA repair. Using direct capture Perturb-seq, we also show that targeting individual genes with multiple sgRNAs per cell improves the efficacy of CRISPR interference and activation, facilitating the use of compact, highly active CRISPR libraries for single-cell screens. Last, we show that hybridization-based target enrichment permits sensitive, specific sequencing of informative transcripts from single-cell RNA-seq experiments.

How cellular and organismal complexity emerges from combinatorial expression of genes is a central question in biology. High-content phenotyping approaches such as Perturb-seq (single-cell RNA-sequencing pooled CRISPR screens) present an opportunity for exploring such genetic interactions (GIs) at scale. Here, we present an analytical framework for interpreting high-dimensional landscapes of cell states (manifolds) constructed from transcriptional phenotypes. We applied this approach to Perturb-seq profiling of strong GIs mined from a growth-based, gain-of-function GI map. Exploration of this manifold enabled ordering of regulatory pathways, principled classification of GIs (e.g., identifying suppressors), and mechanistic elucidation of synergistic interactions, including an unexpected synergy between CBL and CNN1 driving erythroid differentiation. Finally, we applied recommender system machine learning to predict interactions, facilitating exploration of vastly larger GI manifolds.

Seminal yeast studies have established the value of comprehensively mapping genetic interactions (GIs) for inferring gene function. Efforts in human cells using focused gene sets underscore the utility of this approach, but the feasibility of generating large-scale, diverse human GI maps remains unresolved. We developed a CRISPR interference platform for large-scale quantitative mapping of human GIs. We systematically perturbed 222,784 gene pairs in two cancer cell lines. The resultant maps cluster functionally related genes, assigning function to poorly characterized genes, including TMEM261, a new electron transport chain component. Individual GIs pinpoint unexpected relationships between pathways, exemplified by a specific cholesterol biosynthesis intermediate whose accumulation induces deoxynucleotide depletion, causing replicative DNA damage and a synthetic-lethal interaction with the ATR/9-1-1 DNA repair pathway. Our map provides a broad resource, establishes GI maps as a high-resolution tool for dissecting gene function, and serves as a blueprint for mapping the genetic landscape of human cells.

Development of a set of C•G-to-G•C transversion base editors from CRISPRi screens, targetlibrary analysis, and machine learningThe MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters. CitationKoblan, Luke W. et al. "Development of a set of C•G-to-G•C transversion base editors from CRISPRi screens, target-library analysis, and machine learning." Nature Biotechnology (June 2021): dx.

No abstract

Summary SnoN regulates multiple signaling pathways including TGFβ/Smad and p53 and displays both pro-oncogenic and anti-oncogenic activities in human cancer. We have observed previously that both its intracellular localization and expression levels are sensitive to cell density, suggesting that it may crosstalk to Hippo signaling. Here we report that indeed SnoN interacts with multiple components of the Hippo pathway to inhibit the binding of Lats2 to TAZ and the subsequent phosphorylation of TAZ, leading to TAZ stabilization. Consistently, SnoN enhances the transcriptional and oncogenic activities of TAZ, and reducing SnoN decreases TAZ expression as well as malignant progression of breast cancer cells. Interestingly, SnoN itself is downregulated by Lats2 that is activated by the Scribble basolateral polarity protein. Thus, SnoN is a critical component of the Hippo regulatory network that receives signals from the tissue architecture and polarity to coordinate the activity of intracellular signaling pathways.

Language models (LMs) must be both safe and equitable to be responsibly deployed in practice. With safety in mind, numerous detoxification techniques (e.g., Dathathri et al. 2020;Krause et al. 2020) have been proposed to mitigate toxic LM generations. In this work, we show that these detoxification techniques hurt equity: they decrease the utility of LMs on language used by marginalized groups (e.g., African-American English and minority identity mentions). In particular, we perform automatic and human evaluations of text generation quality when LMs are conditioned on inputs with different dialects and group identifiers. We find that detoxification makes LMs more brittle to distribution shift, especially on language used by marginalized groups. We identify that these failures stem from detoxification methods exploiting spurious correlations in toxicity datasets. Overall, our results highlight the tension between the controllability and distributional robustness of LMs.

Cells repair DNA double-strand breaks (DSBs) through a complex set of pathways that are critical for maintaining genomic integrity. Here we present Repair-seq, a high-throughput screening approach that measures the effects of thousands of genetic perturbations on the distribution of mutations introduced at targeted DNA lesions. Using Repair-seq, we profiled DSB repair outcomes induced by two programmable nucleases (Cas9 and Cas12a) after knockdown of 476 genes involved in DSB repair or associated processes in the presence or absence of oligonucleotides for homology-directed repair (HDR). The resulting data enabled principled, data-driven inference of DSB end joining and HDR pathways and demonstrated that repair outcomes with superficially similar sequence architectures can have markedly different genetic dependencies. Systematic interrogation of these dependencies then uncovered unexpected relationships among DSB repair genes and isolated incompletely characterized repair mechanisms. This work provides a foundation for understanding the complex pathways of DSB repair and for optimizing genome editing across modalities.