Quantitative genetic-interaction mapping in mammalian cells

Abstract

Mapping genetic interactions (GIs) by simultaneously perturbing pairs of genes is a powerful tool for understanding complex biological phenomena. Here we describe an experimental platform for generating quantitative GI maps in mammalian cells using a combinatorial RNA interference strategy. We performed ∼11,000 pairwise knockdowns in mouse fibroblasts, focusing on 130 factors involved in chromatin regulation to create a GI map. Comparison of the GI and protein-protein interaction (PPI) data revealed that pairs of genes exhibiting positive GIs and/or similar genetic profiles were predictive of the corresponding proteins being physically associated. The mammalian GI map identified pathways and complexes but also resolved functionally distinct submodules within larger protein complexes. By integrating GI and PPI data, we created a functional map of chromatin complexes in mouse fibroblasts, revealing that the PAF complex is a central player in the mammalian chromatin landscape.

Figure 1

An overview of the mammalian E-MAP pipeline. Flowchart of the experimental setup: esiRNAs to a set of genes are arrayed in a pairwise fashion (in quadruplicate) in tissue culture plates. Reverse transfection is then performed, and the resulting fitness defects are observed using high-content imaging. Raw data are scored and phenotypic signatures are derived for each gene.

Figure 2

Characteristics and quality control of GI data set. (a) GI scores derived from independent biological replicate experiments. Each data point represents a GI score (S score) for the same pairwise knockdown derived from independent experiments. (b) Knockdown efficiencies as measured by quantitative reverse transcription PCR. (c) Relationship between knockdown efficiency and expression: data were split into five groups and plotted against gene expression relative to that of a housekeeping gene.

Figure 3

Comparison of GI and PPI data sets. (a,b) Comparison of individual S scores (a) and GI profiles (b) to the likelihood of the corresponding pairs of proteins being physically associated. (c) Fitness after single gene knockdown versus the number of genetic interactions associated with the same gene. Raw cell counts were used as a proxy for fitness, and a cutoff of S ≥ 2 and S ≤ −2 was used to define informative GIs on the y axis.

Figure 4

A module map of chromatin-related genes. (a) GIs (left) and profile Pearson correlations (right) for members of the PAF transcriptional elongation complex. (b) GIs (left) and profile Pearson correlations (right) for members of the CNOT complex. S scores and profile Pearson correlations were used in a and b. (c) A module map based on a manually curated set of protein complexes (Supplementary Table 4). Modules and inter-module GI bundles are colored according to the enrichment of the observed GIs, with gray signifying no enrichment of a particular interaction type was observed (Supplementary Table 5). (d) Examples of positive and negative GI bundles corresponding to particular edges on the module map.

Figure 5

Validation of observed GIs with an orthogonal phenotypic readout. (a) Micrographs derived from mouse embryo fibroblast cells depleted for Ctr9 in combination with knockdowns of Ruvbl1, Ruvbl2, Actl6a or Morfl1. Scale bars, 20 μm. (b) Phenotype strength represented as a fraction of cells with elongated nuclei (longer bars represent more extreme phenotype). Error bars, s.d. based on 100 random samples of 10% of the data in each experiment.