Limits...
A map of directional genetic interactions in a metazoan cell.

Fischer B, Sandmann T, Horn T, Billmann M, Chaudhary V, Huber W, Boutros M - Elife (2015)

Bottom Line: Gene-gene interactions shape complex phenotypes and modify the effects of mutations during development and disease.The effects of statistical gene-gene interactions on phenotypes have been used to assign genes to functional modules.Our study presents a powerful approach for reconstructing directional regulatory networks and provides a resource for the interpretation of functional consequences of genetic alterations.

View Article: PubMed Central - PubMed

Affiliation: Genome Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany.

ABSTRACT
Gene-gene interactions shape complex phenotypes and modify the effects of mutations during development and disease. The effects of statistical gene-gene interactions on phenotypes have been used to assign genes to functional modules. However, directional, epistatic interactions, which reflect regulatory relationships between genes, have been challenging to map at large-scale. Here, we used combinatorial RNA interference and automated single-cell phenotyping to generate a large genetic interaction map for 21 phenotypic features of Drosophila cells. We devised a method that combines genetic interactions on multiple phenotypes to reveal directional relationships. This network reconstructed the sequence of protein activities in mitosis. Moreover, it revealed that the Ras pathway interacts with the SWI/SNF chromatin-remodelling complex, an interaction that we show is conserved in human cancer cells. Our study presents a powerful approach for reconstructing directional regulatory networks and provides a resource for the interpretation of functional consequences of genetic alterations.

Show MeSH

Related in: MedlinePlus

Experimental design.Re-analysis of the multi-parametric genetic interaction data set of Horn et al. (2010), a square matrix of all pairwise combinations of 93 genes. Matrix columns (playing the role of query genes) were ordered from left to right according to their ability to explain the data in the remaining columns. The explained variance is shown on the y-axis as a function of the number of query genes. The graph illustrates that already 17 suitably selected query genes are sufficient to explain 90% of the variance in the data.DOI:http://dx.doi.org/10.7554/eLife.05464.004
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4384530&req=5

fig1s1: Experimental design.Re-analysis of the multi-parametric genetic interaction data set of Horn et al. (2010), a square matrix of all pairwise combinations of 93 genes. Matrix columns (playing the role of query genes) were ordered from left to right according to their ability to explain the data in the remaining columns. The explained variance is shown on the y-axis as a function of the number of query genes. The graph illustrates that already 17 suitably selected query genes are sufficient to explain 90% of the variance in the data.DOI:http://dx.doi.org/10.7554/eLife.05464.004

Mentions: We generated the largest map of multi-phenotype genetic interaction profiles in metazoan cells to date by co-depleting gene pairs by RNAi in cultured Drosophila S2 cells, high-throughput imaging of single-cell phenotypes, and modelling of gene–gene interactions (Figure 1A). We selected 1367 genes implicated in key biological processes, that is, signalling, chromatin biology, cell cycle regulation and protein turnover control (Supplementary file 1). Each of these 1367 target genes was tested against 72 query genes in all pairwise knockdown combinations (2 × 2 dsRNAs), following previously established approaches (Casey et al., 2008; Horn et al., 2011; Laufer et al., 2013) (Figure 1—figure supplement 1). The 72 query genes were selected from an initial single-gene screen on the 1367 genes, to cover a range of phenotypes, processes and protein complexes (Figure 1—figure supplement 2 and Supplementary file 2). After 5 days, cells were fixed and stained for DNA, α-tubulin, and Ser9-phosphorylated histone 3, a mitosis marker. Cells were imaged by automated whole-well fluorescence microscopy, and phenotypic features were extracted using an image analysis pipeline (see ‘Materials and methods’). On average, 15,962 cells were imaged and analysed per well. Single-cell measurements were aggregated into 328 cell population features per experiment such as cell number, mitotic index, nuclear and cellular area, and other descriptors of shape and morphology (Supplementary file 3). 162 features were highly reproducible between replicates, with Pearson correlation >0.6 (Figure 1B–C). Using a step-wise feature selection algorithm, we determined a subset of 21 features (Supplementary file 4) that non-redundantly captured the range of phenotypes. The algorithm ensured that each feature contained new information not yet covered by the already selected features (Figure 1D and ‘Materials and methods’) (Laufer et al., 2013).10.7554/eLife.05464.003Figure 1.Combinatorial RNAi to map multi-phenotype genetic interactions.


A map of directional genetic interactions in a metazoan cell.

Fischer B, Sandmann T, Horn T, Billmann M, Chaudhary V, Huber W, Boutros M - Elife (2015)

Experimental design.Re-analysis of the multi-parametric genetic interaction data set of Horn et al. (2010), a square matrix of all pairwise combinations of 93 genes. Matrix columns (playing the role of query genes) were ordered from left to right according to their ability to explain the data in the remaining columns. The explained variance is shown on the y-axis as a function of the number of query genes. The graph illustrates that already 17 suitably selected query genes are sufficient to explain 90% of the variance in the data.DOI:http://dx.doi.org/10.7554/eLife.05464.004
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4384530&req=5

fig1s1: Experimental design.Re-analysis of the multi-parametric genetic interaction data set of Horn et al. (2010), a square matrix of all pairwise combinations of 93 genes. Matrix columns (playing the role of query genes) were ordered from left to right according to their ability to explain the data in the remaining columns. The explained variance is shown on the y-axis as a function of the number of query genes. The graph illustrates that already 17 suitably selected query genes are sufficient to explain 90% of the variance in the data.DOI:http://dx.doi.org/10.7554/eLife.05464.004
Mentions: We generated the largest map of multi-phenotype genetic interaction profiles in metazoan cells to date by co-depleting gene pairs by RNAi in cultured Drosophila S2 cells, high-throughput imaging of single-cell phenotypes, and modelling of gene–gene interactions (Figure 1A). We selected 1367 genes implicated in key biological processes, that is, signalling, chromatin biology, cell cycle regulation and protein turnover control (Supplementary file 1). Each of these 1367 target genes was tested against 72 query genes in all pairwise knockdown combinations (2 × 2 dsRNAs), following previously established approaches (Casey et al., 2008; Horn et al., 2011; Laufer et al., 2013) (Figure 1—figure supplement 1). The 72 query genes were selected from an initial single-gene screen on the 1367 genes, to cover a range of phenotypes, processes and protein complexes (Figure 1—figure supplement 2 and Supplementary file 2). After 5 days, cells were fixed and stained for DNA, α-tubulin, and Ser9-phosphorylated histone 3, a mitosis marker. Cells were imaged by automated whole-well fluorescence microscopy, and phenotypic features were extracted using an image analysis pipeline (see ‘Materials and methods’). On average, 15,962 cells were imaged and analysed per well. Single-cell measurements were aggregated into 328 cell population features per experiment such as cell number, mitotic index, nuclear and cellular area, and other descriptors of shape and morphology (Supplementary file 3). 162 features were highly reproducible between replicates, with Pearson correlation >0.6 (Figure 1B–C). Using a step-wise feature selection algorithm, we determined a subset of 21 features (Supplementary file 4) that non-redundantly captured the range of phenotypes. The algorithm ensured that each feature contained new information not yet covered by the already selected features (Figure 1D and ‘Materials and methods’) (Laufer et al., 2013).10.7554/eLife.05464.003Figure 1.Combinatorial RNAi to map multi-phenotype genetic interactions.

Bottom Line: Gene-gene interactions shape complex phenotypes and modify the effects of mutations during development and disease.The effects of statistical gene-gene interactions on phenotypes have been used to assign genes to functional modules.Our study presents a powerful approach for reconstructing directional regulatory networks and provides a resource for the interpretation of functional consequences of genetic alterations.

View Article: PubMed Central - PubMed

Affiliation: Genome Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany.

ABSTRACT
Gene-gene interactions shape complex phenotypes and modify the effects of mutations during development and disease. The effects of statistical gene-gene interactions on phenotypes have been used to assign genes to functional modules. However, directional, epistatic interactions, which reflect regulatory relationships between genes, have been challenging to map at large-scale. Here, we used combinatorial RNA interference and automated single-cell phenotyping to generate a large genetic interaction map for 21 phenotypic features of Drosophila cells. We devised a method that combines genetic interactions on multiple phenotypes to reveal directional relationships. This network reconstructed the sequence of protein activities in mitosis. Moreover, it revealed that the Ras pathway interacts with the SWI/SNF chromatin-remodelling complex, an interaction that we show is conserved in human cancer cells. Our study presents a powerful approach for reconstructing directional regulatory networks and provides a resource for the interpretation of functional consequences of genetic alterations.

Show MeSH
Related in: MedlinePlus