Genome-scale analysis of interactions between genetic perturbations and natural variation

Abstract

Interactions between genetic perturbations and segregating loci can cause perturbations to show different phenotypic effects across genetically distinct individuals. To study these interactions on a genome scale in many individuals, we used combinatorial DNA barcode sequencing to measure the fitness effects of 7,700 CRISPRi perturbations targeting 1,712 distinct genes in 169 yeast cross progeny (or segregants). We identified 460 genes whose perturbation has different effects across segregants. Several factors caused perturbations to show variable effects, including baseline segregant fitness, the mean effect of a perturbation across segregants, and interacting loci. We mapped 234 interacting loci and found four hub loci that interact with many different perturbations. Perturbations that interact with a given hub exhibit similar epistatic relationships with the hub and show enrichment for cellular processes that may mediate these interactions. These results suggest that an individual's response to perturbations is shaped by a network of perturbation-locus interactions that cannot be measured by approaches that examine perturbations or natural variation alone.

Figure 1:. Generating and phenotyping a panel of double-barcoded segregants carrying CRISPRi perturbations.

a) Design and generation of double-barcoded haploid segregants carrying CRISPRi perturbations. Segregant-specific barcodes were first integrated into the genomic landing pad via a first transformation. Then, barcoded CRISPRi constructs were integrated into the genomic landing pad in a second transformation. These integrations produced a double barcode within each cell that uniquely identified its specific segregant-gRNA combination. b) Pooled fitness assays were used to phenotype all segregant-gRNA combinations. These fitness assays involved performing double barcode amplicon sequencing at multiple time points and using the resulting frequency data to estimate fitnesses for all double barcode lineages. c) Density plot showing fitnesses for all lineages in the two replicate experimental fitness assays with induction of gRNA expression (ATC1 and ATC2). Each data point represents a double-barcode lineage that appears in both ATC1 and ATC2, with an independent fitness estimate calculated for each assay. d) Density plot showing fitnesses for all lineages across one experimental fitness assay with induction of gRNA expression (ATC1) and one control assay without induction of gRNA expression (CON). e) Average fitness values of 74 gRNAs that directly overlap polymorphisms between BY and 3S, across both experimental and control assays. gRNAs were designed for the BY genome; segregants with 3S alleles at the binding site will not perfectly match the gRNA. Asterisks and ‘n.s.’ respectively indicate p-values <1×10⁻⁵ and >0.05 via t-test. f) Distribution of all gRNA effects as calculated by the mixed effects linear model. The estimated null distribution is shown by the dashed line. The threshold used for calling gRNAs with mean effects is three standard deviations below the mean of this null distribution, as indicated by the vertical red line. The gRNA effect here is a mean calculated by averaging across all segregants carrying the gRNA. g) Density plot of average fitness values for all gRNAs with mean effects across the two fitness assays (ATC1 and ATC2). For each gRNA, replicate barcodes within each genotype were averaged before taking the mean across all genotypes. h) Density plot of average fitness values for all gRNAs with mean effects across one experimental fitness assay and one control assay (ATC1 and CON). In all density plots, the data points–double barcode lineages in (c) and (d) and gRNAs in (g) and (h)–are organized in bins to generate heatmaps.

Figure 2:. Calculating deviation values for all segregant-gRNA combinations.

a) Schematic representation of how deviation values are calculated from fitness. b) Heatmap of all deviation values. Segregants are shown along the y-axis, ordered by mean fitness, and gRNAs are shown along the x-axis, ordered by mean effect size. Missing combinations are shown in gray. Only rows and columns with 30% or less missing data are shown. c) Relationship between baseline segregant fitness and deviation values for that genotype. Dots indicate average deviation value, and lines indicate 2 standard deviations. The 95% bootstrap confidence interval for the R² value of 0.149 was 0.145–0.152. d) Relationship between mean gRNA effect and the standard deviation (SD) of deviation values for that gRNA, with the best-fit line shown. The 95% bootstrap confidence interval for the R² value of 0.543 was 0.459–0.618.

Figure 3:. Using deviation values as traits in linkage mapping.

a) Histogram of broad sense heritability (H²) values for all gRNAs with background effects. b) Histogram of the ratio between narrow sense heritability (h²) and H² for all gRNAs with background effects, shown as (1 − h²/H²). c) Linkage mapping on deviation values. Upper panel: Visualization of all 2x-log₁₀(pval) drops for the gRNAs with significant peaks. Each row is a unique gRNA. Locations of major fitness loci are indicated by black bars. Lower panel: Number of overlapping 2x-log₁₀(pval) drops at each nucleotide position along the genome. Regions where more intervals overlap than expected by chance (hubs) are shown in unique colors. Vertical lines indicate chromosome boundaries. Linkage mapping results for the replicate assay (ATC2) are available as Supplemental Figure 15. d) Histogram of all effect sizes for the detected loci. These values were obtained from the linear model used for linkage mapping, deviations ~ locus, taking the coefficient of the locus term. This is roughly equivalent to the difference between the mean deviation value of segregants carrying the 3S allele and the mean deviation value of segregants carrying the BY allele. e) Histogram of the proportion of heritability explained by each detected locus, calculated as R²/H². f) Interaction plots connecting gRNA location to the corresponding 2x-log₁₀(pval) drop for each of the four hubs.

Figure 4:. Features of and comparisons between hubs.

Effects of the Chr VII (a), XIV (b), X (c), and XIII (d) hubs on mean fitness and deviation value across gRNAs that interact with each hub. Dot plots in upper panels: Effects of the hub on mean fitness when each interacting gRNA is induced. Control indicates a subset of 1,294 gRNAs with no phenotypic effect in our data set, as determined by gRNAs with p-values greater than 0.8 for both the mean effect term and the background effect term in a mixed effects linear model (see Methods section ‘Identification and quantification of gRNA effects’). Orange points show the mean fitness of all genotypes with the 3S allele at the peak marker for that gRNA’s interacting locus and green points show genotypes with the BY allele. The fitnesses of replicate barcodes were averaged before calculating the mean for each genotype. Vertical bars for each point indicate standard error. The control data used the center of the hub instead of a peak marker. Barplots in lower panels: Effects of a hub on deviation values across interacting gRNAs. The change in deviation values was calculated by taking the mean of all deviation values among genotypes with the 3S allele at the peak marker and subtracting the mean of all deviations among genotypes with the BY allele. The control data has no deviation values.

Figure 5:. Interaction networks for each hub.

Cellmap visualization of the interaction networks for the Chr VII (a), XIV (b), X (c), and XIII (d) hubs. A list of all genes targeted by a gRNA in this hub were used as a query at thecellmap.org to create these images, at a significance of p ≤ 0.05.