Article Figures & Data
Figures
- Figure 1. Targeting multiple ERBSs with synthetic activators can activate gene expression.
(A) Cartoon depicting the targeting of multiple ERBSs in combination to study combinatorial effects on gene expression. (B) Relative locations of ERBSs (blue), ERBS-adjacent regions (red), and genes (green) tested in this study. (C–F) Targeting all 12 ERBSs in combination with dCas9-vp16(10x) (red) or dCas9-p300(core) (green) activated gene expression at MMP17 (C), CISH (D), FHL2 (E), and HES2 (F) to levels that are comparable with an 8-h E2 treatment (light gray). Targeting all ERBSs had significantly higher activation than targeting ERBS-adjacent regions, which is not significantly different than controls that target the IL1RN promoter. Error bars represent SEM. P-values are calculated with respect to control using a one-way ANOVA with Dunnett’s multiple comparisons (*P-value < 0.05, **P-value < 0.01, ***P-value < 0.001).
- Figure S1. dCas9-activator constructs can activate gene expression (related to Fig 1).
(A) A schematic shows dCas9-p300(core) and dCas9-VP16(10x) constructs used in this study. (B) IL1RN activation is induced by targeting either dCas9-p300(core) or dCas9-VP16(10x) to the IL1RN promoter. Error bars represent SEM. (C) Activation from simultaneously targeting all three ERBSs near target genes induces activation to a level correlated with an 8-h estrogen induction.
- Figure 2. Targeting dCas9-activator constructs is specific and induces H3K27ac.
(A, B) The relative HA ChIP-seq signal, an epitope tag on dCas9, is shown for all targeted ERBSs (A) and adjacent regions (B) and compared with non-targeted controls. Points indicate individual replicates and error bars represent SEM. (C) The fold change induction of H3K27ac ChIP-seq signal across all targeted loci shows no significant difference between dCas9-p300(core) and VP16(10x). (D) H3K27ac was induced at all adjacent regions by dCas9-VP16(10x) and at three of six adjacent regions by dCas9-p300(core). Points indicate individual replicates and error bars represent SEM. (E, F) Browser tracks of H3K27ac induced by targeting ERBSs with dCas9-p300(core), dCas9-VP16(10x), or an 8-h E2 treatment are shown at MMP17 (E) and CISH (F). Numbers on the right of the tracks indicate the track height in non-normalized reads per million.
- Figure S2. Targeting dCas9-VP16(10x) to ERBSs and adjacent regions is specific and induces H3K27ac at the intended loci (related to Fig 2).
(A, B) HA ChIP-seq browser tracks show specific targeting of ERBSs (lane 1) and adjacent regions (lane 2) compared with a control with the IL1RN promoter targeted (lane 3) for the CISH locus (A) and MMP17 locus (B). (C) ChIP-seq browser tracks show targeting of dCas9 constructs to the IL1RN promoter (HA—lanes 1 and 2), H3K27ac induced by dCas9-p300(core) (lanes 3 and 4) or dCas9-VP16(10x) (lanes 3 and 5), and RNAPII induced by dCas9-vp16(10x) (lanes 6 and 7). (D, E) H3K27ac ChIP-seq browser tracks for sites targeted by dCas9-p300(core) and dCas9-VP16(10x) show similar histone acetylation as an 8-h estrogen treatment for FHL2 (D) and HES2 (E). (F) Targeting dCas9-p300(core) or dCas9-VP16(10x) to regions adjacent to ERBSs induces H3K27ac. (G) H3K27ac levels before normalization (shown as reads per million) for ERBSs before and after targeting by dCas9-p300(core) (blue/green) or dCas9-VP16(10x) (gray). (H) Limited fold changes in RNAPII levels were observed when targeting individual ERBSs with dCas9-VP16(10x) or dCas9-p300(core).
- Figure 3. Activation of gene expression by targeting CRISPRa to combinations of enhancers.
(A–D) (left) The combination of ERBSs targeted by dCas9-VP16(10x) or bound by ER upon E2 treatment are shown in a schematic. (A–D) (right) The relative fold change in expression as measured by qRT-PCR, when compared with control cells with guides targeting the IL1RN promoter, was determined for each combination of ERBSs. Each data point is shown as a blue dot; error bars represent SEM. Pairwise log2 ratios and significance levels are given in Table S1 (Pairwise t test P-values comparing to control: *** < 0.001, ** < 0.01, * < 0.05).
- Figure S3. Activation from individual ERBSs is specific as well as independent of targeting efficiency and acetylation induced by dCas9-VP16(10x) (related to Figs 2, 3, and 5).
(A, B) dCas9-VP16(10x) was targeted to ERBSs in clonal cell lines in which individual ERBSs have been deleted. When targeting the deleted site at ERBSs surrounding MMP17 (A) and CISH (B), we do not see activation, confirming specificity of activation. P-values are calculated using a holm adjusted ttest (***P-value < 0.001, **P-value < 0.01, *P-value < 0.05). (C–E) Correlation plots showing relationship between activation by dCas9-p300(core) (teal, C–E) or dCas9-VP16(10x) (red, C–E) and fold change ChIP-seq signal of HA (dCas9) (C) H3K27ac (D) and RNAPII (E). r is Pearson’s correlation coefficient.
- Figure 4. Thermodynamic modeling reveals little cooperativity between synthetic activator-bound ERBSs.
(A) Schematic showing the set of modeled parameters. (B–E) Parameters were fit to gene expression data for MMP17, CISH, FHL2, and HES2 (from Fig 3). Plots show the distribution of fitted parameters. Parameter sets were selected if the modeled data correlated with gene expression data within 0.1 of an optimal correlation. Vertical bars represent the mean.
- Figure S4. Thermodynamic modeling of dCas9-p300(core) shows independence between sites (related to Figs 3 and 4).
(A–C) Relative fold change expression levels from targeting combinations of ERBSs with dCas9-p300(core) as measured by qRT-PCR for MMP17 (A) CISH (B) and HES2 (C). (D) Correlation between activation by dCas9-VP16(10x) and dCas9-p300(core) at individual ERBSs. (E–G) Parameters of a thermodynamic model were fit to relative expression data from combinatorial activation of ERBSs with dCas9-p300. (E) Interactions between sites at MMP17 remain mostly neutral. (F, G) CISH (F) and HES2 (G) interaction parameters appear bimodal.
- Figure 5. Sufficiency-necessity comparison shows similar results at individual sites, but differences in cooperation between sites.
(A) Scatter plot shows relative expression, as measured by z-score, for both activation and interference at individual ERBSs. Z-scores were negated for interference so that a higher score is associated with greater necessity. (B, C) Scatter plots show the parameters of thermodynamic models derived from CRISPRa and Enhancer-i for MMP17 (B) and CISH (C). Parameters are shown as mean ± 95% confidence intervals. (D) Clustered correlation matrix showing Pearson correlations between potential predictors and both dCas9-VP16(10x) activation and dCas9-p300(core) activation. (E) Analysis of relative importance of predictors for activation using the Lindeman, Merenda, and Gold method (see the Materials and Methods section). Importance is normalized to sum to one.
- Figure S5. Thermodynamic modeling of Enhancer-I shows cooperativity between sites and correlation of predictors with gene activation (related to Figs 4 and 5).
(A–D) Parameters were fit to relative expression data from combinatorial inhibition of ERBSs with SID(4x)-dCas9-KRAB (Carleton et al, 2017). We see favorable (more negative) interactions between ERBSs 1 and 2 (parameter ⍵1,2) for MMP17 (A), CISH (B), and FHL2 (C). Vertical bars represent the mean. (D) Comparison of modeled parameters from activation to inhibition showing mean fitted parameters in the two studies. We see that the cooperativity of FHL2-1 and FHL2-2 is specific to inhibition, whereas the remaining parameters are relatively similar. Error bars represent 95% confidence intervals around the mean. (E–H) Correlation plots showing relationship between synthetic activation dCas9-p300(core) (teal) or dCas9-VP16(10x) (red) and possible genomic predictors of activation (x-axis).
Supplementary Materials
Table S4 Plasmid construction primers.
Table S5 gRNA sequences.
Table S6 qRT-PCR Primers.
