A high-throughput small molecule screen identifies synergism between DNA methylation and Aurora kinase pathways for X reactivation

Abstract

X-chromosome inactivation is a mechanism of dosage compensation in which one of the two X chromosomes in female mammals is transcriptionally silenced. Once established, silencing of the inactive X (Xi) is robust and difficult to reverse pharmacologically. However, the Xi is a reservoir of >1,000 functional genes that could be potentially tapped to treat X-linked disease. To identify compounds that could reactivate the Xi, here we screened ∼367,000 small molecules in an automated high-content screen using an Xi-linked GFP reporter in mouse fibroblasts. Given the robust nature of silencing, we sensitized the screen by "priming" cells with the DNA methyltransferase inhibitor, 5-aza-2'-deoxycytidine (5azadC). Compounds that elicited GFP activity include VX680, MLN8237, and 5azadC, which are known to target the Aurora kinase and DNA methylation pathways. We demonstrate that the combinations of VX680 and 5azadC, as well as MLN8237 and 5azadC, synergistically up-regulate genes on the Xi. Thus, our work identifies a synergism between the DNA methylation and Aurora kinase pathways as being one of interest for possible pharmacological reactivation of the Xi.

Keywords: Aurora kinase; DNA methyltransferase; X reactivation; high-throughput screen; small molecules.

Fig. 1.

Summary of X-reactivation screen of small molecules. (A) Derivation of Xi-TgGFP cells from TTFs isolated from pups heterozygous for the X-linked GFP transgene. (B) Steps of the screening process. Initial analysis identified 632 compounds (199 active + 433 inconclusive, C) that subsequently failed rescreening. Reanalysis of the primary screen data identified 1,895 compounds not among the initial 632; 1,394 of these were sourced for rescreening along with an additional 2,641 test compounds. Of 15 compounds that passed this stage, confirmation by qRT-PCR was obtained only for VX680, MLN8237, and 5azadC (Methods and Fig. 2). (C) Primary screen results. The two replicates are plotted against each other on a normalized 0–100 scale. Blue lines indicate the threshold of 10 for active compounds on each axis.

Fig. S1.

Well images from the primary high content screen. (A) Neutral control, 0.5 μM 5azadC + DMSO; (B) 8.0 μM 5azadC; and (C) 7.5 μM VX680 + 0.5 μM 5azadC. All images show cells incubated with the indicated compounds for 3 d. (Top) Hoechst staining. (Middle) GFP imaging. (Bottom) Segmentation, a representation of the well after reanalysis by MetaXpress software under the less stringent conditions discussed in the text (the color gray marks Hoechst⁺ cells and green marks the GFP⁺ cells among these). The numbers accompanying these panels are the cell count and percent GFP⁺. Yellow arrows indicate examples of symmetrical, hazy well-bottom artifacts that are not counted as cells.

Fig. 2.

GFP reactivation by Aurora kinase inhibitors, VX680 and MLN8237. (A) Xi-TgGFP female fibroblasts were treated for 3 d with VX680 at 1 μM, 5azadC at 0.5 μM, or both. Expression is relative to X-TgGFP/Y male cells. Means ± SD of three to five biological replicates are shown. Note that the y axis is a logarithmic scale. Fold differences vs. controls are indicated in red; *P = 0.02, ***P < 0.001. (B) The same cells were treated with 1 μM MLN8237 and tested as in A. **P = 0.01; ***P < 0.001. (C) Xi-TgGFP cells treated with control, Aurka, Aurkb, or both Aurka and Aurkb siRNAs. (Left) qRT-PCR of GFP expression, mean RNA levels ± SD, *P = 0.02. Note logarithmic scale on y axis. (Right) Knockdown efficiency of each Aurk gene assessed by qRT-PCR, ***P < 0.001.

Fig. 3.

VX680 effects on cell proliferation and Xist RNA. (A) A total of 20,000 cells were plated per well in 12-well plates (“plated” on x axis). After 3-d treatment with indicated concentrations, viable cells were counted. Means of three experiments ± SD are shown. (B, Left) DNA FISH with probe for GFP transgene (Tg). (Center) RNA FISH with probe for Xist. (Right) Merged images include DAPI stain of nuclei. (Top) FISH of control cells treated with DMSO. (Bottom) FISH of cells treated with 1 μM VX680 for 3 d. (C) Quantification of GFP transgene signals from DNA FISH experiments. ***P < 0.001. (D) Relative amounts of Xist RNA upon 3-d treatment with 0.5 μM 5azadC, 1 μM VX680, or both, shown by qRT-PCR or by a fold-change calculation from RNA-seq data (*FDR < 0.05 for each drug treatment compared with the DMSO control).

Fig. S2.

Xist expression by RNA-seq. This figure is an extension of Fig. 3D. Normalized cas (Xa) and mus (Xi) reads are shown. Xi-TgGFP cells were treated as indicated by labels at Left: 5azadC, VX680, or 5azadC + VX680. Scale: cas, 0–11 and mus, 0–7.5.

Fig. 4.

GFP reactivation as a function of DNA content, shown by FACS. Xi-TgGFP cells were treated for 3 d with (A) 5 μM 5azadC, (B) 1 μM VX680, or (C) 1 μM VX680 + 0.5 μM 5azadC and subject to FACS analysis. (Left) Gating for GFP⁺ cells, after initial gating (Methods). The x axes, side scatter (log scale); y axes, GFP fluorescent signal (log scale). The percent of GFP⁺ cells is indicated at Top Right. (Center) Histogram of DNA content for all cells at Left. The x axes, Hoechst fluorescence (note that ploidy is on linear scale); y axes, cell count. The number of cells is indicated at Bottom Right. (Right) Histogram of DNA content for GFP⁺ cells at Left. The x axes, Hoechst fluorescence; y axes, cell count. The number of GFP⁺ cells is indicated at Bottom Right.

Fig. S3.

GFP reactivation as a function of SSC, shown by FACS. This figure is a continuation of Fig. 4. The x axes show SSC (log scale) and the y axes, GFP fluorescence signal (log scale). Gate indicates GFP⁺ cells; the percent GFP⁺ is indicated at Top Right. Treatments of VX680 are at 1 μM, 5azadC are at 0.5 μM, and high 5azadC are at 5 μM; all treatments are for 3 d. Note that T4 fibroblasts lack the GFP transgene.

Fig. 5.

Allele-specific analysis of X-linked gene expression. (A) RNA-seq, GFP aligned with its cDNA sequence. (B) Mecp2 expression assayed by luciferase. (Left) Normalized activity from Xi-linked Mecp2-Luc (Xi-8 cell line) after 3-d treatment with 1 μM VX680 (VX) or 1 μM MLN8237 (MLN) treatments ±0.5 μM 5azadC. Ctrl, control (DMSO) treatment. (Right) Comparison with a control line, Xa-3, where Mecp2-Luc is on the Xa (note logarithmic scale). Error bars indicate SD. Means of at least three biological replicates are shown. **P = 0.01 (P ≤ 0.005 for 5azadC, 5azadC + VX, and 5azadC + MLN each compared with the control without 5azadC.) ***P = 4.4 × 10⁻⁵. (C) Lamp2, (D) Fhl1, and (E) Msn aligned with mm9 in IGV (Top track for each panel). Note that, for the allelic analysis, reads appear only where there are polymorphisms that enable distinction between cas (Xa) and mus (Xi). For B–D, normalized cas and mus reads are shown. The scale is indicated at Right; note that it is smaller for mus reads. Xi-TgGFP cells were treated as indicated: 5azadC, VX680 (VX), or 5azadC + VX680 as in Fig. 2.

Fig. S4.

Additional RNA-seq tracks for reactivated genes. This figure is an extension of Fig. 5. Normalized cas (Xa) and mus (Xi) reads are shown. The scales are indicated at Right, which differ between genes and the cas and mus alleles. Xi-TgGFP cells were treated as indicated by labels at Left: 5azadC, VX680, and 5azadC + VX680.

Fig. S5.

Extended VX680 and MLN8237 treatment. This figure is an extension of Fig. 5B. Mecp2 expression in Xi-Mecp2-luc cells. Normalized luciferase (LUC) assays after VX680 (VX) or MLN8237 (MLN) treatments at indicated concentrations ± 0.5 μM 5azadC. Ctrl, control (DMSO) treatment. Although there is a trend of higher Mecp2-luc activity upon the addition of VX680 or MLN8237, these conditions are not statistically significant compared with 5azadC alone (ANOVA). Error bars indicate SD. Means of three biological replicates are shown.

Fig. 6.

Reactivated genes on the Xi due to compound treatments. (A) Venn diagram of reactivated Xi-linked genes following indicated treatment. Expression of each gene increased at least threefold with an FDR of <0.05. (B) Distribution of reactivated genes along the Xi. Nucleotide coordinates in megabases are indicated at Left; X-linked RefSeq genes are in blue to the Right of the coordinates. Of these, reactivated genes are named with the color indicated by the small Venn diagram. Locations of Dxz4 and Xist are indicated in gray. (C) RNA-seq results summarized for the three compound treatments for the Xi (mus alleles). In each panel, log₂(fold-change of drug treatment vs. control) on the y axis is plotted against average expression levels across the samples, expressed as log₂(counts per million) on the x axis. Each dot represents one gene. Red dots represent genes where the log₂(fold-change) is significant (FDR < 0.05); the black lines represent a threshold of a threefold change in either direction, i.e., log₂(fold-change) > 1.6 or < −1.6.

Fig. S6.

Changes in Chr13 gene expression. This figure is an extension of Fig. 6C. (Left three panels) RNA-seq results for the mus alleles of Chr13, a representative autosome. Xi-TgGFP cells were treated with 0.5 μM 5azadC, 1.0 μM VX680, and the combination of the two drugs, as indicated in green. In each panel, log₂(fold-change of drug treatment vs. control) on the y axis is plotted against average expression levels across the samples, expressed as log₂(counts per million), on the x axis. Each dot represents one gene. Red dots represent genes where the log2(fold-change) is significant (FDR < 0.05); the black lines represent a threshold of a threefold change in either direction, i.e., log2(fold-change) > 1.6 or < −1.6. (Right) Tables showing the average log₂(fold-change) compared with the control treatment, using the absolute values for logFCs of both up- and down-regulated genes. Averages are shown over all genes or those that are significantly changed upon treatment (FDR < 0.05), on the X chromosome and on Chr13.