Retinoic acid accelerates downregulation of the Xist repressor, Oct4, and increases the likelihood of Xist activation when Tsix is deficient

Abstract

Background: Imbalances in X-linked gene dosage between the sexes are resolved by transcriptionally silencing one of two X-chromosomes in female cells of the early mammalian embryo. X-inactivation is triggered by expression of the non-coding Xist gene. In turn, Xist is dually regulated by the antisense Tsix RNA and by the Oct4 pluripotency factor. Although there is general agreement that Tsix is an inhibitor of Xist, some laboratories have observed ectopic Xist induction in differentiating male ES cells when Tsix is mutated, whereas we have not observed significant changes in Xist. These observational differences have led to fundamentally diverse models of X-chromosome counting. Here, we investigate if different methods of cell differentiation and use of all -trans retinoic acid (RA) could be causative factors and how they might impact Xist expression.

Results: We compared suspension and cell-adhesion cultures in the presence or absence of RA and find that RA significantly impacts Xist expression in Tsix-mutant male cells. Whereas the standard embryoid body method infrequently leads to ectopic Xist expression, adding RA generates a significant number of Xist-positive male cells. However, while normal Xist clouds in wild-type female cells are robust and well-circumscribed, those found in the RA-treated mutant males are loosely dispersed. Furthermore, ectopic Xist expression does not generally lead to complete gene silencing. We attribute the effect of RA on Xist to RA's repressive influence on Oct4, a pluripotency factor recently shown to regulate Tsix and Xist. RA-treated ES cells exhibit accelerated decreases in Oct4 RNA levels and also display accelerated loss of binding to Xist intron 1. When Tsix is deficient, the faster kinetics of Oct4 loss tip the equilibrium towards Xist expression. However, the aberrant Xist clusters are unlikely to explain elevated cell death, as X-linked silencing does not necessarily correlate with the qualitatively aberrant Xist clusters.

Conclusions: We conclude that RA treatment leads to premature downregulation of Oct4 and partial derepression of Xist irrespective of X-chromosome counting. RA-induced Xist clusters in male cells do not result in global or stable silencing, and excess cell death is not observed. These data and RA's known pleiotropic effects on ES transcription networks suggest that RA differentation bypasses normal X-inactivation controls and should be used judiciously. We propose that the likelihood of Xist expression is determined by a balance of multiple Xist activators and repressors, and that levels of Oct4 and Tsix are crucial toward achieving this balance.

Figure 1

Summary of Xist phenotypes in Tsix-mutant male ES cell lines. All mutant ES cell lines are XY diploid male (except where noted) and listed with the names used in the original references. XY Tg, male transgenic line; shaded rectangle represents the 37-kb region re-inserted into a deleted locus; •••, extends beyond the Xic diagram depicted; RA, retinoic acid. SA/tpA, please refer to cited references for details. The 'yes/no' designation for pAA2Δ1.7 refers to the very low number of Xist-positive cells and the fact that there is no excessive cell death or inviability reported in these cells or mice. Differentiation in the presence (+) or absence (-) of retinoic acid is indicated.

Figure 2

Ectopic Xist RNA clouds observed in Tsix-mutant male ES cells. A. Percentage of cells containing Xist RNA clouds. Note the high percentage of Xist-positive cells in X^Δ/Y cells when differentiated in the presence of RA. Days of differentiation are listed as day 2 (d2), day 4 (d4), and day 6 (d6). Day 6 X^Δ/X EB+RA, not determined (n.d.). X/Y, wild-type male ES line; X/X, wild-type female ES line; X^Δ/Y, Tsix^ΔCpG/Y male ES line; X^Δ/X, Tsix^ΔCpG/+ female ES line; n > 200 for each timepoint and cell line examined. Error bars represent the standard deviation from three independent biological replicates. P was calculated by paired, two-tailed Student's t-tests. Differences in the number of Xist⁺ cells for RA-free versus RA-induced X^Δ/Y cells were statistically significant on all differentiation days (each red asterisk denotes P < 0.05 for day 2, day 4, and day 6 comparisons). EB, EB+RA, TC, and TC+RA conditions are described in the Materials and Methods. B. Xist RNA FISH showing ectopic and diffuse Xist cluster formation (arrows) in differentiating Tsix^ΔCpG male cells. Arrows denote Xist RNA clusters. The RNA FISH probe is double-stranded. Pinpoint signals primarily detect mildly elevated Xist expression in mutant cells caused by the Tsix deficiency, though ~10% residual Tsix expression from an upstream promoter may also contribute [13]. Below is a side-by-side comparison at day 6 (in grayscale) of a normal compact Xist RNA cloud (X/X, left) next to the ectopic dispersed Xist cluster observed in X^Δ/Y cells (right). C. X- and Y-chromosome paint confirm that all ES cell lines carry the appropriate number of sex chromosomes. Red, X-chromosome; green, Y-chromosome.

Figure 3

RA-induced differentiation causes sharper decreases in Oct4 mRNA levels. Xist and Oct4 RNA levels in wild-type and Tsix-mutant male ES cells under various differentiation conditions. Error bars represent one standard deviation from the mean. Days of differentiation are listed as day 0 (d0), day 2 (d2), day 4 (d4), and day 6 (d6). Pairwise comparisons were made between corresponding samples at each differentiation time point for EB vs. EB+RA (red asterisks) and TC vs. TC+RA (green asterisks). * P < 0.05. Statistical significance was calculated using paired, two-tailed Student's t-tests.

Figure 4

Oct4-binding to Xist intron 1 decreases faster when differentiated in the presence of RA. Quantitative ChIP analyses of Oct4-binding to Xist intron 1 show faster reductions in Oct4-binding kinetics when wild-type and Tsix-deficient male cells were differentiated by EB+RA versus EB methods. Note that, for each experiment, we started with a single d0 (undifferentiated) culture for both X/Y and X^Δ/Y, and then split them on subsequent days for analysis. Thus, d0 samples were the same for all EB series and for all EB+RA series, and their values were duplicated in the graphs for easier comparisons with d2, d4, d6 samples in the same series. Means ± s.e.m from two independent biological replicates are shown. Pairwise comparisons were made between corresponding samples (EB vs. EB+RA) normalized to the control IgG ChIP (background) for each time point. Statistical significance of each result was calculated using paired, two-tailed Student's t-tests. * p < 0.05.

Figure 5

RA-containing cultures lead to silencing of the X-linked gene Pgk1 and exhibit slower overall growth than cultures without RA. A. Xist/Pgk1 RNA FISH showing the percentage of Pgk1 expression in X^Δ/Y cells differentiated under the EB or EB+RA condition at day 6. EB+RA percentages represent only those cells positive for dispersed Xist RNA clusters. EB percentages represent all cells expressing basal Xist levels. B. Xist/Cot-1 RNA FISH of X/X (EB) and X^Δ/Y (EB+RA) with Xist RNA clouds at day 6. Percentages represent Xist+ cells that were scored to be located within a Cot-1 RNA hole, partially within a Cot-1 hole, or excluded from a Cot-1 hole, as indicated. Note the higher percentage of Xist clouds excluded from a Cot-1 hole in X^Δ/Y cells when compared to X/X cells. Arrows point to Xist clouds/clusters and their corresponding Cot-1 regions. C. Growth curves for wild-type and Tsix-mutant male cells differentiated under EB/EB+RA (left) or TC/TC+RA (right) conditions. Cultures containing RA (denoted in shades of red) had <50% of cells than those cultures grown in the absence of RA (denoted in shades of blue). No differences were seen between X/Y and X^Δ/Y grown under similar conditions. Error bars represent one standard deviation from the mean; days of differentiation are listed as day 0 (d0), day 2 (d2), day 4 (d4), and day 6 (d6).

Figure 6

Model of RA-induced Xist expression in Tsix-mutant male cells. A. Undifferentiated wild-type X/Y cells maintain high levels of Oct4, Sox2, and Nanog, which bind to Xist intron 1. This synergistic binding -- in addition to high Tsix RNA expression -- effectively suppresses Xist. Under the EB differentiation method (left), Oct4 expression remains high early during differentiation, allowing continued binding of Oct4, Sox2, and Nanog to keep Xist off. Although Oct4 levels and binding decrease during late differentiation, Xist continues to remain repressed as the Xist-dependent phase of XCI has passed. Under the RA differentiation method (right), RA binds to its receptor to form RAR complexes, which then bind to and negatively regulate Oct4. However, persistent Tsix expression maintains stable Xist repression. B. In undifferentiated X^Δ/Y cells, high levels of Oct4, Sox2, and Nanog are sufficient to repress Xist in the absence of Tsix. Under the EB differentiation method (left), Oct4 expression remains high early during differentiation, allowing continued binding of Oct4, Sox2, and Nanog to maintain low Xist expression even in the absence of Tsix transcription. Although Oct4 mRNA levels and binding to Xist intron 1 decrease during late differentiation, Xist is stably repressed as the Xist-dependent phase has passed. However, in the RA differentiated state (right), RAR complexes bind to Oct4, causing premature Oct4 downregulation; loss of Oct4 binding to Xist intron 1 and absence of Tsix lead to moderate Xist upregulation during the early (Xist-dependent) stages of differentiation. Due to Xist derepression during this early stage and lack of antisense Tsix transcription, Xist RNA starts to accumulate and diffusely coats the chromosome in cis, leading to formation of visible Xist RNA clusters and incomplete silencing. DE and PE refer to distal and proximal Oct4 enhancers, respectively.