A self-enhanced transport mechanism through long noncoding RNAs for X chromosome inactivation

Abstract

X-chromosome inactivation (XCI) is the mammalian dosage compensation strategy for balancing sex chromosome content between females and males. While works exist on initiation of symmetric breaking, the underlying allelic choice mechanisms and dynamic regulation responsible for the asymmetric fate determination of XCI remain elusive. Here we combine mathematical modeling and experimental data to examine the mechanism of XCI fate decision by analyzing the signaling regulatory circuit associated with long noncoding RNAs (lncRNAs) involved in XCI. We describe three plausible gene network models that incorporate features of lncRNAs in their localized actions and rapid transcriptional turnovers. In particular, we show experimentally that Jpx (a lncRNA) is transcribed biallelically, escapes XCI, and is asymmetrically dispersed between two X's. Subjecting Jpx to our test of model predictions against previous experimental observations, we identify that a self-enhanced transport feedback mechanism is critical to XCI fate decision. In addition, the analysis indicates that an ultrasensitive response of Jpx signal on CTCF is important in this mechanism. Overall, our combined modeling and experimental data suggest that the self-enhanced transport regulation based on allele-specific nature of lncRNAs and their temporal dynamics provides a robust and novel mechanism for bi-directional fate decisions in critical developmental processes.

Figure 1. Regulatory circuits for three models on XCI fate decision system.

The two dashed rectangular boxes in each model illustrate two separate compartments representing the two X chromosomes. (A) Cross inhibition model for XCI. Xist at two X chromosomes are self-enhanced and mutually repressed. Two compartments represent two X chromosomes (X1 and X2) separately. JC represents the complex from the binding of Jpx and CTCF. (B) Self-catalyzed binding model for XCI. The complex JC promotes the binding between Jpx and CTCF, which forms a positive feedback loop. (C) Self-enhanced transport model for XCI. The complex JC between Jpx and CTCF promotes the diffusion for Jpx, which forms a positive feedback loop for Jpx level in one X chromosome. Black arrows represent activation regulation evidenced from experiments, and black short bars represent repression regulation evidenced from experiments. Red dashed links represent hypothetical regulations proposed in three models.

Figure 2. Illustration of the self-enhanced transport mechanism of X chromosome inactivation for multiple X chromosomes, corresponding to the SET model in Fig. 1C.

Jpx, positively regulating XCI through activating Xist, can diffuse among different X chromosomes. In the case of no self-enhancement feedback, all X chromosomes keep to be active because each X has only one Jpx (not enough to trigger XCI). The self-enhancement positive feedback makes the symmetry of Jpx distribution broken, and causes the “reallocation” of Jpx (some X has two Jpx, and some X has 0 Jpx), leading to disparate XCI fates for different X (two active X and two inactive X in this case). The number of Jpx for triggering XCI is not from real data, only for illustration purpose.

Figure 3. The comparison of signal response curves between the model with self-enhanced transport (SET) and the model without feedback.

(A,B) Signal response curves between Jpx and JC, and between Jpx and Xist separately for two X chromosomes (X1 and X2) for the model with self-enhanced transport feedback. (C,D) Signal response curves between Jpx and JC, and between Jpx and Xist separately for two X chromosomes (X1 and X2) for the model without self-enhanced transport feedback. (E) Signal response curves for the binding of CTCF and Jpx. Red line: SET model; green line: the model without self-enhanced transport feedback. (F) Binding isotherm for CTCF-Jpx. Specific binding of Jpx lncRNA to CTCF protein was resolved by RNA gel electrophoresis mobility shift assay (EMSA) (Fig. S3). The binding curve was plotted as the percent bound against CTCF concentration.

Figure 4. Biallelic expression and dispersed Jpx RNA in differentiating ES cells.

(A) RNA FISH detecting Jpx and Xist expression in wild-type female cells during ES cell differentiation at Day 8. Xist cloud (green) is associated with Xi (the inactive X chromosome). Jpx RNA (red) is present on both Xi and Xa (the active X chromosome). Quantitation of cells is shown for the presence of Xist cloud, the Jpx RNA on Xi, and the scattered Jpx RNA in the cell nucleus. (B) Endogenous Xist is upregulated in male ES cells carrying a Jpx-Xist transgene, as examined by RNA FISH in the transgenic male cells during ES cell differentiation at Day 2. DNA FISH was performed for Xpr (X-pairing region, red, bottom panel), which marks the endogenous Xic (X-inactivation center) and therefore the X chromosome. Xist upregulation (green) is associated with the endogenous allele on X. Jpx RNA (red) is present with the active Xist. T, transgene; X, X chromosome. Quantitation of cells from two transgenic male ES clones (#1 and #2) is shown for the presence of Xist upregulation on X, the Jpx RNA associated with Xist, and the scattered Jpx RNA in the nucleus. Arrowheads mark dispersed Jpx RNA. N, sample size.

Figure 5. Trajectories and distributions of Xist RNA for Day 4 and Day 16.

(A,B) Trajectories for Xist RNA expression level at two X chromosomes from day 0 to day 16 obtained from SET model. Fifty cells are plotted, and each trajectory represents a cell in fate decision. The inset shows the trajectories for day 4. X1 represents X chromosome 1, and X2 represents X chromosome 2. (C,D) Distributions of cells for Xist 1 (Xist level at X chromosome 1) and Xist 2 (Xist level at X chromosome 2) at day 4 and day 16. At day 4, the cell population is mostly distributed on the left bottom corner, indicating that most of cells express neither Xist 1 nor Xist 2. For day 16, the cell population is mostly distributed in the right bottom or left top corner, indicating that for most of cells one out of the two X chromosome expresses Xist. (E) The comparison between experiments and simulations from SET model for the percentage of cells with 0 Xist cloud and 1 Xist cloud at day 4. (F) The comparison between experiments and simulations for the relative Xist RNA level for Jpx knockout at Day 12. Cyan bars represent wild- type, and magenta bars represent one Jpx is knocked out.

Figure 6. The comparison between simulations and experiments for Jpx transgenic female ES cells.

(A,B) Trajectories for Xist RNA expression level at two X chromosomes from day 0 to day 8 obtained from SET model for adding transgene (Jpx). (C,D) The comparison between experiments and simulations for the percentage of cells with different number of Xist clouds for adding transgene (Jpx).

Figure 7. Regulatory circuits for the SET model adding Rnf12, Rex1, and Tsix related regulations.

Two compartments represent two X chromosomes (X1 and X2) separately. JC represents the complex from the binding of Jpx and CTCF. (A) The expanded SET model (ESET) with the self-enhanced diffusion for Jpx, assuming that the complex JC between Jpx and CTCF promotes the diffusion for Jpx, which forms a positive feedback loop for Jpx level in one X chromosome. (B) The model without self-enhanced diffusion for Jpx (WSET). Black arrows represent activation regulation evidenced from experiments, and black short bars represent repression regulation evidenced from experiments. Red links represent hypothetical regulations proposed in the model. Rectangle nodes represent long noncoding RNA, and ellipse nodes represent protein.