Targeting microRNA-dependent control of X chromosome inactivation improves the Rett Syndrome phenotype

Abstract

X chromosome inactivation (XCI) is induced by Xist long non-coding RNA and protein-coding genes. However, the role of small non-coding RNA function in XCI remains unidentified. Our genome-wide, loss-of-function CRISPR/Cas9 screen in female fibroblasts identified microRNAs (miRNAs) as regulators of XCI. A striking finding is the identification of miR106a among the top candidates from the screen. Loss of miR106a is accompanied by altered Xist interactome, leading to dissociation and destabilization of Xist. XCI interference via miR106a inhibition has therapeutic implications for Rett syndrome (RTT) girls with a defective X-linked MECP2 gene. Here, we discovered that the inhibition of miR106a significantly improves several facets of RTT pathology: it increases the life span, enhances locomotor activity and exploratory behavior, and diminishes breathing variabilities. Our results suggest that miR106a targeting offers a feasible therapeutic strategy for RTT and other monogenic X-linked neurodevelopmental disorders.

Fig. 1. A CRISPR/Cas9 screen identifies miRNAs as regulators of mammalian XCI.

A Schematic summary of the CRISPR/Cas9 screen in female mouse fibroblast BMSL2 cells. The active X chromosome (X_a) harbors the deletion of Xist. B, C BMSL2 cells expressing a sgRNA against the indicated miRNAs or, as a control, a non-silencing (NS) sgRNA were selected in the HAT medium. The representative images are shown (B), and the results are quantified from three experiments after crystal violet staining (C). Unpaired, two-sided t-test with no multiple adjustments. For miR106a, *p = 0.017; miR363, *p = 0.0303; miR340, *p = 0.026; miR34b, ***p = 4.21 × 10^-5; miR30e, *p = 0.02; miR181a, **p = 0.0032. D–G Two-color RNA FISH monitoring expression of G6pdx (Red) and Mecp2 (Green), and Pgk1 (Red) and Lamp2 (Green) in each of the six miRNAs knockdown H4SV cell lines. DAPI staining is shown in blue. The representative images are shown (D, F), and the results are quantified from three experiments (E, G). H DNA FISH monitoring X chromosome content in the six miRNA KD H4SV cell lines is quantified from three experiments. I qRT-PCR analysis monitoring miRNA levels in the whole-cell, nuclear, and cytoplasmic fractions from H4SV cells. Data were analyzed from three experiments. Data is expressed as mean ± SD (C, E, G, H, I). Scale bars: 5 µm (D, F).

Fig. 2. miR106a physically interacts with RepA in Xist RNA.

A Strategy to capture miR106a-RepA complex (Left). Competitive elution of RepA from miR106a-RepA complex using mismatch, perfect, or imperfect complementary capture oligonucleotides. Samples pretreated with RNase were used as a negative control. The experiment was performed three times; a representative image from one experiment is shown (Middle), and the results are quantified (Right) from three different experiments. For transcript A, Oligo 2, *p = 0.013, Oligo 3, *p = 0.034, Perfect match ^nsp = 0.22; For transcript B, Oligo 4, ^nsp = 0.061, Oligo 5, ^nsp = 0.17, Perfect match ^*p = 0.031; For transcript C, Oligo 1, **p = 0.0034, Oligo 6, ^nsp = 0.083, Perfect match ^nsp = 0.126. B Outline of the modified PARIS2 method. C PARIS2 assay validates predicted miR106a interactions (Colored arcs) in the RepA region. Fractions of duplex groups (DGs) corresponding to each site are shown. Numbers in parentheses are the counts of gapped alignments in each DG. n = 2. D The percentage of mapped reads for each amplicon from the replicates is shown. Data were analyzed from two experiments and expressed as mean ± SD. E qRT-PCR analysis following PARIS2 assay monitoring the enrichment of miR106a-RepA duplex in control and LNA-treated cells. The predicted miR106a non-binding region (NBS) in Xist is a negative control. n = 3. For MRE 1, **p = 0.0072; MRE2, **p = 0.0056; MRE3, **p = 0.0043; MRE4, *p = 0.020. F RIP assay monitoring miR106a-Xist interaction in H4SV cells using biotinylated miR106a (Left) and Xist-specific probes (Right). The predicted miR106a non-binding region (NBS) in Xist and U6 is a negative control. n = 3. For miR106sp RIP, NBS, *p = 0.029; Gapdh, *p = 0.03. For Xist RIP, *p = 0.015. Data is expressed as mean ± SD (A, E, F). Unpaired, two-sided t-test with no multiple adjustments (A, E, F).

Fig. 3. Loss of miR106a-RepA pairing interferes with Xist-X_i association.

A qRT-PCR analysis monitoring Xist levels in H4SV cells expressing miR106sp following treatment with actinomycin D. Gapdh mRNA was used as a normalization control. n = 3. For 0 h, *p = 0.014; 2 h, **p = 0.0013; 4 h, **p = 0.0089. B Schematic showing the in vitro stability assay for P³²-labeled RepA in whole-cell lysates expressing NS, miR106sp, or miR106a mimics (Left). Slot blot assay monitoring stability of in vitro synthesized uncapped RepA in a time-dependent manner. In vitro-synthesized Gapdh is used as an endogenous control. The representative images (Middle) and the results quantified from three experiments (Right) are shown. For RepA, 24 h. miR106a mimics vs. NS, ***p = 0.00049; miR106sp vs. NS, ***p = 0.00035; miR106sp vs. miR106a mimics ***p = 9.39 × 10^-6. C qRT-PCR analysis of EU-labeled nascent Xist RNA prepared from the cells expressing control or miR106sp. Gapdh mRNA was used as a normalization control. n = 3. For 0 h, **p = 0.0091; 2 h, **p = 0.0061; 4 h, **p = 0.0052. D, E Confocal-airyscan sections of nuclei stained for Xist in control and miR106sp- (D) or LNAs (E) treated H4SV cells. The proportion of Xist-positive cells is quantitated (Right). The representative images are shown, and the results are quantified from three experiments D, E. Data is expressed as mean ± SD (A–C). Unpaired, two-sided t-test with no multiple adjustments (A–C). Scale bars: 2 µm (D, Left) and 5 µm (D, Right).

Fig. 4. Loss of miR106a-RepA partnering compromises the assembly of Wtap and m⁶A formation on Xist.

A Volcano plot depicting differential pull-down of 62 Xist-interacting proteins in control and miR106a-depleted cells. Blue indicates proteins downregulated (n = 4), Red indicates proteins upregulated (n = 2), and Gray indicates no change (n = 56) in miR106a-depleted cells relative to control cells. n = 3. B RIP assay monitoring binding of Wtap to Xist and Actin in control or LNAs-treated cells. n = 3. *p = 0.015. C Immunoblot of proteins retrieved by Xist and isogenic control probes from H4SV cells expressing NS or miR106sp by ChIRP. Input is 2% of the total protein. The representative images (Left) and the results quantified from three experiments (Right) are shown. Wtap: Control vs. miR106sp, *p = 0.032; Mettl3: Control vs. miR106sp, *p = 0.023; Mettl14: Control vs. miR106sp, **p = 0.0029. D qRT-PCR analysis monitoring m⁶A, m⁵C, and Ψ enrichment on Xist in H4SV cells expressing control or miR106sp. The predicted non-binding site (NBS) in Xist for m⁶A, m⁵C, and Ψ is used as a negative control. n = 3. For m⁶A: NS vs. miR106sp, **p = 0.0069. E Scheme of in vitro methylation assay with RepA using ³H-labeled SAM (Left). Slot blot assay monitoring methylation of in vitro synthesized RepA incubated with whole cell lysate prepared from cells expressing NS or miR106sp in a time-dependent manner. The representative images (Middle) and the results quantified from three experiments (Right) are shown. RepA pre-treated with RNase was used as a negative control. For NS: 4 h, **p = 0.0024; 8 h, **p = 0.0048; 24 h, **p = 0.0054. For miR106sp: 4 h, *p = 0.027; 8 h, **p = 0.0039; 24 h, **p = 0.01. F RIP assay monitoring binding of Ythdc1 to Xist and Gapdh in H4SV cells ectopically expressing NS or miR106sp or Wtap shRNA. n = 3. Control vs. miR106sp, *p = 0.035; Control vs. Wtap KD, *p = 0.049; Control vs. miR106sp+Wtap KD, *p = 0.033; miR106sp vs. miR106sp+Wtap KD, *p = 0.05; Wtap KD vs. miR106sp+Wtap KD, **p = 0.0037. G qRT-PCR analysis of pSM33 Xist-(BoxB)₃ ES cells expressing DC1 and Wtap shRNA or miR106sp on Xist-mediated gene silencing. Gene expression was normalized to Gapdh and Gpc4 levels prior to Xist induction in cells expressing empty vectors. n = 3. For WTAP KD, Dox^- vs. Dox⁺, *p = 0.012, Dox⁺ vs. Ythdc1⁺, *p = 0.014, dox^- vs. Ythdc1⁺, ***p = 0.00029; For miR106sp, Dox^- vs. Dox⁺, *p = 0.03, Dox⁺ vs. Ythdc1⁺, **p = 0.0029, dox^- vs. Ythdc1⁺, **p = 0.0045. H Schematic model of miR106a-directed regulation of X_i silencing. Data is expressed as mean ± SD (B–G). Unpaired, two-sided t-test with no multiple adjustments (B–G).

Fig. 5. miR106sp-mediated Mecp2 expression in the brain improves the neurobehavioral phenotype of RTT.

A Immunoblot (Top) and quantitation (Bottom) of Mecp2 levels in the whole-brain lysates derived from the AAV9-miR106sp and AAV9-empty-injected Tsix-Mecp2 and control mice at ~16 weeks. n = 3. For AAV9-empty vs. AAV9-miR106sp, ***p = 0.00068; AAV9-miR106sp vs. Control, *p = 0.015; AAV9-empty vs. Control, *p = 0.036. B The lifespan of AAV9-miR106sp-injected mice (n = 15) compared with AAV9-empty-injected (n = 18), control (n = 18), Tsix-Mecp2 (n = 8) and Mecp2^-/Y (n = 7) animals. The data was analyzed using the log-rank test. C Representative images of AAV9-empty-injected (n = 1) or AAV9-miR106sp-injected (n = 1) or control (n = 2) mice at ~12 weeks of age. D The percent changes of in vivo MRI brain volume (Left) and brain area (Right) in AAV9-empty-injected (n = 6) or AAV9-miR106sp-injected (n = 6) or control (n = 2) mice at ~6 weeks of age. Control animals were used as a reference for images and were excluded from the statistical analyses. For brain volume, AAV9-empty vs. AAV9-miR106sp, *p = 0.041. For brain area, AAV9-empty vs. AAV9-miR106sp, *p = 0.035. E, F Representative images (Top) and the quantitation (Bottom) of AAV9-empty-injected (n = 17) or AAV9-miR106sp-injected (n = 14) or control (n = 18) mice showing hind-limb clasping (E) and tail lesions (F). For the hind-limb clasp, AAV9-empty vs. AAV9-miR106sp, *p = 0.033. For the tail lesion, AAV9-empty vs. AAV9-miR106sp, *p = 0.024. G Latency to fall (mean + /- SEM) in rotarod test at 9, 12, and 16 weeks old AAV9-empty-injected (n = 13) or AAV9-miR106sp-injected (n = 12) or control (n = 12) mice. Data for the three consecutive days of evaluation is shown. The data was analyzed using a Welch one-way ANOVA with Dunnett’s post hoc multiple comparisons. H Averaged speed in the Barnes maze test for AAV9-empty-injected (n = 13) or AAV9-miR106sp-injected (n = 12) or control (n = 12) mice at 9, 12, and 16 weeks. Data for the five consecutive days of evaluation is shown. I Apneas (>0.5 sec.) measured by whole-body plethysmograph (mean + /- SEM) per 5 min as measured in AAV9-empty-injected or AAV9-miR106sp-injected or control mice at the indicated times (n = 6 animals per group). J Summary data (mean + /- SD) illustrating changes in short-term (SD1) and long-term (SD2) variabilities in the respiratory cycle period for AAV9-empty-injected or AAV9-miR106sp-injected or control mice at 9, 12, and 16 weeks (n = 6 animals per group). p-values are included in the Source data (G–J). Data is expressed as mean ± SD (A, D, G–J). Unpaired, two-sided t-test with no multiple adjustments (A, D, H–J).

Fig. 6. miR106sp-mediated MECP2 expression in RTT neurons improves the morphological and calcium transient deficits.

A qRT-PCR analysis monitoring the expression of X_i-linked MECP2 in 4- and 12-week-old RTT neurons infected with LTV-empty or LTV-miR106sp. Neurons derived from a clone expressing wild-type MECP2 from the X_a were used as a positive isogenic control (Control), which was set to 1. n = 3. For 4 weeks, LTV-empty vs. LTV-miR106sp, *p = 0.014; LTV-miR106sp vs. Control, **p = 0.0012. For 12 weeks, LTV-empty vs. LTV-miR106sp, *p = 0.018; LTV-miR106sp vs. Control, **p = 0.0068. B Quantification of 4- and 12-week-old LTV-empty or LTV-miR106sp-infected RTT-neurons and control neurons with nuclear MECP2 immunostaining. n = 200 cells per group in three independent experiments. For 4 weeks, LTV-empty vs. LTV-miR106sp, **p = 0.0075; LTV-miR106sp vs. Control, **p = 0.0046. For 12 weeks, LTV-empty vs. LTV-miR106sp, ***p = 4.77 × 10^-05; LTV-miR106sp vs. Control, ***p = 0.00011. C Representative images of MAP2 (Green) and TubIII (Red) staining in ~4-week-old LTV-empty or LTV-miR106sp-infected RTT- and control neurons. DAPI stains the nucleus Blue (Top). Representative reconstructed neuronal images for the quantitative assessment of all orders of branches in each group by Sholl analysis are shown (Bottom). D Quantitative analysis of the soma cross-sectional area (Top) and the number of neuronal branch points (Bottom) in 4- and 12-week-old MAP2 + RTT neurons expressing empty or miR106sp and control neurons. n = 200 cells per group in three independent experiments. The boxed areas span the first to the third quartile, with the central line representing the median expression changes for each group. Soma size: For 4 weeks, LTV-empty vs. LTV-miR106sp, *p = 0.024; LTV-miR106sp vs. Control, **p = 0.001. For 12 weeks, LTV-empty vs. LTV-miR106sp, ***p = 0.00039; LTV-miR106sp vs. Control, ^nsp = 0.425. Branches density: For 4 weeks, LTV-empty vs. LTV-miR106sp, *p = 0.015; LTV-miR106sp vs. Control, ^nsp = 0.491. For 12 weeks, LTV-empty vs. LTV-miR106sp, *p = 0.015; LTV-miR106sp vs. Control, ^nsp = 0.142. E, F Representative fluorescent intracellular calcium transient in RTT neurons expressing empty or miR106sp or control neurons over 5 min (n = 20; E). The quantitation of calcium spikes for each group is shown (F, Left). The boxed areas span the first to the third quartile, with the central line representing the median expression changes for each group. For calcium event frequency, LTV-empty vs. LTV-miR106sp, ***p = 7.43 × 10^-13; LTV-miR106sp vs. Control, ***p = 1.09 × 10^-11. The percentage of GCaMP6-positive neurons in each group is shown (F, Right). (n = 60 per group) G ChIP analysis monitoring binding of MECP2 to the Actin, DLX (LTV-empty vs. LTV-miR106sp, *p = 0.049), JUNB (LTV-empty vs. LTV-miR106sp, *p = 0.040), and SNRPN (LTV-empty vs. LTV-miR106sp, *p = 0.026) promoter in RTT neurons expressing control or miR106sp. Data is expressed as mean ± SD (A, B, F, G). Unpaired, two-sided t-test with no multiple adjustments. Scale bars: 40 µm (D).