Repression of Xist RNA expression is considered a prerequisite to reversal of X-chromosome inactivation (XCI) in the mouse inner cell mass (ICM), and reactivation of X-linked genes is thought to follow loss of Xist RNA coating and heterochromatic markers of inactivation, such as methylation of histone H3. We analyzed X-chromosome activity in developing ICMs and show that reactivation of gene expression from the inactive-X initiates in the presence of Xist coating and H3K27me3. Furthermore, depletion of Xist RNA coating through forced upregulation of NANOG does not result in altered reactivation kinetics. Taken together, our observations suggest that in the ICM, X-linked gene transcription and Xist coating are uncoupled. These data fundamentally alter our perception of the reactivation process and support the existence of a mechanism to reactivate Xp-linked genes in the ICM that operates independently of loss of Xist RNA and H3K27me3 from the imprinted inactive-X.
INTRODUCTION
X-chromosome inactivation (XCI) results in the transcriptional silencing of most genes on one of the two X-chromosomes to equalize X-linked gene dosage in female cells with that in XY male cells (Avner and Heard, 2001). Although the exact mechanism for X-linked gene silencing is unknown, XCI requires expression of the long non-coding RNA X inactive-specific transcript (Xist) (Brown et al., 1991; Marahrens et al., 1997; Penny et al., 1996). Xist RNA is expressed from and preferentially coats the inactive X-chromosome (Clemson et al., 1996). Xist coating is followed by recruitment of multiple chromatin-modifying complexes, such as the polycomb repressive complex 2 (PRC2), which trimethylates lysine 27 on histone H3 to give H3K27me3 (Plath et al., 2002; Silva et al., 2003). H3K27me3 is a chromatin modification associated with gene repression, and H3K27me3 enrichment on the inactive-X is essential to maintain imprinted paternal-X silencing in extra-embryonic lineages (Kalantry et al., 2006; Simon and Kingston, 2009). Together, Xist coating and H3K27me3 are necessary to establish stable and heritable XCI; thus, these epigenetic modifications are considered to be markers of X-linked gene silencing.
Two types of XCI are evident during development of the female mouse embryo. The first type, imprinted XCI, results in exclusive silencing of the paternal X-chromosome (Takagi and Sasaki, 1975; West et al., 1977). By contrast, the second type of XCI is termed random because either maternal-X (Xm) or paternal-X (Xp) can be chosen for silencing (Lyon, 1961). During early mouse embryogenesis, Xist expression is first evident at the two-cell stage and occurs exclusively from the Xp (Kay et al., 1993). Xist RNA gradually coats the Xp during pre-implantation development (Huynh and Lee, 2003; Okamoto et al., 2004). By the early blastocyst stage, prior to specification of extra-embryonic and embryonic lineages, Xist accumulation on the Xp (Xp-Xist coating) is evident in all cells of the embryo (Sheardown et al., 1997). At the late blastocyst stage, when extra-embryonic and embryonic lineages are determined, Xp-Xist coating is maintained in the extra-embryonic trophectoderm and primitive endoderm. By contrast, Xp-Xist coating is lost in epiblast cells, which define the progenitor cells of the embryo proper (Mak et al., 2004; Okamoto et al., 2004; Sheardown et al., 1997).
Based on patterns of Xp-Xist coating in the blastocyst, it is assumed that all cells of the pre-implantation embryo first undergo the imprinted form of XCI and that this imprint is maintained in the extra-embryonic lineages. In the developing ICM, epiblast precursor cells reverse imprinted XCI in order to subsequently initiate random XCI later in development (Mak et al., 2004; Okamoto et al., 2004). However, three recent reports analyzed Xp-linked gene transcription in single cells of pre-implantation embryos. Each report independently confirmed that some degree of Xp-linked gene silencing is evident as early as the four-cell cleavage stage. Yet, despite Xist coating the Xp in all cells, Xp-linked gene silencing is not evident in all cells of the blastocyst, indicating that imprinted XCI is incomplete during pre-implantation development (Kalantry et al., 2009; Namekawa et al., 2010; Patrat et al., 2009). Given these results, it is still unclear whether epiblast progenitor cells, which are specified by the blastocyst stage, exhibit imprinted Xp-linked gene silencing.
Here, we utilized RNA fluorescence in situ hybridization (FISH) techniques to measure directly the kinetics of X-linked gene expression, as opposed to chromosome-wide markers of inactivation, in individual nuclei of developing ICMs. We provide evidence that supports previous assumptions that, prior to epiblast specification, ICM cells exhibit imprinted XCI. We also make the surprising observation that reactivation in the ICM occurs earlier than previously indicated by loss of Xist coating (Mak et al., 2004; Okamoto et al., 2004). Coincident with specification of the Nanog-expressing epiblast lineage, but prior to loss of the Xist domain and H3K27me3 enrichment on the Xp, mid-stage ICMs exhibit an increase in Xp-linked gene expression, indicative of active transcription. Recent results suggest that pluripotency-associated transcription factors NANOG, OCT3/4 (POU5F1 – Mouse Genome Informatics) and SOX2 might directly regulate X-chromosome activity by preventing upregulation of Xist expression in mouse embryonic stem cells (ES cells) (Navarro et al., 2008). We show that forced upregulation of NANOG results in loss of Xp-Xist coating without affecting the kinetics of Xp-linked gene reactivation. These observations argue against a direct role for NANOG-dependent Xist repression as a requirement for initiation of Xp expression. We present an alternative hypothesis that reversal of imprinted X-chromosome inactivation occurs by separable steps, with gene reactivation being independent from loss of the Xist and H3K27me3 domains.
MATERIALS AND METHODS
Embryo collection and immunosurgery
Embryos were collected from natural mating of either Mus musculus domesticus CD1 females (RNA FISH) or Mus musculus molossinus JF1 females (RT-PCR) to CD1 males carrying the X-GFP (D4.EGFP) transgene (Hadjantonakis et al., 1998). Female blastocysts were distinguished from males by GFP expression. The Grb2−/− (Grb2tm2Paw) line was described previously (Cheng et al., 1998). Mice were exposed to light daily between 06.00 h and 18.00 h. The morning of the day the plug was detected was considered to be embryonic day (E)0.5. Early blastocysts were collected between E2.75 and E3.0 (~66-72 hours post-copulation). Mid-stage blastocysts were collected between E3.5 and E3.75 (~84-90 hours post-copulation). Because embryos collected at the same time varied in developmental stage, blastocysts were also staged according to size (a measure of total cell number) and blastoceol cavity expansion (see Fig. S6 in the supplementary material). Immunosurgery procedures were performed as described previously (Solter and Knowles, 1975).
Immunostaining
Embryos were fixed in 3% paraformaldehyde for 10 minutes at room temperature. After washing in 1×PBS, embryos were permeabilized in 0.5% Triton X-100 for 15 minutes. Embryos were then washed in 1×PBS and blocked in 10% serum for 1 hour at room temperature. Primary antibodies were: Nanog (Chemicon, 1:1000), Gata6 (R&D Systems, 1:1000), H3K27me3 (Upstate, 1:250) and EED (Arie Otte, Slater Institute, The Netherlands, 1:200). Primary antibody incubations were performed overnight at 4°C. Embryos were washed in 1×PBS and incubated in secondary antibody for 45 minutes at room temperature. Secondary antibodies coupled to Alexa Flour 594 (Invitrogen A-11037) and Alexa Fluor 488 (Invitrogen A-11039) were diluted to 1:500. The embryos were washed in 1×PBS and mounted into 0.12 mm deep Secure-Seal spacer wells (Invitrogen, S24737) on microscope slides in Vectashield (Vector Labs) diluted 1:1 with DAPI.
RNA FISH
FISH probes were generated from BAC and FOSMID clones (BACPAC Resources Center, CA, USA) and labeled using the Bioprime Labeling Kit (Invitrogen) with FITC-dUTP (Roche), Cy3-dCTP (GE Healthcare) or Cy5-dCTP (GE Healthcare). FOSMIDs and BACs used were: G135P65743A11 (Ube1x), G135P69277C4 (Mecp2), G135P63425C4 (Xist/Tsix), G135P605237C7 (Rnf12), RP24-274B9 (Abcb7), G135P603627A9 (Smcx), G135P60684B7 (Nanog) and RP24329F11 (Gata6). Two FOSMIDs G135P62497G3 and G135P64951A11 were used to detect Atp7a. Mouse Cot-1 DNA (Invitrogen) was used in the Cot-1 RNA FISH experiments. RNA FISH was performed as described previously (Kalantry et al., 2009). All RNA FISH probes used to detect X-linked genes were tested in trophoblast (TS) and embryonic stem (ES) cell lines to determine detection of X-linked sequences with expected expression patterns. Nanog and Gata6 FISH probes were tested in ES and extra-embryonic endoderm (XEN) cell lines, respectively. For combined immunostaining of H3K27me3 and RNA FISH, RNA FISH was performed after immunostaining.
Statistical analysis of RNA FISH
RNA FISH nuclei were classified as biallelic, monoallelic or nonscorable. Nuclei were classified as nonscorable when no signal could be detected. For each X-linked gene, the total number of the three classes of nuclei was compared between early and mid-stage ICMs. Binomial exact test was used to verify that nonscorable data between stages were equivalent. Fisher's exact test compared the three categories between early and mid-stage ICMs. P values <0.01 were considered to be statistically significant. The procedure was repeated for the Cot-1 RNA FISH experiments except nuclei were scored as either excluded, overlapping or nonscorable.
Microscopy
Images for RNA FISH were obtained using a Leica DML fluorescence microscope with a Q-imaging Retiga 200R camera and Q-capture software. Images were processed using SPOT RT Software (Diagnostic Instruments). For Cot-1 RNA FISH and immunostaining, a Zeiss710 confocal microscope was used. Confocal sections were analyzed using Zen software to determine overlap between signals. To determine relative intensity values obtained in the Cot-1 RNA experiments, we generated line traces across nuclei using NIH Image J software and used the plot profile tool to measure the intensity across the line traces.
RT-PCR
RT-PCR was performed as described previously (Kalantry et al., 2009). In brief, mRNA was prepped using the Dynabeads mRNA DIRECT Micro Kit (Invitrogen, 610.21). ICMs were lysed in 100 μl of lysis/binding buffer. For RT-PCR, SuperScript III One-Step RT-PCR Platinum Taq (Invitrogen, 12574-035) was used to amplify the cDNA. A portion of the final RT-PCR product was used to perform one round of amplification in the presence of a trace amount of 32P-dCTP (Perkin Elmer, BLU513H). The final PCR product was digested with the appropriate restriction enzyme, run on a 10% acrylamide gel and exposed to film. All RT-PCR primers were described previously (Kalantry et al., 2009). Band intensities were quantified using Adobe Photoshop CS3 Extended software. Percent paternal expression was calculated as a percentage of total intensity of combined maternal- and paternal-specific bands, i.e. % Xp = [Xp/(Xm+Xp)]×100. In the case of multiple maternal- or paternal-specific bands, the band intensities were added to yield a single value. The average % Xp was compared between early and mid-stage ICMs. P-values were calculated using two-tailed t-tests where the null hypothesis is that the mean % Xp of early and mid-stage samples are the same.
RESULTS
Imprinted XCI is established in early ICMs prior to epiblast specification
To investigate Xp activity in the ICM, we analyzed transcription of X-linked genes by RNA FISH in ICMs isolated by immunosurgery (Solter and Knowles, 1975). We assayed five X-linked genes [Rnf12 (Rlim – Mouse Genome Informatics), Atp7a, Abcb7, Ube1x (Uba1 – Mouse Genome Informatics) and Mecp2] that are expressed at robust levels during the cleavage stages of pre-implantation development (Kalantry et al., 2009; Patrat et al., 2009). Previous allele-specific RT-PCR and RNA FISH data showed a bias towards Xm expression of these genes at the morula stage, indicating that these genes are subject to imprinted XCI (Huynh and Lee, 2003; Kalantry et al., 2009). We assayed expression of the five X-linked genes together with Xist RNA, which marks the Xp, in ICMs of early stage blastocysts (~32-cell stage embryo) (Fig. 1A). If the Xp-allele of the five genes assayed is silenced, RNA FISH will yield a single monoallelic gene signal indicative of transcription from the Xm. If the Xp-linked gene is not silenced, our assay will detect gene expression from both X-chromosomes. As a positive control, we measured expression of Smcx (Kdm5c – Mouse Genome Informatics), a gene known to escape imprinted XCI (Agulnik et al., 1994; Plusa et al., 2008). As expected, Smcx exhibited 97% biallelic expression in cells of early ICMs, demonstrating that our assay is capable of detecting biallelic X-linked gene transcription (see Fig. S1A in the supplementary material). By contrast, Rnf12, Atp7a, Abcb7, Ube1x and Mecp2 demonstrated an average low level of biallelism, ranging from 5% to 23% of cells per ICM (Fig. 1B). This low level of biallelism is comparable to the level of biallelism in trophectoderm nuclei (see Fig. S1B in the supplementary material). A majority (60%) of all early ICMs analyzed exhibited either one or no biallelic nuclei, and a number of individual ICMs exhibited exclusive (100%) monoallelic Rnf12, Atp7a, Abcb7 and Ube1x expression in every nucleus. Together, these data argue against the possibility that escape from imprinted XCI is a general characteristic of ICM nuclei and suggest that early ICMs undergo Xp silencing to the same degree as do cells of the trophectoderm. Therefore, the majority of nuclei in the pre-implantation embryo exhibit Xp-linked gene silencing and undergo imprinted XCI prior to specification of the embryonic lineage.
Mid-stage inner cell masses (ICMs) exhibit increasing Xp-linked gene expression. (A) A representative ICM assayed for Atp7a (red, left panel) and Xist and Tsix (green, right panel) by RNA fluorescence in situ hybridization (FISH). The Xist probe detects Xist RNA and the anti-sense transcript detects Tsix. Xist marks the Xp as a cloud at this stage. In ICMs, Tsix is detected in some nuclei as a pinpoint signal from the Xm (Lee et al., 1999). In the merge image, all nuclei exhibit monoallelic Atp7a expression, localized away from the Xp-Xist coat (green domains). A single nucleus is highlighted in the right corner of each image. (B) Distributions of biallelic (black), monoallelic (gray) and nonscorable (white) nuclei are presented as the average percentage of nuclei per early or mid-stage ICM. In mid-stage ICMs there is a statistically significant (P<0.01) increase in biallelism and decrease in monoallelism compared with early stage ICMs. (C) A representative nucleus illustrates biallelic Abcb7 (red, first panel) expression. H3K27me3 accumulates on the Xp above genomic levels (cyan, second panel). The H3K27me3 enrichment domain colocalizes with Xp-Xist coating (green cloud, third panel). The merge image (fourth panel) indicates that one of the two Abcb7 RNA FISH signals (marked by an arrowhead) colocalizes with both the Xp Xist and H3K27me3 domains. Nuclei are stained with DAPI.
Mid-stage inner cell masses (ICMs) exhibit increasing Xp-linked gene expression. (A) A representative ICM assayed for Atp7a (red, left panel) and Xist and Tsix (green, right panel) by RNA fluorescence in situ hybridization (FISH). The Xist probe detects Xist RNA and the anti-sense transcript detects Tsix. Xist marks the Xp as a cloud at this stage. In ICMs, Tsix is detected in some nuclei as a pinpoint signal from the Xm (Lee et al., 1999). In the merge image, all nuclei exhibit monoallelic Atp7a expression, localized away from the Xp-Xist coat (green domains). A single nucleus is highlighted in the right corner of each image. (B) Distributions of biallelic (black), monoallelic (gray) and nonscorable (white) nuclei are presented as the average percentage of nuclei per early or mid-stage ICM. In mid-stage ICMs there is a statistically significant (P<0.01) increase in biallelism and decrease in monoallelism compared with early stage ICMs. (C) A representative nucleus illustrates biallelic Abcb7 (red, first panel) expression. H3K27me3 accumulates on the Xp above genomic levels (cyan, second panel). The H3K27me3 enrichment domain colocalizes with Xp-Xist coating (green cloud, third panel). The merge image (fourth panel) indicates that one of the two Abcb7 RNA FISH signals (marked by an arrowhead) colocalizes with both the Xp Xist and H3K27me3 domains. Nuclei are stained with DAPI.
Xp reactivation occurs in the presence of the Xp-Xist domain
Reactivation of the Xp has been described to occur in the epiblast of late (fully expanded or hatched) blastocysts (~128-cell stage embryo) after the sequential loss of Xist coating, the PRC2 subunit EED, and H3K27me3 on the Xp (Mak et al., 2004; Okamoto et al., 2004; Silva et al., 2009). To investigate the timing of Xp reactivation, we isolated ICMs from mid-stage female blastocysts (~32- to 64-cell stage embryo). We first performed dual immunofluorescence to detect NANOG, which marks the ICM, and EED in mid-stage blastocysts. On average, 90% of NANOG positive nuclei exhibited an EED Xp domain (Mak et al., 2004; Okamoto et al., 2004; Silva et al., 2009) (see Fig. S2A in the supplementary material). Xp EED domains in ICM nuclei also exhibited accumulation of H3K27me3 (see Fig. S2B in the supplementary material). As H3K27me3 is a hallmark of transcriptionally inactive chromatin domains, these data imply Xp-linked genes are inactive in mid-stage ICMs. We next investigated the transcriptional status of the Xp-linked genes directly using RNA FISH. Assaying the same five X-linked genes that previously showed predominantly monoallelic expression in early ICMs, we detected biallelic expression in 35-50% of cells from mid-stage ICMs (Fig. 1B). Importantly, in all cases of biallelism one of the two X-linked gene signals colocalized with Xp-Xist and H3K27me3 accumulation, and the percentage of nuclei exhibiting Xp-Xist coating did not change significantly (Fig. 1C; see Fig. S2C in the supplementary material). In contrast to early ICMs, we never observed a mid-stage ICM that exhibited 100% monoallelic nuclei. For all five X-linked genes analyzed, we observed a statistically significant increase in biallelism and a corresponding decrease in monoallelism (P<0.01; Fig. 1B; see Table S1 in the supplementary material). Given the cell division timing (~11.5 hours) of ICM nuclei and the low level of biallelism detected at the early stage, it is also unlikely that escape from imprinted XCI, for any of the genes assayed, can account for the increase in biallelic nuclei (Bischoff et al., 2008). These data suggest that reactivation of genic transcription occurs earlier than determined previously by analysis of inactive-X epigenetic markers and, remarkably, occurs in the presence of Xp-Xist coating and H3K27me3 accumulation on the Xp.
Dynamics of X-linked gene expression assayed by allele-specific RT-PCR. F1 hybrid mouse embryos were generated by mating JF1 females to CD1 males. Xm and Xp transcripts were distinguished utilizing single nucleotide polymorphisms between the two strains. (A) Allele-specific expression was analyzed for Xist and the X-linked genes Ube1x, Rnf12, Pdha1 and Ddx3x. Lane 1, Xm allele; Lane 2, Xp allele; Lane 3, equal amounts of Xm and Xp allele; Lanes 4-6, early F1 hybrid inner cell mass (ICM) samples; Lanes 7-9, mid-stage F1 hybrid ICM samples. Analysis of Xist expression illustrates that in both early and mid-stage ICMs, Xist is expressed from the Xp. In early ICMs, X-linked gene expression for Ube1x, Rnf12, Pdha1 and Ddx3x is largely from the Xm allele. In mid-stage ICMs there is increasing Xp expression of Xp-linked genes. (B) Average Xm and Xp contributions were calculated as a percentage of total expression (Xm+Xp). There is a statistically significant increase in average Xp expression in mid-stage ICMs compared with early stage ICMs for each X-linked gene analyzed (P<0.01). The number of ICMs examined is presented in the table.
Dynamics of X-linked gene expression assayed by allele-specific RT-PCR. F1 hybrid mouse embryos were generated by mating JF1 females to CD1 males. Xm and Xp transcripts were distinguished utilizing single nucleotide polymorphisms between the two strains. (A) Allele-specific expression was analyzed for Xist and the X-linked genes Ube1x, Rnf12, Pdha1 and Ddx3x. Lane 1, Xm allele; Lane 2, Xp allele; Lane 3, equal amounts of Xm and Xp allele; Lanes 4-6, early F1 hybrid inner cell mass (ICM) samples; Lanes 7-9, mid-stage F1 hybrid ICM samples. Analysis of Xist expression illustrates that in both early and mid-stage ICMs, Xist is expressed from the Xp. In early ICMs, X-linked gene expression for Ube1x, Rnf12, Pdha1 and Ddx3x is largely from the Xm allele. In mid-stage ICMs there is increasing Xp expression of Xp-linked genes. (B) Average Xm and Xp contributions were calculated as a percentage of total expression (Xm+Xp). There is a statistically significant increase in average Xp expression in mid-stage ICMs compared with early stage ICMs for each X-linked gene analyzed (P<0.01). The number of ICMs examined is presented in the table.
To confirm our RNA FISH results, we analyzed allele-specific X-linked gene expression using RT-PCR in individual ICMs obtained from crosses between M. domesticus (CD1) and M. molossinus (JF1) strains. We first measured expression of Xist RNA. In early F1 ICMs, Xist transcription was detectable exclusively from the Xp (Fig. 2A). At mid-stage, Xist transcription was maintained on the Xp (Fig. 2A) consistent with Xp-Xist coating, which was evident by RNA FISH (Fig. 1C). We next quantified the expression of maternal and paternal alleles for four X-linked genes (Ube1x, Rnf12, Ddx3x and Pdha1), which were previously shown to exhibit imprinted Xp silencing in extra-embryonic tissues (Kalantry et al., 2009). All four X-linked genes exhibited biased Xm expression in early ICMs (Fig. 2A). For Ube1x and Rnf12, very little Xp expression could be detected in early ICMs. Paternal silencing of Pdha1 and Ddx3x was less complete but still favored expression from the Xm. These data agree with our RNA FISH results and verify that imprinted XCI is present in early stage ICMs. By contrast, when individual mid-stage ICMs were analyzed, robust expression from the paternal allele was detected (Fig. 2A). Ube1x, Rnf12, Pdha1 and Ddx3x all exhibited a statistically significant increase in paternal X-linked gene expression in mid-stage compared with early ICMs (P<0.05; Fig. 2B; see Table S2 in the supplementary material). These data illustrate the detection of not only nascent transcripts but also spliced polyadenylated transcripts, confirming an increase in Xp-linked gene expression in mid-stage ICMs.
Together, our RNA FISH and allele-specific RT-PCR results show that reactivation of seven Xp-linked genes occurs in the presence of Xp-Xist coating. In order to determine the broader transcriptional status of the Xp, we next hybridized with a Cot-1 DNA probe and Xist probe in our RNA FISH experiments. Under RNA FISH conditions, Cot-1 DNA hybridizes to transcribed repetitive sequences of the genome (Hall et al., 2002). In combination with a Xist probe, Cot-1 RNA FISH allows visualization of general transcription within the Xist RNA domain. It has previously been shown that the X-chromosome is devoid of Cot-1 hybridization when gene silencing is established (Chaumeil et al., 2006; Namekawa et al., 2010). Consistent with predominant Xp-silencing, we detected mutually exclusive Xist and Cot-1 RNA FISH signals in the majority (~82%) of trophectoderm nuclei of mid-stage blastocysts (Fig. 3A). By contrast, when we performed RNA FISH on mid-stage ICMs, we detected nuclei with overlap of Cot-1 and Xist RNA FISH signals (Fig. 3B; see Fig. S3 in the supplementary material). Cot-1 RNA FISH signal typically overlapped the exterior of the intense Xist domain, although nuclei were also detected that exhibited an Xp-Xist domain that completely overlapped with the Cot-1 RNA FISH signal (see Fig. S3 in the supplementary material). We quantified the number of nuclei per mid-stage ICM that exhibited patterns of mutually exclusive or overlapping RNA FISH signals (Fig. 3A). On average, 45% of nuclei per mid-stage ICM exhibited overlapping signals of Cot-1 and Xist. Compared with the trophectoderm, the ICM is, therefore, enriched with nuclei exhibiting Cot-1 signal overlapping the Xp-Xist domain (Fig. 3A). This pattern of overlap is not simply characteristic of ICM cells, in general, as we detected a significant (P<0.01) increase in the number of nuclei with overlap when we compared early to mid-ICMs (Fig. 3A). This increase in the number of nuclei exhibiting transcription from within the Xist RNA domain is consistent with general Xp reactivation occurring in the presence of Xist coating. In combination with our previous gene-by-gene RNA FISH (Fig. 1B) and allele-specific RT-PCR results (Fig. 2B), these data suggest that the presence of Xist coating and the accumulation of PRC2 and H3K27me3 are not sufficient for X-chromosome silencing.
Mid-stage inner cell masses (ICMs) exhibit transcription within the Xp-Xist domain. Cot-1 DNA (red) and Xist (green) probes were used in RNA FISH experiments to determine general Xp transcriptional status. (A) The average distribution of nuclei with mutually excluded or overlapping Cot-1 and Xist RNA FISH signals are shown as a percentage of total nuclei per mid-stage trophectoderm (TE), early ICM and mid-stage ICM. The bracket represents a statistically significant (P<XXXX) increase in the percentage of cells exhibiting overlapping FISH signals in mid-ICMs compared with early ICMs. (B) Representative mid-stage ICM nuclei indicate mutually excluded or overlapping patterns. Arrowheads indicate the Xist domain. Dashed lines indicate the area of the merged figure represented in the line scans. Line scans indicate the relative intensities of DAPI, Cot-1 and Xist signals.
Mid-stage inner cell masses (ICMs) exhibit transcription within the Xp-Xist domain. Cot-1 DNA (red) and Xist (green) probes were used in RNA FISH experiments to determine general Xp transcriptional status. (A) The average distribution of nuclei with mutually excluded or overlapping Cot-1 and Xist RNA FISH signals are shown as a percentage of total nuclei per mid-stage trophectoderm (TE), early ICM and mid-stage ICM. The bracket represents a statistically significant (P<XXXX) increase in the percentage of cells exhibiting overlapping FISH signals in mid-ICMs compared with early ICMs. (B) Representative mid-stage ICM nuclei indicate mutually excluded or overlapping patterns. Arrowheads indicate the Xist domain. Dashed lines indicate the area of the merged figure represented in the line scans. Line scans indicate the relative intensities of DAPI, Cot-1 and Xist signals.
The mid-stage ICM is composed of cells specified to both epiblast and primitive endoderm (Chazaud et al., 2006). Thus, we next investigated whether the biallelic gene expression evident in mid-stage ICMs corresponded to the epiblast cell fate. Cell fate in the maturing ICM is adopted from mutually exclusive expression of lineage markers Nanog (epiblast) and Gata6 (primitive endoderm), which is resolved from uniform expression of these markers in the early ICM. We analyzed NANOG and GATA6 in mid-stage blastocysts (see Fig. S4A in the supplementary material). As previously reported, at mid-stage NANOG and GATA6 were not expressed mutually exclusively in the ICM (Plusa et al., 2008). However, we reasoned that changes in protein products might lag behind changes in gene expression; consequently, analysis of primary transcripts by RNA FISH might confer a more accurate expression pattern of Nanog and Gata6 to be utilized for lineage markers. We found that Nanog and Gata6 nascent transcripts in mid-stage ICMs were largely detected as mutually exclusive cell populations (see Fig. S4B in the supplementary material). On average, ~47% of cells per ICM exhibited exclusive Nanog expression. By contrast, ~32% of cells exhibited exclusive Gata6 expression. On average, 21% of cells per ICM co-expressed both lineage markers (see Fig. S4B in the supplementary material). The degree of mutually exclusive lineage marker expression appears to depend upon the maturity of the blastocyst with more developed blastocysts exhibiting fewer cells with co-expression. These data correspond to predicted patterns of Nanog and Gata6 expression, based primarily on protein localization at later stages of ICM development, and validate the use of RNA FISH to measure Nanog expression in mid-stage ICMs (Plusa et al., 2008).
Restriction of Nanog expression to epiblast progenitors appeared to be more complete in further developed mid-stage blastocysts. Therefore, in the following experiments, we limited our assays to isolated ICMs from blastocysts of ~58-64 cells. In these ICMs, we assayed Nanog expression in cells that exhibited biallelic expression of X-linked genes Atp7a and Mecp2. We found that most cells expressing Atp7a and Mecp2 from both X-chromosomes also expressed Nanog (Fig. 4A; see Fig. S4C in the supplementary material). The reverse was true for Gata6 expression, which, when assayed in independent experiments, was rarely detected in cells exhibiting biallelic X-linked gene expression (see Fig. S4D in the supplementary material). We also show that in biallelic mid-stage Nanog-positive cells, Xist clearly coated the Xp (Fig. 4B,C). Utilizing Nanog as an epiblast specific marker, we located the cells exhibiting biallelic X-linked gene expression to the epiblast progenitor population. Biallelic gene expression is, therefore, biased to the epiblast cell fate and not simply a function of incomplete Xp silencing during pre-implantation development. We conclude that Xp expression in mid-stage ICMs is a result of reversing imprinted XCI.
Ectopic extinction of the Xist coat has no affect on Xp-linked gene reactivation
The prevailing hypothesis suggests that reactivation of X-linked genes in the ICM results from Xist RNA downregulation. These data are supported by evidence that Xist RNA coating is lost in the late stage ICM (Mak et al., 2004; Okamoto et al., 2004). However, the data presented thus far argue against loss of the Xist coat as a prerequisite to initiate Xp genic reactivation. One possibility that might explain reactivation of X-linked genes in the presence of Xist coating is that Xist expression gradually decreases as the ICM develops. If so, nuclei exhibiting biallelic X-linked gene expression might represent nuclei with lower levels of Xist coating. Thus, Xp-linked gene reactivation could be triggered by decreasing amounts of Xist coating, which would support the previous hypothesis that loss of Xist triggers reactivation. Such subtle decreases in amount of Xist transcripts coating the Xp might also be undetectable by RNA FISH. In order to address this possibility, we sought to deplete Xist coating in mid-stage ICMs. We reasoned that if Xp gene expression were regulated independently of Xist coating then precocious loss of the Xp-Xist domain would have no effect on the kinetics of genic reactivation. By contrast, if decreasing levels of Xist trigger reactivation, then dramatic loss of Xist should result in reactivation of the Xp.
For these experiments, we took advantage of a genetic mutation in Grb2, a key adaptor molecule necessary for MAP kinase signal transduction, which has been shown to regulate Nanog expression (Hamazaki et al., 2006; Pawson and Scott, 1997). NANOG expression was highly variable in the mid-stage ICM, with strong NANOG immunostaining detectable in only a subset of ICM cells (Fig. 5A; see Fig. S4A in the supplementary material). By contrast, Grb2−/− blastocysts exhibited strong NANOG immunostaining in the majority of ICM cells (Fig. 5A) (Chazaud et al., 2006). It has been reported that NANOG directly represses Xist expression in undifferentiated mouse ES cells (Navarro et al., 2008). In support of these results, upregulation of Nanog due to inhibition of MAPK/ERK signaling resulted in loss of the Xp EED domain in all ICM cells at E4.5 (Nichols et al., 2009). We also observed that female mid-stage E3.5 Grb2−/− blastocysts exhibited uniform loss of EED and H3K27me3 Xp enrichment specifically in the ICM (Fig. 5A; see Fig. S5 in the supplementary material). Trophectoderm cells appeared to maintain both inactive-X markers in wild-type and Grb2−/− blastocysts (Fig. 5A; see Fig. S5 in the supplementary material). We also found a significant reduction in the number of ICM cells with Xp-Xist coating in Grb2−/− ICMs compared with littermate controls (Fig. 5B). This was in contrast to Grb2−/− embryos isolated between the 8- and 16-cell stage, which exhibited normal initiation of Xp silencing and Xp-Xist coating. In addition, we noted that even when Xp-Xist coating was detectable in Grb2−/− mid-stage ICMs, the density of the coat was significantly reduced (Fig. 5C). Importantly, when we analyzed Xp-linked gene activity, we found similar patterns of monoallelic and biallelic Xp-linked gene expression in female Grb2−/− ICMs compared with female littermate controls, irrespective of the amount of Xist coating the Xp (Fig. 5C). In fact, for the four genes we analyzed (Ube1x, Rnf12, Atp7a and Abcb7) monoallelic X-linked gene expression was detectable in cells without visible Xp-Xist coating (Fig. 5B,C). Despite widespread loss of Xp-Xist and Xp-EED accumulation, reactivation in Grb2−/− ICMs occurs normally. Analysis of ICM nuclei from late-stage blastocysts (~128 cells) also indicates that in the absence of Xist coating, Xp silencing can be maintained through cell divisions (data not shown). These data rule out the possibility that the reactivation detected in the mid-stage ICM is due to subtle decreases in the levels of Xist and show that genic silencing in the ICM can be maintained independently of cytologically detectable Xist coating.
Nanog expression corresponds to biallelic X-linked gene expression in mid-stage inner cell masses (ICMs). (A) Side-by-side box plots graphically represent the percentage of biallelic nuclei (y-axis) grouped as either positive or negative for Nanog transcripts (x-axis). The box plot on the left represents ICMs assayed for Atp7a expression and the right-hand box plot represents ICMs assayed for Mecp2 expression. The medians of the data are indicated (green triangle). The boxes contain the middle 50% of the data. Whiskers represent maximum and minimum values of the dataset. The box plots are nonoverlapping, illustrating that the population of biallelic nuclei are largely skewed towards being Nanog positive. (B) Representative nuclei of a mid-stage ICM showing detection of Atp7a (red), Xist RNA (blue) and Nanog (green) assayed by RNA FISH. Top row: Biallelic Atp7a (red) expression is marked by arrowheads (first panel) with one allele overlapping Xist transcripts (blue, second panel). Nanog transcripts (green) are marked by arrows (third panel). Bottom row: Monoallelic Atp7a expression is marked by an arrowhead (first panel) and colocalizes with Tsix expression (blue pinpoint signal, second panel) from the Xm. In this nucleus, Nanog is not expressed (third panel). (C) The graph represents the distribution of the percentage of nuclei from mid-stage Nanog positive nuclei that exhibit Xp-Xist coating. The red line represents the average. n, the number of embryos analyzed.
Nanog expression corresponds to biallelic X-linked gene expression in mid-stage inner cell masses (ICMs). (A) Side-by-side box plots graphically represent the percentage of biallelic nuclei (y-axis) grouped as either positive or negative for Nanog transcripts (x-axis). The box plot on the left represents ICMs assayed for Atp7a expression and the right-hand box plot represents ICMs assayed for Mecp2 expression. The medians of the data are indicated (green triangle). The boxes contain the middle 50% of the data. Whiskers represent maximum and minimum values of the dataset. The box plots are nonoverlapping, illustrating that the population of biallelic nuclei are largely skewed towards being Nanog positive. (B) Representative nuclei of a mid-stage ICM showing detection of Atp7a (red), Xist RNA (blue) and Nanog (green) assayed by RNA FISH. Top row: Biallelic Atp7a (red) expression is marked by arrowheads (first panel) with one allele overlapping Xist transcripts (blue, second panel). Nanog transcripts (green) are marked by arrows (third panel). Bottom row: Monoallelic Atp7a expression is marked by an arrowhead (first panel) and colocalizes with Tsix expression (blue pinpoint signal, second panel) from the Xm. In this nucleus, Nanog is not expressed (third panel). (C) The graph represents the distribution of the percentage of nuclei from mid-stage Nanog positive nuclei that exhibit Xp-Xist coating. The red line represents the average. n, the number of embryos analyzed.
DISCUSSION
Here, we present evidence that reversal of imprinted XCI, as evident by an increase in Xp-linked gene activity, initiates in the mid-stage ICM coincident with specification of the epiblast progenitor population. Genic reactivation occurs in the presence of Xist coating, as well as the cytologically detectable PRC2 and H3K27me3 domains. Together, these observations suggest that, particularly at this early stage of embryonic development, these epigenetic marks, classically analyzed at the cytological level to generalize X-chromosome activity, do not accurately reflect the expression state of individual X-linked genes.
Grb2−/− inner cell mass (ICM) cells exhibit loss of Xist transcripts and EED on the Xp. (A) Mid-stage blastocysts immunostained for NANOG (red, top row) and EED (green, second row). NANOG marks the epiblast progenitors. In wild-type blastocysts (left panels), EED accumulates on the Xp in NANOG positive nuclei as seen by the green foci (merge, third row). However, in all but one NANOG positive nuclei of Grb2−/− blastocysts (right-hand panels), Xp EED foci are undetectable (merge, third row). DAPI (blue, fourth row) stains all nuclei. (B) Wild-type (WT, top row) and Grb2−/− (bottom row) mid-stage ICMs assayed by RNA FISH for transcripts of Xist/Tsix (green, left-hand panels) and Ube1x (red, middle panels). Grb2−/− ICMs exhibit a decrease in Xp-Xist coating (green, left panel). A representative nucleus is highlighted and an arrowhead marks monoallelic Ube1x expression, which occurs despite the absence of Xp-Xist coating. The graph indicates the average percentage of cells per ICM that exhibit Xist coating in Grb2−/− compared with wild-type littermate controls. Error bars represent xxx. (C) Nuclei from wild-type (upper rows) and Grb2−/− (bottom rows) ICMs assayed for transcripts of X-linked genes Rnf12, Atp7a and Abcb7 (red signals marked by arrowheads, left panels) and Xist/Tsix (green, middle panels). Grb2−/− ICM nuclei (bottom rows) show monoallelic expression in the absence of the Xp-Xist domain (merge, right panels). The graph indicates the distribution of biallelic X-linked gene expression in Grb2−/− ICMs compared with wild type. The data are presented as a percentage of total cells per ICM. Error bars represent s.d.
Grb2−/− inner cell mass (ICM) cells exhibit loss of Xist transcripts and EED on the Xp. (A) Mid-stage blastocysts immunostained for NANOG (red, top row) and EED (green, second row). NANOG marks the epiblast progenitors. In wild-type blastocysts (left panels), EED accumulates on the Xp in NANOG positive nuclei as seen by the green foci (merge, third row). However, in all but one NANOG positive nuclei of Grb2−/− blastocysts (right-hand panels), Xp EED foci are undetectable (merge, third row). DAPI (blue, fourth row) stains all nuclei. (B) Wild-type (WT, top row) and Grb2−/− (bottom row) mid-stage ICMs assayed by RNA FISH for transcripts of Xist/Tsix (green, left-hand panels) and Ube1x (red, middle panels). Grb2−/− ICMs exhibit a decrease in Xp-Xist coating (green, left panel). A representative nucleus is highlighted and an arrowhead marks monoallelic Ube1x expression, which occurs despite the absence of Xp-Xist coating. The graph indicates the average percentage of cells per ICM that exhibit Xist coating in Grb2−/− compared with wild-type littermate controls. Error bars represent xxx. (C) Nuclei from wild-type (upper rows) and Grb2−/− (bottom rows) ICMs assayed for transcripts of X-linked genes Rnf12, Atp7a and Abcb7 (red signals marked by arrowheads, left panels) and Xist/Tsix (green, middle panels). Grb2−/− ICM nuclei (bottom rows) show monoallelic expression in the absence of the Xp-Xist domain (merge, right panels). The graph indicates the distribution of biallelic X-linked gene expression in Grb2−/− ICMs compared with wild type. The data are presented as a percentage of total cells per ICM. Error bars represent s.d.
In addition, we argue that genic reactivation is not simply a consequence of loss of Xist coating. We present evidence that X-linked gene silencing, at least for a subset of genes, can be maintained even after Xist coating and PRC2 accumulation are removed from the Xp. These data appear to be fundamentally at odds with the prevailing mechanism for X-chromosome reactivation, which suggests that reactivation initiates owing to loss of these epigenetic modifications.
It has long been thought that gene silencing occurs first at X-linked genes located nearest the Xist locus, extends linearly to distant regions of the X-chromosome, and depends on the spread of Xist coating (Huynh and Lee, 2003). The ability to analyze imprinted Xp silencing at the level of single cells has now revealed diverse silencing kinetics between genes and no correlation between the proximity to Xist and the speed or efficiency of silencing could be made (Patrat et al., 2009). Although our study analyzes a subset of X-linked genes, the proximity to Xist had no consequence in the propensity to reactivate. The seven X-linked genes analyzed were chosen at random and are located along the length of the X-chromosome, yet all genes exhibited reactivation in the presence of Xist or PRC2 accumulation. We argue it is unlikely that these genes represent a subset of X-linked genes that exhibit a propensity to reactivate.
Recent data have shown that Xist transcripts first concentrate near the Xist locus, located within the X-chromosome inactivation center (XIC), and then coating spreads to the rest of the chromosome (Sarma et al., 2010); however, how spreading occurs is poorly understood. It is possible that Xist does not coat the entire X-chromosome in a linear fashion. Xist transcripts might favor certain DNA sequences defined as booster elements (Gartler and Riggs, 1983). Xist transcripts can associate with autosomal DNA sequences, suggesting that Xist coating is not dependent on unique X-linked gene sequences. This led to the hypothesis that repetitive sequences might function as booster elements (Lyon, 1998). A possible explanation for ICM reactivation in the presence of Xist coating is that, even though Xist, PRC2 and H3K27me3 are detected cytologically accumulating on the Xp, the spread of these epigenetic modifications could be incomplete. In both early and mid-stage ICMs, Xist coating and PRC2 and H3K27me3 might not have reached all genic loci. In the absence of epigenetic modifications at individual loci, there exists no block to prevent gene specific reactivation from occurring prior to the chromosome-wide removal of epigenetic modifications. Owing to the small number of cells and heterogeneity in the developing ICM, we are limited to cytological detection of these epigenetic modifications, which is low resolution compared with the level of DNA sequence. In the future, analysis of Xist coating and chromatin modifications at high resolution might allow us to determine how these modifications spread to genic loci during pre-implantation development.
The inactive-X in interphase nuclei is organized into a unique heterochromatic structure referred to as the Barr body (Barr and Carr, 1962). The Barr body can be detected cytologically as a RNA polymerase II-depleted domain, with X-linked genes positioned on the outer rim of an interior core composed primarily of repetitive non-coding sequences (Chaumeil et al., 2006; Clemson et al., 2006). Examination of XCI kinetics suggests that formation of the silent, repetitive non-coding core occurs prior to X-linked gene silencing (Chaumeil et al., 2006; Namekawa et al., 2010). Thus, it has been proposed that repression of repetitive non-coding sequences might contribute to X-linked gene silencing by facilitating packaging of the inactive-X into the RNA polymerase II-depleted Barr body. Formation of the silent repetitive non-coding core also appears to coincide with Xist RNA spreading (Chaumeil et al., 2006). However, recent data suggest that during imprinted XCI, repetitive element silencing is initiated independently of Xist coating (Namekawa et al., 2010). Based on our observations that Xp reactivation is uncoupled from Xist coating, we propose that Xp-linked genes reactivate as a consequence of diminished repeat element silencing, which results in breakdown of the silent inactive-X core and inclusion of transcriptional machinery to the imprinted Xp. In support of our hypothesis, we report that mid-stage ICMs are enriched, relative to trophectoderm and early ICMs, with nuclei exhibiting Cot-1 hybridization internal to the Xist domain, suggesting repetitive sequences, within both coding and non-coding sequences, exhibit reactivation prior to removal of Xist coating. These data lead us to hypothesize that during pre-implantation development, imprinted Xp-linked gene silencing might be more dependent on general exclusion of the transcriptional machinery from the Xp than on recruitment of Xist-dependent heterochromatin modifications.
In conclusion, we have presented a detailed analysis of X-chromosome reactivation kinetics in vivo. Taken together, our data lead us to propose a novel explanation for the reversibility of imprinted XCI in the ICM. We suggest that reprogramming requires two separable events. The first step, relief of Xp-linked gene repression, occurs concomitant with loss of repeat element silencing, between the early and mid-stage of ICM development, and results in two active Xs at the mid-stage. The second step, extinction of Xp-Xist coating, occurs in late-stage ICMs by reprogramming of the Xp-Xist locus to suppress Xp-Xist expression. Together, Xp reactivation and Xist reprogramming reset both X-chromosomes for initiation of random XCI in the epiblast. Interestingly, Xist-independent imprinted XCI is evident in marsupials in both extra-embryonic and embryonic tissues (Duret et al., 2006; Rens et al., 2010). By separating Xp genic reactivation from Xist coating, our data highlight Xist-independent mechanisms also occurring in the mouse embryo. How these mechanisms affect characteristics of the inactive-X, such as condensed chromosome structure, depletion of transcriptional machinery, and nuclear organization, to cooperate with Xist in mediating X-linked gene silencing is an important question for future studies.
Acknowledgements
We thank members of the Magnuson Lab for critically reading the manuscript. This work was supported by a developmental biology training grant graduate fellowship awarded to L.H.W., a K99 postdoctoral award to S.K. and a R01 award to T.M., all from NICHD. Deposited in PMC for release after 12 months.
References
Competing interests statement
The authors declare no competing financial interests.