Early investigators discovered an X-linked locus, the X chromosome inactivation center (Xic), required to trigger X chromosome inactivation [10,11,12]. Amazingly, this center is enriched with lncRNAs, most of them functioning in XCI (Figure 1A). The best-studied and most important of these is the 17Kb X-inactivation specific transcript, known as Xist, which is expressed exclusively from the inactivated X-chromosome and essential for the establishment of XCI in early development [13,14,15]. XCI is arbitrarily divided into three stages. At the pre-XCI stage, Xist is expressed at low levels from two active X-chromosomes. During the initiation stage of XCI, Xist is upregulated and begins to “spread” along one of the X-chromosomes and converts that allele into heterochromatin characterized by: (1) genome-wide loss of euchromatic marks; (2) gain of heterochromatic marks; (3) increased DNA methylation; and (4) silencing of gene expression. This very dynamic stage requires Xist. Upon establishment of XCI, the inactive X remains silenced throughout subsequent cell divisions. During this maintenance stage, Xist remains highly expressed from the inactive X. However, deletion of Xist RNA does not result in chromosome-wide gene reactivation [16,17,18]. Therefore, Xist RNA is not required at this stage of XCI [19].

Like proteins, lncRNA also exhibits functional domains. Xist RNA contains six different repeat regions (A to F) (Figure 1B). The most well studied is the repeat A region located at the Xist RNA 5’-end [20]. Transgene studies suggest that Xist lacking the repeat A region cannot initiate gene silencing [20]. The function of that region was subsequently investigated in vivo using a gene targeting approach. Surprisingly, deletion of the A-repeats altered regulation of the mutant Xist allele such that it was not expressed [21], making analysis of that deficiency difficult in vivo. Ex vivo studies using mouse embryonic stem cells (mESCs) revealed at least two independent mechanisms for the silencing function of the A-repeats. One study suggests that the region is required for proper splicing of Xist RNA by directly interacting with splicing factor ASF/SF2 [22]. In another study, we discovered that a shorter transcript within the Xist locus is transcribed through the RepA region in mESCs and we named it RepA RNA [23]. Knockdown of that RepA transcript drastically decreased Xist levels, similar to the phenotype observed following in vivo deletion of the repeat A region of Xist, suggesting that RepA RNA is an important regulator of Xist expression. We found that both Xist and RepA RNA can bind the chromatin repressor Polycomb repressive complex 2 (PRC2) through repeat A region and such interactions are required for the establishment of the chromosome-wide H3K27-3me heterochromatic mark during the initiation of XCI.

Several laboratories have analyzed the structure of the repeat A domain in order to understand the molecular basis for RNA-protein interaction. The A-repeat region contains 7.5–8.5 tandem repeats (variable among different species) of a conserved ~26-mer sequence. Computational analysis indicates that each repeat has a double stem loop structure [20]. However, NMR studies of an in vitro transcribed 26-mer showed that only the first predicted hairpin is formed, while the second predicted hairpin mediates duplex formation among different repeats [24,25]. One limitation of NMR studies is that only a short sequence, such as one of the A repeats, can be analyzed. By carrying out a FRET experiment using chemical and enzymatic probes, another study examined the structure of the whole domain and found that the A region contains two long stem-loop structures, each including four repeats [26]. These studies highlighted the importance of RNA’s structure for its function.

Unlike A repeats, the structure and function of the other Xist repeats are not well understood. Nevertheless, Repeat regions C and F have shown to regulate Xist RNA localization, potentially through interacting with the transcription factor YY1 [20,27,28,29].

Since Xist upregulation is crucial for XCI, much effort has been made to elucidate cis-elements and trans-acting factors that control its expression. The following sections describe how lncRNAs and pluripotency factors regulate Xist expression.

The most extraordinary feature of Xist is its ability to “coat” almost an entire X-chromosome. How Xist RNAs coat and spread to inactivate the X chromosome remains an open question. Interestingly, naturally occurring or induced X:autosome translocations show different degrees of XCI spread from the X into the autosomal DNA segment, suggesting that specific sequences facilitate spreading [52,53]. These observations prompted investigators to look for potential differences in X-chromosome and autosome sequences. LINE (long interspersed elements) are candidates for mediating this effect due to their higher density on the X chromosome compared to autosomes [54]. The theory proposed by Mary Lyon stated that interspersed repetitive LINE elements act as booster elements to promote spread of Xist RNA [54]. A transgenic study in mESCs showed that chromosome regions with higher LINE density are inactivated more efficiently by a Xist transgene [53]. A recent study suggests that silenced LINE elements contribute to formation of a heterochromatic compartment initiated by Xist RNA and that a particular type of active LINEs may participate in local propagation of the XCI into regions that would otherwise escape [55]. Nonetheless, the exact function of LINE elements in XCI remains to be studied.

In terms of trans-acting factors hnRNPU (also known as SAF-A), a known nuclear scaffold protein, is known to be enriched on the inactive X chromosome [56] and to function in Xist RNA loading [57,58]. A recent study showed that YY1, a RNA/DNA binding protein, tethers Xist RNA to the inactive X [27]. However, since YY1 binding sites are highly abundant throughout the mammalian genome, it is unclear where the specificity of YY1-specific guidance of Xist onto the X-chromosome rather than autosomes lies. Factors mediating this activity remain to be identified.

Xist RNA accumulation on Xi leads to chromatin changes, such as DNA hypermethylation, enrichment of macroH2A, loss of the euchromatic mark H3K4-3me and chromosome-wide establishment of heterochromatic mark H3K27-3me and mono ubiquinated H2A (H2AK119u1) [19,59,60,61,62,63]. How can Xist lead to these global changes? Since Xist coats an entire X, it has long been postulated that Xist RNA binds and carries silencing factors, which are deposited as it coats regions during XCI. Thus, much effort has been applied to identifying Xist-interacting proteins.

The best-studied Xist-interacting trans-factors are Polycomb group proteins. These proteins are highly enriched on the inactive X-chromosome during XCI establishment [60,61,64]. Deletion of some Polycomb proteins, such as Eed, leads to reactivation of the silenced X in extra-embryonic tissues, highlighting their essential role in XCI [65]. However, Polycomb proteins do not seem to affect random X-inactivation in embryos and mESCs, suggesting the existence of a functionally redundant mechanism regulating XCI [66]. Polycomb proteins form two major complexes. One of those, Polycomb Repressive Complex 2 or PRC2, is a histone methyltransferase that trimethylates H3 lysine 27 [67]. Upon recognizing H3K27-3me, another complex, PRC1, is recruited to specific genomic loci to establish ubiquinated histone H2A (H2AK119u1) and compact chromatin for gene silencing. Studies using inducible transgenes showed that PRC2 and PRC1 are localized to Xi in a Xist RNA-dependent manner [68,69]. As discussed earlier, it has been suggested that PRC2 is recruited onto Xi by directly interacting with Xist/RepA lncRNAs via repeat A region [23,26,70]. Furthermore, phosphorylation of Ezh2, the catalytic subunit of PRC2, increases PRC2’s RNA binding ability [71]. Notably, a Xist mutant lacking the A-repeats retains the ability to recruit PRC2 and PRC1 to Xi [60,69], indicating that other Xist sequences can recruit Polycomb proteins directly or indirectly. Interestingly, in cells with the depletion of a key component of PRC2, some PRC1 proteins can localize to induced Xist RNA while others cannot, suggesting Xist recruits PRC1 by both PRC2 dependent and independent mechanisms [69]. The PRC2 independent mechanism is supported by a study showing PRC1 can find its genomic targets in PRC2-deficient cells through the interaction with RYBP protein [72]. Surprisingly, trithorax proteins, such as Ash2L, which typically antagonize Polycomb function and are primarily linked to gene activation, also bind Xist, suggesting that either Trithorax proteins also function in gene repression or that Xist RNA can activate specific genes [58].

Studies of Xist RNA in the past two decades have greatly advanced our understanding of how lncRNAs regulate gene expression epigenetically. However, many questions remain unanswered. For example, it is unclear how Xist specifically spreads along the future Xi but not along the activated X. It is also not yet understood how Xist RNA interacts with so many different proteins and what determines the specificity of those interactions. And since Polycomb proteins do not function in establishment of XCI in embryos, the mechanisms used to establish the repressive state of the Xi are not yet identified.

Gtl2 was discovered following an insertional mutagenesis gene trap screen to isolate genes differentially regulated during mouse embryonic development [78]. The gene trap insertion at a region 2.3 kb upstream from Gtl2 promoter caused dwarfism in mouse offspring only through paternal inheritance, not via maternal transmission [78,79]. Subsequent isolation of the Gtl2 gene suggested that its product was a non-coding RNA (ncRNA) based on the observation that the Gtl2 contained no significant open reading frames [79]. Gtl2 was mapped to a region of mouse chromosome 12 (12qF1) thought to contain imprinted genes, due to the observation that in that region uniparental disomy led to embryonic lethality [80]. The paternal-specific growth defect emerging from the insertional mutation plus observation of decreased Gtl2 expression levels in parthenogenetic embryos led to the hypothesis that Gtl2 is paternally imprinted [79]. However, soon after, several independent studies demonstrated that Gtl2 is in fact expressed exclusively from the maternal allele [81,82,83]. The growth phenotype observed following the insertional mutagenesis screen was later suggested to result from perturbations in imprinting of the whole locus, particularly the down regulation of several paternally expressed protein coding genes [84,85].

Not long after the discovery of Gtl2, other imprinted genes that form a cluster and are co-regulated with Gtl2 were also identified [81,83,86,87,88,89]. Interestingly, genes preferentially expressed from the paternally-inherited chromosome are all protein-coding genes, including Dlk1, Rtl1, and Dio3 (Figure 2), whereas genes expressed maternally all encode ncRNAs, namely, Gtl2, Anti-Rtl1, Rian, Mirg, and a large cluster of multiple snoRNAs/microRNAs (Figure 2). All maternally-expressed ncRNAs in this locus are transcribed in the same orientation. This combined with tissue-specific expression of those ncRNAs suggested coordinated expression of the maternally-expressed genes at this locus [90]. It has also been suggested that all of those ncRNAs, with the exception of Gtl2, may come from a long polycistronic transcript [90]. Imprinting is typically regulated by a cis-element called the differentially methylated region (DMR). Three DMRs, located between Dlk1 and Gtl2 (hence intergenic DMR or IG-DMR), at the Gtl2 promoter, and downstream of Dlk1 have been identified [91]. While IG-DMR has been shown to regulate the imprinting status of all genes at this locus, the function of the other DMRs remains poorly understood.

lncRNAs employ diverse mechanisms to regulate gene expression at both the transcriptional and post-transcriptional level (reviewed in [104]). Here, we assess possible molecular mechanisms used by Gtl2 to regulate genes within the Dlk1-Dio3 locus. The modes of regulation considered are not mutually exclusive: Gtl2 function might be exerted through combinatorial mechanisms or carried out differentially in response to specific environmental cues.