The Role of Long Non-coding RNAs in Human Imprinting Disorders: Prospective Therapeutic Targets

Genomic imprinting is a term used for an intergenerational epigenetic inheritance and involves a subset of genes expressed in a parent-of-origin-dependent way. Imprinted genes are expressed preferentially from either the paternally or maternally inherited allele. Long non-coding RNAs play essential roles in regulating this allele-specific expression. In several well-studied imprinting clusters, long non-coding RNAs have been found to be essential in regulating temporal- and spatial-specific establishment and maintenance of imprinting patterns. Furthermore, recent insights into the epigenetic pathological mechanisms underlying human genomic imprinting disorders suggest that allele-specific expressed imprinted long non-coding RNAs serve as an upstream regulator of the expression of other protein-coding or non-coding imprinted genes in the same cluster. Aberrantly expressed long non-coding RNAs result in bi-allelic expression or silencing of neighboring imprinted genes. Here, we review the emerging roles of long non-coding RNAs in regulating the expression of imprinted genes, especially in human imprinting disorders, and discuss three strategies targeting the central long non-coding RNA UBE3A-ATS for the purpose of developing therapies for the imprinting disorders Prader–Willi syndrome and Angelman syndrome. In summary, a better understanding of long non-coding RNA-related mechanisms is key to the development of potential therapeutic targets for human imprinting disorders.


INTRODUCTION
In diploid organisms, most genes are transcribed in an unbiased fashion from both alleles. However, in a small subset of genes, genetically identical alleles can be expressed differentially, a process referred to as ASE. In the mammalian genome, common epigenetic examples of ASE include random X-chromosome inactivation in females (Lee, 2011;Deng et al., 2014), genomic imprinting (Peters, 2014), random MAE (Reinius and Sandberg, 2015), allelic expression of antigen receptor (Bergman and Cedar, 2004;Vettermann and Schlissel, 2010), clustered protocadherin (Chen and Maniatis, 2013), and olfactory receptors (Monahan and Lomvardas, 2015). Imprinted genes are expressed strictly or preferentially from either paternally or maternally inherited alleles (referred to as parent-of-origin) (Barlow and Bartolomei, 2014;Huang et al., 2018;Chen and Zhang, 2020). The ASE of imprinted genes depends on differential epigenetic markings during gametogenesis in germline cells, as opposed to gene sequences. After imprinting patterns become established in mature germlines, genomic imprinting in an individual is maintained until genome-wide erasure of epigenetic modification occurs in gamete precursors. Genomic imprinting has been described in diverse organisms, including marsupials, flowering plants, and insects (Macdonald, 2012). In the human and mouse genome, genomic imprinting has been extensively observed, indicating the conservation and evolutionary significance of this epigenetic regulatory mechanism. While the expression of 1% of human proteincoding genes is estimated to be regulated via genomic imprinting (Im et al., 2005;Patten et al., 2016;Elbracht et al., 2020), many of these imprinted genes are essential for metabolism, development, and the nervous system (Monk et al., 2019;Tucci et al., 2019). Not surprisingly, dysregulated imprinting is closely associated with a broad spectrum of human developmental defects and genetic disorders, including PWS, AS, BWS, SRS, KOS14, and TS14 (Bian and Sun, 2011;Wan et al., 2017). The association between imprinted genes and the clinical features of these human diseases has also been documented in mouse models through the identification of homologous imprinted gene regions corresponding to the imprinted gene regions implicated in human imprinting disorders (Peters, 2014;Tucci et al., 2019).
Long non-coding RNAs are a subgroup of non-coding RNAs defined as having a length longer than 200 nucleotides, and are extensively expressed among the genome (Derrien et al., 2012;Harrow et al., 2012;Knauss and Sun, 2013). The number of lncRNA genes in the human genome has been estimated at 20,000 to 100,000 (Zhao et al., 2016;Fang et al., 2018;Uszczynska-Ratajczak et al., 2018). This number is greater than the canonical protein-coding genes in the human genome (Southan, 2017). lncRNAs are primarily retained in the nucleus, having short half-lives and a rapid turnover rate compared to mRNAs (Clark et al., 2012;Derrien et al., 2012;Yoon et al., 2015). lncRNAs can regulate gene expression in at least three ways: at the transcription level by modulating gene transcription and chromatin structure, at the post-transcription level by affecting splicing and stability of RNA, and at the translation level by modulating protein translation (referred to review Statello et al., 2021). In the human and mouse genome, imprinted genes often reside together within clusters (2-20 genes), called imprinted clusters or imprinted domains (Ferguson-Smith, 2011). In mammals, lncRNAs are generally located in imprinted clusters that contribute to the establishment and maintenance of monoallelic expression at a genome-scale and long time-range (Andergassen et al., 2019). Here, we summarize the roles of lncRNAs in the regulation of genomic imprinting using several well-established imprinted clusters as examples. We also discuss how the expression pattern of lncRNAs and their epigenetic regulatory functions are affected in imprinting disorders and some cancers. Three potential strategies have been developed to target the central long non-coding RNA Ube3a-ATS for the purpose of therapeutically correcting the PWS/AS locus imprinting disorders. We also discuss the functional mechanisms of imprinted lncRNAs in the regulation of mono-allelic imprinted gene expression and how it could help us understand ASE mechanisms and underlying pathological mechanisms of human imprinting disorders, hopefully inspiring additional efficient therapeutic strategies.

GENOMIC BASIS OF IMPRINTING
Along with more profound analysis of patient samples and well-established mouse reciprocal crossing models using highthroughput sequencing, the monoallelic expression of imprinted genes has been observed extensively in mice and humans (Tucci et al., 2019). Methylomes and transcriptomes derived from human peripheral blood and various adult tissue samples have been combined to identify imprinted methylation and the distribution of imprinted genes across the genome (Baran et al., 2015;Zink et al., 2018). In order to identify mouse imprinted genes, parents from strains with different genetic backgrounds were crossed to obtain heterozygotic individuals, permitting the discrimination of parent-of-origin-dependent transcriptional effects from sequence-dependent allelic expression (Babak et al., 2008;Wang et al., 2008). Imprinted genes in mice are identified based on SNPs specific to paternal or maternal genetic backgrounds, thus permitting the quantitation and comparison of expression levels from both alleles. To date, around 160 imprinted genes have been identified in the human genome, and 200 in the mouse genome (Tucci et al., 2019;Chen and Zhang, 2020). Sixty three of these imprinted genes are shared, suggesting that mouse models could be helpful for understanding imprinting regulation in humans.
In the human and mouse genome, imprinted genes often reside together within imprinted clusters (Ferguson-Smith, 2011). More than 80% of the known imprinted genes in the mouse genome are clustered together in multi-gene ranging in size from less than 100 kb to several megabases (Barlow, 2011). Imprinted lncRNAs located in one imprinted cluster are coordinately controlled by shared regulatory factors, including parent-of-origin-dependent differentially methylated regions (DMRs) and lncRNAs (Peters, 2014). In well-studied imprinted clusters, allele-specific DNA methylation occurs in an independent ICR in the germline, referred to as germlinederived DMRs (gDMRs) or primary DMRs, and persists after fertilization. ICRs in imprinted clusters exhibit parent-of-originspecific epigenetic modifications, including DNA methylation, governing different expression patterns of parentally inherited alleles (da Rocha and Gendrel, 2019). Around 35 imprinted gDMRs have been identified in the human genome  and 24 in the mouse genome to date (White et al., 2016). The establishment of gDMRs on paternal or maternal alleles ( Figure 1A) is essential for regulating imprinted gene expression in embryonic development (Barlow, 2011;Kelsey and Feil, 2013;Elbracht et al., 2020). In early primordial germ cells, epigenetic marks are extensively erased genomewide, including DNA methylation and histone modifications. In germline cells, DNA methylation of ICRs is re-established in gametes depending on the parent-of-origin. After fertilization, gDMRs escape secondary global epigenetic reprogramming. DNA methylation information at ICRs of the imprinted regions is retained. In this way, gDMRs of imprinted loci are established robustly during germline development and are resistant to genomic reprogramming after fertilization. Correspondingly, imprinting marks are inherited in a parent-specific manner (Chotalia et al., 2009;Henckel et al., 2012;da Rocha and Gendrel, 2019). gDMRs on the different parent-of-origin alleles are characterized by distinct chromatin configurations, marked with different histone modifications which are corresponding to 'open chromatin' and 'close chromatin' (Singh et al., 2011;Court et al., 2014;Sanli and Feil, 2015). The allele-specific methylation states of gDMRs are recognized by transcription factors with roles in maintaining parent-of-origin specific expression of the imprinted genes, such as ZFP57 protein (Riso et al., 2016). In total, differential methylation states of gDMRs on parental alleles are essential for the establishment of monoallelic gene expression.
Imprinting control regions govern DNA methylation and chromatin organization in early embryonic and adult lineages, resulting in the persistence of imprinting patterns across generations and their maintenance in adult tissues (Monk et al., 2019). After becoming established at early developmental stages in the germline, gDMRs are maintained in most somatic cells throughout life, resulting in the regulation of allelic expression of imprinted gene clusters. gDMRs also direct the rise of 'secondary' DMRs, normally corresponding to repressive chromatin modifications, condensed chromatin structure, and the gene-silencing function of imprinted lncRNAs (Sasaki et al., 1995;Nowak et al., 2011, p. 2;Rao et al., 2014;Tan et al., 2018;Zink et al., 2018). It has been shown that the imprinted expression of some genes is restricted to specific tissues or stages in developmental processes, along with additional allelespecific epigenetic marks further established in somatic cells. The expression patterns of these developmentally expressed imprinted genes are characterized by temporal-and spatialspecific biases Andergassen et al., 2017). For example, UBE3A and IGF2 show imprinted expression patterns in specific human brain cell types (Rougeulle et al., 1997;Vu and Hoffman, 1997;Pham et al., 1998;Yamasaki et al., 2003;Li J. et al., 2020). In a study of ASE in diverse tissues from 178 adult post-mortem donors, paternally silenced IGF2 was reported in the human brain, different from the canonical paternal expression observed in other tissues (Baran et al., 2015). In the mouse E6.5 gastrulating epiblast, it has also been observed that Igf2r is expressed from both alleles and further becomes imprinted in the embryonic lineage at the gastrulation stage (Marcho et al., 2015). Besides, the placenta-specific imprinting has been observed, and the underlying mechanism has been wellunderstood, especially in the potassium voltage-gated channel subfamily Q member 1 (Kcnq1)/Kcnq1 antisense transcript 1 (Kcnq1ot1) cluster and the antisense of Igf2r non-protein coding RNA (Airn)/Igf2r cluster ( Figure 1B; Sleutels et al., 2002;Andergassen et al., 2019;Hanna, 2020). The establishment of the placenta-specific imprinting initiates by allelic DNA methylation in pre-implantation embryos. In the placenta, the genomic profile of DNA methylation in imprinted DMRs is different, likely the result of an overall different pattern of placenta compared to other tissues (Schroeder et al., 2013). After implantation, the silencing of imprinted genes on the paternal allele in the post-implantation placental trophoblast expands and tends to be larger than the post-implantation epiblast. This expansion of gene silencing is mediated by the spreading of H3K27me3 marks along the paternal chromosome (Calabrese et al., 2015;Andergassen et al., 2017).

LONG NON-CODING RNAs AND THEIR ROLES IN REGULATING THE EXPRESSION OF IMPRINTED GENES
Two major mechanisms have been described to explain the regulation of the gene expression within an imprinted cluster (Lee and Bartolomei, 2013;Barlow and Bartolomei, 2014;Chen and Zhang, 2020). The first model is the lncRNA model, which may be more common. In this model, imprinted lncRNAs regulate imprinted gene expression. In the lncRNA model, imprinted lncRNAs intimately associate with ICRs. Imprinted lncRNAs are characterized by their capacity to silence imprinted genes in the same cluster (Rao et al., 2014;Kanduri, 2016;Tan et al., 2018;Zink et al., 2018;Tucci et al., 2019). As illustrated by the Kcnq1/Kcnq1ot1 imprinted cluster (Figure 2A), actively expressed imprinted lncRNA Kcnq1ot1 on the paternal allele can silence multiple imprinted genes bidirectionally along their located gene region . In contrast, a maternally methylated ICR on the paternal directly inhibits Kcnq1ot1 and its silencing effects, leading to the released expression of imprinted genes from the silencing by Kcnq1ot1. Another model, the insulator model is identified in other imprinted regions, in which parental allele-specific epigenetic differences at ICRs contribute to topological alternations of imprinted gene regions, inducing gene silencing or activation of specific alleles. This model is mainly applied to explain how imprinted genes in the insulin-like growth factor 2 (Igf2)/H19 locus are mechanistically regulated ( Figure 2B; Kaffer et al., 2000). H19 is a maternally expressed lncRNA (Bartolomei et al., 1991;DeChiara et al., 1991;Ferguson-Smith et al., 1991). The zinc-finger protein CTCF binds to the unmethylated maternal ICR and creates topologically associating FIGURE 1 | Genomic basis of the regulation of imprinting clusters. (A) The inheritance of allele-specific imprinting epigenetic marks across generations. In the early primordial germ cells, epigenetic modifications are erased at a genomic scale before the formation of germline cells. In the germline, parent-of-origin DNA methylation is established, shown as gDMRs. After fertilization and the formation of the zygote, the gDMRs are further maintained. Established imprinting patterns are maintained in blastocyte and somatic cells in adult tissues. (B) Imprinting in epiblast and placenta in imprinting loci, such as the Kcnq1ot1/Kcnq1 or Airn/Igf2r loci, is shown. In pre-implantation embryos, DNA methylation is inherited in the gDMR on the maternal allele, such as KvDMR1 of the Kcnq1ot1/Kcnq1 imprinting cluster. After implantation, the expression of lncRNA on the maternal allele is repressed by DNA methylation in gDMR, allowing the expression of neighbor genes. On the contrary, lncRNA is expressed from the paternal allele, inducing the spreading of H3K27me3 modifications in adjacent regions in the embryonic lineage (epiblast). In extra-embryonic lineage (placental trophoblast), the extended scale of H3K27me3 marks is longer than that seen in embryonic cells. Adjacent genes further away are also silenced on the paternal allele, indicating placenta-specific imprinting, such as Slc22a18 and Tssc4 genes in the Kcnq1ot1/Kcnq1 imprinting cluster. For simplicity, specific gene names are not shown.
domain boundaries, blocking Igf2 access to the enhancer like an 'insulator' (Schoenherr et al., 2003;Gómez-Marín et al., 2015). On the paternal allele, methylated ICR prevents CTCF binding and leads to secondary methylation of the H19 promoter and therefore silencing of lncRNA expression. The enhancers are then accessible to Igf2, permitting paternal-allele expression of Igf2 (Thorvaldsen et al., 1998;Chen and Zhang, 2020). Different from the lncRNA model, imprinted lncRNAs H19 in the insulator model are not the key regulation elements or whether imprinted lncRNAs affect other genes are not clear.
Here, we discuss on the role of imprinted lncRNAs in epigenetic regulation in the more common model, lncRNA model (Kopp and Mendell, 2018). lncRNA functions can be characterized based on their specific subcellular locations and interactions with DNA, RNA, and proteins, regulation of chromatin structure, expression of nearby and distal In Kcnq1ot1 imprinted cluster, the ICR is unmethylated on the paternal allele, permitting lncRNA Kcnq1ot1 expression. The expression of this lncRNA recruits the PRC1/2 complex and histone methyltransferase G9a, leading to condensed chromatin and silencing of flanking protein-coding genes. The ICR is methylated on the maternal allele, inhibiting lncRNA expression. The expression of Kcnq1 and several paternal silenced genes are activated. (B) The insulator model of imprinted gene expression regulation. The ICR on the maternal allele is unmethylated. CTCF binds to the maternal ICR and functions as an insulator to block Igf2 access to its distal enhancers. In contrast, the ICR on the paternal allele is methylated, preventing binding of CTCF. The expression of Igf2 is activated via enhancer regulation. (C) lncRNA Airn in Airn/Igf2r locus function in two distinctive mechanisms. On the one hand, methylated DMR on the maternal allele inhibits Airn expression, allowing access of transcription factors to the Igf2r promoter. The paternal DMR is unmethylated, permitting Airn transcription. Airn overlaps with the promoter of Igf2r and inhibits Igf2r expression. On the other hand, Airn transcripts recruit PRC2 complex to distal genes, such as Slc22a3 and Slc22a2, where they silence expression. Slc22a1 is a biallelic expressed protein-coding gene between distal regulated imprinted genes and Igf2r gene loci. (D) In Snurf-Snrpn/Ube3a imprinted cluster, the transcription of Ube3a-ATS starts from the exon upstream of the Snurf-Snrpn gene on the paternal allele. A group of non-coding RNAs are expressed, including Snord116 and Snord115 sno-lncRNAs and SnoRNAs. The elongation of this lncRNA overlaps with the Ube3a protein-coding region. A collision occurs between the converging elongation complexes of Ube3a-ATS and Ube3a resulting in the failure of Ube3a transcription elongation. By contrast, on the maternal allele, the ICR of Snurf-Snrpn/Ube3a cluster is methylated in the brain. G9a is recruited to the methylated DMR. This G9a accumulation leads to condensed chromatin and the silencing of flanking imprinted genes near the Snurf/Snrpn gene region. Consequently, the maternal Ube3a allele is expressed.
genes, RNA post-transcription modification, or mRNA translation (St Laurent et al., 2015;Kopp and Mendell, 2018;Statello et al., 2021). Imprinted lncRNAs range from 1.9 to 1,000 kb in length (Guenzl and Barlow, 2012) and regulate the expression of adjacent imprinted genes in cis through interacting with promoters and transcription factor binding sites, modifying chromatin status, or affecting higher-order structures (Barlow, 2011). Two major functional mechanisms of imprinted lncRNAs in the regulation of imprinted gene expression are hypothesized: interacting with promoters or enhancers of nearby target genes to affect transcription initiation, or overlapping imprinted gene regions, covering the gene body, and regulating the chromatin state of adjacent gene regions. We will also discuss the mechanisms underlying the regulation of imprinted gene expression by imprinted lncRNAs using well-characterized imprinted clusters as examples.

Inhibition of Transcriptional Initiation
Transcription of imprinted lncRNAs often overlaps with the promoters or enhancers of imprinted genes and influences their transcription (Lee and Bartolomei, 2013). These imprinted lncRNA transcripts often interfere with the transcription machinery of nearby imprinted genes, influencing the recruitment of transcription factors at their promoters (Latos et al., 2012). Based on an analysis of lncRNA and DNA binding in imprinting clusters from multiple mammalian species, it was suggested that the binding of lncRNAs to promoters of imprinted genes may be common . The Airn/Igf2r imprinted cluster in the mouse genome is a wellstudied example ( Figure 2C; Latos et al., 2012). On the paternal allele, the transcription profile of Airn initiates from its promoter embedded within the ICR in a direction antisense to the transcription of the Igf2r gene (Latos and Barlow, 2009). It was noted that intragenic truncations of the endogenous lncRNA Airn in embryonic stem (ES) cells that do not include the overlapping region are unable to silence the Igf2r paternal allele, thus demonstrating that inhibition of RNA polymerase II recruitment to Igf2r promoter region does not depend on the overlap between Airn transcription and the promoter (Sleutels et al., 2002;Latos et al., 2012;Santoro and Pauler, 2013). Furthermore, during ES cells differentiation, Airn expression was also necessary and sufficient to silence Igf2r . The overlapping regions between Airn transcription and Igf2r promoter and its gene body instead of Airn lncRNA products themselves lead to silencing of Igf2r expression.

The Disturbance of the Transcriptional Elongation
Another mechanism involves a collision between the converging elongation complexes of imprinted lncRNA and imprinted genes, leading to transcription stalling, premature termination, and subsequent degradation of the imprinted gene transcript ). An example is the UBE3A/UBE3A-ATS imprinted domain on human chromosome 15q11-13, in which imprinted genes, including MAGEL2, NDN, SNRPN, SNORD115, and SNORD116, are silenced on the maternal allele (Horsthemke and Wagstaff, 2008). In contrast, UBE3A, which encodes an E3 ubiquitin ligase, is expressed from the maternal allele, especially in neurons in the brain. The homologous imprinted locus in mice has also been identified and studied, locating at a syntenic loci chromosome 7qC (Yang et al., 1998; Figure 2D). In this imprinted cluster, the ICR embedded within the Snurf-Snrpn gene is unmethylated on the paternal allele. In mouse neurons, Ube3a-ATS lncRNA is expressed specifically from its promoter embedded in the unmethylated ICR (Yin et al., 2012;Meng et al., 2013). Notably, the Ube3a promoter region is not methylated differently like Ube3a-ATS. This, combined with the observation that Ube3a-ATS transcription initiates from an exon region upstream of the Snurf-Snrpn gene and elongates approximately 1,000 kb as far as the intronic region of Ube3a between exons 4 and 5 (Landers et al., 2004;Lewis et al., 2019), it was hypothesized that the two opposing polymerases of Ube3a and Ube3a-ATS collide ( Figure 2D). This transcriptional collision may lead to premature termination of Ube3a transcription inside its exon region on the paternal chromosome. In neurons from the monoallelic genetically engineered mouse model with the transcription of paternal Ube3a-ATS allele being terminated, Ube3a allele expression was activated on the paternal allele (Meng et al., 2013), resulting in increased expression comparable to maternal Ube3a (Meng et al., 2012). In cultured AS mouse neurons with biallelic silenced Ube3a expression, Antisense oligonucleotides (ASOs) targeting Ube3a-ATS rescued the expression of Ube3a efficiently (Meng et al., 2015). Consistently, in human induced pluripotent stem cells (iPSC)-derived neuron cells with biallelic silenced UBE3A expression, ASOs targeting UBE3A-ATS lncRNA transcripts lead to transcriptional termination by displacement of RNA Polymerase II, releasing the transcription of UBE3A (Germain et al., 2021). Recently, in human iPSCs, both sufficient expression of UBE3A-ATS lncRNA and two newly identified boundary elements were located inside the IPW gene and the PWAR1 gene (Martins-Taylor et al., 2014;Hsiao et al., 2019). These two genes are located between SNORD115 and SNORD116. In human iPSCs with the boundary elements deleted using gene editing technology, the expression of UBE3A was not silenced by upregulated UBE3A-ATS expression (Hsiao et al., 2019). Mapping RNAPII density showed that reduced active RNAPII across the 3 half of UBE3A corresponding to silenced UBE3A. These results together further support the hypothesized collision between UBE3A-ATS and UBE3A transcription complexes, leading to premature termination of the latter. In summary, the overlap between Airn and Igf2r promoter region disrupts the initiation of Igf2r transcription, while Ube3a-ATS silences the expression of Ube3a by disturbing its transcriptional elongation.

Chromatin Modification
Another lncRNA-related imprinting mechanism involves coating the bidirectionally flanking chromosomal region and recruiting repressive chromatin modification factors (Lee and Bartolomei, 2013;Sanli and Feil, 2015;Statello et al., 2021). The interactions between lncRNAs and these chromatin factors facilitate transcriptional silencing of target genes. The repressive chromatin-modification factors methylate DNA and produce histone modifications resulting in condensed chromatin structure and repressed gene expression. Among well-known repressive chromatin-modification factors, PRCs bind and spread across targeted chromatin facilitated by lncRNAs (Kotzin et al., 2016;Marín-Béjar et al., 2017). lncRNAs, genome structures, and CpG islands are essential factors in recruiting these PRCs, which have the capacity to catalyze lysine 119-mono-ubiquitinated histone H2A (H2AK119ub1) and H3K27me3 to repress gene expression through chromatin compaction and antagonization of transcriptional activators (Schwartz and Pirrotta, 2013;Simon and Kingston, 2013;Calabrese et al., 2015;Pintacuda et al., 2017;Colognori et al., 2019;Schertzer et al., 2019;Gil and Ulitsky, 2020;MacDonald and Mann, 2020). In genomic imprinting, some imprinted lncRNAs can bidirectionally direct repression of flanking neighbor imprinted gene region, such as KCNQ1OT1 lncRNA. Some lncRNAs can target distal gene regions in the same imprinted clusters they locate, such as Airn.

Locally Recruiting Condensed Chromatin Structure to Neighbor Gene Region
The Kcnq1/Kcnq1ot1 ICR, also known as KvDMR1 (KvLQT1 differentially methylated region 1), with the embedded lncRNA Kcnq1ot1 promoter, is unmethylated on the paternal allele (Figure 2A; Lee et al., 1999, p. 1;Smilinich et al., 1999;Beatty et al., 2006;Ager et al., 2008). lncRNA Kcnq1ot1 transcripts from the promoter region recruit several epigenetic factors such as the Polycomb group proteins RING1B (Polycomb Repressive Complex 1, PRC1), EZH2 (PRC2), and histone methyltransferase euchromatic histone lysine N-methyltransferase-2 (EHMT2 or G9a) to neighboring gene regions, forming repressive histone modifications such as H3K27me3 and H3K9me2 (Figure 2A II; Umlauf et al., 2004;Pandey et al., 2008). The chromatin state around the flanking regions of this lncRNA becomes condensed and results in silencing of flanking multi-protein coding genes such as Cdkn1c, Slc22a18, and Tssc4. On the maternal allele, DNA methylation of KvDMR1 silences the activation of the Kcnq1ot1 promoter and represses the transcription, releasing the transcription of neighboring genes.

Recruiting Chromatin Modification Factors to Distal Imprinted Genes
A typical example of imprinted lncRNA regulating distal imprinted genes through epigenetic silencing is Airn and recruitment of PRCs in the placenta ( Figure 2C; Latos et al., 2012;Lee and Bartolomei, 2013). As mentioned before, the transcription of Airn represses the expression of flanking imprinted gene Igf2r by transcriptional interference of the overlapping Igf2r promoter without repressive chromatin modification involved. In contrast, distal imprinted genes, such as Slc22a2 (about 100 kb to Airn locus) and Slc22a3 (about 300 kb to Airn locus), are also silenced by Airn in the extra-embryonic lineage, where Airn mediates the recruitment of PRC1 and PRC2 to distal targets on the paternal alleles (Terranova et al., 2008;Zhao et al., 2010;Schertzer et al., 2019). Recently, Airn was found to silence Slc22a3 in mouse trophoblast stem cells (Andergassen et al., 2019). Allele-specific chromosome conformation capture studies have suggested that Airn transcription throughout the enhancer of Slc22a3 may silence Slc22a3 expression by disrupting its promoter-enhancer interactions. However, with monoallelic deletion of the entire Airn gene, no essential enhancers for the distal silenced genes were found in the Airn gene region. Nonetheless, it has also been shown that Airn lncRNA is enriched on the Slc22a3 promoter together with an H3K9 dimethylase, G9a (Nagano et al., 2008). These results illustrate that Airn may target the promoters of distal imprinted genes by recruiting PRCs and G9a. The enrichment of these histone modification factors may lead to condensed chromatin in distal imprinted regions and silence imprinted genes.

THE ROLE OF IMPRINTED LONG NON-CODING RNAs IN HUMAN IMPRINTING DISORDERS AND CANCER
Long non-coding RNAs play essential roles in many biological processes and are related to various human diseases. Altered expression of imprinted loci has been linked to various neurodevelopmental disorders and cancers (Schaller et al., 2010;Huang et al., 2011;Riordan et al., 2013;Peters, 2014). Since imprinted regions are inherited in a parent-of-origin way, defects in one allele may be sufficient to lead to imprinting disorders (Kopp and Mendell, 2018). More specifically, silencing of parentally expressed imprinted genes can lead to the ultimate loss of its expression. Under abnormal conditions, DNA methylation status, allelic expression, and the biological functions of imprinted lncRNAs may be affected. These alterations may relate to human imprinting disorder-related disease phenotypes (Lee and Bartolomei, 2013). Here, we examine several well-studied imprinting disorders and emphasize the roles of imprinted lncRNAs in pathophysiological processes of imprinting-related diseases and cancers.

Common Molecular Mechanisms of Imprinting Disorders
Appropriate expression patterns of imprinted genes are important to growth and development. Correspondingly, imprinting disorder-related human diseases can be caused by genetic or epigenetic abnormalities on paternally or maternally inherited alleles (Lee and Bartolomei, 2013). Several common molecular mechanisms behind imprinting disorders have been defined, including molecular changes or genetic abnormalities, UPD, and epigenetic alterations (Figures 3A-D; Soellner et al., 2017;Carli et al., 2020). Firstly, genetic alterations, including SNPs and copy number variants on one imprinting allele, can affect imprinting ( Figure 3B). Another mechanism is UPD, in which the inheritance of two copies of chromosomes or chromosomal regions are both from either the paternal or maternal allele, resulting in synchronous expression or silencing ( Figure 3C; Robinson, 2000). Different from genetic alterations, epigenetic changes known as epimutations in DNA or histone modification without obvious genetic mutations have also been documented in imprinting disorders ( Figure 3D; Horsthemke, 2010). Hypermethylation at imprinted DMRs can silence the active allele of the original monoallelic expressed imprinted genes. In contrast, hypomethylation can result in overexpression of the original silenced allele. Epimutations can arise randomly or be driven by their environment during the inheritance of germline epigenetic imprinting marks. DNA methylation in DMRs can thus be abnormally inherited in the absence of genetic sequence alterations (Robertson, 2005). Moreover, as with molecular or genetic alterations, epimutations can be permanently maintained in somatic tissues for life and cause developmental phenotypes (Ioannides et al., 2014;Gillessen-Kaesbach et al., 2018;Monk et al., 2019). Besides the imprinted disorder caused by variations in a single imprinted gene, imprinting disorders with epigenetic alterations at loci across the genome have also been observed in many imprinting diseases, referred to as MLID (Horsthemke, 2010;Fontana et al., 2018). Instead of changes at specific genetic loci, MLID may be caused by a globally disturbed imprinting inheritance process across the genome. However, since current research is mostly limited to a subset of imprinted genes and the mosaic character of MILD (Azzi et al., 2014;Eggermann et al., 2021), the role of MLID in imprinting disorders is still poorly understood.

Congenital Imprinting Disorders and Related Imprinted Long Non-coding RNAs
Molecular disturbances, like loss or gain of methylation at ICRs, and subsequent loss or gain of imprinted gene expression have been described in various congenital human disorders. The frequencies of different molecular abnormalities vary among imprinting disorder-related diseases (Eggermann et al., 2015). More details about typical clinical syndromes and the pathological mechanisms are summarized in Table 1. However, details underlying imprinting disorder mechanisms and how they might impact adult neurobiology and developmental processes remain to be clarified. Fortunately, multiple mouse models of human imprinting disorders have been generated based on genomic conservation in most imprinted clusters. Strong correlates have been shown between these two genomes in imprinting loci, imprinting disorder phenotypes, and underlying molecular mechanisms (Perez et al., 2016). Spatial-and temporalspecific expression of allele-specific genes have been observed in several imprinting clusters in humans and mice. Here, imprinting disorders in three well-studied imprinting clusters are introduced as examples to demonstrate the roles of imprinted lncRNAs in imprinting-related congenital human disorders.

UBE3A-ATS in Prader-Willi Syndrome and Angelman Syndrome
Recent RNA-Seq data revealed strong allele-biased expression in the adult mouse brain, especially in imprinted regions (Perez et al., , 2016, where many of these genes are expressed in cell type-specific manners. Importantly, mutations or disruptions in imprinted genes are linked with extensive neurobehavioral phenotypes, demonstrating that brain-specific imprinted genes may play important roles in neurodevelopmental disorders (Tucci et al., 2019). PWS and AS are two neurodevelopmental disorders caused by oppositely inherited deficiencies occurred in the same imprinted cluster (Peters, 2014;Kalsner and Chamberlain, 2015;Buiting et al., 2016). These two syndromes perform common phenotype characters, including hypotonia at the newborn stage, abnormal sleep patterns, and the deficiency in intellectual development (Buiting, 2010;Kalsner and Chamberlain, 2015). Children affected by PWS exhibit poor suck phenotypes with reduced muscle tone and mental abilities (Buiting et al., 1995;Buiting, 2010;Fontana et al., 2017), while AS is characterized by deficient motor function, intellectual development, and speech abilities (Buiting, 2010;Eggermann et al., 2015). These two disorders are caused by imprinting disorder in the imprinted PWS/AS locus (UBE3A/UBE3A-ATS imprinted cluster) on human chromosome 15q11-13 ( Figure 4A). Similar to the mouse homologous locus mentioned previously in Section 3, the E3 ubiquitin ligase-encoding UBE3A gene is specifically imprinted in the brain (Vu and Hoffman, 1997). On the maternal allele, the methylated DMR encompasses the promoter of the SNRPN gene, silencing the SNURF/SNRPN gene and a series of downstream non-coding RNA genes (Rougeulle et al., 1997). In contrast, actively expressed UBE3A-ATS and the non-coding SNORD gene clusters are expressed from the paternal allele.
On the paternal allele of imprinted human PWS/AS locus, the unmethylated PWS-ICR is the region upstream to a proteincoding gene SNRPN and a lncRNA SNHG14 (small nucleolar RNA host gene 14) (Sutcliffe et al., 1994;Buiting et al., 1995;Rougeulle et al., 1997;Runte et al., 2001;Vitali et al., 2010;Chamberlain, 2013;Stanurova et al., 2018; Figure 4A). The neuron-specific non-coding transcript SNHG14 is processed to give rise to a series of non-coding RNA products, such as repeated C/D box small nucleolar RNAs (snoRNAs) and lncRNAs including 116HG, 115HG, and the antisense transcript to UBE3A (Mendiola and LaSalle, 2021; Figure 4A). The most studied RNA product from the host transcript SNHG14 is SNORD116 snoRNA, embedded within intronic regions of SNORD116 gene locus (Cavaillé et al., 2000;de los Santos et al., 2000;Stanurova et al., 2018;Mendiola and LaSalle, 2021). SNORD116 snoRNA present in ribonucleoprotein complexes (snoRNPs) and may participate in splicing, ribosomal RNA maturation, RNA modifications, and regulation of prohormone processing-related gene expression (Bazeley et al., 2008;Burnett et al., 2017). Meanwhile, SNORD116 locus encoded 116HG lncRNA was discovered recently (Vitali et al., 2010). 116HG is stably retained in the nucleus 'RNA cloud' at its transcription site (Powell et al., 2013a). 116HG potentially regulates transcript levels of circadianrelated genes in the cortex and energy-related metabolism through in a time-of-day-dependent manner (Coulson et al., 2018b). Similarly, SNORD115 locus encodes lncRNA 115HG and SNORD115 snoRNAs. While on the maternal allele, the methylated PWS-ICR occurs the upstream of the SNRPN gene. It silences the expression of the paternally expressed transcripts while allows the expression of UBE3A (Vu and Hoffman, 1997). Prader-Willi syndrome is the first human disease identified to be caused by the abnormal expression of non-coding RNAs (Sahoo et al., 2008;de Smith et al., 2009;Duker et al., 2010). All cases of PWS in humans involve a deletion in the SNORD116 On the paternal allele, UBE3A-ATS is expressed from the SNURF gene region, overlapping the exon region of UBE3A gene. On the paternal allele, lncRNA host transcript are processed to give rise to snoRNAs (SNORD115 and SNORD116), lncRNAs (116HG, 115HG, and UBE3A-ATS). Three different spliced non-coding transcripts are produced from SNORD116 gene locus, including 116HG lncRNA, snoRNAs, and sno-lncRNAs. SNORD116 sno-lncRNAs with snoRNAs on two ends are produced after splicing. (B) Imprinting disorders occur in PWS. PWS-related molecular alterations in UBE3A imprinted gene cluster. Line I indicates deletion in the imprinted region; Line II shows double maternal alleles are inherited, losing the paternal copy; Line III shows that the epigenetically mutated DNA methylation in DMR of the ICR leads to the silencing of lncRNA expression. Line IV: small deletion within the ICR. (C) Sno-lncRNAs transcribed from paternal allele can recruit Fox proteins and other related proteins, regulating Fox protein distribution and related alternative splicing functions. However, in PWS patients, loss of the UBE3A-ATS and other noncoding gene expression lead to the accumulation of Fox proteins in the nucleus and global abnormal splicing patterns. (D) Imprinting disorders occur in AS. Line I: deletions of the maternal imprinted regions containing the UBE3A and surrounding genes; Line II: both alleles are inherited from paternal chromosome; Line III: UBE3A mutations lead to transcript loss of function; Line IV: epimutations in the maternal allele lead to lncRNA expression from the maternal allele, preventing normal UBE3A expression.
Furthermore, an alternative RNA species (sno-lnRNAs) processed from SNORD116 host non-coding transcript has been described in human (Yin et al., 2012;Powell et al., 2013a; Figure 4A III). The role of SNORD116 sno-lncRNAs in RNA processing and decay of their target mRNAs is not well-understood but may facilitates our understanding of the connection between imprinting disorder and pathological mechanism of PWS ( Figure 4C). SNORD116 exon transcript is retained between two snoRNAs, forming sno-lncRNAs with two small nucleolar ribonucleoprotein ends (Yin et al., 2012). These sno-lncRNAs accumulate near the synthesis site together with a type of lncRNAs that are 5 capped by snoRNAs and 3 polyadenylated (SPAs) . These lncRNAs may interact with RNA binding proteins including TDP43 (TAR DNA-binding protein 43), RBFOX2 (RNA Binding Fox-1 Homolog 2), and hnRNP M (Heterogeneous nuclear ribonucleoprotein M). Especially, splicing regulator RBFOX2 are required for the neuron-specific splicing of Snord116 transcript to produce 116HG lncRNA and Snord116 snoRNA (Yeo et al., 2009;Coulson et al., 2018a). Since immunoprecipitation coupled with high-throughput sequencing (CLIP-seq) and RT-PCR assays confirmed that RBFOX2 directly binds to Snord116 snoRNA, it is hypothesized that Snord116 snoRNA may reduce the availability of these splicing-related proteins and regulate alternative splicing in the nucleus (Yin et al., 2012;Wu et al., 2016). Therefore, the disruption of SNORD116 in PWS may lead to more uniform distribution of RBFOX2 protein and global changes in normal alternative splicing patterns, contributing to PWS phenotypes.
In contrast to the paternal-allelic imprinting disorder in PWS, AS, is mainly caused by the lack of maternal UBE3A gene expression ( Figure 4D; Buiting et al., 2016). The brain-specific and maternally biased expression of UBE3A has been shown to function in regulating dendritic growth and influencing behavior and neurotransmitters (Avagliano Trezza et al., 2019). In AS patients, the expression of UBE3A or functional UBE3A protein is lost. These alternations can be caused by various imprinting disorder mechanisms including deletions of the maternally imprinted regions containing the UBE3A and surrounding genes. Besides pathological variants in the UBE3A gene, loss of SNURF DMR methylation has also been observed in AS cases (2-3%), in which the expression of UBE3A is silenced by UBE3A-ATS as discussed previously (Dagli et al., 1993).

KCNQ1OT1 and H19/IGF in Beckwith-Wiedemann Syndrome and Silver-Russell Syndrome
Beckwith-Wiedemann syndrome and SRS are clinically opposite growth-affecting disorders (Õunap, 2016). The underlying pathological mechanisms involve genetic and epigenetic perturbations of two imprinting clusters on human chromosome 11p15, the KCNQ1/KCNQ1OT1 and H19/IGF loci (Carli et al., 2020;Chang and Bartolomei, 2020;Figures 5A-C). BWS is one of the most common congenital overgrowth conditions (Mussa et al., 2013), with common phenotypes including postnatal overgrowth, placenta mesenchymal dysplasia, and congenital and childhood cancer predisposition. In contrast, SRS patients exhibit postnatal growth failure with body hemihypoplasia, lower birth weight, fetal undergrowth and poor feeding predisposition (Wakeling et al., 2017).
Approximately 50% of BWS patients lose DNA methylation accompanied by loss of H3K9me2 on maternal KvDMR1 (Figure 5A; Robertson, 2005). This epigenetic disturbance results in biallelic expression of the KCNQ1OT1 lncRNA. As a consequence, expression of this lncRNA silences adjacent imprinted genes on both alleles (Soejima and Higashimoto, 2013). Among these silenced genes, CDKN1C is linked to the development of BWS phenotypes (Yan et al., 1997;Zhang et al., 1997;Tunster et al., 2011). These epigenetic mutations in maternal KvDMR1 and biallelic expressed KCNQ1OT1 lncRNAs lead to loss of CDKN1C expression and fetal overgrowth, thus contributing to BWS syndrome (Eggermann et al., 2016;Wakeling et al., 2017). Therefore, after the establishment of DMRs on imprinted alleles, monoallelic expression of KCNQ1OT1 lncRNA is a crucial regulator of adjacent proteincoding genes, which have essential roles in maintaining normal growth processes during early development. Another major abnormal imprinted cluster identified in BWS patients is H19/IGF2 (Figure 5B). Under normal conditions, H19/IGF2 ICR is methylated on the paternal chromosome, controlling the expression of H19. In BWS patients, mutations or hypermethylation of the H19/IGF2 ICR can lead to H19 silencing and subsequent overexpression of IGF2, a circulating hormone and tissue growth factor. The upregulated expression of IGF2 is linked to BWS overgrowth-related phenotypes (Pollak et al., 2004;Brioude et al., 2018a,b;Duffy et al., 2019). As for SRS, loss of H19/IGF2 ICR methylation on the paternal chromosome 11p15 accounts for 40-60% of patients (Wakeling et al., 2017). ICR hypomethylation is bound by the insulator CTCF. The interaction of the IGF2 promoter with its enhancer on both alleles is disrupted, resulting in decreased IGF2 expression and subsequent growth and development delays (Figure 5C; Abi Habib et al., 2017).
Although some BWS and SRS patients can be identified based on clinical features alone, diagnosing imprinting disorders can be complicated by complex molecular alternations (Ibrahim et al., 2014;Wakeling et al., 2017). In addition to the two imprinted loci primarily relevant to BWS and SR phenotypes, MLID has also been observed in an increasingly growing fraction of patients with methylation abnormalities at other imprinted loci (Rossignol et al., 2006;Azzi et al., 2009;Eggermann et al., 2011;Fontana et al., 2018). In addition, symptoms vary widely in patients with imprinting disorders (Wakeling et al., 2017;Brioude et al., 2018b;Mantovani et al., 2018). Therefore, additional insights into the relationship between the epigenetic mechanisms of imprinting disorders and neurological diseases can help clarify more accurate diagnostic guidelines and appropriate clinical therapies.
The distribution of imprinted genes and regulatory mechanisms of DLK1/DIO3 locus are highly conserved between humans and mice. The regulation of this imprinted locus has been revealed in mouse models established with genetic alterations in the Dlk1/Dio3 locus on chromosome 12 ( Figure 5D; Paulsen et al., 2001;da Rocha et al., 2008). Three paternally expressed imprinted protein-coding genes are Dlk1, Rtl1, and Dio3. lncRNA Meg3 (also called Gtl2), the Rtl1-antisense Rtl1as, the C/D-box snoRNA cluster Rian, and the microRNA cluster Mirg are transcribed from the maternal allele (da Rocha et al., 2008;Kota et al., 2014). The regulation of imprinted gene expression in this locus relies on an intergenic DMR (IG-DMR). On the maternal allele, AFF3 protein binds to an upstream enhancer of Meg3, activating lncRNA expression. In contrast, on the paternal allele, AFF3 binds instead to the methylated IG-DMR, leading to silencing of Meg3 and other non-coding genes Wang et al., 2017). It has also been suggested recently that maternally expressed lncRNA Meg3 is involved in the regulation of the Dlk1/Dio3 imprinted cluster (Sanli et al., 2018). The maternal expression of the Meg3 lncRNA may play a role in preventing maternal Dlk1 activation through interaction with the lysine methyltransferase (KMT) Ezh2 and PRC2 in the maternal Dlk1 gene region (Kaneko et al., 2014;Sanli et al., 2018). Remarkably, Meg3 lncRNA's regulation of imprinted protein-coding gene Dlk1 is restricted to a developmental window as follows. In embryonic stem cells, the Dlk1 gene is expressed biallelically at a low level. Upon neuronal differentiation, Dlk1 expression is upregulated on the paternal allele. Conversely, the activation of the Dlk1 gene on the maternal allele is prevented by the overlap of Meg3 lncRNA in cis and the recruitment of Ezh2 to the Dlk1 gene region (Sanli et al., 2018). Although the Meg3 lncRNA is necessary for the silencing of Dlk1 expression, the mechanisms underlying the connection between the Meg3 lncRNA and repressed Dlk1 expression on the maternal allele are unknown.
The DLK1/DIO3 locus is predominantly imprinted in the human brain (Davis et al., 2005;Ferrón et al., 2011). Proteincoding genes DLK1, RTL1, and DIO3 are expressed on the paternal allele; lncRNAs (MEG3, MEG8, RTL1as, DIO3OS), snoRNAs, and miRNAs are transcribed on the maternal allele. Importantly, the DLK1 gene plays essential functions in regulating development and metabolism. In KOS14 patients, gain of DNA methylation on the maternal ICR leads to MEG3 silencing ( Figure 5E I; Sato et al., 2011). However, maternal micro-deletions of the MEG3 promoter that don't affect ICR methylation are also observed in some cases ( Figure 5E II; Kota et al., 2014). In another case, a maternal micro-deletion has been detected in the MEG3 gene body instead of the IG-DMR or MEG3 promoter ( Figure 5E III; van der Werf et al., 2016).
In summary, in these conditions, imprinted lncRNAs play essential roles as upstream regulators of protein-coding genes in the same imprinted cluster. However, the detailed mechanisms are diverse and complicated in different imprinting disorders and remain to be further investigated.
H19 is one of the most commonly implicated tumorigenesispromoting lncRNAs (Zheng et al., 2020). The expression of H19 occurs during embryonic development and decreases after birth in most tissues. However, H19 is abnormally upregulated in various cancers, including breast, liver, lung, esophageal, pancreatic, ovarian, and bladder (Vennin et al., 2015;Zhang et al., 2016). H19's tumor-promoting effects include the inhibition of cell death, promotion of proliferation, downregulation of growth suppressors, and promotion of invasion and metastasis (reviewed in Matouk et al., 2015;Lecerf et al., 2019). Moreover, high H19 expression may be a molecular marker to predict cancers and prognoses after clinical treatment, including the rate of posttherapeutic relapse in hematological cancer patients . Increased risk of developing congenital and childhood tumors seen in BWS is also associated with aberrant H19. H19 is also associated with growth suppression (Yoshimizu et al., 2008;Lecerf et al., 2019;Zhou et al., 2019). H19's contribution to tumorigenesis varies by tissue and developmental windows and requires clarification in future investigations.

MODULATION OF THE LONG
NON-CODING RNA UBE3A-ATS TO RESCUE ABNORMAL IMPRINTING IN PRADER-WILLI SYNDROME/ANGELMAN SYNDROME IMPRINTED CLUSTER Although our understanding of the mechanisms of imprinting disorders has grown, efficient molecular diagnosis and effective treatments are limited to nonexistent (Elbracht et al., 2020).
Modulation of imprinted lncRNAs has been proposed as a potential therapeutic strategy to target imprinted genes and rescue imprinting disorders (Peters, 2014;Statello et al., 2021). As the epigenetic regulatory mechanisms of the Ube3a/Ube3a-ATS imprinted cluster are understood best, attempts have been made to rescue Ube3a expression through modulating the collision between the transcriptional machinery of Ube3a and Ube3a-ATS in an allele-specific manner. Herein, three state-ofthe art therapeutic strategies by targeting Ube3a-ATS lncRNA, editing Ube3a-ATS gene region, or modulating chromatin transcriptional state by small molecules are discussed along with recent preclinical studies of UBE3A/UBE3A-ATS imprinted cluster-related diseases.

Antisense Oligonucleotides for Imprinted Long Non-coding RNAs
Antisense oligonucleotides are single-stranded DNA oligos designed using sequence homology with their RNA targets that hybridize with the targeted RNA region based on complementary base pairs, and induce subsequent RNA degradation at the ASO-RNA heteroduplex part (Mishra et al., 2019;. ASOs can be used to alter splicing or gene expression. ASOs have been designed as potential therapies for various diseases, including AS, spinal muscular atrophy (SMA), Duchenne muscular dystrophy, Huntington disease, and hyperlipidemia (Beaudet and Meng, 2016;Dhuri et al., 2020). Several ASO-based therapies, such as Nusinersen (Spinraza) for SMA treatment, have received approval by the United States Food and Drug Administration (FDA) and other regional regulatory agencies (Karaki et al., 2019). Nusinersen is quite effective in rescuing protein deficiency by altering pre-mRNA splicing (Hoy, 2017;Groen et al., 2018;Claborn et al., 2019). The capacities of ASOs to access targeted RNAs through homology base pairing and in inducing RNase H-mediated cleavage at the pairing regions by exonucleases make them suitable to decrease lncRNA levels post-transcriptionally (Chan et al., 2006).
As mentioned before, on the paternal allele of Ube3a/Ube3a-ATS imprinted cluster, Ube3a-ATS represses Ube3a expression by prematurely terminating the elongation of Ube3a transcripts. Therefore, a potential strategy is to rescue the defective Ube3a transcription by targeting Ube3a-ATS transcripts using ASOs (Figure 6A I). To avoid influencing the transcription of sno-lncRNAs essential for neuronal development and PWS, ASOs were designed to be complementary to Ube3a-ATS transcripts downstream of the Snord115 cluster. These ASOs were provided to cultured AS mouse neurons with deficient Ube3a expression (Meng et al., 2015). The treatment achieved sustained ectopic paternal expression of Ube3a, partially rescued UBE3A brain protein levels, and alleviated some cognitive deficits. Remarkably, other splicing products derived from Ube3a-ATS like Snrpn and Snord116 were unaffected. Consistently, ASOs were designed to rescue the expression of UBE3A in AS iPSC-derived neuron cells with a large deletion of maternal 15q11-q13. ASOs targeting UBE3A-ATS transcripts at SNORD115 and SNORD109B, or targeting the snoRNA located between SNORD 115 locus and UBE3A gene region, cleave UBE3A-ATS and release the transcription of UBE3A on the paternal chromosome (Germain et al., 2021). UBE3A-ATS transcription is terminated by displacing RNA Polymerase II several kilobases downstream of the ASO targeting site. Therefore, targeting the lncRNA UBE3A-ATS by ASOs could be a potential strategy for rescuing UBE3A expression and related imprinting disorders. Besides, ASOs have several unique features in treating imprinted disorders, including high in vivo efficacy, broad tissue distribution, low adverse events, and long duration of action (Smith et al., 2006;Kordasiewicz et al., 2012). Considering that several mRNA-targeting ASOs have been approved (Dhuri et al., 2020), targeting lncRNAs using ASOs to treat imprinting diseases could achieve wide application. However, robust delivery systems devoid of associated toxicity should be carefully developed and evaluated.
Modulation of Imprinted Long Non-coding RNA Expression Using the CRISPR/Cas9 System The CRISPR/Cas9 system permits in vitro and in vivo gene editing tool and is another novel strategy to modulate imprinted lncRNA expression (Deltcheva et al., 2011;Jinek et al., 2012;Konermann et al., 2015;Suzuki et al., 2016;Knott and Doudna, 2018). A series of CRISPR/Cas-engineered systems can be designed to manipulate lncRNAs, including deletion of the lncRNA encoding gene region (pre-transcription level), inhibition or activation of the expression of the lncRNA (transcription level), or direct degradation of the lncRNA transcripts (post-transcriptional level) (Perez-Pinera et al., 2013;Qi et al., 2013;Ran et al., 2013;Abudayyeh et al., 2017). For example, CRISPRi and CRISPRa can modulate lncRNA expression by recruiting transcriptional repressors or activators without inducing genetic mutations (Bester et al., 2018;Kampmann, 2018). At the same time, CRISPR/Cas9 is being studied as a strategy of in vivo genome editing therapy in neurological diseases like schizophrenia and Alzheimer's disease Kuruvilla et al., 2018;Park et al., 2019;Sun et al., 2019). It is hoped that effective manipulation of the non-coding regions achieved in human cell lines and animal models could result in novel strategies to eliminate obstacles in developing therapies for lncRNA-related imprinting diseases (Cho et al., 2013;Cong et al., 2013;Jiang et al., 2013;Doudna and Charpentier, 2014, p. 9). Furthermore, when taking into account brain-specific expression of imprinted clusters, CRISPR/Cas9 could be designed to correct abnormal imprinting patterns (Han et al., 2014). Indeed, recently Cas9 gene therapy has shown promise in trapping Ube3a-ATS to activate paternal Ube3a expression ( Figure 6A II; Wolter et al., 2020). In addition, a CRISPR/Cas9 system targeting the Snord115 locus in cultured mouse cortical neurons and human neural progenitorderived neurons was able to successfully increase total Ube3a protein expression while decreasing Snord115 expression. Using a neuron-specific saCas9 and guide RNAs packaged in an adenoassociated virus delivering system and administered to an AS mouse brain during the embryonic and early postnatal stages led to silencing of paternal Snord115 expression with longlasting effects.
FIGURE 6 | Three state of the art strategies for imprinted disorders via targeting imprinted lncRNAs. (A) Therapeutic strategies for AS. Molecular alterations such as deletions in Ube3a/Ube3a-ATS imprinted cluster can cause the loss of effective Ube3a expression. Line I: ASOs are designed to target the overlapping regions of Ube3a-ATS transcripts and Ube3a, releasing paternal Ube3a expression; Line II: the human synapsin 1 (hSYN1) gene promoter is used drive neuron-specific expression and Cas9 packaged with adeno-associated virus delivering system is inserted into the gene region of the Snord115, leading to disrupted transcription of Ube3a-ATS before extending to the Ube3a gene encoding region. Line III: Top I inhibitors disrupt the elongation of the Ube3a-ATS at the Snord116 region. Ube3a paternal expression is released from transcriptional collision. (B) Therapeutic strategy for PWS. G9A inhibitors prevent G9A recruitment to flanking regions near the ICR, releasing Ube3a-ATS lncRNA expression from its promoter in the ICR.
In summary, the CRISPR-Cas9 system offers promising therapeutic strategies with the potential to permanently alter imprinted gene expression with high specificity and low toxicity. Nevertheless, since lncRNAs lack open reading frames and functional protein products, the use of CRISPR-Cas9 system to achieve efficient lncRNA manipulation needs to be further improved (Statello et al., 2021). In addition, an optimal sgRNA design and an effective delivery mechanism to penetrate the blood-brain barrier need further investigation Hana et al., 2021).

Small Molecules Targeting Histone Modifiers
Small molecules have been screened to target histone modification proteins involved in imprinted lncRNA regulation. As mentioned before, PWS and AS are two imprinting disorders related to the same imprinted cluster. In AS patients, UBE3A expression is decreased. Through high-content screening in mouse-derived primary cortical neurons, about 10 topoisomerase I (Top I) inhibitors have been identified with the capacity to downregulate Ube3a-ATS expression and induce reactivation of UBE3A expression from the paternal allele (Huang et al., 2011;Powell et al., 2013b). The Top I inhibitor topotecan blocks the elongation of the Ube3a-ATS transcription complex in cultured mouse neurons (Powell et al., 2013b). It inhibits sno-lncRNA transcription throughout the Ube3a encoding gene region by stabilizing the formation of R loops between RNA and DNA within paternal Snord116, leading to chromatin decondensation (Liu and Wang, 1987;Belotserkovskii et al., 2010;El Hage et al., 2010;Belotserkovskii and Hanawalt, 2011;French et al., 2011;Skourti-Stathaki et al., 2011;Aguilera and García-Muse, 2012;Ginno et al., 2012; Figure 6A III). The Ube3a-ATS transcription complex stalled before transcription of Sno-lncRNAs completed. Subsequent Ube3a expression was reactivated on the paternal allele. Additional candidates of other Top I inhibitors have also been assessed to identify inhibitors with better pharmacological profiles of Ube3a activation . Prospective therapeutic safety and central nervous system (CNS) bioavailability studies have also been performed recently in AS mouse neurons .
A therapeutic strategy for PWS based on the induction of SNORD116 expression has been proposed. SNORD116 is normally silenced on the maternal allele, but its expression can be induced by modulating 'closed' chromatin condition into an 'open' state (Kim et al., 2017). The methylation of histone H3K9 performs allele-specific pattern in the ICR located upstream of SNRPN gene (PWS-ICR). On the maternal chromosome, histone methyltransferase euchromatic histone lysine N-methyltransferase-2 (G9a) locates at the methylated PWS-ICR and recruits repressive histone modifications (H3K9me2) along the PWS-ICR in a bidirectional manner. This leads to condensed chromatin structure and silencing of PWS-associated genes ( Figure 6B). The inactivation of histone H3K9 methyltransferase G9a in mouse embryonic stem (ES) cells leads to reduced DNA methylation in PWS-ICR, and the expression of Snrpn was activated on both chromosome in vitro (Xin et al., 2003). However, in in vivo mouse model, two inhibitors of G9a selected lead to the activation of maternal copy of Snord116 and improve survival of the PWS mouse without effect on the methylation state of the PWS-ICR or Ube3a expression on the maternal allele (Kim et al., 2017). Thus, further studies are needed to clarify the association between DNA methylation of PWS-ICR and allele-specific distribution of G9a. Meanwhile, the reactivation of SNRPN and SNORD116 was recently achieved by preventing the recruitment of H3K9me3 repressive histone modification-related protein factor to SNORD116 locus in PWS-derived iPSCs . In summary, small molecules related epigenetic therapy for PWS through modulating the condition of specific chromatin regions could be a potential strategy to be translated in clinical relevance (Crunkhorn, 2017;Chung et al., 2020).

CONCLUDING REMARKS
Several imprinting gene clusters have been discovered and studied since the middle of the last century. These studies have shown that lncRNAs play crucial roles in regulating imprinted gene clusters and individual imprinted genes related to human health and diseases. However, from a genomic perspective, the characteristics of gene regulation among imprinting loci remain to be fully elucidated. This is despite the advancement in knowledge of the epigenetic regulatory mechanisms of a subset of genes in imprinted regions. In the three imprinted clusters (Airn/Igfr2, Kcnq1/Kcnq1ot1, and Ube3a/Ube3a-ATS), imprinted lncRNAs which play essential regulatory roles in silencing other imprinted genes are all expressed on the paternal allele. It has been reported that maternal expressed imprinted genes are prominent with protein-coding genes, while paternal expressed genes exhibit consistent distribution between non-coding and protein-coding sequences (Hutter et al., 2010). However, the difference between the establishment of maternal and paternal imprinted genes in lncRNA mechanisms remains unclear. Thus, comprehensive investigations are needed to understand further the mechanisms of imprinted lncRNAs in the epigenetic regulation of imprinted clusters. With technological advancements, studies on lncRNA-associated human imprinting disorders will lead to needed therapies.
Pharmacological treatments for congenital imprinting disorders are limited to symptomatic therapies, which are inefficient in promoting the patients' quality of life . Fortunately, the biological role of lncRNAs in the etiology of congenital imprinting disorders has been revealed thanks to the advancement in high-throughput genome-wide sequencing technologies. Therapeutic approaches based on disease-related lncRNAs have been investigated. In a recent study, lncRNA mimics were designed to restore the tissue-specific lncRNA HULC in mice, essential for phenylalanine metabolism (Li et al., 2021). In addition, three strategies mentioned above targeting Ube3a-ATS have efficiently rescued imprinting disorders in PWS/AS imprinted cluster in mouse models and human cell lines. Although the three strategies mentioned here targeting Ube3a-ATS have efficiently rescued imprinting disorders of PWS/AS imprinted cluster in mouse models and human cell lines, therapies for other disease-related clusters have not been investigated. Long non-coding RNA-based and lncRNA-targeting therapies have some unique advantages. For instance, in lncRNA-targeting methods like ASOs, synthesized RNA can be designed with organ-targeting peptides to achieve tissue-specific targeting of endogenous lncRNAs. Besides, synthesized RNA products could be modified to promote in vivo stability. Further translation of these strategies to real clinical tools will require further investigation to overcome related challenges. In vivo delivery of synthesized RNA molecules, cellular permeability, immunogenicity, and potential of organ toxicity also deserve further investigation (Perry and Ulitsky, 2021). Another challenge to extend the lessons learned in PWS and AS into other imprinting disorders is the epigenetic and molecular complexities in different imprinting disorders-related imprinted loci. Considering the complexity of the regulatory network of genomic imprinting, further efforts are needed to reveal underlying pathological mechanisms linked to imprinting disorder phenotypes and support continuous improvement of clinical management and therapeutic strategies.

AUTHOR CONTRIBUTIONS
TW and QM wrote and edited the manuscript and drew the pictures. JL, LY, and MW assisted in manuscript collation and review. QM provided critical inputs as the corresponding authors and obtained funds. All authors contributed to the article and approved the submitted version.