IPSC Models of Chromosome 15Q Imprinting Disorders: From Disease Modeling to Therapeutic Strategies

The chromosome 15q11-q13 region of the human genome is regulated by genomic imprinting, an epigenetic phenomenon in which genes are expressed exclusively from one parental allele. Several genes within the 15q11-q13 region are expressed exclusively from the paternally inherited chromosome 15. At least one gene UBE3A, shows exclusive expression of the maternal allele, but this allele-specific expression is restricted to neurons.

The appropriate regulation of imprinted gene expression across chromosome 15q11-q13 has important implications for human disease. Three different neurodevelopmental disorders result from aberrant expression of imprinted genes in this region: Prader–Willi syndrome (PWS), Angelman syndrome (AS), and 15q duplication syndrome. These three disorders each occur at an estimated frequency of approximately 1/15,000–1/30,000 live births [1, 2]. The Prader–Willi and Angelman syndromes most commonly result from large, 5–7 Mb deletions that include both imprinted and non-imprinted genes. These deletions occur during meiosis and are mediated by local repetitive sequences [3-5]. 15q duplication syndrome results when the same repetitive sequences mediate duplication, rather than deletion, by unequal homologous recombination [6]. In addition to interstitial duplication, an extra isodicentric chromosome 15q can also result, yielding two extra copies of the chromosome 15q11-q13 region [7]. Induced pluripotent stem cells (iPSCs) have been generated from individuals with each of these disorders by multiple groups.

A map of the chromosome 15q11-q13 region is shown in Fig. 1. MKRN3, MAGEL2, NDN, C150RF2, SNURF-SNRPN (heretofore called SNRPN), SNORD107, SNORD64, SNORD108, SNORD109A, SNORD116, IPW, SNORD115, and SNORD109B are exclusively expressed from the paternally inherited allele. UBE3A is expressed from the maternally inherited allele, and shows brain-specific imprinted expression. ATP10A is similarly imprinted and expressed from the maternally inherited allele in brain, although not in all individuals [8]. Several genes in the 15q11-q13 region lie within the region deleted or duplicated in PWS, AS, or Dup15q syndrome but are not imprinted. Proximal to the imprinted domain are GOLGA8E, TUBGCP5, CYFIP1, and NIPA2 [9]. Distal to the imprinted domain, lie a cluster of three non-imprinted GABA_A receptor subunit genes, GABRB3, GABRA5, and GABRG3, and the genes OCA2, HERC2, and GOLGA8G [10]. Each of the three disorders is described more fully below.

Fig. 1.

Diagram of the human 15q11-q13 genetic region. Individual genes are indicated by rectangles. Paternally expressed imprinted genes are denoted by blue rectangles, the maternally expressed imprinted gene is denoted by a red rectangle. Black rectangles represent silences alleles. Gray rectangles represent genes expressed from both parental alleles. BP1, BP2, and BP3 indicate common deletion/duplication break points. Dashed line represents the SNHG14 non-coding RNA. The large dashes represent SNHG14 expressed in all cell types, while the short dashes represent neuron-specific SNHG14/UBE3A-ATS expression. The PWS critical region is highlighted in blue. The neuron-specific portion of SNHG14 is highlighted in pale orange

1. Prader–Willi Syndrome

Prader–Willi syndrome (PWS) is a multifaceted disorder that first presents as hypotonia and failure to thrive in affected infants, which gives way to hyperphagia and morbid obesity. Phenotypic hallmarks also include small stature with small hands and feet, mild to moderate cognitive deficit, and behavioral problems, including obsessive-compulsive disorder [1].

PWS is caused by the loss of expression from the paternal allele of 15q11-q13. Approximately 70% of PWS patients have a ~ 5–7 Mb deletion of the entire region; approximately 25% of PWS patients have inherited both copies of chromosome 15 from their mother, a condition called maternal uniparental disomy (UPD); and approximately 5% of PWS can be attributed to an imprinting defect whereby the individual has inherited one chromosome 15 from each parent, however, the paternally inherited allele of 15q11-q13 behaves as if it were the maternally inherited allele (Fig. 2). Originally thought to be a true contiguous gene syndrome because no single gene mutation was found to lead to the disorder, recently identified patients have shown that PWS may result from the loss of one of the snoRNA clusters that reside in the region. Individuals lacking the SNORD116 snoRNA cluster and IPW suffer the same failure to thrive, hypotonia, and hyperphagia that is observed in patients with larger deletions [11, 12]. Therefore, this region may encompass the PWS critical region. Despite the further refinement of the PWS critical region to a defined set of transcripts, their function and contribution to the PWS disease mechanism remain a mystery.

Fig. 2.

Genetic etiologies of Prader–Willi, Angelman, and Dup15q syndromes. Maternal and paternal chromosomes 15 are depicted in pink and blue, respectively. Circle represents the centromeres, and the horizontal lines represent the 15q11-q13 region. The green arrow represents the active paternal PWS-IC and ensuing transcription. Approximate percentages of each genetic etiology is reported for PWS and AS. UPD uniparental disomy, Imprinting imprinting defect, Mat. int. dup. maternal interstitial duplication, Mat.int.trip maternal interstitial triplication, and idic15 isodicentric chromosome 15

PWS mice produced by targeted mutation of the PWS-IC or by a paternally inherited transgene insertion that results in the physical deletion of the paternally expressed genes, recapitulate the early failure-to-thrive phenotype and show perinatal lethality, depending on their genetic background [13-15]. Although the regulation of imprinted gene expression seems similarly regulated between mouse and human, murine models of PWS have yet to show hyperphagia and obesity. Deletion or mutation of individual paternally expressed genes within the chromosome 15q11-q13 (i.e. Ndn [16, 17], Magel2 [18, 19], Snrpn [13], Snord116 [20]) has also been generated. While one group reported that deletion of Snord116 caused hyperphagia in mice [20], another group did not observe that phenotype in a similar murine model. A third group reports hyperphagia in mice with adult-onset deletion of Snord116 in the mediobasal hypothalamus, as well as obesity in a subset of those mice [21]. None of the individual gene disruptions has recapitulated the fully penetrant obesity seen in PWS individuals. The absence of an obesity phenotype in murine models has propelled the scientific community to investigate neuronal deficits in human iPSC-derived hypothalamic neurons.

Schaaf–Yang syndrome (SYS) has features both overlapping with PWS and distinct from PWS [22, 23]. Individuals with SYS have feeding difficulties, global developmental delay, intellectual disability ranging from mild to severe, hypotonia, joint contractures, and autism spectrum disorder [23]. SYS is caused by loss-of-function mutations in the paternal allele of MAGEL2. Mouse models with Magel2 deletions show growth deficits, failure-to-thrive, reproductive phenotypes, circadian rhythm deficits, and behavioral phenotypes [18, 19, 24, 25].

At least six different groups have developed human iPSC models of PWS [26-31]. iPSC models harboring large deletions of chromosome 15q11-q13 [26, 28, 30, 31], two different atypical smaller deletions [29, 32], two different translocations that cause PWS [27, 28], and maternal uniparental disomy [33] have been generated. For each of these cases, multiple cell lines appropriately modeling PWS gene expression and epigenotype have been generated. PWS is a disorder primarily affecting hypothalamic function. Generation of highly enriched populations of mature hypothalamic neurons from human pluripotent stem cells is challenging and is a limiting factor for PWS iPSC research.

2. Angelman Syndrome

The human neurogenetic disorder, Angelman Syndrome (AS), is caused by loss of function from the maternally inherited allele of UBE3A. UBE3A is an E3 ubiquitin ligase that is expressed from both parental alleles in most tissues. However, UBE3A is expressed exclusively from the maternally inherited allele in the brain, so mutation or deletion of the maternal copy of the gene results in complete loss of the ubiquitin ligase in this tissue [34].

Loss of UBE3A expression in AS patients occurs due to one of five classes of genetic abnormalities (see Fig. 2). 70% of patients suffer from a ~ 5–7 Mb deletion of the maternal allele of chromosome 15q11-q13 and 10% of patients harbor a loss-of-function mutation in the UBE3A gene. These two classes comprise the most common genetic etiologies of AS. The remainder of patients attribute AS to either paternal uniparental disomy or an imprinting defect in which the maternal allele behaves as if it were paternally inherited [35]. A small but significant portion of individuals with a clinical diagnosis of AS has none of the aforementioned genetic etiologies. In some of these cases, mutations in other genes such as SLC9A6 or TCF4 have been identified. Indeed, there are a handful of rare neurodevelopmental disorders that can present as Angelman-like [36].

Phenotypic hallmarks of AS include motor dysfunction leading to an ataxic gait, frequent seizures that can be severe, profound learning disability coupled with a short attention span, absent speech, and characteristic happy demeanor [35]. Mouse models of AS have been generated by targeted mutation of the murine Ube3a gene [37, 38]. AS mice have increased incidence of seizures, poor performance on rotarod assays, and defects in long-term potentiation (LTP), a measure of learning and memory [37-40]. These phenotypes, while similar to human AS symptoms, are less severe in mice and some are dependent on particular inbred genetic backgrounds [38, 39]. The ubiquitylation targets of UBE3A that cause the neuronal deficits in the AS mouse model and whether UBE3A ubiquitylates the same proteins in human neurons is not known.

At least four different groups have generated iPSC models of AS [26, 41-43]. Genetic subtypes of the iPSCs generated include large deletion of chromosome 15q11-q13 [26, 43], non-sense mutations in UBE3A [44], missense mutation in UBE3A [41], and paternal uniparental disomy [42]. In each of these instances, multiple clonal AS iPSC lines have been generated with the appropriate epigenotype and have been successfully differentiated into neurons. Many different neuronal subtypes are thought to be affected in AS, since UBE3A is imprinted in nearly all neuronal subtypes investigated. iPSC models of AS have yet to delve into the effect loss of UBE3A has on different neuronal subtypes.

3. 15q11-q13 Duplication

Dup15q syndrome results from either interstitial duplication or triplication of 15q11-q13 or an extra, isodicentric chromosome 15 [2] (see Fig. 2). Individuals with Dup15q syndrome commonly present with hypotonia, delay in motor skills and language development, cognitive and learning disabilities, epilepsy, and characteristic facial features. Affected individuals also usually meet the diagnostic criteria for autism. In fact, one study of autistic children identified 15q aberrations as the single most common cause, and estimates place the prevalence of 15q duplication in autism at 1–3% [45, 46]. Some individuals also present with anxiety, hyperactivity, and short stature [2]. Dup15q syndrome is more severe in individuals with idic15, presumably because they have three copies of maternal 15q11-q13, compared with two copies in interstitial duplication patients [7].

The phenotypic 15q11-q13 duplications are often of maternal origin. In contrast, individuals who inherit a paternal chromosome with an interstitial 15q11-q13 duplication usually display a normal phenotype [47, 48], although there have been reports of individuals with a paternally inherited duplication of 15q and developmental delay and/or intellectual disability [49]. It is possible that individuals with paternally inherited interstitial duplications have a sub-clinical phenotype or present with other neuropsychiatric disorders. Idic15 is almost exclusively of maternal origin, likely due to the mechanisms underlying the construction of this chromosomal aberration, thus the pathology of individuals with an analogous supernumerary chromosome 15 of paternal origin is unclear. Since the phenotypic duplications are usually of maternal origin, studies suggest that maternal duplication of UBE3A underlies many of the phenotypes associated with 15q11-q13 duplications [45, 50, 51], although no mechanism has been proposed, and more than 20 genes have been identified within the duplication region.

A sophisticated mouse model of Dup15q syndrome has been used to investigate the roles of maternal versus paternal duplications [52]. This mouse model used chromosomal engineering to generate an interstitial duplication of a 6.3 Mb region encompassing the entire imprinted domain, as well as several non-imprinted genes hypothesized to play a role in Dup15q syndrome. The duplicated allele assumes the appropriate epigenotype when transmitted either maternally- or paternally; maternal duplication resulted in increased expression of Ube3a and paternal duplication results in increased expression of Snord116 and Snord115 in brain. Unexpectedly, the presence of the duplicated allele in the mice led to autistic-like behaviors, such as reduced social interactions and vocalizations, when paternally transmitted. There was no observable phenotype when maternally transmitted. Other mouse models of Dup15q syndrome are focused on overexpression of Ube3a. At least two different transgenic lines overexpressing Ube3a have been reported. The large duplication mouse model has excellent construct validity, but tenuous face validity, while the Ube3a overexpression models show some phenotypes that may be analogous to human Dup15q phenotypes, but do not appropriately model the genetics of the disorder. Human iPSC models of Dup15q syndrome may help determine the relative roles of individual genes in Dup15q syndrome and contribute to the development of optimal mouse models.

Only one group has published iPSC models of Dup15q syndrome [53]. iPSCs from individuals with maternal and paternal interstitial duplications as well as multiple individuals with idic15 were generated. The iPSCs reported by this group also maintained the expected epigenotype in the iPSCs compared to the somatic cells sampled from the patients. Forebrain cortical neurons were also successfully generated from these iPSCs, although the disorder likely affects other neuronal subtypes.

4. Special Considerations for Disease Models of Imprinting Disorders

Appropriate parent-of-origin specific gene expression is essential for faithful modeling of chromosome 15q imprinting disorders. Germline imprints are erased and re-established in the germline, while somatic imprints are often established post-zygotically, and in some cases can be tissue-specific. Since cells do not transit through a germ cell-like state during the reprogramming process, germline imprints are not frequently disrupted during reprogramming. On the other hand, somatic imprints can be disrupted by simple maintenance in culture. Since epigenetic reprogramming disrupts both DNA methylation and histone modifications, attention to both germline and somatic imprints would be optimal, although somatic imprints can depend on the particular differentiated tissue.

The germline imprint for the chromosome 15q11-q13 locus is comprised of a region of differential DNA methylation between the parental alleles termed the PWS imprinting center [54]. The PWS-IC encompasses exon 1 of the SNRPN gene and extends into the first intron. The CpG island on the paternal allele is unmethylated, while the same region on the maternally inherited allele is methylated (Fig. 1). The PWS-IC is a promoter for the paternally expressed coding SNRPN and non-coding SNHG14 RNAs [54, 55]. The latter includes all of the snoRNAs and non-coding RNA species mentioned above (SNORD107, SNORD64, SNORD108, SNORD109A, SNORD116, IPW, SNORD115, and SNORD109B) as well as UBE3A-ATS. How the PWS-IC influences expression of the proximal cluster of genes, including MKRN3, MAGEL2, and NDN, more than 1 Mb proximal to it, is not understood. The PWS-IC is functionally conserved between mouse and human. The DNA methylation imprint at the PWS-IC is highly stable in murine and human cell culture [56-58] and even in mouse iPSCs and cloned embryos [59, 60].

Another, lesser-known regulatory element in the region is the Angelman syndrome imprinting center (AS-IC). It represses the PWS-IC on the maternal allele in the maternal germline [54] by driving transcription through the PWS-IC in oocytes. The expression leads to gene body methylation at the PWS-IC, which later is interpreted as repressive promoter methylation when the PWS-IC becomes the major SNRPN promoter in the zygote. Disruption of the AS-IC leads to a paternal epigenotype (loss of DNA methylation), and maternal transmission of this epigenetic abnormality leads to AS [61-63] by epigenetic silencing of maternal UBE3A. The human AS-IC has been mapped to a small region approximately 35 kb upstream of SNRPN. Although deletions of DNA analogous to the AS-IC fail to cause AS in the mouse [64], the function of the AS-IC is conserved between human and mouse [65]. In mouse, multiple upstream SNRPN exons promote expression through the PWS-IC. In humans, the AS-IC is most likely a single oocyte-specific promoter, but this hypothesis will require testing in human cell types. While the AS-IC acts in the maternal germline, oocytes are not finished developing until after fertilization [66]. Therefore, the activity of this germline imprint may not be fully realized in all early zygotes. This may contribute to the prevalence of individuals with AS who are mosaic for an imprinting center defect.

Somatic imprints are not well understood. Within the chromosome 15q11-q13 region, the best known somatic imprint is the repression of paternal UBE3A in neurons. Unlike most imprints, the UBE3A imprint does not involve a repressive DNA or histone modification, to our current understanding. Rather, paternal UBE3A is repressed by transcription of a paternally expressed transcript, UBE3A-ATS, in the antisense direction. This somatic imprint has been shown to be intact in human iPSCs upon neuronal differentiation [26, 41, 53]. The somatic imprints governing NDN, MAGEL2, and MKRN3 involve two differentially methylated regions at NDN and MKRN3, which are thought to be downstream of the methylation at the PWS-IC. A recent paper suggested that these somatic imprints may either be unfaithfully re-established during the reprogramming process or not stably maintained during iPSC culture [43]. Another recent paper revealed an unexpected somatic imprint that represses maternal SNRPN in human neurons. Knockdown of ZNF274 was shown to de-repress coding and non-coding transcripts of SNRPN in neural derivatives of PWS iPSCs. This de-repression involved activation of upstream SNRPN exons and did not change the methylation imprint at the PWS-IC. Together, these data suggest that we do not fully understand the somatic imprints in the chromosome 15q11-q13 region and how they are regulated in different tissues.

5. Physiological and Morphological Phenotypes

In order to understand the functional consequences of the genetic disruptions in AS, PWS, and Dup15q, it is necessary to examine the intrinsic excitability and network activity of iPSC-derived neurons. Such studies are important, as understanding the functional consequence of genetic disruptions on neuronal communication will ultimately better inform our understanding of neural pathophysiology. Characterizing the spectrum of functional deficits and unraveling primary phenotypes are critical for identifying pathways and molecules for therapeutic targets. Patient-specific iPSC-derived neurons have been used to characterize changes in electrophysiological properties, network activity, and neuronal morphology. Below we summarize results using AS and Dup15q patient-derived cell lines; however, similar studies have not yet been carried out on PWS-derived neurons.

5.1. Cellular Phenotypes in Angelman Syndrome-Derived Neurons

The functional maturation of neurons from AS patients has been characterized in a recent study using multiple control and patient-derived lines [67]. Electrophysiological analysis of control iPSC-derived neurons over 20 weeks in culture revealed maturation of physiological properties including hyperpolarization of the resting membrane potential (RMP), increases in action potential (AP) firing pattern, decreases in AP duration, increases in AP amplitude and transient potassium current density, and increases in the frequency of synaptic activity as well as the percent of synaptically active cells. Spontaneous AP activity as assayed by calcium imaging was also present throughout in vitro development in control neurons. Thus, these data provide a robust timeline for electrophysiological maturation of iPSC-derived neurons.

Compared to control neurons, AS-derived neurons failed to show the same degree of maturation across 20 weeks in culture. AS neurons showed immature resting membrane potentials and AP firing, as well as significantly fewer spontaneous AP-dependent calcium transients. Levels of spontaneous synaptic activity were also decreased in AS neurons, along with decreased dendritic branching and decreased density of dendritic spine-like protrusions. These cellular deficits resulted specifically from the loss of UBE3A. The same cellular phenotype that was seen in cells derived from AS patients with a large deletion was also seen in a line derived from an AS patient with a specific UBE3A mutation. To further confirm the critical role of UBE3A loss, knocking out UBE3A in an isogenic CRISPR-Cas9 gene-edited cell line completely replicated the AS electrophysiological phenotype. Moreover, this phenotype could be rescued by unsilencing UBE3A expression from the paternal allele with the topoisomerase inhibitor topotecan [67].

It is known from mouse models of AS that there is a critical period (early in development) in which rescue of UBE3A can also rescue behavioral phenotypes [40]. To determine whether acute loss of UBE3A causes similar phenotypes, UBE3A was knocked down using antisense oligonucleotides against UBE3A at early (6–9 weeks in culture) and late (18–20 weeks in culture), which reduced UBE3A message and protein ~50%. While both early and late UBE3A knockdown in control neurons could recapitulate the depolarization of RMP, only early knockdown altered AP firing and synaptic activity. This suggests the possibility that UBE3A loss results in a primary change to RMP that then drives changes in firing and synaptic activity, as changes in resting potential can disrupt neuronal excitability and function and thus trigger compensatory changes. To test this, control neurons were depolarized (~10 mV) with potassium chloride for the duration of their development followed by electrophysiological analysis. As with AS and UBE3A KD/KO neurons, potassium-treated control neurons displayed the entire spectrum of AS phenotypes, despite the normal expression of UBE3A.

These phenotypes are not all observed in mouse models of AS, though there has been a report of a hyperpolarized RMP later in development [68]. Such differences might be due to the differences in the developmental time periods being studied in mouse models vs iPSC-derived neuronal cultures. Changes in synaptic activity and plasticity commonly observed at mature ages in mouse models of AS may be a consequence of earlier changes in RMP, as suggested in our study. Relevant targets of UBE3A that may relate to changes in synaptic activity and plasticity have been hard to identify. It is known that ubiquitin ligases are expressed in different cell types and change their targets in a brain region- and developmental stage-specific manner. Thus, relevant targets of UBE3A may need to be identified during critical developmental periods.

Because of the relevance of neuronal and synaptic plasticity to the behavioral phenotypes in AS, Dup15q, and other neurodevelopmental disorders, one of the critical challenges to the use of human stem cell models is generating neurons with the capacity to undergo activity-dependent plasticity. Pharmacologically induced plasticity protocols, which have successfully been used in mouse brain slices and cultured neurons to elicit LTP [69], have also been used in studies of AS iPSC-derived neurons [67]. For example, plasticity induction via increased cAMP levels and enhanced activation of NMDA receptors caused a long-term elevation of spontaneous synaptic event frequency in control neurons which was absent in AS-derived neurons, paralleling LTP deficits in AS mouse models. The deficits in plasticity of synaptic transmission were paralleled by deficits in plasticity of AP firing as revealed by calcium imaging.

5.2. Cellular Phenotypes in Dup15q Syndrome-Derived Neurons

iPSC-derived neurons have also been used to characterize functional development and morphology in neurons derived from Dup15q patients. In contrast to AS-derived neurons, development of a mature RMP in Dup15q neurons was similar to controls over 20 weeks of in vitro development, shifting to more hyperpolarized potentials over time in culture. Development of AP firing was delayed in Dup15q neurons compared to controls, but was not as strongly impaired as observed in AS neurons. Dup15q neurons also showed significant increases in synaptic event frequency and amplitude compared to controls, which is maintained over 20 weeks of in vitro development, but no differences in dendritic branching. Similar increases in synaptic event amplitude have been observed in AS-derived cultures, though these cells show significant decreases in synaptic frequency [70]. Neurons derived from a patient with a maternal interstitial triplication showed a phenotype similar to isodicentric-derived Dup15q neurons, and neurons from a paternal interstitial duplication individual were similar to unaffected controls.

Synaptic plasticity deficits were also seen in the Dup15q neurons. Whereas control neurons display a long-term increase in the frequency and amplitude of synaptic events in response to pharmacologically induced plasticity, Dup15q neurons fail to show similar long-term plasticity. Another common form of plasticity is induced by long-term changes in network activity. Blocking AP firing (with the sodium channel blocker tetrodotoxin) or increasing AP firing (via GABA receptor blockade) results in changes to AMPA receptor expression. Such plasticity, termed homeostatic synaptic scaling, has been tied to a variety of neurodevelopmental syndromes in mouse models. In a study of Dup15q using human iPSC-derived neurons, control cells displayed the expected increases and decreases in AMPA event amplitude in response to manipulations in network activity as measured by both immunocytochemistry and electrophysiology. Interestingly, Dup15q neurons were not able to scale AMPA events either up or down, which could contribute to a hyperexcitable phenotype in these cells [70]. These differences were reflected in both synaptic activity and spontaneous AP firing. The goal of homeostatic scaling and plasticity is to establish a mechanism to avoid positive-feedback loops that continually increase AP firing in response to long-term potentiation. Therefore, an impaired ability of Dup15q neurons to decrease synaptic amplitudes in response to changes in network activity represents a potential mechanism for hyperexcitability in these neurons [70].

Given the seizure phenotype commonly associated with Dup15q, spontaneous AP firing in control, AS, and Dup15q neurons was also monitored. Control and AS neurons showed low levels of baseline spontaneous firing; however, Dup15q neurons fired at triple the rate of controls. Increased firing rate in Dup15q neurons was confirmed with population calcium imaging. Together these data suggest an additional mechanism of hyperexcitability specific to Dup15q neurons. Pharmacological manipulations suggest that KCNQ2 channels may be impaired in Dup15q neurons. KCNQ2 channels are potassium channels that are modulated by muscarinic receptor activation and act at subthreshold voltages as a brake on repetitive AP firing. Moreover, mutations in KCNQ2 result in benign familial neonatal convulsions, a genetic form of epilepsy. Interestingly, Dup15q neurons failed to respond to either pharmacological activation or blockade of KCNQ2 channels, suggesting impaired function and/or a reduction of these channels in Dup15q neurons. Interestingly, blockade of KCNQ2 in control neurons resulted in firing rates that were similar to the baseline firing of Dup15q neurons. In line with these data, KCNQ2 expression was significantly diminished in neurons from Dup15q patients as measured by immunostaining and flow cytometry [70].

These results establish three mechanisms of hyperexcitability in Dup15q neurons: increased synaptic event frequency and amplitude, impaired synaptic scaling, and increased AP firing due to KCNQ2 disruptions. Although primary vs. secondary targets have not been established, it is interesting to speculate that the impaired down-scaling in Dup15q neurons may contribute to the development of excessive firing and synchronous activity. Likewise, impaired up-scaling may be occluded by excessive baseline firing. Deficits in KCNQ2 channels may also contribute to synaptic scaling deficits.

5.3. Cellular Phenotypes in Prader–Willi Syndrome-Derived Neurons

Unlike Angelman and Dup15q syndromes, where many different brain regions likely have deficits that contribute to the phenotypic presentation of the disorder, PWS is thought to be caused almost entirely by deficits in the hypothalamus. Although deficits in PWS-derived cortical neurons cannot be entirely ruled out, disease relevant phenotypes are most likely confined to hypothalamic neurons. Wang et al. reported the differentiation of human pluripotent stem cells into hypothalamic-like neurons [71]. Recently, Rajamani et al., reported an improved protocol to generate hypothalamic neurons that respond to hormones like ghrelin and leptin [72]. With these improvements in hypothalamic neuron differentiation, assessment of function in this neuronal subtype should be possible for PWS-derived neurons.

6. Regulation of 15Q11-Q13 Gene Expression and Therapeutic Strategies

While many pharmaceutical therapies currently exist or are in the pipeline to treat the various symptoms of AS, PWS, and Dup15q (reviewed elsewhere), in vitro neuronal models are often not appropriate for testing their efficacy given the complex nature of the systems the drugs target. In this chapter, we instead focus on the use of iPSC models to advance therapeutic approaches aimed at targeting the genetic and epigenetic causes of these disorders.

6.1. Angelman Syndrome

In non-neuronal tissues, UBE3A is expressed from both parental alleles. However, in neurons, a long non-coding antisense transcript (UBE3A-ATS) originating from the SNURF-SNRPN promoter silences paternally inherited UBE3A [73, 74]. Therefore, loss of maternal UBE3A in AS results in complete absence of the protein in neurons. Gene replacement therapy has been proposed to introduce a functional copy of UBE3A using an adeno-associated virus (AAV) vector into AS patients and is currently under development independently by two different groups, PTC therapeutics and the Gene Therapy Center at the University of Pennsylvania. However, the intact, but silenced, paternal copy of UBE3A that already exists in AS neurons is an attractive therapeutic target since reactivation of this copy could potentially restore proper levels of UBE3A expression and function.

Studies in the AS mouse model have shed light on the mechanism whereby UBE3A-ATS silences paternal UBE3A. The Beaudet lab first showed that by engineering deletions of the Snrpn promoter, Ube3a-ATS levels were significantly reduced and paternal Ube3a expression was increased [75]. This data supported the hypothesis that Ube3a-ATS expression in cis is required for silencing paternal Ube3a. Further experiments determined that reduction specifically in Ube3a-ATS—and not in the several other transcripts that are processed from the long non-coding RNA initiating at the Snrpn promoter—is responsible for Ube3a unsilencing. By inserting a transcriptional stop cassette between Snord115 and Ube3a on the paternal chromosome, the Beaudet lab prematurely truncated the transcript and reduced expression of Ube3a-ATS [76]. This resulted in the complete unsilencing of paternal Ube3a and rescue of some AS behavioral phenotypes in the mouse model. These studies set the groundwork for developing therapeutic approaches centered on the hypothesis that reducing UBE3A-ATS activity could restore paternal UBE3A expression.

Imprinting of paternal UBE3A by the UBE3A-ATS transcript is hypothesized to be the result of transcriptional interference. The transcriptional interference model for imprinting of paternal UBE3A purports that active transcription of UBE3A-ATS on the plus strand prevents full transcription of UBE3A from the minus strand. This competition between transcripts should be amenable to manipulation by adjusting the level of transcription from either strand. In fact, preliminary work from the Chamberlain lab shows that drastically increasing the level of UBE3A-ATS transcription in AS iPSCs can completely silence paternal UBE3A in non-neuronal cells [77]. Using CRISPR-cas9 technology to delete the region between SNRPN intron 1 and SNORD115-47, directly placing UBE3A-ATS under control of the SNURF/SNRPN promoter, they were able to significantly increase UBE3A-ATS expression and silence paternal UBE3A in iPSCs. By comparing these iPSCs to other iPSC lines in which they had engineered smaller deletions to remove boundary elements regulating UBE3A-ATS expression, they determined that increasing UBE3A-ATS levels past a certain threshold resulted in complete silencing of the UBE3A sense transcript. This evidence strongly implicates a competition between the levels of UBE3A-ATS and UBE3A sense transcription in establishing the UBE3A imprint.

Two main therapeutic paradigms for unsilencing the imprinted paternal UBE3A allele in AS have been under development over the past several years. High-throughput screening experiments have identified small molecules, such as topoisomerase inhibitors, that are capable of restoring Ube3a expression. The other approach involves targeting cleavage of the Ube3a-ATS transcript using RNA-based therapeutics.

The Philpot lab used primary cortical neurons derived from mice with a Ube3a-GFP reporter knocked in to the paternal allele in high-throughput screens to identify small molecules capable of inducing expression of paternal Ube3a [78]. These studies identified irinotecan and topotecan, two FDA approved camptothecin-derived topoisomerase inhibitors, as well as 14 other compounds that inhibit either topoisomerase I or II, as drugs that could induce paternal Ube3a expression. Topotecan treatment restored paternal Ube3a expression in various neuronal populations in the mouse brain and spinal cord as well as in human AS iPSC-derived neurons. Because of the limited bioavailability of topotecan to the central nervous system, the Philpot group subsequently identified 13 indenoisoquinoline-derived topoisomerase inhibitors that all restored paternal Ube3a to levels similar to topotecan [79]. They suggest that one compound, indotecan, may be more efficient at reactivating Ube3a; however, whether indotecan has improved bioavailability over topotecan remains to be determined. Mechanistic studies determined that topotecan likely works to inhibit Ube3a-ATS by a mechanism independent of the enzymatic function of topoisomerase I [80]. Instead, topotecan works by forming stable complexes between topoisomerase I and the DNA to which it is bound (called the TOPI cleavage complex) thereby inhibiting transcription elongation.

Given that these candidate topoisomerase inhibitors are already FDA approved and widely used as cancer therapeutics, they represent an encouraging therapeutic avenue for AS. However, there are also some weaknesses to their application for AS. For example, topoisomerase inhibition was shown to not only reduce expression of Ube3a-ATS but of many other genes, including several autism candidate genes, which are important for neuronal development [81]. Additionally, inhibition of topoisomerase I was shown to impair synapse formation and function in cultured primary cortical neurons [82]. Topotecan and indotecan also have high cellular toxicity. Ongoing efforts to identify and screen additional topoisomerase inhibitors may identify candidates with safer toxicity profiles.

RNA-based therapeutics are under development or are already in clinical trials to treat a variety of genetic disorders. These approaches aim to reduce expression of target genes by reducing target mRNA levels or inhibiting protein translation. RNA-therapeutics can also be used to correct aberrant mRNA splicing caused by genetic mutations. Antisense oligonucleotides (ASOs) are one such RNA-based therapeutic currently under development for AS by several pharmaceutical companies including Ionis Pharmaceuticals, Biogen, and Roche. ASOs are single-stranded oligonucleotide sequences that can bind to and target specific RNAs for degradation [83]. ASOs reduce RNA levels by sterically inhibiting transcription or translation, by triggering RNase H-mediated degradation of the RNA in the DNA/RNA hybrid formed by the ASO and its target, or by inhibiting RNA splicing [83]. Ease of delivery to the CNS by intrathecal injection into the spinal fluid and high bioavailability make ASOs a promising approach [84]. In fact, several ASOs are currently FDA approved to treat a host of diseases including neurodegenerative disorders [85].

Building upon previous findings that reduction of Ube3a-ATS could unsilence paternal Ube3a, the Beaudet lab collaborated with Ionis Pharmaceuticals to develop ASOs to knock down Ube3a-ATS [86]. Using cultured cortical neurons from mice with YFP knocked in to the paternal allele of Ube3a, ASOs targeting the region downstream of Snord115 were screened for their ability to activate paternal Ube3a expression. Select ASOs reduced Ube3a-ATS up to 90% and increased Ube3a protein levels in primary neurons from maternal Ube3a knock-out mice up to 90% of wild type levels. When injected into the lateral ventricle of adult paternal Ube3a-YFP mice, ASOs produced a two- to fivefold increase in Ube3a-YFP RNA. Ube3a protein levels were also significantly increased throughout the brain following ASO injection; however, these levels did not reach that of normal maternal allelic expression. Encouragingly, a single ASO injection was capable of sustaining increased paternal Ube3a expression for up to 4 months. When ASOs were administered to maternal Ube3a-deficient mice, moderate restoration of paternal Ube3a protein was able to rescue two AS phenotypes—the contextual fear conditioning deficit and obesity.

One important observation from the mouse ASO studies is that only Ube3a-ATS-targeting ASOs which trigger an RNase H response were effective at restoring paternal Ube3a expression [86]. Recently, two separate studies demonstrated that the 5′- > 3′ exonuclease XRN2 is required for the depletion of RNase H cleavage products generated after ASO binding to pre-mRNAs and nuclear retained RNAs [87, 88]. XRN2 is implicated in the torpedo model of transcription termination where it degrades remaining RNA cleavage products and disengages RNA Pol II to terminate transcription [89]. Together these findings suggest that ASOs targeting UBE3A-ATS may work by disengaging RNA Pol II. Other RNA-based therapeutics that terminate UBE3A-ATS transcription by this mechanism are also likely to be effective approaches for restoring paternal UBE3A expression.

It is important to note that with any of the approaches to restore UBE3A expression in AS neurons, care must be taken to regulate the amount of UBE3A expressed since overexpression of UBE3A is associated with Dup15q syndrome. While overexpression is likely to be a concern with proposed gene replacement approaches, it is less likely to be a problem with approaches that aim to unsilence the paternal allele. In addition, approaches that unsilence the paternal allele by reducing levels of the UBE3A-ATS must not also reduce levels of the upstream transcripts including SNORD116 sufficient to cause PWS.

6.2. 15q Duplication Syndrome

While several studies support the hypothesis that maternal duplication of UBE3A underlies the majority of the Dup15q symptoms, including the autism phenotype [45, 50, 90, 91], the duplicated region also contains several other biallelically expressed genes, which may contribute to the pathophysiology of Dup15q (Fig. 1). Several non-imprinted genes with known neuronal functions are duplicated in some int. dup15q individuals and all cases of Idic15. These include a cluster of GABA receptor genes; CYF1P1, a gene whose protein product binds to and antagonizes FMRP, the gene disrupted in Fragile X syndrome [92]; TUBGCP5, a gene encoding a member of the gamma tubulin complex [93]; HERC2, another E3A ubiquitin ligase and binding partner of UBE3A [94, 95]; and NIPA1 and NIPA2, genes encoding putative magnesium transporters involved in seizures, schizophrenia, and hereditary spastic paraplegia [96, 97].

In contrast to AS, which is largely a single gene disorder, less is known about the individual contributions of the many duplicated genes to the various Dup15q symptoms. Furthermore, combinations of duplicated genes may interact to produce the overall phenotype. At the gene regulation level, sets of genes such as UBE3A and the GABA receptor cluster may be co-regulated, making targeted modulation of any individual gene difficult. Due to this complexity, far fewer therapeutic approaches have been investigated to treat Dup15q at the genetic or epigenetic level. This is expected to improve as we gain understanding of the various gene contributions through modeling of Dup15q with human iPSC-derived neurons. Theoretically, approaches to reduce the expression levels of individual duplicated genes or combinations of genes to those of normal neurons—for example, using ASOs—could be a therapeutic strategy for Dup15q.

6.3. Prader–Willi Syndrome

PWS is caused by loss of a group of paternally imprinted genes, minimally the SNORD116 host gene, which generates a set of C/D box snoRNAs, and IPW [11, 12]. Similar to the case for AS, reactivation of the silent second allele—in PWS this is the maternal allele—represents a promising epigenetic therapeutic approach.

While the germline imprint of the PWS-IC has been well studied (discussed earlier), recent work from the Lalande and Chamberlain labs suggests that a repressive complex comprised of the zinc finger protein ZNF274 and the histone methyltransferase SETDB1 represses expression of SNORD116 on the maternal allele [33, 98]. Using PWS iPSCs and iPSC-derived neurons, they showed that ZNF274 and SETDB1 preferentially bind the maternal allele at a set of six binding sites within SNORD116. This binding was associated with maternal allele-specific deposition of the repressive H3K9me3 chromatin mark. Knockdown of SETDB1 and ZNF274 by shRNA or deletion of ZNF274 using CRISPR/Cas9 technology in PWS iPSCs resulted in reduced H3K9me3 and concomitant unsilencing of maternal SNORD116.

Importantly, ZNF274-knockout-induced transcription of SNORD116 was shown to be initiated from the upstream exon promoters of SNRPN and not at the SNRPN exon 1 promoter that is regulated by the PWS-IC. Consistent with findings that transcription of the SNRPN long non-coding RNA is driven predominantly from these upstream promoters in the brain, activation of SNORD116 was even more robust following neuronal differentiation of PWS iPSCs lacking ZNF274 [33]. Ongoing work aims to further test the therapeutic capacity of reducing ZNF274 levels or disrupting binding of the ZNF274/SETDB1 repressive complex to the maternal allele in order to treat PWS.

Studies in mouse ESCs determined that an additional histone methyltransferase G9a is required to maintain maternal-specific H3K9 and CpG island methylation at the PWS-IC. Xin et al. [99] showed that homozygous deletion of Ehmt2/G9a in mouse ESCs resulted in 70% reduction of H3K9 methylation and complete loss of CpG island methylation at the PWS-IC. G9a-null ESCs also showed biallelic expression of Snrpn. Restoration of G9a function in these ESCs, through expression of a transgene, was able to restore H3K9 methylation but did not result in re-methylation of CpG islands at the PWS-IC or re-silencing of maternal Snprn. This data suggested that G9a is necessary for maintaining the already established imprint of Snrpn but is not sufficient to set up the initial imprinting.

A more recent study used mouse embryonic fibroblast cultures, which carried a maternally inherited Snrpn-GFP reporter, in a high-throughput screen to identify small molecules capable of inducing transcription of Snrpn from the maternal allele [100]. A screen of over 9000 molecules identified four compounds, which are catalytic inhibitors of G9A, that were capable of activating maternal Snprn-GFP expression. These compounds were confirmed to induce maternal expression of SNRPN mRNA as well as the snoRNA host transcripts from the SNPRN long non-coding RNA (SNHG14), including SNORD116 and SNORD115, in human PWS fibroblast cultures. When tested in a mouse model of PWS, which have a paternal deletion from Snrpn to Ube3a, G9A inhibition induced long lasting expression (12 weeks following last drug injection) of maternal Snrpn and Snord116 to almost half of normal levels in brain and liver. Drug treatment was also able to partially rescue a growth deficit and premature death phenotype of the PWS mouse. In contrast to the G9A deletion studies, activation of the maternal allele by G9A inhibitors was associated only with reduced H3K9 methylation at the PWS-IC and not a reduction in CpG island methylation. The authors suggest a model whereby G9A inhibition reduces H3K9 methylation at the PWS-IC opening up the chromatin structure to allow transcription from the normally repressed maternal allele [100].

One important consideration with epigenetic therapies for PWS is that since the distal portion of SNHG14 (the UBE3A-ATS transcript) represses paternal UBE3A in cis, any approach to induce expression of maternal SNORD116 must not also increase UBE3A-ATS levels sufficient to silence maternal UBE3A. Encouragingly, data from both the ZNF274 and G9A targeting studies do not suggest any detrimental effect on maternal UBE3A expression.

7. Synopsis

Using iPSCs to model the genomic imprinting disorders on chromosome 15q11-q13 presents unique opportunities and challenges. Three distinct neurodevelopmental disorders result from copy number variation at this locus. This necessitates diverse understanding of neuronal function in different regions of the brain. Angelman and Dup15q syndrome likely affect multiple regions of the brain. It remains to be determined whether the presentations of these syndromes result predominantly from a single brain region, or whether they are the result of an amalgamation of deficits affecting multiple brain regions. Prader–Willi syndrome, on the other hand is primarily a disorder of the hypothalamus. These disorders also necessitate broad understanding of gene regulation and gene function. Angelman syndrome is caused by loss of function from a single protein-coding gene. Prader–Willi syndrome is principally caused by the loss of a cluster of non-coding RNAs. The precise gene(s) underlying Dup15q syndrome has not yet been determined. The genetic aberrations associated with Dup15q are complex and difficult to replicate in mice. Thus, iPSC models of Dup15 syndrome will help inform development of appropriate mouse models. Faithful models of all three disorders depend heavily on maintenance of genomic imprints (both germline and somatic) during the reprogramming, maintenance, and differentiation processes. Finally, each of these disorders present a unique opportunity for therapeutic development. Individuals with Angelman and Prader–Willi syndromes may soon benefit from activation of an epigenetically silenced version of the gene involved in their respective disorders. However, careful consideration for all three disorders is essential to ensure that therapeutic interventions do not unintentionally disrupt SNORD116 or UBE3A, or lead to too much UBE3A and cause deficits associated with another neurodevelopmental disorder.