Autism

Autism is a developmental brain disorder whose definition has been extended to autism spectrum disorders (ASDs), including non-pervasive development disorder and Asperger syndrome (Levitt and Campbell 2009). The prevalence of autism is increasing and currently stands at one in every 150 individuals. Its influence on modern society is increasing, especially in developed countries, where birthrates are low and the management of children with these disorders is a serious social issue. However, the cause of autism and its pathophysiology remain unknown and require further research from a number of different approaches. Among the neuropsychiatric diseases, autism is considered as an appropriate disorder that can be approached from biological aspects. This is partly because genetic studies show that the concordance rate of autism in identical twins is over 90%, which is higher than other neuropsychiatric diseases such as schizophrenia and mood disorders (Abrahams and Geschwind 2008).

Most autism cases are sporadic and their causes are still unknown; however, 10–20% have known causes, such as (1) comorbidity with other diseases, such as fragile X syndrome, tuberous sclerosis, and Rett syndrome; (2) genome abnormalities, including copy number variations (CNVs), such as 15q11–q13 and 16q11.2; and (3) single gene mutations such as NLGN3/4, NRXN1, and SHANK2/3 (Abrahams and Geschwind 2008; Berkel et al. 2010; Cook and Scherer 2008; Kumar et al. 2008; Marshall et al. 2008; Pinto et al. 2010; Toro et al. 2010; Vorstman et al. 2006; Weiss et al. 2008). Recent genetic screening using cutting-edge technologies has revealed more CNV cases than previously expected (Cook and Scherer 2008). Given the rapid progress in current technologies, such as array comparative genomic hybridization (array CGH) and next generation sequencers, and their availability for clinical use, it is likely that more CNV cases will be diagnosed in the near future.

CNVs are important not only in explaining human heterogeneity but also because they are responsible for several diseases, especially neuropsychiatric disorders including autism. Human chromosome 15q11–q13 is the second most frequent region to contain chromosome abnormalities, probably because it contains large repeat sequences. Depending on its copy variant, this region causes different diseases, such as Prader–Willi syndrome (PWS), Angelmann syndrome (AS), and autism, all sharing ASD features. Comparison of mouse models of 15q copy variants with phenotypes of genetically engineered mice of each gene, with or without imprinting, will provide some clues to the pathophysiology of autism.

Mouse chromosome seven is syntenic to human chromosome 15q11–q13

Chromosomal abnormalities were known as causes of autism before the era of current genome technology. Among them, human 15q11–q13 is the most notable chromosome region (Vorstman et al. 2006) (Fig. 1). This region is also known to be an imprinting region (Chamberlain and Lalande 2010a, b; Nicholls and Knepper 2001). Thus, there are three classified regions: paternally, maternally, and biallelically expressed regions. Human chromosome 15q11–q13 is conserved in mouse chromosome seven and it is believed that the imprinting status is conserved between mouse and human (Yang et al. 1998). Knockout (KO) mice for each gene corresponding to a human counterpart have been generated and studied (Table 1).

Fig. 1
figure 1

Genes on human chromosome 15q11–q13. Clusters of non-coding RNA, such as HBII-52 and HBII-85, are omitted. The original data are taken from UCSC Genome Browser (http://genome.ucsc.edu)

Full size image
Table 1 Genes on human chromosome 15q11–q13 and the phenotypes of their genetically engineered mice
Full size table

Paternally expressed genes

In the paternally expressed regions of the mouse chromosome, Mkrn3, Magel2, Necdin (Ndn), Snrpn-Snurf, and huge clusters of small nucleolar RNAs (snoRNAs) are present. Magel2 is involved in hypothalamic functions, and its deficient mice show abnormal circadian rhythm, body weight, and hormonal secretion, such as testosterone (Bischof et al. 2007; Kozlov et al. 2007; Mercer and Wevrick 2009). Ndn was first identified as a gene related to neural differentiation (Maruyama et al. 1991). Ndn KO mice were generated from multiple laboratories; however, they display different phenotypes, from no abnormality to respiratory distress and lethality, similar to PWS, depending on the strain of mouse (Gerard et al. 1999; Kuwako et al. 2005; Muscatelli et al. 2000; Tsai et al. 1999a). Muscatelli et al. observed a neonatal lethality of variable penetrance in Ndn mutants, and they showed a reduction in both oxytocin-producing and luteinizing hormone-releasing hormone (LHRH)-producing neurons in the hypothalamus and emotional and cognitive-related behaviors reminiscent of PWS (Muscatelli et al. 2000). Interestingly, oxytocin is known to be related to social behavior and is expected to be a candidate for therapeutics of ASD (Insel 2010). Ndn deficiency induces perinatal serotonergic alterations that affect the maturation and function of the respiratory network, inducing respiratory problems (Zanella et al. 2008). Serotonergic and catecholaminergic projections are reduced in the Ndn null embryo (Lee et al. 2005). Provided that serotonin is involved in symptoms of autism, Ndn is a candidate for further studies to reveal its involvement in abnormal social behavior. Additionally, from studies using Ndn KO mice, it is reported that Ndn is involved not only in TrkA signaling (Kuwako et al. 2005) but also promotes tangential migration of neocortical interneurons from the basal forebrain (Kuwajima et al. 2010). The fact that the neocortical interneurons are important for development of inhibitory neural circuits also makes Ndn a possible candidate for causing an imbalance of the excitatory/inhibitory ratio. Thus, although it is still controversial whether Ndn is a cause of PWS, it is an intriguing target gene for further analyses in terms of developmental brain disorders, including ASDs. Snrpn is a small nuclear ribonucleoprotein N and the Snrpn promoter is located in a maternally methylated CpG island within an imprinting center (IC). Mutations of Snrpn are not sufficient to induce PWS in mice, but a larger deletion including the PWS-IC and Ndn show phenotypes common to PWS infants (Yang et al. 1998). Paternal transmission of IC deletions results in a PWS-like phenotype in mice (Bielinska et al. 2000). The human PWS-IC could acquire a methylation imprint in mouse oocytes but the imprint is lost in somatic cells, suggesting that the human imprinting center demonstrates conserved acquisition, but diverged maintenance, of imprinting in a mouse model for AS imprinting defects (Johnstone et al. 2006). The main promoter of Snrpn is not required for the function of the PWS-IC, and the 4.8-kb deletion induces a severe phenotypic effect and a partial imprinting defect; however, this defect is less severe than that with the 42-kb deletion of the PWS-IC (Bressler et al. 2001; Yang et al. 1998). Paternal deletion from Snrpn to Ube3a in the mouse causes hypotonia, growth retardation, and partial lethality (Tsai et al. 1999b). The non-coding RNA region includes HBII-52 and HBII-85. Knockout mice of MBII-85, a mouse homolog of HBII-85, show growth retardation, increased anxiety, and hyperphagia (Ding et al. 2008; Skryabin et al. 2007). Recently, HBII-85 was reported as a cause of PWS (Duker et al. 2010; Sahoo et al. 2008). On the other hand, knockout mice of MBII-52, a mouse homolog of HBII-52, display increased RNA editing of the serotonin 2c receptor (5-HT2cR) and less locomotor activity (Doe et al. 2009). In terms of the relationship between serotonin and autism, MBII52 is an interesting candidate gene.

Maternally expressed genes

Maternally expressed genes in chromosome 15q11–q13 include Ube3a and Atp10a. Ube3a is an E3 ubiqutin ligase and is known to be a causative gene for AS (Kishino et al. 1997; Matsuura et al. 1997). Ube3a knockout mice show impaired motor function, inducible seizures, learning deficits (Jiang et al. 1998; Miura et al. 2002), and sleep disturbance (Colas et al. 2005). The ataxic and long potentiation phenotypes are rescued by reduction of αCaMKII inhibitory phosphorylation (van Woerden et al. 2007). Knock-in (KI) Ube3a-YFP fusion reporter mice show that Ube3a is imprinted in all regions of the brain, but is biallelically expressed in glial cells (Dindot et al. 2008). Maternally Ube3a-deficient mice show altered linking behavior, which has been associated with cerebellar control (Heck et al. 2008). Recently, Ube3a was extensively characterized in neurons and found to be involved in synaptic plasticity (Greer et al. 2010; Margolis et al. 2010; Sato and Stryker 2010; Yashiro et al. 2009). A genetic study also reported the involvement of Ube3a in ASD (Glessner et al. 2009). Ube3a is the primary candidate gene for causing the phenotypic features in duplication 15q autism. In contrast, Atp10a, an aminophospholipid translocase, has been less comprehensively studied than Ube3a, and its imprinting status remains unclear (DuBose et al. 2010; Kayashima et al. 2003). Mice with a radiation-induced maternal deletion, including Ube3a and Gabrb3, are a model of obesity and type two diabetes (Dhar et al. 2004). Mice with a large maternal deletion from Ube3a to Gabrb3, produced by chromosome engineering, show altered ultrasonic vocalization, impaired learning and memory, and display parent of origin effects (Jiang et al. 2010).

Non-imprinting genes

In the conventional biallelic region, there is a cluster of GABAA receptor subunits (Gabrb3, Gabra5, Gabrg3), Oca2, and Herc2. The GABAA receptor is one of the major neurotransmitter receptors and forms the most notable target in this region because an imbalance between excitatory and inhibitory neurons during development is a pathophysiological hypothesis of developmental brain disorders, including ASD (Rubenstein and Merzenich 2003). Of the GABAA receptor subunits, the GABAA β3 subunit (encoded by the Gabrb3 gene) is well studied. Its deficient mice show epileptic seizures, deficits in learning and memory, poor motor skills (DeLorey et al. 1998; Homanics et al. 1997), as well as sleep disturbance, such as lower REM sleep time and greater EEG delta power than the wild-type (WT) (Wisor et al. 2002). The level of norepinephrine in the spinal cord of Gabrb3 KO mice was less than that of WT mice, suggesting that β3 subunit-containing GABAA receptors mediate the trophic effects on a subpopulation of spinally projecting noradrenergic neurons (Ugarte et al. 2001). Furthermore, Gabrb3 N265M KI mice display altered general anesthetic actions (Jurd et al. 2003). Gabrb3 KO mice show expansion of the plexus of locus coeruleus dendrites, with a concurrent reduction in muscle tone and marble burying behavior (Hashemi et al. 2007). They also exhibit hypoplasia of cerebellar vermal lobules and impaired social and exploratory behaviors in addition to deficits in non-selective attention (DeLorey et al. 2008). Survivors of pan-neuronal Gabrb3 knockouts are overtly normal, whereas those of forebrain selective knockouts have reduced fertility, are hyperactive, and some become obese (Ferguson et al. 2007). An association between a specific GABAA β3 receptor defect and autism has been reported recently (Delahanty et al. 2010). Thus, Gabrb3 is also an intriguing candidate gene in the region with respect to autism. Gabra5 KO mice display enhanced spatial learning, decreased amplitude of inhibitory post-synaptic current (IPSC), and enhanced amplitude of paired-pulse facilitation of excitatory postsynaptic potential (EPSP) (Collinson et al. 2002). While Gabra5 KO mice exhibit a normal behavioral phenotype for fear-associated contextual learning (Cheng et al. 2006) and no differences in long-term potentiation (LTP) in hippocampal slices (Collinson et al. 2002), they show reduced threshold for the induction of LTP via mechanisms that are independent of synaptic inhibition (Martin et al. 2010). On the other hand, α5 (H105R) KI mice, in which a partial reduction of α5 GABAA receptor restricted to CA1-3 subfield of the hippocampus was achieved, exhibit facilitated trace fear conditioning (Crestani et al. 2002). Moreover, H105R mutant mice show attenuated prepulse inhibition (PPI) and increased spontaneous locomotor activity (Hauser et al. 2005). Herc2 contains a COOH-terminal HECT domain and functions as an assembly factor for the complex of an E3 ubiquitin ligase (Bekker-Jensen et al. 2010). Mutant mice of Herc2 show defects in growth (Lehman et al. 1998), but the physiological functions of Herc2 in the brain remain unknown.

Genes between BP1 and BP2

In addition to genes between breakpoints BP2 and BP3, there are four highly homologous genes (Nipa1, Nipa2, Cyfip1, and Tubgcp5) in between BP1 and BP2 (Chai et al. 2003), which appear to result in developmental behavioral problems when disturbed (Doornbos et al. 2009; Tam et al. 2010; van der Zwaag et al. 2010; von der Lippe et al. 2011). Cyfip1 (cytoplasmic FMRP interacting protein, also known as Sra-1) is an intriguing candidate among them because it binds with the Fragile X protein (Schenck et al. 2001). Microarray analysis using lymphoblastoid cells from males with autism due to a 15q11–q13 duplication reveal that Cyfip1 is upregulated (Nishimura et al. 2007). Although there is no mouse model of Cyfip1 at this moment, it would be interesting to make both Cyfip1 deletion and transgene animals since individuals with 15q duplications and AS deletion patients have a larger deletion that includes Cyfip1 (Sahoo et al. 2006).

Genomic diseases in human chromosome 15q11–q13

In humans, a paternal deletion of 15q11–q13 causes PWS in which the patient shows hypotonia, obesity, cognitive deficit, and behavioral problems, including obsessive-compulsive disorder (Cassidy and Driscoll 2009; Chamberlain and Lalande 2010b). Maternal deletion of this region causes AS, which includes motor dysfunction, frequent seizures, learning disability, absent speech, and characteristic happy demeanor (Van Buggenhout and Fryns 2009). Both PWS and AS share ASD features (Schanen 2006; Sutcliffe and Nurmi 2003). On the other hand, duplication of this region is known to be the most frequent cytogenetic abnormality in ASDs (Hogart et al. 2010; Vorstman et al. 2006). Autism with 15q duplication syndrome was first reported as partial trisomy of chromosome 15 (Gillberg et al. 1991). It is widely believed that maternally duplicated cases cause autism, but paternally duplicated ones do not (Bolton et al. 2001; Browne et al. 1997; Cook et al. 1997; Mao et al. 2000; Repetto et al. 1998; Roberts et al. 2002). On the other hand, cases with developmental delay are documented with paternal duplications (Bolton et al. 2004; Mao et al. 2000; Mohandas et al. 1999; Roberts et al. 2002; Veltman et al. 2005). A recent screening using multiplex ligation-dependent probe amplification (MLPA) did not detect maternal duplication, suggesting that ASD cases in maternal duplication are not as frequent as previously assumed (Depienne et al. 2009). Recent advances in genetic analyses, such as array CGH, and their clinical application are expected to detect larger numbers of clinical cases for 15q duplication.

A mouse model for 15q duplication

Using a chromosome-engineering technique (Mills and Bradley 2001; van der Weyden and Bradley 2006; Yu and Bradley 2001), a model for human chromosome 15q duplication has been generated (Nakatani et al. 2009). The duplication was confirmed in ES cells by Southern blotting and in mice by fluorescence in situ hybridization (FISH) and CGH arrays. Southern blot analysis using methylation sensitive and insensitive restriction enzymes, and a probe near the IC, indicated that allele-specific methylation is conserved in mice with the duplication. As the region is imprinted, the duplication is classified into two types: patDp/+ mice, in which the duplicated region is derived from the paternal allele; and matDp/+ mice, in which the duplication is derived from the maternal allele.

Both patDp/+ and matDp/+ mice breed normally and are fertile (Nakatani et al. 2009). The patDp/+ male mice begin to show an increase in body weight after 15 weeks and the body weight of patDp/+ mice is significantly greater than that of the WT after 20 weeks. The fact that both decreased (PWS) and increased (15q duplication) gene expressions of PWS regions lead to obesity is contradictory, but rather intriguing. As described above, Atp10a seems to be involved in metabolism and body weight (Dhar et al. 2004). Further, some of forebrain specific KO mice of Gabrb3 also become obese (Ferguson et al. 2007). The PWS region is involved in aspects of metabolism and above a certain critical threshold, the amounts of the genes and their products, regardless of any increase or decrease, may result in an abnormal pathway of metabolism.

Histological analyses, either at the macroscopic level or at the microscopic level, using simple staining such as hematoxylin–eosin or Bodian staining, detect no significant difference between mice with the duplication and WT mice (Nakatani et al. 2009). Recent studies reveal that synaptic adhesion molecules, and their associated proteins, are known to be rare mutation cases in autism (Betancur et al. 2009; Bourgeron 2009; Toro et al. 2010). For example, Shank3 mutant mice display abnormality in spines (Bozdagi et al. 2010; Peca et al. 2011). Thus, it will be quite interesting as to whether spine or synaptic abnormality is observed in 15q duplication mice or other models.

Autistic-like behavior of 15q duplication mice

Behavioral analyses are essential for animal models of autism because major phenotypes of ASDs are behavioral abnormalities and the diagnosis is made by three major phenotypes: impairment in social interaction, impaired social communication, restricted interest and behavior, or resistance to change. Mouse behavioral tests corresponding to human autistic phenotypes have been reported (Moy and Nadler 2008; Silverman et al. 2010). The systematic behavioral analyses were performed for 15q duplication mice (Nakatani et al. 2009). First, in the three-chamber test for social interaction, WT mice display more time spent near a novel mouse than an empty cage, while no difference of time spent around the novel mouse and the empty cage is observed in patDp/+ mice. In comparison with time between an inanimate object and an empty cage, both patDp/+ and WT mice spend more time near the object than empty cage, indicating that patDp/+ mice are able to see the object. Interestingly, WT mice spend more time around a novel mouse compared with an inanimate object, whereas there was no difference in time spent between the novel mouse and the inanimate object in patDp/+ mice. On the other hand, there was no significant difference between matDp/+ and WT mice in the three-chamber test. These results indicate that patDp/+ mice exhibit impaired social interaction. Similarly, Gabrb3−/− mice spend less time in the proximity of unfamiliar mice than do their control counterparts in a similar social behavior test, although some aspects of social curiosity appear to remain intact (DeLorey et al. 2008).

Second, patDp/+, but not matDp/+, mice display abnormal ultrasonic vocalizations (USVs) (Nakatani et al. 2009). When the pups are separated from their dams, they produce USVs. In WT pups, the numbers of USVs reach a peak around postnatal day five (P5) and then gradually decrease until P14, when their eyes are open and they can communicate through the visual pathway. In contrast, the numbers of USVs remain high at P7 or even at P14 in patDp/+ mice, suggesting that the developmental course of USVs in patDp/+ mice is delayed compared with WT. These results demonstrate that patDp/+ mice may be developmentally abnormal in comparison with WT, although the larger number of USVs in patDp/+ pups may reflect higher anxiety and fear in response to isolation stress. No difference of USVs between matDp/+ and WT mice was observed. The increased USVs were also observed in mice with a maternal deletion from Ube3a to Gabrb3 (Jiang et al. 2010).

Third, ritualistic behavioral inflexibility seen in autistic patients is examined by reversal tasks using learning mazes. In the Barnes maze, WT, patDp/+, and matDp/+ mice exhibit intact spatial learning and memory. In a reversal test, in which the target is altered to the opposite position, WT and matDp/+ mice learn the new target position, whereas patDp/+ mice return to the original position more often than WT and matDp/+ mice, indicating that patDp/+ mice display resistance to change or perseveration (Nakatani et al. 2009). patDp/+ mice also show similar behavior in the Morris water maze. These results suggest that patDp/+ mice exhibit behavioral inflexibility.

In addition to these phenotypes, patDp/+ mice display novelty-induced anxiety and decreased exploratory activity (Tamada et al. 2010). Interestingly, anxiety is also observed in paternal duplication of 15q11–q13 in humans (Mao et al. 2000; Veltman et al. 2005). Mice with maternal deficiency for just the Ube3a gene also show decreased exploratory behavior (Allensworth et al. 2011). In several tests involving food deprivation, such as the novelty suppressed test, radial maze, and Y-maze, patDp/+ mice display peculiar behavior (Tamada et al. 2010). Even though patDp/+ mice must be hungry after food deprivation, they do not rush to feed and take more time to proceed to feed than WT mice, and finally perform some tasks that are equivalent to the WT. This behavior is likely to reflect a lack of desire. patDp/+ mice also show less marble burying behavior (Tamada et al. 2010), a method commonly used to assess obsessive-compulsive behavior, which is also seen in Gabrb3 KO mice (Hashemi et al. 2007).

Abnormal serotonin in 15q duplication mice

The mouse model for 15q duplication can be considered an artificial founder mouse to unravel which gene and/or region in 15q11–q13 is critical for autistic phenotypes. A series of bacterial artificial chromosome (BAC), yeast artificial chromosome (YAC), or other engineered mice with longer insertions that cover the duplicated region will provide useful materials for further investigation. Alternatively, genetic rescue experiments crossed with a series of knockout or deficient mice will be required to identify the gene.

The primary benefit of the mouse model is that it allows us to study abnormality at the molecular and cellular levels. Major behavioral phenotypes appear in patDp/+ mice but not in matDp/+ mice (Nakatani et al. 2009); therefore, genes in the maternally imprinted region can be regarded as candidates for abnormal behaviors observed in patDp/+ mice. Among them, MBII-52, a mouse brain specific snoRNA 52 in a huge cluster of non-coding RNAs, would be an interesting candidate for the following reasons. First, MBII-52 is a paternally expressed gene. Second, HBII-52, a human homolog of MBII-52, is reported to be involved in posttranscriptional modification of the serotonin 2c receptor (5-HT2cR) (Kishore and Stamm 2006). Third, serotonin (5-HT) is well known to be related to the pathophysiology of autistic phenotypes (Whitaker-Azmitia 2005). Fourth, 5-HT2cR is the only G-protein coupled receptor that undergoes A to I RNA editing, resulting in an amino acid substitution in the second loop of the cytoplasmic domains (Seeburg 2002), which may affect G-protein coupling efficiency. MBII-52 RNA displays an approximately two-fold increase in expression in the patDp/+ mice brain compared with matDp/+ and WT mice (Nakatani et al. 2009). Pyrosequencing analyses revealed that the RNA editing ratio in potential editing sites of 5-HT2cR in patDp/+ mice was higher than that in matDp/+ or WT mice. This suggests that increased snoRNA in the patDp/+ brains may affect the RNA editing of 5-HT2cR. Furthermore, a specific agonist for 5-HT2cR induces higher intracellular Ca2+ concentration in primary cultured neurons derived from patDp/+ mice compared with matDp/+ and WT neurons, indicating that serotonergic signals in patDp/+ neurons are different from matDp/+ and WT neurons. This result suggests that alterations in RNA editing of 5-HT2cR may affect serotonergic signals in neurons and that this alteration in serotonergic signals in patDp/+ neurons may recapitulate abnormal behavior in patDp/+ mice.

Neurochemical analyses revealed abnormal levels of 5-HT and its metabolites in the adult cerebellum and midbrain of the duplication mouse. Interestingly, 5-HT in patDp/+ mice was also observed to decrease in all brain regions tested during postnatal weeks 1–3 (Tamada et al. 2010). These results suggest that abnormal 5-HT levels during developmental stages affect development of neurons and neural circuits, and may recapitulate the abnormal behavior seen in patDp/+ mice. Interestingly, the mean 5-HT concentrations are larger in Ndn KO mice than in WT, and the morphologies of 5-HT fibers are altered (Zanella et al. 2008). Combined with these findings, Ndn in the region may be involved in the metabolism and development of 5-HT. Ndn facilitates TrkA signaling for sensory neuron survival (Kuwako et al. 2005). The molecular mechanisms by which Ndn might affect 5-HT metabolism and development remain unknown.

Phenotypes (in)dependent on copy variants

Mice with a 15q duplication and other genetically engineered mice for genes in 15q11–q13 share some phenotypes, such as obesity, social behavior, and 5-HT concentration. Serotonin concentration seems to correlate with gene dosage of Ndn (Tamada et al. 2010; Zanella et al. 2008); however, complex phenotypes such as metabolism, weight control, or social behavior are unlikely to be dependent on simple gene dosage or copy number variants, but above a certain critical threshold, increases or decrease of gene expression may produce the abnormal pathway of these phenotypes.

The idea that phenotypes are dependent on copy variants may not be correct and their underlying mechanisms are unlikely to be simple in humans. In contrast to the interstitial duplication of human chromosome 15q, termed idic (15) syndrome individuals, the presence of a supernumerary marker chromosome formed by the inverted duplication of proximal chromosome 15, resulting in partial tetrasomy 15q, often display increased, yet inappropriate, social behaviors (Battaglia et al. 2010).

A gap between the mouse model and the human patient

The mouse model for 15q duplication as an animal model for human neuropsychiatric disease fulfills construct validity, in which it possesses the same chromosomal abnormality as human patients. Thus, the model can be regarded as a humanoid mouse model for autism in one aspect (Takumi 2010). In terms of humanized models, Tabuchi et al. report a KI mouse for an NLGN3 R541C mutant (Tabuchi et al. 2007). However, the same KI mouse independently generated by Dr. Heintz’s group displays no significant or mild social abnormality (Chadman et al. 2008). Further humanized model mice with rare mutations of other molecules, such as serotonin transporter G56A KI mice (Veenstra-Vanderweele et al. 2009), will be expected to display the phenotypes seen in humans.

The allele-specific phenotypes that appeared in 15q model mice will imply some biological significance that can provide a general concept. However, it is surprising that patDp/+ mice, but not matDp/+ mice, display abnormal social behavior (Nakatani et al. 2009) because maternally duplicated cases in humans are likely to result in an ASD phenotype, while paternal duplications do not (Cook et al. 1997). Although allele-specific methylation seems to be conserved, at least around the IC, in mice with a duplication, the possibility that the imprinting status in other regions is different between human and mouse is not excluded. The timing and mechanism of the expression of the downstream non-coding Snurf-Snrpn exons seem to be different between human and mouse (Chamberlain et al. 2010). Recent systematic analyses of the mouse epigenome reveal that Ube3a and a GABA receptor subunit are differentially imprinted in different brain areas (Gregg et al. 2010). Alternatively, epigenetic status other than DNA methylation may be different in mice from that in humans. Quantitative analysis of GABRB3 transcripts in PWS and AS brain samples show evidence for a paternal expression bias when a biparental 15q11–q13 contribution is lacking (Hogart et al. 2007). The Ube3a-ATS is thought to regulate expression by turning off Ube3a on the paternal allele. This may cause difference between maternal and paternal duplication animals as well and should be investigated. At the very least, there is a significant variability and diversity of clinical phenotypes and gene expression in human chromosome 15q11–q13 duplication syndrome (Hogart et al. 2009; Piard et al. 2010). Further detailed epigenetic analyses both in humans and mice will provide an insight into the imprinting mechanisms.