Inhibition of histone deacetylation rescues phenotype in a mouse model of Birk-Barel intellectual disability syndrome

Mutations in the actively expressed, maternal allele of the imprinted KCNK9 gene cause Birk-Barel intellectual disability syndrome (BBIDS). Using a BBIDS mouse model, we identify here a partial rescue of the BBIDS-like behavioral and neuronal phenotypes mediated via residual expression from the paternal Kcnk9 (Kcnk9pat) allele. We further demonstrate that the second-generation HDAC inhibitor CI-994 induces enhanced expression from the paternally silenced Kcnk9 allele and leads to a full rescue of the behavioral phenotype suggesting CI-994 as a promising molecule for BBIDS therapy. Thus, these findings suggest a potential approach to improve cognitive dysfunction in a mouse model of an imprinting disorder.

B irk-Barel intellectual disability (ID) dysmorphism syndrome (BBIDS or KCNK9 imprinting syndrome, OMIM 612292) is typically associated with congenital central hypotonia, developmental delay, intellectual disability, severe feeding problems, and hyperactivity 1,2 . The disease is inherited autosomal dominantly with maternal-only transmission 1 , as the KCNK9 gene is embryonically paternally silenced (imprinted) in man and mouse. It encodes the potassium channel subunit TASK3, which dimerizes to form two-pore domain potassium (K2P) leak channels 3 . The mouse Kcnk9 gene maps to an imprinted cluster on mouse chromosome 15 together with further imprinted genes, i.e., the brain-specific maternally expressed genes Ago2, Chrac1, and Trappc9 and the paternally expressed Peg13 gene 4 .
Kcnk9 mRNA expression is widespread in the central nervous system 5,6 ; in rodents with notably high levels in cerebellar granule neurons, the locus coeruleus (LC), the dorsal raphe nuclei, hippocampal CA1 and CA3 pyramidal neurons, and several hypothalamic nuclei 7,8 . Homozygous deletion of Kcnk9 in the mouse (Kcnk9KO hom ) reduces the resting potassium conductance and enhances spike firing accommodation in adult cerebellar granule neurons 9 . Kcnk9KO hom mice further display increased nocturnal motor activity, cognitive deficits, as well as a reduced sensitivity to inhalation anesthetics and the cannabinoid receptor agonist WIN55212-2 mesylate [10][11][12] . More recently, RNAi-based knockdown of Kcnk9 and expression of a dominant-negative mutant KCNK9, which had been associated with the human disease phenotype were shown to impair neuronal migration during mouse cortical development 13 . However, the phenotype of mice with heterozygous deletion of the active maternal Kcnk9 allele (Kcnk9KO mat ) thus mimicking BBIDS has not yet been characterized.
DNA methylation is well established as a key player in imprinting regulation by multiple studies, which characterized the kinetics and mechanisms of methylation reprogramming at imprinting control regions (ICRs). The critical involvement of post-translational histone modifications in transcriptional regulation, in general, is also well accepted, but much less explored for ICRs 14 . It is assumed that an interplay between DNA methylation and histone acetylation is essential for proper erasure and resetting of imprints in the germline as well as selective imprint maintenance during postzygotic reprogramming 15 . A differentially methylated region (DMR) which is methylated on the maternal allele and assumed to be involved in Kcnk9 gene regulation has been only identified in the promoter region of Peg13 16,17 . Promoter CpG islands of Kcnk9 are unmethylated, but display high levels of active histone H3 lysine 4 monomethylation (H3K4me1) and histone H3 lysine 27 acetylation (H3K27ac) chromatin marks in brain tissues 4 .
Here, we characterize the behavioral and neuronal phenotype of mice with heterozygous deletion of the active maternal Kcnk9 allele (Kcnk9KO mat ). We demonstrate partial behavioral rescue in these animals compared with full knockout animals and show that epigenetic manipulation stimulates Kcnk9 pat expression sufficiently, to rescue the behavioral phenotypes, thereby opening new avenues for treatment of cognitive dysfunctions in BBIDS.

Results
Deletion of Kcnk9 leads to impaired behavior. To assess behavioral deficits along the BBIDS phenotype in Kcnk9KO mice, we performed in vivo experiments in adult wild type (WT) as well as Kcnk9KO mat and Kcnk9KO hom mice with a deletion of Kcnk9 exon 2 as previously described 18 (Fig. 1a). Following the strictly monoallelic expression pattern of Kcnk9 in mouse brain (<1% paternal expression) 3 , we expected largely concordant phenotypes in mice carrying a deletion of both Kcnk9 alleles (Kcnk9KO hom ) or only the maternal Kcnk9 allele (Kcnk9KO mat ), respectively.
No differences in behavior between Kcnk9KO hom , Kcnk9KOmat , and WT littermates were seen in elevated plus maze and open field testing for anxiety ( Supplementary Fig. 1a, b) and in the rotarod-test for motor coordination ( Supplementary Fig. 1c). Given the described intellectual disability in BBIDS patients, we then examined whether working memory was affected in Kcnk9KO mice. Spontaneous alternation relies on the natural tendency of rodents to explore a novel environment and can be quantified using a Y-maze task. The ability to remember the immediately preceding choice is considered an indicator for active working memory. In the Y-maze, spontaneous alternation was defined as consecutive entries into all three arms without revisiting an arm. The average percentage of spontaneous alternation made by Kcnk9KO hom , Kcnk9KO mat , and WT mice across a 10-min trial was analyzed and revealed a significant reduction of spontaneous alternation by about 10% in both Kcnk9KO hom and Kcnk9KO mat mice compared with WT littermates (Fig. 1b). There was no significant difference in the extent of the working memory impairment between Kcnk9KO hom and Kcnk9KO mat mice (Fig. 1b). Our findings indicate an impaired working memory in Kcnk9KO hom and Kcnk9KO mat mice, which is consistent with previously reported observations in Kcnk9KO hom mice 10 .
Circadian rhythms in mammals are endogenously coordinated oscillations of biological parameters such as the sleep/wake cycles with an overall period length of about 24 h 19 . KCNK9 channels are expected to contribute to the in vivo electrical activity of neurons in those brain regions associated with the regulation of circadian rhythms and arousal [10][11][12]20 . Spontaneous motor activity analysis during light (resting) and dark (active) phase was performed, resembling day and night in diurnal humans, respectively. A significantly increased overall locomotor activity during the dark phase was found in Kcnk9KO hom and Kcnk9KO mat mice compared with WT controls, describing an exaggerated nocturnal activity in Kcnk9KO animals (Fig. 1c), which is in consensus with previous observations 10 . Surprisingly, an intermediate phenotype for dark phase activity was observed in the Kcnk9KO mat animals with statistically significant differences to both WT and Kcnk9KO hom animals. No significant differences between genotypes were observed in the overall locomotor activity during the light phase (Fig. 1c, Supplementary  Table 1).
Together, these data demonstrate that the loss of Kcnk9 in mice results in the impairment of behavioral parameters resembling components of the BBIDS phenotype. Interestingly, we found an intermediate phenotype for nocturnal locomotor activity in animals with a loss of only the actively expressed maternal allele (Kcnk9KO mat ) suggesting the involvement of the silenced paternal allele.
Non-canonical imprinting of Kcnk9 in the mouse brain. In human and mouse brain KCNK9 was reported to be monoallelically expressed from the maternal allele 3,4 while the paternal allele is silenced. To elucidate if the intermediate phenotype of Kcnk9KO mat animals was due to residual expression from the paternal allele, we analyzed the parent-of-origin allele-specific expression pattern of Kcnk9 in different brain regions of F1 hybrid animals from crosses between C57BL/6 (B6) and Mus musculus castaneus (Cast/Ei) mouse strains [(C57BL/6xCast/Ei) F1] 3 . As expected, we observed a predominant expression of the maternal Kcnk9 allele in all analyzed brain regions (Fig. 1d). However, we also detected significant expression from the repressed paternal allele, which represented 1-14% of all    transcripts depending on the brain region analyzed (Fig. 1d, suppl. Fig. 2c). Highest paternal expression was observed in the LC (Fig. 1d). Accuracy of the LC tissue-punches was demonstrated through elevated gene expression levels of tyrosine hydrolase (TH) in LC samples ( Supplementary Fig. 2a, b). These data demonstrate a brain region-specific leakiness of the imprint on the Kcnk9 pat allele with a peak in the LC. This leakiness might be responsible for the observed intermediate phenotype in Kcnk9KO mat animals and suggested a particular importance of LC Kcnk9 expression for locomotor activity during the active (dark) phase.
Kcnk9 knockdown in the LC induces elevated nocturnal activity. To test the role of functional Kcnk9 expression in LC neurons for the control of nocturnal activity, we bilaterally infused AAV vectors for expression of eGFP and shRNAmir sequences, either scrambled (pAAV-Syn-shRNAmir-scrambled-EF1a-eGFP) or specifically targeting Kcnk9 mRNA (pAAV-Syn-shRNAmir-Kcnk9-EF1a-eGFP), into the LC of WT mice under stereotactic control (Fig. 2a). Efficiency of the used shRNA sequence was pre-tested in vitro (Fig. 2b) and validated by RT-qPCR from mouse brain tissue (Fig. 2c). Targeting efficiency and selectivity of stereotactically guided infusion was demonstrated by eGFP immunohistochemistry (Fig. 2a). Twenty-one days after injection, animals were tested for circadian activity and working memory. A significant and selective increase of nocturnal activity was detected in the shRNAmir-Kcnk9 injected animals compared with age-matched control animals injected with the scrambled shRNAmir virus (Fig. 2d). These results demonstrated that altering local expression of Kcnk9 in the LC was sufficient to selectively affect dark-phase activity in mice. Furthermore, it identified LC as an important neural hub for mediating the behavioral effects of altered Kcnk9 expression, which warranted further mechanistic analysis of these neurons. Interestingly, a clear trend toward impaired working memory was also observed in shRNAmir-Kcnk9 injected animals (Fig. 2e), suggesting that Kcnk9 expression also controls working memory-related activity of LC neurons.
Increased dark-phase LC pacemaking activity in Kcnk9KO hom but not Kcnk9KO mat mice. Electrical activity of LC neurons drives periods of wakefulness and arousal [21][22][23] . Indeed, selective optogenetic activation of LC neurons revealed their causal role in sleep-to-wake transitions and locomotor arousal by demonstrating that sustained 3-Hz LC neuronal stimulation enhanced spontaneous locomotion (total track length) by about 50% 24 . As we detected a similar increase of spontaneous locomotion during the active (dark) phase of Kcnk9KO hom mice compared with WT controls, we reasoned that KCNK9-dependent differences in LC pacemaking activity might contribute to this phenotype.
By recording from synaptically isolated (including inhibition of somatodendritic alpha2-autoreceptors), spontaneously active LC neurons in brainstem slices from adult mice, we observed a dark phase-selective, about 70% increase of pacemaker frequency in LC neurons from Kcnk9KO hom mice ( Fig. 3a-d). In contrast, no significant dark-phase increase of LC pacemaker frequency was observed in WT animals (Fig. 3c, d). These results demonstrate that some degree of KCNK9 channel expression is necessary to selectively dampen enhanced pacemaker activity during the dark phase, which is present in Kcnk9KO hom mice. Interestingly, LC neurons from Kcnk9KO mat animals displayed no significant increase in pacemaker activity in the dark phase (Fig. 3c, d), but a significant difference in pacemaker activity in the dark was observed between Kcnk9KO hom and Kcnk9KO mat animals. This indicated that a certain level of functional expression of KCNK9 channels from the Kcnk9 pat allele in LC is present in Kcnk9KO mat animals to prevent the full nocturnal in vitro pacemaker frequency increase. This level of Kcnk9 expression might at the same time not be sufficient to dampen in vivo LC activity, which is also driven by synaptic inputs from neuronal networks, in line with the observed intermediate behavioral phenotype regarding dark-phase locomotion in Kcnk9KO mat mice. In line with this argument, we show below that boosting further expression of KCNK9 channel subunits from paternal alleles indeed also restores the WT motor activity phenotype.
CI-994 activates the paternally repressed Kcnk9 allele. The paternally inherited Kcnk9/KCNK9 gene is epigenetically silenced yet structurally unimpaired. Our behavioral and gene expression experiments suggest a contribution of paternally expressed Kcnk9 to the intermediate phenotype of the Kcnk9KO mat mice and the possibility of upregulation of the paternal allele in the case of loss of the maternal allele. We, therefore, speculated that exogenous application of epigenetic modulators might alter the structure at the Kcnk9 promoter region and result in further derepression of the Kcnk9 pat allele. We hypothesized that this could fully compensate for the loss of the maternal allele and rescue the BBIDSlike phenotype of Kcnk9KO mat mice.
To investigate the paternally derived Kcnk9 transcript levels after epigenetic drug treatment, we isolated E14 mouse primary cortical neurons (mPCNs) from WT and Kcnk9KO mat animals from crosses between WT males and Kcnk9KO hom females (Fig. 4a). We then chose six different compounds representing different classes of epigenetic modulators for treatment (Fig. 4b).
For further investigations, we focused on the benzamide-based second-generation histone deacetylase inhibitor (HDACi) CI-994, a selective inhibitor of class I HDACs, which is a potent inhibitor of HDAC1 and 3 isoenzymes relative to HDAC6 and HDAC8 25 . CI-994 showed the most efficient upregulation of Kcnk9 pat expression in Kcnk9KO mat mPCNs (Fig. 4c). Furthermore, a previous study reported that intraperitoneal administration of CI-994 in WT mice resulted in long-lasting CI-994 levels in the brain without affecting the overall behavioral phenotype of mice 26 .
The effect of CI-994 in Kcnk9KO mat mPCNs was further tested in a dose-response experiment. Kcnk9KO mat mPCNs were treated with increasing concentrations of CI-994 for 24 h, and Kcnk9 expression was analyzed using RT-qPCR in comparison to DMSO-treated Kcnk9KO mat and WT control cells (Fig. 4d). As expected, the Kcnk9 pat expression in DMSO-treated Kcnk9KO mat mPCNs was significantly reduced compared with that of WT mPCNs. After CI-994 treatment of Kcnk9KO mat mPCNs, we observed a strong linear correlation between the increase of Kcnk9 pat expression and CI-994 dosage suggesting a specific, dose-dependent effect of CI-994. Notably, Kcnk9 expression in 80 µM CI-994-treated Kcnk9KO mat mPCNs exceeded that of DMSO-treated WT mPCNs in vitro (Fig. 4d). Treatment of Kcnk9KO mat mPCNs with 20 µM CI-994 showed a consistent increase of Kcnk9 pat expression even after 10 days suggesting a prolonged and stable in vivo effectiveness (Fig. 4e). Cell viability analysis after treatment of Kcnk9KO mat mPCNs with 10 or 20 µM CI-994 did not reveal any toxic effect ( Supplementary Fig. 4c). Intriguingly, CI-994 treatment did not affect the expression of other nearby imprinted genes within the imprinted cluster on mouse chromosome 15 in vitro ( Supplementary Fig. 3b).
CI-994 rescues the behavioral phenotype of Kcnk9KO mat animals. The detection of Kcnk9 pat allele expression in several brain regions, prominently among them the LC, as well as the observation of intermediate phenotypes in Kcnk9KO mat animals led us to hypothesize that epigenetic manipulation could further stimulate paternal gene expression and thereby boost the phenotypical rescue.
For testing, either DMSO (100%) or CI-994 (30 mg/kg of body weight in 100% DMSO) were injected daily over 14 days in the peritoneum (Fig. 5a). After injections, Kcnk9 pat expression was induced up to~3-fold compared with DMSO-treated controls in several brain regions including the cerebellum, hippocampus, pons, hypothalamus, olfactory bulb, and the LC (Fig. 5b). Importantly, this drug-induced rescue also included brain regions like the hippocampus, where only a small degree (<1%) of spontaneous expression from paternal alleles was observed in the Kcnk9KO mat mice. In an allele-specific assay using WT (C57BL/ 6xCast/Ei)F1 hybrids, we found that CI-994 affects specifically expression of the Kcnk9 pat allele ( Supplementary Fig. 4b). Only a very weak upregulation of global (maternal plus paternal) Kcnk9 expression was seen in WT animals after CI-994 treatment (Fig. 5c, Supplementary Fig. 4c). Importantly, and in line with the in vitro observations, expression of only Kcnk9 but no other of the genes located within the imprinting cluster on mouse chromosome 15 was found to be increased after CI-994 injection in vivo (Fig. 5d).
To assess whether CI-994-mediated enhancement of Kcnk9 expression also promotes behavioral recovery, WT, Kcnk9KO mat , and Kcnk9KO hom mice were subjected to behavioral testing after treatment with either DMSO (100%) or CI-994 (30 mg/kg of body weight). In the Y-maze task, CI-994 treatment led to a significant increase of spontaneous alternation in the Kcnk9KO mat mice compared with DMSO-treated Kcnk9KO mat control mice (Fig. 5e). Post hoc testing for multiple comparisons showed no difference between CI-994-treated Kcnk9KO mat animals and either DMSO-or CI-994-treated WT animals (Fig. 5e). WT mice treated with either DMSO or CI-994 did not exhibit deficits in working memory and significant differences in the Y-maze. By contrast, no increase of spontaneous alternation was seen in the CI-994 treated Kcnk9KO hom animals suggesting that the paternal allele silenced in the Kcnk9KO mat animals carries the effect of the treatment (Supplementary Fig. 5a).
We then asked whether CI-994 treatment also alters nocturnal activity in Kcnk9KO mat or in Kcnk9KO hom mice. Indeed, Kcnk9KO mat mice showed a significant decrease in horizontal nocturnal activity during the 12-h dark phase after CI-994 treatment (Fig. 5f). Again, post hoc analysis of multiple comparisons showed no significant difference between CI-994 -treated Kcnk9KO mat and DMSO-or CI-994-treated WT animals suggesting full rescue of the nocturnal hyperactivity phenotype in the Kcnk9KO mat animals. No drug effect was observed in Kcnk9KO hom animals and total locomotor activity in the light phase was not altered by CI-994 treatment irrespective of the genotype of the animals (Fig. 5f, Supplementary Fig. 5b).
CI-994 interferes with H3K27 acetylation at the Kcnk9 locus. Imprinting of the Kcnk9/KCNK9 gene is thought to be regulated by a maternally methylated germline differentially methylated region (DMR) in the promoter of the Peg13/PEG13 gene 3 . Acetylation of histone 3 lysine 27 (H3K27ac), monomethylation of histone 3 lysine 4 (H3K4me1) as well as DNA methylation at the maternally methylated DMR in the human PEG13 promoter region have been suggested to control human KCNK9 expression 4 (Fig. 6a, Supplementary Fig. 6a-d). CI-994 is a second-generation class I histone deacetylase inhibitor inhibiting HDACs1 and 3 with high specificity 25 . To clarify the epigenetic mechanism underlying the induction of the Kcnk9 pat allele through CI-994 we investigated the methylation status of two subregions of the Peg13-DMR (Peg13 DMR1 and Peg13 DMR2) in adult hippocampus and LC of DMSO-and CI-994-treated Kcnk9KO mat mice using bisulfite pyrosequencing. Methylation levels of about 40-50% typical for imprinted gene DMRs in somatic cells were determined for both DMRs in the two brain regions of Kcnk9KO mat mice. These methylation levels were not significantly altered by CI-994 treatment of Kcnk9KO mat mice ( Supplementary Fig. 6c, d). Our results indicate that the upregulation of Kcnk9 mRNA in various brain regions of Kcnk9KO mat mice after CI-994 treatment was independent of the DNA methylation status at the Peg13-DMR.
The human PEG13-DMR has been further demonstrated to bind CTCF-cohesin which conveys chromatin looping between a brain-specific enhancer region, marked by H3K27ac and H3K4me1, and the PEG13 and KCNK9 promoters to supposedly control brain-specific KCNK9 expression 4 . To determine the effect of CI-994 on the Kcnk9 locus in the hippocampus and LC of Kcnk9KO mat mice, we performed Chromatin immunoprecipitation (ChIP)-qPCR for H3K27ac and H3K4me1 marks following CI-994 or vehicle treatment. Using several public mouse brain ChIP-seq datasets, we did not find an orthologous region in the  mouse displaying a brain-specific enhancer chromatin signature similar to that identified at the human imprinted domain on chromosome 8q24. Thus, we focused our analysis on the promoter and intronic region of the Kcnk9 locus, which are enriched for both marks in the mouse brain ChIP-seq datasets (Fig. 6a). Due to the small size of the LC, we opted to use a novel low-input ChIP-method 27 for both regions (LC and hippocampus) and validated this by conventional ChIP-qPCR in the hippocampus ( Supplementary Fig. 7b). Upon CI-994 treatment, we observed a significant more than two-fold increase in H3K27ac depositions at the two Kcnk9 regions both in the hippocampus and LC of Kcnk9KO mat mice (Fig. 6b). The levels of H3K27ac were also increased in the hippocampus of WT littermates however the increase was not as prominent as in the Kcnk9KO mat animals ( Supplementary Fig. 7i, j). An intergenic region, lacking H3K27ac or H3K4me1 modifications, was used as negative control ( Supplementary Fig. 8c-j). Looking at allelespecific H3K27ac deposition in the hippocampus of (C57BL/ 6xCast/Ei)F1 hybrid animals using two different SNPs in intron 1 of Kcnk9, we found that, as expected, the maternal allele had a higher degree of H3K27ac association than the paternal allele. CI-994 treatment of Kcnk9KO mat mice, however, led to a higher degree of H3K27ac deposition at the paternal than at the maternal Kcnk9 allele (Fig. 6c).
In addition, we tested the deposition of H3K4me1 both in the LC and hippocampus to assess if histone acetylation inhibition leads to additional changes in other histone marks. Treatment with CI-994 led to only a slight, but not significant increase of H3K4me1 deposition at the Kcnk9 promoter region (Supplementary Fig. 7f-h). Our findings indicate that CI-994 treatment affects histone acetylation consistently at the promoter and intronic regions of Kcnk9 of the investigated brain tissues and has diverging effects on H3K4me1 deposition at the investigated Kcnk9 locus.
Taken together, our findings demonstrate that the paternal allele of Kcnk9 is not fully silenced in the brain, further derepressed in the case of loss of the maternal allele and substantially activated upon exogenous epigenetic modulation. Our data provide evidence for a promising therapeutic effect of the class I HDAC inhibitor CI-994, in a mouse model for BBIDS with maternal loss of Kcnk9.

Discussion
Kcnk9/KCNK9 is an imprinted gene in mouse and man that is expressed from the maternal allele. Pathogenic variants present on only the maternal allele, therefore, cause disease in patients 1,2 . In a study comparing full Kcnk9 knockout animals (Kcnk9KO hom ) with maternal Kcnk9 knockout animals (Kcnk9KO mat ) and WT littermates in a behavioral battery, we surprisingly found an intermediate phenotype in nocturnal locomotor activity in Kcnk9KO mat animals. Furthermore, electrophysiological analysis of tyrosine hydroxylase (TH)-positive neurons in the LC, a region involved in regulation of arousal and locomotor activity in mice, displayed significantly increased spontaneous firing frequencies only during the active dark phase in full knock-outs in contrast to WT, while it was fully rescued in the Kcnk9KO mat animals. Allele-specific expression analysis of Kcnk9 mRNA in (C57BL/6JxCast/Ei)F1 hybrid animals revealed significant residual paternal expression of up to 14% of total Kcnk9 expression in the LC in the Kcnk9KO mat animals. Daily i.p. injection of the second-generation HDAC inhibitor CI-994 could increase this expression up to three-fold and resulted in a full rescue of the behavioral phenotypes of Kcnk9KO mat animals.
We show here that an increase in dark-phase locomotion in the Kcnk9KO hom animals is associated with an increase of in vitro pacemaker frequency in the LC. Deletion of the active maternal Kcnk9 allele in the mouse is supposed to lead to a phenotype comparable to the full knockout. Surprisingly, however, our study identified an intermediate behavioral phenotype in nocturnal (hyper)activity in the Kcnk9KO mat animals, as well as a fully rescued cellular phenotype (LC firing frequency) (Figs. 1c, 3). Extended allele-specific expression analysis using (C57BL/ 6JxCast/Ei)F1 mouse hybrids demonstrated residual Kcnk9 expression from the paternal allele in several brain regions with a significant peak in the LC of 14% of total Kcnk9 expression coming from the paternal allele.
The LC plays an essential role in arousal and sleep-wake transitions 28 and, as the origin of noradrenergic neurons projecting into the hippocampus and the prefrontal cortex, is also involved in the regulation of long-term and working memory, two essential building blocks of higher cognition/brain functions 29 . In a knockdown experiment using shRNAmir containing AAVs targeting the Kcnk9 mRNA specifically injected into the LC we found a similar behavior phenotype in circadian activity as seen in the full and maternal Kcnk9 knockout animals demonstrating an essential and causal contribution of Kcnk9 expression within the LC in the control of nocturnal locomotor activity. Interestingly, our electrophysiological data show full rescue of in vitro pacemaker firing frequency through Kcnk9 pat expression in the LC in Kcnk9KO mat animals on the one hand and only a partial rescue of dark-phase activity on the other. This might indicate different levels of Kcnk9 expression in order to control on one hand isolated pacemaker function, which typically operates with very small ionic currents of a few picoamps, compared   with in vivo function for behavioral control, where neurons are embedded in active networks and larger currents are necessary to alter their firing patterns. Our data suggest that the influence of KCNK9 on pacemaker frequency and potentially in vivo activity could be feasible as a therapeutical strategy for attention deficit and hyperactivity disorder (ADHD). However, future studies should directly record from LC neurons in vivo in awake behaving mice as well as from Kcnk9KO mat and Kcnk9KO hom animals to better define the functional role of KCNK9/TASK3 channels.
Nocturnal activity values (measured as distance traveled, in cm) of WT mice showed considerable variability between experiments. This variability may be attributed to various factors such as different treatments (naïve vs. injected) of the analyzed mice as well as different timepoints and environmental conditions at which the experiments were performed. It is well-known, that animal behavior heavily relies on environmental conditions within animal facilities and that these conditions can slightly change with time 30 . For example, in our animal facility, maintenance work was carried out in the time between the circadian locomotor activity experiments for the first manuscript version and those including treatment of Kcnk9KO hom mice with CI-994 for the revised manuscript version. In light of these limitations, we strongly argue that a comparison of nocturnal activity values makes only sense within the same experiment but not across the different experiments performed.
A recent study has suggested histone methylation as a potential therapeutic target for another imprinting disorder, the Prader-Willi syndrome (PWS) 31 . The authors showed that UNC0642, a selective inhibitor of euchromatic histone-lysine N-methyltransferase-2, derepressed the maternal copies of paternally expressed PWS candidate genes and promoted the growth and survival of mouse pups with a paternal deletion of the PWS region without altering DNA methylation at the PWS-ICR. We show here that derepression of the paternal allele in Kcnk9KO mat animals leads to a significant increase in Kcnk9 expression in selected brain regions and that this is enough to substantially influence the behavioral phenotype in domains likely to be regulated by LC activity. Furthermore, we found that the secondgeneration HDAC inhibitor CI-994 is another attractive option for treatment of imprinting disorders such as BBIDS. CI-994 treatment of Kcnk9KO mat mice as a mouse model of BBIDS substantially activated the repressed paternal allele of Kcnk9 in several brain areas resulting in a significant increase in Kcnk9 expression. This led to the full rescue of specific domains of brain function impaired in BBIDS. Our data for the first time shows successful phenotypic rescue of impaired brain function in a mouse model for an imprinting disorder using an epigenetic modulator. Similar to the findings in the UNC6042-treated mouse model of PWS 31 , activation of the Kcnk9 pat allele was not accompanied by changes of DNA methylation at the Peg13 DMR, which is assumed to play a key role in regulation of Kcnk9 imprinting also in the mouse 3 . Notably, the bisulfite technique did not allow us to discriminate between 5-methylcytosine and 5-hydroxymethylcytosine levels possibly affected by CI-994. Though, our data strongly argue that CI-994-mediated derepression of the Kcnk9 pat allele is primarily caused by an increase of H3K27ac at regulatory elements in the promotor and intronic region of the Kcnk9 gene. However, the treatment did not display significant differences in H3K27ac deposition between the LC and hippocampus tissues suggesting a region-independent effect of the HDAC inhibitor. Such effects have commonly been reported in multiple genome-wide studies investigating the effect of HDAC inhibitors 32 . Interestingly, CI-994 as a histone deacetylase inhibitor did not only selectively increase H3K27ac but also slightly increased H3K4me1 at the Kcnk9 promoter region. This is not unexpected since crosstalk mechanisms between histone-modifying enzymes often affect the binding and activity of further histone-modifying enzymes. For instance, treatments with HDAC inhibitors were reported to affect globally the histone methylation marks specifically of H3K4me1 32 . These observations are repeatedly described in the literature suggesting a broad effect of HDAC inhibitors on chromatin structure and transcription 33,34 .
In general, epigenetic modulation is supposed to rather unspecifically target numerous genes and their expression in the genome. Side effects of all sorts including tumor development and metabolic dysregulation are believed to be likely with such substances. Also, interference with brain function leading to cognitive impairment and mental dysfunction cannot be excluded. However, valproic acid, a first generation HDAC inhibitor has been used in anti-epileptic therapy for decades without severe tumorigenic or metabolic long-term effects. CI-994 is a second-generation, specific class I HDAC inhibitor. Although a relatively broad effect of CI-994 on gene expression has been demonstrated in vivo, it is being used in several clinical trials for anti-cancer therapy having proven inhibitory potency of epithelial-mesenchymal transition processes 35 . Accumulating evidence also suggests that CI-994 provides neuroprotective effects in the central nervous system and cell survival in vitro 36 . Regarding brain function in WT mice no effect of CI-994 treatment has been seen in the open field 26 . In our experiments, no significant influence of two weeks CI-994 injection was seen on working memory or circadian locomotor activity in WT animals. Furthermore, an increase of Kcnk9 pat expression after CI-994 injections was only seen in selected brain regions, but not in, e.g., cortex or prefrontal cortex, and mainly in the maternal knockout animals (only very weakly in WT animals). Allele-specific ChIP-qPCR of (C57BL/6JxCast/Ei)F1 hybrids showed that the CI-994 effect is more pronounced on the paternal allele (Fig. 6c). This leads to a visible increase in expression of the paternal but not the maternal allele (Supplementary Fig. 4a) and might be due to the high acetylation status of the active maternal allele. Furthermore, CI-994 treatment did not affect the expression of other nearby imprinted genes within the imprinted cluster on mouse chromosome 15 in vitro and in vivo. Taken together these data suggest a rather specific epigenetic action of CI-994 in Kcnk9KO mat animals, which might make CI-994 safe for usage in BBIDS patients. However, so far all CI-994 related studies have been carried out in vitro or in adult animals. If used in patients, early treatment during development might be more efficient. In order to make this possible, further studies on CI-994 specificity and possible adverse and teratogenic effects in young mice have to be conducted.
In summary, we show here that brain region-specific upregulation of the paternally silenced Kcnk9 allele particularly in the LC of Kcnk9KO mat animals modulates nocturnal hyperactivity, a central phenotype of Kcnk9 knockout animals. We further show that the class I HDAC inhibitor CI-994 leads to a significant additional increase of Kcnk9 pat expression in several brain regions and fully rescues behavioral alterations in Kcnk9KO mat animals. This rescue is associated with an increase of H3K27 histone acetylation at the promotor and intronic region of the Kcnk9 gene but not with changes in DNA methylation at the differentially methylated region in the Peg13/Kcnk9 locus. Our data suggest epigenetic modulation by CI-994 as a promising therapeutic strategy in patients with BBIDS and, more generally, derepression of imprinted gene alleles as a sustainable approach for the treatment of imprinting disorders.

Methods
Mice. All experimental procedures were performed in accordance with institutional animal welfare guidelines and were approved by the ethical committee of the state government of Rhineland-Palatinate, Germany (ID: 23 177-07/G 17-1-022). Kcnk9KO hom mice were provided by Florian Lesage, Institut de Pharmacologie Moléculaire et Cellulaire, Valbonne, France. The gene targeting strategy of Kcnk9KO hom mice was based on a cre-mediated deletion of exon 2 encoding pore domains P1 and P2, transmembrane domains M2-M4 as well as the cytoplasmic C-terminus 18 . Breeding of two heterozygous animals revealed homozygous, heterozygous and WT animals at a proportion of 1:2:1. WT littermates were used as controls. WT and heterozygous Kcnk9KO mice with inactivation of the maternal Kcnk9 allele (Kcnk9KO mat ) were obtained by crossing male WT mice with female Kcnk9KO mat mice. Finally, breeding of homozygous female and heterozygous male mice resulted in Kcnk9KO mat and Kcnk9KO hom mice. The mice were kept under specific-pathogen-free (SPF) conditions on a 12 h light/12 h darkness cycle in standard polystyrene cages with free access to water and food. Using tail tip DNA, mice were genotyped by assessing exon 2 excision using the primers flanking this region (F, 5′-TGCGAGCTTCAGAGAGAGGATG-3′ and R, 5′-ATGCTC-TAATCTCCAGTCTG-3′) producing fragments for WT and mutant alleles. Additional primers within exon 2 were applied to generate a control product for the WT allele (F, 5′-CACCACGCCATGTACTTCCT-3′ and R, 5′-GGACCG-GAAGTAGGTGTTCC-3′). Male mice at 8-10 weeks of age were used for expression analysis and to investigate the behavioral phenotype with and without drug treatment. Allele-specific expression analyses were carried out with total RNA from F1 offspring derived from crosses between female WT C57BL/6J or Kcnk9KO hom mice and male WT Mus musculus castaneus (Cast/Ei) mice.
Behavioral testing. Littermate WT and Kcnk9KO mice were tested in 4-8 cohorts of mice, with testing beginning at 8-10 weeks of age. All experimenters were blinded to the genotype of the mice throughout the studies and behavioral analyses. Immediately after the behavioral tests, animals were sacrificed and whole brains were rapidly removed and incubated in RNAlater® (Sigma) for subsequent expression analysis.
Drug administration in vivo. Nine weeks old male Kcnk9KO mat and WT mice were injected intraperitoneally once daily with CI-994 (ApexBio) or dimethyl sulfoxide (DMSO) (control vehicle) for 7 consecutive days before behavior experiments were initiated (Fig. 5a). On days of testing, mice were injected 2 h before behavior experiments. A single injection contained either 20 µl DMSO or 20 µl CI-994 (35 mg/kg) dissolved in 100 % DMSO.
Toxicity/viability test in CI-994 treated mPCNs. mPCNs were washed with PBS and fixed with 4% paraformaldehyde for 2 h. Subsequently, cells were incubated for 20 min in 0.4% Triton X-100 in PBS at 37°C and further rinsed three times with PBS. After blocking with 10% sheep serum at 37°C for 1 h, cells were washed for a further three times with PBS. Next, neurons were incubated overnight at 4°C with anti-neuron-specific nuclear protein NeuN monoclonal antibody (cat. # ab104225; dilution 1:200 in PBS), followed by five washes with PBS for 5 min each. The neurons were subsequently incubated with an Alexa Fluor 488-conjugated goat anti-mouse secondary antibody (cat # ab150077; dilution 1:200 in PBS) for 1 h at 37°C and washed for a further five times in PBS. After staining, cells on coverslips were mounted onto microscope slides and counterstained with mounting medium containing 2 μg/mL Hoechst 33258 (Cat # H3569).
Tissue collection. Whole mouse brains were removed and incubated in RNAlater® (Sigma) and stored for 2 days at 4°C. After the dehydration process, several brain regions were dissected and further processed or stored at −80°C. For LC samples, coronal brain sections (80 μm) between bregma −5.40 and −5.52 37 were generated, and 0.51 µm punches bilaterally of each brain region were collected with the Brain Punch Tissue Set (Leica). The accuracy of the LC tissue collection was validated by performing tyrosine hydroxylase (TH) expression analyses and comparing the expression levels to those of WT cerebellum and hippocampus samples. Samples with non-adequate TH levels were excluded from analysis. Tyrosine hydroxylase, which serves as a norepinephrine marker, was expressed~400-fold higher compared with hippocampus samples, indicating a precise tissue dissection (Supplementary Fig. 2).
Cell culture. Kcnk9KO mat primary cortical neurons (mPCNs) were isolated from E14 embryos derived from matings between female Kcnk9KO hom mice and WT males. After collection of brains, cortices were dissected out and mechanically separated into single cortical cells through resuspension. The mPCNs were plated on Poly-L-Ornithine (Sigma)-and Laminin (Sigma)-coated plates and maintained in culture medium containing Neurobasal medium (Gibco), supplemented with 2% B27 plus vitamin A (Gibco) and 1% Glutamax (Gibco), in a humidified incubator at 37°C and 5% CO 2 .
In vitro drug treatment. We cultured murine primary cortical neurons (mPCNs) for 4 days in 6-or 12 well-plates and treated them with epigenetic compounds diluted in culture medium for 24 or 72 h. To investigate the duration of Kcnk9 unsilencing, mPCNs were cultured for 3 days and treated with 20 μM CI-994. After 2 days of treatment, cells were washed with 37°C culture medium. After 1 day, some cells were harvested. Medium change was conducted after additional three days. Six days later, on 10 days, remaining cells were harvested and analyzed via RT-qPCR.
AAV generation and Kcnk9 knockdown in WT mice. Viral constructs were purchased from Sirion Biotech (Martinsried, Germany). Generation and evaluation of shRNAmir sequence targeting the gene Kcnk9 were performed using SIRION's RNAiONE platform as previously described 38 . Ten different 21mer shRNAmir sequences against Kcnk9 were selected using algorithms listed in Fig. 2a. Corresponding shRNA template oligonucleotide cassettes were cloned into a shuttle plasmid under the control of the human U6 promoter. The coding region of the cDNA was PCR-amplified and cloned into the validation vector pVal downstream of the EGFP coding region resulting in pVal-target. The promoter-shRNA regions of the shRNA shuttle plasmids along with a shuttle plasmid encoding for a NT-shRNA as control were then transferred into pVal-target by recombinational cloning. Cells were then transfected with the validation plasmids and were incubated for 48 h. Total RNA was subsequently isolated and 1 μg was reverse transcribed using a mixture of random hexamer and oligo-dT primer. The silencing efficiency of each shRNA was determined by quantification of the EGFP-target cDNA expression levels relative to NT-shRNA control vector. The expression cassette was then cloned into an AAV transfer vector (pAAV-Syn-shRNAmir-Kcnk9-EF1a-eGFP and pAAV-Syn-shRNAmir-scrambled-EF1a-eGFP) and ITRs integrity was then validated by restriction analysis. 293T cells were utilized for AAV production, which were generated via co-transfection and transferred to 1E13 VG in vivo buffer. We infused 8-week-old C57BL/6J mice (CLR) with either AAV1-hSyn-shRNAmir-Kcnk9-EF1a-eGFP (Sirion, Germany) to knockdown the Kcnk9 expression or AAV1-hSyn-shRNAmir-scrambled-Kcnk9-EF1a-eGFP (Sirion, Germany) as a control bilaterally in the LC. Animals where anesthetized with isoflurane (1-2%) and received a subcutaneous injection of carprofen (4 mg/kg). Three hundred nanoliters of virus was infused with a rate of 100 nl/min in the LC (AP: −5.4, ML: ±1.0, DV: 3.0, in mm, relative to bregma) under stereotaxic control (Kopf Instruments, USA) using a 1 µl Hamilton syringe with a glass capillary. Twenty-one days after surgery spontaneous alteration and circadian rhythm were tested.
RNA Isolation, cDNA synthesis, and RT-qPCR. The total RNA extraction with TRIzol reagent from brain tissue was prepared as suggested by Invitrogen Life Technologies. High Pure RNA Tissue Kit (Roche) and High Pure RNA Isolation Kit (Roche) were applied to extract total RNA from small tissue and cell samples using spin columns. The purity, quantity, and integrity of the RNA were measured with a NanoDrop 1000 spectrophotometer and an Agilent 2100 Bioanalyzer. The cDNA samples were synthesized from 200 to 1000 ng total RNA using the Pri-meScript TM RT Master Mix cDNA (Takara) as recommended by the manufacturer. Quantitative real-time PCR (qRT-PCR) was carried out using SYBR® Premix Ex Taq™ II (Tli RNaseH Plus) and 10 μM primers (final concentration), according to the manufacturer's instructions. Allele-specific RT-qPCR reactions were performed on an ABI StepOnePlus TM Real-Time PCR System using exon-exon junction primers with the following conditions: 95°C/30 s, 40 cycles of 95°C/5 s, 68.5°C/30 s, 72°C/30 s. All additional RT-qPCR reactions were performed with the following conditions: 95°C/30 s, 40 cycles of 95°C/5 s, 60°C/30 s, 72°C/30 s. All reactions were measured in triplicates, and median cycles to threshold (Ct) values were used for analysis. The housekeeping gene Impa2 was used to normalize against experimental genes, and relative gene expression was determined using the 2 −ΔΔCT methods 39 . Impa2 forward primer 5′-CGTGCGGGACAAATCATCA G-3′ and reverse primer 5′-CGTGCGGGACAAATCATCAG-3′.
Chromatin Immunoprecipitation (ChIP)-qPCR. ChIP experiments were performed as described in Akhtar et al. with minor modifications for brain tissue processing 27 . Briefly, brain tissues were dissected in cold PBS and homogenized with a pestle. Following dissection, tissues were fixed with 1% formaldehyde in PBS for 10 min at room temperature, followed by quenching of the fix with 125 mM glycine for 5 min. Fixed cells were resuspended in 140 mM RIPA (10 mM Tris-Cl pH 8.0, 140 mM NaCl, 0.1 mM EDTA pH 8.0, 1% Triton X-100, and 0.1% SDS) and subjected to 18 cycles of sonication on Bioruptor Pico (Diagenode), with 30 secs "ON"/ "OFF" at high settings. After sonication, samples were centrifuged at 14,000 × g for 10 min at 4°C and supernatant was transferred to a fresh tube. The extracts were incubated overnight with 1 µg (1:500 dilution) of H3K27ac (Abcam, Cat # ab4729), H3K4me1 (Active motif, Cat # 39297), H3 (Abcam, Cat # ab1791) or IgG control (Diagenode, Cat # C15410206) antibody at 4°C with head-over tail rotations. After overnight incubations, 20 µl of blocked Protein A and G Dynabeads (Diagenode) were added to the tubes and further incubated for 3 h to capture the antibodies. Bead separation was carried out using a magnetic rack and washed as following; once with 140 mM RIPA (10 mM Tris-Cl pH 8.0, 140 mM NaCl, 0.1 mM EDTA pH 8.0, 1% Triton X-100, and 0.1% SDS), four times with 250 mM RIPA (10 mM Tris-Cl pH 8.0, 250 mM NaCl, 0.1 mM EDTA pH 8.0, 1% Triton X-100, and 0.1% SDS) and twice with TE buffer pH 8.0 (10 mM Tris-Cl pH 8.0 and 0.1 mM EDTA pH 8.0). For reversal of cross-linking, samples were RNase-treated (NEB) and subjected to Proteinase K treatment for 12 h at 37°C and at least 6 h at 65°C after the immunoprecipitation. After proteinase K treatment, the DNA was extracted using the phenol-chloroform method. After precipitating and pelleting, DNA was dissolved in 30 µl of TE buffer pH 8. The real-time PCR was performed using SYBR Green chemistry (ABI) on the ABI StepOnePlusTM Real-Time PCR System. The input and chromatin immunoprecipitated material were processed identically across the samples. The fluorescent signal of the amplified DNA was normalized to H3 binding or input with the following primers: Kcnk9 promoter F 5′-CGTGTGCGCTACATCTCCTA-3′ and R 5′-ATTCGCCGGTTCCTTCTACT-3′; Kcnk9 intron F 5′-AGGGCAGATGCTTAAGAGGA-3 and R 5′-CATCTGTT CTGTACCCCATCC-3′; Intergenic region 3 (Ig3) F 5′-ATGCCCCTCAGCTATC ACAC-3′ and R 5′-GGACAGACATCTGCCAAGGT-3′. The low amount chromatin immunoprecipitation (TAF-ChIP) was performed according to Akhtar et al. 27 .
Allele-specific ChIP followed by pyrosequencing. For allele-specific ChIP analysis, DNA of WT and Kcnk9KO mat (C57BL/6JxCast/Ei)F1 hybrid mice, treated either with DMSO or CI-994, was amplified in a PCR applying one non-biotinylated forward primer: 5′-AAATTCGCCGGTTCCTTCTAC-3′ and one biotinylated reverse primer: 5′-GAGATGTAGCGCACACGAAGC-3′. To distinguish the origin of the alleles we subsequently analyzed the PCR products in an allele-specific manner, using two neighboring SNPs, rs240891437 and rs47242380 (Ensembl; GRCm38.p5). The quantification of enrichment was performed using the pyrosequencing technique. A Kcnk9-specific pyrosequencing primer: 5′-AAGGAATG GGTGTGC-3′ was designed and pyrosequencing was carried out on a PyroMark Q96 instrument (Qiagen).
Behavioral analyses. Brief descriptions of the behavioral tests are provided below. Male mice between 8 and 10 weeks of age were used for behavioral tests, as described below.
Circadian locomotor activity. Mice were individually placed in transparent polypropylene cages (38 × 22 × 15 cm) with standard beddings. Food and water were available and accessible at all times. The cages were placed in a sound-prove room under dimed light during the light phase and under infrared LED lights during the dark cycle. The horizontal locomotor activity of the mice was recorded in the light and in the dark phase using a video tracking software (EthovisionXT, Noldus software). The circadian rhythm activity was measured after 48 h acclimatization for 24 h and analyzed with EthovisionXT, Noldus software. The results show the distance traveled (in cm) in the light and in the dark phase.
Spontaneous alternation. Spatial working memory was assessed in a Y-maze test as described previously 40 . During a 10-min session in a Y-maze the spontaneous alternation behavior was recorded. The spontaneous alternation test is based on the willingness of rodents to explore new environments, therefore to enter the arm of a Y-maze that had not been recently explored, i.e., the arm that was not entered in the previous choice. Mice were placed in a gray plexiglass Y maze in a sound proof room. Each maze consisted of three arms. Each arm was 40-cm long, 20-cm high, and 10-cm wide, and the arms converged to an equilateral triangular central area. Each mouse was placed at the end of the start arm, in which the mouse starts to explore and move freely through the maze during the 10-min session. The series and order of arm entries was observed manually and considered to be completed when the hind paws of the mouse had completely entered the arm. Transitions between arms were scored as either correct or incorrect alternations by the observer. Healthy animals (controls) show a high change rate. This shows that the animal can remember which arm was last entered. The percentage of alternation was calculated as the ratio of actual to possible alternations multiplied by 100%, either for a 10 min session or 25 initial decisions. Each of the arms of the Y maze was cleaned with 70% ethanol solution between trials.
Statistical analyses and graphical illustrations. All analyses and graphical illustrations were performed using Prism 7 (Graphpad Software, San Diego, CA) Assumptions concerning normal distribution were applied (Kolmogorov-Smirnov test) before statistical analyses were implemented. Statistical analyses and approaches are provided in Supplementary Table 1. Concerning annotation, the sample/cell number is expressed with n and number of mice is noted with N. For animal studies and RT-qPCR expression analysis, data were compared using a one-way ANOVA and two-way ANOVA with and without repeated measures by multiple comparisons followed by Tukey's or Bonferroni's multiple comparison post hoc test. The numbers of biologically independent experiments, sample size, statistical results and ages of the animals are all indicated in the main text or figure legends and Supplementary Table 1. Values were only excluded if they were identified as outliers based on a Grubbs' test. Unless stated otherwise, the data given in figures and text are expressed as mean ± SEM. P < 0.05 was considered to be statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.