Antisense oligonucleotide therapy rescues disturbed brain rhythms and sleep in juvenile and adult mouse models of Angelman syndrome

UBE3A encodes ubiquitin protein ligase E3A, and in neurons its expression from the paternal allele is repressed by the UBE3A antisense transcript (UBE3A-ATS). This leaves neurons susceptible to loss-of-function of maternal UBE3A. Indeed, Angelman syndrome, a severe neurodevelopmental disorder, is caused by maternal UBE3A deficiency. A promising therapeutic approach to treating Angelman syndrome is to reactivate the intact paternal UBE3A by suppressing UBE3A-ATS. Prior studies show that many neurological phenotypes of maternal Ube3a knockout mice can only be rescued by reinstating Ube3a expression in early development, indicating a restricted therapeutic window for Angelman syndrome. Here we report that reducing Ube3a-ATS by antisense oligonucleotides in juvenile or adult maternal Ube3a knockout mice rescues the abnormal electroencephalogram rhythms and sleep disturbance, two prominent clinical features of Angelman syndrome. Importantly, the degree of phenotypic improvement correlates with the increase of Ube3a protein levels. These results indicate that the therapeutic window of genetic therapies for Angelman syndrome is broader than previously thought, and electroencephalogram power spectrum and sleep architecture should be used to evaluate the clinical efficacy of therapies.


Generation of a new Ube3a null allele in mice 115
Our goal was to assess the effect of Ube3a-ATS-targeted ASOs in a mouse model of Angelman 116 syndrome. The widely used mouse model is a Ube3a knockout allele (Ube3a tm1Alb , referred to as 117 Ube3a De5 here to be distinguished from the new allele) that deletes exon 5 (previously named as 118 exon 2), resulting in a premature stop codon in exon 6 (Jiang et al., 1998) (Figure 1B). We 119 performed reverse transcription droplet digital PCR (RT-ddPCR) analyses on this Ube3a 120 knockout allele with primer sets targeting different exons. Exons 4, 6, and other exons 121 downstream of the deleted exon 5 were still transcribed in the brains of adult heterozygous 122 maternal (Ube3a mDe5/p+ ) and homozygous (Ube3a mDe5/pDe5 ) mutant mice at a level comparable to 123 wild type (WT) mice (Figure 1C), possibly due to an escape from nonsense-mediated mRNA 124 decay or an alternative start site. Similarly, Ube3a mRNA was only modestly reduced in the 125 livers of Ube3a mDe5/p+ and Ube3a mDe5/pDe5 mice ( Figure 1C). Although this knockout allele 126 produces very little full-length functional Ube3a proteins in the brain (Judson et al., 2014;Grier 127 et al., 2015), we sought to create a new Ube3a null allele with diminished Ube3a mRNA to 128 facilitate the evaluation of ASO efficacy at the transcript level. CRISPR/Cas9 was used to delete 129 the largest Ube3a coding exon, exon 6. The resulting allele (Ube3a De6 ) carries a premature stop 130 codon in exon 7 ( Figure 1B). RT-ddPCR analyses of adult heterozygous maternal mutant mice 131 (Ube3a mDe6/p+ ) showed that Ube3a mRNA was diminished in the brain and reduced in the liver 132 as compared to WT mice ( Figure 1D). Western blots revealed that Ube3a protein levels in 133 different brain regions of Ube3a mDe6/p+ mice were 2-17% of those in WT mice when they were at 134 6 weeks of age or older (see below). Thus, both Ube3a mRNA and proteins are diminished in the 135 Ube3a mDe6/p+ mouse brains. Furthermore, Ube3a mDe6/p+ mice showed similar rotarod and marble 136 burying phenotypes to previously reported deficits in Ube3a mDe5/p+ mice (Shi et al., 2022). 137 138 ASOs targeting Ube3a-ATS non-coding RNA up-regulate paternal Ube3a-YFP expression 139 To increase the paternal expression of Ube3a in mice, we used two mouse-specific antisense 140 oligonucleotides to downregulate the Ube3a-ATS levels. The first one (Ube3a-as) is 141 complementary to a region downstream of the Snord115 small nuclear RNA cluster, and this 142 sequence was also targeted by the "ASO B" used in a previous study (Meng et al., 2015). The 143 second one (Snord115) is complementary to a sequence that repeats 110 times in the Snord115 144 RNAs. To test the effect and visualize the distribution of these two ASOs in the brain, we first 145 used paternal Ube3a YFP mice (Ube3a m+/pYFP ) carrying a yellow fluorescent protein (YFP)-tagged 146 Ube3a (Dindot et al., 2008), as downregulating Ube3a-ATS is expected to reactivate the paternal 147 Ube3a-YFP allele. A non-targeting control ASO, Ube3a-as ASO, or Snord115 ASO was 148 administered to the brains of 3-month-old Ube3a m+/pYFP mice by a single unilateral 149 intracerebroventricular (ICV) injection. We visualized Ube3a-YFP expression by 150 immunostaining of YFP 18 days post ASO injection. Maternal Ube3a YFP mice (Ube3a mYFP/p+ ) 151 exhibited strong Ube3a-YFP in the brain, whereas Ube3a m+/pYFP mice receiving the control ASO 152 showed little expression (Figure 2A,B). Ube3a-as ASO and Snord115 ASO caused a robust 153 increase in Ube3a-YFP expression throughout the brains of Ube3a m+/pYFP mice (Figure 2A,B). 154 Although we did not specifically examine different cell types, YFP was observed in several types 155 of GABAergic neurons including cerebellar Purkinje cells, olfactory bulb granule cells, striatal 156 neurons, and interneurons in cortical layer 1, hippocampal stratum oriens, and cerebellar 157 molecular layer ( Figure 2B) of both hemispheres at 3, 6, or 10 weeks post ASO injections ( Figure 3A). We used the 170 hemispheres ipsilateral to the ASO injection site for Western blot analyses and the corresponding 171 contralateral hemispheres for reverse transcription quantitative real-time PCR (RT-qPCR) 172 analyses. At 3 weeks post ASO injections, Ube3a-as ASO and Snord115 ASO increased Ube3a 173 protein levels in Ube3a mDe6/p+ mice to about 28-71% of the WT levels in different brain regions 174 Ube3a-ATS levels were downregulated by 29-73% (Figure 4A,B, Figure 4

-supplement 1) and 176
Ube3a mRNA levels were increased to about 22-57% of the WT levels ( Figure 4A,B, Figure 4-177 supplement 1). The effects of Ube3a-as ASO remained stable for at least 10 weeks. However, 178 the Ub3a protein and Ub3a mRNA levels in Snord115 ASO-treated Ube3a mDe6/p+ mice markedly 179 decreased at 6 weeks post injections and reached to the levels of control ASO-treated 180  injection timepoints from the ipsilateral and contralateral hemispheres, respectively, these 203 correlations indicate a broad distribution of ASOs in the mouse brains and a spatiotemporally 204 comparable pattern between the changes in transcripts and proteins. Taken together, our results 205 demonstrate that a single unilateral ICV injection of ASO targeting Ube3a-ATS in Ube3a mDe6/p+ 206 mice leads to a long-lasting down-regulation of this transcript and reactivation of the paternal 207 Ube3a allele throughout the brains, and the up-regulation of Ube3a proteins by Ube3a-as ASO 208 can last at least 10 weeks. 209 210

Reactivation of paternal Ube3a expression alleviates abnormal EEG rhythmic activity in 211
Ube3a mDe6/p+ mice 212 Maternal Ube3a deficiency in mice causes altered brain rhythms, sleep disturbance, and 213 epileptiform activity (e.g., cortical poly-spikes), all of which can be examined by chronic video-214 EEG and electromyogram (EMG) recordings. Thus, to determine if up-regulation of paternal 215 Ube3a expression can reverse these phenotypes in Ube3a mDe6/p+ mice, we injected male and 216 female Ube3a mDe6/p+ mice with control, Ube3a-as, or Snord115 ASO and their sex-and age-217 matched WT littermates with control ASO in parallel at the juvenile (postnatal days 21.5 ± 0.1 218 (mean ± s.e.m.), range 21-24, n = 35) or adult (postnatal days 62.5 ± 0.6 (mean ± s.e.m.), range 219 56-66, n = 28) age. Intracranial EEG from the frontal, somatosensory, and visual cortices and 220 EMG from the neck muscles of each mouse were recorded at 3, 6, and 10 weeks post ASO 221 injections ( Figure 3A, Figure 5A). To avoid bias, we evenly sampled 6 out of 24 hours of the 222 EEG/EMG data for power spectrum and poly-spikes analyses and used 24 hours of data for sleep 223 scoring (see Materials and Methods). 224

225
We first removed artifacts and then computed the absolute power spectral densities (PSDs) of 226 EEG signals including all brain states (Figure 5-supplement 1). To control for the variations 227 caused by different impedances across electrodes and mice, we normalized PSDs by the total 228 power within 1-100 Hz to obtain the relative PSDs. The relative PSDs from the frontal cortex of 229 control ASO-treated Ube3a mDe6/p+ mice were higher at 4-25 Hz and lower at 40-80 Hz than 230 those of control ASO-treated WT mice (Figure 5B,E). Thus, we further computed the relative 231 power in the frequency bands of delta (d, 1-4 Hz), theta (q, 4-8 Hz), alpha (a, 8-13 Hz), low 232 beta (b1, 13-18 Hz), high beta (b2, 18-25 Hz), low gamma (g1, 25-50 Hz), and high gamma (g2, 233 50-100 Hz) (Figure 5C,F). To capture the concurrent changes in both low and high frequency 234 ranges, we calculated the ratio of the total power in the alpha, low beta, and high beta bands over 235 the power in high gamma band (i.e., (a+b1+b2)/g2). This ratio represents the relative 236 distribution of power between the low and high frequency bands and importantly, is independent 237 from the use of PSD or relative PSD. We discovered that the power ratio (a+b1+b2)/g2 was 238 higher in control ASO-treated Ube3a mDe6/p+ mice than control ASO-treated WT mice across all 239 time points (Figure 5D,G). Similar phenotypes were also observed in the somatosensory cortex 240 ( Figure 5-supplement 2), but the EEG rhythmic activity in the visual cortex was not 241 significantly altered in control ASO-treated Ube3a mDe6/p+ mice ( Figure 5-supplement 3). These 242 results indicate that maternal Ube3a deficiency alters EEG rhythms in the frontal and 243 somatosensory cortices, and the power ratio (a+b1+b2)/g2 can be a robust measure of the effects 244 of Ube3a-as and Snord115 ASOs. 245 246 Treating Ube3a mDe6/p+ mice with Ube3a-as ASO at the juvenile age caused a decrease of the 247 power in the alpha, low beta, and high beta bands and an increase of the power in the high 248 gamma band, particularly at 3 and 6 weeks post ASO injections (Figure 5C), which led to the 249 normalization of the ratio (a+b1+b2)/g2 in the frontal and somatosensory cortices, as the ratios 250 in Ube3a-as ASO-treated Ube3a mDe6/p+ mice were indistinguishable from those in control ASO-251 treated WT mice ( Figure 5D; Figure 5-supplement 2C). In contrast, Snord115 ASO only 252 showed such effects in the frontal cortex at 3 weeks post ASO injections, and the effects waned 253 at later timepoints ( Figure 5C,D). This difference between Ube3a-as ASO and Snord115 ASO 254 generally correlates with their difference in up-regulating Ube3a proteins (see below). When 255 Ube3a mDe6/p+ mice were treated with ASOs at the adult age, Ube3a-as ASO and Snord115 ASO 256 were also able to reduce the power in the low frequency bands and increased the power in the 257 high gamma band in the frontal cortex ( Figure 5E,F), thereby reducing the ratio (a+b1+b2)/g2 258 ( Figure 5G). These effects also waned over time, consistent with the change of Ube3a protein 259 levels ( Figure 3F) To study the sleep architecture, we used the EEG and EMG signals and a convolutional neural 267 network-based algorithm SPINDLE (Miladinović et al., 2019) to classify the brain states into 268 rapid eye movement (REM) sleep, non-rapid eye movement (NREM) sleep, and wake 269 throughout 24 hours (Figure 6-supplement 1A). The difference in the EEG PSDs did not affect 270 the accuracy of SPINDLE (Figure 6-supplement 1B,C). Overall, mice spent more time in REM 271 and NREM sleep and less time in wake during the light phase than the dark phase (Figure 6). 272 The time in wake was similar between control ASO-treated WT and Ube3a mDe6/p+ mice across 273 ages ( Figure 6C,F). Control ASO-treated Ube3a mDe6/p+ mice spent significantly less time in 274 REM sleep than control ASO-treated WT mice in the light phase, but this phenotype was more 275 variable in adult mice (Figure 6A,D). Correspondingly, control ASO-treated Ube3a mDe6/p+ mice 276 spent slightly more time in NREM sleep than control ASO-treated WT mice because REM sleep 277 is a small fraction of the total sleep (Figure 6B,E). Thus, the sleep disturbance in Ube3a mDe6/p+ 278 mice manifests as a selective reduction in REM sleep, which recapitulates the observation in 279 Angelman patients (Miano et al., 2004;. Ube3a protein levels. Therefore, the power ratio (a+b1+b2)/g2 also negatively correlated with 329 the Ube3a protein levels ( Figure 8A). However, the relative power in the delta (d, 1-4 Hz) band 330 did not correlate with the Ube3a protein levels ( Figure 8A) Meanwhile, an increasing number of studies in mouse models of Angelman syndrome 342 demonstrate that Ube3a must be reinstated in late embryonic and early postnatal development to 343 correct most neurological phenotypes. Among the previously tested phenotypes, only a small 344 subset (i.e., synaptic transmission, plasticity, and spatial memory) can be improved upon 345 increasing Ube3a at the age of 6 weeks or older, and slightly a few more (i.e., rotarod 346 performance and susceptibility to seizure induction) when increasing Ube3a at postnatal day 21 347

(Supplementary File 1). Despite limited prior successes in rescuing juvenile and adult maternal 348
Ube3a deficiency mice, we chose these two ages to examine the effects of ASO therapy on 349 cortical hyperexcitability, altered EEG power spectrum, and sleep disturbance because these ages 350 are more translationally relevant than the neonatal period. Our study reveals that a single ICV 351 injection of Ube3a-ATS-targeted ASOs to Ube3a mDe6/p+ mice, a new rodent model of Angelman 352 syndrome, restores the EEG power spectrum and sleep pattern for at least 6 weeks, particularly 353 upon treatment at the juvenile age. Therefore, our results significantly expand the range of 354 phenotypes that can be reversed by restoring Ube3a expression in juvenile and adult mice. 355 Interestingly, we were not able to reduce the frequency of poly-spikes in Ube3a mDe6/p+ mice at 356 either age (Figure 7). It is possible that suppression of poly-spikes requires up-regulation of 357 Ube3a starting at a younger age or reaching to a higher level than what we have achieved, both 358 of which should be tested in future studies. Nevertheless, this result indicates that poly-spikes are 359 independent from the EEG power spectrum and sleep pattern deficits, and probably involve a 360 different mechanism. 361 362 A critical finding of our study is that the improvement in the EEG power spectrum and sleep 363 pattern tracks the increase in Ube3a protein levels across different ASOs, injection ages, and 364 timepoints post injection (Figure 8). This suggests that following a bolus injection of ASOs, 365 both phenotypes are acutely modulated by the Ube3a levels that decrease over time due to ASO 366 clearance. Future studies should determine if repeated administration of ASOs can generate a 367 long-lasting improvement of the phenotypes beyond the period when Ube3a levels are 368 sufficiently up-regulated, as the outcome can help inform the ASO treatment schedule in clinical 369 trials. The ASO treatment in adult mice is less effective than juvenile mice, possibly due to two 370 reasons. First, the reversibility of these two phenotypes may decrease over age, just like other 371 neurological deficits (Supplementary File 1). Second, the ASO treatment in adult mice causes a 372 smaller increase of Ube3a protein than juvenile mice (Figure 3). Given the strong correlation 373 between Ube3a levels and phenotypic improvement, we speculate that the latter is more likely 374 the reason and a higher dose of ASO or an ASO with a higher efficacy in down-regulating 375 Ube3a-ATS should further increase Ube3a protein and improve these two phenotypes in adult 376 mice. that the relative delta power of cortical surface EEG was reduced in Ube3a mDe5/p+ mice on the 390 C57BL/6J background during NREM sleep in the night, but not in the day (Ehlen et al., 2015). 391 Our absolute PSD results from the new Ube3a mDe6/p+ mice are qualitatively similar to the 392 previous results (Figure 5-supplement 1), but relative PSD analysis reveals an increase of 393 relative power in the theta, alpha, or beta frequency bands and a decrease in the low or high 394 gamma frequency bands. These differences could be due to different mutations, genetic 395 backgrounds, mouse ages, experimental conditions, or different brain states included in the 396 analyses. Nevertheless, we observed a robust and consistent increase in the power ratio 397 (a+b1+b2)/g2 across timepoints (Figure 5). In fact, inspection of previous results suggests a 398 common pattern that EEG power is relatively higher in maternal Ube3a deficiency mice than 399 WT mice in the lower frequency bands (i.e., beta or lower) and relatively lower in the gamma 400 bands, although the results were not always statistically significant (Born et  were grouped together. ASO solution was injected at a rate of 407 nl/s using an UltraMicroPump 501 III and a Micro4 controller (World Precision Instruments). A total of 10 µl ASO solution (50 502 mg/ml) was administered for a total dosage of 500 µg/mouse except 6 mice that were injected 503 with 5 µl ASO solution for a total dosage of 250 µg/mouse at the age of 3 weeks and used in the 504 Western blot experiments. The results from these 6 mice were similar to other mice (Figure 3-505 supplement 3) and grouped together. After injection, the pipette was held in place for 10 min 506 before withdrawal. The skin was sutured, and mice were allowed to recover from anesthesia in a 507 cage placed on a heating pad. When the recovery takes longer than 1 hour, the duration on the 508 heating pad should not exceed 1 hour, as longer exposure of mice on the heating pad 509 significantly reduces post-surgery survival rates (less than 1 hour: 1 out of 102 injected mice 510 died, more than 1 hour: 31 out of 106 injected mice died, P < 0.0001).

Video-electroencephalogram(EEG) and electromyogram (EMG) recordings 562
Video-EEG/EMG recordings were performed as previously described (Chen et al., 2020). 563 Briefly, one week after ASO injection, mice were anesthetized with 2.5% isoflurane in oxygen, 564 and craniotomies were performed as described above for ICV injection. recording and an EMG reference electrode were inserted into the neck muscles. All electrodes 571 were soldered to an adaptor prior to the surgery. The electrodes and adaptor were secured on the 572 skull by dental acrylic. The skin was sutured and attached to the dried dental acrylic. Mice were 573 singly housed to recover for at least one week after the surgeries. Before recording, mice were 574 individually habituated in the recording chambers (10-inch diameter of Plexiglas cylinder) for 24 575 hours. EEG/EMG signals (5000-Hz sampling rate with a 0.5-Hz high-pass filter) and videos (30 576 frames per second) were recorded synchronously for more than 48 continuous hours using a 4-577 channel EEG/EMG tethered system and Sirenia 1. spikes is defined as a cluster of three or more spikes on any of the EEG channels. PSD analyses 585 of EEG data were performed using custom scripts in Python. Prior to PSD calculation, data were 586 detrended by subtracting the mean of the data. The data segments containing artifacts on any of 587 the EEG channels were first excluded, and then an 8th order Butterworth filter was applied to 588 each channel to bi-directionally notch filter around 60 Hz (± 2 Hz bandwidth) to remove power-589 line noise. The PSDs were then estimated for each channel using a Welch's periodogram [1] with 590 a 2-s Hanning window (achieving a frequency resolution of 0.5 Hz) and 50% overlap between 591 windows. To account for the effect of notch filtering, the PSD was linearly interpolated between 592 58-62 Hz using the 10 points before and after the mentioned ranges. To analyze different 593 frequency bands, the PSD was segmented into 7 bands: delta (1-4 Hz), theta (4-8 Hz), alpha (8-594 13 Hz), low beta (13-18 Hz), high beta (18-25 Hz), low gamma , and high gamma 595 (50-100 Hz). The power within a frequency band (the area under the PSD curve for a band) was 596 then computed for each band. The relative power in a frequency band is the ratio of the power 597 within the band over the total power within 1-100 Hz. The normalized PSD curves were 598 obtained by dividing the PSD curves with the total power within 1-100 Hz. The low-to-high 599 frequency band ratio was calculated as the ratio of the total power in the alpha and beta bands 600 experts was similar to that of the experts compared among each other (Figure 6-supplement 1). 617

618
Statistics 619 All reported sample numbers (n) represent independent biological replicates that are the numbers 620 of tested mice. Statistical analyses were performed with Prism 9 (GraphPad Software). Student's 621 t-test or ANOVA with multiple comparison test for all pairs of groups were used to determine if 622 there is a statistically significant difference between two groups or among three or more groups, 623 respectively. One-way or two-way ANOVA was applied for one or two independent variables, The same summary data presented in Figure 3D and  comparison test for all pairs of groups, * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. 764