Ank3-1b dose-dependent increases in slow gamma oscillations correlate with epilepsy phenotype severity
Previous characterization of Ank3-1b KO mice showed that both homozygous (Ank3-1bKO/KO) and heterozygous (Ank3-1bKO/+) KOs have frequent seizure episodes, with SWDs occupying approximately 4% of total EEG activity from Ank3-1bKO/KO mice16. Since SWD spiking occurs at 6–8 Hz, creating pronounced peaks in power spectra, and because these seizures are accompanied by behavioral arrest, it was important to remove these events from our analyses. To do this, we utilized a supervised learning algorithm to automatically identify seizures in the EEG data30. This algorithm identifies SWDs based on the weighed scores of three properties, 1.) ~ 6 Hz spiking frequency, 2.) ~ 16–32 Hz harmonic, and 3.) sharpness of spikes utilizing the DB4 wavelet. We trained this algorithm on EEG from the Ank3-1b model to identify SWDs (Fig. 1A & 1B) according to scores most resembling a subset of hand scored data and then visually verified the automated calls (Fig. 1C). This data was then used to index EEGs and remove SWDs before spectral analysis.
Spectral analysis of all EEG recordings, indiscriminate of behavior, was first conducted to look for potential large effects between genotypes (Fig. 1D). Statistically significant increases in slow gamma power were seen in Ank3-1bKO/KO mice compared to WT and Ank3-1bKO/+ animals (main effect of genotype: F = 7.422, df = 2, p = 0.006; Ank3-1bKO/KO vs WT: std error: ±1.50 a.u., p = 0.024; Ank3-1bKO/KO vs Ank3-1bKO/+: std error: ±1.16 a.u., p = 0.015; Fig. 1E). This suggests a gene dose dependent effect of slow gamma, similar to that previously described with regard to seizure frequency16. Thus, we ran linear regression analysis to test the relationship between slow gamma power and seizure classification scores and found a statistically significant positive correlation (R2 = 0.30, p = 0.02; Fig. 1F). These data indicate a relationship between seizure activity and slow gamma power in an Ank3-1b dosage-dependent manner.
Ank3-1b KO mice exhibit sleep disturbances co-occurring with increased slow gamma power
Sleep disturbance is symptomatic of BD and because sleep disturbances are present throughout all mood states of BD, we used an existing classifier29 to analyze sleep patterns in Ank3-1b KO mice. We further used an algorithm that distinguishes REM and NREM in sleep EEG using the theta (6–10 Hz) to delta (2–5 Hz) power ratio33,34. Significant differences in overall sleep duration (i.e., number of minutes of sleep per hour) were not observed between genotypes (Supplementary Fig. 2A, generalized linear mixed model, main effect of genotype: F(2,81) = 1.4, p = 0.3). The number of sleep bouts per hour also were not significantly different across genotypes (Supplementary Fig. 2B, generalized linear mixed model, main effect of genotype: F(2,81) = 0.5, p = 0.6). However, we found that the effects of Ank3-1b deletion on sleep patterns differed across sleep stages, with sleep patterns more strongly affected during REM than non-REM (Fig. 2A, D; significant genotype by sleep stage interaction effects: F(3,168) = 35.7, p < 0.001 for time spent in sleep and F(3,169) = 23.1, p < 0.001 for number of sleep bouts). Ank3-1b deletion did not affect the duration of NREM sleep epochs (Fig. 2A; no significant main effect of genotype on sleep duration: F(2,85) = 1.6, p = 0.20), although the number of NREM bouts per hour was significantly lower in Ank3-1bKO/KO mice than in wildtype mice (significant main effect of genotype: F(2,85) = 5.6, p = 0.005; Ank3-1bKO/KO vs wildtype: t(85) = 3.4, p = 0.004; other pairwise comparisons were non-significant). In contrast, the duration of sleep epochs during REM was significantly lower in Ank3-1b KO mice than in wildtype mice (Fig. 2D; main effect of genotype on sleep duration: F(2,84) = 6.8, p = 0.002; post-hoc pairwise comparisons: Ank3-1bKO/KO vs wildtype: t(84) = 2.5, p = 0.03; Ank3-1bKO/+ vs wildtype: t(84) = 3.6, p = 0.002). The number of REM bouts per hour was also significantly lower in Ank3-1b KO mice than in wildtype mice (Fig. 2D; main effect of genotype on number of sleep bouts: F(2,84) = 13.5, p < 0.001; post-hoc pairwise comparisons: Ank3-1bKO/KO vs wildtype: t(84) = 5.1, p < 0.001; Ank3-1bKO/+ vs wildtype: t(84) = 3.4, p = 0.002). These findings suggest that Ank3-1b deletion more strongly disrupts REM sleep than NREM sleep. It is important to note that due to lack of simultaneous electromyography (EMG) recordings, we were unable to definitively determine if overall REM time and number of REM bouts was reduced, or if this data simply reflects disruption of normal theta-delta ratios characteristic of classic REM sleep. Nonetheless, disrupted REM properties are indicative of many different sleep disorders and are implicated in BD35–42. These data warrant further investigation of sleep deficits in this model, which stands to provide important insights and implications for BD treatments.
We next examined whether disturbances in REM and NREM sleep patterns were accompanied by disturbances in EEG rhythms during sleep. We found that Ank3-1b deletion affected oscillatory power differently during REM and NREM sleep (generalized linear mixed model: significant genotype by sleep stage interaction: F(3,342) = 92.3, p < 0.001; significant genotype by sleep stage by rhythm type interaction: F(12,342) = 1178.1, p < 0.001), with stronger effects of Ank3-1b deletion on power spectra observed during REM than NREM (Fig. 2B, E). Therefore, we next analyzed EEG rhythms during NREM and REM separately.
During identified NREM sleep epochs, Ank3-1b deletion did not produce generalized effects on power across all frequencies (generalized linear mixed model: significant genotype by rhythm type interaction: F(6,168) = 53.7, p < 0.001). Instead, effects of Ank3-1b deletion on rhythmic power were most pronounced for slow gamma rhythms (Fig. 2B). Slow gamma power was significantly greater in Ank3-1b KO mice than in wildtype mice in a gene-dosage dependent manner (Fig. 2C; main effect of genotype on slow gamma power: F(2,56) = 38.5, p < 0.001; post-hoc pairwise comparisons: Ank3-1bKO/KO vs wildtype slow gamma: t(56) = 8.8, p < 0.001; Ank3-1bKO/+ vs wildtype slow gamma: t(56) = 3.5, p = 0.001; Ank3-1bKO/KO vs Ank3-1bKO/+: t(56) = 7.0, p < 0.001). Ank3-1b deletion also modestly affected delta rhythms, a type of rhythm that is predominant in EEG recordings during NREM43. However, in contrast to gamma results, power in the delta frequency range was significantly lower in homozygous KOs than in wildtype mice (generalized linear mixed model: main effect of genotype on delta power during NREM: F(2,56) = 5.5, p = 0.007; post-hoc pairwise comparisons: Ank3-1bKO/KO vs wildtype delta: t(56) = 2.4, p = 0.04; Ank3-1bKO/+ vs wildtype delta: t(56) = 1.0, p = 0.34) and significantly lower in homozygous KOs than in heterozygous KOs (Ank3-1bKO/KO vs Ank3-1bKO/KO: t(56) = 3.3, p = 0.005).
We next investigated how Ank3-1b deletion altered EEG rhythms during identified REM sleep epochs. We again assessed slow gamma rhythms and also assessed theta, a rhythm type that dominates EEG recordings during REM sleep43. Ank3-1b deletion differentially affected the power of the different rhythm types during REM (Fig. 2E; generalized linear mixed model: significant genotype by rhythm type interaction: F(6,174) = 2302.5, p < 0.001). The power of slow gamma rhythms during REM significantly differed across genotypes (Fig. 2F, main effect of genotype: F(2,58) = 404.0, p < 0.001). Specifically, slow gamma power increased with decreasing levels of Ank3-1b (post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+ slow gamma: t(58) = 2.5, p = 0.02; Ank3-1bKO/+ vs Ank3-1bKO/KO slow gamma: t(58) = 21.6, p < 0.001; Ank3-1bKO/KO vs wildtype slow gamma: t(58) = 27.3, p < 0.001). Theta power during REM also significantly differed across genotypes (main effect of genotype: F(2,58) = 94.8, p < 0.001). However, in contrast to slow gamma results, homozygous Ank3-1b deletion significantly decreased theta power during REM (post-hoc pairwise comparisons: Ank3-1bKO/KO vs wildtype theta: t(58) = 6.2, p < 0.001). Also, theta power during REM did not significantly differ between wildtype mice and heterozygous KOs (wildtype vs Ank3-1bKO/+ theta: t(58) = 1.9, p = 0.07). Taken together, this collection of results suggests a potential relationship between increased slow gamma power and disrupted REM sleep.
Previous characterization of these mice showed that Ank3-1bKO/KO mice have premature mortality rates16, and meet the criteria for sudden death in epilepsy (SUDEP), so we also looked at seizure activity during REM and NREM sleep. Our data indicates that seizures are approximately 10X more common during REM (1.24 seizures/hr) than during NREM (0.13 seizures/hr) sleep, and this increased seizure frequency was statistically significant in REM (n = 9, mean rank = 12.11) compared to NREM sleep (n = 9, mean rank = 6.89; p = 0.040, std error = ± 9.957; Supplementary Fig. 4A & B). This result was striking because seizures triggered during REM sleep are more likely to cause sudden death compared to other times of the day and NREM sleep.44 Then we looked at the effects of seizure occurrence on gamma power in Ank3-1bKO/+ and Ank3-1bKO/KO mice and found that seizure occurrence affected gamma power differently between genotypes during REM sleep (generalized linear mixed model: significant genotype by seizure occurrence interaction: F(2,38) = 15.3, p < 0.001). We found that seizure occurrence had no significant effect on gamma power in Ank3-1bKO/+ (seizure occurrence vs. no seizure occurrence: F(1,19) = 0.5; t(19) = 0.7, p = 0.5), but seizure occurrence was associated with significantly increased gamma power in Ank3-1bKO/KO mice (seizure occurrence vs. no seizure occurrence: F(1,19) = 404; t(19) = 20.1, p < 0.001; Supplementary Fig. 4C). This highlights again a positive correlation between seizure phenotype severity and increased slow gamma power and suggests that seizure activity and slow gamma rhythms may be associated. When we looked at the effect of seizure occurrence on theta power in Ank3-1bKO/+ and Ank3-1bKO/KO mice during REM sleep, we found that seizure occurrence affected theta power differently between genotypes (significant genotype by seizure occurrence interaction: F(2,38) = 114.7, p < 0.001). We found that seizure occurrence had no significant effect on theta power in Ank3-1bKO/+ (seizure occurrence vs. no seizure occurrence: F(1,19) = 4.1; t(19) = 2.0, p = 0.06), but seizure occurrence was associated with significantly decrease theta power in Ank3-1bKO/KO mice (seizure occurrence vs. no seizure occurrence: F(1,19) = 228.5; t(19) = 15.1, p < 0.001; Supplementary Fig. 4D). Seizures were extremely rare during NREM; only a single mouse exhibited seizures during NREM. Thus, we were unable to explore potential differential effects of seizure occurrence on gamma power during NREM.
Again, it is important to note that seizures may be unaccounted for due to potentially unidentified sleep bouts, since thorough analysis of sleep requires simultaneous EMG analysis. Mice may be having seizures during REM and NREM periods that our algorithm was unable to identify due to disrupted theta-delta-ratios. Thus, it is critical for future studies to further explore these effects in order to better understand the risk and underlying mechanisms leading to SWDs during sleep in this model. This connection between disrupted REM sleep and increased epileptic activity is another shared feature of the Ank3 link to bipolar disorder and epilepsy.
Ank3-1b KO mice exhibit hyperactivity during awake behaviors
Since activity levels are altered in depressive and manic states of BD, we looked at awake rest and walking behaviors in Ank3 KO mice (Fig. 3A, D). Awake rest and walking behaviors were classified using LDA on the kinematic information from the videos taken for bouts lasting longer than 0.5 seconds. Awake rest states were defined as periods during which the mouse was not moving but was not identified as asleep by the sleep classifier. The number of awake rest (Fig. 3A) and walking (Fig. 3B) bouts per hour were differentially affected by Ank3-1b deletion (significant genotype by behavior type interaction: F(2,168) = 4.8, p = 0.001). Compared to wildtype mice, Ank3-1bKO/KO mice exhibited significantly fewer bouts of awake rest per hour (generalized linear mixed model, significant main effect of genotype: F(2,84) = 5.4, p = 0.006; post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(84) = 2.2, p = 0.07; Ank3-1bKO/KO vs Ank3-1bKO/+: t(84) = 0.9, p = 0.4; Ank3-1bKO/KO vs wildtype: t(84) = 3.2, p = 0.006) and spent less time (minutes per hour) in the awake rest state (Fig. 3A, significant main effect of genotype: F(2,84) = 3.9, p = 0.002; post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(84) = 2.1, p = 0.7; Ank3-1bKO/KO vs Ank3-1bKO/+: t(84) = 0.2, p = 0.9; Ank3-1bKO/KO vs wildtype: t(84) = 2.6, p = 0.04). Regarding walking behavior, significantly more walking bouts per hour were observed for Ank3-1b KO mice compared to wildtype mice (Fig. 3D; significant main effect of genotype: F(2,84) = 12.8, p < 0.001; post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(84) = 4.9, p < 0.001; Ank3-1bKO/KO vs wildtype: t(84) = 3.4, p = 0.002). Also, Ank3-1b KO mice spent significantly more time walking compared to wildtype mice (Fig. 3D; significant main effect of genotype: F(2,84) = 8.6, p < 0.001; post-hoc pairwise comparisons: Ank3-1bKO/KO vs wildtype: t(84) = 3.2, p = 0.04; Ank3-1bKO/+ vs wildtype: t(84) = 4.1, p < 0.001). These data support the claim for a manic-like phenotype in Ank3-1b KO mice previously reported26.
Ank3-1b KO mice exhibit increased slow gamma power during awake rest and walking states
Delta rhythms are often evident during periods of awake rest in rodents45–47, and low frequency theta can be associated with non-moving states48. Therefore, we included measurements of these rhythm types, together with slow gamma rhythms, in our model when testing for differences in the power of EEG rhythms between genotypes during awake rest. During awake rest, the different rhythm types were differentially affected by Ank3-1b deletion, with largest effects observed for slow gamma rhythms (Fig. 3B, generalized linear mixed model, significant genotype by rhythm type interaction effect: (F(6,159) = 173.8, p < 0.001). Slow gamma power increased as Ank3-1b levels decreased across the three genotypes (Fig. 3C, significant main effect of genotype: F(2,53) = 94.0, p < 0.001). Ank3-1bKO/KO and Ank3-1bKO/+ mice had significantly increased slow gamma power compared to wildtype mice, and Ank3-1bKO/KO mice had significantly higher slow gamma than Ank3-1bKO/+ mice (post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(53) = 8.4, p < 0.001; Ank3-1bKO/KO vs Ank3-1bKO/+: t(53) = 7.2, p < 0.001; Ank3-1bKO/KO vs wildtype: t(53) = 12.8, p < 0.001). In contrast, power in the theta frequency range was not significantly affected by Ank3-1b deletion (no significant main effect of genotype on theta power: F(2,53) = 1.7, p = 0.2), and delta power was lower in Ank3-1bKO/KO mice than in wildtype mice (significant main effect of genotype: F(2,53) = 67.8, p < 0.001; post-hoc pairwise comparison: Ank3-1bKO/KO vs wildtype: t(53) = 3.2, p < 0.005).
We next assessed the effects of Ank3-1b deletion on EEG rhythms during walking behavior. Active walking behavior is associated with prominent theta and gamma activity in local field potential recordings from rodents49–51. Thus, during walking, we tested for theta and slow gamma power differences across genotypes. We found that theta and slow gamma rhythms during walking were differentially affected by Ank3-1b deletion (generalized linear mixed model, significant genotype by rhythm type interaction effect: F(3,112) = 75.5, p < 0.001). Slow gamma rhythms during walking were significantly increased by Ank3-1b deletion (Fig. 3F, significant main effect of genotype on slow gamma: F(2,56) = 189.3, p < 0.001). As was observed during awake rest, slow gamma power increased with decreasing Ank3-1b (post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(56) = 4.9, p < 0.001; Ank3-1bKO/KO vs Ank3-1bKO/+: t(56) = 13.2, p < 0.001; Ank3-1bKO/KO vs wildtype: t(56) = 18.9, p < 0.001). In contrast, theta power during walking was significantly decreased in Ank3-1b knockout mice compared to wildtype mice (significant main effect of genotype: F(2,56) = 56.8, p < 0.001; post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(56) = 9.4, p < 0.001; Ank3-1bKO/KO vs wildtype: t(56) = 9.2, p < 0.001).
Ank3-1b KO mice exhibit increases in repetitive behaviors
Another endophenotype of BD is obsessive and compulsive thoughts and behaviors52,53. One way to test for such characteristics in mice is to look at their propensity for repetitive behaviors, such as repetitive grooming and digging. Repetitive behaviors are another characteristic related to BD that has not previously been explored in Ank3-1b mice. Thus, we compared the frequency of grooming and digging behaviors in Ank3-1b KO and wildtype mice. We found a significant effect of Ank3-1b deletion on repetitive grooming bouts and amount of time spent grooming per hour (Fig. 4A; generalized linear mixed model, main effect on genotype: bouts: F(2,84) = 36.3, p < 0.001; time: F(2,84) = 44.5, p < 0.001). However, only heterozygous but not homozygous Ank3-1b deletion significantly altered grooming bouts (post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(84) = 7.4, p < 0.001; Ank3-1bKO/KO vs Ank3-1bKO/+: t(84) = 7.6, p < 0.001; Ank3-1bKO/KO vs wildtype: t(84) = 0.05, p = 1.0). Also, only heterozygous but not homozygous Ank3-1b KO mice showed significantly increased time grooming (generalized linear mixed model, main effect on genotype: wildtype vs Ank3-1bKO/+: t(84) = 8.7, p < 0.001; Ank3-1bKO/KO vs Ank3-1bKO/+: t(84) = 8.1, p < 0.001; Ank3-1bKO/KO vs wildtype: t(84) = 0.7, p = 0.5). Digging may provide a better measure of repetitive behaviors in Ank3-1b mice because, unlike grooming, it does not require animals to balance on their hind paws which may be more difficult for Ank3-1bKO/KO mice due to their mild ataxia. A significant effect of Ank3-1b deletion on digging behaviors was observed (Fig. 4D) when digging was measured in bouts per hour (generalized linear mixed model, main effect of genotype: F(2,84) = 6.6, p = 0.002). As was the case with grooming, only heterozygous Ank3-1b KO mice and not homozygous Ank3-1b KO mice showed significantly more digging bouts per hour in compared to wildtype mice (post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(84) = 3.6, p = 0.002; Ank3-1bKO/KO vs Ank3-1bKO/+: t(84) = 2.5, p = 0.03; Ank3-1bKO/KO vs wildtype: t(84) = 1.3, p = 0.21). We also found a significant effect of Ank3-1b deletion on repetitive digging minutes per hour (generalized linear mixed model, main effect on genotype: F(2,84) = 5.5, p = 0.006). Unlike with grooming, both heterozygous Ank3-1b KO mice and homozygous Ank3-1b KO mice showed significantly increased digging time compared to wildtype mice (generalized linear mixed model, main effect on genotype: wildtype vs Ank3-1bKO/+: t(84) = 2.8, p = 0.02; Ank3-1bKO/KO vs Ank3-1bKO/+: t(84) = 0.2, p = 0.9; Ank3-1bKO/KO vs wildtype: t(84) = 2.9, p = 0.02). It is interesting to note that while one would expect the magnitude of behavioral effects to increase with reduced Ank3-1b dosage, Ank3-1bKO/+ mice exhibited the strongest repetitive behaviors phenotype. This may be due to the confounding factor of ataxia in Ank3-1bKO/KO mice27, making it difficult for Ank3-1bKO/KO mice to engage in prolonged bouts of coordinated grooming and digging movements.
Ank3-1b KO mice show increased slow gamma rhythms during repetitive behaviors
Power spectra during grooming behaviors showed peaks in the theta and slow gamma bands, so we next examined whether Ank3-1b deletion affected theta and slow gamma power during grooming. As was observed for other awake behaviors, different effects of Ank3-1b deletion on theta and slow gamma power were observed during grooming behaviors (Fig. 4B, generalized linear mixed models, significant genotype by rhythm type interaction effect on grooming behaviors: F(3,136) = 114.8, p < 0.001). Specifically, slow gamma power was increased by Ank3-1b deletion in a dosage-dependent manner (Fig. 4C, generalized linear mixed model, significant main effect of genotype: F(2,60) = 59.6, p < 0.001; post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(60) = 3.8, p < 0.001; Ank3-1bKO/+ vs Ank3-1bKO/KO: t(60) = 7.6, p < 0.001; Ank3-1bKO/KO vs wildtype: t(60) = 10.8, p < 0.001). In contrast, Ank3-1b deletion decreased theta power (generalized linear mixed model, significant main effect of genotype: F(2,60) = 5.3, p = 0.008), but the only post-hoc pairwise comparison that was significantly different was the comparison of theta power in Ank3-1bKO/+ mice and wildtype mice (t(60) = 3.2, p = 0.006). Similar to slow gamma power during grooming, slow gamma power during digging was increased by Ank3-1b deletion in a dosage-dependent manner (Fig. 4E-F, generalized linear mixed model, significant main effect of genotype: F(2,68) = 118.2, p < 0.001; post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(68) = 5.8, p < 0.001; Ank3-1bKO/+ vs Ank3-1bKO/KO: t(68) = 10.0, p < 0.001; Ank3-1bKO/KO vs wildtype: t(68) = 15.4, p < 0.001). Indeed, analyses revealed that Ank3-1b deletion affected slow gamma power similarly during the different types of repetitive behaviors (i.e., grooming and digging; generalized linear mixed model, no significant main effect of repetitive behavior type on slow gamma power: F(1,128) = 0.41, p = 0.525; no significant genotype by repetitive behavior type interaction effect: F(1, 128) = 0.27, p = 0.76). Therefore, we included slow gamma measurements from grooming and digging behaviors together and again assessed the effect of Ank3-1b deletion on slow gamma rhythm power. As was observed for grooming and digging behaviors alone, Ank3-1b deletion increased slow gamma power during grooming and digging behaviors analyzed together (generalized linear mixed model, main effect of genotype: F(2,128) = 91.9, p < 0.001). During grooming and digging behaviors, slow gamma power in both Ank3-1bKO/+ and Ank3-1bKO/KO mice was larger than slow gamma power in wildtype mice (post-hoc pairwise comparisons: wildtype vs Ank3-1bKO/+: t(128) = 3.9, p < 0.001; Ank3-1bKO/KO vs wildtype: t(128) = 13.2, p < 0.001). Also, slow gamma power during grooming and digging behaviors was greater in homozygous KO mice than in heterozygous KO mice (Ank3-1bKO/+ vs Ank3-1bKO/KO: t(128) = 9.7, p < 0.001). Taken together, these findings suggest that increased slow gamma rhythms accompany increased repetitive behaviors in the Ank3-1b mice. Interestingly, increased gamma power (20–50 Hz) has been shown to be associated with repetitive behaviors in autism spectrum disorder (ASD) patients54, and ANK3 rare variants have been found in patients with ASD55–57. Thus, not only is this phenotype in Ank3-1b mice relevant for BD, but it may also have implications for ASD.