Early Developmental Electroencephalography Abnormalities, Neonatal Seizures, and Induced Spasms in a Mouse Model of Tuberous Sclerosis Complex

Summary/Abstract

Objective:

Tuberous sclerosis complex (TSC) is one of the most common genetic causes of epilepsy. Seizures in TSC typically first present in infancy or early childhood, including focal seizures and infantile spasms. Infantile spasms in TSC are particularly characteristic in its strong responsiveness to vigabatrin. While a number of mouse models of epilepsy in TSC have been described, there are very limited EEG or seizure data during the pre-weanling neonatal and infantile-equivalent mouse periods. Tsc1^GFAPCKO mice are a well-characterized mouse model of epilepsy in TSC, but whether these mice have seizures during early development has not been documented. The objective of this study was to determine whether pre-weanling Tsc1^GFAPCKO mice have developmental EEG abnormalities or seizures, including spasms.

Methods:

Longitudinal video-EEG and EMG recordings were performed serially on Tsc1^GFAPCKO and control mice from postnatal days 9–21 and analyzed for EEG background abnormalities, sleep-wake vigilance states, and spontaneous seizures. Spasms were also induced with varying doses of N-methyl-D-aspartate (NMDA).

Results:

The interictal EEG of Tsc1^GFAPCKO mice had excessive discontinuity and slowing, suggesting a delayed developmental progression compared with control mice. Tsc1^GFAPCKO mice also had increased vigilance state transitions and fragmentation. Tsc1^GFAPCKO mice had spontaneous focal seizures in the early neonatal period and a reduced threshold for NMDA-induced spasms, but no spontaneous spasms were observed.

Significance:

Neonatal Tsc1^GFAPCKO mice recapitulate early developmental aspects of EEG abnormalities, focal seizures, and an increased propensity for spasms. This mouse model may be useful for early mechanistic and therapeutic studies of epileptogenesis in TSC.

Introduction

Tuberous sclerosis complex (TSC) is a relatively common genetic cause of epilepsy, attributed to mutations in the TSC1 or TSC2 genes leading to excessive activation of the mechanistic target of rapamycin (mTOR) pathway^{1, 2}. Seizures occur in up to 90% of TSC patients and are often intractable to medication and non-medical treatments³. The typical age of onset of epilepsy in TSC is during infancy or early childhood, with a majority of TSC patients having their first seizure within the first year of life⁴. In particular, infantile spasms occur in about one-third of TSC patients, who frequently respond specifically to treatment with vigabatrin⁵. In addition, focal seizures or other seizure types can occur prior to or after infantile spasms and often persist chronically despite treatment, even after spasms resolve. Developing more effective treatments for seizures in TSC depends on having a better mechanistic understanding of epileptogenesis in TSC during this early developmental period.

The pathophysiology of epilepsy in TSC is incompletely understood, but is strongly associated with developmental brain pathological abnormalities, particularly cortical tubers, and cellular and molecular changes related to mTOR pathway activation. Animal models of TSC have been helpful in identifying potential mechanisms of epileptogenesis and providing preclinical evidence for novel therapeutic approaches for epilepsy in TSC^6–11. For example, multiple mouse models of TSC demonstrated evidence of rapamycin’s efficacy against epilepsy^11–14, supporting clinical trials of everolimus, an analogous mTOR inhibitor, and leading to its recent FDA approval for drug-resistant epilepsy in TSC patients^{15, 16}.

Given the early onset and prevalence of seizures in infants and young children with TSC, animal models specifically focusing on seizures or infantile spasms during early postnatal brain development are needed to understand mechanisms of epileptogenesis. However, to our knowledge, only one mouse model of TSC has been reported to have spasms or other seizures in young, pre-weanling mice^{17, 18}. One of the most characterized mouse models of epilepsy in TSC is Tsc1^GFAPCKO mice, involving conditional homozygous Tsc1 inactivation in astrocytes and neurons^{7, 19–25}. The progression of epilepsy and interictal EEG abnormalities in post-weanling juvenile and adult Tsc1^GFAPCKO mice is well-documented and correlates with progressive cellular and pathological abnormaities^{7, 12, 19}, but whether pre-weanling Tsc1^GFAPCKO mice, which have minimal pathological abnormalities, have EEG abnormalities or seizures, including spasms, has not been previously investigated. In this study, we performed detailed longitudinal video-EEG and EMG analysis of Tsc1^GFAPCKO mice between postnatal days 9–21 to characterize developmental seizure-related phenotypes in this TSC model.

Methods

Animals

Care and use of mice were conducted according to an animal protocol approved by the Washington University School of Medicine Animal Studies Committee, and consistent with National Institutes of Health (NIH) guidelines on the Care and Use of Laboratory Animals. In addition, NIH guidelines on Rigor and Reproducibility were followed, including use of randomization, blinding, both sexes, and statistical/power analyses. No sex differences were identified, although the study was not powered to detect small sex differences. Tsc1^flox/flox-GFAP-Cre knock-out (Tsc1^GFAPCKO) mice, with a mixed C57Bl/6 and SV129 genetic background and conditional inactivation of the Tsc1 gene in both neurons and glia driven by a GFAP promoter using the Cre-LoxP system, were generated as described previously^{7, 12}. Although GFAP is often viewed as an “astrocyte-specific” promoter as originally reported⁷, due to GFAP expression in embryonic neuroprogenitor cells, our GFAP-Cre mice have been shown to also involve extensive neuronal Cre-recombination^{26, 27}. Tsc1^flox/+-GFAP-Cre and Tsc1^flox/flox littermates have previously been found to have no abnormal phenotype and were used as control animals⁷. Mice were genotyped in the first week of life.

Surgery and Video-EEG monitoring

Mouse pups received surgery for EEG electrode placement on postnatal day seven (P7), as previously described²⁸. Briefly, custom wire electrode sets were constructed consisting of four Teflon coated stainless steel wires soldered to a pin header. Isoflurane anesthesia was maintained at 2–3% and midline vertical incision made to expose the skull. A hole for the frontal reference electrode was placed in the skull (anterior +0.8mm, lateral 0.5mm; bregma) using the tip of a 29g needle. The tip of the wire was positioned under the skull approximately 1mm anterior into the burr hole in contact with the cortical surface and the exposed coated portion of the wire secured to the skull using Locktite 454 followed by a small amount of dental cement. Two bilateral “active” recording electrodes were individually inserted and secured in place over the parietal cortex (posterior −2.3mm, lateral +/− 2.0; bregma) using the same techniques as the reference electrode. The final electrode wire was inserted into the neck muscle for nuchal electromyography (EMG) recordings. The exposed skull and visible wires were covered in a layer of dental cement. Tissue glue (Vetbond, 3M) was used to close the remainder of the incision. Buprenorphine (0.1mg/kg) was administered and pups recovered in a warmed chamber for one hour before returning to the dam.

Tsc1^GFAPCKO and control pups were separated from the dam and recorded in three-hour video-EEG monitoring sessions at P9, P10, P12, P14, P17 and P21 during the light phase of a 12hr light/dark cycle, as previously described²⁸. Control and knockout pups were placed in separate heated Plexiglas recording chambers and fitted with a flexible cable connected to single channel AC amplifiers (P511, Grass). EEG and EMG signals were digitized at 500Hz (MP150; Biopac) and pups were recorded in 3-hour sessions. Bilateral cortical EEG signals were acquired using a referential montage and amplified at 10,000X with highpass (1Hz) and lowpass (100Hz) filters. EMG signals were filtered with highpass (10Hz) and lowpass (300Hz) filters. After the recording session, pups were weighed and returned to the dam.

Vigilance state scoring and EEG background analysis

Bilateral EEG and nuchal muscle EMG files (.edf) were imported into ADInstruments LabChart software and a digital bandpass (1–35Hz EEG; 10–100Hz EMG) filter applied for manual review. Video-EEG and EMG recordings of control and Tsc1^GFAPCKO mice were initially screened for seizures and obvious interictal abnormalities. Later, vigilance states were manually scored using 4 second epochs and labeled as awake, non-REM sleep (NREM) or REM sleep (REM) using a combination of video, EEG, EMG and respective spectral power representations, as described previously²⁸. Vigilance state scoring parameters were assessed in older mice based on standard criteria for adult rodents and similarly applied to younger neonatal mice when such states became recognizable at P10. Wakefulness was defined as periods of cyclic lower amplitude mixed frequency EEG and high tone EMG activity for greater than half of the epoch. Brief arousal periods around ~2sec with high muscle tone during sleep transitions were also scored as awake. EEG periods dominated by higher amplitude delta wave activity with nuchal muscle atonia were scored as NREM sleep epochs, with younger mice often having low voltage discontinuous EEG surrounded by higher amplitude delta. REM sleep consisted of periods of semi-uniform theta activity or mixed frequency EEG with muscle atonia with or without brief myoclonic twitches. In younger neonatal mice prior to development of theta rhythms, the above REM criterion was used along with the absence of slow wave delta activity. EMG and video data were used to help exclude artifacts.

Percentage of time spent in each vigilance state was tabulated using LabChart software. Vigilance state transitions were calculated using the scored comment and averaged per hour. Scored transitions were used to determine bout durations by calculating the state specific time spent between vigilance state changes. The percentage of total power in delta or theta ranges were calculated using sorted vigilance state spectral data in the 1–4 Hz and 4–8.5Hz frequency ranges, respectively, and excluded epochs containing obvious artifact.

Similar to human neonates, young neonatal mice have periods of low voltage discontinuous EEG depending on state. Qualitative assessments of discontinuous EEG were manually selected using standardized two-hour raw EEG traces from P9 to P17 pups. The duration of discontinuous periods was selected from non-artifact EEG with a greater than 50% reduction in amplitude from baseline and with an average voltage below 10μV. A return in amplitude to baseline levels indicated the end of the discontinuous period. The duration of suppressed discontinuous periods were averaged for each animal and the frequency of events normalized per hour.

Total power fast Fourier transforms (FFT) were calculated using LabChart software from 1–20Hz with 512 bin size and a Hann (cosine-bell) data window using spectral data extracted from each epoch. Epochs with obvious movement artifact were excluded. FFTs were sorted by vigilance state and left and right EEG power data were averaged and displayed with a 1Hz frequency resolution.

Seizure analysis and NMDA-induced spasms

Spontaneous focal electrographic seizures were identified by their characteristic pattern of discrete periods of rhythmic spike discharges that evolved in frequency and amplitude lasting at least 10 seconds and typically ended with repetitive burst discharges, followed by brief voltage suppression. Video showing behavioral changes during electrographic seizures were generally subtle, but included forelimb/hindlimb clonus, rhythmic myoclonic twitching, and occasional brief generalized convulsive activity. Mean seizure frequency was quantified for the three-hour recording sessions. The EEG was also screened for evidence of electrographic patterns associated with spasms in rodents and people, such as bursts of high-amplitude slow or sharp waves and generalized electrodecrement of the background, as well as hypsarrhythmia^{29, 30}.

Spasms induced by N-methyl-D-aspartate (NMDA) are a well-characterized acute model of spasms in rodents^{31, 32}. To investigate the potential interaction of an induced-spasms model on the chronic genetic background of Tsc1^GFAPCKO mice, control and Tsc1^GFAPCKO mice at P12 were injected intraperitoneally (i.p.) with 5, 7.5, or 10mg/kg NMDA (Sigma, St. Louis, MO) and video recorded for one hour. Spasm-like behaviors including rapid extension of all four limbs, trunk flexion, and falling over with failure to right occurred in all groups. Videos were analyzed for latency to spasm-like events as well as total number of spasm-like events within the 60 minute recording window. Latency was recorded from the time of injection to the time of the first spasm-like event, and total number of events was recorded in ten minute bins starting after the first event. To confirm the presence of an EEG correlate to the spasm-like behaviors and to monitor for longer-lasting effects of NMDA, a subset of control and Tsc1^GFAPCKO mice was implanted with intracranial EEG electrodes on P7 as described above. These animals were injected with 5 or 10mg/kg NMDA i.p. at P12 and video-EEG recorded for three hours per day on P12–P15 and at least one additional recording period on P17 or P21. In a separate pilot study, vigabatrin (250 mg/kg, i.p.; Sigma) or vehicle was injected 24 hours prior to NMDA (5 mg/kg i.p.) and the latency and number of spasms was monitored by video in Tsc1^GFAPCKO mice^{33, 34}.

Histology/Immunohistochemistry

Histological and immunohistochemical studies were performed on P9 control and Tsc1^GFAPCKO mice to examine for abnormalities in neuronal organization and astrogliosis as previously described in older Tsc1^GFAPCKO mice^{7, 12, 13}. Briefly, mice were perfusion fixed with phosphate-buffered saline followed by 4% of paraformaldehyde. Brains were frozen and coronal sections of 45 um thickness were cut by a microtome (Microm, HM400). For neuronal organization, cresyl violet staining was performed as previously described. For astrogliosis, brain sections were stained with a primary anti-GFAP antibody (1:500, Sigma G3893) and then incubated in the secondary antibody (Alexa 488, Life Technologies A11034). Images were acquired with a Nanozoomer HT system (Hamamatsu, Bridgewater, NJ). For quantification, images of comparable regions of interest were taken and ImageJ software was used for measurements of cortical thickness, CA3 pyramidal cell layer width, and number of GFAP-positive cells.

Statistics

All statistical analysis was performed using SigmaStat 3.5 software. Quantitative differences between control and Tsc1^GFAPCKO across developmental period were analyzed by two-way repeated measures ANOVA with multiple comparisons using Tukey multiple comparisons posttests. Quantitative differences in FFT generated frequency bins were analyzed by two-way ANOVA with Tukey post hoc tests. 1 Hz bins in the frequency range of 1–20 Hz were analyzed. A linear mixed effects model analysis was also used to confirm the FFT analysis. NMDA dose responses were quantified using a two-way ANOVA with Tukey. Quantitative data are expressed as mean ± SEM. Statistical significance was defined as p < 0.05.

Results

The EEG background of Tsc1^GFAPCKO has excessive discontinuity during the neonatal period.

Longitudinal assessment of EEG background and wake-sleep states was performed by serial EEG and EMG recordings from control and Tsc1^GFAPCKO mice in daily three-hour recording sessions from P9–P14, P17, and P21, returning to the dam between sessions. As previously described²⁸, the EEG background activity of control mice at P9 is uniformly discontinuous (featuring alternating periods of higher amplitude EEG activity and periods of voltage suppression) with no state differentiation. The background activity becomes progressively more continuous with age, first in awake and REM states at P10 and then NREM by P12 (Fig. 1A). In conjunction with the development of high amplitude slow-wave activity during NREM sleep, the number and duration of discontinuous periods per hour declines with age (Fig. 1C, D). Compared with control mice, Tsc1^GFAPCKO mice exhibit excessive discontinuity at P9 overall, as well as at P10 during NREM sleep (Fig. 1B, C), suggesting a delayed maturation of EEG continuity in the Tsc1^GFAPCKO mice. Correspondingly, the duration of the suppressed EEG periods are also increased in developing Tsc1^GFAPCKO mice at P9 and P10 (Fig. 1D).

Figure 1.

The EEG background of Tsc1^GFAPCKO mice exhibits excessive discontinuity during the neonatal period. A, Representative 25 second epochs of EEG and nuchal EMG traces from a control mouse demonstrate maturation of vigilance states and continuity from P9 to P14. At P9, the background activity exhibits uniform discontinuity with no state differentiation. At P10, the awake state is characterized by a continuous low amplitude, mixed frequency background EEG and increased EMG activity. REM sleep involves similarly continuous low amplitude, mixed frequency activity but suppressed EMG. NREM demonstrates a discontinuous higher amplitude, predominantly slow-wave (delta) activity and suppressed EMG. By P12–P14, NREM gradually becomes continuous. B, Representative epochs of EEG and nuchal EMG from a Tsc1^GFAPCKO mouse demonstrate excessive discontinuity at P9 and during NREM at P10, but otherwise normal state differentiation. C,D Quantitative analysis of discontinuous periods show that Tsc1^GFAPCKO mice have increased number of suppressed periods per hour and longer durations of discontinuous EEG periods compared to control mice during the neonatal development period. p<0.001 for age by two-way repeated measures ANOVA; ***p<0.001; **p<0.01 between Tsc1^GFAPCKO and control mice at the specified age with Tukey multiple comparisons posttest; n = 12–14 mice per group.

Tsc1^GFAPCKO mice have increased vigilance state fragmentation during the neonatal period

In control mice, wake-sleep state differentiation becomes apparent by P10 (Fig. 1A). As previously described²⁸, wakefulness consists of periods of mixed frequency lower amplitude EEG with high nuchal EMG tone. REM sleep is characterized initially by a similar EEG background at P10, but can be distinguished from wakefulness by muscle atonia and sporadic myoclonic twitches. Later development of a stereotypic rhythmic theta activity, typically emerging by P14, further distinguishes REM sleep. The NREM state is characterized by development of higher amplitude slow-wave EEG activity, which initially is discontinuous at P10 and subsequently becomes continuous by P12, along with decreased muscle activity on EMG. Vigilance state maturation during the neonatal period in control mice results in increasing wakefulness and NREM sleep with less REM sleep as pups near weaning age (Fig. 2A). Overall, Tsc1^GFAPCKO display similar trends in sleep development with percentage of time spent in awake and NREM sleep increasing during the developmental period, though there are isolated differences in percentage of time in REM sleep compared with control mice (Fig. 2A). Representative hypnograms from control and Tsc1^GFAPCKO mice exemplifying these trends across P10 to P21 are shown in Fig. 3.

Figure 2.

Tsc1^GFAPCKO have relatively normal sleep-wake maturation, but increased vigilance state fragmentation. A, During neonatal development, control mice demonstrate increasing % of time in wakefulness and NREM sleep and declining REM sleep between P10–P21. Tsc1^GFAPCKO mice show similar trends in state maturation, although had isolated statistically-significant differences in REM sleep at P10 and P17, of unclear clinical significance. B, In control mice, the number of sleep-wake transitions decrease with age, as sleep cycles lengthen. Tsc1^GFAPCKO mice have increased sleep-wake transitions and NREM/REM transitions between P10–P14, indicating increased fragmentation of vigilance states. C, In control mice, average bout duration of awake and NREM sleep gradually increases between P10–P21, with a biphasic (initial increase, then decrease) in REM. Tsc1^GFAPCKO mice have a transient decreased bout duration length in NREM and REM from P10–P14, as well as increased bout duration in REM at P17, suggesting a delayed maturation in Tsc1^GFAPCKO mice compared to controls during the neonatal development period. p<0.001 for age by two-way repeated measures ANOVA; ***p<0.001, **p<0.01, *p<0.05 between Tsc1^GFAPCKO and control mice at the specified age with Tukey multiple comparisons posttest; n = 11–14 mice per group.

Figure 3.

Representative hypnograms of 3 hour monitoring periods from control and Tsc1^GFAPCKO mice during early development. Control mice show a gradual increase in awake time and decrease in REM sleep between P10 and P21, with decreasing number of transitions between states and longer bout durations. Tsc1^GFAPCKO mice show similar trends in state development, but show increased state transitions and fragmentation compared with control mice.

In control mice, the frequency of transitions between different vigilance states declines with age, as mice consolidate sleep and establish more mature, longer sleep cycles throughout neonatal development (Fig. 2B, Fig. 3). By comparison, Tsc1^GFAPCKO mice exhibit increased vigilance state fragmentation during the P10–P14 period, reflected in an increase in the number of awake-sleep, sleep-awake and NREM-REM transitions. Correspondingly, in control mice, mean bout durations of each state increase with longer sleep cycles during the early developmental period (Fig. 2C, Fig. 3). In contrast, Tsc1^GFAPCKO mice have decreased NREM and REM bout durations between P10-P14, although this difference is transient and resolves by P21, with the overall trend suggesting a delayed maturation of sleep consolidation in Tsc1^GFAPCKO mice.

Tsc1^GFAPCKO mice exhibit abnormalities in EEG spectral power during the neonatal period.

Total power spectra and fast Fourier transforms (FFTs) can be useful tools to quantify background EEG development in the neonatal mouse²⁸. In control mice, background EEG generally progresses from lower power and lower frequencies to steady increases in power and a shift to higher frequencies throughout the developmental period regardless of vigilance state (Fig. 4A). Tsc1^GFAPCKO mice display increased slow wave power in the 2–5Hz range compared to controls at P12 and 6–7Hz theta activity at P14 during NREM sleep. Similarly, Tsc1^GFAPCKO mice develop delayed power increases in theta during REM sleep compared with controls from P12–P17.

Figure 4.

Spectral EEG power analysis demonstrate maturational abnormalities in Tsc1^GFAPCKO mice. A, In control mice, Fast Fourier transforms (FFT) of spectral EEG display increased overall power and relative power in higher frequencies with age regardless of vigilance state (P10 and P21 not shown for clarity). Tsc1^GFAPCKO mice (lighter line) mice demonstrate less overall power in the 3–5Hz range during slow-wave development at P12 and 5–6Hz at P14 during NREM sleep periods. Tsc1^GFAPCKO mice also have delayed development of theta rhythms (5–7Hz) during REM sleep from P12–P17. p<0.001 for age by two-way repeated measures ANOVA; **p<0.01, *p<0.05 at P12; #p<0.05 at P14; ▲p<0.05 at P17 between Tsc1^GFAPCKO and control mice at the specified age with Tukey multiple comparisons posttest; n = 11–14 mice per group. B, As control mice mature, a decline in the percentage of total delta power (1–4Hz) and an increase in the percentage of total theta power (4–8Hz) occurs in the background EEG throughout neonatal development. Tsc1^GFAPCKO mice EEG (red lines) show a delayed decline in percent of delta power during awake, NREM and REM states and a delayed increase in percent of theta during NREM throughout the developmental period compared to control (blue lines). p<0.001 for age by two-way ANOVA; ***p<0.001, **p<0.01, *p<0.05; percent delta power, ##p<0.01 percent theta power between Tsc1^GFAPCKO and control mice at the specified age with Tukey multiple comparisons posttest; n = 11–14 per group.

When examining the developmental changes in delta and theta power with age, the percent of total power in the delta (1–4Hz) range and theta (4–8.5Hz) range have an inverse relationship across development. In control mice, slower frequency delta power is replaced by increasing higher frequency theta in awake and REM periods as the EEG becomes mature (Fig. 4B). Tsc1^GFAPCKO mice show a similar transition between delta and theta power, but have a delayed decline in the percentage of delta power regardless of vigilance state from P10–P17. Tsc1^GFAPCKO mice also have delayed percent theta increases in NREM sleep from P12–P14.

Spontaneous seizures are limited during development in pre-weanling Tsc1^GFAPCKO mice

Tsc1^GFAPCKO mice were monitored by video-EEG for evidence of spontaneous seizures. Stereotypical electrographic seizures occurred in a majority of Tsc1^GFAPCKO mice (61.5%), characterized primarily by a paroxysmal, evolving increase of spike frequency and amplitude along with rhythmic spike discharges lasting more than 10 seconds (mean duration = 91.5 ± 12.2 sec), followed by brief postictal suppression period (Fig. 5A). The seizures often started in one hemisphere, but then spread bilaterally, suggesting they were focal-onset seizures. On video analysis, behavioral changes during electrographic seizures were generally subtle and variable, but often included forelimb/hindlimb clonus, rhythmic myoclonic twitching, and occasional brief generalized convulsive activity. Seizures in Tsc1^GFAPCKO mice were limited to P9 and P10 pups with no electrographic seizures observed after the initial neonatal period through P21 (Fig. 5B). No control mice had seizures. Furthermore, no ictal EEG patterns or behavioral episodes consistent with spasms, such as bursts of high-amplitude slow or sharp wave or generalized electrodecrement of the background, were seen in control or Tsc1^GFAPCKO mice. To investigate for potential pathological correlates to the EEG abnormalities in neonatal Tsc1^GFAPCKO mice, which have been identified in older mice^{7, 12, 13}, we examined but found no significant changes in neuronal organization (cortical thickness, CA3 pyramidal cell layer thickness) and astrogliosis (increased GFAP-positive cells) in neocortex and hippocampus of P9 Tsc1^GFAPCKO mice (Supporting Information Fig. S1).

Figure 5.

Tsc1^GFAPCKO mice have spontaneous focal seizures during an age-limited period of neonatal development. A, A representative electrographic seizure on EEG and EMG from a P9 Tsc1^GFAPCKO mouse evolves with increasing frequency and amplitude of spikes and brief post ictal suppression. The seizure appears to start first in the left hemisphere but then spreads bilaterally, suggesting a focal-onset seizure. EMG from nuchal muscle shows a modest increase in myogenic activity. B, Mean seizure frequency is increased in P9 Tsc1^GFAPCKO and subsides by P12. ***p<0.001 % in Tsc1^GFAPCKO compared to control by two-way repeated measures ANOVA with Tukey multiple comparisons posttest; n = 11–14 mice per group.

Tsc1^GFAPCKO mice have a decreased threshold and increased frequency of NMDA-induced infantile spasms

Given the lack of spontaneous spasms, the effect of NMDA in inducing spasms was tested in Tsc1^GFAPCKO mice. Consistent with previous studies^{31, 32}, pilot studies in control P12 mice identified spasm-like behaviors, including rapid simultaneous extension of all four limbs, trunk flexion, and falling over with failure to right, which occurred in clusters following intraperitoneal injections of NMDA (Supporting Information Video S1). Video-EEG recording in a subset of mice showed almost immediate suppression of EEG following NMDA injection regardless of dose, followed by the emergence of periodic clustered ictal sharp and slow waves, correlating with these spasm-like behaviors (Fig. 6A). We then tested whether these NMDA-induced spasm-like behaviors varied in a dose-dependent manner and differed between control and Tsc1^GFAPCKO mice, quantifying the latency and frequency of these behaviors on video. At 7.5 and 10mg/kg NMDA, behavioral spasms reliably appeared around 8 minutes after injection in both control and Tsc1^GFAPCKO mice and remained nearly constant for about one hour. However, at 5mg/kg, the latency to spasms onset was significantly longer in controls compared to Tsc1^GFAPCKO mice (Fig. 6B). At 7.5 and 10 mg/kg, spasms appeared continuous and individual behaviors were difficult to isolate and identify; however, at 5mg/kg, clusters of spasm-like behaviors could be individually identified and quantified, showing that Tsc1^GFAPCKO mice had a significantly greater number of spasm-like events in the first thirty minutes after spasms onset (Fig. 6C). In the subset of mice recorded by video-EEG, no evidence of additional spontaneous spasms was seen beyond the acute period on P12 as monitored on subsequent days through P17 or P21 (n=3–5 Tsc1^GFAPCKO mice with 5 mg/kg and 10 mg/kg NMDA injections on P12). In a separate cohort of Tsc1^GFAPCKO mice, vigabatrin treatment (250 mg/kg, i.p., 24 hours prior to 5 mg/kg NMDA at P12) significantly increased the latency to spasms onset compared with vehicle but had a non-significant decrease in spasm number (Supporting Information Fig. S2).

Figure 6.

Tsc1^GFAPCKO mice have a decreased threshold and increased frequency of NMDA-induced spasms. A, Bilateral EEG traces showing a representative cluster of spike-wave discharges, correlating with behavior spasms, and background suppression following NMDA (10mg/kg) injection. B, Video analysis shows decreased latency to spasm-like events for Tsc1^GFAPCKO mice compared to control at 5mg/kg NMDA, as well as increased latency for control animals at 5mg/kg compared to 7.5mg/kg and 10mg/kg. **p<0.01, *p<0.05; for Tsc1^GFAPCKO compared to control by two-way ANOVA with Tukey multiple comparisons posttest; n=10–12 mice per group. C, Further video analysis shows that 5mg/kg NMDA elicits increased frequency of spasm-like events in Tsc1^GFAPCKO mice compared to controls over the first 30 minutes after the initiation of spasms. *p<0.05; Tsc1^GFAPCKO compared to control by two-way repeated measures ANOVA with Tukey multiple comparisons posttest; n = 10–12 mice per group.

Discussion

In this study, we performed a longitudinal assessment of background EEG, vigilance states, and seizures in Tsc1^GFAPCKO mice during pre-weanling neonatal development. The main findings are: 1) the background EEG of Tsc1^GFAPCKO mice demonstrates excessive discontinuity, increased slow wave (primarily delta) power and delayed developmental maturation compared with control mice, 2) Tsc1^GFAPCKO mice have normal vigilance state differentiation and maturation, but increased transitions between vigilance states and lower bout durations, indicating increased state fragmentation, 3) Tsc1^GFAPCKO mice exhibit spontaneous focal-onset seizures during the age-limited neonatal period of P9-P10, and 4) Tsc1^GFAPCKO mice had no evidence of spontaneous spasms or pathological abnormalities, but did have a reduced threshold and increased frequency of NMDA-induced spasms. Overall, these findings extend the previous characterization of the epilepsy phenotype in juvenile and adult Tsc1^GFAPCKO mice to the neonatal/infantile period and recapitulate some clinical or EEG features of human TSC during early development, but did not identify a spontaneous or chronic infantile spasms phenotype.

A significant technical strength of this study is the utilization of chronic, serial EEG and EMG recordings in pre-weanling neonatal mice, allowing a detailed longitudinal analysis of EEG background activity, vigilance state maturation, and seizures during early development in Tsc1^GFAPCKO mice. Using similar methods, we previously documented maturational changes in EEG patterns and sleep-wake states in control mice²⁸. By comparison, the excessive discontinuity, increased delta power, and delayed transition between delta and theta activity in Tsc1^GFAPCKO mice suggest an immature or delayed maturation of the EEG. These abnormalities likely represent early, albeit transient, EEG biomarkers of encephalopathy or other neurological dysfunction, as juvenile and adult Tsc1^GFAPCKO mice have previously been documented to have progressive seizures and learning deficits^{12, 19, 35}.

Neonatal mice normally undergo rapid maturational changes in sleep-wake cycles, with increasing awake and NREM sleep, decreasing REM sleep, and a corresponding decrease in state transitions and increase in bout durations. Tsc1^GFAPCKO mice display similar trends in vigilance state maturation, but demonstrate increased number of transitions between different sleep and awake states and decreased bout durations at early developmental ages, indicating excessive state fragmentation and again suggesting delayed developmental maturation. This may also represent an early biomarker of later sleep abnormalities demonstrated in adult Tsc1^GFAPCKO mice³⁶, as well as sleep disorders commonly seen in TSC patients^{37, 38}.

Previous work documented that one hundred percent of adult Tsc1^GFAPCKO mice develop seizures, typically starting around 4 weeks of age and becoming progressively more frequent over time^{12, 19}. Seizure progression in Tsc1^GFAPCKO mice correlates with development of cellular and pathological abnormalities, such as astrogliosis and neuronal disorganization⁷. However, seizures in pre-weanling Tsc1^GFAPCKO mice, which have minimal pathological abnormalities (Supporting Information Fig. S1), have not previously been reported. In the present study, we documented focal seizures in approximately sixty percent of neonatal Tsc1^GFAPCKO mice, with seizures limited to P9 and P10, remitting thereafter for unknown reasons. Although control mice did not exhibit seizures, it is possible that the transient stress of surgery superimposed on the seizure-susceptible background of Tsc1^GFAPCKO mice had a synergistic or additive (i.e., “two-hit”) effect in causing neonatal seizures in these mice. In comparison, another mouse model of TSC, Tsc1+/− heterozygous mice, has been reported to have seizures limited to between P9–P18, which subsequently remitted¹⁷. The susceptibility to seizures during an early developmental time window in both Tsc1+/− mice and Tsc1^GFAPCKO mice does partially mimic the natural history of epilepsy in TSC patients, which most commonly presents with seizures within the first 1–3 years of life⁴.

As infantile spasms occurs in about one-third of TSC patients and are a particularly clinically significant seizure type, it is disappointing that spontaneous spasms were not identified in Tsc1^GFAPCKO mice. There are multiple mouse models of TSC and epilepsy⁶, but surprisingly only one model has been reported to have spasm-like seizures¹⁷, with no subsequent studies confirming or further characterizing that model’s spasms phenotype in more detail. In general, there are only a handful of models of infantile spasms related to other etiologies and no model perfectly recapitulates all clinical and electrographic features of infantile spasms²⁹. It is possible that rodent brains have intrinsic differences that limit their capacity to generate realistic models of spasms observed in people.

As only a subset of TSC patients develop spasms despite a larger number developing other seizure types, there are likely other environmental or genetic modifying factors that influence the propensity to develop spasms on the background of a TSC1 or TSC2 gene mutation. Given the lack of spontaneous spasms in Tsc1^GFAPCKO mice, we attempted to generate a “two-hit” model of IS by superimposing a pharmacological agent known to induce spasms in normal rodents upon the genetic defect of Tsc1^GFAPCKO mice^{31, 32}. We found that Tsc1^GFAPCKO mice did have a reduced threshold and increased frequency of NMDA-induced spasms, suggesting an increased propensity for spasms, though no spontaneous spasms were observed on subsequent days beyond the acute period after NMDA injection. Other “two-hit” approaches, such as combining an inflammatory insult or other genetic mutations with Tsc1^GFAPCKO mice, could potentially be used to generate a chronic model of IS^{33, 39, 40}.

This study has a number of limitations. Although the method of performing longitudinal serial EEG and EMG recordings in the same mice from P9 to P21 is relatively unique and advantageous²⁸, individual recording sessions were limited to three hours due to dependence on feeding by the dam, so a comprehensive evaluation of sleep-wake cycles was not obtained and some seizures could have been missed. While Tsc1^GFAPCKO mice were originally thought to target Tsc1 inactivation specifically in astrocytes⁷, it was subsequently determined that the GFAP driver also targets neurons^{26, 27}, thus the model recapitulates neurological aspects of TSC in involving both glia and neurons. However, Tsc1^GFAPCKO mice have relatively diffuse pathological abnormalities, such as astrogliosis and neuronal disorganization, and do not recapitulate focal, tuber-like abnormalities. Although the importance of tubers in the neurological phenotype of TSC is increasingly under scrutiny, more realistic models of TSC that recapitulate tubers may increase the clinical relevance and validity of these models^{10, 41}. Despite these limitations, however, Tsc1^GFAPCKO mice have been useful in elucidating mechanisms of epileptogenesis in TSC and testing novel, clinically-translatable treatments, such as mTOR inhibitors¹². Initial evaluation of vigabatrin demonstrated a limited effect on induced NMDA spasms in Tsc1^GFAPCKO mice, but future studies will include more comprehensive, rigorous preclinical drug studies on vigabatrin and other novel therapeutic agents. As the trend of therapeutic approach has shifted toward early treatment and prevention of neurological manifestations of TSC, the findings from this study extend the potential utility of this pre-clinical model to early developmental periods.

Key Points.

• The interictal EEG of neonatal Tsc1^GFAPCKO mice has excessive discontinuity and slowing, indicating delayed developmental maturation.

• Neonatal Tsc1^GFAPCKO mice have increased vigilance state transitions and fragmentation.

• Tsc1^GFAPCKO mice have spontaneous focal seizures in the early neonatal period.

• Tsc1^GFAPCKO mice have a reduced threshold and increased severity of NMDA-induced spasms, but no spontaneous infantile spasms.