Abstracts

Rationale: Post-traumatic epilepsy (PTE) is a common acquired epilepsy seen in about 10-20% of patients with Traumatic Brain Injury (TBI). Latency to PTE can vary in individuals. While some imaging, genetic and electrophysiological biomarkers like HFOs have been explored, none accurately predict who develops PTE. It is vital to identify biomarkers to stratify those at risk to reduce morbidity and mortality associated with PTE. Here we explored the association of sleep spindle morphology and PTE in a mouse model of severe TBI. Methods: We performed severe TBI (controlled cortical impact-CCI; n=40) or Sham injury (craniotomy without TBI; n=22) or no craniotomy (n=6) along with epidural EEG and EMG electrode placement under isoflurane anesthesia in CD-1 mice. TBI and sham cohorts underwent video-EEG recording (using Natus Neuroworks) from days 1-7; 60-67, 90-97 and 180-187 after surgery, whereas those with no craniotomy were recorded from days 1-7 only. All EEG files were manually scored for seizures. TBI cohort was further divided into those with seizures (TBI-S) and without seizures (TBI-NS). Day 4 and 5 of some EEG recordings were manually scored for sleep-wake patterns in 4sec epochs using Sirenia Sleep (n=9, 14 for TBI group; n=6, 6 for Sham in week 1 and month 2 respectively; n=6 for no craniotomy controls). NREM sleep epochs were extracted from sleep scored files and used for sleep spindle detection. We implemented an automated spindle detection algorithm based on Ferrarelli et al (2007) in Wonambi package, Python. The parameters used to detect spindles were frequency between 9-15Hz; duration of 0.3-2sec; detection threshold of 2 and a selection threshold of 2 (Figure 1). The selection and detection thresholds were chosen after comparing various thresholds in automated analysis to manually scored spindles (1hour NREM segments, n=6). The Sensitivity and F1 scores (an indicator of accuracy and precision) for automated scoring were calculated at 0.84+-0.04 and 0.75+-0.05 (Mean+-SD) respectively. We compared differences in time spent in NREM and spindle characteristics including density (number of spindles per 60sec epochs); frequency (Hz); duration (sec); mean peak amplitude (µV) and absolute spindle power (µV2) between groups across time using single factor ANOVA with post-hoc correction. Results: 22% (9/40) of animals with CCI had seizures recorded anywhere between days 1 to 187, but none were seen in sham group. On sleep analysis, the time spent in NREM sleep was not different between groups in week 1. At month 2 sham controls spent significantly more time in total NREM sleep compared to TBI-NS or TBI-S (Table 1). Spindle characteristic analysis showed that mice with no craniotomy had significantly higher spindle power, mean peak amplitude and frequency compared to those with sham injury or CCI at week 1 (Table 1). The mean peak amplitude and power of spindles remained unchanged in sham group between week 1 and month 2, but TBI groups had a progressive increase in both from week 1 to month 2. Moreover, spindle power and peak amplitude were much higher in TBI-S group compared to TBI-NS (Table 1). No differences were noted in sleep spindle density or duration between groups.  Conclusions: Sleep spindle characteristics may serve as potential biomarkers for PTE after TBI. Further confirmatory studies are needed in other TBI models. References:Ferrarelli F, Huber R, Peterson MJ, Massimini M, Murphy M, Riedner BA, Watson A, Bria P, Tononi G. Reduced sleep spindle activity in schizophrenia patients. Am J Psychiatry. 2007;164(3):483-92  Funding: Department of Defense: W81XWH-17-1-0049