Abstracts

Rationale: A major challenge in the field of temporal lobe epilepsy (TLE) research is to separate pathological high frequency oscillations (pHFO), which are biomarkers for both epileptogenesis and seizure focus, from normal high frequency oscillations (ripples), which have functional roles in cognition. It is unclear whether in TLE the hippocampus generates both pHFO and normal ripple, and if so, what criteria should be used to properly distinguish between them. Methods: We perform hippocampal single unit and local field potential recordings in rats with chronic epilepsy (low-dose KA) (n=4) and in control rats (n=4) across several behavioral brain states. We analyze temporal features of fast oscillations, their brain state dependence, and their recruitment of single units in order to determine what type of events are generated in the hippocampus of rats with TLE, and the best criteria for separating them.  Results: We find that in TLE, the CA1 network of hippocampus generates fast oscillations that cluster into two types based on waveform properties. Type 1 are faster and associated with large amplitude slow waves (n = 923, medians; 245.9 Hz, 520.0  μV), while type 2 are slower and associated with low amplitude slow waves (n = 1166, 184.9 Hz; 253.8  μV). To test whether one type was similar to ripples recorded in control rats, we performed ROC analysis. Type 1 are well distinguished from ripple, and will henceforth be called pHFO, while type 2 are indistinguishable from ripple, suggesting that CA1 in TLE is capable of generating ‘ripple-like’ events. However, shape parameters are highly dependent on exact recording location, so we sought additional criteria for separating pHFO and ripple-like oscillations. Analysis of foraging sessions, in which animal position was tracked, reveal that both ripples in control rats and ‘ripple-like’ events in TLE rats mostly occur during periods of immobility (ctrl; n = 691, 0.8 cm/s; ripple-like; n = 293, 0.5 cm/s). In contrast, pHFOs occur over the entire range of running speeds (n = 490, 4.7 cm/s), suggesting that brain state dependence could be used to classify normal and pathological activity in TLE (ctrl vs ripple-like, n.s.; ctrl vs to pHFO, p < 0.05; Kruskal-Wallis ANOVA). Finally, some CA1 pyramidal neurons participate in both pHFO and ripple-like events, but many are modulated by only one event type (60 % modulated by ripple-like only, 8 % modulated by pHFO only) suggesting that sub-networks in CA1 are more or less pathological, and that observing individual neurons could additionally help separate normal and abnormal activity.  Conclusions: Our results show that the hippocampal network in TLE generates two types of fast oscillations; one type is more normal and the other type is pathological. Brain state dependence and single unit modulation are features that could be used to properly classify fast oscillations. These results highlight that memory networks in epilepsy are not completely pathological, and interventions that selectively target pathological activity should be the goal of treatment. Funding: NIH MH100349, Epilepsy Foundation 157927, Hellman Family Foundation, NRSA 1F32MH096526-01A1