The findings described here suggest that clinical seizures begin from sub-millimetre scale epileptiform activity that spreads to neighbouring regions before a sufficient population of synchronously firing cells is recruited to be detectable on the macroelectrodes.
The lack of statistical specificity of microperiodic epileptiform discharges for epileptic brain, and their similarity to activity described after local tissue trauma, suggests that the mechanism of seizure generation involving microperiodic epileptiform discharges may be different than for microseizures, and perhaps responsible for generating post-traumatic seizures.
In previous studies, microelectrode recordings capable of probing epileptic microdomains were limited to isolated regions (<1
cm diameter) of hippocampus or neocortex, and included patients with a limited range of tissue pathologies, particularly mesial temporal sclerosis and non-specific gliosis (Bragin et al., 2002b
; Schevon et al., 2008
; Worrell et al., 2008
). Because these studies were limited to patients with epilepsy, the specificity of interictal microdomain discharges to epileptic brain could not be explored. Additionally, the spatiotemporal and spectral evolution of microdomain discharges in relation to seizures recorded on clinical macroelectrodes could not be evaluated because sufficient simultaneous recordings were not available. In this study, microseizures were observed in all patients with epilepsy and were increased in the ictal onset zone. This suggests that microseizures are pathological and probably ubiquitous within the ictal onset zone, because the microelectrode arrays dramatically undersample the volume of tissue under an implanted grid. Although the observation of microperiodic epileptiform discharges in control brain could be associated with electrode-related tissue damage, rare microseizures were also observed in control brain, suggesting that potentially pathological microdomain activity can be present in normal brain. It is unlikely that either microseizures or microperiodic epileptiform discharges arise solely from tissue damage, because microseizure and microperiodic epileptiform discharge rate and duration were increased in the ictal onset zone and ~20% of clinical seizures were preceded by evolving microdomain activity.
In a rat model of epilepsy created by intra-hippocampal kainic acid injection (Bragin et al., 2000
), pathological high-frequency oscillations emerged in microdomains (<1
) weeks to months before spontaneous seizures developed. Epileptogenesis was proposed to be initiated by local cellular injury, resulting in small clusters of pathologically interconnected neurons. Bragin et al. (2000)
hypothesize that pathologically interconnected neurons generate hypersynchronous discharges that kindle the brain through the creation of new pathological microdomains, and the emergence of an interacting network of pathologically interconnected neuron clusters. The results presented here, from human epileptic brain, are consistent with this hypothesis. The fact that they are independent of tissue pathology suggests that the topographically fractured functional organization may underly the process of epileptogenesis.
The progression of normal brain tissue to epileptic tissue capable of generating spontaneous seizures (epileptogenesis) may reflect a continuum of increasing density or connectivity of pathological microdomains. Similarly, the transition from normal brain activity to seizure (ictogenesis) may involve the interaction and spread of pathological microdomain activity (). In this model of seizure generation, the earliest local field potential oscillations of seizures are multifocal, asynchronous microseizures and microperiodic epileptiform discharges that recruit surrounding microdomains until a critical volume or network of tissue progresses into a large-scale seizure. Our observation of both microseizures and microperiodic epileptiform discharges in normal brain supports the hypothesis that epileptiform activity can occur in non-epileptic tissue but is controlled by homoeostatic mechanisms or is insufficient in spatial or temporal density to initiate a seizure. That focal seizures arise in normal individuals after exposure to conditions such as hyperglycaemia, electrolyte abnormalities and toxic exposures strengthens the hypothesis that an individual's ‘seizure threshold’ may be a function, in part, of the volume of tissue generating microseizures and microperiodic epileptiform discharges.
Figure 4 Hypothetical model of the ictal onset zone composed of sparse, non-contiguous, pathological microdomains that are characterized by the ability to generate microseizures, microperiodic epileptiform discharges and high-frequency oscillations. Left: the (more ...)
Although these results implicate pathological microdomains in epileptogenesis and ictogenesis, the anatomical and cellular substrate of epileptic microdomains cannot be elucidated from our data. We can suggest that the microseizures and microperiodic epileptiform discharges we recorded from our neocortical microwires are probably generated in superficial cortical layers given their proximity to the pial layer and the density of synaptic currents known to occur there. We cannot speak to the origins of these events from our hybrid depth electrodes that are typically implanted in the three-layered archicortex of the mesial temporal structures.
The extracellular local field potentials recorded by microwires is primarily a manifestation of the co-operative activity of the local neuronal population. Until recently, the local field potential was thought to exclusively reflect the summation of post-synaptic currents because of their relatively slow dynamics. This is the reason that extracellularly recorded action potentials—with the fast Na+
current being the largest contributor—are detected only if the microwire is close to the cell. The amplitude of the extracellular action potential falls off rapidly with distance, and the events are unlikely to constructively sum because of their brief duration. However, it has been recognized that there are additional sources of local field potentials not associated with synaptic currents and they can be significant [reviewed in Buzsaki et al. (2003)
]. They include Ca2+
-mediated action potentials generated in dendrites (Wong et al., 1979
), slow long-lasting calcium-mediated potassium currents, voltage-dependent intrinsic oscillations in neurons (Leung and Yim, 1991
) and currents related to glia–neuron interactions (Tian et al., 2005
). Understanding the cellular mechanisms underlying the generation of microseizures and microperiodic epileptiform discharges is important and a focus of our current research but is beyond the scope of the work described here.
In support of epileptiform activity inherently originating at the microscale, previous work from our group and others (Schevon et al., 2008
; Worrell et al., 2008
) has demonstrated in situ
epileptiform activity on scales as small as 1
or less; similar dimensions have been observed in animal models (Bragin et al., 2002a
; Supplementary Fig. 6
). Recent work in resected human epileptogenic cortex demonstrated runs of epileptiform spikes in 0.5
mm in vitro
slices, the approximate width of a human cortical column (Mountcastle, 1978
). This spontaneous activity bears morphologic similarity to the microseizures and microperiodic epileptiform discharges described in this work. Furthermore, the in vitro
activity was necessarily generated by highly localized neuronal networks, and was dependent on gap junction connections between neurons, suggesting a non-synaptic generator of some forms of epileptiform activity (Roopun et al., 2010
). Also, realistic computational models of single cortical columns have been shown to be capable of generating a rich array of physiological and epileptiform discharges (Traub et al., 2005
A potential concern regarding these phenomena is that, because of their frequent restriction to single microwires, they are a form of artefact. There are several observations that make this unlikely however.
- On occasions we see spread to adjacent microwires (A4, and Supplementary Fig. 6). Such spread has also been shown by others with more tightly placed microwires e.g. ~400µm apart (Schevon et al., 2008).
- As can be seen by example in A2, we observe microperiodic epileptiform discharges occurring simultaneously, yet asynchronously, with different periods on separate microwires effectively excluding an exogenous source of this artefact. Also arguing against exogenous artefact sources is the fact that microseizures occur asynchronously on independent microwire channels.
- Between events we record ‘morphologically conventional’ EEG on these channels making electrode damage, high impedance and channel-restricted electronic failure quite unlikely. Furthermore, we have recorded from a saline–gelatin solution simultaneously with a patient recording, thus exposing the electrodes and electronics to all the same noises present in the patient recording milieu. No microseizures or microperiodic epileptiform discharges were detected over 4 days of recording in the saline–gelatin, while these events were detected in the patient.
- These signals localize the clinically determined ictal onset zone, a finding whose most conservative probability estimate of occurring by chance was <0.03, as shown in .
- These phenomena display morphologic and spectral structures that are not seen as common sources of artefact in electrophysiologic recording such as 60Hz line noise and movement artefact. We do see these types of artefact and they are easily identified visually or algorithmically and excluded from the analysis.
Multiple investigators have hypothesized that the functional cortical column may play a fundamental role in the initiation and propagation of seizures (Ebersole and Levine, 1975
; Gabor et al., 1979
; Reichenthal and Hocherman, 1979
). In this study, we observe microscale electrophysiology whose spatial extent is consistent with the scale of cortical columns. Cortical columns have sufficient recurrent excitatory inter-connections (Ayala et al., 1973
) to provide the cortical substrate for pathologically interconnected neuron clusters. Therefore, we hypothesize that relatively sparse pathological cortical columns are the anatomical substrate of focal neocortical epilepsy, ‘the sick column hypothesis’. Although archicortex does not exhibit columnar organization, pathologically interconnected neurons could serve as the ‘sick column’ substrate in these structures.
The observation that focal seizures begin on spatial and temporal scales not probed by current clinical intracranial EEG systems may explain the difficulty in identifying a focal discrete region of seizure onset, the apparent random nature of seizure occurrence and the limited success of first-generation responsive stimulation devices that attempt to detect and abort seizures. Microseizures could provide interictal biomarkers of epileptic tissue, possibly improving the efficacy of epilepsy surgery. They may also illuminate the process of ictogenesis, and thereby open new therapeutic windows for seizure warning and preventive stimulation devices.