Spontaneous High-Frequency Oscillations

In limbic structures of several mammalian species, including rodents, bats, non-human primates and humans, spontaneous HFOs termed ripples contain spectral power between 80 and 200 Hz and have a duration of ten’s to hundreds of milliseconds.^1–4 The largest amplitude ripples are found within the pyramidal cell layer of hippocampal subfield CA1 (Figure 1A), and are also present in CA3, subiculum, entorhinal cortex and amygdala.^5,6 In vivo studies in rodents have not observed normal ripples in the granule cell layer of dentate gyrus, although a study in cats observed ripple-frequency ‘minispindles’ that occurred in CA1, but were also present in the hilus of the dentate gyrus.⁷

Figure 1

Examples of normal HFO and pHFOs. A) Spontaneous ripple recorded in CA1 of normal rat. Ripple is present in wide bandwidth (0.1 Hz–1 kHz; top) and corresponding bandpass filter (80–200 Hz; middle) traces recorded just above the pyramidal (more...)

Hippocampal ripples occur more frequently during awake immobility and slow wave sleep (SWS) than states characterized in animals by prominent theta rhythm such as exploratory behavior and rapid eye movement (REM) sleep. During episodes of SWS, hippocampal ripples can occur bilaterally and are often associated with large amplitude slow waves called sharp waves (SPWs) generated in hippocampal CA3 (Figure 1A). ⁸ Multi-electrode studies show there is strong coherence between ripples recorded in CA1 over distances of 4 or 5 mm that suggests neuronal activity coordinated on a temporal scale of several milliseconds can occur across large areas of hippocampus.^5,9,10

Neocortical HFOs

Spontaneous ripple-frequency HFOs are also present in neocortex. In cats, neocortical HFOs occur most frequently during episodes of SWS and states of ketamine anesthesia, and their appearance in isolated cortical tissue suggests intracortical circuits can support the generation of HFOs.¹¹ Neocortical HFOs are often associated with the electroencephalogram (EEG) depth-negative component of the neocortical slow wave oscillation (0.5 – 1Hz). Coherence among neocortical HFOs is strong over distances up to 10 mm within individual gyri, while the coherence weakens between HFOs in different gyri.¹¹

In rat and human neocortex, somatosensory evoked potentials (SEPs) are associated with HFOs that contain spectral frequencies between 200 and 600Hz (Figure 1B). In rats, rapid mechanical stimulation of the vibrissae or electrical stimulation of the thalamic ventrobasal nuclei can evoke HFOs in somatosensory barrel cortex.^12–14 In humans, peripheral nerve stimulation elicits SEPs that contain HFOs with different components of the HFO arising from thalamic and cortical circuits.^15–20

Sensory-evoked neocortical HFOs are typically superimposed on the earliest components of the biphasic positive-negative SEP (Figure 1B) that can propagate in-phase over several millimeters in rat barrel cortex.¹³ Simultaneous stimulation of individual vibrissa evokes HFOs that propagate across barrel cortex and can constructively interact in a fashion that results in a supra-linear summation of HFOs within sites of interaction.^21,22 These latter studies suggest that the locally facilitated HFO response could reflect a recruitment of additional neurons that discharge due to the in-phase interactions between propagating HFO.

Neuronal Correlates of HFOs in Normal Mammalian Brain

Much of what is known regarding the neuronal activity associated with ripples derives from in vivo wide bandwidth recordings in rat hippocampus. In CA1 of behaving rats, the SPW-ripple complex is associated with a significant increase in firing from pyramidal cells and several types of interneurons. Pyramidal cells discharge during the trough of the extracellular ripple oscillation, but do not fire during each cycle of the ripple.²³ Interneurons such as basket cells that provide perisomatic inhibition to pyramidal cells can fire at ripple-frequencies and the timing of their discharges coincide closely with those from pyramidal cells.²⁴ Furthermore, recording from pyramidal cells reveals the presence of an intracellular oscillatory potential that corresponds with extracellular ripple oscillation, and is likely mediated by chloride ions due to the activation of GABA_A receptors from discharging basket cells.⁹ These latter data suggest hippocampal ripples reflect fast inhibitory postsynaptic potentials (IPSPs) that can powerfully regulate the firing and timing of pyramidal cell discharges.

Excitatory and inhibitory neurons are also involved in the generation of neocortical HFOs. Fast spiking cells, presumably GABA-containing neurons, discharge bursts of action potentials at intra-burst intervals that correspond with the frequency of the extracellular HFO.¹¹ Fast spiking cell discharges can precede those from regular spiking cells, which occur during the trough of the extracellular HFO.²⁵ Similar to the coordinated pattern of firing between pyramidal cells and interneurons during hippocampal ripples, neocortical HFOs could reflect IPSPs from fast spiking cells that regulate the timing and firing of regular spiking cells.

Mechanisms Generating Normal HFOs

The single neuron studies in the preceding section indicate inhibitory networks could play an important role during hippocampal ripples and neocortical HFOs. Inhibitory interneurons can entrain pyramidal cells or other interneurons into a rhythmic pattern of firing,²⁶ and it is possible that phasic activation of hippocampal interneurons by external input could produce rhythmic IPSPs that regulate the timing of pyramidal cell discharges.⁹ In contrast, some studies have shown that blocking GABA_A receptor activation does not abolish spontaneous hippocampal ripples or evoked neocortical HFOs.^27,28 However, under conditions of reduced inhibition, ripples contain higher spectral frequencies, significantly longer durations, and can be associated with epileptiform discharges, all of which raises questions whether ripples under these conditions in vitro are the same as in vivo. Furthermore, blocking GABA_A-mediated inhibition in hippocampal slice can generate ripple-frequency HFOs in the isolated dentate gyrus that reflect synchronously bursting granule cells.²⁹ While there is no direct evidence for ripples in the normal dentate gyrus, studies show that abnormal ripple-frequency HFOs do occur in the epileptogenic dentate gyrus of epileptic rats.^30,31

Electrotonic coupling mediated through gap junctions is another mechanism that could support the generation of HFOs. In vitro studies show chemical agents that interfere with gap junction communication between hippocampal neurons suppress ripples, while manipulations intended to increase gap junction conductance are associated with an increase in ripples.³² Furthermore, SPW-ripple complexes occur less frequently in hippocampal slices from mice deficient in a gene encoding a neuronal gap junction subunit than in tissue slices from littermate controls.³³ These in vitro data are consistent with the suppression of CA1 ripples in vivo in rats administered halothane anesthesia, which can reduce electronic coupling.⁹ In addition, computer simulations predict the generation of ripples in hippocampal networks that contains axon-to-axon gap junctions between pyramidal cells.³⁴ These data suggest that electrical signaling through gap junctions may contribute to the synchronous neuronal discharges associated with ripples.

Non-synaptic mechanisms such as ephaptic or field effects could also be involved with the generation of HFOs.^35,36 Ephaptic interactions arise when depolarizing currents associated with neuronal action potentials generate an electric field in the extracellular space that depolarizes adjacent neurons.³⁷ The probability that ephaptic interactions could trigger action potentials in nearby neurons increases if neurons are in close proximity to each other and neighboring neurons are already close to firing threshold. Field effects occur on a larger spatial scale and can arise when local populations of neurons generate large amplitude field potentials, e.g. SPWs, that depolarize nearby inactive neurons.³⁸ Given the organized parallel arrangement of principal neurons in hippocampus and neocortex, the summation of extracellular currents across local principal cell networks could be sufficient to depolarize and coordinate the firing during hippocampal ripples and neocortical HFOs.

Physiological Role of HFOs

Spontaneous ripples have been implicated in information transfer between hippocampal and neocortical structures during sleep.³⁹ During SWS, the large scale neuronal activation associated with SPWs and the increase in precisely timed discharges of pyramidal cells during ripples provide ideal conditions for synaptic transmission to extra-hippocampal sites. In vitro evidence indicates that SPW-like stimulation is effective in inducing long-term potentiation in deep layers of entorhinal cortex,⁴⁰ which are cellular layers that receive hippocampal output and project to neocortical sites. In addition, there is evidence that ripples are temporally coupled with forebrain spindle oscillations that occur regularly during episodes of SWS.⁴¹ Coordinated discharges along hippocampal and neocortical pathways could be related to processes associated with memory consolidation when short-term hippocampal memory traces are transferred to neocortical networks for long-term storage. A recent study in patients with medically refractory epilepsy found a significant correlation between ripples in hippocampus and rhinal cortex,⁴² similar to the coherence of ripples between hippocampus and parahippocampal structures in rats.⁵ In this patient study, rates of ripples in rhinal cortex measured after a brief daytime nap correlated with number of successfully recalled items learned during a prior cognitive task. While these latter data are consistent with the possible role of normal ripples in memory consolidation, not all HFOs in the epileptic brain are normal.

Spatiotemporal Properties of Pathologic High-Frequency Oscillations

In animal models of chronic limbic epilepsy, HFOs with spectral frequencies between 250 and 600 Hz, termed fast ripples, occur in dentate gyrus, subfields CA1 and CA3, subiculum and entorhinal cortex (Figures 1C, 1D). Fast ripples are found in rats that exhibit recurrent spontaneous seizures, but not in rats that have been subjected to an epileptogenic insult, e.g. status epilepticus, that do not exhibit spontaneous epileptic seizures.⁴³ These findings and the unique association between interictal fast ripples and sites of seizure onset indicate fast ripples are pathological.⁴⁴

In the intrahippocampal kainate model of mesial temporal lobe epilepsy (MTLE), in addition to fast ripples, interictal ripple-frequency HFOs can be found in the epileptogenic dentate gyrus, an area where normal ripples do not occur in normal rats (Figure 2).³⁰ Similar to fast ripples, ripple-frequency HFOs appeared within days to weeks after kainate injection, but only in rats that later exhibit spontaneous seizures. In addition, shorter latencies to the first appearance of fast ripples or ripple-frequency HFOs in dentate gyrus correlated with shorter latencies to first spontaneous seizure.³⁰ These data indicate that spectral frequency alone is not sufficient for distinguishing normal ripples from ripple-frequency pHFOs, but like fast ripples, ripple-frequency oscillations in epileptogenic dentate gyrus are also pathological. Importantly, these finding indicate that the occurrence and location of pHFOs can predict, although are not required for, the development of seizures after an epileptogenic insult and thereby provide a biomarker for epileptogenesis.

Figure 2

Multi-contact electrode array recording spontaneous pHFOs in dentate gyrus of epileptic rat. Local fast ripple and ripple-frequency HFO occur on electrode contacts numbered 2 and 6 in granule cell layer (GrL) and CA3 area respectively. There is no evidence (more...)

It has been proposed that pHFO-generating neuronal clusters represent a basic mechanism underlying limbic epilepsy.⁴⁴ Support for this hypothesis derives from studies that show the size and location of pHFO-generating sites in dentate gyrus become stable over time,⁴⁵ but application of bicuculline, an antagonist to inhibitory GABA_A-receptors, causes these areas to increase in size.⁴⁶ In addition, the number of pHFO-generating clusters in dentate gyrus correlates with seizure frequency and the spectral power of these oscillations increases during the transition to hippocampal hypersynchronous seizures.^45,47 These findings suggest that spontaneous seizures arise when reduction in tonic inhibitory influences results in the coalescence and synchronization of pHFO-generating sites.

Differences between normal and pathologic HFOs

Studies in the rat hippocampus have identified several important differences between normal ripples and pHFOs.⁴⁸ Normal ripples arise from a relatively large area of tissue that involves the controlled firing of pyramidal cells and interneurons, and the extracellular ripple largely reflects summated IPSPs from precisely timed interneuron discharges. In contrast, pHFO-generating sites can be localized to small discrete neuronal clusters often embedded within tissue that does not generate pHFOs,⁴⁶ although HFOs observed in scalp EEG recordings suggest neocortical sites could be larger than those found in epileptogenic hippocampus.⁴⁹ Figure 2 shows an example of the local generation of spontaneous pHFOs in epileptic dentate gyrus. The pHFO recorded on electrode contact 2 positioned within the granule cell layer does not appear on contact 1 located 200 μm from contact 1, and the voltage of the pHFO reverses in polarity in the depth beginning on contact 3. During pHFOs, there is no consistent firing pattern between interneurons and pHFOs, although principal cell discharges occur during the trough of the extracellular pHFO.^43,50 It is believed pHFOs reflect bursts of population spikes arising from synchronized discharges of abnormally bursting principal cells and possibly interneurons (compared Figures 1E and 1F).³¹

It is not clear how abnormal synchrony is generated during pHFOs, but recording in CA3 of normal hippocampal slices bathed in high potassium medium shows that blockade of ionotropic glutamatergic signaling reduces synchronous discharges between pyramidal cells and attenuates pHFO-like field events.⁵¹ The same study demonstrated that increasing the fidelity of pyramidal cell discharges increases pHFO amplitude, whereas disrupting intrinsic action potential generating mechanisms decreases pHFO amplitude. These data suggest that increased chemical transmission through recurrent excitatory synapses between pyramidal cells can generate pHFOs. However, another study recording in CA3 of hippocampal slices from epileptic rats offers an alternative explanation.⁵² In this study, a reduction in neuronal spike timing was associated with a decrease in ripple power and increase in fast ripple power, although the amplitude of fast ripples was much lower, not higher compared to ripples. Furthermore, increasing spike timing in epileptic CA3 restored ripples and decreased fast ripples, suggesting that in vitro fast ripples emerge from an abnormal reduction in neuronal synchrony, not an increase in synchrony. It could be argued that ripples and fast ripples from the epileptogenic CA3 in this latter study were both pathologic and each HFO reflects bursts of population spikes with differing degrees of neuronal synchrony.⁴⁸ Clearly there is a need for additional studies that will differentiate normal from pathologic HFOs in vitro and in vivo.

HFOs in Patients with Epilepsy

The first human studies on interictal HFOs were carried out using microelectrodes positioned in hippocampus and entorhinal cortex of patients with medically refractory MTLE.^2,43 Several of the properties of fast ripples recorded in patients resemble fast ripples in epileptic rats, including spectral frequency, duration, association with the seizure-onset zone,^43,53 and generation in cellular layers of entorhinal cortex that support evoked population spike discharge and abnormal synchrony of burst firing.⁵⁴ Microelectrode studies in patients also demonstrated that rates of fast ripples are highest during SWS and remain elevated during REM sleep.⁵⁵ With respect to pathological substrates and the hippocampal sclerosis often observed in patients with MTLE, higher rates of fast ripples in hippocampus correlate with severity of local atrophy,⁵⁶ and higher rates of fast ripples and lower rates of ripples correlate with lower neuron densities in Ammon’s horn and dentate gyrus.⁵⁷ These data suggest that rates of fast ripple and ripple discharge reflect the severity of epileptogenicity and thus have clinical value in determining treatment for patients with MTLE. In addition, morphological abnormalities associated with hippocampal sclerosis could promote the generation of fast ripples, although a recent study showed fast ripples are present in an nonlesional animal model of temporal lobe epilepsy.⁵⁸

It is not yet possible to distinguish normal from pathologic ripple-frequency HFOs in human mesial temporal lobe limbic structures, but presumably normal hippocampal ripples share several important characteristics with normal ripples found in the non-primate hippocampus. Human ripples have similar spectral frequencies, occur bilaterally with highest rate during SWS and lowest during REM sleep,^2,53,55 and arise from broad areas of tissue supporting generation and synchrony of neuronal discharges.⁵⁴ Single neuron studies in humans show that putative interneurons reach their highest rate of firing during the onset of the ripple, whereas pyramidal rates peaked during the maximum amplitude of the ripple event (Figures 3A–C).⁵⁹ During individual cycles of the ripple, human pyramidal cells fire during the trough of the ripple while interneurons discharge between 0.0 and 1.0 msec of pyramidal cells, which is similar to the well-timed firing between pyramidal cells and some interneurons during rat hippocampal ripples (Figure 3D).

Figure 3

Single-neuron correlates of human hippocampal ripples. A) Raster plot of putative pyramidal cell (PYR) firing (horizontal axis) during individual ripple events (vertical axis). The trace at the top shows an example of a ripple oscillation. The cumulative (more...)

Interest in pHFOs as markers of epileptogenic areas in presurgical patients has driven the development of many clinical electrophysiology systems used with standard clinical depth and subdural grid electrodes to support higher sampling rates and provide greater bandwidth. Studies using these systems confirm the strong association between pHFOs and epileptogenic regions, particularly fast ripples compared to ripple-frequency HFOs.^60–63 In addition, a patient study found pHFOs associated epileptogenic lesions that included focal cortical dysplasia and nodular heterotopia, but rates of pHFOs were more strongly linked to areas of seizure onset than anatomical lesion.⁶⁴ These latter data suggest that pHFOs are not limited to a specific type of epilepsy, but could be a fundamental property of epileptogenicity common to many types of epilepsy. Moreover, these results suggest that in some types of epilepsy, electrophysiological disturbances can be remote to anatomical abnormalities.

Interictal Spikes and HFOs

Recording of EEG interictal spikes can be useful for delineating the epileptogenic region, but the spatial extent of interictal spikes is usually more widespread than the seizure onset zone. Studies demonstrate a significant spatial and temporal association between interictal spikes and HFOs, although 40 to 50% of HFOs occur independently of interictal spikes.⁶¹ Thus, abnormal HFOs may occur on the rising component, peak or descending component of an interictal spike, or not be associated with a spike at all. ^60,65–67 One study found that the occurrence of interictal spikes containing abnormal HFOs along with HFOs alone was more sensitive in predicting the seizure-onset area than the occurrence of interictal spikes alone with no HFOs.⁶¹ Furthermore, in a separate study that used surgical outcome to verify the boundaries of the epileptogenic region, patients with good surgical outcome had a significantly larger proportion of abnormal HFO-generating sites removed than patients with poor surgical outcome.⁶⁸ In contrast, there was no difference in the number of interictal spike (with no HFOs) or seizure onset sites removed between good and poor surgical outcome groups. Results from these latter studies need to be confirmed with studies using a greater number of patients. Nevertheless, results suggest that pHFOs not only delineate the epileptogenic region better than interictal spikes with no pHFOs, but also better than EEG identification of the site of seizure onset.

Seizure Generation and pHFOs

In animals and patients with epilepsy, the power of HFOs can increase several seconds to tens of minutes before seizure onset. ^30,69–76 Focal cortical seizures that secondarily generalize are characterized by a relatively small area of pHFO generation at seizure onset that can move along the cortex and increase in size as the seizure progresses.⁷⁷ It is hypothesized that an increase in size and synchrony among pHFO-generating sites can trigger seizures, ^30,44 but not all seizures are associated with an increase in HFO power and some seizures can be accompanied by a reduction in HFO power during onset.⁷⁸ It is possible that a pre-ictal increase in HFO power that reflects bursts of population spikes could indicate enhanced synchronization of principal cell networks that reach threshold supporting spread of ictal discharges. In contrast, a pre-ictal increase in HFOs that reflect IPSPs could be an indication of interneuronal network activation to suppress propagation of epileptiform discharges, whereas a reduction in these latter type HFOs might fail to prevent spread of ictal discharges. A more thorough understanding of the neuronal mechanisms during the ictal transition will help clarify the role of pHFOs during seizure genesis.