An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Noebels JL, Avoli M, Rogawski MA, et al., editors. Jasper's Basic Mechanisms of the Epilepsies. 5th edition. New York: Oxford University Press; 2024. doi: 10.1093/med/9780197549469.003.0010
The subiculum has for a long time been neglected and solely considered as the output region of the hippocampus, conducting or relaying inputs to the entorhinal cortex. However, significant evidence does not support such a passive role. The subiculum is a three-layer cortex structured associated with highly organized substructures. Preeminent bursting neuronal behavior and local microcircuits favor the production of strong neuronal activities. All hippocampal rhythms such as theta, gamma, sharp-wave ripples, and ripples can be generated by the subiculum. As a result, the subiculum is an important limbic structure involved in spatial location and memory. The role of the subiculum in epilepsy was highlighted by studies on human postoperative tissues, showing that it is a major and autonomous site of genesis of epileptic interictal and ictal activities. In these studies, a leading role of interneurons in the buildup of epileptic activities was suggested, and chloride dysregulation leading to depolarizing and potentially excitatory effects of GABAA signaling was unraveled. Evidence indicates that the subiculum not only anatomically “supports” hippocampal outputs, but that it also actively participates to the generation of epileptic activities in the temporal lobe.
The term subiculum comes from the Greek for “support,” suggesting, anatomically speaking, that this structure supports the hippocampal formation. The subiculum receives projections from the hippocampal CA1 subfield and projects to the entorhinal cortex (EC), acting as a relay between the hippocampal formation and the parahippocampal cortices. The subiculum has therefore been considered as a part of the circuitry of episodic memory and spatial orientation, with no dedicated function.
However, some properties of the subiculum suggest that it was more than a bridge in mesial temporal lobe structures. First, the anatomical structure of the subiculum is complex, comprising specific nuclei such as the prosubiculum, the presubiculum, the parasubiculum, and the postsubiculum (Ding, 2013; Witter & Groenewegen, 1990), which are characterized by an evolutionary intermediate three-layer cortex. Second, the rate of bursting pyramidal cells is higher than in most cortices, suggesting that specific activities may be generated by this area rather than only transmitting them. Further, the local neuronal microcircuits do favor the production of bursting rhythms.
In an unusual manner, epilepsy allowed the subiculum to reach nobility. In pioneer work on human postoperative temporal lobe tissues, the subiculum was shown to generate epileptic activities. New mechanisms were identified such as the triggering role of interneurons and depolarizing effects of GABA signaling, which contributed to reshape our knowledge on epileptic local networks. This review attempts to show how the subiculum has long been neglected, and only considered as a bridge in the temporal lobe, until it came out as a key structure for spatial orientation, memory, and epilepsy.
The subiculum is located in the mesial temporal lobe of the mammalian brain where it is embedded in the parahippocampal gyrus. The subiculum is anatomically (and functionally) positioned as a principal output structure of the hippocampal formation (Eichenbaum & Cohen, 2001; O’Keefe & Nadel, 1978), connecting it with the EC and other subcortical areas. The subiculum is therefore considered as a bridge, or a relay station, between the hippocampus and parahippocampal cortices, and more widely with several limbic structures of the brain.
The EC supplies a major synaptic input to the hippocampus proper through the dentate gyrus (DG), the first synaptic connection in its well-known trisynaptic circuitry (Aggleton & Christiansen, 2015) (Fig. 10–1). Major output from the hippocampus proper goes back to the EC, through the subiculum, adding a fourth synapse to this circuitry (Aggleton & Christiansen, 2015; Hoesen et al., 1979; Hoesen, 1982; O’Mara et al., 2001 & 2006; Rosene & Hoesen, 1977; Witter & Groenewegen, 1990). Apart from receiving main inputs from CA1 area of the hippocampus proper (60% of all subicular inputs in primates and 40% in rodents; Roy et al., 2017; Saunders & Aggleton, 2007), subicular neurons are also directly innervated by axons arising from layer II/III of the EC (Fig. 10–1). Interestingly, inputs from different portions of EC are segregated; axons from the lateral EC project to the proximal part of the subiculum, while the medial EC sends axons to innervate the distal part of the subiculum (reviewed in Witter et al., 2000). Moreover, there is a dorsoventral organization of the inputs from the medial and lateral EC, the dorsal subiculum being preferentially innervated by its dorsal and lateral areas, while axons of the medial parts of these cortices terminate in the ventral subiculum (Witter et al., 1989). Therefore, the subiculum acts as a transition zone between the hippocampus proper and the EC, playing an important role in the modulation of hippocampal-cortical information flow. The importance of the hippocampal formation neuronal circuitry is highlighted by its highly conservative organization among different species of mammals, including rodents, primates, and humans (Ding, 2013). In this circuitry, the subiculum represented the least investigated area; nevertheless, it has been implicated not only in important physiological functions, such as spatial relation and working memory, but also in pathophysiological process such as Alzheimer disease and most notably epilepsy.
Schematic representation of neuronal connections within human hippocampal formation and entorhinal cortex.
In addition to the EC, the subiculum also receives inputs from other regions such as the parasubiculum, the presubiculum, the perirhinal cortex, and the amygdala (Fig. 10–1) (Kloosterman et al., 2003; Kosel et al., 1982; Witter et al., 1989). These afferent projections also show a topographic organization. In particular, the presubiculum projects to the dorsal part of the subiculum, while the amygdala and the perirhinal cortex project to the proximal portion of the subiculum (Ding et al., 2020). Subicular outputs terminating in temporal cortices have a topographic organization as well. Similar to inputs from the EC, the distal subiculum sends efferent projections to the medial EC and the proximal subiculum innervates its lateral portion (Witter, 2006). In addition, the distal subiculum projects to the presubiculum and parasubiculum, while the proximal subiculum sends efferent fibers to the amygdala, the perirhinal, and the prefrontal cortices. At the dorsoventral axis, the presubiculum receives inputs from the entire subiculum, while the amygdala is preferentially targeted by the ventral subiculum and the perirhinal cortex is innervated by the dorsal subiculum (Canteras & Swanson, 1992; Naber & Witter, 1998; Namura et al., 1994). Such distinct dorsoventral organization leads to functional differences: the ventral subiculum participates in the modulation of stress response, while the dorsal subiculum is involved in memory formation.
As already stated, the subiculum is actually a part of the subicular complex composed of the prosubiculum, the presubiculum, the parasubiculum, and the postsubiculum, although this division varies among species (Ding, 2013; van Groen & Wyss, 1990). The entire subicular complex is positioned distally in relation to CA1 and more proximal than the retrosplenial cortex, and its regions seem to have not only distinct anatomy but also different functions. In particular, while the subiculum itself receives main afferents from CA1 and EC, principal input to the presubiculum and the parasubiculum comes from the neocortex (Burwell et al., 1995; Lopes da Silva et al., 1990). Similar to the hippocampus proper, the subiculum has only three layers (molecular, pyramidal, and polymorphic fiber layers), characteristic of allocortex structures, while other parts of the subicular complex have a six-layered structure, which is a typical feature of neocortex, and are thus more similar to the EC (Aggleton & Christiansen, 2015). Several recent studies confirmed the heterogeneity of the subicular complex both on anatomical and molecular levels of organization, using viral labeling and single-cell RNA sequencing (Cembrowski, Phillips, et al., 2018; Cembrowski, Wang, et al., 2018; Ding et al., 2020). As the heterogeneity of subicular complex becomes more apparent, it is important to note, that in the past, most studies investigating role of the subiculum in epilepsy were conducted on the dorsal part of this structure.
Beginning of the subiculum can be visually identified as widening of the CA1 pyramidal cell layer, with relatively loose neuronal packing compared to the adjacent CA1. At the same time, stratum oriens disappears while the subicular molecular layer is a continuation of the stratum radiatum and of the stratum lacunosum-moleculare of CA1. Thus, apical dendrites of subicular pyramidal neurons occupy the molecular layer, while their basal dendrites are located within the pyramidal layer itself (Harris, Witter, et al., 2001). Similar to the connections with temporal cortices, reciprocal connections of the subiculum with CA1 follow strict topographic organization. The proximal subiculum, which is closest to the CA1 area, receives input from distal portions of CA1, while distal areas located further from CA1 are innervated by proximal CA1 neurons, and finally the middle portion of the subiculum is targeted by axons arising from the middle of CA1. Therefore, CA1 projections create a series of so-called nested loops (Amaral et al., 1991).
Importantly, there is also a spatial segregation of synaptic inputs within pyramidal and molecular layers of the subiculum. Axonal inputs from hippocampal CA1 terminate in the pyramidal layer of the subiculum and proximal portions of its molecular layer, thus primarily innervating basal dendrites of pyramidal neurons, their soma, and apical dendrites (Amaral et al., 1991). In contrast, projections from the EC form synaptic contacts on the distal dendrites of subiculum pyramidal cells in the distal part of the molecular layer (Baks-te Bulte et al., 2005; Witter & Groenewegen, 1990). This particular synaptic organization may play a key role in the regulation of epileptic discharges by the subiculum.
Neuronal circuits within the subiculum are composed of three main neuronal types: regular and bursting pyramidal cells, as well as interneurons (Harris, Hirase, et al., 2001; Menendez de la Prida, 2003). Several types of interneurons have been described in the subiculum. The most prevalent are parvalbumin (PV) expressing fast-spiking cells, but interneurons expressing somatostatin, calretinin, and neuropeptide Y are also present (Ding, 2013). These GABAergic interneurons provide feedback and feedforward inhibitory inputs onto subicular pyramidal cells to modulate their firing (Finch et al., 1988; Menendez de la Prida, 2003; Panuccio et al., 2012). Relevant asymmetry of local connections between two types of pyramidal cells and interneurons has been described (Böhm et al., 2015). In particular, while synaptic connection probability between pyramidal cells is relatively low (only 3.7% within population of bursting neurons and 4.7% within regularly firing neurons), it is almost twice higher for regular to bursting cell inputs (7.3%). Moreover, signaling of regular spiking cells on bursting neurons is unidirectional. In addition, inhibitory interneurons including fast-spiking cells preferentially form synaptic connections with regular firing neurons. This topographic organization leads to the spread of the excitation from regular firing pyramidal neurons to burst-firing neurons, while, at the same time, strong feedback inhibition will rapidly diminish activity of regular spiking neurons (Böhm et al., 2015). However, reduced inhibitory control on bursting cells results in an overall larger and longer bursting excitation.
Principal pyramidal cells are considered to represent a relatively homogenous population. However, in the subiculum, pyramidal cells have been known for almost 30 years to fall into two different categories based on their firing pattern. As mentioned above, regular firing and bursting pyramidal cells coexist, identified by in vitro intracellular recordings in response to depolarizing current steps (Mason, 1993; Mattia et al., 1993; Stewart & Wong, 1993; Taube, 1993), as well as by in vivo registration in freely moving animals (O’Mara et al., 1995). Bursting neurons compose the majority of pyramidal cells, and in some regions of subiculum they are twice more prevalent than regular spiking cells, which is a unique condition. In addition, the ratio between bursting and regular firing neurons changes along proximal-distal axis, with a number of bursting neurons increasing in the distal part of the subiculum (Jarsky et al., 2008; Kim & Spruston, 2012), as well as the dorsoventral axis, with a higher number of bursting neurons in deeper ventral layers (Greene & Totterdell, 1997). Such organization, combined with the change in the subicular projection targets along the same axis, suggests that bursting and regular firing neurons send information to different downstream brain regions (Kim & Spruston, 2012). However, a recent study denied such location specific for bursting neurons in the subiculum (Simonnet & Brecht, 2019).
In response to depolarization, bursting neurons discharge with a short burst of 2–5 action potentials at up to 200 Hz frequency, and regular firing cells generate a train of action potentials with much lower frequency (Lisman, 1997). Burst generation could provide the mechanism for more reliable information transfer as it increases the probability of synaptic release (Lisman, 1997; Staff et al., 2000). Apart from obvious differences in firing patterns, these two types of pyramidal cells also have differences considering morphological and electrophysiological properties. For instance, the length of apical dendrites was found to be higher in bursting neurons, while basal dendrites were shorter (Fiske et al., 2020). Furthermore, bursting neurons have more negative membrane potential, lower input resistance, and larger sag currents in response to hyperpolarization (Graves et al., 2012; Mattia et al., 1997a). Regular firing and bursting neurons also exhibit distinct pharmacological responses and express synaptic plasticity with different underlying mechanisms (Greene & Mason, 1996; Kintscher et al., 2012; Pandey & Sikdar, 2014; Wozny et al., 2008). Regular firing neurons can switch to a bursting behavior (Mattia et al., 1997b). Similar to hippocampal pyramidal neurons, they also show place field firing, although with much larger areas and lower resolution, compared to CA1 cells (O’Mara et al., 2000; Sharp & Green, 1994).
The subiculum constitutes the main hippocampal output. But it does not work just as a door allowing information flow toward cortical and subcortical areas. In the subiculum, an intrinsic processing and coding of the information take place, as a final step before dispatching the output data. Generally speaking, the dorsal subiculum appears principally involved in the processing of information about space, movement, and memory, whereas the ventral subiculum modifies the response to stress by regulating the hypothalamic–pituitary–adrenal (HPA) axis (Kitanishi et al., 2021a; O’Mara, 2005).
Hippocampus rhythms have been widely investigated. By contrast, the role of the subiculum is underexplored. Different groups of investigators have observed that the subiculum can also generate, as the hippocampus, oscillatory rhythms such as theta and gamma. Theta and gamma oscillations are produced during activity periods and are thought to store recent memories into the hippocampal circuitry (Buzsáki, 1996). On the other hand, during “off-line” periods, sharp wave-ripples complexes arise. They possibly transfer memory information toward cortical areas for long-term storage (Ego-Stengel & Wilson, 2011).
In the subiculum, locally produced theta rhythms are controlled by PV interneurons (Amilhon et al., 2015). They can follow two different modalities. They can first get out from the hippocampus, through the subiculum toward cortical areas, under the control of excitatory drive, or travel backward to modulate CA1 and CA3 rhythms and spike timing, under GABAergic control, especially during REM sleep (Jackson et al., 2014). Theta oscillations may represent a timing mechanism to organize movement sequences temporally, memory encoding, or planned trajectories for spatial navigation (Nuñez & Buño, 2021).
An intrinsic gamma activity can be recorded in an in vitro rat subiculum preparation, comprising both slow gamma (25–50 Hz) and fast gamma (100–150 Hz) oscillations. Gamma rhythms are phase-modulated by theta rhythms, generated during their rising phase and their activity peak. Not only the mean amplitude of slow and fast gamma but also the strength of phase amplitude coupling was unchanged, compared to preparations with the whole hippocampus (Jackson et al., 2011). Moreover, fast GABAergic inhibition is required for the generation of fast gamma, whereas slow gamma is generated by excitatory and inhibitory mechanisms.
Sharp wave-ripple complexes, recorded in the hippocampus, also play a pivotal role in memory consolidation. Sharp wave-ripple suppression provokes memory deficits, whereas the intensification of spontaneous ripples by optogenetic stimulation increases memory performances during maze task in rats (Fernández-Ruiz et al., 2019). Even if the principal generator of sharp wave-ripples is CA3, experimental evidence shows that the subiculum can also produce sharp wave-ripples. Using multineuronal calcium imaging, Norimoto and collaborators found that immediately before hippocampal sharp wave-ripples a group of subicular neurons was activated (Norimoto et al., 2013). The results highlight the active role of the subiculum in modifying neural information related to sharp wave-ripples. These results were recently confirmed by showing that subicular-generated sharp wave-ripples spread to the EC but also backward to the CA1 and CA3 regions (Imbrosci et al., 2021).
In pathological conditions, such as temporal lobe epilepsy (TLE), both physiological and pathological ripples can be recorded in the subiculum. In vitro, in human epileptic subiculum, ripple oscillations appear in local field potential (LFP) recordings during interictal discharges (IIDs) but also in preictal discharges (PIDs) that precede inducted ictal-like events (Alvarado-Rojas et al., 2015).
Memory processing is a complex task that involves the hippocampal formation and neocortical areas (Moscovitch et al. 2016). Most likely, new memories are initially stored in the hippocampus during activity periods, in the form of theta and gamma rhythms. Once the animals get into a resting or sleep phase, the memories travel in the form of sharp wave-ripples to the neocortex.
Little attention has been paid to the subiculum’s role in memory processing. But experimental evidence has already shown that the subiculum can be a generator of theta, gamma, and sharp wave-ripple activities. Moreover, it can communicate “backward” with CA1/3 regions, modulating hippocampal circuitry function (Jackson et al., 2011). In addition, subiculum’s connections make it an information hub in memory processing. For instance, the projection into the nucleus accumbens is probably involved in instrumental learning (Phillips et al., 1988), and entorhinal connection takes part in spatial learning (Taube et al., 1992).
Pathology also points toward a key role of subiculum in memory consolidation. Lesions in the hippocampus can provoke amnesia and spatial orientation deficits (Benson et al., 1974; Ott & Saver, 1993; Taube et al., 1992). Besides, functional magnetic resonance imaging (MRI) studies have found an increase in the function of the subiculum during memory recall activities (Eldridge et al., 2005). In Alzheimer patients, structural MRI studies have found that atrophy in the subicular complex is detected as an early sign (Carlesimo et al., 2015). Further, senile plaques are found in greater concentration in the subicular complex compared to the hippocampus (Davies et al., 1988).
Like the hippocampus, the subiculum also includes “place cells” (Barnes et al., 1990; Brotons-Mas et al., 2010; O’Keefe, 1976). Place cells fires when the animal is placed in a determined region and remains silent when placed in another location. Every animal position presents discharges in a different group of place cells, generating a cognitive map of the environment (Sharp, 2006). In the postsubicular region, Rank and colleagues founded head-direction cells, harboring a firing when the rat orientates in a particular direction independently from animal’s location, working as a compass (Taube et al., 1990).
Subicular neuronal activity is detected during object exploration tasks, and subicular cell firing appears to correlate with the concurrent location and speed of the rats within the task environment (O’Mara, 2005). It has been suggested that the dorsal subiculum is a site of integration between hippocampal spatial information and whole-body, movement-related information. Besides, place cell spiking is organized in time-compressed sequences inside the ongoing theta rhythm (7–14 Hz) (O’Keefe & Recce, 1993). Place fields of subicular place cells are supposed to be larger than in CA1. Recently, the key role of dorsal subiculum in transferring navigation-associated information has been studied using optogenetics and challenged such reports. Subicular neurons demonstrate a representation of place, speed, and trajectory, which was as accurate as or even more accurate than that of hippocampal CA1 neurons. Theta oscillations and sharp wave-ripples tightly controlled the firing of projection neurons depending on the output area (Kitanishi et al., 2021b).
In addition to widely studied memory and spatial orientation, the subiculum may also regulate the stress response via a modulation of the HPA axis (Roessler et al., 2019). Inhibitory inputs are projected into the hypothalamus via the postcommissural fornix, the medial cortico-hypothalamic tract, and the amygdala. The inhibitory effect is thought to be mediated by inhibitory GABAergic neurons. It plays a key role in ending or limiting the response of the HPA axis to stress (Lowry, 2002). Fine regulation of stress response is a crucial factor in memory processing as prolonged stress inhibits long-term potentiation, leading to hippocampal atrophy and impairing learning (Sapolsky, 2003).
Research on TLE is facilitated by the availability of animal models in vivo, mostly relying on the injection of convulsant drugs such as pilocarpine and kainic acid. Systemic administration of both chemicals in rodents or intrahippocampal injection of kainate provokes a status epilepticus, which is followed by seizure-free latent phase and subsequent spontaneous seizures reoccurring for the rest of the animal’s life (Ben-Ari et al., 1979; Cavalheiro et al., 1982; Curia et al., 2008; Nadler, 1979; Turski et al., 1983). These models allow for reproducing both recurrent focal seizures characteristic of human patients with medial TLE and histological changes observed in their hippocampus, including sclerosis (Curia et al., 2008; Lévesque & Avoli, 2013). It was recently reported that the kainate model was not only artificially generated in experimental labs. In California, sea lions are poisoned by eating red algae producing the kainate agonist domoic acid, resulting in status epilepticus and subsequent TLE with hippocampal sclerosis, close to the human pattern (Buckmaster et al., 2014).
Data obtained from animal models of epilepsy provide evidences for an activation of the subiculum during seizures and for its important role in the generation of epileptiform activities. Several studies reported that spontaneous seizures could be observed in the subiculum weeks after pilocarpine administration (Behr et al., 2017; Lévesque et al., 2012; Toyoda et al., 2013). In particular, Levesque et al. (2012) performed chronic LFP recordings in CA3, EC, subiculum, and DG; within 2 weeks after pilocarpine treatment, they observed that hypersynchronous-onset seizures were mostly initiated in CA3 and subiculum (Fig. 10–2A). Moreover, high-frequency oscillations in the fast ripple band (250–500 Hz), which are characteristic of epileptic neuronal network activity in the seizure-onset zone (Jefferys et al., 2012), were also observed in the subiculum after status epilepticus induced by pilocarpine, comprising ripples or fast ripples, depending if they occurred during the first 2 weeks after SE or around 5 weeks after, respectively (Behr et al., 2017). Importantly, multielectrode recordings in various rat brain regions revealed that earliest seizure activities most frequently appear in the ventral subiculum and hippocampus up to 26 weeks after pilocarpine administration (Toyoda et al., 2013).
Role of the subiculum in rodent models of epilepsy. A. Example of hypersynchronous-onset seizure recorded from the subiculum using implanted LFP electrodes in rat model of TLE. The initial phase of the seizure on an expanded time scale is shown in 1; (more...)
Interestingly, a complex interaction between the phase of slow neuronal oscillations and the amplitude of faster rhythms could be studied using phase-amplitude coupling (PAC) (Tort et al., 2010). PAC has been proposed as a biomarker of epileptogenesis (Samiee et al., 2018), as it reflects the interactions between different neuronal subpopulations likely affected by epileptic disorders. Indeed, strength of PAC between slow delta frequency (0.18–4 Hz) and fast beta to ripple oscillations (20–250 Hz) are larger in the CA3 and subiculum of pilocarpine treated rats compared to control animals. Moreover, there is a positive correlation between PAC strength and high seizure rates 2 weeks after pilocarpine treatment, suggesting neuronal network remodeling, accompanied by changes in the neuronal excitability and their synaptic connections (Samiee et al., 2018). Thus, epileptic activities in the subiculum could be reproduced in animal models of epilepsy, which allows for studying the mechanism of their development.
Activity changes during the preictal period could be an important indicator for the upcoming seizures and could help to understand how those seizures are initiated. Simultaneous LFP and single-cell recordings in the hippocampal formation of epileptic rats during spontaneous seizures showed higher average firing rate of subicular, CA1, and DG neurons 2–4 minutes before seizures, and increased preictal activity in these regions (Fujita et al., 2014). This could be related to more prominent preictal theta activity, as average firing rates of neurons during theta activity were higher for both bursting and regular spiking subicular neurons (Fujita et al., 2014). Interestingly, it has been shown in acute in vivo experiments that in the subiculum systemic injections of the K+ channel blocker 4-aminopyridine (4-AP) trigger predominantly low-voltage, fast-activity seizures associated with ripples, while GABAA receptor antagonist picrotoxin evokes hypersynchronous seizure-onset patterns with higher rate of fast ripples (Fig. 10–2B) (Salami et al., 2015). These data suggest that GABAA receptor-mediated signaling could be involved in the low-voltage fast seizures, whereas hypersynchronous seizures would be dependent on excitatory activity. Thus, initiation and propagation of different seizure types are likely dependent on the activity changes in distinct neuronal networks.
The role of interneurons in seizure initiation has been investigated using LFP and unit recordings in the hippocampus of epileptic rats (Toyoda et al., 2015). According to this study, most hippocampal interneurons increased their firing rates before seizures. This preictal activation was beginning the earliest in the subiculum and was most abundant in the subicular interneurons. In addition, it also correlated with theta activity occurring before seizure onset. Interestingly, just before seizure onset, many interneurons display short pause in their activity, while during seizures, frequency of intraneuronal firing rates drops down. Increased preictal interneuron activity might indicate attempt of neuronal circuitry to balance excessive activation of principal excitatory cells, which ultimately fails and leads to the seizure onset, suggesting critical involvement of inhibitory neurons (Toyoda et al., 2015).
Optogenetic studies in the mouse kindling model of TLE showed that photoactivation of GABAergic neurons in the subiculum delays development of secondary seizure through inhibition of pyramidal cells. However, expression of secondary seizures was exacerbated by activation of GABAergic cells due to the changes in chloride homeostasis resulting in depolarizing GABA signaling (Wang et al., 2017). Different types of interneurons had distinct contribution to the seizure development. Thus, activation of both fast-spiking PV-expressing interneurons and somatostatin-positive interneurons delayed development of generalized seizures. Interestingly, there was an opposite effect on the expression of generalized seizures, since activation of PV neurons aggravated seizures, while activation of somatostatin neurons reduced the severity of seizures (Wang et al., 2017). Additionally, the important role of PV-expressing interneurons has been confirmed using inhibition of GABA release from these cells in the ventral subiculum using viral vector injection. Permanent silencing of PV interneurons resulted in the development of clusters of spike-wave discharges and spontaneous recurrent seizures in mice, while transient silencing did not induce any acute or chronic seizure (Drexel et al., 2017).
In vitro studies using mouse and rat brain slices indicate substantial neuronal network remodeling within subiculum in various epileptic models. First of all, even though overall neuronal loss in the subiculum is substantially smaller compared to CA1/CA3 hippocampal areas, it was shown to be up to 30% in pilocarpine-treated rats and 15% in mice (He et al., 2010; Knopp et al., 2005). Interestingly, there is a substantial difference in the neuronal loss between distal and proximal parts of the subiculum, the later showing up to 50% reduction in neurons (Drexel et al., 2012). Despite these differences regarding the proportion of neuronal loss within epileptic subiculum, evidence from different laboratories indicates overall increase in the number of bursting pyramidal neurons weeks after epilepsy induction (Knopp et al., 2005; Wellmer et al., 2002). Whether regular spiking cells are transformed into bursting or simply less susceptible to the cell loss, therefore being more abundant in epileptic tissues, is unclear. However, at least one study found that pharmacological blockade of GABAAR did not result in the switch from regular firing to bursting mode (Panuccio et al., 2012).
While pyramidal neurons largely stay intact, significant loss of GABAergic interneurons has been reported in the pilocarpine model of TLE in rats (Knopp et al., 2008). The density of GAD-positive interneurons revealed using immunostaining was significantly reduced in all subicular layers. In addition, the density of GAD65 expressing synaptic terminals in the pyramidal cells layer was reduced, while both frequency of mIPSCs and evoked IPSCs was decreased. All together these data indicate weaker perisomatic GABAergic inhibition of principal cells in epileptic subiculum, which might contribute to the disruption of excitatory-inhibitory balance within subicular neuronal networks and promote epileptic activity.
Finally, synaptic reorganization including axonal sprouting has been well documented in the epileptic hippocampus, but mostly in the DG and CA3 due to the advantage of using zinc staining in visualization of zinc-enriched mossy fiber synaptic terminals (Pitkänen et al., 2002). Sprouting of CA1 neuronal axons within the subiculum has been studied in five different models of TLE using Timm staining combined with retrograde labeling (Cavazos et al., 2004). In the normal brain, projections from CA1 neurons are topographically organized in the subiculum in a lamellar fashion, but in epileptic rats retrograde labeling was more extensive, spreading to the adjacent lamellas. CA1 axonal sprouting in the subiculum could provide additional amplification and synchronization of epileptic activity generated in the hippocampus. Nevertheless, these models of epilepsy (Cavazos et al., 2004) show less pronounced hippocampal neuronal loss, in contrast to the sclerotic hippocampus in epileptic patients. Thus, it still has to be investigated how axonal sprouting of remaining CA1 neurons affects overall connectivity between CA1 and subiculum and its relative synaptic strength.
The role of the subiculum in the generation and propagation of epileptic activity has been extensively studied using in vitro models (for review, see Lévesque & Avoli, 2021). Early experiments suggested that the subiculum is important for the synchronization of epileptiform activity between hippocampus and EC as well as that inhibition within the subiculum controls epileptiform activity propagation from the hippocampus to adjacent areas (Avoli et al., 1996; Barbarosie & Avoli, 1997; Benini & Avoli, 2005; de Guzman et al., 2006).
Experiments in rat brain slices incubated with 4AP indicate that interactions between the subiculum and the EC play an important role in the generation and synchronization of epileptic activity (Fig. 10–2C). In particular, subiculum has a significantly higher number of low-voltage, fast-onset ictal discharges, while sudden hypersynchronous onset events are more frequently initiated in the EC. Destruction of the connection between the subiculum and the EC desynchronized not only ictal but also interictal discharges, resulting in the overall decrease of epileptiform activity (Fig. 10–2C2; Herrington et al., 2015).
Recently, Fiske et al. (2020) investigated local excitatory neuronal networks within the subiculum and their role in the generation of epileptic activity. Paired patch-clamp recording in the mouse subicular slices revealed functional connectivity between similar and different types of principal neurons (regular and bursting pyramidal cells). Importantly, isolated and pharmacologically inhibited subicular slices were able to generate hypersynchronous neuronal firing, thus suggesting that local excitatory circuits in the subiculum are sufficient to trigger epileptic activity (Fiske et al., 2020).
Emerging evidence confirms the role of the subiculum in the generation of epileptic activity. What are the possible molecular mechanisms involved? Hyperexcitability of subicular pyramidal neurons is associated with changes in sodium channel currents, as was shown by Barker et al. (2017) in the “continuous hippocampal stimulation” rat model of TLE. In particular, epileptic subicular neurons show significant increase in the resurgent and persistent sodium currents, facilitating action potential generation. Nav1.6 channels seem to be responsible for this effect. Another mechanism regulating epileptiform activity in the subiculum is through the inhibition of T-type calcium channel by serotonin (Petersen et al., 2017). Serotonin acts on the 5-HT2C receptors and decreases bursting of subicular pyramidal neurons, thus controlling in vitro epileptiform activity.
The subiculum plays a key role in focal TLE, although this has been recognized relatively recently. Such late identification might be related to the fact that the subiculum is relatively spared by the cell loss process of hippocampal sclerosis as reported in early (Bratz, 1899) and more recent studies (Fig. 10–3A; Thom et al., 2005). The prosubiculum, however, may be affected (Andrade-Valença et al., 2008), and a limited cell loss, peaking at ~10%, occurs in cases of hippocampal sclerosis (Furtinger et al., 2001). In an exceptional case report, an encephalitis-related status epilepticus, a selective subicular and amygdala degeneration, was found (Yamashita & Yamamoto, 1999). To note as, in an unprecedented way, light has been shed on the subiculum after ex vivo studies of human postoperative TLE tissues revealed that both interictal and ictal epileptic discharges can originate from this area (Fig. 10–3B and C).
Role of the subiculum in human epilepsy. A. Schematic drawing of hippocampal reorganization during epilepsy. B1. Synchronous rhythmic activity recorded in the human temporal lobe slices using tungsten extracellular electrodes. Interictal activity recorded (more...)
While several studies reported that spontaneous epileptiform activities could be recorded in human temporal lobe postoperative epileptic tissues, either in the hippocampus or from the adjacent neocortices (Kohling et al., 1998, 1999; Schwartzkroin & Haglund, 1986; Schwartzkroin & Knowles, 1984), a key electrophysiological study of postoperative slices from hippocampal sclerosis patients, comprising the hippocampus connected to the subiculum and the EC, showed that spontaneous rhythmic population events were generated by the subiculum (Cohen et al., 2002). These activities were similar to those recorded in situ by intracranial stereo-EEG electrodes and were therefore considered to represent interictal-like discharges (Fig. 10–3B and C). They synchronized a limited portion of the subiculum and were still recorded in isolated subiculum, indicating that the subiculum was sufficient to produce such epileptic activities on its own. This study was a pioneer in the identification of excitatory signaling of GABA in mature tissues. Interictal-like activities were controlled by AMPA glutamatergic receptors antagonists but also by GABAA receptors blockade, indicating an active involvement of GABAergic inputs (Fig. 10–3B2 and B3). Sharp electrode recordings allowed identification of their cellular mechanism: interneurons were firing before and at the onset of the event and were considered to generate it. In addition, most pyramidal cells were hyperpolarized during these interictal-like activities, but up to 20% of them were rather depolarized (Fig. 10–3C3). A similar proportion of pyramidal neurons had depolarized values of the GABAA currents reversal potential, with a driving force that, at rest, resulted in depolarizations. This study was replicated in patients with and without hippocampal sclerosis (Wozny et al., 2003, 2005). Subicular pyramidal cells activated during interictal events were found to be more excitable, with reduced after-hyperpolarization, while another study did not observe differences in anatomy or intrinsic properties (resting potential, input resistance, firing pattern, existence of an Ih current) in cells that were depolarized versus hyperpolarized during these events (Huberfeld et al., 2007).
Such depolarizing effects of GABA are presumably related to Cl− dysregulations. The Cl− extruder KCC2 is down-regulated in these epileptic tissues (Huberfeld et al., 2007; Munoz et al., 2004; Palma et al., 2005, 2006). However, such impairment in Cl− outflow process is not the only mechanism at play since, at a single-cell level, while in neurons hyperpolarized by GABA, a membranous expression of KCC2 was systematically identified, in a minority of cells depolarized by GABA, an expression of KCC2 was detected, suggesting that even with functional Cl− extrusion through KCC2, GABA could be depolarizing. The role of KCC2 was later stressed by computational studies, which showed that reduction of KCC2 effects results in the emergence of population events similar to IIDs (Buchin et al., 2016). KCC2’s counterpart, the Cl− loader NKCC1, normally repressed in mature neurons, was suggested to be re-expressed in the epileptic subiculum (Huberfeld et al., 2007; Munoz et al., 2004; Palma et al., 2005, 2006; Sen et al., 2007), although using unspecific antibodies. NKCC1 was definitively proven to be involved since its pharmacological blockade by the antagonist bumetanide suppresses IIDs and restores GABAA reversal potential in the human tissue (Fig. 10–3B4 and B5; Huberfeld et al., 2007; Palma et al., 2006).
Specific proepileptic neuronal properties have also been shown in the human subiculum. A persistent (or slowly inactivated) Na+ current, activated at shifted hyperpolarized values, has been identified in half of the neurons of the subiculum, and its density was shown to be increased in tissues from pharmacoresistant patients (Vreugdenhil et al., 2004). Such current may increase excitability by favoring bursting behaviors.
More recently, the human epileptic subiculum has been maintained in organotypic cultures, together with the hippocampus (Eugène et al., 2014; Le Duigou et al., 2018). While epileptic activities are still recorded in the subiculum, they also appear to be generated in CA3 and in the DG after a few days of cultures and rapidly predominate. Possible neurogenesis was identified in the subiculum (and in CA3) by Nestin labeling.
The human subiculum has also been shown to generate ictal-like discharges in ex vivo preparations. Such events are induced by modifications of the perfusion media in order to increase neuronal excitability (Fig. 10–3C). Seizures arise after a transition period lasting tens of minutes, characterized by the progressive buildup of new transient population events known as PIDs (Huberfeld et al., 2011). These PIDs are sustained by neuronal networks that are different from those involved in generating interictal events: (1) they are triggered by pyramidal cells instead of interneurons; (2) they synchronize most neurons of the area and propagate faster and at larger distances; and (3) they are blocked only by AMPA receptors antagonists, although they require functional NMDA signaling to buildup (Fig. 10–3C3 and C4). Once fully established, recurring PIDs precede seizure onset, as in vivo (Bartolomei et al., 2004; Huberfeld et al., 2011; Lagarde et al., 2019). The seizure itself may start by nonsynchronized, low-voltage fast activities or by a hypersynchronous pattern. As for IIDs, the subiculum is sufficient to generate such seizure-like events. Similar ictal-like events generated by the subiculum were observed by other groups (Gabriel et al., 2004; Jandova et al., 2006; Klaft et al., 2016; Kovács et al., 2011; Remy et al., 2003).
High-frequency oscillations of the ripple band (150–250 Hz) have been recorded in the subiculum ex vivo, both during interictal and PIDs (Alvarado-Rojas et al., 2015). Cellular recordings allowed assessment of their distinct neuronal basis. Interictal ripples are sustained by mixed rhythmic GABAergic and glutamatergic synaptic inputs but with limited action potentials contribution. In contrast, preictal events are associated with ripples composed of depolarizing synaptic inputs that frequently trigger a bursting discharge in pyramidal cells.
From a histological perspective, an up-regulation of Neuropeptide Y-Y2 receptors that were identified in the hippocampus extending to the subiculum may play a regulatory role (Furtinger et al., 2001). Activated microglia is detected in the subiculum (Milior et al., 2020; Morin-Brureau et al., 2018), with a specific phenotype characterized by multiple ramifications, like surveilling microglia, more pronounced in the subiculum than in hippocampal fields. Glutamine synthetase—whose depletion in the hippocampus has been proposed to be involved in the hyperexcitability process by reducing extracellular glutamate uptake—is well expressed in the subiculum (Eid et al., 2004). Glutamate decarboxylase 67, involved in the synthesis of GABA, is clearly expressed in the subiculum (Sperk et al., 2012). Note that no increase in GFAP expression is detected in the subiculum, suggesting limited gliosis (Eid et al., 2004).
After ex vivo evidence of the key involvement of the subiculum in TLE, some studies have highlighted its role in epileptic patients studied in vivo. First, high-resolution (7 Tesla) MRI tractography has revealed an increased connectivity and fiber density in the presubiculum and parasubiculum (and in the subiculum considering fiber density) in the ipsilateral temporal lobe in patients with left-sided TLE, with reduced fiber density in the contralateral subiculum, contrasting with a lack of difference for right-sided TLE patients (Rutland et al., 2018). The laterality of these findings is unexpected, but these results are consistent with a local network reorganization. In TLE patients characterized by a gliosis predominating on atrophy processes, maximal MRI abnormalities (combining volume, T2 signal intensity, and diffusion markers) were restricted to the subiculum (Bernhardt et al., 2016). Second, in cases of secondary generalization of temporal lobe seizures, the subiculum was shown to be hypotrophic, suggesting a role for the spread of epileptic activities (Wang et al., 2017). These human imaging data are supported by experimental data showing that, before seizure generalization, activation of subicular interneurons by optogenetic activation limits the spread of epileptiform activities; in contrast, once generalization has occurred multiple times, their activation favors generalization of seizures (Wang et al., 2017). Such spreading facilitation effect may be caused by “acquired” depolarizing responses to GABA, related to PV interneurons activation.
In vivo recordings obtained from the subiculum are limited in humans, mostly due to the fact that clinical macro stereo-EEG electrodes record signals obtained from ~1 cm radius make it difficult to distinguish activities from the hippocampus or the EC. Spatial resolution to ascertain subiculum recording requires microelectrodes and such studies are therefore limited. Orthogonal stereo-EEG macroelectrodes associated with a bundle of 40 µm microelectrodes at its tip have provided subicular recordings (Staba et al., 2002). These data have shown that the frequency of unitary spikes and bursts of spikes was higher in epileptic than in contralateral structures, together with an increased synchrony. Spike and burst rates were higher in the subiculum than in the hippocampus and the EC, and they were enhanced during sleep. Microelectrode recordings have also been performed using a laminar electrode under anesthesia (Fabo et al., 2008). These recordings are supported by previous studies showing that interictal spikes were generated in the subiculum (Cohen et al., 2002; Huberfeld et al., 2007, 2011). According to their laminar profile, two different categories of interictal events were identified, with a variable source and variable high-frequency oscillation component, influenced by the hippocampal histological pattern. Recordings with two electrodes encompassing the whole subiculum indicated that discharges were simultaneous within the whole structure. Some of these spikes could spread to hippocampal structures and were then initially detected in the subiculum, highlighting its possible leading role.
We have reviewed in this chapter several findings that support the hypothesis that the subiculum not only “supports” the hippocampal formation but is rather a complex network of interconnected structures comprising several neuronal populations that generate oscillatory rhythms, which play key roles in cognitive processes, such as learning and memory, and pathological activities, such as interictal spikes and seizures. In vitro, several studies suggest that weaker perisomatic GABAergic inhibition of principal cells might contribute to the disruption of excitatory-inhibitory balance in the subiculum, which could favor ictogenesis. In vivo, the subiculum is often the seizure-onset zone of spontaneous seizures in pilocarpine-treated animals, and optogenetic or chemogenetic manipulation of subicular PV-interneurons modulates ictogenesis, therefore indicating that dysfunctional GABA-receptor signaling in this region could contribute to the generation of pathological network activities in focal epileptic disorder. We have finally reviewed studies using ex vivo recordings of human epileptic postoperative tissue showing that the subiculum contributes to the generation of interictal and ictal discharges. However, there is still limited in vivo evidence on the role of the subiculum in TLE in humans, due to the difficulty to selectively record from the subiculum in situ.
According to Xu et al. (2019), the subiculum could play a role in pharmacoresistance in focal seizure disorders. They reported that optogenetic inhibition of pyramidal subicular neurons can reverse phenytoin resistance, whereas activation of these neurons could induce resistance. Moreover, a similar modulation of CA1 pyramidal neurons had no effect on phenytoin resistance. It is noteworthy that similar effects were reproduced with chemogenetic procedures (Xu et al., 2019). According to these authors, pharmacoresistance and subiculum activity could be linked through voltage-gated Na+ channels. Therefore, subicular pyramidal neurons may represent a target to control pharmacoresistance in medial TLE.
This work was supported by grants from the European Research Council (consolidator grant no. 865592) and from the Canadian Institutes of Health Research (PJT153310, PJT166178, and MOP130328).
The authors declare no relevant conflicts.
This is an open access publication, available online and distributed under the terms of a Creative Commons Attribution-Non Commercial-No Derivatives 4.0 International licence (CC BY-NC-ND 4.0), a copy of which is available at https://creativecommons.org/licenses/by-nc-nd/4.0/. Subject to this license, all rights are reserved.
Your browsing activity is empty.
Activity recording is turned off.
See more...
