Thorny Excrescences as a Defining Feature—With a Caveat

The first mention of the term “mossy cells” was based on the seminal work of David Amaral, who published a review of the cell types he found based on Golgi analysis of the rat hilus (Amaral, 1978). As the term “mossy” implies, he found cells in the hilus that appeared to be covered in moss. The mossy appearance was due to the dense clusters of large spines on the proximal dendrites and somata (Fig. 23–1A; Amaral, 1978). The clusters of spines made the cell appear to be covered in moss. The only other cell type that has these unique clusters of large spines, called thorny excrescences, is the CA3 pyramidal cells. At the level of the electron microscope (EM), a thorny excrescence has numerous synapses, and they are complex in their shape (Fig. 23–1B; Chicurel and Harris, 1992). Presynaptic to the thorny excrescences are the massive boutons of GCs, which are packed with glutamatergic vesicles (Fig. 23–1B; Laatsch and Cowan, 1966; Blackstad and Kjaerheim, 1961). The anatomical characteristics suggest the potential for strong excitatory effects. Consistent with that idea, the quantal size of the excitatory postsynaptic potential (EPSP) in a MC produced by a single GC action potential is extremely large (Fig. 23–1C; Scharfman et al., 1990). When multiple GC APs occur, EPSPs show strong frequency facilitation (Fig. 23–1C; Scharfman et al., 1990). Even 1 AP in a GC can trigger APs in MCs (Fig. 23–1C; Scharfman et al., 1990). Furthermore, a train of GC APs can lead to a response in MCs that is similar to a paroxysmal depolarization shift (PDS) in epilepsy (Fig. 23–1C; Scharfman et al., 1990).

Figure 23–1.

MC thorny excrescences and the large excitatory effect of GC giant boutons. A. A hallmark of MCs is thorny excrescences (red arrows). Double red arrowheads mark the axon. 1. A MC filled with biocytin in a rat hippocampal slice using intracellular recording. (more...)

As more studies were conducted in hippocampal slices using intracellular labeling with dyes like biocytin, it became clear that the MCs with thorny excrescences varied with respect to the size of thorny excrescences (Fig. 23–1A; Scharfman and Schwartzkroin, 1988; Scharfman and Myers, 2012). In guinea pig and rat, some thorny excrescences were not as exuberant as others (Fig. 23–1A; Scharfman and Schwartzkroin, 1988). Moreover, some cells that had characteristics of MCs did not have detectable thorns at all. Instead, there were dense spines (Fig. 23–1A; Scharfman, 1993). Therefore, there is a caveat to the idea that all MCs have thorny excrescences. Some may simply have very dense spines. Indeed, early studies used the term “spiny hilar cell” and “MC” synonymously (Scharfman and Schwartzkroin, 1988; Livsey and Vicini, 1992; Scharfman, 1993). The term “MC” became more popular after higher resolution imaging of GABAergic neurons revealed that they can have spines. This led to the adoption of more than one characteristic, besides thorns/spines, to define MCs (Scharfman and Myers, 2012; Scharfman, 2016).

Somatic and Dendritic Characteristics

Most investigators assume that MCs have relatively large somata compared to other hilar neurons. While many MCs do have a large soma, like CA3 pyramidal cells (PCs), other MCs have smaller somata, similar to the size of the DG GABAergic neurons. These MCs have thorny excrescences when recorded intracellularly and labeled by injecting a marker like biocytin (Scharfman and Schwartzkroin,1988; Scharfman, 1993; Scharfman and Myers, 2012) The reason they are called MCs is due to their thorns and electrophysiological characteristics which are distinct from hilar GABAergic neurons (Scharfman, 1992b; Scharfman and Myers, 2012). The MC soma is also varied in shape, sometimes triangular like CA3 PCs (Scharfman, 1993) and otherwise round or oval (Scharfman and Schwartzkroin, 1988; Scharfman, 1993; Scharfman and Myers, 2012), like the hilar GABAergic neurons. For these reasons, it is hard to determine hilar cell type based on the morphology of the soma.

MCs have very long dendrites that are thick near the cell body and then become much thinner (Fig. 23–1A; Amaral, 1978; Ribak et al., 1985; Frotscher et al., 1991  Buckmaster et al., 1996). The dendrites can be long and extend from one end of the hilus to the other (Fig. 23–1A).

A subset of MCs have dendrites in the molecular layer (ML; Fig. 23–3; Scharfman and Schwartzkroin, 1988; Scharfman, 1991; Blackstad et al., 2016). These cells have a lower threshold than adjacent GCs to input from the lateral perforant path (LPP; Scharfman, 1991). Consistent with those data, a recent study using rabies virus to label inputs to MCs showed that MCs receive monosynaptic input from the lateral entorhinal cortex (LEC; Azevedo et al., 2019). This pathway may allow a subset of MCs to be sensitive to object information and other sensory input that is relayed from to the superficial layers of the LEC to MCs (Fig. 23–3). That pathway may allow MCs cells to detect novelty in the environment and relay that information to GCs. In vivo data from the rat suggest these MCs may be ventral primarily (Fig. 23–3), because the ventral MCs preferentially expressed the activity marker c-fos protein after exposure to novelty (Duffy et al., 2013; Bernstein et al., 2019). Behavioral data are consistent with this general idea, because specific MC inhibition or excitation using chemogenetics led to altered effects in a novel object recognition task (Botterill et al., 2021b). Other studies of MCs vary in their ability to show effects in this task but do show some effects of behaviors involving objects or changes in the environment (Jinde et al., 2012; Bui et al., 2018; Fredes et al., 2021). Some of the differences among studies may be technical because the novel object tests are not conducted in the same way from one study to another. In addition, sex and other factors influence the ability to detect an effect of MCs on behavioral tasks (Botterill et al., 2021b).

Figure 23–3.

Circuitry that could lead to a functional role of MCs related to novelty recognition and environmental context. A. The left and right hippocampus from a rodent brain are shown side by side (lateral view). The PP input from the LEC sends direct projections (more...)

This topic is highly relevant to comorbidities in epilepsy because of the loss of superficial layers of the EC in TLE (Du et al., 1993; 1995), and the loss of MCs in TLE (Scharfman, 1999; Scharfman and Myers, 2012). Thus, the loss of some cognitive ability in TLE may in part be due to loss of neurons in critical circuits that provide information about the environment to the hippocampus.

Axon

Initial studies of hilar neurons, before they were divided into subtypes, noted that many large hilar cells projected to the contralateral DG with axons that traveled in the hippocampal commissure (Zimmer, 1971; Buzsaki and Czeh, 1981; Douglas et al., 1983; Ribak et al., 1985). Later it became clear that MCs formed a significant proportion of the commissural projection, which terminated mostly in the inner molecular layer (IML).

After methods to distinguish MCs from other hilar neurons developed, more information was obtained specifically about MCs. The experiments showed that the majority of the axon leaves the local area and terminates at distal locations in the DG IML (Fig. 23–2A; Buckmaster et al., 1996). The local area does have some axon collaterals, however (Scharfman and Schwartzkroin, 1988; Buckmaster et al., 1996; Scharfman and Myers, 2012), and hilar axons may also be present distal to the soma. Quantitatively, it appears that the vast majority of synapses distant to the MC soma are in the IML on GCs (Buckmaster et al., 1996). The remainder is on GABAergic neurons (Buckmaster et al., 1996). In addition, two groups have now shown that MCs can innervate other MCs, and this appears to occur local to the MC soma by MC axon collaterals in the hilus (Sun et al., 2017; Shi et al., 2019; Ma et al., 2021).

Figure 23–2.

Characteristics of the MC axon. A. A diagram illustrates the location of the two hippocampi in the rodent brain from a side view. B. A mouse was used to inject AAV dorsally in one DG to evaluate the dorsal MC axons. Near the site of injection there is (more...)

New methods have revealed variations in MC axons across the septotemporal axis, and this suggests the axon is different in dorsal MCs relative to ventral MCs (Fig. 23–2B). These methods take advantage of mouse lines with Cre recombinase (Cre) primarily in MCs and use stereotaxic injection of adeno-associated virus (AAV) in the DG to label MCs with fluorescent markers. Two studies have now shown that dorsal MC axons terminate in ventral DG in a way that is different than expected (Botterill et al., 2021a; Houser et al., 2021). The dorsal MCs have axons in all subdivisions of the ventral ML rather than being restricted to the IML (Fig. 23–2B). In contrast to dorsal MCs, ventral MCs have an axon that is consistent with past views that the ML axon terminals are restricted to the IML in all locations of the septotemporal axis (Fig. 23–2A). These findings have some precedent in the rat, where intracellular injection of biocytin in vivo led to labeling of MCs (Buckmaster et al., 1996). Notably, the cells were relatively dorsal. These dorsal MCs are those that appear to have more axon terminals in the ventral ML, and ventral axons indeed had collaterals in the outer two-thirds of the ML, although the IML axon still was a terminal zone. The Buckmaster et al. study was in rats, which is notable because the more recent work of Botterill et al., and Houser et al. were in mice. Therefore, the dorsal MC axon in rats could have fewer axons terminate in the ventral middle molecular layer (MML) and outer molecular layer (OML). On the other hand, the mouse studies visualized the ventral axon using membrane labeling and confocal microscopy that might be more sensitive than the cytoplasmic marker biocytin observed in brightfield using rats.

Neurotransmitter

Initial views suggested that MCs were inhibitory and potentially GABAergic. The primary evidence for this view was from recordings in the GC layer in anesthetized rats where extracellular recordings were used to analyze subsets of GCs that had APs in response to a perforant path (PP) stimulus. A commissural stimulus triggered before the PP stimulus inhibited the extracellularly recorded GC APs, called the population spike (Buzsaki and Eidelberg, 1981; Douglas et al., 1983). However, later studies provided strong evidence that MCs use glutamate as a neurotransmitter (Soriano and Frotscher, 1994; Scharfman, 1995b). The explanation for the inhibitory effects of MCs on GCs that were observed earlier is that MCs use glutamate to activate GABAergic neurons in the DG. These GABAergic neurons innervate GCs, leading to inhibition of GCs (Fig. 23–4). The two studies that initially showed MCs use glutamate as a neurotransmitter were distinct in methods. The first study used a technique to identify glutamate in Golgi-stained neurons with the appearance of MCs (Soriano and Frotscher, 1994). The second study used simultaneous recording of monosynaptically- connected MCs and GCs in slices (Scharfman, 1995b). Unitary EPSPs in GCs were elicited by MC APs. The evidence that MC activated DG GABAergic neurons also came from simultaneous recordings of connected cells, this time from MCs and GABAergic neurons in the hilus (Larimer and Strowbridge, 2008). Other studies supported these findings. For example, labeling the commissural projection so that a light-activated opsin (channelrhodopsin; ChR2) would express ChR2 in contralateral MCs, slices could be made of the contralateral DG to study light activation of MCs while recording in GABAergic neurons. These studies showed that several types of GABAergic neurons were activated by ChR2 as well as GCs (Hsu et al., 2015). However, ChR2 activation is difficult to use to define monosynaptic connections because typically light activation makes it hard to know when the presynaptic cell released transmitter. More recently, evidence that MC activation of hilar GABA- and somatostatin (SOM)-expressing cells was shown (Li et al., 2021). Notably, there is an exceptional study which suggested that MCs express GABA (Gonzalez-Reyes et al., 2019).

Figure 23–4.

MC excitation vs. inhibition of GCs. A. A diagram of the normal DG circuitry. MCs innervate GCs in the IML and DG GABAergic neurons (black). Besides the GCs, the other glutamatergic cells are MCs in the hilus (green hilar cells). The prevailing view (more...)

These fundamental studies of MCs and their circuitry are important when considering how MC loss contributes to epileptogenesis and chronic seizures in TLE, as discussed below.

Cell-Specific Markers of MCs

MCs can be labeled by antibodies to a number of proteins they express, although the specificity is often imperfect, so several antibodies need to be used or caveats about the specificity of expression need to be made. For example, calretinin is a calcium-binding protein that is found in MCs of rodents and is mostly expressed in ventral MC somata, although in dorsal hilus there is faint somatic labeling of MCs and robust IML labeling of the axons of the ventral MCs (Fujise et al., 1998; Liu et al., 1996; Blasco- Ibanez and Freund, 1997). Its specificity for MCs is limited because calretinin expression occurs at a specific stage of development of young adult GCs that are born throughout life. Although these young adult-born GCs are mainly in the subgranular zone and GC layer, MC somata are present in the subgranular zone. Moreover, young adult-born GCs that express calretinin are scattered throughout the hilus (Bermudez-Hernandez et al., 2017). One way to discriminate between the adult-born GCs and MCs is with two markers: Prox1, which is specific for GCs (but not MCs), and calretinin. A Prox1-negative, calretinin-positive cell in the hilus would be a potential MC. It is notable that originally the calretinin-positive hilar cells were considered to be GABAergic (Gulyas et al., 1996), but these cells were in the rat and do not appear to be in mouse (Bermudez-Hernandez et al., 2017). In the mouse, hilar calretinin cells that are glutamatergic appear to only be young adult-born GCs (Bermudez-Hernandez et al., 2017).

Based on the neurotransmitter of MCs, the GluR2/3 receptor discriminates MCs from the hilar GABAergic neurons. A Prox1-negative GluR2/3 positive cell would be the best approach to identify MCs, because the young-adult GCs of the hilus will be labeled by both Prox1 and GluR2/3.

Other markers of MCs include antibodies to the neuromodulator calcitonin gene-regulated peptide (CGRP) and the cannabinoid receptor type 1 receptor (CB1R; see section below on Endocannibinoids; Scharfman, 2016). Both antibodies label the MC terminals primarily. A caveat is that CGRP labeling can be somewhat weak, and the CB1 receptors also exist on other DG neurons. Additional potential markers are discussed more below where they are used in transgenic mice to allow Cre expression preferentially in MCs. A more complete list of markers without regard to transgenic mice is provided elsewhere (Scharfman, 2016). An important consideration is how these markers change with epilepsy, when many changes in gene expression occur not only inside but also outside the DG.

Electrophysiology

Most of our past understanding of the electrophysiological characteristics of MCs is based on recordings in hippocampal slices, first with sharp microelectrodes in guinea pig and rat, and later using patch electrodes in mouse. However, unit recordings of putative MCs in vivo, and Ca²⁺ imaging of MCs in vivo, have also added to our understanding. Together the data suggest that MCs have characteristics similar to other principal cells that have been called regular spiking (McCormick et al., 1985; Scharfman, 1992b; Thomson and Deuchars, 1997). The regular spiking neurons were named originally because of their broader AP relative to GABAergic neurons, which historically have been called fast spiking. It is notable, however, that some data suggest the division is not so simple. For example, some GABAergic neurons in the DG have a relatively long AP (Scharfman, 1995a; Hosp et al., 2014). In these cases, the MC and GABAergic neurons can be distinguished by other differences, such as the dv/dt ratio of the AP (Scharfman, 1992b; 1995a; Scharfman et al., 2000). When elicited at threshold by intracellular current, the rate of rise of the AP (dv/dt of the rising phase) relative to the rate of decay of a GABAergic neuron is close to 1. In contrast, the MC AP has a much sharper rate of rise than the rate of decay, leading to a dv/dt ratio much greater than 1, like GCs. More recently it has become clear that some GABAergic neurons have slower As, so the nomenclature has become more complex. Despite the caveat about slow spiking GABAergic neurons, in vivo unit recordings often use the broad spike to distinguish MCs from GABAergic neurons.

Other distinguishing intrinsic properties of MCs include a very long time constant, like CA3 PCs and very different from GCs and DG GABAergic neurons. However, this difference is greatest using sharp electrodes and diminishes with whole-cell recordings. In general, the intrinsic properties are greatly influenced by the recording methods, so care must be taken when comparing studies. Similarly, the way the intrinsic property is tested is important, meaning if cells were all held at the same potential regardless of RMP or if cells varied because they were recorded at their RMP and RMP could be as depolarized as –50 or as hyperpolarized as –90 mV. Similarly, an AP that is triggered by a rectangular pulse at threshold will exhibit differences from an AP of the same cell that is spontaneous or triggered by a ramp, and so on. The AHP varies depending on whether one or a train of spikes was triggered. In general, the amplitude of the MC AP, AHP, R_in, and voltage “sag” (referred to by various names, such as anomalous rectification, I_Q, etc.) vary across investigations of MCs, often for these reasons. Whole-cell recordings in vivo have the additional difficulty of not being able to make the best recordings, and often the variable of anesthesia adds further differences to in vitro recordings. The reader is referred to excellent papers that show some of the variable intrinsic properties of MCs in the literature (Scharfman and Schwartzkroin, 1988; Lübke et al., 1998; Buckmaster and Amaral., 2001; Howard et al., 2007).

Another hallmark of MCs is that there is a very frequent barrage of excitatory synaptic input and if excitatory and inhibitory neurotransmission is not blocked by antagonists, the firing of MCs will be extremely variable from one current or voltage command to the next. Firing can occur in bursts or trains, and it typically shows less spike frequency adaptation than GCs but more than GABAergic neurons (Scharfman and Schwartzkroin, 1988; Scharfman, 1993).

The spontaneous EPSPs of MCs are very large (Scharfman and Schwartzkroin, 1988; Strowbridge et al., 1992; Scharfman, 1993), and the frequency can be so great that the EPSPs are overlapping. Many of these large events may be due to spontaneous glutamate release from the giant boutons of MCs that oppose MC thorny excrescences (Scharfman et al., 1990). These data were originally obtained in slices, and similar findings have been made in vivo recently with Ca²⁺ transients reflecting MC activity (Danielson et al., 2017). GABAergic neurons also can have frequent EPSPs, and these events can be distinguished from MCs because they are smaller than MC EPSPs and their kinetics are typically faster (Livsey and Vicini, 1992).

Interestingly, the frequency of spontaneous EPSPs varies even in the same cell, so that for periods of seconds to minutes there is high frequency and AP firing, followed suddenly by a cessation of firing and lower frequency. Several studies examined this phenomenon and found that APs of MCs could trigger the periods of high-frequency activity (Strowbridge et al., 1992; Scharfman, 1993; Strowbridge et al., 1996; Anderson and Strowbridge 2014). Later it was identified that MCs innervate other MCs (Shin et al., 2017; Ma et al., 2021), suggesting that positive feedback between synaptically connected MCs may be a contributing factor. Additionally, MC excitation of GCs may lead to reactivation of the same MC in slices where all the necessary MCGC connectivity exists.

The intrinsic properties and firing behavior of hippocampal cell types change in TLE (Beck and Yaari, 2008). Although less is known about MCs than GCs, there is considerable plasticity of intrinsic firing in GCs in TLE models. Therefore, predictions about the characteristics of the DG cells in TLE, based on what is known about the characteristics of DG cells in the normal brain, need to be made cautiously.

Dorsal and Ventral MCs

The MC axon projection of the mouse varies, depending on the location of the soma. As described above, dorsal MCs have a projection in ventral DG that extends into the MML and even the OML. In contrast, ventral MCs project to the IML throughout the septotemporal axis. These differences are not the only distinctions that have been described for dorsal versus ventral MCs. For example, MCs from ventral DG exhibit bursts when the slice is exposed to a buffer that increases excitability (Jinno et al., 2003). The expression of calretinin in MCs is strong in the somata of ventral MCs but not dorsal MCs (Liu et al., 1996; Blasco-Ibanez and Freund, 1997; Fujise et al., 1998). MCs in the part of the hilus near the inferior blade of the DG are more frequently lost relative to the MCs near the superior blade—at least after pilocarpine-induced status epilepticus (SE) in the rat (Scharfman et al., 2002). Functionally, more distinctions between the MCs in different parts of the DG have been shown and these are elaborated below in the section about behavior.

Summary

In summary, MCs form a discrete cell type in the hilus that is characterized by dense clusters of spines, a hilar and IML projection primarily, and electrophysiology that is distinct from the other cell types. Notably, all three characteristics are necessary to be sure a cell is a MC because some MCs have less exuberant thorny excrescences than others. The regular spiking nature of MC electrophysiology is subject to error if the electrophysiology is recorded from a cell that is unhealthy. The axon is variable in the degree it collateralizes in the hilus and its restriction to the IML and may not be fully visualized if the method involves slices or weak staining. This often leads to studies that mention MCs were sampled but actually may not have been. Conversely, there also can be underreporting of MCs in other studies.

Inputs and Outputs of MCs in the Normal DG

To consider how MCs influence the DG functionally, it is important to know their inputs and outputs. Many of these synaptic connections are reviewed elsewhere (Scharfman, 2016).

The main afferents to MCs are glutamatergic and GABAergic. Glutamatergic input includes the terminals of GCs which cause very large unitary EPSPs in monosynaptically-connected GCs and MCs (Scharfman et al., 1990). In addition, the LEC innervates MCs (Azevedo et al., 2019) and so do CA3 PCs (Scharfman, 1994b). These inputs and others are much weaker than the GC input (Scharfman et al., 1990; Scharfman, 1995b). The diverse GABAergic neuronal input has not been quantified well, but strong inhibitory inputs have been known for a long time (Soltesz and Mody, 1994; Livsey and Vicini, 1992). In our experience these inhibitory inputs are often masked by the very large excitatory input (Scharfman, 1992a), but this is not the experience of all laboratories, and the differences have not been fully explained. Semilunar GCs (Williams et al., 2007) and adult-born GCs provide excitatory input to MCs (Drew et al., 2016; Toni et al., 2008).

Inputs from extrinsic sources are abundant. Cholinergic input has been well documented, arising from the basal forebrain and additional cholinergic nuclei. The brainstem pathways that are modulatory (noradrenergic, serotoninergic, dopaminergic) influence MCs, but whether the terminals of these ascending pathways innervate MCs or diffusely release transmitter in the hilus, potentially spreading to MCs, is not clear. One report does provide support for the idea of diffuse release of dopamine in the hilus, which then could influence dopamine receptors on MCs (Etter and Krezel, 2014). Release of dopamine from the locus coeruleus could influence dopamine receptors on GCs also (Kempadoo et al., 2016). Other direct inputs are debated, such as the supramammillary input to the IML, which could influence MC axons.

MCs have two main outputs, to GCs and GABAergic neurons. For GCs, MCs primarily innervate GCs in the IML that are distal to the soma (Buckmaster et al., 1996). Of the types of GCs that are innervated, it is notable that MCs innervate the adult-born GCs at a particularly influential time in development, when adult-born GCs lack input from other sources. At this time, the young adult-born GC mainly has an apical dendrite with a tuft in the IML, but dendrites are not yet in the ML (Chancey et al., 2014). At this time the young adult-born GC is only starting to make synaptic connections, and the role of MCs appears to be a nurturing role (Gonzalez-Reyes et al., 2019; Uemura et al., 2021) with some reports indicating the MCs are critical to adult neurogenesis (Yeh et al., 2018).

Of the GABAergic neurons that MCs target, evidence in the rat suggests the primary output is to the basket cell of the GC layer which is typically pyramidal in morphology (Scharfman, 1995a). However, MCs also innervate hilar GABAergic neurons (Larimer and Strowbridge, 2008) and in the mouse, SOM neurons of the hilus. Given the MCs have both ML and hilar projections, it seems likely that MCs innervate diverse types of GABAergic neurons.

Are MCs Excitatory or Inhibitory to GCs?

One of the debates about MCs is their circuit function. For decades, it has been unclear whether MCs primarily excite or inhibit GCs (Fig. 23–4). One hypothesis is that MCs primarily are excitatory to GCs and this serves an important purpose to enhance the ability of GCs to perform tasks related to memory (Buckmaster and Schwartzkroin, 1994). At the core of the hypothesis is quantitative EM data showing that the primary projection of MCs, to areas distal to the soma, terminates on spines in the IML (Buckmaster et al., 1996). The spines are highly likely to represent GC dendritic spines, given those are the vast majority of spines in the IML. In addition to these data, MCs make en passant synapses in the IML, which could lead to synchronization of numerous GCs (Buckmaster et al., 1992; Buckmaster and Schwartzkroin, 1994; Wenzel et al., 1997). Other studies also support the hypothesis that MCs primarily excite GCs. In hippocampal slices, MCs were individually killed using a controlled approach, GCs excitability did not increase, suggesting MCs do not inhibit GCs normally and (Ratzliff et al., 2004).

Thus, the other hypothesis for MC function in the normal brain is that MCs are primarily inhibitory, and this is due to their activation of DG GABAergic neurons which innervate GCs (Fig. 23–4). Initial evidence for this hypothesis was obtained using a method to stimulate the PP intermittently in vivo under anesthesia. After many hours, MCs were killed, and GC population spikes elicited by PP stimulation increased in number, suggesting disinhibition (Sloviter, 1991a, 1991b). However, there was damage to hilar SOM-expressing neurons as well as MCs, and it is likely that there were other changes in response to cell death. Therefore, it was valuable when additional studies were conducted with other methods suggesting that the hypothesis was correct, although there were counterarguments. Many of the studies supporting the idea that MCs were mainly inhibitory arose after the advent of optogenetics and chemogenetics. For example, when commissural axons of hilar neurons were activated by ChR2, GCs population spikes elicited by PP stimulation were inhibited (Hsu et al., 2015). In other studies, ChR2 activation of MC axons evoke IPSCs that are blocked by glutamate receptor antagonists and occur at a latency consistent with a disynaptic (MC→GABAergic neuron → GC) pathway (Hashimotodani et al., 2017). In in vivo recordings that discriminated MCs from GCs using the characteristics of the unitary spikes, more evidence for a net inhibitory effect was obtained (Senzai and Buzsaki, 2017).

Some studies suggest that both an excitatory and inhibitory role exist, and each effect is used under different conditions. For example, one study using Drd2-Cre mice showed that ChR2 leads to EPSPs in more GCs than inhibitory postsynaptic potentials (IPSPs; Bernstein et al., 2020). The same study then examined subsets of GCs and found that when the PP is stimulated, the net effect was inhibition of GC population spikes in slices. However, when timed appropriately, there was an excitatory effect on the PP input. The timing was critical, so that there was minimal delay between the PP stimulus and onset of MC-mediated excitation of GCs. That is notable because there was a significantly longer delay in ventral slices. That would make the dorsal MCs potentially have less of an excitatory effect. A different study suggested that the ventral MCs synergize with the PP input to promote GC excitation (Houser et al., 2021), which is consistent with the idea that ventral MCs would be more able to facilitate the dorsal GCs than dorsal MCs. Another study suggested that dorsal MCs are more “inhibitory” than ventral MCs based on in vivo behavior and c-fos staining (Fredes et al., 2021). These ideas are potentially relevant to TLE where dorsal MCs were able to inhibit seizures more than ventral MCs (Bui et al., 2018).

A complication to the normal circuitry is the idea that MCs innervate each other. This idea is based on relatively new methods such as rabies virus tracing, Ca²⁺ signals in MCs, and Cre-dependent voltage sensors (Sun et al., 2017; Shi et al., 2019; Ma et al, 2021). However, these methods do not discriminate EPSPs from APs and IPSPs are difficult to study. Finding MC→ MC connections would be notable because the MC→MC connections would make the MCs like CA3 cells, meaning that memory storage in MCs could lead to amplification and storage functions like pattern completion, analogous to CA3 recurrent collaterals.

Plasticity

Consistent with the role of MCs in learning and memory, many studies have provided evidence that MC → GC synapses exhibit long-term potentiation (LTP). One of the first used recordings in vivo from the IML (Hetherington et al., 1994), but the ability to selectively stimulate the IML and record the MCGC synapses selectively was not possible. Later studies stimulated the PP at high frequency in vivo and showed that MCs ipsilateral to the stimulus develop a high level of GAP43, a plasticity-related protein linked to memory (Namgung et al., 1997). Remarkably, GCs and pyramidal cells did not show the elevation in GAP43. The authors suggested MCs play a pivotal role in PP LTP. More recently, slice studies have also suggested a critical role of MCs in LTP, and of the potential for LTP of GC inputs to influence CA3 PCs (Namgung et al., 1997). More recently, selective MC activation of GCs showed clear evidence of LTP at the MC→GC synapses (Hashimotodani et al., 2017). That study also showed dependence of LTP on molecules linked to memory such as BDNF and its receptor TrkB. Other factors that have been shown to be important in activity-dependent synaptic plasticity have been identified at the MC terminals in the IML, such as CB1 receptors (Chiu and Castillo, 2008).

During high-frequency activation of MCs, it appears that they are more likely to produce LTP at the excitatory synapse on GCs rather than excitatory synapses on GABAergic neurons. These synapses are now known to be different pharmacologically and during trains of stimulation to the DG (Hashimotodani et al., 2017; Hedrick et al., 2017; Li et al., 2021). This is not only critical to understanding the neuronal circuitry under normal conditions, but it is also relevant to epilepsy. For example, the MC→GC synapse may be potentiated preferentially during conditions of high-frequency input by the PP. That could contribute to the idea that during the initial insult that precipitates TLE, MCs are primarily proconvulsant, whereas later, in chronic epilepsy, the circuit changes could make the role of MCs much different. These views are discussed further below.

Hippocampal EEG

Of the many studies about MC activity and its role in hippocampal electroencephalography (EEG), two topics merit emphasis because they have potentially important ramifications in epileptogenesis and epilepsy. One topic is the role of MCs in theta rhythm. It has been known for some time that MCs participate in theta rhythm (Soltesz et al., 1993), and these findings contributed to the increasing interest in how MCs are important to behavioral functions associated with theta rhythm such as spatial exploration and encoding of experience. During theta rhythm in normal rodents, MCs typically fire during the phase of theta similar to CA3 PCs (Soltesz et al., 1993; Henze and Buzsáki, 2007; Senzai and Buzsaki, 2017). However, the role of MCs in theta is not necessarily the same as CA3 PCs. The behavior of MCs in theta rhythm, its regulation, and its relevance to function are still being established (Senzai and Buzsaki, 2017; GoodSmith et al., 2017, 2019). Moreover, how MCs may be involved in other oscillations such as gamma is only starting to be revealed.

In epilepsy, theta and gamma oscillations have been of great interest because investigators have reported (or shown in their data) increased or decreased theta rhythm before, during, or after seizures. In one study where MCs were inhibited chemogenetically, there were changes in theta as well as other frequency bands (Botterill et al., 2019). Theta oscillations occurring for several seconds have been suggested to be seizures, but this is controversial (Dudek and Bertram, 2010). One reason to be cautious about theta oscillations in epileptic animals, and if they reflect seizure activity, is that theta oscillations appear in hippocampus when spike-wave discharges occur in normal animals (Pearce et al., 2014). The spike-wave discharges are generated distant from the hippocampus in thalamus and are similar in frequency to theta in rodents (Pearce et al., 2014). Regarding gamma oscillations, they are of interest in epilepsy because they appear to precede seizures (Traub et al., 2001). Very fast oscillations (high-frequency oscillations; HFOs, typically >200 Hz) are very relevant to epilepsy, where they may influence the seizures or mark the seizure-onset zone (Levesque and Avoli, 2019; Frauscher et al., 2017; Gloss et al., 2017). They are rare in normal animals, but common in both animal models of epilepsy (Levesque and Avoli, 2019; Lisgaras and Scharfman, 2021a; Bragin et al., 2004) and clinical epilepsy (Cimbalnik et al., 2016). Very little is understood about MCs in these oscillations or their modulation by HFOs.

A role of MCs in theta is interesting because theta appears to be inhibitory to seizures and has an anticonvulsant effect in animal models (Miller et al., 1994). If MCs contribute to theta rhythm, it would be consistent with the view that MCs are primarily inhibitory in the normal brain. The inhibitory effects of MCs on GCs during theta oscillations could be part of the reason that few seizures occur.

Another topic related to hippocampal oscillations which has relevance to epilepsy is sharp waves. Unfortunately, “sharp waves” is a term that is used to refer to diverse phenomena in normal and epileptic conditions. Regrettably, the same term is used despite different meaning. In most cases, there is a sharp wave caused by synchronous firing of a subset of principal cells near the recording site. The field potential that is recorded adjacent to the principal cells is a downward (negative) deflection because Na⁺ ions rush into the principal cells during AP generation. Extracellularly, there is a loss of Na⁺ ions, leading to a negativity. The negativity is relatively fast in onset and decay, like the APs it reflects, leading to the name “sharp wave.” In epileptic animals, the term “sharp wave” is also used to refer to an event with a different mechanism. If there is synchronous activity of a subpopulation of GABAergic neurons that inhibit a subset of principal cells, a sharp positivity occurs as GABA_A receptors open and Cl⁻ leaves the principal cells (Muldoon et al., 2015). The field potential that is generated outside the principal cells reflects IPSPs, not APs. Still, the term “sharp wave” may be used. Thus, the term “sharp wave” is used for sharp deflections with a range of amplitudes and the entire wave can be as long as several hundred milliseconds. The variations in meaning lead to confusion about what a sharp wave is.

For the classic hippocampal sharp waves of CA3 PCs in the normal brain, they are generated near the PC cell bodies where APs are generated. They are caused by synchronous bursts in a subset of CA3 PCs (Buzsaki, 1986). The EPSPs generated in CA1 apical dendrites by the CA3 projection to CA1 can also be called sharp waves. Near the CA3 PCs there is additional activity in adjacent GABAergic neurons which cause small field potential oscillations in the CA3 PCs called ripples. The complex of sharp wave and ripples are called sharp wave-ripples (SPW-R). They are a normal part of the hippocampal EEG that occurs in the rodent during immobility and sleep (Buzsaki, 1986). They are thought to be important to memory consolidation of what occurred during the awake state (Buzsaki, 1989; Roux et al., 2017). The consolidation is thought to be caused by the synaptic plasticity initiated by the SPW-R in its downstream targets, such as CA1 (Buzsaki, 1989). Thus, after the SPW-R in CA3, there is potentiation of the input to CA1 dendrites, and this is thought to cause the consolidation. CA1 also generates activity that can initiate potentiation in downstream areas such as cortex, and these downstream areas are considered to be important to long-term memory storage (Buzsaki, 1989). More recently several aspects of SPW-R and memory have been elucidated such as the role of CA2 (Roux et al., 2017).

In the epilepsy literature, “sharp waves” are events that are usually considered to be interictal spikes and pathological. They are usually quite large relative to hippocampal SPW-R in the normal brain, but it depends on the recording system and other methods. If multiple electrodes are used to record the EEG in cortex and hippocampus, these interictal spikes are usually recorded nearly simultaneously at different electrodes. Notably, there is evidence that interictal spikes disrupt memory during epileptogenesis rather than improve it (Kleen et al., 2010). In clinical literature, interictal spikes or epileptiform activity (less uniform than interictal spikes and more complex) impairs memory also (Kleen et al., 2013). Interictal spike-like events can also occur in other diseases where there are seizures such as Alzheimer disease and are considered pathological (Vossel et al., 2013, 2017). Thus, reducing the spikes improved memory in one study of an animal model of Alzheimer disease (Sanchez et al., 2012). However, in other diseases with comorbid seizures such as autism, the functional impact of spikes is unclear (Ghacibeh and Fields, 2015). Why a SPW-R-like event in epilepsy impairs and in the normal brain it improves memory may simply be that in epilepsy the event is so much larger than normal hippocampal SPW-Rs, and often reflects synchronous activity in many locations of the brain rather than hippocampus only.

During normal and “pathological” SPW-Rs, MCs are activated. The mechanism appears to be the CA3 PC backprojection to MCs (Scharfman, 1994b, 2007). Thus, in normal slices when CA3 PCs are stimulated, MCs are activated monosynaptically (Scharfman, 1994b, 1996). The backprojection also innervates DG GABAergic neurons, so stimulation of CA3 PCs activates both MCs and GABAergic neurons in the DG at about the same time (Scharfman, 1994a, 1994c; Kneisler and Dingledine, 1995). In the hands of some investigators, slices from normal rodents exhibit small SPW-R spontaneously, and MCs are usually activated (Swaminathan et al., 2018). It should be noted, however, that these small SPW-R may occur in slices that have not preserved inhibitory circuits because in our experience they occur when GABAergic neurons are poorly preserved during slice preparation. In these slices, very small SPW-Rs occur in the CA3 PCL spontaneously. They also occur when extracellular Mg²⁺ is reduced to nominal 0 mM, and in this case there are clusters of unit activity that cause small SPW-like events which recruit more and more PCs until seizure activity occurs (Lu and Scharfman, 2021). Therefore, spontaneous SPW-Rs in normal slices may reflect hyperexcitability. One reason to propose these slices do not reflect normal conditions is because the normal rodent slices with the small SPW-Rs are associated with poor synaptic plasticity rather than improved synaptic plasticity (Colgin et al., 2004b).

In summary, MCs are quite relevant to normal hippocampal SPW-Rs (Scharfman, 1994a, 1994c, 1996; Penttonen et al., 1997; Henze and Buzsáki, 2007; Swaminathan et al., 2018). In the normal animal without hyperexcitability, the interactions of the DG with CA3 and MCs/GABAergic neurons provide an advanced level of information processing by the DG. A seminal role of this circuitry has been examined in DG-dependent memory-associated functions (Lisman et al., 2005; Myers and Scharfman, 2009; Myers and Scharfman, 2011).

In animal models of TLE, CA3 PCs develop spontaneous interictal-like spikes which are bursts of a subset of PCs, like SPW-Rs, but larger because more PCs participate (Scharfman et al., 2000). In an epileptic animal, slice studies showed virtually all residual CA3 PCs that survived the epilepsy participated in the bursts (Scharfman et al., 2000). Interictal spikes/bursts were an order of magnitude greater than spontaneous SPW-R in normal slices (Swaminathan et al., 2018; Colgin et al., 2004a). Some investigators do not find these giant events are spontaneous but can evoke them by stimulation of CA3 in an epileptic animal (Hedrick et al., 2017). The strong activity of CA3 activates MCs more than normal (Scharfman et al., 2001; Hedrick et al., 2017). The large excitatory events of MCs have also been observed in an animal model of head trauma where epilepsy can develop with time (Santhakumar et al., 2000). In these animal models, the increased excitation of MCs may lead to stronger MC→GC excitation by a CA3→MC→GC pathway because of weaker GABAergic neuron→GC inhibition by a CA3→GABAergic neuron→GC pathway (Scharfman, 1994a). Reduced inhibition could occur because GABAergic neurons are lost or dysfunctional, or for other reasons. If there is more MC excitation than normal during interictal spikes in the epileptic brain, one would predict strong activity in GCs, and this pathway would depend on MCs (Scharfman, 1994a, 1994b, 1994c; Scharfman et al., 2001). Indeed, large excitatory events in GCs have been shown in hippocampal slices of epileptic rats following the spontaneous, large interictal spikes/bursts that produce large excitatory events in MCs (Scharfman et al., 2000; Scharfman, 2007). Interestingly, part of the reason GCs exhibit large excitatory events is that the input from MCs occurs at the same time as increased excitation of GCs by ectopic GCs (Scharfman et al., 2000).

Behavior

New data have shed a great deal of light on the role of MCs in behavior, a topic that is very important in epilepsy because of behavioral comorbidities that might be related to MC loss.

One long-standing question is how MCs might contribute to spatial memory in the normal behaving animal. Therefore, it was important when MCs were found to have place fields and that there was considerable plasticity (GoodSmith et al., 2017, 2019; Senzai and Buzsaki, 2017). These data and others have made a compelling case that MCs are involved in the coding and processing of information related to spatial location.

In the DG, this topic is often raised in the context of a specific function attributed to the DG called pattern separation. Thus, the DG contributes to hippocampal-dependent memory by separating or orthogonalizing the excess of input from the EC PP input relative to the numbers of GCs. The GCs have been described as filtering or separating the patterns of input and then passing the separated input to CA3. This filtering or processing may become deficient in epilepsy, leading to both a deficit in memory and also excess excitatory input to CA3, leading to seizures.

Early studies by Marr and others lay the theoretical foundation for these ideas, which were later found to have empirical support using laboratory rodents. However, we recognized that the majority of these studies failed to incorporate any role for hilar neurons in pattern separation or the interaction of the DG with CA3. An exception was the laboratory of John Lisman who described the potential importance of the backprojection from CA3 to MCs and then GCs (Lisman et al., 2005). More recently the role of MCs in pattern separation has been specifically raised by in vivo studies of the Knierim laboratory (GoodSmith et al., 2019). To address the few studies about hilar neurons in pattern separation, and also to address the issue with the benefits of a network model, we developed a network model and then tested the contribution of MCs in the model using an approach that selectively manipulated MCs and also simulated pattern separation computationally (Myers and Scharfman, 2009). The results showed impaired pattern separation-like function (Myers and Scharfman, 2009).

Subsequently the model was expanded and then implemented to study the role of the backprojection from CA3 to MCs (Myers and Scharfman, 2011), complementary to the work of Lisman, where he invoked the backprojection to address DG-CA3 interactions in behavior (Lisman et al., 2005). Our results suggested a very important role of the backprojection (Myers and Scharfman, 2011). Next we began to study changes in MCs and other cell types in a simulation of the epileptic brain (Myers et al., 2013). It was shown that the altered circuitry in the model, including loss of MCs, had substantial deleterious effects (Myers et al., 2013). One of the contributing cell types was the ectopic hilar GC born after SE (Scharfman et al., 2000, 2007). These cells had been predicted to contribute to persistent seizures and cognitive comorbidities in epilepsy (Scharfman et al., 2000; Scharfman, 2004). The prediction was supported later experimentally (Cho et al., 2015) when it was shown that reducing the ectopic GCs improved chronic seizures and a cognitive task that tests the memory for object location.

Since that time, additional empirical and modeling studies of the normal and epileptic DG have further identified pattern separation functions that are potentially mediated by MCs. However, it is also important to note that numerous cell types other than MCs are involved in pattern separation.

Another hypothesis that has substantial support is the role of MCs in novelty recognition, and novel objects located in the environment, which has been discussed in the preceding text. Additional discussion of this topic is valuable to consider its validity and potential importance.

A role of MCs in novelty recognition is consistent with the understanding that the DG, particularly ventral DG, is important to contextual fear conditioning, originally identified based on studies of ventral DG lesions where there was reduced contextual fear conditioning, although not cued fear conditioning (Phillips and LeDoux, 1992). Further studies with more specificity reinforced the view that MCs are involved in behaviors related to contextual discrimination (Jinde et al., 2012; Oh et al., 2019; Jung et al., 2019; Fredes et al., 2021).

Other studies also have added important information. Using two mouse lines, the Drd2-Cre and Crlr-Cre, Botterill et al. (2019) used excitatory or inhibitory Designer receptors exclusively acting on designer drugs (DREADDs) to test the role of MCs on diverse behaviors. The excitatory (hM3Dq-Gq) DREADDs and inhibitory (hM4Di-Gi) DREADDs have two point mutations in the human muscarinic cholinergic receptors that make them insensitive to acetylcholine but very sensitive to clozapine-N-oxide (CNO), which has little effect otherwise. DREADDs were expressed in MCs across the septotemporal axis of both hemispheres by injection of AAV at four sites. The results showed that MC inhibition led to improvement in a number of tasks associated with novelty, object memory, and context; furthermore, MC inhibition appeared to be anxiolytic. It was noted that the anxiolytic effects may have contributed to the improvements in behaviors usually attributed to cognition alone, such as object memory. Specifically, MC inhibition improved novel object recognition tests, contextual fear conditioning tests, novelty suppressed feeding, and the light-dark box task, and there was a weak effect on the elevated plus maze. Notably, effects were often in females or males; only one task showed effects in both sexes. This was a test developed previously (Bernstein et al., 2019) where animals are kept in their home cage while being tested for novel object exploration. The task was designed to help lower stress after it was noticed that the typical length of acclimation used for most behavioral tests did not seem to reduce stress as much as expected. Therefore, animals were allowed to explore novel objects added to the cage with the cagemate present. Use of the home cage prevented social isolation produced by removing an animal to a new cage, a typical way tests are conducted. This task was highly sensitive to MC manipulations.

The study of contextual fear conditioning contrasted with results of Jinde et al. (2012), who reported ablation of MCs, rather than inhibition, impaired contextual discrimination. However, the tasks were quite different, as well as other methods. Jinde et al. (2012) also reported that MC ablation led to effects in the open field task, whereas Botterill et al. (2021b) did not report significant effects of MC manipulations in the open field. The results of Botterill et al. (2021b) were similar to a prior study using chemogenetics by Oh et al. (2019), suggesting ablation leads to very different effects relative to chemogenetic inhibition. Taken together, the data suggest that MC inhibition can benefit the DG in object exploration and memory and can reduce anxiety-like behavior.

The results of the work by Botterill et al. (2021b) also suggest cognitive functions of the DG could be gated by anxiety, and the gate could involve MCs. MCs are suited to act as a gate because MCs project throughout the DG. The idea that anxiety plays a role in cognitive functions of the DG has been raised before (Anacker and Hen, 2017). There is a potential for effects of young adult-born GCs to be mediated by MCs because MCs form the earliest excitatory input to the young GCs (Chancey et al., 2014; Scharfman and Bernstein, 2015). Also, Oh et al. (2019) suggested that MCs are important to the functions of the DG to regulate mood, relevant to depression and its treatment. It is not surprising that the DG plays a role in anxiety and mood regulation, given its role in the adverse effects of behavioral stress (DiSabato et al., 2020).

These data are valuable because of the comorbidities accompanying TLE which include anxiety and depression. These are discussed further below.

How Studies of MCs in Epilepsy Led to Predictions about Their Role in the Disease

How Studies of the Normal Brain Suggested Additional Ways MCs Influence Epileptogenesis and Epilepsy

Comorbidities

Studies of the role of MCs in normal behavior have implications for epilepsy, specifically the comorbidities in TLE. Based on studies of behavior with chemogenetics, the data suggested the intriguing idea that MC inhibition was anxiolytic and improved some types of cognitive tasks (Botterill et al., 2021b). What does this mean for comorbidities in TLE? First of all, there are many comorbidities, and they do include cognitive impairment (Allone et al., 2017; Chauviere, 2020) and anxiety/depression (Kanner et al., 2012; Mikulecka et al., 2019) Fig. 23–5 suggests a relationship between the data from normal animals and epilepsy. If one first considers that normally the GCs are relatively inhibited, in part because MCs activate GABAergic neurons that innervate GCs, why would MC inhibition be anxiolytic and improve some cognitive functions? One possibility is by reducing MC inhibition of GCs there is more activity in GCs, and slightly more activity can increase LTP and GC-dependent behaviors that require memory. But if that is true, what about epilepsy? There MCs are not inhibited chemogenetically, but many are lost. That also would lead to more activity of GCs, and with other circuit changes GCs may be susceptible to hyperexcitability. This hyperexcitability could worsen cognition and anxiety.

Figure 23–5.

A role for MCs in anxiety in the normal brain and patients with epilepsy. A. A diagram of the normal DG circuitry as in the prior figure but more cells are shown. B. Using iDREADDs, Botterill et al. (2021b) inhibited about half of MCs with systemic injection (more...)

Advent of New Methods Using Mice Expressing Cre in MCs and AAV

The Era of Mouse Models of Epilepsy: What It Has Suggested about the Role of MCs

Studies of Chemogenetics in the Pilocarpine Model Supports a Distinct Role of MCs in the Initial Insult and Epileptogenesis Compared to the Influence of MCs in Epilepsy

A subsequent study proposed that both excitatory and inhibitory effects of MCs on GCs play important roles in the context of epilepsy (Fig. 23–6; Botterill et al., 2019). In that study the hypothesis that was proposed suggested that MC excitatory effects on GCs are strong during the initial insult that starts the process of epileptogenesis. However, MC inhibition of GCs dominates the actions of MCs once epilepsy is established. This study does not argue against the dormant basket cell hypothesis but does suggest that there is a time during epileptogenesis when MC excitation of GCs is strong. If true, inhibition of MCs could be helpful in blunting epileptogenesis and activation of MCs would ameliorate chronic seizures. A key difference in the studies of Bui et al. (2018) and Botterill et al. (2019) was that the latter was timed to examine the initial insult at the start of epileptogenesis, when the network starts out normally but then enters a very unusual condition during the initial insult. Because the network begins without cell loss and the circuit changes accompanying epilepsy, effects of MCs appear to be distinct from the epileptic condition.

Figure 23–6.

MCs influence the response to an initial precipitating insult in a mouse model of TLE. A. 1. The timeline of experiments in A–D. Top: Drd2-Cre mice were injected with AAV to express iDREADDs in MCs. On Day 1, iDREADDs were activated before pilocarpine-induced (more...)

The data used experimental SE induced by systemic pilocarpine injection to simulate an initial insult and initiate SE in adult mice. In these experiments, the Drd2-Cre mouse line was used. In these mice, studies were done to determine if any hippocampal GABAergic neurons expressed Cre, and it appeared that they did not (Botterill et al., 2019). In contrast, MCs had robust expression. Adult Drd2-Cre mice were initially anesthetized to inject AAV in the DG encoding either excitatory or inhibitory DREADDs stereotaxically. Afterward electrodes were implanted over two cortical sites and over the dorsal hippocampus of the left and right hemisphere. There were four sites for the injection, one dorsal and one relatively caudal in each hemisphere, the latter to target ventral DG. The aim was to express DREADDs in the most MCs as was possible. To retain specificity, injection volumes were relatively low. To ensure trauma did not inadvertently injure hilar neurons, the gauge of the syringe used for injection was high and the sites of injection were made so that injection would be above the hilus. A slow rate of injection was used to reduce trauma also. These efforts were valuable because the track of the syringe was small. An important part of this procedure was how many MCs expressed Cre. Based on quantitative estimates, about 50% of MCs showed robust expression, and this was distributed mainly across the dorsal 2/3 of the DG. One might think expression of MCs in only 50% of the population would have been a caveat of the study, and it could have been. However, one might find deleterious effects if 100% of all MCs were excited or inhibited by DREADDs at the same time. Another important factor is that the IML axon of MCs were labeled extremely well across the entire septotemporal axis, even though the expression in MC somata suggested only 50% of MC somata expressed DREADDs. The reason why this is important is that a widespread assumption is that DREADDs act by regulating AP firing at the soma, and the change in firing leads to a change in transmitter release. However, DREADDs may act at the terminal to either hyperpolarize it (for inhibitory DREADDs) or depolarize it (for excitatory DREADDs). Those effects could regulate transmitter release independent of the soma. Evidence for this idea comes from making slices and transecting the soma and axons of MCs in the process. Then slices with only the MC axons containing DREADDs were activated by exposing the slice to the DREADD activator CNO. Prominent effects were found but not in controls (Botterill et al., 2019). Therefore, regardless of the effects on the soma, DREADDs could act on terminals of MCs in the IML.

After approximately 4 weeks to allow animals to recover from surgery and for the viral expression to occur, animals were recorded during a baseline period and then induction of SE occurred (Fig. 23–6). Mice were either Cre^+/– or Cre^–/–, and all mice received CNO.

SE lasted for several hours and then frequent generalized spiking occurred in all electrodes overnight and into the next day (Botterill et al., 2019). There also were intermittent seizures, with and without convulsive behavior overnight and during the next day. Therefore, experiments examined not only SE but events during the day after SE. When MCs were inhibited 30 min before pilocarpine administration using a systemic injection of CNO, SE was reduced in the initial 30 minutes, but SE was still quite clear. Later that day a second dose of CNO was administered because the kinetics of CNO in the brain suggests it will be elevated for only a few hours (Botterill et al., 2019). The next day, the mice that were Cre^+/– had fewer seizures than the Cre^–/– mice (Fig. 23–6), suggesting MCs normally excited the network, contributing to seizures after pilocarpine administration, even those that outlasted CNO injection. To examine neuronal loss, mice were examined 3 days after SE, and there was preservation of neurons in the hilus and CA3 in Cre^+/– mice compared to Cre^–/– mice (Fig. 23–6). Those data suggested that fewer seizures during SE and the subsequent days led to less MC excitation of GCs, leading to less excitotoxicity of GC targets in the hilus and CA3.

In the long term, no additional doses of CNO were used. Nevertheless, Cre^+/– mice had reduced chronic seizures compared to Cre^–/– mice (Fig. 23–6). These data suggest that MC inhibition can protect the brain sufficiently during SE and in subsequent days that there is reduced damage and other sequelae which ultimately reduce chronic epilepsy. The data argue that MC inhibition can blunt epileptogenesis, a very important finding and one that has therapeutic implications.

There were several interesting findings in addition to those revealed by the use of inhibitory DREADDs. For example, MC excitation by excitatory DREADDs did not protect against SE or excitotoxicity of hilar and CA3 neurons. However, one might expect that if excitatory DREADDs had the opposite effect of inhibitory DREADDs, they would exacerbate SE, and that did not occur. One reason may be that MCs are so activated during SE that further activation may not have additional effects. Data from hippocampal slices supported this idea (Botterill et al., 2019). For these experiments, MCs were patched and then SE was simulated. Thus, a baseline was recorded and then the slice was exposed to agents that would produce an environment similar to the start of SE. Theoretically, as the agents continued to be present, insight would be gained into SE itself. CNO was added and comparisons were made of Cre^+/– and Cre^–/– mice. The data showed several notable findings. First, MCs that were recorded from Cre^–/– depolarized rapidly during “SE” and did not recover when the agents causing simulated SE were removed. In contrast, MCs expressing iDREADDs in Cre^+/– mice depolarized but recovered. These data are consistent with studies of MCs in rats where intermittent PP stimulation was triggered in slices (Scharfman and Schwartzkroin, 1989, 1990a, 1990b). MCs depolarized during the stimulation and after stimulation continued for some time, and they did not recover (Scharfman and Schwartzkroin, 1989, 1990a, 1990b). However, using a Ca²⁺ chelator in the intracellular recording electrode preserved MCs, suggesting Ca²⁺ entry mediated the excitotoxicity and Ca²⁺ overload led to deterioration of MCs (Scharfman and Schwartzkroin, 1989).

Together our results do not negate the studies of other investigators suggesting that the primary effects of MCs are to inhibit GCs in the normal network. However, during the unusual conditions of a brain insult, MCs may lose this effect. In addition, their excitatory effects on GCs may strengthen. The result would be that MCs exacerbate the injury and promote epileptogenesis (Fig. 23–7). In contrast, during chronic epilepsy MCs may activate residual GABAergic neurons as they do under normal conditions (Fig. 23–7). Their effects may strengthen or MCs may sprout onto surviving GABAergic neurons to compensate for the loss of GABAergic neurons. Due to stronger inhibitory effects, MC could inhibit GCs sufficiently to reduce seizure duration and convulsive behavior. Thus, we propose that there are diverse roles of DG MCs depending on the state of the network (Fig. 23–7). In epilepsy, MCs appear to play distinct roles in the effects of the initial insult, influencing epileptogenesis. However, during chronic epilepsy, their role seems to be the opposite. This hypothesis reconciles some of the discrepancies in published data sets and serves as way to bridge different hypotheses about the role of MCs in the DG.

Figure 23–7.

Different roles of MCs during the initial insult and TLE. A. A diagram of the normal DG circuitry. This diagram is similar to prior figures with a simplified circuit rather than all cells and their connections. B. During the initial insult, MCs appear (more...)

This work was supported by the National Institutes of Health and the New York State Department of Health.

The author declares no relevant conflicts.