Cannabinoid receptor 1-labeled boutons in the sclerotic dentate gyrus of epileptic sea lions

Abstract

Pathology in the dentate gyrus, including sclerosis, is a hallmark of temporal lobe epilepsy, and reduced inhibition to dentate granule cells may contribute to epileptogenesis. The perisomatic-targeting axonal boutons of parvalbumin-expressing interneurons decrease in proportion with granule cells in temporal lobe epilepsy. In contrast, dendrite-targeting axonal boutons of somatostatin-expressing interneurons sprout exuberantly in temporal lobe epilepsy. A third major class of GABAergic interneurons expresses cannabinoid receptor type 1 (CB1) on their terminal boutons, but there is conflicting evidence as to whether these boutons are increased or decreased in temporal lobe epilepsy. Naturally occurring temporal lobe epilepsy in California sea lions, with unilateral or bilateral sclerosis, offers the benefit of neuroanatomy and neuropathology akin to humans, but with the advantage that the entirety of both hippocampi from control and epileptic brains can be studied. Stereological quantification in the dentate gyrus revealed that sclerotic hippocampi from epileptic sea lions had fewer CB1-labeled boutons than controls. However, the reduction in the number of granule cells was greater, resulting in increased CB1-labeled boutons per granule cell in sclerotic hippocampi at temporal levels. This suggests that although CB1-expressing boutons are decreased in sclerotic dentate gyri, surviving cells have enhanced innervation from these boutons in epileptic sea lions.

1. Significance statement

Plasticity of GABAergic circuits in epilepsy may provide clues into mechanisms of seizure initiation and control. Insufficient innervation from GABAergic interneurons in the dentate gyrus may contribute to seizure initiation. Perisomatic-targeting (parvalbumin-expressing) boutons track with the number of granule cells, dendrite-targeting (somatostatin-expressing) boutons sprout copiously, and reports on perisomatic and dendritic-targeting CB1-containing boutons have been variable. The present results demonstrate that CB1-labeled boutons are increased on a per granule cell basis in sclerotic dentate gyri, arguing against CB1-expressing interneurons providing insufficient innervation.

2. Introduction

In temporal lobe epilepsy the dentate gyrus is a region of interest due to pathology in human patients that includes mossy fiber sprouting, hilar cell loss, and relative granule cell survival (in contrast to the often considerable loss of hippocampal pyramidal cells). The low firing rate of granule cells in healthy controls has given rise to the long-standing hypothesis that this ‘dentate gate’ to the epilepsy-prone hippocampus is disrupted in temporal lobe epilepsy (Heinemann et al., 1992; Lothman et al., 1992). There is evidence in patients that granule cells are hyperexcitable (Dietrich et al., 1999; Gabriel et al., 2004; Williamson et al., 1995), inhibition is weakened (Franck et al., 1995; Williamson et al., 1999) and hilar GABAergic interneurons are reduced (Arellano et al., 2004; de Lanerolle et al., 1989).

A previous study found that among numerous anatomical changes in the dentate gyrus, the only correlation with seizure frequency in mice was with decreased interneurons in the subgranular zone, suggesting that loss of these interneurons relates to ictogenesis (Buckmaster et al., 2017). In that study reduced parvalbumin (PV)-labeled interneurons in the subgranular zone did not correlate with seizure frequency. Another study found that although PV labeling of cell bodies is reduced in the sclerotic dentate gyri of epileptic sea lions, their axon terminals remain proportionate with the number of granule cells (Cameron et al., 2019). This contrasts with hilar-dwelling somatostatin-expressing interneurons that are severely depleted in temporal lobe epilepsy (de Lanerolle et al., 1989; Mathern et al., 1995; Robbins et al., 1991; Sloviter, 1987), but axons from somatostatin-expressing interneurons in potentially both CA1 oriens and the dentate hilus sprout profusely into the dentate molecular layer as demonstrated in rodent (Peng et al., 2013; Zhang et al., 2009), sea lion (Buckmaster et al., 2014) and human (Mathern et al., 1995). An additional major interneuron population in the subgranular zone lacks a somatic label but expresses cannabinoid receptor type 1 (CB1) on their axon terminal boutons and targets granule cells from their somata out to their distal dendrites (Yu et al., 2015). The impact of temporal lobe epilepsy on these interneurons is uncertain. Do axons of these cells sprout like dendrite-targeting somatostatin-expressing cells? Are their terminals reduced in proportion to granule cells like perisomatically-targeting PV-expressing boutons? Or are CB1-expressing boutons reduced more than granule cells, in line with a cell type whose reduction might correlate with enhanced seizure frequency?

There is evidence that CB1 receptor binding, its corresponding G-protein activation, and CB1 immunolabeling are reduced in the inner molecular layer of the rat pilocarpine model of temporal lobe epilepsy (Falenski et al., 2007, 2009). Consistent with this there is evidence for reduced inhibition of granule cells by CB1-modulated interneurons in the continuously stimulated hippocampus rat model of temporal lobe epilepsy (Sun et al., 2014). By contrast, another study found more CB1-labeled terminals in the dentate molecular layer of the chronic mouse pilocarpine model, notably with an increased density of CB1 receptors on terminals which made GABAergic symmetric synapses (Karlocai et al., 2011). In tissue resected from human temporal lobe epilepsy patients the density of CB1-labeled terminals which form symmetric synapses in the dentate inner molecular layer was unchanged compared to autopsy controls, while the density of CB1-labeled terminals forming glutamatergic asymmetric terminals was reduced (Ludanyi et al., 2008). A subsequent study using an antibody favoring CB1 on GABAergic terminals found increased CB1-labeled terminals in the dentate molecular layer of tissue resected from temporal lobe epilepsy human patients with sclerotic CA1 versus autopsy controls and non-sclerotic temporal lobe epilepsy patients, a finding they corroborated in the mouse pilocarpine model (Magloczky et al., 2010). Thus, there are reports that CB1-labeled interneuron terminals are decreased, unchanged or increased in epileptic tissue from rodents and humans.

California sea lions (Zalophus californianus) with naturally occurring temporal lobe epilepsy arising from domoic acid excitotoxicity consequent to their exposure to toxic algal blooms provide a unique opportunity to study epilepsy in a large mammalian brain. Like humans, sea lions’ gyrencephalic brains exhibit classic temporal lobe epilepsy pathology including principal and interneuron cell loss, and sprouting by mossy fibers and somatostatin-expressing axons; furthermore, hippocampal sclerosis can be unilateral or bilateral and often includes some degree of granule cell loss (Buckmaster et al., 2014; Silvagni et al., 2005). Sea lions therefore provide a rare opportunity to study the complete extent of both hippocampi in cases of naturally occurring temporal lobe epilepsy that is often unilateral. We took advantage of these features using stereology to quantify the number of CB1-labeled terminals across the septo-temporal extent of control and sclerotic dentate gyri. In the same sea lions, we stereologically quantified the loss of their granule cell targets along the septo-temporal axis of the dentate gyrus. This allowed us to calculate the number of CB1-expressing boutons per granule cell in epileptic versus control sea lions to test for loss in a major interneuron population in the subgranular zone.

3. Materials and methods

3.1. Animals

Stranded California sea lions (Zalophus californianus) from central and northern California were brought to The Marine Mammal Center in Sausalito for diagnosis and treatment, 9 % (~70/year) with domoic acid intoxication (Goldstein et al., 2008). Domoic acid-induced temporal lobe epilepsy in sea lions was reviewed by Paul Buckmaster (2017). Clinical signs of acute domoic acid toxicosis include head weaving, muscle tremors, ataxia, seizures (1–30 in 24 h), status epilepticus and coma (Gulland et al., 2002). Acute symptoms can be continuous for a week, but seizure frequency can increase leading to death, or resolve over a month (Goldstein et al., 2008). However, sea lions under care have developed epilepsy in 16–85 days after admittance with acute toxicosis. Approximately 50 % of sea lions survive the acute syndrome and can appear clinically normal (Scholin et al., 2000; Gulland et al., 2002), but many sustain brain damage and develop chronic domoic acid toxicosis with clinical signs including memory impairment (Cook et al., 2015), maladaptive migration (Thomas et al., 2010), abnormal aggression, cardiac lesions, repetitive behaviors like chewing on a flipper, swimming in tight circles, pacing or rocking, and seizures were observed in 85 % with intervals of hours-weeks (Goldstein et al., 2008). Those successfully treated and released spent time further from shore making shorter, shallower dives and spent less time on shore which suggested less successful foraging (Thomas et al., 2010). Consistent with this, on average 71 % restrand after 4 months (Goldstein et al., 2008), often underweight or emaciated (Thomas et al., 2010), compared to only 0.5 % for cases not involving domoic acid toxicity (Gulland et al., 2002). As such, sea lions with a poor prognosis were euthanized out of compassion. This study used tissue from 13 control and 31 epileptic sea lions collected between 2010 and 2015 for a previous study (Cameron et al., 2019). Control subjects (9 females, 3 males, 1 unknown) included a pup (<1 y), a yearling (1–2 y), a juvenile (2–4 y), a subadult and eight adults (females >5 y, males >8 y) who were euthanized due to infection, trauma, or gunshot wounds. Three were observed to have an acute seizure, but crucially no hippocampal sclerosis was evident, ruling out temporal lobe epilepsy. Sclerotic hippocampi were defined by neuron loss in the dentate gyrus (on average 80 % reduction from controls (Buckmaster et al., 2014)). Similar to human temporal lobe epilepsy, neuron loss in sea lions is less severe in the granule cell layer, but granule cell loss is the most variable among hippocampal principal cells, ranging from 0 % to 80 %. Epileptic sea lions (18 females, 13 males) included six juveniles, ten subadults, and fifteen adults. Twenty-one of the epileptic sea lions had at least one seizure while in treatment, and nine had multiple seizures.

3.2. Tissue preparation

Sea lions were euthanized with an overdose of intravenous pentobarbital and perfused immediately after death through the ascending aorta, first with 0.9 % NaCl for 2 min, followed by 4 % paraformaldehyde in 0.1 M phosphate buffer, pH 7.4 at 4 °C for 30 min. Brains were hemisected, cut coronally into five ~2 cm blocks, post-fixed and cryoprotected (4 % PFA with 30 % sucrose at 4 °C) until equilibrated, ~1 week. Blocks were frozen in isopentane and stored at − 80 °C until cryosectioned at 40 μm on a sliding microtome, then cryopreserved (30 % ethylene glycol with 25 % glycerol in 50 mM phosphate buffer, −20 °C).

3.3. Antibody

The CB1 antibody was previously produced in rabbit by author Ken Mackie against the last 15 amino acids of the intracellular C-terminus of rat CB1, and specificity was reported in CB1 knockout mice (Nyiri et al., 2005). As a major function of CB1 receptors is to restrict neurotransmitter release, it is particularly targeted to presynaptic axons and terminals (Nyiri et al., 2005).

3.4. Immunohistochemistry

Starting with a random section at the temporal pole, every 40th section was processed for CB1 immunohistochemistry. Sections were incubated in 1 % hydrogen peroxide for 2 h to quench endogenous peroxidase, then blocked with 3 % normal goat serum (NGS), 2 % bovine serum albumin (BSA) and 0.3 % triton X-100 in 0.1 M Tris-buffered saline (TBS, pH 7.4) for 1 h. Sections were incubated in primary solution containing anti-CB1 (1:500) with 1 % NGS, 0.2 % BSA, 0.3 % triton in 0.1 M TBS for 7 days at 4 °C to allow penetration into the middle of the section for accurate counting. After rinsing, sections were incubated in secondary solution containing biotinylated anti-rabbit (1:500, Vector Laboratories, BA-1000) with 2 % BSA and 0.3 % triton in TBS for 2 h, rinsed, and the signal amplified with the Vectastain Elite ABC kit (1:500, Vector Laboratories, PK-6100) in the same diluent for 2 h. Sections were rinsed and developed in 0.02 % diaminobenzidine (Sigma-Aldrich, D5905) with 0.04 % ammonium chloride, 0.015 % glucose oxidase and 0.1 % β-D-glucose in 0.1 M Tris buffer, pH 7.4. Finally, sections were rinsed, mounted, lightly Nissl-stained with 0.25 % thionin to visualize the granule cell layer and coverslipped. Images were taken on a Zeiss Axiophot.

3.5. Stereology

A 1-in-40 series (approximately 10 sections per dentate gyrus) was evaluated by the optical fractionator method (West et al., 1991) by a blinded researcher. To estimate the number of CB1-labeled boutons, the granule and molecular layers were outlined under a 10x objective using Stereo Investigator (MBF Bioscience). Boutons at least ~0.5 μm in diameter were counted in 5 × 5 μm counting frames on a 300 × 300 μm grid using the 100x objective. Dissector height was the full thickness of the section and boutons not cut at the superficial surface were counted. On average, 292 boutons per dentate gyrus were counted. The mean coefficient of [within sea lion] error (0.097) was less than an eighth of the coefficient of variation across all sea lions (0.78) indicating that less than an eighth of group variance was due to the within animal estimation procedure, confirming sufficient sampling. Granule cells from the same hippocampi were previously quantified by the optical fractionator method (Cameron et al., 2019). For this study, boutons and cells were determined to be in the septal or temporal half of the hippocampus based on their position in the coronal sections. In sections where the hippocampus curves to include septal and temporal hippocampus, half the dentate gyrus was allocated septally, the other temporally.

3.6. Statistical procedures

Statistical analyses were performed using Sigmaplot 12.0. Parametric and nonparametric tests were used as appropriate, so averages are given as mean ± standard error or median with a 25–75 % confidence interval, and graphed accordingly.

4. Results

4.1. Granule cells were reduced across the extent of sclerotic hippocampi

Sea lion hippocampi were previously characterized as sclerotic based on evidence of neuron loss in the dentate gyrus upon Nissl staining: 14 subjects had bilateral sclerosis, 17 had unilateral sclerosis (12 on the right, 5 on the left;(Cameron et al., 2019)). While sclerotic hippocampi were defined by neuron loss in the dentate gyrus (Fig. 1A,B), the entire sclerotic sea lion hippocampus was generally shrunken, as in humans, consistent with neuronal loss being substantial across the septotemporal extent of all fields: dentate gyrus, CA3, CA1 (Buckmaster et al., 2014). In these sea lions, the number of granule cells was reduced by 73 % from 2.3 ± 0.07 million per nonsclerotic (control & unilateral) dentate gyrus to an average of 0.6 ± 0.09 million per sclerotic (unilateral and bilateral) dentate gyrus (Cameron et al., 2019). Furthermore, the number of granule cells in controls, 2.3 ± 0.08 million, was similar to unilaterally nonsclerotic dentate gyri, 2.3 ± 0.11 million, and the number of granule cells in unilaterally sclerotic dentate gyri, 0.7 ± 0.14 million, was similar to bilaterally sclerotic dentate gyri, 0.5 ± 0.10 million (Cameron et al., 2019). The granule cell loss in sea lions is consistent with reduced granule cells in humans with sclerotic temporal lobe epilepsy (Babb et al., 1989; Blumcke et al., 2007; Kim et al., 1990; Mathern et al., 1995; Sass et al., 1990), but is generally inconsistent with rodent models of temporal lobe epilepsy (Buckmaster and Lew, 2011; Thind et al., 2010).

Fig. 1.

Nissl stain demonstrates unilateral hippocampal sclerosis in a sea lion with temporal lobe epilepsy. (A) Numerous cell bodies in the hilar polymorphic layer (H) and the dense granule cell layer (arrowheads) indicate this dentate gyrus in an epileptic sea lion is nonsclerotic. (B) The contralateral dentate gyrus is smaller with few cells labeled in the hilus (H) and partial cell loss in the granule cell layer (arrowheads), therefore this sea lion was classified as being unilaterally sclerotic. (C) Plot of the number of granule cells counted along the length of the septo-temporal axis the hippocampi from control sea lions and sea lions with unilateral or bilateral sclerosis. (D) The number of granule cells in both the septal and the temporal halves were reduced in sclerotic hippocampi relative to controls. *p < 0.001 v. control and nonsclerotic (one-way ANOVA); n = hippocampi from 13 control, 17 unilaterally sclerotic, and 14 bilaterally sclerotic sea lions.

In the current study we analyzed the number of granule cells across the septo-temporal axis of the dentate gyrus. A severe reduction in granule cells in bilaterally sclerotic hippocampi relative to control hippocampi was evident across the length of the dentate gyrus (Fig. 1C). In sea lions with unilateral sclerosis, the number of granule cells along the sclerotic side resembled bilaterally sclerotic hippocampi, and the nonsclerotic side shadowed controls. Analysis of the septal and temporal halves confirmed stark granule cell loss in sclerotic hippocampi at both poles (Fig. 1D). At the septal pole, unilaterally sclerotic and bilaterally sclerotic hippocampi had less than half of the granule cells in control and unilaterally nonsclerotic hippocampi (Table 1). At the temporal pole unilaterally sclerotic and bilaterally sclerotic hippocampi had less than a quarter of the granule cells in control and unilaterally nonsclerotic hippocampi (p < 0.001, one-way ANOVA).

Table 1.

The number of granule cells in the septal and temporal halves of the dentate gyrus.

Granule cells (millions):	Septal	Temporal
Control	1.21 ± 0.07	1.07 ± 0.06
Unilaterally Nonsclerotic	1.27 ± 0.07	1.07 ± 0.06
Unilaterally Sclerotic	0.48 ± 0.09a	0.25 ± 0.07b
Bilaterally Sclerotic	0.38 ± 0.07a	0.13 ± 0.03b

^a,bp < 0.001 v. control, unilaterally nonsclerotic (one-way ANOVA, Holm-Sidak method of comparison). n = hippocampi from 13 control, 17 unilaterally sclerotic, and 14 bilaterally sclerotic sea lions

4.2. CB1 labeling was reduced in sclerotic dentate gyri

The CB1 antibody used was selected in order to investigate changes to CB1-expressing GABAergic boutons. Previous studies with the L15 antibody found chromogenic labeling in a pattern consistent with interneuron boutons, and examination at the ultrascopic level confirmed CB1-labeling in symmetric (GABAergic) axon terminals but not asymmetric (glutamatergic) terminals following pre-embedding immunolabeling in macaque (Eggan and Lewis, 2007) and post-embedding labeling in rat hippocampus (Nyiri et al., 2005). In this study immunostaining for brightfield microscopy labeled a plexus of axon terminals in the dentate gyrus (Fig. 2). However we were unable to optimize immunofluorescent labeling to test for colocalization of the L15 antibody with GABAergic or glutamatergic markers: with immunofluorescent processing the antibody labeled boutons unevenly making bouton delineation difficult, punctate background labeling could not be eliminated, and autofluorescence was particularly significant in sclerotic dentate gyri. Thus, although we cannot show that the L15 CB1 antibody definitively labels GABAergic boutons exclusively, previous work indicates that the vast majority of boutons are GABAergic.

Fig. 2.

CB1 labeling is decreased in the sclerotic dentate gyrus of epileptic sea lions. (A,B) CB1 labels axonal boutons in the granule cell layer (G, demarcated with light Nissl stain), inner (I), middle (M) and outer (O) molecular layers of the dentate gyrus (DG) in a control sea lion. Note the subtly enhanced labeling near the border of the inner and middle molecular layers. (C,D) In an epileptic sea lion with bilateral sclerosis (Bilat) CB1-labeling is reduced overall. (E,F) CB1-labeling in the non-sclerotic dentate gyrus (Uni-Nonscler) of an epileptic sea lion with unilateral sclerosis is similar to the control. (G,H) CB1-labeling in the contralateral sclerotic dentate gyrus (Uni-Scler) of the same sea lion is reduced by comparison. Scale bars in panel A apply to panels C, E, G; scale bars in panel B apply to panels D, F and H.

In humans and rodents CB1-immunoreactive boutons are profuse in the molecular layer, with the highest density in the inner molecular layer adjacent to the granule cell layer (Ludanyi et al., 2008; Magloczky et al., 2010; Steindel et al., 2013). In control and unilaterally non-sclerotic sea lion hippocampi, the CB1 antibody labeled punctate boutons in the granule cell layer; the number of puncta proliferated in the inner and middle molecular layers, then diminished out to the edge of the outer molecular layer (Fig. 2A,B,E,F). This contrasts somewhat with a more densely labeled outer molecular layer in macaques (Eggan and Lewis, 2007). A band at the border of the inner and middle molecular layers was frequently conspicuously labeled at mid-levels along the septo-temporal axis, often more prominently than the inner molecular layer (Fig. 2A,B). In sclerotic dentate gyri, CB1 labeling was usually decreased relative to nonsclerotic dentate gyri (Fig. 2C,D,G,H), although occasionally appeared stronger. A pronounced band at the inner/middle molecular layer border was somewhat less common in sclerotic than nonsclerotic hippocampi: about half the time labeling in the inner molecular layer remained predominant throughout the extent of the sclerotic dentate gyrus (Fig. 2G,H).

High magnification of the inner molecular layer clearly illustrates the punctate labeling of axon boutons in the nonsclerotic dentate gyrus (Fig. 3A,C), and fewer labelled boutons in sclerotic dentate gyri (Fig. 3B, D). Stereological quantification of the dentate granule cell and molecular layers found that the median number of CB1-expressing boutons for all nonsclerotic dentate gyri was 194 million (Table 2). For all sclerotic dentate gyri the median number of CB1-containing boutons was halved (p = 0.005, Mann-Whitney).

Fig. 3.

Decreased CB1-labeled puncta in the inner molecular layer (iml) of sclerotic dentate gyri from sea lions with temporal lobe epilepsy. (A) CB1 labels interlaced punctate terminal boutons in the inner molecular layer of a control sea lion. Inset shows a montage of an axon in the inner molecular layer with terminal swellings. (B) Labeling of CB1 boutons is reduced in the inner molecular layer of a sea lion with bilateral sclerosis (Bilat). Cell bodies are labeled indigo by Nissl staining. Inset shows a montage of an axon and its boutons in the inner molecular layer. (C) A dense network of CB1-labeled boutons and axon fibers is evident in the inner molecular layer of the nonsclerotic dentate gyrus (Uni-Nonscler) from an epileptic sea lion with unilateral sclerosis. Inset shows a montage of an axon and its boutons in the inner molecular layer. (D) In the contralateral sclerotic dentate gyrus (Uni-Scler) labeling of CB1 boutons in the inner molecular layer more closely resembles labeling in the sea lion with bilateral sclerosis. Inset shows a montage of a CB1-labeled axon and its boutons in the inner molecular layer. Scale bars in panel A apply to panels B, C and D. (E) Stereological quantification of the number of CB1-labeled boutons across the entire extent of the granule cell and molecular layers of the dentate gyrus with box plots for the median and 25–75 percentiles. *p = 0.05 v. control (ANOVA on ranks, Dunn’s Method of comparison). For controls and epileptic sea lions with bilateral sclerosis (Bilat) the averaged value for the left and right dentate gyrus are plotted. For epileptic sea lions with unilateral sclerosis, the nonsclerotic (Uni-Norm) and sclerotic (Uni-Scler) dentate gyri are paired. (F) Plot of the number of CB1-labeled terminals (CB1-T) quantified along the septo-temporal length of the dentate gyrus. (G) Box plots of the median number (and 25–75 % confidence interval) of CB1-labeled boutons in the septal and temporal halves of control and epileptic dentate gyri. Asterisks indicate significant reductions relative to controls: p < 0.05 (ANOVA on ranks, Dunn’s Method of comparison). n = hippocampi from 13 control, 17 unilaterally sclerotic, and 14 bilaterally sclerotic sea lions.

Table 2.

The number of CB1 boutons in the granule cell and molecular layers of the dentate gyrus.

CB1-T (millions):	Whole HC	Septal	Temporal
Nonsclerotic	194 (106–377)
Control	211 (133–361	112 (74–179)	103 (67–191)
Uni. Nonscler.	167 (61–411)	70 (34–233)	102 (29–174)
Sclerotic	88 (50–185)a
Uni. Scler.	108 (60–215)	55 (32–91)	43 (19–94)
Bilat. Scler.	60 (48–134)b	35 (22–78)c	31 (25–49)d

Values are given as median (25–75 percentiles).

^ap = 0.005 v. Nonsclerotic (Mann-Whitney).

^b,c,dp < 0.05 v. Control (ANOVA on ranks, Dunn’s method of comparison).

Abbreviations: T: terminals, HC: hippocampus, Uni: unilaterally, Nonscler: nonsclerotic, Scler: sclerotic, Bilat: bilaterally. n = hippocampi from 13 control, 17 unilaterally sclerotic, and 14 bilaterally sclerotic sea lions.

Breaking down the groups (Fig. 3E), in the granule cell and molecular layers of control sea lions the median number of CB1-labeled boutons was 211 million. In sea lions with bilateral sclerosis, CB1-immunoreative boutons were decreased to only 28 % of controls (Table 2). In sea lions with unilateral sclerosis, the number of CB1-labeled boutons on the sclerotic side was decreased to 65 % of the nonsclerotic side. Analysis of the four groups found that only the number of CB1-labeled boutons in bilaterally sclerotic dentate gyri was significantly reduced from controls (p < 0.05, ANOVA on ranks, Dunn’s Method of comparison). As is evident from the graph, for select unilaterally sclerotic pairs (5/17) the sclerotic dentate gyrus had more CB1 boutons than the nonsclerotic dentate gyrus.

Analysis of CB1-labeled boutons along the length of the dentate gyrus revealed a decrease at all levels in bilaterally sclerotic hippocampi compared to controls (Fig. 3F). In sea lions with unilateral sclerosis the number of CB1-labeled boutons along the nonsclerotic dentate gyrus closely tracked the number in controls, while the sclerotic side showed a tendency toward reduction, especially at temporal levels. The septal and temporal halves were therefore analyzed (Fig. 3G). At the septal pole, the number of CB1-labeled boutons in the granule cell and molecular layers of bilaterally sclerotic hippocampi was a third of controls (p < 0.05), while in sea lions with unilateral sclerosis the values on the sclerotic side trended lower but was not significantly different from controls or the nonsclerotic side (Table 2, ANOVA on ranks, Dunn’s Method of comparison). Similarly, at the temporal pole CB1-labeled boutons in bilaterally sclerotic dentate gyri were sharply decreased to 30 % of controls (p < 0.05) while the unilaterally nonsclerotic and sclerotic sides were not significantly different from controls or each other (ANOVA on ranks, Dunn’s method of comparison).

4.3. The ratio of CB1 boutons per granule cell was increased in sclerotic dentate gyri

Parvalbumin-expressing boutons were drastically reduced in the sclerotic dentate gyrus of epileptic sea lions (by roughly 75 %), however, this was on a similar scale as the reduction in granule cells, suggesting that the degree of innervation was maintained (Cameron et al., 2019). To understand the loss of CB1-positive boutons in the context of remaining granule cells, the number of CB1 boutons per granule cell was calculated for each sea lion. As a group, nonsclerotic dentate gyri had a median of 84 CB1-immunoreactive boutons per granule cell (Table 3). By comparison in sclerotic dentate gyri the number of CB1-labeled boutons per granule cell was increased more than two and a half times (p < 0.001, Mann-Whitney).

Table 3.

The number of CB1 boutons per granule cell in the dentate gyrus.

CB1-T /GC:	Whole HC	Septal	Temporal
Nonsclerotic	84 (41–152)
Control	102 (60–152)	99 (51–205)	117 (73–168)
Uni. Nonscler.	78 (33–155)	58 (41–149)	93 (28–165)
Sclerotic	217 (111–566)a
Uni. Scler.	237 (100–516)b	177 (81–451)c	298 (170–708)d
Bilat. Scler.	180 (107–574)b	160 (75–375)c	361 (161–1336)d

Values are given as median (25–75 percentiles).

^ap < 0.001 v. Nonsclerotic (Mann-Whitney).

^b,cp < 0.05 v. Uni. Nonscler. (ANOVA on ranks, Dunn’s method of comparison).

^dp < 0.05 v. Control and Uni Nonscler. (ANOVA on ranks, Dunn’s method of comparison).

Abbreviations: T: terminals, GC: granule cell, HC: hippocampus, Uni: unilaterally, Nonscler: nonsclerotic, Scler: sclerotic, Bilat: bilaterally. n = hippocampi from 13 control, 17 unilaterally sclerotic, and 14 bilaterally sclerotic sea lions.

In particular, control sea lions had a median of 102 CB1-positive boutons per granule cell (Table 3, Fig. 4A). On average, sea lions with sclerosis trended toward more CB1-immunoreactive boutons per granule cell than controls (76 % increase for the bilateral group, 2.3 fold increase for the unilateral group), however, they were not significantly different from controls (ANOVA on ranks). Sea lions with unilateral sclerosis had triple the number of CB1-expressing boutons per granule cell on the sclerotic side as the contralateral nonsclerotic side, which was significant (p < 0.05, ANOVA on ranks with Dunn’s method of comparison). In sea lions with unilateral sclerosis the number of CB1-containing boutons per granule cell was always increased on the sclerotic versus the nonsclerotic side (ranging from 7 % to a factor of 27 greater, with a median factor of 2.8).

Fig. 4.

Calculation of CB1-labeled boutons per granule cell in epileptic and control sea lions. (A) For each individual, the number of CB1-labeled boutons was divided by their previously determined number of granule cells (GC; (Cameron et al., 2019). Box plots for the median and 25–75 percentiles are superimposed over individual values. Bilat: bilaterally sclerotic; Uni-Norm: unilaterally nonsclerotic; Uni-Scler: unilaterally sclerotic (B) The number of CB1-labeled terminals (CB1-T) per granule cell was calculated across the septo-temporal length of the dentate gyrus. (C) Box plots of the median and 25–75 % confidence intervals reveal that the number of CB1-labeled boutons per granule cell was increased in the temporal half of sclerotic dentate gyri relative to controls. *p < 0.05 v. control and unilaterally non-sclerotic dentate gyri (ANOVA on ranks, Dunn’s method of comparison); n = hippocampi from 13 control, 17 unilaterally sclerotic, and 14 bilaterally sclerotic sea lions.

The stereological quantification allowed comparison of the number of CB1-labeled boutons per granule cell along the length of the hippocampus (Fig. 4B). The ratio of CB1-labeled boutons per granule cell was fairly constant throughout the dentate gyrus in control and unilaterally non-sclerotic hippocampi. At septal levels the values for all four groups were fairly similar, but they diverged temporally. This was borne out in analysis of the septal and temporal halves (Fig. 4C). At the septal pole, the number of CB1-labeled boutons per granule cell in controls was not significantly different from either the unilaterally nonsclerotic dentate gyri, or the unilaterally or bilaterally sclerotic dentate gyri (Table 3; ANOVA on ranks). At the temporal pole however, the ratio of CB1-labeled boutons per granule cell was significantly increased in unilaterally and bilaterally sclerotic dentate gyri relative to both controls and unilaterally nonsclerotic dentate gyri (Table 3, Fig. 4C, p < 0.05, ANOVA on ranks with Dunn’s Method of comparison).

4.4. Correlation between granule cell loss and CB1-expressing bouton number

Granule cell loss is one of the most variable aspects of hippocampal pathology measured in sea lions with temporal lobe epilepsy (Buckmaster, 2017; in this study sclerotic hippocampi have 2–75 % of the control mean). The number of CB1-labeled boutons in sclerotic dentate gyri is also variable, ranging from 13 % to 146 % of controls (Fig. 3E). Therefore we conducted correlation analyses to determine whether there was any predictive value between granule cell loss and the number of CB1-labeled boutons. Indeed, for all dentate gyri, the number of granule cells was positively correlated with the number of CB1-labeled boutons (Spearman rank order correlation coefficient = 0.55, p < 0.0001; Fig. 5A). Sclerotic dentate gyri were on the lower end of each measure.

Fig. 5.

Correlations between granule cell loss and CB1-labeled boutons in nonsclerotic and sclerotic dentate gyri. (A) The number of CB1-labeled boutons per dentate gyrus was positively correlated with the number of granule cells in both (control and unilaterally) nonsclerotic and (unilaterally and bilaterally) sclerotic dentate gyri (p < 0.0001, Spearman rank order correlation). (B) In nonsclerotic dentate gyri the ratio of CB1-labeled boutons per granule cell was positively correlated with the number of CB1-labeled boutons per dentate gyrus (p < 0.0001, Spearman rank order correlation). However, in sclerotic hippocampi there was no correlation as many hippocampi had abnormally high ratios of CB1-labeled boutons per granule cell (p = 0.2, Spearman rank order correlation). (C) In the septal halves of nonsclerotic hippocampi there was no correlation between the ratio of CB1-labeled boutons per granule cell and the number of granule cells as the former remains in a narrow range despite variability in the latter (p = 0.5, Spearman rank order correlation). However, in the septal halves of sclerotic hippocampi there is a negative correlation: progressive granule cell loss is correlated with increased CB1-labeled boutons per granule cell (p < 0.0001, Spearman rank order correlation). (D) Similar to their septal counterparts, the ratio of CB1-labeled boutons per granule cell is not correlated with the number of granule cells in the temporal halves of sclerotic hippocampi (p = 0.6, Spearman rank order correlation). However, in the temporal halves of sclerotic hippocampi, the highest ratios of CB1-labeled boutons per granule cell are found in the hippocampi with the most severe granule cell loss (p < 0.0001, Spearman rank order correlation). n = individual hippocampi from 13 control, 17 unilaterally sclerotic, and 14 bilaterally sclerotic sea lions.

Like the loss of granule cells and the change in CB1-labeled boutons, their ratio in sclerotic dentate gyri was also quite variable (ranging from 0.6 to 20 times the control median, Fig. 4A). In nonsclerotic hippocampi, the ratio of CB1-labeled boutons per granule cell scaled up with the number of CB1-labeled boutons rather than remaining constant (Spearman rank order correlation coefficient = 0.94, p < 0.0001; Fig. 5B), suggesting that in dentate gyri with more CB1-expressing boutons, granule cells are more highly innervated by them. By contrast, the correlation was lost in nonsclerotic hippocampi (Spearman rank order correlation coefficient = 0.21, p = 0.2) where numerous dentate gyri had an unusually high ratio of CB1-labeled boutons per granule cell (Fig. 5B). Although the highest numbers of CB1-containing boutons were in nonsclerotic dentate gyri, sclerotic dentate gyri had the highest ratios of CB1-containing boutons on a per granule cell basis.

This is further explained by examination of the relationship between the ratio of CB1-expressing boutons per granule cell and the number of granule cells. In nonsclerotic hippocampi at the septal (Fig. 5C) and temporal poles (Fig. 5D), the ratio of CB1-expressing boutons per granule cell remained relatively consistent despite variability in the number of the granule cells, and the two factors were uncorrelated (Spearman rank order correlation coefficient = −0.1 (septal, p = 0.5), 0.1 (temporal, p = 0.6)). However, in sclerotic hippocampi they were negatively correlated. At the septal pole, dentate gyri with progressively severe granule cell loss correlated with an increased ratio of CB1-labeled boutons per granule cell (Spearman rank order correlation coefficient = −0.7, p = 0.0001). And at the temporal pole, dentate gyri with increased ratios of CB1-labeled boutons per granule cell were concentrated in hippocampi with the most granule cell loss (Spearman rank order correlation coefficient = −0.6, p = 0.0001). This suggests that sprouting of CB1-expressing axons increased with the severity of granule cell loss even though sclerotic dentate gyri had fewer CB1-expressing boutons overall.

All in all, despite the decrease in CB1-labeled boutons in sclerotic dentate gyri (45 % on average), the granule cell loss (73 % on average; (Cameron et al., 2019) exceeded the reduction in CB1-labeled terminals). Thus, in sclerotic hippocampi the number of CB1-labeled boutons normalized to the remaining granule cells was 2.6 times that of non-sclerotic dentate gyri. Septo-temporal analysis indicated that the changes in ratio were especially pronounced at temporal levels where the median was increased 3 fold in sclerotic dentate gyri versus 1.8 fold at septal levels. Correlation analyses found that the number of CB1-expressing boutons was most disproportionately high in epileptic sea lions with the most severe granule cell loss.

5. Discussion

The principal finding of this study is that although the number of CB1-labeled boutons of putative interneurons was reduced in the sclerotic hippocampi of epileptic sea lions relative to controls, the remaining number of CB1-labeled boutons per granule cell was actually increased. This describes a different response than parvalbumin-expressing basket and chandelier cells in the epileptic circuitry of the dentate gyrus: labeling of these perisomatic-targeting cells is decreased more than granule cells, but their axons maintain proportionate innervation (Cameron et al., 2019). The finding for CB1-labeled boutons is somewhat more in line with the prolific sprouting of dendritic-targeting somatostatin-expressing boutons into the dentate gyrus molecular layer.

In this study where the septo-temporal location of granule cells was determined from coronal sections, the reduction in sclerotic hippocampi was evident along the extent of the hippocampus, though stronger at temporal levels. In a similar study where the hippocampus was isolated and sectioned transversely, granule cell loss was only significant at temporal levels (Buckmaster, 2014). Nevertheless, these findings are in line with our ultimate conclusion that the ratio of CB1-labeled boutons per granule cell is particularly elevated at the temporal pole. Interestingly, the temporal hippocampus is the most frequent seizure onset zone in the rat pilocarpine model (Toyoda et al., 2013; Wyeth et al., 2020) and in humans (King et al., 1997; King and Marsan, 1977; Masukawa et al., 1995) where the changes in this study were most pronounced.

The majority of CB1 mRNA is expressed on GABAergic neurons with low expression on glutamatergic neurons (Marsicano and Lutz, 1999) even though there is evidence that CB1 receptors are more efficiently coupled to G-proteins in glutamatergic neurons than GABAergic neurons (Steindel et al., 2013). Only a quarter of hippocampal CB1 protein is estimated to reside in glutamatergic neurons (Steindel et al., 2013). Immuno-electron microscopy indicated that the antibody used here preferentially labels GABAergic terminals forming symmetric synapses in rat hippocampus and macaque neocortex, but not glutamatergic terminals forming asymmetric synapses (Eggan and Lewis, 2007; Nyiri et al., 2005). Immunohistochemistry in mice with CB1 conditionally knocked out in GABAergic or glutamatergic cells confirmed that the vast majority of terminals labeled by this antibody are GABAergic, with only faint labeling of glutamatergic terminals in the inner molecular layer (Steindel et al., 2013). We were unable to confirm the specificity for GABAergic boutons with immunofluorescent colocalization due to what we deemed are technical shortcomings. Immunofluorescence employs a different secondary labeling and amplification process than chromogenic staining. With brightfield microscopy CB1-labeled boutons were clearly defined, but with fluorescent microscopy partial CB1-labeling of boutons was difficult to discern from background labeling, and autofluorescence further interfered. This is not unprecedented as demonstrating immunofluorescent colocalization for GABAergic markers with another GABAergic selective CB1 antibody proved similarly challenging (Szabo et al., 2014). Thus, although we cannot rule out that some boutons labeled with the L15 antibody in our study are glutamatergic, it is reasonable to conclude that the large majority are GABAergic. Moreover, immunopositive glutamatergic terminals likely derive from hilar mossy cells which express CB1 message (Monory et al., 2006) and functional protein (Jensen et al., 2021; Uchigashima et al., 2011), but are some of the most vulnerable cells in temporal lobe epilepsy. Therefore, in sclerotic hippocampi CB1-labeled terminals are even more likely to be GABAergic, while in nonsclerotic hippocampi the number of CB1-labeled interneuron terminals may be slightly overstated. Weak CB1 labeling of glutamatergic terminals in the nonsclerotic inner molecular layer would further emphasize the present finding of increased GABAergic CB1-expressing terminals per granule cell in the sclerotic dentate gyrus.

CB1 labels the boutons of total molecular layer (TML) cells and hilar commissural-associational pathway-associated (HiCAP) cells. Both have cell bodies in the subgranular zone, but TML axons innervate granule cell dendrites across the depth of the molecular layer and HiCAP boutons concentrate in the granule cell and inner molecular layers (Hefft and Jonas, 2005; Yu et al., 2015). The increased number of CB1-labeled boutons per granule cell in sclerotic hippocampi found here is in general agreement with anatomical studies in human and mouse that found unchanged or increased density of CB1-expressing GABAergic terminals in the dentate gyri (Karlocai et al., 2011; Ludanyi et al., 2008; Magloczky et al., 2010). HiCAP cells express CCK in addition to CB1, and there is anatomical and functional evidence that labeling of their cell bodies (Khan et al., 2018; Sayin et al., 2003) but see (Sun et al., 2007) and boutons (Sayin et al., 2003; Sun et al., 2014) are reduced in epileptic rats (but see (Sloviter, 1987)). This raises the possibility that CB1/CCK-expressing HiCAP boutons may be decreased in the epileptic dentate gyrus, while CB1-expressing TML boutons sprout on a per granule cell basis. TML cells are currently identified molecularly by their expression of axonal CB1 in the absence of CCK, so no somatic label is known. We were interested in calculating the number of TML boutons in our tissue through subtraction of CCK-labeled terminals, however in addition to the expected band of CCK-expressing boutons in the inner molecular layer, this was hindered by a second prominent CCK-labeled band in the middle molecular layer that has been visualized in other species (Gall, 1990; Lotstra and Vanderhaeghen, 1987; Sayin et al., 2003; Schwarzer et al., 1995; Stengaard-Pedersen et al., 1983; Sun et al., 2014), and may belong to terminals from the medial perforant path (Fredens et al., 1984).

HiCAP cells provide dense perisomatic innervation to granule cells and release a barrage of asynchronous GABA that produces prolonged inhibition even beyond presynaptic activity (Hefft and Jonas, 2005). TML cells are perhaps the most abundant interneuron in the subgranular zone (Mott et al., 1997; Yu et al., 2015). Thus, axonal sprouting of either CB1-expressing interneuron could have a substantial effect and may be interpreted to reflect a compensatory effort to increase GABAergic innervation of granule cells in a hyperexcitable circuit.

However, functional studies indicate decreased receptor binding and CB1-mediated G-protein activity in the inner molecular layer (Falenski et al., 2007) and reduced inhibition in dentate granule cells of epileptic rats (Sayin et al., 2003) from CB1-expressing terminals (Sun et al., 2014). Thus, the functionality of CB1-expressing interneurons may be impaired despite their increased number relative to granule cells. An additional caveat is that although the number of CB1-labeled boutons per granule cell was increased in sclerotic hippocampi, the ultimate targets of sprouted CB1 boutons may not be granule cells, but rather mossy cells (Acsady et al., 2000) or parvalbumin-expressing cells (Medrihan et al., 2017) which are already targets in controls. However, CB1-sensitive inhibition of parvalbumin-expressing cells is decreased in the epileptic dentate gyrus (Yu et al., 2016a), perhaps to reduce disinhibition. Finally, we cannot exclude the possibility that CB1-expressing interneurons increase their innervation of each other, akin to the synaptic strengthening seen in a rat model of temporal lobe epilepsy one week post status epilepticus (Yu et al., 2016b). That instance would be disadvantageous, enhancing disinhibition of granule cells and making the network more permissive to seizures.

As compensatory measures, CB1 (and somatostatin) interneuron sprouting seems unable to perpetually hold an epileptic network in check. However, it has been suggested that CCK-expressing cells, and arguably TML cells, may be a fruitful therapeutic target given evidence that they underperform in epileptic networks and their temporal imprecision may not affect network rhythms underlying cognitive function (Dudok et al., 2022). Nevertheless, the relative increase in CB1-labeled boutons is not consistent with a cell type whose loss correlates with increased seizure frequency as originally hypothesized in this study.