Glucagon-like peptide-1 receptor differentially controls mossy cell activity across the dentate gyrus longitudinal axis

Abstract

Understanding the role of dentate gyrus (DG) mossy cells (MCs) in learning and memory has rapidly evolved due to increasingly precise methods for targeting MCs and for in vivo recording and activity manipulation in rodents. These studies have shown MCs are highly active in vivo, strongly remap to contextual manipulation, and that their inhibition or hyperactivation impairs pattern separation and location or context discrimination. Less well understood is how MC activity is modulated by neurohormonal mechanisms, which might differentially control the participation of MCs in cognitive functions during discrete states, such as hunger or satiety. In this study, we demonstrate that glucagon-like peptide-1 (GLP-1), a neuropeptide produced in the gut and the brain that regulates food consumption and hippocampal-dependent mnemonic function, might regulate MC function through expression of its receptor, GLP-1R. RNA-seq demonstrated that most, though not all, Glp1r in hippocampal principal neurons is expressed in MCs, and in situ hybridization revealed strong expression of Glp1r in hilar neurons. Glp1r-ires-Cre mice crossed with Ai14D reporter mice followed by co-labeling for the MC marker GluR2/3 revealed that almost all MCs in the ventral DG expressed Glp1r and that almost all Glp1r-expressing hilar neurons were MCs. However, only ~60% of dorsal DG MCs expressed Glp1r, and Glp1r was also expressed in small hilar neurons that were not MCs. Consistent with this expression pattern, peripheral administration of the GLP-1R agonist exendin-4 (5 μg/kg) increased cFos expression in ventral but not dorsal DG hilar neurons. Finally, whole-cell patch-clamp recordings from ventral MCs showed that bath application of exendin-4 (200 nM) depolarized MCs and increased action potential firing. Taken together, this study adds to known MC activity modulators a neurohormonal mechanism that may preferentially affect ventral DG physiology and may potentially be targetable by several GLP-1R pharmacotherapies already in clinical use.

INTRODUCTION

The use of cutting-edge tools for recording and manipulating neurons has led to a rapid expansion in understanding the in vivo activity and functional roles of dentate gyrus (DG) mossy cells (MCs). Mossy cells are glutamatergic neurons with cell bodies in the DG hilus that provide both monosynaptic feedforward excitation and disynaptic feedforward inhibition of granule cells (H. E. Scharfman, 2016; H. E. Scharfman & Myers, 2012). Recording or imaging activity in vivo revealed that MCs are more active than granule cells, have multiple place fields, and strongly remap their spatial activity in response to contextual changes (Danielson et al., 2017; GoodSmith et al., 2017; GoodSmith et al., 2022; GoodSmith, Lee, Neunuebel, Song, & Knierim, 2019; Jung et al., 2019; Senzai & Buzsaki, 2017). Mossy cell ablation, inhibition, or excitation has shown their involvement in not only cognitive processes such as pattern separation and novelty detection that are critical for contextual and spatial memory (Bauer et al., 2021; Bui et al., 2018; Fredes et al., 2021; Jinde et al., 2012; X. Li et al., 2021), but also in a diverse repertoire of other behaviors and processes, including anxiety and avoidance (Botterill, Vinod, et al., 2021; Wang et al., 2021), neurogenesis (Oh et al., 2020; Yeh et al., 2018), and food intake (Azevedo et al., 2019).

This strong foundation establishing that carefully tuned MC activity is necessary for aspects of episodic memory supports the need for a more complete understanding of how neuromodulatory systems regulate MC excitability. Previous work has shown regulation of MC activity by monoamine (Etter & Krezel, 2014; Oh et al., 2020), glucocorticoid (Patel & Bulloch, 2003), and cannabinoid receptors (Chiu & Castillo, 2008; Hofmann, Nahir, & Frazier, 2006; Monory et al., 2006), in addition to several others (reviewed in (H. E. Scharfman, 2016)). Elucidating neurohormonal signaling mechanisms might provide additional insight into how MC circuit function differs in distinct motivational states.

Neurohormonal systems regulating feeding have consistently been shown to also regulate hippocampal function with consequences for learning and memory (reviewed in (Suarez, Noble, & Kanoski, 2019)). As such, we were intrigued that cell-specific RNA sequencing (RNA-seq) of hippocampal excitatory neurons revealed transcriptional enrichment of Glp1r, the gene encoding the glucagon-like peptide-1 receptor (GLP-1R) (Cembrowski, Wang, Sugino, Shields, & Spruston, 2016), in MCs. Other feeding-relevant endocrine receptors, including ghrelin, leptin, and insulin, were not reported as enriched in MCs. GLP-1 is produced in the distal gut and in the brainstem, where its role in the central regulation of feeding has been extensively described (Muller et al., 2019). Interestingly, GLP-1R signaling also promotes hippocampal-dependent spatial and associative learning (During et al., 2003; Isacson et al., 2011) as well as DG adult neurogenesis (Gault, Lennox, & Flatt, 2015; H. Li et al., 2010). However, the specific neuronal substrate within the hippocampus on which GLP-1 acts is not well defined. In this study, we characterize the expression and function of GLP-1R on MCs in the murine DG. We find that hippocampal GLP-1Rs are strongly and selectively expressed on ventral DG MCs and that ventral MC GLP-1Rs are functional both ex vivo and in vivo. Our results support future investigation of how ventral MC GLP-1Rs regulate cognitive and non-cognitive functions shown previously to be mediated by MCs.

MATERIALS AND METHODS

Animals

Male and female Glp1r-ires-Cre mice (RRID:IMSR_JAX:029283) (Williams et al., 2016) were crossed with Ai14D tdTomato reporter mice (RRID:IMSR_JAX:007914) (Madisen et al., 2010) to yield Glp1r-ires-Cre x Ai14D mice. All other mice used were male and female C57BL/6J. Animals were group housed with a 12-hour light/dark cycle at 72 ± 2 °F with ad libitum access to food and water. All procedures were approved by the Vanderbilt Institute Animal Care and Use Committee.

Drugs

Exendin-4 acetate was purchased from Cayman Chemical Company (catalogue no. 11096, Ann Arbor, MI) (Yu, Park, & Beyak, 2019).

cFos expression following exendin-4 administration

Mice were habituated to the testing room for at least 1 hr, then administered exendin-4 (5 μg/kg in saline, s.c.) or saline, returned to their home cage, and transcardially perfused 90 mins later, a timecourse previously shown to be appropriate for cFos evaluation in mice (Bernstein, Lu, Botterill, & Scharfman, 2019). There was no prolonged deprivation of food or water prior to exendin-4 or saline administration.

Immunostaining

Terminal anesthesia, transcardial perfusion, and tissue sectioning were performed exactly as previously described ((Biddinger, Lazarenko, Scott, & Simerly, 2020) for Glp1r-ires-Cre x Ai14D mice and (Bauer et al., 2021) for all others). For GluR2/3 immunostaining: two brain sections from each animal in each region of interest were selected. Dorsal DG slices were at approximately anterior/posterior (AP): −1.94 mm and ventral DG slices at approximately AP: −3.40 mm. Sections were permeabilized and blocked in 0.3% Tx-100 and 3% normal donkey serum (Jackson ImmunoResearch, West Grove, PA) in phosphate-buffered saline (PBS) for 2 hours. Sections were incubated overnight in rabbit anti-GluR2/3 (AB1506, Millipore Sigma, Burlington, MA, RRID:AB_90710, 1:200 dilution) (Danielson et al., 2017) at 4 °C. Sections were washed 3 × 10 mins in PBS, then incubated in donkey anti-rabbit Alexa 488 (Jackson ImmunoResearch, 1:1,000 dilution) at room temperature for 2 hrs. Sections were washed 3 × 10 mins in PBS, then incubated in 4′,6-diamidino-2 -phenylindole (DAPI, Millipore, 1:5,000 dilution) in PBS at room temperature for 5 mins, washed in PBS, then mounted on slides using Fluoromount G (Electron Microscopy Sciences, Hatfield, PA). For GluR2/3 and calretinin costaining, supporting the specificity of the GluR2/3 antibody (Figure S1), the same protocol was used except mouse anti-calretinin (MAB1568, Millipore, RRID:AB_94259, 1:1,000 dilution) (Fuentes-Santamaria, Alvarado, Taylor, Brunso-Bechtold, & Henkel, 2005) labeling was performed in series after GluR2/3 labeling. cFos immunostaining was performed identically, except permeabilization and blocking was done using 0.1% Tx-100, 1% normal donkey serum in PBS. Primary antibody was rabbit anti-cFos (226 003, Synaptic Systems, Goettingen, Germany, RRID:AB_2231974, 1:1,000 dilution) (Zhou et al., 2019), and secondary antibody was donkey anti-rabbit Alexa 488 (Jackson ImmunoResearch, 1:500 dilution).

Microscopy and image quantification

Imaging was performed using an LSM 880 (Zeiss, White Plains, NY) equipped with a 20x Plan-Apochromat objective (NA = 0.8) and Zeiss Zen software for acquisition. All image analysis was performed blinded to treatment condition. For Glp1r-ires-Cre x Ai14D and GluR2/3 co-labeling, total numbers of tdTomato+ (i.e., Ai14D reporter), GluR2/3+, and tdTomato+GluR2/3+ neurons were manually counted within the hilus of each DG of each section and averaged across at least 2 sections per mouse. For cFos activation, the total number of cFos+ neurons within the DG hilus were counted blind to treatment condition and averaged across at least two sections per mouse. A hilar neuron was considered cFos+ when fluorescent signal was observed in a characteristic nuclear pattern overlying the DAPI signal and exceeded surrounding background signal. Images were processed using Fiji (Schindelin et al., 2012).

Acute slice preparation

Acute brain slices were prepared from male and female juvenile (P19 to P35) C57BL/6 mice. Mice were decapitated under isoflurane, and their brains were removed quickly and placed in an ice-cold sucrose-rich slicing artificial cerebrospinal fluid (ACSF) containing (in mM): 85 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 75 sucrose, 25 glucose, 0.01 DL-APV, 100 kynurenate, 0.5 Na L-ascorbate, 0.5 CaCl₂, and 4 MgCl₂. Sucrose-ACSF was oxygenated and equilibrated with 95% O2/5% CO2. Horizontal slices (300–350 µm) were prepared using a vibratome (model VT1200S, Leica Biosystems). Slices were transferred to a holding chamber containing sucrose-ACSF warmed to 30°C and slowly returned to room temperature over the course of at least 30 min. Slices were then transferred to oxygenated ACSF at room temperature containing (in mM): 125 NaCl, 2.4 KCl, 1.2 NaH₂PO₄, 25 NaHCO₃, 25 glucose, 2 CaCl₂, and 1 MgCl₂, and were maintained under these incubation conditions until recording.

Electrophysiological recordings

Slices were transferred to a submerged recording chamber continuously perfused at 2.0 ml/min with oxygenated ACSF maintained at 30 °C. Putative hilar mossy cells were identified using infrared differential interference contrast on a microscope (Slicescope II, Scientifica) and recordings made as previously reported (Hedrick et al., 2017). Whole-cell patch-clamp recordings were performed using borosilicate glass micropipettes with tip resistance between 3 and 6 MΩ. Signals were acquired using an amplifier (Axon Multiclamp 700B, Molecular Devices). Data were sampled at 10 kHz and low-pass filtered at 10 kHz. Access resistances ranged between 16 and 22 MΩ and were continuously monitored before switching to current-clamp configuration. Changes greater than 20% from the initial value that were recorded at the end of an experiment were excluded from data analyses. Series resistance was uncompensated. Data were recorded and analyzed using pClamp 11 (Molecular Devices). Current clamp was performed using a potassium gluconate-based intracellular solution containing the following (in mM): 135 K-gluconate, 5 NaCl, 2 MgCl₂, 10 HEPES (pH 7.0), 0.6 EGTA, 4 Na₂ATP, and 0.4 Na₂GTP, pH 7.3, at 281 mOsm). Input resistance was measured immediately after breaking into the cell and was determined from the peak voltage response to a 5 pA current injection. Following stabilization and measurement of the resting membrane potential, current was injected to hold all cells at a membrane potential between 60 – 65 mV, maintaining a common membrane potential that is within the reported resting membrane potential of hilar mossy cells to account for intercell variability. Hilar mossy cells were selected for recording based on the presence of a large, multipolar soma in the hilus. After achieving whole-cell configuration, hilar mossy cells were verified by a large whole-cell capacitance (>45 pF), a high frequency of sEPSCs (>5 Hz), and baseline action potential firing. Action potentials were characterized as having a mean amplitude of 83.988 ± 4.578 mV, a mean duration of 1.64 ± 0.09 ms, and a ratio of rising slope:decay slope greater than 2 (n = 7 cells from 6 animals), which is similar to that reported in the literature (H. E. Scharfman, 1992, 1995; H. E. Scharfman & Myers, 2012; H. E. Scharfman & Schwartzkroin, 1988).

Statistical analysis

t tests (paired or unpaired, as appropriate) were used to compare two groups. To compare three or more groups, one- or two-way analysis of variance (ANOVA) as appropriate with Sidak’s post test was used. All tests were two-tailed. Analyses were performed using Prism 9 (GraphPad, San Diego, CA). Error bars depict standard error of the mean (SEM) unless otherwise noted.

RESULTS

Using cell-specific RNA-seq, Cembrowski et al. previously reported that expression of Glp1r was enriched in MCs (Cembrowski et al., 2016). We further investigated this enrichment using HippoSeq, a publicly available tool to analyze the RNA-seq data generated by this study. Glp1r expression was detectable in MCs and ventral CA3 pyramidal neurons (Figure 1a). Gene expression of the receptors for ghrelin, insulin, and leptin, which like GLP-1 are other feeding-relevant hormones previously shown to be important in hippocampal function (Suarez et al., 2019), was markedly lower than Glp1r expression and not enriched in MCs (Figure 1b). HippoSeq did not divide MCs into dorsal and ventral DG MCs, which is important because dorsal and ventral MCs differ not only molecularly (Blasco-Ibanez & Freund, 1997; Fujise, Liu, Hori, & Kosaka, 1998), physiologically (Bui et al., 2018; Fredes et al., 2021; Jinno, Ishizuka, & Kosaka, 2003), and anatomically (Botterill, Gerencer, Vinod, Alcantara-Gonzalez, & Scharfman, 2021; Houser, Peng, Wei, Huang, & Mody, 2020), but also in their role in cognitive function (Bauer et al., 2021; Botterill, Vinod, et al., 2021; Yassa & Stark, 2011). Therefore, to corroborate RNA-seq data and examine potential dorsal and ventral MC expression differences, we examined in situ hybridization (ISH) data for Glp1r from the Allen Mouse Brain Atlas, which studied 56-day old male C57BL/6J mice (Lein et al., 2007) (Figure 1c,d). These data demonstrated Glp1r expression in neurons in both dorsal and ventral DG hilus, with markedly stronger staining in ventral DG, and modest expression in CA3. Together, these data reveal that Glp1r in hippocampal excitatory neurons is enriched in MCs, and that Glp1r expression is strongest in hilar neurons of the ventral DG.

Figure 1. Glp1r gene expression in the hippocampus.

a,b) HippoSeq, a publicly available mouse hippocampal principal neuron RNA-seq gene expression database, demonstrates that expression of Glp1r, the gene encoding GLP-1R, is enriched in MCs (a). Data in (a) are represented as mean ± 95% confidence intervals. Gene expression of other feeding-relevant hormone receptors in the hippocampus is much weaker than Glp1r expression and not enriched in MCs (b). Lepr, leptin receptor; Ghsr, growth hormone secretagogue receptor (ghrelin receptor); Insr, insulin receptor. FPKM, Fragments Per Kilobase of Exon Per Million Reads Mapped. c,d) In situ hybridization for Glp1r from Allen Mouse Brain Atlas corroborates RNA-seq data demonstrating most Glp1r expression is in hilar neurons consistent with MC expression, with strongest expression in the ventral DG hilus. There is less expression in CA1 or CA3 pyramidal cell layers. From bregma, depicted dorsal hippocampus is ~−1.9 mm and depicted ventral hippocampus is ~−3.5 mm. DG’ denotes the extreme ventral DG pole.

Because HippoSeq only includes excitatory neurons and did not differentiate MCs between dorsal and ventral DG, the above RNA-seq and ISH data are unable to determine 1) whether all MCs express Glp1r, 2) whether all Glp1r is confined to MCs (since interneurons were not included in HippoSeq), and 3) whether these expression patterns differ between dorsal and ventral MCs. All of these might have important functional consequences. To approach these questions, we examined Glp1r-ires-Cre x Ai14D mice, which express the fluorescent reporter tdTomato in Cre-positive neurons (Figure 2). This genetic reporter approach differs from ISH in that it allows visualization of not only neuronal soma but also projections and permits colabeling with MC markers. Unlike ISH however, genetic reporter intensity cannot be used as a proxy for relative expression. In dorsal hippocampus, tdTomato was almost exclusively expressed in hilar cells, with almost no expression in the granule cell layer, area CA3, and area CA1. In ventral hippocampus, tdTomato was expressed in hilar cells, including at the extreme ventral pole. Expression was sparse in ventral CA3, though much more prevalent than in dorsal CA3. Ventral CA1, as in dorsal CA1, was largely devoid of tdTomato expression. In both dorsal and ventral sections, a strong band of tdTomato immunoreactivity was found in the inner molecular layer, the site of MC terminals. Finally, strong expression of tdTomato was found throughout the brain in blood vessels, consistent with known localization of GLP-1R in brain arteriolar smooth muscle and endothelial cells (Nizari et al., 2021).

Figure 2. Genetic reporter for Glp1r-ires-Cre expression in dorsal and ventral hippocampal formation.

Confocal microscopy of dorsal and ventral hippocampal sections from Glp1r-ires-Cre mice crossed with Ai14D genetic reporter line (Glp1r-ires-Cre x Ai14D), in which the red fluorescent protein tdTomato is expressed in Cre-positive neurons, revealed hilar expression in both dorsal and ventral hippocampus, along with sparse ventral CA3 expression. There was also strong tdTomato expression in the DG inner molecular layer, consistent with MC terminals. Glp1r is also strongly expressed in the brain vasculature, exemplified by structures marked by asterisks. Insets are single optical sections at 40X magnification of the indicated DG region. Scale bar in inset images = 50 microns.

We next tested whether hilar neurons expressing the fluorescent genetic reporter were MCs by co-labeling with GluR2/3, a marker of hilar MCs in both dorsal and ventral DG (Jiao & Nadler, 2007; Leranth, Szeidemann, Hsu, & Buzsaki, 1996) (Figure 3a). These studies revealed that MC expression of Glp1r and Glp1r specificity for MCs differed between dorsal and ventral MCs (Figure 3b,c). In the ventral DG, 95.7% ± 1.6% of GluR2/3+ MCs expressed Glp1r and 94.2% ± 1.1% of Glp1r-expressing hilar neurons expressed GluR2/3. However, in the dorsal DG, only 60.8% ± 5.6% of dorsal DG GluR2/3+ MCs expressed Glp1r, and 66.3% ± 4.4% of Glp1r-expressing neurons also expressed GluR2/3+, suggesting about one-third of dorsal Glp1r-expressing neurons were not MCs. Altogether, RNA-seq, ISH, and genetic reporter strategies suggest that Glp1r expression overall is more prevalent in ventral hippocampus, where it is expressed in a sparse population of CA3 pyramidal neurons and expressed in essentially all ventral MCs, while Glp1r is not universally expressed in dorsal MCs and is also expressed in non-MC hilar neurons. Furthermore, dense tdTomato-positive terminals in the inner molecular layer throughout the DG dorsoventral axis suggests that Glp1r-positive MCs innervate granule cells across all hippocampal lamellae.

Figure 3. Hilar mossy cell Glp1r expression differs across the DG longitudinal axis.

a) Glp1r-ires-Cre mice crossed with Ai14D reporter mice (Glp1r-ires-Cre x Ai14D) express tdTomato in Cre-positive neurons. Representative images from the dorsal and ventral DG are shown in the top and bottom rows of images, respectively, revealing tdTomato expression in hilar somata as well as a dense band of projections in the DG inner molecular layer. Sections were stained for GluR2/3, which in the hilus is a marker for MCs. Images on the far right are magnification of the boxed area. White arrowheads denote example Glp1r-positive/GluR2/3-negative neurons, black arrowheads denote example Glp1r-negative/GluR2/3-positive neurons, and white arrows denote example Glp1r-positive/GluR2/3-positive neurons. b,c) In the ventral DG, almost all MCs were Glp1r-positive, and almost all Glp1r-positive neurons were MCs. However, in the dorsal DG, only about 61% MCs were Glp1r-positive and 66% of Glp1r-expressing neurons were MCs. N = 6 mice. Paired t test, dorsal versus ventral: % Glp1r+GluR23+/Glp1r+: t(5) = 7.77, p < 0.001; % Glp1r+GluR23+/GluR23+: t(5) =8.19, p = < 0.001.

Glucagon-like peptide-1 and several GLP-1R agonists, including exendin-4, readily cross the blood-brain barrier (Kastin & Akerstrom, 2003; Kastin, Akerstrom, & Pan, 2002) where they have centrally mediated effects on cognition, feeding, and neurogenesis (During et al., 2003; Gault et al., 2015; Isacson et al., 2011; Kanoski, Fortin, Arnold, Grill, & Hayes, 2011). To investigate whether DG activity was changed by pharmacological activation of GLP-1Rs, we administered the GLP-1R agonist exendin-4 (5 μg/kg) peripherally, sacrificed the animal 90 mins later (Bernstein et al., 2019), and performed immunohistochemistry for the immediate early gene cFos. Consistent with stronger expression of Glp1r in ventral MCs, we found a significant increase in cFos following exendin-4 administration in the ventral but not dorsal DG (Figure 4). We also examined dorsal DG cFos expression at 30 min after vehicle or exendin-4 administration to test whether we may have missed a more transient change in cFos (Moretto, Duffy, & Scharfman, 2017). However, we again found no difference in dorsal DG cFos expression at this time point (Figure S2). Finally, while both male and female mice were used, the study was not formally powered to detect sex differences.

Figure 4. Peripheral administration of GLP-1R agonist increases ventral but not dorsal DG hilar cFos expression.

a) Mice were administered the GLP-1R agonist exendin-4 (5 micrograms/kg) or vehicle and returned to their home cage. 90 mins later mice were sacrificed. Immunohistochemistry for the immediate early gene cFos revealed that expression of hilar cFos was increased by exendin-4 treatment in the ventral DG but was unchanged by exendin-4 treatment in the dorsal DG, , quantified from horizontal and coronal sections, respectively. b) Higher magnification images of the regions depicted in (a). c) Two-way ANOVA: dorsoventral region x treatment interaction: F(1,17) = 6.154, p = 0.024; dorsoventral region: F(1,17) = 4.348, p = 0.052, treatment: F(1,17) = 4.999, p = 0.039. Pairwise comparison p values from Sidak’s multiple comparison tests are shown in the figure. Dorsal vehicle: N = 2M/2F; Dorsal exendin-4: N = 2M/1F; Ventral vehicle: N = 4M/2F; Ventral exendin-4: N = 4M/4F.

Along with Glp1r expression data, these findings suggest that GLP-1R activation might act directly on MCs to increase their firing. However, cFos cannot differentiate this possibility from network effects, such as activation of MCs and hilar GABAergic interneurons via GLP-1R-mediated activation of CA3c neurons and back projections to these DG hilar cell types, a process known to be particularly strong in the ventral hippocampus (H. E. Scharfman, 2007). To quantify the effects of GLP-1R activation on MC physiology, we performed whole-cell current-clamp recordings from ventral DG MCs to examine the effect of bath application of exendin-4 (200 nM) (Figure 5a), a dose used recently to characterize neuronal responses to GLP-1R activation ex vivo (Povysheva, Zheng, & Rinaman, 2021). We specifically examined ventral MCs because our expression data demonstrated that almost all ventral hilar neurons expressing Glp1r are MCs (Figure 3), which, in addition to our electrophysiological criteria (see methods), contributes to the likelihood that we were indeed recording from MCs. Bath application of 200 nM exendin-4 significantly increased the rate of spontaneous action potential firing (Figure 5b-d) compared to baseline. This was associated with a small but significant membrane depolarization compared to baseline (Figure 5e,f) and is consistent with actions of GLP-1R agonists in neurons of the hypothalamic paraventricular nucleus and the bed nucleus of the stria terminalis (Cork et al., 2015; Povysheva et al., 2021).

Figure 5. GLP-1R agonist depolarizes hilar MCs and increases action potential firing.

a) Image of recording electrode in the hilus from a horizontal ventral DG slice. b) Representative current clamp recording of hilar MC following bath application of GLP-1R agonist exendin-4 (200 nM) showing increase in action potential firing and membrane depolarization. c) Mean action potential (AP) firing following exendin-4 (200 nM) bath application reveals overall increase in firing. Data are normalized to mean firing during the 2 min period prior to onset of bath application. One-way repeated measures ANOVA: F(11,66) = 2.194, p = 0.025. **p < 0.01 versus −1 min baseline (Sidak’s multiple comparison test). N = 7 cells from 6 mice. d) Maximal firing rate in the 9 min period following exendin-4 bath application is significantly increased from baseline. Paired t test versus −1 min baseline: t(6) = 2.62, p = 0.039. e) Change in membrane potential (V_m) also increased significantly following exendin-4 bath application. One-way repeated measures ANOVA: F(11,66) = 3.112, p = 0.0020. *p < 0.05 versus −1 min baseline (Sidak’s multiple comparison test). N = 7 cells from 6 mice. f) Maximal V_m change in the 9 min period following Exendin-4 bath application is significantly increased from baseline. Paired t test versus −1 min baseline: t(6) = 2.71, p = 0.035.

DISCUSSION

Identifying non-synaptic mechanisms that regulate MC activity is critically important to bridge our current understanding of MC function, primarily gleaned from activity manipulation studies, with MC function during specific physiological and behavioral states. In this study, we demonstrate that GLP-1 signaling via GLP-1Rs on MCs may be an important neurohormonal mechanism regulating MC activity, as well as a cellular target for GLP-1 analogues already in clinical use. Using RNA-seq, ISH, and genetic reporter lines, we found that 1) ventral MCs strongly express Glp1r, 2) that GLP-1R agonist depolarizes MC membrane potential and increases action potential firing ex vivo, and 3) that peripheral administration of GLP-1R agonist increases ventral DG hilar neuron activity in vivo. Interestingly, dorsal MCs differed markedly from ventral MCs in their expression of Glp1r, where it is expressed in only about two-thirds of MCs, and only about 60% of Glp1r-positive hilar neurons were MCs. Glp1r expression as revealed by ISH in dorsal hilar neurons was weaker than in ventral hilus, which may contribute to our inability to detect cFos activation in the dorsal hilus by peripheral GLP-1R agonist. Alternatively, dorsal hilar Glp1r-positive neurons that were not MCs may play an active role in inhibiting local hilar neurons following GLP-1R agonist administration, also accounting for an absence of detectable cFos increase. There remain several important anatomical questions for future study. It will be critical to determine the identity of non-MC dorsal hilar neurons that express Glp1r, which will require careful differentiation between interneuron subtypes. Relatedly, a small number of murine granule cells, labeled by Prospero homeobox 1 (Prox1), are located within the hilus, which in adult C57BL/6J mice are comprised of approximately half mature (NeuN+) and half immature (NeuN-) neurons (Bermudez-Hernandez et al., 2017; H. Scharfman, Goodman, & McCloskey, 2007). Additional studies are necessary to determine whether these hilar granule cells express Glp1r and what its functional effects are on neuronal physiology. Finally, a basic understanding of sex differences in Glp1r localization using a rigorous, well-powered experimental design would be potentially of great benefit for contextualizing future behavioral studies using drugs active at GLP-1Rs.

As is the case for the hippocampal formation as a whole (Fanselow & Dong, 2010; Strange, Witter, Lein, & Moser, 2014), an appreciation of meaningful differences between dorsal and ventral DG MCs continues to evolve. These include marked differences in protein expression (Blasco-Ibanez & Freund, 1997; Cembrowski et al., 2016; Fujise et al., 1998), activity (Bui et al., 2018; Fredes et al., 2021; Jinno et al., 2003), connectivity (Botterill, Gerencer, et al., 2021; Houser et al., 2020), and effects on cognitive and behavioral function (Bauer et al., 2021; Botterill, Vinod, et al., 2021; Yassa & Stark, 2011). For instance, relevant to our Glp1r expression findings, in mice the calcium binding protein calretinin is strongly and selectively expressed in MCs in the ventral and intermediate DG whereas most dorsal MCs are calretinin-negative (Blasco-Ibanez & Freund, 1997; Fujise et al., 1998). Ventral MCs show markedly greater intrinsic bursting than dorsal MCs due to differential expression of persistent sodium currents (Jinno et al., 2003). Finally, ventral MCs are significantly more active in novel contexts and this novelty detection can gate contextual fear conditioning whereas dorsal MC activity plays a less specific role in this form of learning (Fredes & Shigemoto, 2021; Fredes et al., 2021). Thus, our findings that ventral MCs and the ventral DG are activated by GLP-1R agonist may have several behavioral consequences. Whether GLP-1R signaling improves or degrades performance in DG-dependent cognition is likely task-dependent and difficult to predict. For example, either optogenetic inhibition (Bui et al., 2018) or chemogenetic excitation (Bauer et al., 2021) of ventral MCs impairs spatial encoding during an object location memory task, suggesting that disruption of distinct activity bidirectionally has a degradative effect on encoding. However, GLP-1R signaling may enhance contextual fear conditioning in familiar environments (Fredes et al., 2021) and perhaps restore MC activity that is necessary for mnemonic function in pathological conditions in which MCs are lost, such as epilepsy (Blümcke et al., 2000; Bui et al., 2018).

It is important to note that MC protein expression may be highly species dependent, which has previously been appreciated in calretinin staining between mice, rats, and monkeys (Blasco-Ibanez & Freund, 1997; Gulyás, Miettinen, Jacobowitz, & Freund, 1992; Miettinen, Gulyás, Baimbridge, Jacobowitz, & Freund, 1992; Seress, Nitsch, & Leranth, 1993). Along these lines, recent work in rats identified Glp1r expression in ventral CA1 neurons and showed that GLP-1R signaling through these neurons had functional effects on food intake and operant responding (Hsu, Hahn, Konanur, Lam, & Kanoski, 2015; Hsu et al., 2018). In contrast, murine RNA-seq, ISH, and a Glp1r-ires-Cre x Ai14D reporter line cross did not demonstrate appreciable ventral CA1 Glp1r expression (Figures 1 and 2). Even within species, different transgenic approaches to visualize neuronal Glp1r show notable differences. For example, the Glp1r-ires-Cre knock-in line crossed with Ai14D mice shown in the present study exhibits similar dorsal and ventral DG hilar neuronal expression to a Glp1r-Cre BAC transgenic line (Cork et al., 2015; Richards et al., 2014) and a GLP-1R-mApple BAC transgenic line (Graham et al., 2020). However, these three lines differ in their reporter expression in other hippocampal cell types, including DG granule cells and pyramidal neurons in CA1 and CA3, which contrast with RNA-seq and ISH data. That three different transgenic lines demonstrate hilar neuronal expression adds confidence to the expression of Glp1r in MCs. However, variable expression elsewhere suggests a need for caution when making claims as to the degree of Glp1r expression in these other hippocampal fields.

Our findings set the stage for several avenues of future study for understanding healthy brain function and for the development of therapeutic interventions. Basic motivational states, such as hunger or thirst, interact with cognitive processes to guide adaptive behavior. For instance, in a state of hunger, an adaptive response is to seek and consume food. This process is facilitated by pairing the hungry state with prioritized recall of food-paired contexts. Indeed, such fundamental motivational states are coded in hippocampal firing (P. J. Kennedy & Shapiro, 2004; Pamela J. Kennedy & Shapiro, 2009; Wood, Dudchenko, Robitsek, & Eichenbaum, 2000), but how state is communicated to the hippocampus is not well understood. As a satiety signal (Hsu et al., 2018; Muller et al., 2019; Trapp & Richards, 2013), GLP-1 signaling is one potential mechanism that might couple satiety or absence of satiety with distinct MC function. Indeed, because GLP-1Rs are G-protein-coupled receptors, even relatively short-lived prandial increases in central GLP-1 levels might be well situated to have long-lasting effects on MCs that far outlast the presence of the hormone itself. Supporting this notion, MCs were recently shown to be more active in the fed than fasted state (Azevedo et al., 2019). Finally, the unique vulnerability of MCs in disorders such as epilepsy and the cognitive consequences of their loss coupled with the current widespread clinical use of GLP-1R pharmacotherapies encourages translational investigation into whether targeting MC GLP-1Rs can preserve or enhance pattern separation function to facilitate episodic memory in neurological or psychiatric disease.

Data availability statement:

The data that support the findings of this study are available from the corresponding author upon reasonable request.