Target selectivity of septal cholinergic neurons in the medial and lateral entorhinal cortex

Significance

Acetylcholine is a key modulator of hippocampal and entorhinal cortex (EC) function. The majority of cholinergic projections targeting these structures originate in the basal forebrain complex, specifically the medial septum. Many studies focused on the behavioral effects involving these projections, but there still is a paucity regarding their connectivity in the target area. Here we provide this missing link. By combining optogenetics with whole-cell recordings in superficial EC layers, we identified the synaptic target cells of septal cholinergic neurons. This level of analysis is an important step toward a better understanding of the modulatory action of acetylcholine in EC in vivo.

Abstract

The entorhinal cortex (EC) plays a pivotal role in processing and conveying spatial information to the hippocampus. It has long been known that EC neurons are modulated by cholinergic input from the medial septum. However, little is known as to how synaptic release of acetylcholine affects the different cell types in EC. Here we combined optogenetics and patch-clamp recordings to study the effect of cholinergic axon stimulation on distinct neurons in EC. We found dense cholinergic innervations that terminate in layer I and II (LI and LII). Light-activated stimulation of septal cholinergic projections revealed differential responses in excitatory and inhibitory neurons in LI and LII of both medial and lateral EC. We observed depolarizing responses mediated by nicotinic and muscarinic receptors primarily in putative serotonin receptor (p5HT₃R)-expressing interneurons. Hyperpolarizing muscarinic receptor-mediated responses were found predominantly in excitatory cells. Additionally, some excitatory as well as a higher fraction of inhibitory neurons received mono- and/or polysynaptic GABAergic inputs, revealing that medial septum cholinergic neurons have the capacity to corelease GABA alongside acetylcholine. Notably, the synaptic effects of acetylcholine were similar in neurons of both medial and lateral EC. Taken together, our findings demonstrate that EC activity may be differentially modulated via the activation or the suppression of distinct subsets of LI and LII neurons by the septal cholinergic system.

The superficial layers of entorhinal cortex (EC) play a quintessential role in processing and conveying sensory information from the neocortex to the hippocampus via the perforant and temporoammonic pathways (1, 2). Commonly, EC is subdivided into medial EC (MEC) and lateral EC (LEC), which display both functional and organizational differences (3–5). MEC conveys mostly spatial information (6, 7), while LEC mediates nonspatial inputs to the hippocampus (8–10). Based on anatomical and electrophysiological studies, distinct principal cell types have been described in layer II (LII) of MEC and LEC (11–14). These principal cell types, together with local interneurons, constitute the complex local microcircuit in LII of EC (13, 15), which received increasing attention ever since the discovery of grid cells and other spatially tuned cells in the superficial layers of MEC (6, 7). In vivo recordings from this area in freely moving animals reveals that the most striking rhythmic activity is theta oscillations (4–12 Hz) (16), which is a prerequisite for the temporal organization of neurons within local networks. Of note, MEC exhibits stronger theta rhythmicity than LEC, and so far, it appears that in contrast to MEC, neurons in LEC are only weakly modulated by theta (17). Although local networks can support the generation and maintenance of theta oscillations (18, 19), the main source for theta oscillations in EC is the medial septum/vertical limb of the diagonal band of Broca (MS/DBB) (19–21).

It has been known for decades that MS/DBB, which is a part of the basal forebrain complex, is the main external pacemaker of hippocampal-EC theta rhythm, thereby coordinating several distant brain areas (19, 22). Pharmacological inactivation of MS strongly reduces theta oscillations and grid cell firing in MEC, leading to spatial memory deficits (23–25). The septo-entorhinal pathway comprises glutamatergic, GABAergic, and cholinergic projections (26). The target selectivity of septal glutamatergic and GABAergic projections in LII of MEC has been previously studied. While the former targets pyramidal cells (PCs) and fast-spiking (FS) interneurons (27), the latter selectively inhibits interneurons in LII of MEC (13, 28). Selective lesions of septal cholinergic neurons or their optogenetic activation have indicated that acetylcholine (ACh) plays an important role in regulating theta rhythmic activity in the hippocampus, thereby augmenting the dynamics of memory encoding (29–32). Optogenetic activation of septal cholinergic fibers in the CA1 region of the hippocampus evoked responses in both PCs and in a variety of interneurons (33–35), but the identity of cholinergic MS/DBB target cells in EC has remained elusive. In vitro investigations using bath application of cholinergic receptor agonists (36, 37) gave first indications as to differential effects of ACh in EC neurons. However, technical limitations precluded specific axonal activation that is nowadays offered by virus-supported optogenetics, enabling both clear-cut identification of source cells and mimicking of endogenous release.

Here we analyzed the septal cholinergic projections to MEC and LEC, focusing on the identification of target cells and characterization of optogenetically induced responses. We characterized the projections based on antero- and retrograde tracing. Following virus-mediated ChR2-mCherry expression in cholinergic MS/DBB neurons, we electrophysiologically studied cholinergic responses in identified neurons in acute slices and demonstrate that they activate or suppress distinct subsets of LI/LII neurons. Measurements in MEC and LEC were directed toward answering the question of whether common connectivity rules hold true in the two brain areas. This indeed appears to be the case. However, based on our anatomical and electrophysiological data, we infer that cell type-specific septal cholinergic modulation is more pronounced in MEC than in LEC neurons.

Results

Septal Cholinergic Neurons Project to the MEC and LEC.

To investigate the extent of MS/DBB cholinergic projections to EC, we injected AAV-DIO-ChR2-mCherry virus (in four instances AAV-DIO-eGFP virus was used; for details see SI Materials and Methods) into MS/DBB of 8-wk-old male ChAT^Cre mice (Fig. 1A). Viral expression, as indicated by mCherry labeling of cell bodies, was observed across MS/DBB (Fig. 1B), with an infection rate of 50.76 ± 2.32% within MS/DBB cholinergic cells (Fig. 1C, 573 double⁺ cells from a total of 1,148 ChAT⁺ cells in 12 slices from 4 mice). We found that mCherry was virtually exclusive in MS/DBB cholinergic neurons, as 98.92 ± 0.22% of the cells positive for mCherry expression were also ChAT⁺ (Fig. 1C, 573 double⁺ cells from a total of 579 mCherry⁺ cells in 12 slices from 4 mice). Furthermore, within MEC and LEC, we observed mCherry⁺ cholinergic axons predominantly in LI/LII, with fewer axons in LEC (Fig. 1G and Fig. S1). To corroborate the results of our anterograde tracing experiments and to estimate the percentage of EC projecting septal cholinergic neurons, we injected the retrograde tracer fluorogold (FG; 0.5%) unilaterally into MEC and LEC (Fig. 1D). We found more MS/DBB cholinergic neurons projecting to MEC (5.70 ± 0.75%) than LEC (3.37 ± 0.51%) (Fig. 1 E and F, 93 double⁺ cells from a total of 1,786 ChAT⁺ cells in 14 slices from 5 mice for MEC and 83 double⁺ cells from a total of 2,270 ChAT⁺ cells in 15 slices from 6 mice for LEC; P < 0.05). As indicated by vesicular acetylcholine transporter (vAChT) stainings, the mCherry⁺ axons within the EC can pack and release ACh (Fig. 1H). To test whether there is a difference in synapse density, we quantified the number of vAChT punctae per axon length. Indeed, the synapse density with 2.1 ± 0.1 vAChT punctae/10 µm of axon segment was significantly higher in MEC compared with LEC, with 1.5 ± 0.2 vAChT punctae/10 µm of axon segment (Fig. 1I, n = 32 axon segments from 5 mice for MEC and n = 20 axon segments from 4 mice for LEC, respectively; P < 0.05).

Fig. 1.

MS/DBB cholinergic neurons project to EC. (A) Schematic representation indicating the injection site (red) in MS/DBB of a ChAT^Cre mouse for anterograde tracing experiments with AAV-DIO-ChR2-mCherry. (B) Maximum intensity confocal image showing selective expression of ChR2-mCherry in ChAT-expressing neurons upon stereotaxic virus injection into MS/DBB. Higher magnification of a representative mCherry⁺/ChAT⁺ cell from the boxed area is shown below. (C) Percentage of double⁺ cells relative to mCherry (gray) and ChAT (green) expression (n = 12 slices from 4 mice). (D) Schematic representation of a horizontal section indicating the unilateral injection site into MEC (D₁, blue) and LEC (D₂, orange) for retrograde FG (0.5%) tracing experiments. (E) Maximum intensity confocal image of MS/DBB showing retrograde FG labeling in ChAT-expressing neurons injected unilaterally in MEC (E₁) and LEC (E₂). Higher magnification of a representative FG⁺ and ChAT⁺ cell is shown as below. (F) Percentage of double⁺ cells relative to ChAT expressions after unilateral injection of the retrograde tracer FG into MEC (red; n = 14 slices from 5 mice) and LEC (blue; n = 15 slices from 6 mice). (G) Confocal images showing mCherry⁺ axonal projections in superficial layers of MEC (Top) and LEC (Bottom). Red dashed lines indicate the border between LI and LII as ascertained based on DAPI staining. (H) vAChT punctae are localized within mCherry⁺ cholinergic axons in MEC (Top) and LEC (Bottom). Images below show higher magnification of the boxed areas. (I) Number of vAChT punctae per micrometer of axon length is shown for MEC (red; n = 32 axon segments from 5 mice) and LEC (blue; n = 20 axon segments from 4 mice). Error bars indicate mean ± SEM. *P < 0.05. [Scale bars: (B and E) 150 µm, (G) 50 µm, and (H) 10 µm.]

Identification and Characterization of Postsynaptic Targets in LI/LII of MEC and LEC.

We employed acute slices and performed optogenetic stimulation of MS/DBB-derived cholinergic axons and patch-clamped adjacently located neurons. We recorded from LII EC excitatory and LI/LII EC inhibitory neurons (223 and 289 cells, respectively). In MEC, the four principal cell types—namely, PCs, intermediate pyramidal cells (IMPCs), stellate cells (SCs), and intermediate stellate cells (IMSCs)—were classified based on their intrinsic electrophysiological properties as described before (ref. 13; Fig. S2 A–F and Table S1; estimated classification error of ∼5%; see Materials and Methods). In LEC, in line with previous reports, the three major principal cell types—that is, PCs, fan cells (FCs), and multiform cells (MFCs)—cannot be distinguished based on electrophysiological properties (Fig. S2 G–J and Table S2; estimated classification error of ∼50%) but purely by morphology (11, 14). Therefore, only biocytin-filled and visually identified LEC principal cells were used in this study. LI interneurons can be morphologically segregated into neurogliaform cells (LI NGCs) and single bouquet-like cells (LI SBCs), which constitute the two major cell types (38–40). We could clearly distinguish them based on their intrinsic electrophysiological properties (Fig. S3 A–C and Table S3; estimated classification error of ∼5%). As in many other brain areas, EC LII interneurons can be subdivided into three by-and-large nonoverlapping subgroups (13, 41, 42). Their electrical signature corresponds to the expression of distinct neurochemical markers, which are hence considered in this study as putative parvalbumin (pPV⁺), putative somatostatin (pSOM⁺), and putative serotonin receptors (p5HT₃R)-expressing interneurons (Fig. S3 D–G and Table S4, estimated classification error of ∼2%; see Materials and Methods).

Nicotinic Receptor-Mediated Responses Are Elicited in LI and LII p5HT₃R⁺ Interneurons.

Nicotinic acetylcholine receptor (nAChR)-mediated inputs were tested by stimulating septal cholinergic axons with 5-ms LED pulses, while clamping the cells at their approximate resting membrane potential of −70 mV. The nicotinic nature of the resulting excitatory postsynaptic currents (EPSCs) was verified by mecamylamine application (Fig. 2A, nonselective nAChR blocker, n = 17/17). For some cells, before mecamylamine, we applied CNQX/d-AP5 (glutamatergic receptor antagonists), which did not block the response (n = 12/12; amplitude comparison P > 0.1). The latency, peak amplitude, rise, and decay time constants for all responding cells in MEC and LEC were 4.1 ± 0.2 ms, 3.2 [1.8, 7.2] pA, 1.5 [1.1, 2.0] ms, and 5.9 [3.5, 8.7] ms, respectively (Fig. 2B, reported as mean ± SEM for normally distributed data and median [25th, 75th percentile] for other cases; no difference was observed between MEC and LEC). Cell type-specific response probabilities were statistically estimated using a Bayesian approach and are depicted as violin plots (for details, see SI Materials and Methods). Responses were elicited in LI and LII p5HT₃R⁺ interneurons of MEC (Fig. 2C). In LEC, we observed only a few responses in LI/LII interneurons (Fig. 2C). Hence septal cholinergic neurons excite LI and LII p5HT₃R⁺ interneurons via activation of nAChRs.

Fig. 2.

MS/DBB cholinergic neurons elicit nicotinic receptor-mediated responses in interneurons of EC. (A₁) Representative average current trace of a cell held at −70 mV showing nAChR-mediated EPSC upon single 5-ms LED stimulation. (A₂) Response was not blocked by glutamatergic receptor antagonists (CNQX/d-AP5; n = 12/12) (A₃) but was completely blocked by a nonselective nAChR antagonist (mecamylamine; n = 12/12). Baselines and LED stimulations are indicated by red dashed lines and blue lines, respectively. (B) Box and jittered-dot plots showing the latency and peak amplitude (on a logarithmic scale) of EPSCs elicited in all responding cells in MEC (red) and LEC (blue). (C) Cell type-specific responses in MEC (Left) and LEC (Right) are depicted as violin plots that reflect the number of sampled cells of a given cell type and the number of responding cells within that group of cells (for details, see SI Materials and Methods). Red lines indicate the median of the distribution. Vertical dashed line in each plot segregates excitatory (Left) from inhibitory (Right) cell types. Numbers above the violin plots denote responding and patched cells for every cell type.

Septal Cholinergic Neurons Evoke Different Muscarinic Receptor-Mediated Responses.

Muscarinic acetylcholine receptor (mAChR)-mediated inputs were tested by stimulating cholinergic axons with ten 5-ms LED stimuli at 5-Hz frequency, while clamping the cells at a subthreshold membrane potential of −50 mV. Two main types of mAChR-mediated responses—namely, hyperpolarizing (Fig. 3A₁) and depolarizing (Fig. 3A₂) responses—were elicited in superficial EC neurons. The resulting EPSCs or IPSCs were verified by bath-application of atropine (nonselective mAChR blocker, n = 8/8 for hyperpolarizing and n = 7/7 for depolarizing responses). A third type of mAChR-mediated response (biphasic) was rarely observed (Fig. 3A₃, 2/488 cells patched) and was not further analyzed. In MEC and LEC, hyperpolarizing responses were predominantly observed in LII excitatory cells [Fig. 3C, excitatory cells (58/151 and 44/98; MEC and LEC) vs. interneurons (5/107 and 2/73): P < 10⁻¹⁰ and P < 10⁻¹⁰], whereas depolarizing responses were preferentially elicited in interneurons [Fig. 3D, interneurons (20/107 and 5/73; MEC and LEC) vs. excitatory cells (0/151 and 0/98): P < 10⁻⁸ and P < 0.05]. Different from other interneurons, pPV⁺ interneurons in MEC exhibited a hyperpolarizing and not a depolarizing response (P < 0.05). Latency to peak, peak amplitude, and charge for all hyperpolarizing responses were 1.7 [1.2, 1.9] s, 2.8 [2.1, 4.7] pA, and 6.9 [4.1, 13.0] pC (Fig. 3B₁) and for depolarizing responses were 2.8 [1.4, 4.0] s, 2.0 [1.1, 2.9] pA, 4.5 [3.2, 9.7] pC (Fig. 3B₂), respectively (P < 10⁻³, P < 10⁻³, and P < 0.05, respectively).

Fig. 3.

MS/DBB cholinergic neurons evoke hyperpolarizing or depolarizing muscarinic receptor-mediated responses in a cell type-specific fashion. (A) Representative average current traces (gray) of cells held at −50 mV showing mAChR-mediated hyperpolarizing (A₁, Left), depolarizing (A₂, Left), and biphasic (A₃) response upon ten 5-ms LED stimulations at 5 Hz. In black, Savitzky–Golay-filtered data are shown. Baselines and LED stimulations are indicated by red dashed lines and blue lines, respectively. (A₁, Right; A₂, Right) Responses were completely blocked by a nonselective mAChR antagonist (atropine; n = 8/8 and 7/7, respectively). (B) Box and jittered-dot plots depicting the latency to peak and peak amplitude (on a logarithmic scale) of currents elicited by all MEC (red) and LEC (blue) cells that show either (B₁) a hyperpolarizing response or (B₂) a depolarizing response. (C and D) Cell type-specific responses are depicted as violin plots that reflect the number of sampled cells of a given cell type and the number of responding cells within that group of cells (for details, see SI Materials and Methods). The violin plots indicate the response probability for all target cells in superficial layers of MEC (Left) and LEC (Right) that show (C) a hyperpolarizing response and (D) a depolarizing response. Red lines indicate the median of the distribution. Vertical dashed line in each plot segregates excitatory (Left) from inhibitory (Right) cell types. Numbers above the violin plots denote responding and patched cells for every cell type.

In a subset of experiments, we stimulated with 10 stimuli not only at 5 Hz but also at 2, 10, and 20 Hz to test for a frequency preference (Fig. S4A). While peak amplitude plateaued in the theta range, the overall transferred charge did not appear to be dependent on the stimulation frequency (Fig. S4B). Thus, we used a simple model (linear rise and decay of each individual EPSC and no short-term plasticity) to predict the observed behavior. As the prediction fit our data nearly perfectly (Fig. S4B), it is most likely that there is no strong frequency preference at this synapse.

Notably, previous studies performed in the MEC (36, 37) and other brain regions (43–45) reported opposite results to ours. Exogenous bath application of carbachol (CCh; nonselective AChR agonist) induced slow mAChR-mediated depolarizations, whereas we find that optogenetic stimulation evoked hyperpolarizing responses in excitatory cells. To follow up on this apparent contradiction, we conducted a set of experiments in which we tested optogenetically and CCh-induced responses on the same cell. Nine out of 22 LII MEC excitatory cells showed a mAChR-mediated hyperpolarizing response upon optogenetic stimulation, as opposed to 21 of 22 cells that exhibited mAChR-mediated depolarizations after puff application of 10 mM CCh (Fig. S5). We also tested responses evoked by longer stimulations—that is, 20 Hz for 2 s. All eight tested LII MEC excitatory cells showed a hyperpolarizing response.

MS/DBB Cholinergic Neurons Corelease GABA.

Of note, in a substantial fraction of cells (61/488 cells recorded), we observed time-locked fast IPSCs that corresponded most likely to GABA_A receptor (GABA_AR) activation. Hence we wondered whether the IPSCs reflected indirect recruitment of local interneurons, in analogy to what was shown for the hippocampus (34), or if MS/DBB cholinergic neurons corelease GABA in EC, as previously suggested in the neocortex (46). We first performed fluorescence in situ hybridization for Gad1/Gad2 and ChAT mRNA and found that 88.84 ± 3.73% of MS/DBB cholinergic neurons express Gad2 (Fig. 4 A and B, 215 double⁺ cells from a total of 242 ChAT⁺ cells in four slices from two mice) but not Gad1 mRNA (Fig. 4 A and B, seven double⁺ cells from a total of 266 ChAT⁺ cells in four slices from two mice, P < 0.0001). Moreover, MS/DBB cholinergic axons expressed vesicular GABA transporter (vGAT) along with vAChT in EC, indicating that GABA may be coreleased (Fig. 4C).

Fig. 4.

MS/DBB cholinergic neurons elicit GABA_AR-mediated responses in EC neurons. (A) Maximum intensity confocal images of fluorescence in situ hybridization showing expression of Gad1/Gad2 (Left/Right) and ChAT mRNA in MS/DBB neurons. Higher magnification images of the boxed areas below show ChAT⁺ cells that do not express Gad1 but express Gad2. (B) Percentage of ChAT⁺ cells that are Gad1⁺/Gad2⁺ in MS/DBB (n = 4 slices from 2 mice). (C) vGAT punctae colocalizing with vAChT punctae inside mCherry⁺ axonal termini in MEC (Top) and LEC (Bottom). Images below show higher magnification of the boxed areas. (D₁) Representative average current traces of a cell held at −50 mV showing GABA_AR-mediated IPSCs upon single 5-ms LED stimulation. (D₂–D₄) Response was not blocked by CNQX/d-AP5, mecamylamine, and/or atropine but was blocked by application of the GABA_AR antagonist (gabazine, n = 21/21). In six cells, monosynaptic inputs were tested before gabazine application: (D₅) First, the response was abolished by the voltage-gated sodium channel blocker (TTX) and (D₆) subsequently revived by the voltage-gated potassium channel blocker (4-AP). (D₇) The recovered response was again fully blocked by gabazine (n = 5/5). Baselines and LED stimulations are indicated by red dashed lines and blue lines, respectively. (E) Histogram of the response latency with a bimodal behavior is shown. Red line is a fit of two weighted Gaussian distributions, which come close to the data. (F) Cell type-specific responses in MEC (Left) and LEC (Right) are depicted as violin plots that reflect the number of sampled cells of a given cell type and the number of responding cells within that group of cells (for details, see SI Materials and Methods). Red lines indicate the median of the distribution. Vertical dashed line in each plot segregates excitatory (Left) from inhibitory (Right) cell types. Numbers above the violin plots denote responding and patched cells for every cell type. (G) The relative frequencies of putative mono- (red), mono-/poly- (blue), and putative polysynaptic (yellow) GABA_AR-mediated responses, estimated from the latencies of the responses obtained from D₁ experimental conditions, are shown by stacked bar graphs for all cell types in MEC (Left) and LEC (Right). Error bars indicate mean ± SEM. ***P ≤ 0.0001. [Scale bars: (A) 150 µm and (C) 10 µm.]

We pursued this idea and tested electrophysiologically whether optogenetic stimulation of cholinergic axons in EC directly evokes GABA_AR-mediated responses. Cells were voltage-clamped at −50 mV, and a single 5-ms LED stimulus was applied. Resulting IPSCs were preferentially elicited in LI and LII pSOM⁺ and p5HT₃R⁺ interneurons (Fig. 4F) and were not blocked by CNQX/d-AP5 (n = 12/12; amplitude comparison P > 0.05), mecamylamine (n = 21/21; amplitude comparison P < 0.05, increased amplitude), or atropine (n = 17/17; amplitude comparison P > 0.05), but only by gabazine (Fig. 4 D₁–D₇, GABA_AR antagonist; n = 21/21). In six cells, we tested whether the evoked IPSCs resulted from monosynaptic input. We employed consecutive pharmacological application of tetrodotoxin (TTX; voltage-gated sodium channel blocker) and 4-aminopyridine (4-AP; voltage-gated potassium channel blocker), thereby allowing selective optogenetic depolarization only in ChR2-expressing cells (Fig. 4 D₅ and D₆, 6/6 cells). Finally, GABA_AR-mediated responses were verified by gabazine application (Fig. 4D₇, 5/5 cells; one cell was lost during the wash in process of gabazine).

Next, we tested if all recorded responses (n = 69) could be subdivided into mono-/polysynaptic responses based on the latency of pharmacologically tested monosynaptic responses (n = 6). The mean latency of confirmed monosynaptic inputs was found to be 3.62 ± 0.24 ms and was significantly faster than the overall response latency 4.89 ± 0.18 ms (P < 0.05). Moreover, the overall latency distribution showed a partially overlapping bimodal distribution (Fig. 4E, Akaike Information Criterion smallest for a bimodal mixed Gaussian model; uni- to decamodal tested). Thus, we used a k-mean clustering algorithm (run 1,000 times) to divide the dataset into three groups: putative monosynaptic (41%; 28/69 cells; latency: ≤4.1 ms), putative polysynaptic (28%; 19/69 cells; latency: ≥5.8 ms), and either mono-/polysynaptic (32%; 22/69 cells; latency: between 4.1 and 5.8 ms). The mean latency, rise, and decay time constants, as well as the median peak amplitude, and charge were 3.40 ± 0.10 ms, 2.11 ± 0.26 ms, 9.35 ± 1.01 ms, 6.3 [4.2, 12.3] pA, 0.082 [0.033, 0.137] pC, and 6.53 ± 0.14 ms, 3.70 ± 0.44 ms, 16.93 ± 2.72 ms, 2.7 [2.1, 4.9] pA, 0.037 [0.016, 0.115] pC for putative mono- and polysynaptic responses, respectively. Not only was the latency shorter for putative monosynaptic responses (P < 10⁻²³), but the rise and decay time constant and the amplitude were also larger (P < 0.01 for all) compared with putative polysynaptic responses. This difference in amplitude might account for the divergent estimates for monosynaptic responses based on the pharmacological approach or latency. Thus, in all pharmacologically tested cells, responses were monosynaptic. However, a bias was introduced, as all six cells had been selected for high amplitude to ensure that the frequently observed small amplitude following 4-AP revival is still detectable. Furthermore, the bimodal latency distribution (Fig. 4E) might also reflect presynaptic differences in evoked GABA release from MS/DBB axonal terminals. Using latency as a criterion resulted in an approximation of 41–73%. Additionally, we found a clear tendency of more frequent putative monosynaptic responses in interneurons (22/45) compared with excitatory cells (6/24; P > 0.05), which was not the case for putative polysynaptic responses (Fig. 4G, P > 0.25).

MS/DBB Cholinergic Cells Elicit Differential Responses in Specific EC Neuron Populations.

Next, we questioned whether nAChR-, hyper-/depolarizing mAChR-, and mono-/polysynaptic GABA_AR-mediated responses occurred within the same neuron. Fig. 5 A₁–A₃ shows sample current traces from a cell in which both nAChR- and GABA_AR-mediated EPSC and IPSC, respectively, were observed. Similarly, Fig. 5 B–F shows sample current traces depicting different response combinations. Note the individual time-locked EPSCs and IPSCs corresponding to nAChR- and GABA_AR-mediated responses, respectively, observed at 5-Hz LED stimulation (Fig. 5 A₃, C₃, E₃, and F₃). Fig. 5G shows sample current traces of a cell responding to nAChR and GABA_AR activation. At −50 mV, a clear and distinct biphasic response mediated via both nAChR and GABA_AR was observed (Fig. 5G₂). The nAChR-mediated response was abolished not by atropine but only by mecamylamine, and the GABA_AR-mediated response was abolished only after the application of gabazine (Fig. 5G), thus reconfirming the specificity of the responses. The relative frequency of the different responses elicited in the indicated cell types of MEC and LEC is shown in Fig. 5H. The estimated overall response rate of LI (LII) cells was higher than 55% (40%) in MEC and more than 25% (35%) in LEC. Please note the more restricted repertoire of response patterns of excitatory cells in contrast to the large variety of response combinations in interneurons, especially in LI and LII p5HT₃R⁺ interneurons.

Fig. 5.

EC neurons exhibit different receptor-mediated response combinations. (A–F) Combinations of nAChR-, GABA_AR-, and hyperpolarizing and depolarizing mAChR-mediated responses. Every column depicts measurements from a single cell (A and F—p5HT₃R⁺ interneuron; B—MFC; C—pPV⁺ interneuron; D—SC; E—LI SBC). Representative average current traces of (A₁–F₁) nicotinic (cells held at −70 mV), (A₂–F₂) GABAergic (cells held at −50 mV), and (A₃–F₃) hyperpolarizing or depolarizing muscarinic responses (cells held at −50 mV). In A₃–F₃, Savitzky–Golay filtered data are shown in black. In A₃, C₃, E₃, and F₃, note the individual time-locked EPSCs/IPSCs (seen in overlaid gray traces) corresponding to nAChR/GABA_AR-mediated responses. Baselines and LED stimulations are indicated by red dashed lines and blue lines, respectively. (G) Representative average current traces of a cell, in which nAChR responses (recorded at −70 mV) and GABA_AR-mediated responses (recorded at −50 mV) are elicited. This cell was identified as a p5HT₃R⁺ interneuron. (G₁) The nAChR-mediated response is not blocked by atropine but is completely blocked by mecamylamine. (G₂) The GABA_AR-mediated response is blocked neither by atropine nor by mecamylamine but is blocked by gabazine. (H) Stacked bar graphs depicting the relative response frequency of individual cell types, in which either a single response or combinations of responses in superficial neurons of MEC (Top) and LEC (Bottom) was detected. The number of patched cells for every cell type is mentioned above. GABA_AR-mediated responses include both mono- and polysynaptic responses. Responses are specified based on the receptors mediating the corresponding currents. mAChR-dep, depolarizing mAChR-mediated response; mAChR-hyp, hyperpolarizing mAChR-mediated response.

Discussion

Using whole-cell patch-clamp recordings combined with optogenetics, we demonstrate here that the synaptic release of ACh from MS/DBB neurons selectively recruits specific subclasses of superficial EC neurons via the activation of different AChRs. Viral tracing experiments revealed dense MS/DBB cholinergic projections in the superficial layers of MEC and to a lesser extent in LEC. Upon optogenetic stimulation of these axons locally in EC, we observed both fast nAChR- and slow mAChR-mediated responses in LI/LII neurons. While pSOM⁺ and p5HT₃R⁺ interneurons are depolarized via the activation of mAChRs and/or nAChRs, principal cells and pPV⁺ are hyperpolarized via the activation of mAChRs. In a fraction of cholinergic axons, light stimulation led to corelease of GABA, which activates GABA_ARs and evokes inhibitory currents predominantly in LI/LII non-FS interneurons (pSOM⁺ and p5HT₃R⁺).

How can the plethora of MS/DBB cholinergic actions affect EC neuronal activity? First, we observed depolarizing nAChR-mediated responses specifically in p5HT₃R⁺ interneurons and not in pPV⁺ and pSOM⁺ interneurons. This scenario is akin to what was reported for the cortex (45, 47, 48). Notably, nAChR-activated LI and LII p5HT₃R⁺ interneurons in the cortex were repeatedly shown to inhibit preferentially, if not exclusively, GABAergic interneurons (39, 40, 49). Some studies identified these GABAergic interneurons as pSOM⁺ (50, 51) or pPV⁺ interneurons (48, 52). Based on our results, we envisage a scenario in which LI and LII p5HT₃R⁺ interneuron activation via nAChRs causes disynaptic disinhibition in neighboring excitatory cells. Given the response kinetics, one can infer that the majority of evoked nicotinic responses were very likely mediated by fast α7-subunit–expressing nAChRs (48).

Second, it is striking that the slow mAChR-mediated depolarizations affect only GABAergic neurons in EC. Cholinergic excitation through mAChRs has been previously reported for hippocampal interneurons (53, 54). Although the function at the network level remains to be established, the implication of mAChRs in theta oscillations was often highlighted in the past (19, 55, 56). In CA1 hippocampal slices, mAChR-induced depolarization in interneurons is essential for generating and sustaining theta oscillations, presumably by changing their active conductances, making them responsive to theta frequency input (57). This in turn would allow interneurons to provide rhythmic inhibition onto PCs (58).

Finally, slow optogenetically induced mAChR-mediated hyperpolarizing responses were predominantly observed in LII excitatory neurons of EC. Of note, the same cells responded to bath/puff application of ACh/CCh with a membrane depolarization, as was often reported by others before (36, 37, 43–45). For instance, Widmer et al. (59) reported that hippocampal interneurons that are unresponsive to synaptically released ACh could be recruited following bath application of CCh. We can only speculate that this difference might result from different concentrations of agonist activating predominantly synaptic rather than extrasynaptic mAChRs or due to different ACh sources (e.g., MS/DBB versus local cholinergic interneurons).

The muscarinic depolarizing and hyperpolarizing responses were likely mediated by M1- and M2-type receptors, respectively (60, 61), that, as indicated by our functional results, appear to be differentially expressed in EC interneurons and excitatory cells. The distinct excitatory cell types in the superficial layers of MEC and LEC did not differ with respect to the strength and frequency of hyperpolarizing mAChR-mediated responses. Hence our results cannot explain the in vivo report of stronger theta rhythmicity in PCs compared with that in SCs (62), and other factors must be considered. Furthermore, while we did not observe a robust and direct synaptic mechanism selective for theta, the different kinetics, and therefore timing, of hyperpolarizing excitatory and depolarizing inhibitory cells could potentially enhance and/or restrict the overall network oscillation to a certain frequency (e.g., in the theta range).

There is ample evidence that septal lesions affect not only theta in both EC and hippocampus (24, 25) but also spatial learning (23). Less clear though appears the contribution of the different septal projections in sustaining theta and in particular the mechanisms underlying different forms of theta. MS inactivation and cholinergic blockade led to the differentiation of two forms of theta—namely, urethane- and movement-related theta (21, 63). Theta activity under urethane anesthesia is slower than that seen in freely moving animals, but most importantly, the two forms differ with respect to their response following cholinergic receptor blockade. Atropine, for instance, abolishes urethane-related theta and theta-correlated activity but has only a minor or no effect on movement-related theta (64, 65). Notably, cholinergic and noncholinergic components of theta have been observed both in hippocampus (31, 64, 65) and in MEC (16, 21).The latter may be supported by septal GABAergic neurons that project to both brain areas. Indeed, pharmacological interventions support the notion that septal GABAergic neurons may contribute to both forms of theta (29, 66, 67). These septal GABAergic neurons can be subdivided into pPV⁺ neurons that inhibit preferentially FS interneurons and calbindin-positive neurons that inhibit pSOM⁺ interneurons (13). We show here that MS/DBB cholinergic neurons can also corelease GABA, thereby adding a third GABAergic input that could potentially be involved in theta rhythmicity. The exact contribution of any of these projections to EC theta remains to be addressed by cell type-specific genetic manipulations.

The finding that some septal neurons corelease ACh and GABA further increases the number of neuronal projections exhibiting a dual neurotransmitter phenotype. The notion of a neuron releasing one fast-acting neurotransmitter has been questioned lately by several studies (46, 68–71). For example, supramamillary nucleus to dentate gyrus and basal ganglia to lateral habenula projections have the capacity to corelease glutamate and GABA (72, 73). More relevant to this study, some basal forebrain cholinergic neurons projecting to the neocortex (46) and hippocampus (74) were demonstrated to corelease GABA and ACh. Here we provide both anatomical and electrophysiological evidence that MS/DBB cholinergic neurons corelease GABA in superficial EC layers. It is very likely that the two neurotransmitters are packed into separate vesicles within the same axonal terminals, as was already shown for the supramamillary nucleus to dentate gyrus projections (72) and the MS to hippocampus projection (74). Even more interesting and functionally relevant is the question of if and how signaling via the two coreleased neurotransmitters is regulated. It is of note in this context that, at least for the basal ganglia to lateral habenula projection, it was shown that in a mouse model of depression, the balance of coreleased GABA and glutamate is altered and can be restored by antidepressant treatment (73).

Upon activation of cholinergic axons, we observed a similar response profile in LEC, but the response probability was much lower than that in MEC. This is in line with the anatomical difference in our study, as we found fewer cholinergic projections in LEC. Alternatively, AChR expression may be lower in LEC than MEC. In addition, the number of vAChT punctae per micrometer of axon was significantly lower in cholinergic fibers projecting to LEC than that on fibers targeting MEC, indicating fewer ACh release sites. Indeed, anatomical evidence suggests that the extent of cholinergic innervation can vary substantially between different cortical structures (20, 75). Nevertheless, tracing studies revealed an organization of basal forebrain neurons that supports concerted cholinergic regulation of cortical areas that are spatially apart but functionally linked (75). Given the physical closeness of MEC and LEC, but in particular the distinct yet overlapping function of the two brain structures in spatial memory, it is not surprising that upon anterograde labeling of septal cholinergic neurons, we could visualize axons in both areas. At this point, one can only speculate that the quantitative difference in cholinergic innervation might account for the more pronounced theta oscillations and the stronger theta modulation of neurons in MEC compared with LEC (17).

Yet another interesting and related question pertains to the issue as to what extent several downstream areas are simultaneously coordinated by the same septal projections. Thus, in the context of this study, one wonders whether an individual cholinergic projection targets both MEC and LEC. At least for septal GABAergic projections targeting the hippocampus and MEC, we showed that this was the case. Thus, injecting cholera toxin subunit B and FG in the hippocampus and MEC, respectively, we found septal GABAergic neurons that were positive for the two retrograde tracers (13).

In sum, based on viral tracing and optogenetically aided patch-clamp recordings, we here reported and characterized cell type-specific septal cholinergic synaptic input to MEC and LEC. Functional data at this level of analysis—that is, in small circuits—are a prerequisite if we are to understand the action of ACh in these brain areas in vivo. Thus, our data bridge the gap that links neuronal and behavioral levels and provide the necessary information to further probe the contribution and function of different ACh-mediated effects for the generation of rhythmic theta and gamma activity in EC.

Materials and Methods

Experiments were carried out on 8–11-wk-old male wild-type C57/BL6 and ChAT^Cre mice [cre recombinase expressed in all choline acetyltransferase (ChAT) positive cells; Chat^tm2(cre)Lowl/J, purchased from The Jackson Laboratory]. Animals were housed in a 12/12 h light/dark cycle with food and water ad libitum. All experiments were performed according to the German protection of animals act and after obtaining approval from the Regierungspräsidium Karlsruhe, Germany. All chemicals were obtained from Sigma-Aldrich unless mentioned otherwise.

Surgical Procedures.

Animals were anesthetized with isoflurane, mounted on a stereotaxic apparatus, and kept under isoflurane anesthesia during surgery. Briefly, for AAV injections, AAV-DIO-ChR2-mCherry and AAV-DIO-eGFP viral vectors were obtained from Penn Vector Core (76, 77). Eight-week-old ChAT^Cre mice were subjected to 300 nL viral injections in the MS/DBB as reported before (13). For retrograde tracer injections, 100 nL FG (0.5%) was injected in MEC or LEC of one hemisphere in 8-wk-old wild-type male mice. For this retrograde labeling, surgical procedures were followed as described previously (13, 14). Please see SI Materials and Methods for further information.

Immunohistochemistry, Fluorescence in Situ Hybridization, and Reconstruction of Biotin-Labeled Cells.

These methods involved standard procedures described in the SI Materials and Methods.

Imaging and Image Analysis.

Images for immunohistochemistry or in situ hybridization experiments were taken according to a standard protocol using a Zeiss LSM 700 confocal microscope (Carl Zeiss) and analyzed using ImageJ software (ImageJ; US National Institutes of Health). Please see SI Materials and Methods for further information.

In Vitro Electrophysiology.

Acute coronal, horizontal, and sagittal sections containing MS/DBB, LEC, and MEC, respectively, were 300 µm thick following slicing on a vibratome (Slicer HR2; Sigmann Elektronik). The different planes of sectioning were used to increase the area of the region of interest. Also the axonal projection patterns did not indicate that the plane of sectioning influenced the results (Fig. S1). Sections were incubated at ∼30–32 °C in carbogenated extracellular solution containing (in mM) 125 NaCl, 25 NaHCO₃, 1.25 NaH₂PO₄, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, and 25 glucose. For patch-clamp experiments, cells in EC were viewed with DIC optics, and recordings were performed using HEKA PatchMaster EPC 10 amplifier (HEKA). The injection site and the extent of axonal fibers were visualized based on neuronal mCherry expression. Recording pipettes with a tip resistance of ∼4–6 MΩ were filled with a low Cl⁻ potassium-based solution containing (in mM) 130 K-gluconate, 10 Na-gluconate, 10 Hepes, 10 phosphocreatine, 4 NaCl, 4 Mg-ATP, and 0.3 GTP, and sometimes 3 mg/mL biocytin with pH adjusted to 7.2 with KOH.

The LED optic fiber (200 µm in diameter, cat. no. M470F1; Thorlabs) was placed above LI to stimulate long-range cholinergic projections with blue light (wavelength, 470 nm). After obtaining a whole-cell configuration, a “test pulse” (TP) and “firing pattern” (FP) were recorded. TP and FP were analyzed offline by a custom-written MATLAB script (The Mathworks) to obtain passive and active properties of the cell.

To test for nicotinic receptor (nAChR)-mediated responses, cells were voltage-clamped at −70 mV and a single 5-ms, 470-nm LED stimulus was applied to depolarize ChR2-containing axonal fibers. To test for muscarinic receptor (mAChR)-mediated responses, cells were voltage-clamped at −50 mV and ten 5-ms LED stimuli at 5 Hz (in some cases also at 2, 10, and 20 Hz) were applied. To test for indirect and direct GABAergic responses upon cholinergic fiber stimulation, cells were voltage-clamped at −50 mV and a single 5-ms, 470-nm LED stimulus was given. Please see SI Materials and Methods for further information regarding this section.

Identification of Cell Types.

LI/LII EC excitatory and inhibitory neurons were classified based on their morphological characteristics and/or electrophysiological properties similar to previous descriptions (12–14, 38–40, 78). Please see SI Materials and Methods for further information.

Violin Plots for the Visualization of Proportions.

To get an estimation of the response probability, we used a Bayesian approach. Please see SI Materials and Methods for further information.

Quantification and Statistical Analysis.

Standard statistical tests were used to analyze the data. Please see SI Materials and Methods for further information.