1. Neuroscience
Download icon

Somatostatin interneurons activated by 5-HT2A receptor suppress slow oscillations in medial entorhinal cortex

  1. Roberto de Filippo  Is a corresponding author
  2. Benjamin R Rost
  3. Alexander Stumpf
  4. Claire Cooper
  5. John J Tukker
  6. Christoph Harms
  7. Prateep Beed
  8. Dietmar Schmitz  Is a corresponding author
  1. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Neuroscience Research Center, Germany
  2. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Cluster of Excellence NeuroCure, Germany
  3. German Centre for Neurodegenerative Diseases (DZNE), Germany
  4. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Department of Experimental Neurology, Germany
  5. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Center for Stroke Research Berlin, Germany
  6. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Einstein Center for Neurosciences Berlin, Germany
Research Article
  • Cited 2
  • Views 1,080
  • Annotations
Cite this article as: eLife 2021;10:e66960 doi: 10.7554/eLife.66960

Abstract

Serotonin (5-HT) is one of the major neuromodulators present in the mammalian brain and has been shown to play a role in multiple physiological processes. The mechanisms by which 5-HT modulates cortical network activity, however, are not yet fully understood. We investigated the effects of 5-HT on slow oscillations (SOs), a synchronized cortical network activity universally present across species. SOs are observed during anesthesia and are considered to be the default cortical activity pattern. We discovered that (±)3,4-methylenedioxymethamphetamine (MDMA) and fenfluramine, two potent 5-HT releasers, inhibit SOs within the entorhinal cortex (EC) in anesthetized mice. Combining opto- and pharmacogenetic manipulations with in vitro electrophysiological recordings, we uncovered that somatostatin-expressing (Sst) interneurons activated by the 5-HT2A receptor (5-HT2AR) play an important role in the suppression of SOs. Since 5-HT2AR signaling is involved in the etiology of different psychiatric disorders and mediates the psychological effects of many psychoactive serotonergic drugs, we propose that the newly discovered link between Sst interneurons and 5-HT will contribute to our understanding of these complex topics.

Introduction

5-HT is one of the most important neuromodulators in the central nervous system. Projections originating from the Raphe nuclei, the brain-stem structure that comprises the majority of 5-HT releasing neurons in the brain, innervate all cortical and sub-cortical areas (Descarries et al., 2010). 5-HT levels in the brain are closely linked to the sleep-wake cycle: the activity of serotonergic raphe neurons is increased during wakefulness, decreased during slow-wave sleep (SWS) and virtually silent during REM sleep (McGinty and Harper, 1976; Oikonomou et al., 2019; Unger et al., 2020). Cortical activity is also influenced by the behavioral state of the animal: SWS is generally associated with ‘synchronized’ patterns of activity, characterized by low-frequency global fluctuations, whereas active wakefulness and REM sleep feature ‘desynchronized’ network activity, in which low-frequency fluctuations are absent. The shifting of cortical networks between different patterns of activity is controlled, at least in part, by neuromodulators (Tukker et al., 2020; Lee and Dan, 2012). For instance, Acetylcholine (ACh) can profoundly alter cortical network activity by inducing desynchronization via activation of Sst interneurons (Chen et al., 2015). However, ACh is not solely responsible for suppressing cortical synchronized activity, as lesions of cholinergic neurons are not sufficient to abolish desynchronization (Kaur et al., 2008). On the other hand, blocking ACh and 5-HT transmission at the same time causes a continuous ‘synchronized’ cortical state, even during active behavior, thus suggesting that 5-HT plays an important role in mediating transitions between different network states (Vanderwolf and Baker, 1986). In agreement with this line of thought, electrical and optogenetic stimulation of the Raphe nuclei reduce low frequency (1–10 Hz) power in the cortex, implying a reduction in neuronal synchronization at these frequencies (Puig et al., 2010; Grandjean et al., 2019). Moreover, optogenetic stimulation of serotonergic neurons is sufficient to awaken mice from SWS (Oikonomou et al., 2019). These data suggest a relationship between 5-HT levels and patterns of cortical activity (Lee and Dan, 2012; Harris and Thiele, 2011). The exact mechanism by which 5-HT modulates network activity in the cortex however, is still not fully understood.

Here, we used electrophysiological techniques together with pharmacology, optogenetics, and pharmacogenetics to investigate the effect of 5-HT on slow oscillations (SOs), a network oscillation characterized by synchronized transitions (<1 Hz) between periods of high activity (upstates) and relative quiescence (downstates) (Steriade et al., 1993; Neske, 2015; Isomura et al., 2006). SOs are a global phenomenon observed throughout the cerebral cortex and are considered to be the default emergent activity of cortical networks during SWS and anesthesia (Neske, 2015; Sanchez-Vives et al., 2017; Wolansky et al., 2006). We performed our experiments in the medial entorhinal cortex (mEC), a region where SOs can be recorded both under anesthesia and in acutely prepared cortical slices (Tahvildari et al., 2012; Beed et al., 2020). Pyramidal neurons located in L3 of mEC provide the excitatory drive underlying each upstate (Namiki et al., 2013; Beed et al., 2020). Similarly to other mammalian cortical areas, mEC comprises different types of inhibitory GABAergic neurons that can be grouped into three main classes according to immunoreactivity: parvalbumin (PV), somatostatin (Sst), and 5-HT3 (Miao et al., 2017; Rudy et al., 2011). Most PV neurons target the soma and the spike initiation zone, have low input resistance and minimal spike frequency adaptation. Sst neurons are divided into two groups, one showing features similar to PV interneurons and a second one (i.e. Martinotti cells) that, in contrast, tend to form synapses onto the dendritic trees of their target cells, have high input resistance and show a considerable adaptation. 5-HT3 neurons are usually located in the superficial layers, have high input resistance and mixed adaptation. While these different classes of interneurons are all depolarized during upstates, PV interneurons receive decidedly the strongest excitation (Tahvildari et al., 2012; Neske et al., 2015). Recurrent excitation and temporal summation of inputs contribute to the transition from downstate to upstate (Sanchez-Vives et al., 2017; Tukker et al., 2020). This excitation propagates to L5 and differentially entrains L5a and L5b excitatory neurons. L5a neurons do not participate in SOs, whereas L5b neurons are steadily synchronized. Both intrinsic and synaptic mechanisms have been implicated in upstate termination (Neske, 2015; Tukker et al., 2020). Activity-dependent K+ channels decrease excitability of neurons over time causing a generalized reduction of facilitation (Neske, 2015; Harris and Thiele, 2011). At the same time, blockage of GABAB receptors significantly extends upstate duration (Craig et al., 2013) and inhibitory drive has been observed to increase during upstate termination (Lemieux et al., 2015). Sst interneurons, and in particular Martinotti cells, characterized by strongly facilitating synapses (Beierlein et al., 2003; Gibson et al., 1999) have been proposed as an important source of inhibition in the termination of upstates (Melamed et al., 2008; Krishnamurthy et al., 2012).

Our results show that (±)3,4-methylenedioxymethamphetamine (MDMA) and fenfluramine (Fen), two substitute amphetamines that induce robust 5-HT release, suppress SOs in anesthetized mice and, concurrently, activate a small group of neurons characterized by an intermediate waveform shape. Using cortical slices in vitro, we demonstrate that 5-HT inhibits SOs in the mEC and that Sst interneurons, activated by 5-HT2AR, are involved in the modulatory effect of 5-HT. While previous studies have shown that parvalbumin (PV) interneurons are excited by 5-HT2AR (Puig et al., 2010; Athilingam et al., 2017), our results identify cortical Sst interneurons as novel targets of the 5-HT neuromodulatory system via 5-HT2AR.

Results

Pharmacological release of 5-HT inhibits SOs in anesthetized mice

We investigated the effect of 5-HT on network activity in anesthetized mice using multisite silicon microelectrodes placed in the mEC L3, a region located in the medial temporal lobe interconnected to a variety of cortical and subcortical areas, including the Raphe nuclei (Figure 1A; van Strien et al., 2009; Muzerelle et al., 2016). Under urethane anesthesia, EC, like the rest of the cortex, displays SOs (Figure 1B). As expected, we found that upstates were present synchronously in the local field potential (LFP) of all the recording channels (Figure 1—figure supplement 1), and that every upstate coincided with large increases in population spiking activity (Figure 1B). 5-HT does not cross the blood brain barrier (Hardebo and Owman, 1980), therefore, to understand the effect of 5-HT on SOs we used MDMA, a potent presynaptic 5-HT releaser and popular recreational drug (Green et al., 2003) that has shown promising results in the treatment of post-traumatic stress disorder (PTSD) (Inserra et al., 2021).

Figure 1 with 3 supplements see all
MDMA/Fen inhibit SOs in anesthetized mice.

(A) Immunocytochemical analysis of an ePet-YFP mouse showing serotonergic fibers in medial entorhinal cortex, horizontal slice (M = medial, L = lateral, P = posterior, A = anterior). Scale bar: 20 µm. (B) LFP (top) and instantaneous population activity (bottom) of a representative in vivo recording during SOs (spikes/s units in thousands), cyan rectangles represent detected upstates. (C) 3D visualization of the microelectrode location of the recording shown in E. EC represented in gray. (D) Left: microelectrode tracks (red) of the recording shown in (E). Right: EC position represented in green. (E) Top: Cyan lines represent detected upstates. Bottom: LFP (black) and average upstate incidence per minute (cyan). Pink dotted line represents MDMA application time. (F) Fourier transformation and (G) instantaneous population activity for the recording shown in E. (H) Mean upstate incidence after saline (control), Fen or MDMA application (control: n = 5, Fen: n = 6, MDMA: n = 7; p<10−4, unpaired t test with Holm-Šidák correction). (I) Mean normalized spike rate after saline (control), Fen or MDMA application (control: n = 5, Fen: n = 6, MDMA: n = 7; p<10−4, unpaired t test with Holm-Šidák correction).

Intraperitoneal injections of MDMA (1.25 mg/kg) caused a strong suppression of upstate incidence (Figure 1E–H), a decrease in power of low frequencies (1–20 Hz) (Figure 1F), and a reduction of population spiking activity (Figure 1G–I). In addition to 5-HT, MDMA has been shown to cause the release of other neuromodulators, such as dopamine and noradrenaline (NE), although to a much lesser extent (Green et al., 2003). To test whether the effect of MDMA was mediated specifically by 5-HT, we repeated the experiment using Fen (5 mg/kg), a more selective 5-HT releaser (Rothman and Baumann, 2002; Heifets et al., 2019; Baumann et al., 2000). Intraperitoneal injection of Fen had a comparably strong suppressive effect on both the occurrence of upstates and population spiking activity (Figure 1H–I). Given the similarity of the observed effects, we grouped the results of Fen and MDMA application and found that both of these drugs significantly decreased LFP power in the delta and theta frequency ranges (Figure 1—figure supplement 2). Furthermore, the duration and area of upstates were also significantly reduced (Figure 1—figure supplement 3). These results point to a likely involvement of 5-HT in modulating ongoing oscillatory activity and suppressing low-frequency fluctuations.

MDMA and Fen activate a subgroup of cortical neurons

In addition to the LFP signal, we recorded the activity of 355 single units within the EC. Because of the similar effect of MDMA and Fen on spike rates (Figure 1I), we pooled all units recorded in both types of experiments. We found that drug injections differentially affected spiking rates (Figure 2A) of recorded units: while spiking decreased in most units (‘non-activated’), a small group of units (‘activated’) responded in the opposite fashion (n = 31/355, 8.7%).

Figure 2 with 2 supplements see all
Divergent unit responses to MDMA/Fen application.

(A) Spike rate of the activated units versus all the other units during MDMA/Fen application (activated: n = 31, Non-activated: n = 324). (B) Top: TP latencies color-coded by group. Middle: cumulative distribution of TP latencies (Kolmogorov-Smirnov test, p Activated vs Int <10–4, p Activated vs Exc <10–4). Bottom: bar plot representing probability distribution of TP latencies, on the right y axis dashed line representing the percentage of ‘activated’ units per TP latency bin. (C) Distribution of units according to trough-to-peak (TP) latencies and repolarization time. Units were classified as putative interneurons (Int, blue) and putative excitatory neurons (Exc, dark gray) according to a threshold at 0.55 ms; activated units (red) could belong to either group but were mostly intermediate as shown by the covariance (2 STD) of each group (Ellipses). Units recorded during control experiments are represented by empty circles. (D) Waveforms of recorded units (n = 355). Units were divided into ‘putative excitatory’ (black) and ‘putative inhibitory’ (blue) neurons according to TP latencies. Units activated by either MDMA or Fen application are represented in red. Inset shows the average waveform for each group. Scale bars: 0.5 ms.

We then calculated the trough-to-peak (TP) latency of the spike waveforms, which has been consistently used as a metric to classify units. In accordance with previous studies (Senzai et al., 2019; Roux et al., 2014), we found a clear bimodal distribution of TP latencies distinguishing putative excitatory (Exc) and fast-spiking inhibitory (Int) groups. Analysis of cross-correlograms confirmed the inhibitory nature of a subset of putative FS units (Figure 2—figure supplement 1; Barthó et al., 2004). The cumulative distribution of TP latencies of ‘activated’ units was significantly different from both Exc and Int groups (Figure 2B). Specifically, the average TP latency of ‘activated’ units was situated in between the Int and Exc groups (Figure 2B–D, Figure 2—figure supplement 2), possibly suggesting a non-FS interneuron identity (Trainito et al., 2019; Kvitsiani et al., 2013). Notably, the intermediate TP latency of ‘activated’ units is not the result of an equal distribution between high and low values. Units with intermediate TP latency (between 0.4 and 0.6 ms) are unique in showing a 40–80% likelihood of being ‘activated’ by drug application (Figure 2B).

5-HT suppresses SOs and activates Sst interneurons via 5-HT2AR in vitro

To understand the mechanism underlying the suppression of SOs by MDMA/Fen, we performed in vitro experiments combining electrophysiology and pharmacology. First, we recorded simultaneously from up to four neurons in the superficial layers of the EC (Figure 3A). Brain slices were perfused with an extracellular solution containing Mg2+ and Ca2+ in concentrations similar to physiological conditions. With this method we could reliably detect SOs reminiscent of the in vivo network activity (Tahvildari et al., 2012). Release of 5-HT in vitro, induced by Fen application (200 µM), caused a suppression of SOs similar to what we observed in vivo (Figure 3—figure supplement 1). Likewise, application of low concentrations of 5-HT (5 µM) resulted in the suppression of SOs (Figure 3B,C). This effect was highly consistent across different slices and was readily reversible (Figure 3D). Similarly to spontaneous upstates, electrically evoked upstates (Neske et al., 2015) were also suppressed by 5-HT (Figure 3—figure supplement 2). Increasing the stimulation intensity did not rescue upstate generation, indicating that a lack of excitation alone cannot explain the suppressive effect of 5-HT on SO. In accordance with these results, we decided to focus solely on 5-HT, however, we acknowledge that other neurotransmitters might be involved in the effect of MDMA/Fen we observed in vivo.

Figure 3 with 9 supplements see all
5-HT suppresses SOs and activates Sst interneurons.

(A) Biocytin staining of four simultaneously recorded cells shown in (B) WFS1 expression (in red) delimits L2/3 border. (B) Intracellular recordings showing synchronous upstate events in four simultaneously recorded cells before (left) and after (right) 5-HT application. Scale bars: 1: 7.5 mV, 2: 25 mV, 3: 25 mV, 4: 10 mV; 10 s. (C) Upstate raster plot before and after 5-HT application, orange box represents 5-HT application (n = 17, p<10−4, Wilcoxon signed rank test). (D) Representative recording showing the temporary inhibitory effect of 5-HT on SOs in two simultaneously recorded cells. Scale bars: 5 min, 20 mV. (E) PCA projection plot of all the cells recorded (n = 48). Cells are color-coded according to group identity: Exc (black), FS (light blue), or LTS (green). Typical voltage responses to current injection (−150 and +250 pA) are plotted for each group. Inset shows the average spike waveform for each group. (F) Representative recording of an excitatory (black) and a low-threshold (green) neuron simultaneously recorded during 5-HT application. Scale bars: 10 mV, 30 s. (G) Average change of RP before and after 5-HT application, across excitatory, fast-spiking and low-threshold neurons (Exc: n = 34, FS: n = 6; LTS: n = 9; p<10−4, unpaired t test with Holm-Šidák correction). (H) Biocytin staining of cells recorded in Sst-tdTomato mouse. Biocytin in green, tdTomato in red. Scale bar: 50 µm. (I) Representative recording of a Sst interneuron during 5-HT application. Scale bars: 10 mV, 30 s. (L) Average RP of Sst interneurons during 5-HT (red) and ketanserin + 5-HT (blue) application, orange bar represents 5-HT (5-HT: n = 19, ketanserin + 5-HT: n = 22).

Suppression of activity can have either an intrinsic or synaptic origin (Turrigiano, 2011). A substantial subset of EC excitatory neurons is known to express 5-HT1A receptor (5-HT1AR) and hyperpolarize upon 5-HT application via activation of G protein-coupled inwardly rectifying potassium (GIRK) channels (Schmitz et al., 1998; Chalmers and Watson, 1991). The suppression of SOs by 5-HT, however, was not influenced by blocking 5-HT1AR (WAY 100635, 100 nM) (Figure 3—figure supplement 3A,E). Blocking 5-HT3R (tropisetron, 1 µM), the receptor that characterizes one of the three main groups of interneurons, also did not have any impact on SOs suppression induced by 5-HT. Similarly, application of the 5-HT3R agonist m-CPBG (50 µM) did not have any effect on SOs (Figure 3—figure supplement 4). In contrast, blocking 5-HT2AR with the selective antagonist ketanserin (1 µM) (Preller et al., 2018) strongly reduced the suppression power of 5-HT on SOs from 95 ± 4% to 57 ± 10.1% (Figure 4—figure supplement 1B,E). The remaining suppression can be possibly explained by the activation of 5-HT1AR on excitatory cells, as is reflected by the reduced spiking activity of putative excitatory cells (Figure 3—figure supplement 5).

Selective activation of 5-HT2AR by α-methyl-5-HT (5 µM) mimicked the suppression of SOs observed after 5-HT wash-in (Figure 3—figure supplement 3C,E). Together, these results point to the importance of 5-HT2AR in the suppression of SOs. 5-HT2AR activation is known to cause an increase in intracellular calcium and consequent depolarization of the resting potential (RP) (Nichols and Nichols, 2008). Accordingly, after 5-HT application, we found that a small group of neurons was depolarized (n = 6/48, 12.5%) (Figure 3—figure supplement 6B–C). Using a soft clustering approach with six electrophysiological parameters (see ‘Materials and methods’) we divided the recorded cells into three groups: Excitatory (Exc), fast-spiking (FS), and low-threshold spiking (LTS). Strikingly, the cells excited by 5-HT belonged exclusively to the LTS group (Figure 3G, Figure 3—figure supplement 6). A substantial proportion of LTS neurons express Sst (Tremblay et al., 2016; Gibson et al., 1999); therefore, we performed targeted patch-clamp recordings using a mouse line expressing tdTomato specifically in Sst-expressing interneurons (Figure 3H). A subgroup of Sst interneurons are FS cells (Urban-Ciecko et al., 2018), only LTS Sst interneurons were considered in the following analysis. Sst interneurons depolarized upon 5-HT application (n = 19, ∆RP: 7.5 ± 1.23 mV) (Figure 3I–L) and in some cases continued to spike while SOs were suppressed (n = 8/17, 47.05%, mean spiking rate = 3.03 ± 0.39 spikes/s). This effect was blocked by ketanserin (n = 22) (Figure 3L). We confirmed the presence of 5-HT2AR in Sst interneurons using immunohistochemistry in mice expressing EGFP under the 5-HT2AR promoter. In accordance with a previous study we found the majority of 5-HT2AR positive cells in the deep layers of EC, with a peak in L6 (Weber and Andrade, 2010). We observed that 11.8 ± 2.9% of the 5-HT2AR positive cells in EC colocalized with Sst (n = 7 mice) (Figure 3—figure supplement 9). These results suggest that Sst interneurons may provide the synaptic inhibition required for the suppression of SOs.

Sst interneurons mediate the suppression of SOs by 5-HT in vitro

To evaluate the contribution of Sst interneurons to the 5-HT-mediated silencing of SOs we used an opto- and pharmacogenetic approach. First, we transgenically expressed channelrhodopsin-2 (ChR2) in Sst interneurons (Figure 4A). Light-stimulation of ChR2- expressing Sst interneurons in vitro suppressed SOs consistently (Figure 4D–F). Expectedly, upstate-associated spiking was also diminished (Figure 4C,E,F). At the end of the light stimulation, spontaneous upstates immediately reoccurred (Figure 4G–H), in line with a critical role of Sst interneurons in the modulation of SOs (Fanselow et al., 2008; Funk et al., 2017; Niethard et al., 2018). We acknowledge that activation of PV interneurons can cause a similar suppression (Zucca et al., 2017). While this experiment establishes the ability of Sst interneurons to suppress SOs, it does not causally link Sst interneuron activation to the suppression of SOs induced by 5-HT. Therefore, we generated a transgenic mouse line carrying a Cre-conditional expression cassette of the pharmacogenetic silencer hM4Di (Figure 4—figure supplement 1; Armbruster et al., 2007). Homozygous Cre-conditional hM4Di transgenic mice and Sst-Cre mice were bred to obtain heterozygous Sst-Cre/hM4Di offspring, which allow specific inhibition of Sst interneuron activity using Clozapine-N-Oxide (CNO). Following application of 5-HT we observed a strong reduction of upstate incidence, which was partially restored by subsequent application of CNO (Figure 5A–B).

Figure 4 with 1 supplement see all
Sst interneurons activation suppresses SOs.

(A) Experimental protocol: Sst interneurons expressing ChR2 are activated by light during intracellular recording of L3 neurons in EC. (B) Representative recordings from a L3 neuron during Sst interneuron activation. Scale bars: 10 mV, 0.5 s. (C) Spikes raster (top) and density plot (bottom) during light stimulation. (D) Upstate raster (top) and density plot (bottom) during light stimulation. (E) Left: spike frequency during baseline light stimulation (n = 14; p<0.001, Wilcoxon signed rank test). Right: upstate incidence during baseline and light stimulation (n = 14; p<0.001, Wilcoxon signed rank test). Patches represent 95% confidence interval, lines represent standard deviation. (F) Left: spike probability polar plot during Sst interneurons light activation. Right: upstate probability polar plot during Sst interneurons light activation. Note the absence of both spiking activity and upstates during Sst interneurons activation.

Figure 5 with 2 supplements see all
Sst interneurons mediate the effect of 5-HT on SOs.

( A) Upstate raster plot during 5-HT and subsequent CNO application. Orange box represents 5-HT, purple boxes represent CNO. Note the appearance of upstates after CNO application. (B) Upstate incidence during 5-HT and 5-HT+CNO application (n = 15; p (baseline vs 5-HT)<10–4, p (baseline vs CNO)=0.0482, p5-HT vs CNO = 0.0405, Kruskal-Wallis test). Patches represent 95% confidence intervals, lines represent standard deviation.

Activation of 5-HT1AR on excitatory cells by 5-HT and the resulting decreased network excitation drive might account for the fact that the incidence of upstates did not completely return to the baseline level upon CNO wash-in. Additionally, in further experiments in which CNO was applied before 5-HT, the significant reduction in upstate incidence typically seen after 5-HT wash-in was not observed (Figure 5—figure supplement 1). CNO did not show any significant effect in both wild-type littermates and hM4Di-PV mice (Figure 5—figure supplement 2), indicating that the results observed are due to its specific effect on the activity of Sst interneurons. In summary, while activation of Sst interneurons either via 5-HT or directly by ChR2 suppresses SOs, the pharmacogenetic inactivation of Sst interneurons weakens the effect of 5-HT on SOs.

Discussion

In this study, we show that the substitute amphetamines MDMA and Fen suppress default cortical network oscillations in vivo in the mEC. Furthermore, using an opto- and pharmacogenetic approach in vitro, we demonstrate that Sst interneurons, activated by 5-HT2AR, contribute to this suppression.

Organization of cortical activity is brain state-dependent, ranging continuously from ‘synchronized’ to ‘desynchronized’ states (Harris and Thiele, 2011). SOs are on one end of this continuum, representing the prototypical synchronized state. Our results, in line with previous studies (Puig et al., 2010; Grandjean et al., 2019), show that 5-HT can suppress synchronized cortical activity; in addition, we identify Sst interneurons as contributing to this suppression.

Sst interneurons have been previously proposed to provide the inhibition necessary for the termination of upstates, due to their strongly facilitating synapses (Krishnamurthy et al., 2012; Melamed et al., 2008). Additionally, increased inhibition, as shown both in computational models and experimental data, can counteract temporal summation of inputs and reduce correlation due to tracking of shared inputs between inhibitory and excitatory populations (Sippy and Yuste, 2013; Renart et al., 2010; Stringer et al., 2016). In the case of Sst interneurons, this would especially impact summation in distal compartments of pyramidal cells, due to the axonic targeting of Sst cells onto pyramids. Sst interneurons are specifically known to be sufficient to cause desynchronization in V1 (Chen et al., 2015). While it is well known that Sst interneurons are potently excited by ACh in various cortical areas (Chen et al., 2015; Obermayer et al., 2018; Fanselow et al., 2008), including mEC (Desikan et al., 2018) and NE in frontal cortex (Kawaguchi and Kubota, 1998), our work identifies them as a novel target of 5-HT regulation via 5-HT2AR. Recently, it has been proposed that inhibitory interneurons play a key role in mediating the effect of ACh and NE on cortical state transitions (Cardin, 2019). Our results add a new level of complexity to this picture.

The excitation of Sst interneurons by 5-HT possibly contributes to the net inhibitory effect of 5-HT release observed in many cortical areas (Grandjean et al., 2019; Seillier et al., 2017; Azimi et al., 2020), and could explain why the inhibition strength is linearly correlated to 5-HT2AR expression (Grandjean et al., 2019). Giving further support to this idea, Sst interneurons in the somatosensory cortex show increased cFos levels following 5-HT2AR activation (Martin and Nichols, 2016). Previous studies have reported either 5-HT2AR-dependent inhibition or 5-HT2AR-dependent activation of interneurons in the prefrontal cortex (PFC) (Abi-Saab et al., 1999; Ashby et al., 1990; Athilingam et al., 2017), piriform cortex (Marek and Aghajanian, 1994; Sheldon and Aghajanian, 1990), cingulate cortex (Zhou and Hablitz, 1999), cochlear nucleus (Tang and Trussell, 2017), amygdala (Sengupta et al., 2017), olfactory bulb (Petzold et al., 2009; Hardy et al., 2005), visual cortex (Michaiel et al., 2019; Azimi et al., 2020), and hippocampus (Wyskiel and Andrade, 2016). However, none of these studies identified interneurons using molecular markers, so we do not exclude that different interneuron classes in other cortical areas might mediate the inhibitory downstream effects of 5-HT2AR. For example, in the PFC a subgroup of PV interneurons has been reported to be activated by this receptor (Athilingam et al., 2017; Puig et al., 2010). Furthermore, we know that PV neurons can reliably induce up to down state transitions (Zucca et al., 2017). 5-HT modulation is also involved in gain regulation. In the olfactory cortex, 5-HT has a selective subtractive effect on stimulus evoked firing (Lottem et al., 2016), and a recent study has shown in the visual cortex that the reduced gain of evoked responses is dependent on 5-HT2AR activation (Azimi et al., 2020). Intriguingly, Sst interneurons have been shown to regulate subtractive inhibition (Sturgill and Isaacson, 2015; Wilson et al., 2012).

5-HT levels in the brain, similarly to other neuromodulators, are synchronized to the sleep wake cycle, with higher levels present during the waking state (Oikonomou et al., 2019). How does this notion relate to activation of Sst interneurons by 5-HT2AR? A limitation of our study is the absence of data in naturalistic conditions, with anesthesia potentially being a confounding factor (Adesnik et al., 2012). We can nonetheless speculate that Sst interneurons should be more active during states with higher 5-HT levels (wake>SWS>REM). A previous study directly measuring the activity of various neuronal classes across different states seems to support this idea. Sst interneurons in the dorsal cortical surface display their highest activity during waking states, lower activity during SWS and lowest activity during REM (Niethard et al., 2016). Following pharmacological release of 5-HT in vivo, we observed increased spiking activity in a subgroup of neurons whose waveform features are compatible with Sst interneurons. Our in vitro data using 5-HT at a concentration commonly used in the field (Gorinski et al., 2019; Wang et al., 2016; Huang et al., 2009) shows that Sst interneurons can be activated by 5-HT2AR, however, recordings during natural sleep and wake conditions in combination with optotagging will be needed to demonstrate conclusively that Sst interneurons are activated by 5-HT in physiological conditions in vivo. Potential differences between physiological and pharmacological 5-HT release are especially of interest considering the reported efficacy of MDMA in the treatment of PTSD and its recently approved status as breakthrough therapy (Inserra et al., 2021; Mithoefer et al., 2019).

While our data relative to MDMA application supports a role of 5-HT in modulating SOs in vivo we cannot exclude that other neuromodulators might play a role. For example, it is known that dopamine can suppress SOs in vitro (Mayne et al., 2013). Fen, however, has been reported to selectively increase 5-HT concentration in the brain (Rothman and Baumann, 2002) and has been used previously to specifically disentangle the effect of MDMA on different neuromodulatory systems (Heifets et al., 2019). This, together with our in vitro results, suggests that 5-HT is most likely mediating the effect induced by MDMA/Fen.

Besides its involvement in various physiological brain processes, 5-HT is also associated with the etiology of various psychiatric disorders, as are Sst interneurons (Pantazopoulos et al., 2017; Lin and Sibille, 2015). Furthermore, 5-HT is linked to the psychological effect of many psychotropic drugs; specifically, 5-HT2AR activation has been reported to be essential for the psychological effects induced by various psychedelics (Nichols, 2016), and in the case of MDMA, has been linked to perceptual and emotional alterations (Liechti et al., 2000; Kuypers et al., 2018). Broadband reduction in oscillatory power, triggered by 5-HT2AR, seems to be linked to the subjective effect of serotonergic drugs (Carhart-Harris et al., 2016; Carhart-Harris and Friston, 2019) and has been consistently observed in humans and rodents following administration of MDMA (Frei et al., 2001; Lansbergen et al., 2011) or various other 5-HT2AR agonists (Kometer et al., 2015; Muthukumaraswamy et al., 2013; Carhart-Harris et al., 2016; Wood et al., 2012). The link between 5-HT2AR and perception is further supported by the fact that several routinely used antipsychotic drugs are potent 5-HT2AR antagonists (Marek et al., 2003; Meltzer, 1999). Although the most recent attempts to explain the psychological effects of 5-HT2AR activation focus on the increased spiking of cortical pyramidal neurons in the deep layers (Carhart-Harris and Friston, 2019; Nichols, 2016), our study suggests that Sst interneurons may also play a role. Sst interneurons, in contrast to PV interneurons, form synapses on the dendrites of their target cell (Tremblay et al., 2016). A wealth of evidence suggests that active dendritic processing in cortical pyramidal neurons has a critical influence on sensory perception (Takahashi et al., 2016; Murayama et al., 2009; Smith et al., 2013; Ranganathan et al., 2018) and, in accordance with their unique anatomical properties, Sst interneurons strongly influence dendritic computations and directly modulate perceptual thresholds (Takahashi et al., 2016).

We propose that the novel link between 5-HT2AR and Sst interneurons might help elucidate the mechanisms underlying a host of psychiatric disorders and contribute to our understanding of how serotonergic drugs exert their psychological effects.

Materials and methods

All experiments were conducted according to regulations of the Landesamt für Gesundheit und Soziales (Berlin [T 0100/03], Berlin [G0298/18]) and the European legislation (European Directive 2010/63/EU).

Animals

Data for the in vivo part of the study was collected from C57Bl/6 mice (aged 6–10 weeks). Data for the in vitro part was collected from C57Bl/6 (P10-P17), Sst-tdTomato (P10-P30), Sst-Chr2-EYFP (P10-P16), hM4Di-Sst (P10-P15), hM4Di-Sst (+/-) (P10-P15) and hM4Di-PV (P10-P15) mice. Immunostainings to localize 5-HT2AR were performed on 5-HT2AR-EGFP mice (P20-P90) and immunostainings to localize 5-HT fibers were performed on an ePet-YFP mouse (P35). Sst-Cre mice (RRID:IMSR_JAX:013044) have Cre recombinase targeted to the Sst locus; they were obtained from Jackson Laboratory (ME, USA). PV-Cre mice (RRID:IMSR_JAX:008069) have Cre recombinase targeted to the Pvalb locus; they were obtained from Charles River Laboratory (MA, USA). tdTomato mice (Ai9, RRID:IMSR_JAX:013044) were obtained from Jackson Laboratory. Chr2-EYFP mice (Ai32, RRID:IMSR_JAX:012569) were obtained from Jackson Laboratory. ePet-Cre mice (RRID:IMSR_JAX:012712) have Cre recombinase targeted to the Fev locus; they were obtained from Jackson Laboratory. 5-HT2AR-EGFP mice (RRID:MMRRC_010915-UCD) express EGFP reporter protein under the control of the Htr2a gene, they were obtained from the Mutant Mouse Resource and Research Centers (MMRRC, CA, USA). Generation of hM4Di mice is described in the paragraph ‘Generation of Cre-conditional hM4Di mice’. The animals were housed in a 12:12 hr light-dark cycle in singularly ventilated cages with ad libitum access to food and water. SOs in vitro recordings were performed on P12-P16 mice.

Generation of Cre-conditional hM4Di mice

Request a detailed protocol

We produced a transgenic mouse line carrying a Cre-conditional hM4Di expression cassette in the Rosa26 locus. The transgene construct was inserted by recombination-mediated cassette exchange (RMCE). RMCE relies on recombination events between attB and attP recognition sites of the RMCE plasmid and genetically modified acceptor embryonic stem (ES) cells, mediated by the integrase of phage phiC31 (Hitz et al., 2007). The RMCE construct is thereby shuttled into the Rosa26 locus of the ES cells, along with a Neomycin resistance cassette (Figure 1—figure supplement 2A). The acceptor cell line IDG3.2-R26.10–3 (I3) was kindly provided by Ralf Kühn (GSF National Research Centre for Environment and Health, Institute of Developmental Genetics, Neuherberg, Germany).

We subcloned a Cre-conditional FLEX (flip-excision) cassette (Schnütgen et al., 2003) into pRMCE, and inserted a strong CAG promoter (CMV immediate early enhancer/modified chicken β-actin promoter, from Addgene Plasmid #1378) in front of the FLEX-cassette to create pRMCE-CAG-Flex. The coding sequence of hM4Di-mKateT was inserted into the FLEX cassette in reverse orientation to the promoter (Figure 1—figure supplement 2A). Finally, a rabbit globulin polyA cassette including stop codons in every reading frame was placed downstream of the FLEX cassette, in the same direction as hM4Di, in order to prevent unintended transcriptional read-through from potential endogenous promoters. The construct was completely sequenced before ES cell electroporation.

Electroporation of the RMCE construct together with a plasmid encoding C31int was performed by the transgene facility of the ‘Research Institute for Experimental Medicine’ (FEM, Charité, Berlin) according to published protocols (Hitz et al., 2009; Hitz et al., 2007). Recombinant clones were selected by incubation with 140 µg/ml G418 for at least 7 days. To activate hM4Di expression by recombination of the FLEX switch, selected clones were further transfected transiently with pCAG-Cre-EGFP using Roti-Fect (Carl Roth, Karlsruhe, Germany). G418-resistant clones were analyzed by PCR for successful integration and recombination of the construct (Figure 1—figure supplement 2B), using the following primers:

  • GT001 PGK3’-fw: CACGCTTCAAAAGCGCACGTCTG;

  • GT002 Neo5’-rev: GTTGTGCCCAGTCATAGCCGAATAG;

  • GT005 PolyA-fw: TTCCTCCTCTCCTGACTACTCC;

  • GT006 Rosa3’-rev: TAAGCCTGCCCAGAAGACTC;

  • GT013 hM4Di3’rec-rev: CAGATACTGCGACCTCCCTA

After verification of correct integration and functional FLEX-switch recombination, we generated chimeras by blastocyst injection of I3 ES cells. Heterozygous offsprings were mated with a Flpe deleter mouse line in order to remove the neomycin resistance cassette by Flp-mediated recombination.

Mice homozygous for the Rosa-CAG-FLEX-hM4Di-mKateT allele are viable and fertile and show now obvious phenotype. Importantly, application of CNO to these mice does not induce any behavioral effects. Homozygous Cre-conditional hM4Di transgenic mice and Sst-Cre mice (Taniguchi et al., 2011) were maintained on a C57Bl/6 genetic background and were bred to obtain heterozygous Sst-Cre / hM4Di offsprings.

Drugs

Urethane (U2500, Merck), Fenfluramine ((+)-Fenfluramine hydrochloride, F112, Merck), 5-HT (Serotonin creatinine sulfate monohydrate, H7752, Merck), m-CPBG (1-(3-Chlorophenyl)biguanide hydrochloride, C144, Merck), tropisetron (Tropisetron hydrochloride,Y0000616, Sigma), WAY-100635 (W108, Merck), α-Methylserotonin (α-Methylserotonin maleate salt, M110, Merck), MDMA ((±)3,4-methylenedioxymethamphetamine, 64057-70-1, Merck), CNO (Clozapine N-oxide dihydrochloride, 6329, Tocris) were dissolved in water for in vitro application and in 0.9% normal saline for in vivo application. Ketanserin (Ketanserin (+)-tartrate salt, S006, Merck) was dissolved in Dimethyl sulfoxide (DMSO).

Surgery and in vivo recording

Request a detailed protocol

Before recordings mice were briefly anesthetized with isofluorane (2%) and then injected intraperitoneally with urethane (1.2 g/kg, Sigma Aldrich, Munich, Germany). The level of anesthesia was maintained so that hindlimb pinching produced no reflex movement and supplemental doses of urethane (0.2 g/kg) were delivered as needed. Upon cessation of reflexes the animals were mounted on a stereotaxic frame (Kopf Instruments, Tujunga, California), and body temperature was maintained at 38°C. The scalp was removed, and the skull was cleaned with saline solution. A craniotomy was performed at +3 mm ML, −3 mm AP, +3.25 mm DV to reach mEC.

Extracellular recordings from EC (Figures 12) were performed using a Cambridge Neurotech (Cambridge, United Kingdom) silicon probe (64-channels) (n = 15) or 32-channels (n = 3). The recording electrode was painted with the fluorescent dye DiI (Thermo Fisher Scientific, Schwerte, Germany) and then slowly lowered into the craniotomy using micromanipulators (Luigs and Neumann, Ratingen, Germany) at a 25° angle AP (toward the posterior side of the brain). The exposed brain was kept moist using saline solution. A ground wire connected to the amplifier was placed in the saline solution covering the skull to eliminate noise. Brain signals were recorded using a RHD2000 data acquisition system (Intan Technologies, Los Angeles, California) and sampled at 20 kHz. Recording quality was inspected on-line using the open-source RHD2000 Interface Software. A supplementary dose of urethane (0.2 g/kg) was injected right before the start of the recording to standardize anesthesia level across different experiments and avoid the arise of theta activity in the first 30 min of recording. Recordings began after a 10 min waiting period in which clear upstates could consistently be seen at a regular frequency.

In vivo analysis

Request a detailed protocol

We selected the channel to use for upstate detection based on the standard deviation (STD) of the trace during baseline (first 5 min of recording): the channel with the highest STD was selected, as larger voltage deflection increases detection algorithm accuracy. Given the highly synchronous nature of SOs (Figure 3—figure supplement 1) the spatial location of the channel selected was not considered. Upstates were detected comparing threshold crossing points in two signals: the delta-band filtered signal (0.5–4 Hz) and the population spike activity. Candidate upstates were identified in the delta-band filtered signal using two dynamic thresholds ‘a’ and ‘b’:

a=m+σ1.5
b=m+σ0.8

where σ is the standard deviation of the signal during the first five minutes of recording (baseline) and m is the centered moving median calculated using 60 s windows (Matlab function movmedian). The median was used instead of the mean to account for non-stationaries in the data. A candidate upstate was identified at first using the threshold crossings of the signal compared to ‘a’: candidates shorter than 200 ms were deleted and multiple candidates occurring within a window of 300 ms were joined together. Subsequently, the threshold ‘b’ was used to separate upstates occurring in close proximity: if the signal within one candidate crossed the threshold ‘b’ in more than one point then the candidate upstate was split in two at the midpoint between the two threshold crossings. Candidate upstates were finally confirmed if the population spike activity (calculated in 100 ms windows) within the candidate crossed a threshold of 1 σ (calculated during the baseline).

Units detection and classification

Request a detailed protocol

Spike detection was performed offline using the template-based algorithm Kilosort2 https://github.com/MouseLand/Kilosort2; Filippo, 2021a; copy archived at swh:1:rev:a1fccd9abf13ce5dc3340fae8050f9b1d0f8ab7a, with the following parameters:

  • ops.fshigh = 300

  • ops.fsslow = 8000

  • ops.minfr_goodchannels = 0

  • ops.Th = [8 4]

  • ops.Iam = 10

  • ops.AUCsplit = 0.9

  • ops.minFR = 1/1000

  • ops.momentum = [20 400]

  • ops.sigmaMask = 30

  • ops.ThPre = 8

  • ops.spkTh = −6

  • ops.nfilt_factor = 8

  • ops.loc_range = [3 1]

  • ops.criterionNoiseChannel = 0.2

  • ops.whiteningrange = 32

  • ops.ntbuff = 64

Manual curation of the results was performed using Phy https://github.com/cortex-lab/phyFilippo, 2021b; copy archived at swh:1:rev:6ffe05a559bd0302e98ec60d5958ace719544713. Each Isolated unit satisfied the following two criteria: Refractory period (2 ms) violations < 5%, fraction of spikes below detection threshold (as estimated by a gaussian fit to the distribution of the spike amplitudes)<15%. Units with negative maximal waveform amplitude were further classified as putative excitatory if the latency (TP latency) was >0.55 ms or putative inhibitory when TP latency <0.55 ms. The value 0.55 was chosen in accordance with previous studies (Senzai et al., 2019; Antoine et al., 2019). In pharmacological classification, units were classified as ‘activated’ if their firing rate in the 25 min following drug injection was 2 σ (standard deviation) above the baseline rate for at least 5 min. Remaining units were pulled together in the category ‘non-activated’.

Cross-correlogram analysis

Request a detailed protocol

Cross-correlogram based connectivity analysis was performed for every unit to identify inhibitory connections. Units with a spiking rate smaller than 0.3 spikes/s were discarded from the analysis. We used total spiking probability edges (TPSE) algorithm https://github.com/biomemsLAB/TSPEFilippo, 2021c; copy archived at swh:1:rev:b780c753039a2f48201a6bb77dd8f5e65551a845, De Blasi et al., 2019 to identify in a computationally efficient manner putative inhibitory connections between units and all clusters recorded. The parameters used were:

  • d = 0,

  • neg_wins = [2, 3, 4, 5, 6, 7, 8],

  • co_wins = 0,

  • pos_wins = [2, 3, 4, 5, 6],

  • FLAG_NORM = 1.

The connectivity vectors of each unit resulting from TSPE were sorted by inhibition strength. The top 20 connections were further analyzed using custom Matlab (RRID:SCR_001622) code. A connection was classified as inhibitory if the cross correlogram values (x) were smaller than the mean of x by more than one standard deviation (x < mean(x) – std(x)) in at least four consecutive bins (bin size = 1 ms) in a window 4–9 ms after the center of the cross-correlogram.

Slice preparation

Request a detailed protocol

We prepared acute near horizontal slices (∼15° off the horizontal plane) of the mEC from C57Bl/6 mice. Animals were decapitated following isoflurane anesthesia. The brains were quickly removed and placed in ice-cold (∼4° C) ACSF (pH 7.4) containing (in mM) 85 NaCl, 25 NaHCO3, 75 Sucrose, 10 Glucose, 2.5 KCl, 1.25 NaH2PO4, 3.5MgSO4, 0.5 CaCl2, and aerated with 95% O2, 5% CO2. Tissue blocks containing the brain region of interest were mounted on a vibratome (Leica VT 1200, Leica Microsystems), cut at 400 μm thickness, and incubated at 35°C for 30 min. The slices were then transferred to ACSF containing (in mM) 85 NaCl, 25 NaHCO3, 75 Sucrose, 10 Glucose, 2.5 KCl, 1.25 NaH2PO4, 3.5 MgSO4, 0.5 CaCl2. The slices were stored at room temperature in a submerged chamber for 1–5 hr before being transferred to the recording chamber.

In vitro recording

Request a detailed protocol

In order to perform whole-cell recordings slices were transferred to a submersion style recording chamber located on the stage of an upright, fixed-stage microscope (BX51WI, Olympus) equipped with a water immersion objective (×60, Olympus) and a near-infrared charge-coupled device (CCD) camera. The slices were perfused with ACSF (∼35°C bubbled with 95% O2–5% CO2) at 3–5 ml/ min to maintain neuronal health throughout the slice. The ACSF had the same composition as the incubation solution except for the concentrations of calcium and magnesium, which were reduced to 1.2 and 1.0 mM, respectively. Recording electrodes with impedance of 3–5 MΩ were pulled from borosilicate glass capillaries (Harvard Apparatus, Kent, UK; 1.5 mm OD) using a micropipette electrode puller (DMZ Universal Puller). The intracellular solution contained the following (in mM): 135 K-gluconate, 6 KCl, 2 MgCl2, 0.2 EGTA, 5 Na2- phosphocreatine, 2 Na2-ATP, 0.5 Na2-GTP, 10 HEPES buffer, and 0.2% biocytin. The pH was adjusted to 7.2 with KOH. Recordings were performed using Multiclamp 700A/B amplifiers (Molecular Devices, San Jose, California). The seal resistance was >1 GΩ. Capacitance compensation was maximal and bridge balance adjusted. Access resistance was constantly monitored. Signals were filtered at 6 kHz, sampled at 20 kHz, and digitized using the Digidata 1550 and pClamp 10 (Molecular Devices, San Jose, California). Activation light was delivered by a 460 nm laser (DPSS lasers, Santa Clara, California) using a 460–480 nm bandpass excitation filter. Stimulation consisted of 500 ms pulses at 1 Hz.

Stimulation experiments were performed using a bipolar micro-electrode (glass pipette filled with ACSF solution, wrapped by a fine grounding wire) connected to an isolated voltage stimulator (ISO-Flex, A.M.P.I., Israel). A 4x objective (Olympus) was used to visually guide the stimulating electrode into the mEC. Stimulation power was adjusted to achieve consistent upstate generation during baseline (>95%). Each stimulus had a duration of 50 μs and the inter-stimulus interval was 8–10 s.

In vitro analysis

Request a detailed protocol

In vitro upstates were detected in Matlab using an algorithm similar to the one described in the in vivo analysis method section. We used a coincident detection in two signals. In multicellular recordings, we used the membrane potential of two cells, and in single-cell recordings, we used membrane potential and the envelope of the gamma filtered trace (50–250 Hz), as upstates are characterized by an increase in gamma activity (Neske, 2015).

The baseline condition was defined as the last 120 s before drug application, while the post-drug application condition was defined as the 120 s of recording after drug application (Total recording duration: 600 s).

Excitatory (Exc), fast-spiking (FS), and low-threshold spiking (LTS) neurons were classified using Gaussian mixture models (GMM) with a soft clustering approach in Matlab. Input resistance (Rin), Δafter-hyperpolarization (ΔAHP), sag, rheobase, spike width and resting potential (RP) were extracted from each neuron and used in the classification. The first two components of the principal component analysis (PCA) were used to fit the data to a Gaussian mixture model distribution. Initial values were set according to the k-means algorithm with centroid positioned at x and y position: 5, 0; −15,–15; −15, 10. This centroid were placed according to the loadings of the PCA to identify three clusters with the following main features:

  • Cluster 1 (putative Exc): high spike width, low AHP, low rheobase.

  • Cluster 2 (putative FS): low spike width, low SAG, high rheobase, low Rin.

  • Cluster 3 (putative LTS): low spike width, high SAG, high AHP, high Rin.

Covariance matrices were diagonal and not shared. Neurons with a posterior probability of belonging to any of the three clusters of <90% were discarded from further analysis (1/49).

While the majority of Sst-interneurons display LTS features, a minority (~10%) belong to the FS group (Urban-Ciecko et al., 2015). To distinguish FS and LTS interneurons in the Sst-Td Tomato mice, we employed the GMM with posterior probability threshold of 90%. Only LTS Sst neurons were considered for further analysis.

Histological analysis

Request a detailed protocol

For the postmortem electrode track reconstructions of the in vivo recordings, mice were not perfused; rather, brains were extracted from the skull, post-fixed in 4% PFA overnight at 4°C and afterwards cut with a vibratome (Leica Microsystems, Wetzlar Germany) in 100 μM thick sequential sagittal slices. Images were taken using a 1.25x objective and stitched together using the microscope software (BX61, Olympus). Afterwards, we used AllenCCF code https://github.com/cortex-lab/allenCCFFilippo, 2021d; copy archived at swh:1:rev:80fb1326ccaf6ae765944173c7465f650adebabe, to identify electrode shank location (Shamash et al., 2018).

For the anatomical reconstructions of recorded cells in vitro, brain slices were fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (PB) for at least 24 hr at 4°C. After being washed three times in 0.1 M PBS, slices were then incubated in PBS containing 1% Triton X-100% and 5% normal goat serum for 4 hr at room temperature (RT). To visualize biocytin-filled cells we used Streptavidin Alexa 488 conjugate (1:500, Invitrogen Corporation, Carlsbad, CA, RRID:AB_2315383). WFS1 (1:1000, Rabbit, Proteintech, IL, USA, RRID:AB_2880717) was used in a subset of experiments to visualize the L2/L3 border, and Sst (1:1000, Rat, Bachem, Switzerland, RRID:AB_2890072) was used in the 5-HT2AR localization analysis. Slices were incubated with primary antibodies for 48 hr at RT. After rinsing two times in PBS, sections were incubated in the PBS solution containing 0.5% Triton X-100, Alexa fluor 488, Alexa fluor 555 and Alexa fluor 647 (Invitrogen Corporation, Carlsbad, CA) according to the number of antibodies used. Slices were mounted in Fluoroshield (Sigma-Aldrich) under coverslips 2–3 hr after incubation with the secondary antibodies and stored at 4°C.

Labeled cells were visualized using 20x or 40x objectives on a confocal microscope system (SP8, Leica, RRID:SCR_018169). For the 5-HT2AR localization analysis, images of the whole EC were acquired and stitched together using the auto stitching method, without smoothing. Z stacks were acquired every 30 µM. The image stacks obtained were registered and combined in Fiji (RRID:SCR_002285) to form a montage of the sections. Cell counting was executed using the Fiji multi-point tool. X-Y-Z coordinates of each 5-HT2AR-EGFP-positive cell were exported to Matlab and subsequently, using custom written code in Matlab, we semi-automatically inspected each cell for colocalization between EGFP(5-HT2AR) and Sst.

Statistical analysis

Request a detailed protocol

All datasets were tested to determine normality of the distribution either using D’Agostino-Pearson omnibus normality test or Shapiro-Wilk normality test. Student’s t-test and one-way ANOVA were used for testing mean differences in normally distributed data. Wilcoxon matched-pairs signed rank test and Kruskal-Wallis were used for non-normally distributed datasets. Dunn-Sidak multiple comparison test was used to compare datasets with three or more groups. Kolmogorov-Smirnov test was used to compare cumulative distributions. Statistical analysis was performed using Prism (6.01, RRID:SCR_002798) and Matlab. All data are expressed as mean ± SEM. Asterisks in figures represent p-values smaller than 0.05, unless stated otherwise in the legend.

Data availability

A database for the in vivo data is available at http://dx.doi.org/10.6084/m9.figshare.14229545. Custom code used for the analysis is available at https://github.com/Schmitz-lab/De-Filippo-et-al-2021 (copy archived at https://archive.softwareheritage.org/swh:1:rev:eda027f20c4f267762bc46c47f976fff49a0aa84/). Code used for spike sorting is available at https://github.com/MouseLand/Kilosort2 (copy archived at https://archive.softwareheritage.org/swh:1:rev:a1fccd9abf13ce5dc3340fae8050f9b1d0f8ab7a/). Code used for manual curation of sorted spikes is available at https://github.com/cortex-lab/phy (copy archived at https://archive.softwareheritage.org/swh:1:rev:6ffe05a559bd0302e98ec60d5958ace719544713/). Code used to estimate neuronal connectivity is available at https://github.com/biomemsLAB/TSPE (copy archived at https://archive.softwareheritage.org/swh:1:rev:b780c753039a2f48201a6bb77dd8f5e65551a845/).

The following data sets were generated
    1. de Filippo R
    (2021) figshare
    Somatostatin interneurons activated by 5-HT2A receptor suppress slow oscillations in medial entorhinal cortex.
    https://doi.org/10.6084/m9.figshare.14229545

References

    1. Ashby CR
    2. Jiang LH
    3. Kasser RJ
    4. Wang RY
    (1990)
    Electrophysiological characterization of 5-hydroxytryptamine2 receptors in the rat medial prefrontal cortex
    The Journal of Pharmacology and Experimental Therapeutics 252:171–178.
  1. Book
    1. Descarries L
    2. Riad M
    3. Parent M
    (2010) CHAPTER 1.4 - Ultrastructure of the Serotonin Innervation in the Mammalian Central Nervous System
    In: Müller CP, Jacobs BL, editors. Handbook of Behavioral Neuroscience. Amsterdam, Netherlands: Elsevier. pp. 65–101.
    https://doi.org/10.1016/S1569-7339(10)70072-2

Decision letter

  1. Laura L Colgin
    Senior Editor; University of Texas at Austin, United States
  2. Martin Vinck
    Reviewing Editor; Ernst Strüngmann Institute (ESI) for Neuroscience in Cooperation with Max Planck Society, Germany
  3. Antonio Fernandez-Ruiz
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This paper identifies a mechanism through which serotonin, one of the brain's major neuromodulators, affects network activity, through 5HT2a-mediated activation of a major subclass of GABAergic interneurons – Somatostatin-positive interneurons. These findings advance our understanding of how brain state affects cortical circuits through the action of specific cell classes.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Serotonin suppresses slow oscillations by activating somatostatin interneurons via the 5-HT2A receptor" for consideration by eLife. Your article has been reviewed by four peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Antonio Fernandez-Ruiz (Reviewer #3).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife at present. However given the interest of the reviewers in the topic, we welcome a resubmission to eLife if it is possible to run additional experiments requested by the reviewers; this submission would be treated as a new submission but would be send to the same reviewers. Although the reviewers made several positive comments, they identified several major issues, in particular: 1) The interpretation and framing of the paper in terms of slow oscillations, which was questioned by several reviewers. 2) The selectivity of pharmacological manipulation (see Reviewer 2), and 3) Data analysis. The reviewers agreed that while recordings without anesthesia would enhance the manuscript, these experiments are beyond the scope of the paper, although this limitation needs to be discussed more clearly.

Reviewer #1:

This paper shows that slow oscillations in entorhinal cortex are affected by serotonin, in particular acting through 5HT2AR receptor, and that this effect is mediated by Somatostatin interneurons. Although the generality of this finding for other cortical areas and natural (i.e. not anesthesized) conditions remains to be established, I find this is a set of novel findings that could be important to understand control of brain state and also clinical applications. Critique:

1) A clear short-coming of the paper is the lack of recordings during naturally occurring slow waves. If it is reasonable to conduct these experiments, this would greatly enhance the paper. In particular potential effects of anesthesia on SSt interneurons need to be taken into account (see Adesnik and Scanziani, 2012).

2) A second critical question is whether there is at all an effect on slow oscillations. It appears to me that there is a massive suppression of spiking activity such that the circuit is essentially in a sustained downstate that is not at all representative of the desynchronized state as one normally finds it, and what would have happened with stimulating the Raphe Nucleus. So the conclusion that there's an effect on slow oscillations is not entirely clear to me. This pertains to Figure 1 and 4, for instance.

Would blocking 5HT2aRs during normal wake-to-sleep or arousal-to-SWS transitions abolish slow oscillations? If one would stimulate the Raphe and blocks 5HT2aR what would one predict? I understand these are difficult experiments to do, but does the paper bear relevance for these more naturalistic scenarios?

3) For completeness it would be useful to comment on the other GABAergic cells that have 5HT3Rs and explain why their contribution is not relevant for different pharmacological agents or more in general.

Reviewer #2:

The manuscript by De Filippo et al. presents a combined in vivo / in vitro study of the cellular mechanisms of network dynamics in the entorhinal cortex (EC). The main findings support 5-HT suppression of slow oscillatory activity, mediated by 5-HT2AR activation of SOM interneurons. This is a novel finding, potentially of general interest. My major concerns relate to the strength of the in vivo evidence and the framing of the study.

Concerns related to in vivo data

1) The authors present clear demonstration that the pharmacological agents used suppress the slow oscillation (SO) in EC. Given that none of these drugs are truly selective and that they are presented i.p., the authors provide no direct evidence that these effects are 1) mediated by 5-HT release and 2) mediated by 5-HT release in the entorhinal cortex. This can be addressed by preventing drug-mediated SO reduction with local application of 5-HT2AR antagonists directly in the EC.

2) The in vivo single unit data does not support the activation of SOM interneurons.

a) The activated neurons appear to include both putative inhibitory (fast spiking) and putative excitatory (non-fast spiking) neurons (Figure 2C), rather than an intermediate, potential SOM population. If the activated neurons include both FS inhibitory and non-FS excitatory neurons, one would expect TP latency of the population to be intermediate as shown in Figure 2B,C.

b) Given the inability to identify SOM interneurons based on waveform alone, additional characterization of the activated units is necessary. First, do the CCGs of the activated units indicate inhibitory connections? These data are notably missing (or not easily extracted from Figure 2—figure supplement 1).

c) Second, definitely identifying SOM interneurons in vivo would require some type of tagging.

Concerns related to the Introduction

3) Motivating this study by discussing psychological processes, psychiatric disorders, sleep/wake cycles or psychoactive drugs is highly tenuous and misleading. This study does not address any of these processes. Alternatively, the Introduction lacks even a superficial description of topics that are highly relevant to this study, including neuroanatomy and neural circuitry of the EC, physiological properties of GABAergic interneurons in the EC, previous studies of the slow oscillation and involvement of different neuronal cell types in the EC, and neuromodulation of different interneuron subtypes, especially known patterns of neuromodulatory receptor expression on SOM interneurons.

Reviewer #3:

The manuscript by De Filipo et al. presents a well-executed and novel study on the role of somatostatin positive (Som) interneurons in the generation of cortical slow oscillations (SO) in the entorhinal area. The authors employ a combination of mouse in vivo recordings under anesthesia, slice physiology, together with pharmacological and chemo/optogenetic manipulations to shown that entorhinal SO are suppressed by serotonin activation of Som interneurons. The present work provides new data on an interesting topic that can have broader implications to understand the relationship between neuromodulators, state-dependent brain oscillations and different interneuron subtypes. The experimental approaches are rigorous and adequately support the authors claims. However, the study has some limitations that cast doubts on the specificity of the effects reported. Additional analysis may be sufficient to clarify these doubts.

1) The authors need to show more convincing evidence that the effects 5-HT and the manipulations reported in Figure 1 are specific of the SO and not a general dampened of entorhinal activity. More extensive analysis of the current data will suffice for that purpose. Figure 1H suggests that there may be a reduction in power in other frequencies than δ, but is not clearly appreciated (using a log frequency axis will facilitate it). A power spectrum comparison before, during and after drug application is necessary to evaluate the changes in all frequencies. Entorhinal cortex displays prominent γ oscillations, even during sleep and urethane anesthesia, are those also affected? This should be specifically quantified. Under urethane there is an alternation between a SO and a slow theta (~3-4Hz) state. Are theta oscillations also affected by the pharmacological manipulations?

2) More comprehensive analysis of the effect of the different manipulations on SO reported through the manuscript should be performed, not only the rate of up states. For example, distribution of durations, inter-event intervals, amplitude of the remaining up states, etc. These will help to better understand the nature of the effects of serotonin release on SO.

3) Expanding the analysis and discussion of the contribution of PV interneurons will help establish the specificity of the role of Som interneurons. The results of the experiment with hM4Di-PV are encouraging in this regard. The authors do have fast-spiking, narrow-waveform cells (thus putative PV+ interneurons) in their dataset; however, there is very little discussion about this through the text. The results in Figure 4 provide evidence of the ability of Som interneurons to suppress SO. However, I would expect a very similar effect from PV activation. Unless the authors have evidence of the opposite, they should acknowledge this.

Reviewer #4:

In this manuscript, De Filippo et al. identify somatostatin-containing interneurons (SOM) of the entorhinal cortex (EC) as targets of serotonergic neurons in the dorsal raphe and describe their involvement in the mediation of slow oscillations in the EC, using an array of techniques such as pharmacology, optogenetics and DREADD.

The manuscript is well and the analysis is solid. While there is no doubt this study demonstrates convincingly the involvement of SOM neurons in the regulation of SO, it fails to convincingly rule out other cell types which could equally contribute to the cortical dynamics under investigation. In that respect, I am left with some lingering doubts as to whether or not something was missed in the study and would therefore appreciate some clarifications on the part of the authors (see major point below).

1) Some previous studies (for example, Ferezou et al., 2002) have shown that VIP/CCK GABAergic cells express 5-HT3 receptors, which are ionotropic receptors and can therefore respond very fast (with a depolarization, therefore an activation) to serotonergic neuromodulation. How come the authors do not mention this study (and others) and could they give an explanation as to why none of these neocortical cells (presumably also present in the EC?) were identified in their study as activated? Alternatively, using mCPBG in a similar fashion to how α-methyl5-HT was used in Figure 3 to rule out involvement of 5-HT3R would be valuable. The authors also mention the fact that only 11% of the cells expressing 5-HT2AR promoter were SOM. Could they please elaborate on the remaining 89% of cells?

2) Studies in the somatosensory cortex of awake mice (presumably therefore at a time when 5-HT release is highest) showed that L2/3 SOM neurons were not correlated with other neuronal cell types at the level of their membrane potentials and that they hyperpolarized and therefore reduce their spiking during active wake (when the mouse movedits whiskers (Gentet et al., 2012, Muñoz et al., 2017)). The authors should discuss their findings in the light of these papers considering that they argue that serotonin activates SOM neurons in the EC and 5-HT release is highest during wake, and therefore presumably, even more so during active wake.

https://doi.org/10.7554/eLife.66960.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Reviewer #1:

This paper shows that slow oscillations in entorhinal cortex are affected by serotonin, in particular acting through 5HT2AR receptor, and that this effect is mediated by Somatostatin interneurons. Although the generality of this finding for other cortical areas and natural (i.e. not anesthesized) conditions remains to be established, I find this is a set of novel findings that could be important to understand control of brain state and also clinical applications. Critique:

1) A clear short-coming of the paper is the lack of recordings during naturally occurring slow waves. If it is reasonable to conduct these experiments, this would greatly enhance the paper. In particular potential effects of anesthesia on SSt interneurons need to be taken into account (see Adesnik and Scanziani, 2012).

Recording during natural sleep together with application of either MDMA or fenfluramine (Fen) would be problematic as both these compounds act as sleep suppressant in both rodents and humans (Myers et al., 1993, Fornal and Radulovacki, 1983, Balogh et al., 2004, Randall et al., 2009). We agree that more effort should be put in studying the effect of 5-HT on Som interneurons in vivo and we plan on doing so in a following work in awake head-fixed condition. In accordance with the reviewing and senior editors’ opinions, we believe that these experiments would go beyond the aims of our current study. We modified the text to explicitly state that the absence of data in naturalistic conditions represent a limitation of our work.

2) A second critical question is whether there is at all an effect on slow oscillations. It appears to me that there is a massive suppression of spiking activity such that the circuit is essentially in a sustained downstate that is not at all representative of the desynchronized state as one normally finds it, and what would have happened with stimulating the Raphe Nucleus. So the conclusion that there's an effect on slow oscillations is not entirely clear to me. This pertains to Figure 1 and 4, for instance.

Would blocking 5HT2aRs during normal wake-to-sleep or arousal-to-SWS transitions abolish slow oscillations? If one would stimulate the Raphe and blocks 5HT2aR what would one predict? I understand these are difficult experiments to do, but does the paper bear relevance for these more naturalistic scenarios?

The choice of the word “desynchronization” was unfortunate as what we observe in our data does not reflect in vivo awake classic “desynchronization”(Chen et al., 2015), we revised the Discussion to claim instead that 5-HT suppresses synchronized activity. We agree that the effect we see, both in vivo and in vitro, can be described as a general dampening of spiking activity. We consider suppression of upstates to be equal to suppression of slow oscillations (SO), as we do not believe that a constant downstate can be categorized as slow oscillation. Our usage of the terms upstate and downstate follow what, in our view, is the field convention (Harris and Thiele, 2011), some works use a different definition (see (Niethard et al., 2018)).

We use SO as a proxy to understand the effect of 5-HT on networks of cortical neurons but we do not aim to make conclusive statements about the relevance of our findings in natural sleep. For example, during natural sleep (both slow-wave sleep and REM) the level of 5-HT in the brain are at the lowest (Oikonomou et al., 2019, Unger et al., 2020, McGinty and Harper, 1976), thus, the increase of 5-HT during sleep would be of itself an artificial condition. We investigated in a new set of experiments (Author response image 1 and Figure 3) the effect of Raphe 5-HT neurons optogenetic activation during anesthesia in the same cortical region (medial entorhinal cortex). We did not see suppression of SO but only a decrease in spiking rate during upstates. Decrease in spiking following the same manipulation has been reported in a previous study in somatosensory and motor areas (Grandjean et al., 2019). Our explanation for this difference is that the amount of 5-HT released using pharmacological or optogenetic release is probably considerably different. As we explain in the revised text, MDMA and Fen (5 mg/kg) have been reported to induce a ~20-fold increase in peak 5-HT concentration both in monkey and rodent brain (Udo de Haes et al., 2006, Gołembiowska et al., 2016). On the other hand, light activated 5-HT neurons in ePet mice seems to fire at a physiological rate (≈ 3 Hz) (Ranade and Mainen, 2009, Sakai and Crochet, 2001) even when stimulated at higher frequencies. (Grandjean et al., 2019). In consequence it is likely that optogenetic activation is not eliciting the release of enough 5-HT to cause SO suppression and Som interneurons activation during anesthesia. We decided to include these data as we think that the different effect of pharmacological and optogenetic release of 5-HT might be of interest to the community especially considering the potential usage of MDMA as medication for post-traumatic stress disorder (Mithoefer et al., 2019, Inserra et al., 2021).

We do not believe that 5-HT is solely responsible for SO suppression between sleep to wake transition, it is known that acetylcholine (Chen et al., 2015) and noradrenaline (Steriade et al., 1993) can both suppress SO, and both of them are also synced with the wake cycle (higher levels during wake compared to slow wave sleep (Watson et al., 2012, Mitchell and Weinshenker, 2010). This makes pharmacology experiments in sleeping condition difficult to interpret in our opinion. In the future we will focus on 5-HT2A pharmacological manipulations in vivo awake conditions, we believe that these experiments go beyond the aims of the current study.

Author response image 1
SO are not suppressed by optogenetic activation of Raphe 5-HT neurons during anesthesia.

(A) DRN immunohistochemistry image showing ChR2-YFP infection in DRN serotonergic neurons. 5-HT in red, YFP in green. Scale bars: 50 µm. (B) 3D visualization of microelectrodes location for all experiments and optic fiber (black) location. EC represented in grey, dorsal Raphe nuclei (DRN) in pink. (C) Top: upstates raster plot. Bottom: upstates density using Gaussian kernel density estimation. (D) Histogram of upstates incidence during baseline and laser stimulation (n animals = 4onedr, baseline: 2.25 ± 0.13 upstates/10s, laser: 2.11 ± 0.14 upstates/10s). (E) Left: normalized spike population activity during baseline (black) and laser stimulation (blue) in mEC L3/5 (n = 108, p < 0.05, unpaired t-test with Holm-Šidák correction). Right: heatmap showing average spike rate difference per cluster between stimulation and baseline during upstates, some units show reduced spiking during stimulation. (F) Left: normalized spike population activity during baseline (black) and laser stimulation (blue) in CA1 (n = 140, p > 0.05, unpaired t-test with Holm-Šidák correction). Right: heatmap showing average spike rate difference per cluster between stimulation and baseline during upstates, no clear difference in spiking activity.

3) For completeness it would be useful to comment on the other GABAergic cells that have 5HT3Rs and explain why their contribution is not relevant for different pharmacological agents or more in general.

We performed a new set of in vitro experiment to complement some already available preliminary data on 5-HT3R and we can confirm that in vitro application of either 5-HT3R agonist and antagonist do not have any effect on SO (Figure 3—figure supplement 4). We integrated these results in the main text.

Reviewer #2:

The manuscript by De Filippo et al. presents a combined in vivo / in vitro study of the cellular mechanisms of network dynamics in the entorhinal cortex (EC). The main findings support 5-HT suppression of slow oscillatory activity, mediated by 5-HT2AR activation of SOM interneurons. This is a novel finding, potentially of general interest. My major concerns relate to the strength of the in vivo evidence and the framing of the study.

Concerns related to in vivo data

1) The authors present clear demonstration that the pharmacological agents used suppress the slow oscillation (SO) in EC. Given that none of these drugs are truly selective and that they are presented i.p., the authors provide no direct evidence that these effects are 1) mediated by 5-HT release and 2) mediated by 5-HT release in the entorhinal cortex. This can be addressed by preventing drug-mediated SO reduction with local application of 5-HT2AR antagonists directly in the EC.

To address whether the suppression we see in vivo is selectively mediated by 5-HT we performed new in vivo anesthetized recordings in ePet-cre mice where we activated, using an optogenetic approach, 5-HT neurons in the dorsal raphe nucleus (DRN) (Author response image 1/Figure 3), this experimental paradigm enabled us to have better temporal control on 5-HT release.

Our in vitro data show that suppression of SO is not mediated exclusively by 5-HT2AR and this is reflected by the fact that even in presence of ketanserin (5-HT2AR antagonist) 5-HT is still able to significantly suppress SO (Figure 4—figure supplement 3 B-E, ≈ 60% incidence reduction). This, however, is not surprising because in mEC pyramidal cells are known to express 5-HT1AR, a receptor that causes hyperpolarization via activation of G protein-coupled inwardly rectifying potassium (GIRK) channels (Schmitz et al., 1998, Chalmers and Watson, 1991). We decided not to inject ketanserin together with MDMA/Fen in vivo because the potential action of 5-HT1AR would make the result difficult to interpret. For example, the amount of 5-HT released by these compounds (MDMA/Fen), vastly greater than physiological, could activate 5-HT1AR to an extent that would suppress SO without the need of 5-HT2AR.

In contrast to the application of MDMA/Fen, after optogenetic 5-HT release, we did not observe suppression of upstate incidence but only a reduction in spiking rate during upstates. A possible explanation might be that the amount of 5-HT released via optogenetic activation is considerably less compared to the pharmacological release. These results prevent us to claim that the effect we see in vivo is due solely to 5-HT release, nonetheless, we like to underline that Fen has been used before as a 5-HT selective releaser with the specific purpose of disentangle the effect of MDMA (Heifets et al., 2019). To strengthen the link between suppression and 5-HT2AR we performed a new set of in vitro experiments showing that ketanserin (5-HT2AR antagonist) dampens the suppressive power of fenfluramine (Figure 3—figure supplement 1). In this revised version we are careful to avoid strong statement such as “5-HT is sufficient to suppress slow oscillations in vivo”, nonetheless, considering our in vitro data, we believe it is legitimate to state that 5-HT plays a role in the suppression. We discuss this in the penultimate paragraph of the Discussion. We agree that more effort should be put in studying the effect of 5-HT on Som interneurons in vivo and we plan on doing this in a following work in awake head-fixed condition. We would prefer to focus on more naturalistic conditions to disentangle the relationship between Som interneurons and 5-HT especially because anesthesia is known to be a confounding factory particularly in the case of Som interneurons (Adesnik et al., 2012).

We believe, however, that these experiments would go beyond the aims of our current study as they would also require a new animal license.

2) The in vivo single unit data does not support the activation of SOM interneurons.

a) The activated neurons appear to include both putative inhibitory (fast spiking) and putative excitatory (non-fast spiking) neurons (Figure 2C), rather than an intermediate, potential SOM population. If the activated neurons include both FS inhibitory and non-FS excitatory neurons, one would expect TP latency of the population to be intermediate as shown in Figure 2B,C.

b) Given the inability to identify SOM interneurons based on waveform alone, additional characterization of the activated units is necessary. First, do the CCGs of the activated units indicate inhibitory connections? These data are notably missing (or not easily extracted from Figure 2—figure supplement 1).

c) Second, definitely identifying SOM interneurons in vivo would require some type of tagging.

a) We agree that our data do not prove conclusively the involvement of Som Interneurons in SO suppression in vivo. We also agree that if the activated neurons were to be equally putative FS and putative excitatory the final TP latency average would be in between these two groups. To understand whether this is the case we looked at the cumulative probability distribution (Figure 2B, middle panel) where we can see a clear difference between the three groups. This is reflected by the fact that units with TP latency between 0.4 and 0.7 ms have much higher likelihood (from ~40% to ~80%, red line, right y-axis) of being part of the “activated” group compared to units with any other TP latency (Figure 3B bottom). If the “activated” units were to be equally distributed across TP latencies we should not have such peak in our opinion. We added a clarification in the text. In conclusion, we believe that these data support our in vitro data regarding Som interneurons.

b) CCG based connectivity analysis requires a minimum baseline firing rate to be effective, now we explain this detail in the legend and not only in the Materials and methods section. The threshold we used is 0.3 spikes/s, units missing from the analysis have a lower firing rate. We acknowledge that this analysis does not help in characterizing the intermediate population as inhibitory therefore we are open to remove the image as superfluous.

c) We agree and we revised the text to clarify that our data do not demonstrate conclusively the involvement of Som interneurons in vivo. We consider our in vitro experiment to show conclusively the involvement of Som interneurons activated by 5-HT2A in SO suppression. Our in vivo data merely support this notion as Som interneurons in vivo might have a TP latency in between FS and excitatory neurons (Trainito et al., 2019). Optotagging in vivo would answer the question conclusively however the location of the entorhinal cortex (EC) makes this experiment particularly challenging, to have a decent chance of finding responsive units we should use Neuropixels probes together with an external optic fiber. The number of responsive units In Sst-cre mice found by the Allen institute using Neuropixels has also discouraged us from performing optotagging (8/350 units, example found at https://allensdk.readthedocs.io/en/latest/_static/examples/nb/ecephys_optotagging.html). Moreover anesthesia is a known confounding factor especially in the case of Som interneurons (Adesnik et al., 2012), for this reason we would prefer to focus on non-anesthetized conditions, something we plan to do in our next work.

Concerns related to the Introduction

3) Motivating this study by discussing psychological processes, psychiatric disorders, sleep/wake cycles or psychoactive drugs is highly tenuous and misleading. This study does not address any of these processes. Alternatively, the Introduction lacks even a superficial description of topics that are highly relevant to this study, including neuroanatomy and neural circuitry of the EC, physiological properties of GABAergic interneurons in the EC, previous studies of the slow oscillation and involvement of different neuronal cell types in the EC, and neuromodulation of different interneuron subtypes, especially known patterns of neuromodulatory receptor expression on SOM interneurons.

We agree and changed the Introduction adding a description of EC neuroanatomy and role in slow oscillations. We also removed from the Introduction references to psychological processes, psychiatric disorders. We kept the description of the relationship between sleep/wake cycles and 5-HT as this provides a frame of reference to the relationship between 5-HT and slow oscillations. We address neuromodulation in the Discussion.

“Som interneurons are specifically known to be sufficient to cause desynchronization in V1 (Chen et al., 2015), while it is well known that they are potently excited by ACh in various cortical areas (Chen et al., 2015, Obermayer et al., 2018, Fanselow et al., 2008) including mEC (Desikan et al., 2018) and NE in frontal cortex (Kawaguchi and Kubota, 1998), our work identifies them as a novel target of 5-HT regulation via 5-HT2AR. Recently it has been proposed that inhibitory interneurons play a key role in mediating the effect of ACh and NE on cortical state transitions (Cardin, 2019), our results add a new level of complexity to this picture.”

Reviewer #3:

The manuscript by De Filipo et al. presents a well-executed and novel study on the role of somatostatin positive (Som) interneurons in the generation of cortical slow oscillations (SO) in the entorhinal area. The authors employ a combination of mouse in vivo recordings under anesthesia, slice physiology, together with pharmacological and chemo/optogenetic manipulations to shown that entorhinal SO are suppressed by serotonin activation of Som interneurons. The present work provides new data on an interesting topic that can have broader implications to understand the relationship between neuromodulators, state-dependent brain oscillations and different interneuron subtypes. The experimental approaches are rigorous and adequately support the authors claims. However, the study has some limitations that cast doubts on the specificity of the effects reported. Additional analysis may be sufficient to clarify these doubts.

1) The authors need to show more convincing evidence that the effects 5-HT and the manipulations reported in Figure 1 are specific of the SO and not a general dampened of entorhinal activity. More extensive analysis of the current data will suffice for that purpose. Figure 1H suggests that there may be a reduction in power in other frequencies than δ, but is not clearly appreciated (using a log frequency axis will facilitate it). A power spectrum comparison before, during and after drug application is necessary to evaluate the changes in all frequencies. Entorhinal cortex displays prominent γ oscillations, even during sleep and urethane anesthesia, are those also affected? This should be specifically quantified. Under urethane there is an alternation between a SO and a slow theta (~3-4Hz) state. Are theta oscillations also affected by the pharmacological manipulations?

A comprehensive analysis of the effect of MDMA/Fenf to different frequency bands has been added in Figure 1—figure supplement 2. Prominent theta oscillations were not observed in our experiments (Figure 1-2), at the start of every recording session an additional dose of urethane (0.2 g/kg) was injected to stabilize upstates frequency, in the following 30 min we never observed strong theta oscillations.

2) More comprehensive analysis of the effect of the different manipulations on SO reported through the manuscript should be performed, not only the rate of up states. For example, distribution of durations, inter-event intervals, amplitude of the remaining up states, etc. These will help to better understand the nature of the effects of serotonin release on SO.

Additional analysis has been performed in: Figure 1—figure supplement 3, Figure 3—figure supplement 1 and Figure 3—figure supplement 5.

3) Expanding the analysis and discussion of the contribution of PV interneurons will help establish the specificity of the role of Som interneurons. The results of the experiment with hM4Di-PV are encouraging in this regard. The authors do have fast-spiking, narrow-waveform cells (thus putative PV+ interneurons) in their dataset; however, there is very little discussion about this through the text. The results in Figure 4 provide evidence of the ability of Som interneurons to suppress SO. However, I would expect a very similar effect from PV activation. Unless the authors have evidence of the opposite, they should acknowledge this.

We agree with the reviewer that PV neurons activation have an effect on upstates that is comparable to Som neurons, we explicitly state this in the revised text citing a recent work that investigated effect of Som and PV neurons on upstates (Zucca et al., 2017). Our in vivo data (Figure 2) show a small number of FS units (TP latency < 0.4 ms) activated by 5-HT. However, in vitro, we did not see any activation from FS neurons (Figure 4—figure supplement 6). We acknowledge that the sample size of FS neurons is too small to conclusively talk about the role of PV neurons, moreover, PV neurons have been shown to be activated by 5-HT in prefrontal cortex (Athilingam et al., 2017, Puig et al., 2010). In conclusion, we do not exclude that a minority of FS cells are activated by 5-HT, even in entorhinal cortex, but in our opinion, this does not detract from the finding that Som neurons are activated by 5-HT, a discovery not previously reported in any brain area. Following the reviewer suggestion, we are open to perform additional analysis on PV neurons of our in vivo dataset.

Reviewer #4:

In this manuscript, De Filippo et al. identify somatostatin-containing interneurons (SOM) of the entorhinal cortex (EC) as targets of serotonergic neurons in the dorsal raphe and describe their involvement in the mediation of slow oscillations in the EC, using an array of techniques such as pharmacology, optogenetics and DREADD.

The manuscript is well and the analysis is solid. While there is no doubt this study demonstrates convincingly the involvement of SOM neurons in the regulation of SO, it fails to convincingly rule out other cell types which could equally contribute to the cortical dynamics under investigation. In that respect, I am left with some lingering doubts as to whether or not something was missed in the study and would therefore appreciate some clarifications on the part of the authors (see major point below).

1) Some previous studies (for example, Ferezou et al., 2002) have shown that VIP/CCK GABAergic cells express 5-HT3 receptors, which are ionotropic receptors and can therefore respond very fast (with a depolarization, therefore an activation) to serotonergic neuromodulation. How come the authors do not mention this study (and others) and could they give an explanation as to why none of these neocortical cells (presumably also present in the EC?) were identified in their study as activated? Alternatively, using mCPBG in a similar fashion to how α-methyl5-HT was used in Figure 3 to rule out involvement of 5-HT3R would be valuable. The authors also mention the fact that only 11% of the cells expressing 5-HT2AR promoter were SOM. Could they please elaborate on the remaining 89% of cells?

We thank the reviewer for this concern. We have performed additional experiments to address the role of the 5HT3 expressing interneurons. Our results show that their involvement in the modulation of upstates in the medial entorhinal cortex is negligible (Figure 3—figure supplement 4). The 5HT3agonist, m-CPBG caused no reduction in the upstate frequency and the 5HT3 antagonist, Tropisetron failed to block the suppression of upstates by serotonin.

The majority of 5HT2A cells were found in L6 as also reported by Weber and Andrade (2010). This is reflected by the peak density at ~800 µm in Figure 4—figure supplement 9 E (left). We did not analyze further the dataset using the 5-HT2AR-EGFP mouse (Weber and Andrade, 2010) as we discovered from personal communication with Dr. Andrade that in this mouse not all cells expressing 5-HT2AR are expressing the GFP marker. Future studies will be conducted using 5-HT2AR-cre mice.

2) Studies in the somatosensory cortex of awake mice (presumably therefore at a time when 5-HT release is highest) showed that L2/3 SOM neurons were not correlated with other neuronal cell types at the level of their membrane potentials and that they hyperpolarized and therefore reduce their spiking during active wake (when the mouse movedits whiskers (Gentet et al., 2012, Muñoz et al., 2017)). The authors should discuss their findings in the light of these papers considering that they argue that serotonin activates SOM neurons in the EC and 5-HT release is highest during wake, and therefore presumably, even more so during active wake.

We thank the reviewer for this question. We now discuss the relationship between sleep-wake and Som interneurons in the text:

“5-HT levels in the brain, similarly to other neuromodulators, are synchronized to the sleep wake cycle, with higher levels present during wake state (Oikonomou et al., 2019). How does this notion relate to activation of Som interneurons by 5-HT2AR? A limitation of our study is the absence of data in naturalistic conditions, with anesthesia potentially being a confounding factor (Adesnik et al., 2012). We can nonetheless speculate that Som interneurons should be more active during states with higher 5-HT levels (wake>SWS>REM). A previous study directly measuring the activity of various neurons classes across different states seem to support this idea. Som interneurons in the dorsal cortical surface display highest activity during wake, lower during SWS and lowest during REM (Niethard et al., 2016)… “

In a recent review from our lab (Tukker et al., 2020) we have discussed important differences between the somatosensory and entorhinal cortices in the initiation, propagation and regulation of up down states. Although we have not investigated Som neurons in the somatosensory cortex and their modulation by serotonin in this paper regarding up-down states, we predict that there might be differences compared to entorhinal cortex. Further in the paper from Muñoz et al., 2017 it is evident that Som has layer specific modulation / participation during whisker stimulation. Further studies will hopefully investigate the modulation of Som interneurons in a layer specific manner during brain states in different cortices.

https://doi.org/10.7554/eLife.66960.sa2

Article and author information

Author details

  1. Roberto de Filippo

    1. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Neuroscience Research Center, Berlin, Germany
    2. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Cluster of Excellence NeuroCure, Berlin, Germany
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Investigation, Visualization, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    roberto.de-filippo@charite.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4085-9114
  2. Benjamin R Rost

    German Centre for Neurodegenerative Diseases (DZNE), Berlin, Germany
    Contribution
    Conceptualization, Resources, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Alexander Stumpf

    Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Neuroscience Research Center, Berlin, Germany
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Claire Cooper

    Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Neuroscience Research Center, Berlin, Germany
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  5. John J Tukker

    1. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Neuroscience Research Center, Berlin, Germany
    2. German Centre for Neurodegenerative Diseases (DZNE), Berlin, Germany
    Contribution
    Resources, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4394-199X
  6. Christoph Harms

    1. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Department of Experimental Neurology, Berlin, Germany
    2. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Center for Stroke Research Berlin, Berlin, Germany
    3. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Einstein Center for Neurosciences Berlin, Berlin, Germany
    Contribution
    Resources, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2063-2860
  7. Prateep Beed

    Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Neuroscience Research Center, Berlin, Germany
    Contribution
    Conceptualization, Supervision, Investigation, Methodology, Writing - review and editing
    Contributed equally with
    Dietmar Schmitz
    Competing interests
    No competing interests declared
  8. Dietmar Schmitz

    1. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Neuroscience Research Center, Berlin, Germany
    2. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Cluster of Excellence NeuroCure, Berlin, Germany
    3. German Centre for Neurodegenerative Diseases (DZNE), Berlin, Germany
    4. Charité–Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health; Einstein Center for Neurosciences Berlin, Berlin, Germany
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Writing - original draft, Writing - review and editing
    Contributed equally with
    Prateep Beed
    For correspondence
    dschmitz-office@charite.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2741-5241

Funding

Bundesministerium für Bildung und Forschung (SFB1315-327654276)

  • Dietmar Schmitz

Deutsche Forschungsgemeinschaft (SPP1926)

  • Benjamin R Rost

Deutsche Forschungsgemeinschaft (SPP1665)

  • Dietmar Schmitz

Deutsche Forschungsgemeinschaft (01GQ1420B)

  • Dietmar Schmitz

Deutsche Forschungsgemeinschaft (HA5741/5-1)

  • Christoph Harms

Deutsche Forschungsgemeinschaft (TRR295)

  • Christoph Harms

Bundesministerium für Bildung und Forschung (01EO0801)

  • Christoph Harms

NeuroCure Exzellenzcluster (EXC-2049 - 390688087)

  • Dietmar Schmitz

Charité – Universitätsmedizin Berlin (BIH Fellowship)

  • Prateep Beed

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Senior Editor

  1. Laura L Colgin, University of Texas at Austin, United States

Reviewing Editor

  1. Martin Vinck, Ernst Strüngmann Institute (ESI) for Neuroscience in Cooperation with Max Planck Society, Germany

Reviewer

  1. Antonio Fernandez-Ruiz

Publication history

  1. Received: January 27, 2021
  2. Accepted: March 10, 2021
  3. Version of Record published: March 31, 2021 (version 1)
  4. Version of Record updated: April 9, 2021 (version 2)

Copyright

© 2021, de Filippo et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,080
    Page views
  • 116
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Neuroscience
    Stanley Heinze
    Insight

    Studying neurons and their connections in the central complex of the fruit fly reveals new insights into how their structure and function shape perception and behavior.

    1. Neuroscience
    Amit Kumar et al.
    Research Article

    The piriform cortex (PCx) is essential for learning of odor information. The current view postulates odor learning in the PCx is mainly due to plasticity in intracortical (IC) synapses, while odor information from the olfactory bulb carried via the lateral olfactory tract (LOT) is 'hardwired'. Here we revisit this notion by studying location and pathway dependent plasticity rules. We find that in contrast to the prevailing view, synaptic and optogenetically activated LOT synapses undergo strong and robust long-term potentiation (LTP) mediated by only few local NMDA-spikes delivered at theta frequency, while global spike timing dependent plasticity protocols (STDP) failed to induce LTP in these distal synapses. In contrast, IC synapses in apical and basal dendrites undergo plasticity with both NMDA-spikes and STDP protocols but to a smaller extent compared with LOT synapses. These results are consistent with a self-potentiating mechanism of odor information via NMDA-spikes which can form branch-specific memory traces of odors that can further associate with contextual IC information via STDP mechanisms to provide cognitive and emotional value to odors.