Gamma-frequency oscillatory activity plays an important role in information integration across brain areas. Disruption in gamma oscillations is implicated in cognitive impairments in psychiatric disorders, and 5-HT3 receptors (5-HT3Rs) are suggested as therapeutic targets for cognitive dysfunction in psychiatric disorders. Using a 5-HT3aR-EGFP transgenic mouse line and inducing gamma oscillations by carbachol in hippocampal slices, we show that activation of 5-HT3aRs, which are exclusively expressed in cholecystokinin (CCK)-containing interneurons, selectively suppressed and desynchronized firings in these interneurons by enhancing spike-frequency accommodation in a small conductance potassium (SK)-channel-dependent manner. Parvalbumin-positive interneurons therefore received diminished inhibitory input leading to increased but desynchronized firings of PV cells. As a consequence, the firing of pyramidal neurons was desynchronized and gamma oscillations were impaired. These effects were independent of 5-HT3aR-mediated CCK release. Our results therefore revealed an important role of 5-HT3aRs in gamma oscillations and identified a novel crosstalk among different types of interneurons for regulation of network oscillations. The functional link between 5-HT3aR and gamma oscillations may have implications for understanding the cognitive impairments in psychiatric disorders.
The gamma wave, defined as oscillations with a frequency between 20 and 80 Hz (Fisahn et al. 2004, 2005), is commonly observed in many brain regions of mammals during wakefullness and sleep. Gamma oscillations are believed to be essential for binding neuronal assemblies in the brain for various cognitive processes such as memory (Tallon-Baudry et al. 2004; Popescu et al. 2009; Fell and Axmacher 2011; Carr et al. 2012) and sensory responses (Cardin et al. 2009), which could be significantly influenced by emotional states (Phelps 2006). Evidence suggests that abnormalities in gamma-frequency synchrony may contribute to the cognitive control impairments in some neurological and psychiatric disorders, including schizophrenia (Spencer et al. 2004; Lewis et al. 2005; Cho et al. 2006). Reduced gamma band power and synchronization are reported during working memory tasks in schizophrenia (Uhlhaas and Singer 2010). It is well established that oscillatory activity is dependent on the network of inhibitory interneurons (Whittington et al. 1995; Traub et al. 1997; Fisahn et al. 1998). At present, a broad consensus is that rhythmic inhibition by PV interneurons results in synchronous postsynaptic consequences for pyramidal cell (PC) membrane potential, leading to network gamma oscillations (Whittington and Traub 2003; Bartos et al. 2007; Freund and Katona 2007). Besides PV cells, cholecystokinin (CCK)-containing basket cells are another major type of interneuron and have been suggested as a key player in fine-tuning gamma oscillations (Freund and Katona 2007). Consistent with this hypothesis, CCK interneurons fire action potentials that are phase-locked to field oscillations (Whittington et al. 1995). Furthermore, cannabinoid receptor 1 (CB1) agonists, known to inhibit GABA release from CCK interneurons, reduce gamma oscillations (Hajos et al. 2000). Recent reports that CCK interneurons regulate firings of PV interneurons through GABAergic synaptic transmission between them (Karson et al. 2009) also raise the possibility that CCK interneurons modulate gamma oscillations via PV interneurons.
A notable feature of CCK interneurons is their unique expression of 5-HT3a receptors (5HT3aRs), which mediate emotional effects of subcortical serotonergic afferents (Freund and Katona 2007; Armstrong and Soltesz 2012). 5-HT3aR is the only 5-HT-gated cation channel (Chameau and Van Hooft 2006; Faerber et al. 2007) in the 5-HT receptor family. Activation of these receptors that are expressed in the soma, dendrites, and axons of CCK interneurons modulates membrane excitation and neurotransmitter release (Sudweeks et al. 2002; Fukushima et al. 2009) of these interneurons, to relay the information carried by serotonergic ascending afferents from midbrain raphe 5-HT neurons and mediate fast synaptic transmission in the hippocampal network (Varga et al. 2009).
In addition, converging evidence shows that 5-HT3 receptors (5-HT3Rs) modulate learning and memory processes (Maeda et al. 1994; Staubli and Xu 1995; Buhot et al. 2000; Gonzalez-Burgos and Feria-Velasco 2008). 5-HT3R antagonist ondansetron has been found to improve memory in patients of >50 years of age (Thompson and Lummis 2007). Human data also support the potential therapeutic role for 5-HT3R antagonists in cognitive dysfunction in schizophrenia. For example, administration of the selective 5-HT3R antagonists ondansetron and tropisetron improved deficits in auditory P50 suppression in schizophrenic patients (Adler et al. 2005; Koike et al. 2005). Evidence from a recent study identified two 5-HT3 sequence variations (R344H and P391R) in a small group of patients with bipolar disorder and schizophrenia (Niesler et al. 2001). Moreover, the P391R mutation may play a role in the pathology of schizophrenia (Thompson et al. 2006). However, the mechanisms involved in the therapeutic effect for 5-HT3R antagonists in cognitive dysfunction in psychiatric disorders and the roles of 5-HT3Rs in gamma oscillations remain to be clarified.
In this study, using a transgenic mouse line in which 5-HT3aR promoter drives GFP expression in CCK cells, and by pharmacological activation of 5-HT3aRs, we simultaneously recorded single-cell firing and carbachol-induced field gamma oscillations in hippocampal slices to investigate how 5-HT3aRs modulate gamma oscillations via interneurons. We show that selective suppression and desynchronization of the firings of 5-HT3aR-expressing interneurons via 5-HT3aR activation disinhibited PV interneurons and desynchronized their firings, leading to impairment of gamma oscillations. Our results reveal an important role of 5-HT3aRs in gamma oscillations and identify an inhibitory pathway among different subtypes of interneurons as a critical pathway in controlling network oscillations.
Materials and Methods
5-HT3a-GFP (Gensat BAC transgenic line DH30 expressing eGFP under the control of 5-HT3a receptor gene promoter, MMRRC strain: STOCK Tg(Htr3a-EGFP)1Gsat/Mmnc) and 5-HT3aKO (Zeitz et al. 2002) male mice were used according to the procedures approved by NIMH Animal Care and Use Committee.
Mice (6–7 weeks old) were decapitated under isoflurane anesthesia. Brains were rapidly removed and immersed in ice-cold cutting solution containing (in mm): 252 sucrose, 2.5 KCl, 4 MgCl2, 0.5 CaCl2, 1.2 NaH2PO4, 26 NaHCO3, and 10 glucose, saturated with 95% O2 and 5% CO2. Horizontal slices of 400 μm were prepared using Microslicer (DSK). The slices were incubated in ACSF containing (in mm): 124 NaCl, 3.5 KCl, 2 MgCl2, 2 CaCl2, 1.25 NaH2PO4, 25 NaHCO3, and 10 glucose, saturated with 95% O2 and 5% CO2 at room temperature for at least 1 h before recording.
Slices were transferred to a submerged recording chamber (33°C) double surfaces perfused (Hajos et al. 2004) at a rate of 4.5–5.5 mL/min with ACSF identical with the incubation solution except for the concentration of MgCl2, which was reduced to 1.5 mm. Extracellular and whole-cell recordings were performed in the hippocampal area CA1 using a Muticlamp 700B amplifier (Molecular Devices). Data were acquired at 10 kHz and filtered at 2 kHz.
In extracellular recording, 2.5- to 3.5-MΩ glass pipettes filled with ACSF were used to record field oscillations. Extracellular spikes were recorded in cell attached mode with 3- to 5-MΩ pipettes filled with ACSF containing 0.2% biocytin. After recording in a cell attached mode, negative pressure was given through the pipette to obtain whole-cell configuration for biocytin labeling. Alternatively, a different pipette filled with K-gluconate internal solution was used to obtain whole-cell configuration for biocytin labeling. During simultaneous recording of cell spike and field oscillations, the two electrodes were placed within 250 μm of each other.
Whole-cell recordings were performed using 3- to 5-MΩ pipettes with internal solution containing (in mm): 125 K-gluconate, 10 KCl, 1 MgCl2, 10 HEPES, 0.2 EGTA, 10 phosphocreatine-Na2, 4 ATP-Na2, 0.5 GTP-Na, and 0.2% biocytin. Membrane potential was tested under current clamp mode, and injected current was 0. Afterhyperpolarization (AHP) was evoked from the resting membrane potential by applying a train of four 2-ms 600-pA depolarizing current pulses given at 15-ms intervals. To investigate spike-frequency accommodation, cells were given 500 ms of depolarizing current pulses ranging from 50 to 800 pA in 50-pA increments. Spike-frequency accommodation was measured by counting the number of spikes during pulse at different membrane potentials. For sIPSCs and sEPSCs recording, Cs+ was used instead of K+ (the pH adjusted to 7.3 by CsOH) in the internal solution. Membrane potential was corrected by a junction potential and held at 0 mV in sIPSCs and –70 mV in sEPSCs recording. Picrotoxin or CNQX were added into perfusion solution at the end of recording to confirm that the recorded currents were sIPSCs or sEPSCs, respectively. Series resistance ranged from 10 to 25 MΩ and was not compensated, but if its value changed significantly (>15%) during recording, the data were not included in the analysis.
The power spectra and integral power for 20- to 50-Hz frequency range (Fisahn et al. 1998, 2004), corresponding to the peak frequency of the carbachol-induced oscillations fell within this range, were calculated for 60-s-long recording segments using fast Fourier transformation in Clampfit 9.2 (Molecular Devices).
The analysis of correlations between neuronal spikes and field oscillation was performed generally as described (Hajos et al. 2004) with minor modifications using custom-written MATLAB routine. First, field signals were band-pass filtered with cutoff frequencies of 20 and 50 Hz. Finite impulse response filter was applied in both directions to preserve the phase of oscillations, and the negative peaks of oscillations were detected. The oscillation period was the mean time between adjacent negative peaks. The 0° phase of 360° cycle was assigned to the negative peak of oscillations. Action potentials were detected using threshold-based algorithm. Phase of each spike was represented as the relative time of spike in the oscillation cycle. It was calculated as 360° × (Δti/Ti), in which Δti is the time difference between peak of spike and negative peak of oscillation and Ti is the time of oscillation cycle period.
The synchrony of firing was quantified by comparing the distribution of spike phase in oscillation cycle. Circular statistics (Hajos et al. 2004; Oren et al. 2006) was used to perform the quantification. Each spike was represented by a unit vector with a phase angle. A mean vector with the magnitude r and the direction φ was used to describe circular distribution. The and components of the mean vector were calculated as follows:
The range of r is from 0 to 1, and it reflects dispersion of the spikes phase. If the phases of spikes are clumped in 1 direction, then r is near 1.0. If the phases of spikes are randomly distributed in oscillation cycle, then r is close to 0. The angle, φ, of the mean vector was determined by solving the following equations:
Data are presented as mean ± SEM. Statistical significance was determined using paired, or unpaired t-test.
Post hoc Immunostaining
After recording, the slices were fixed in 4% paraformaldehyde overnight, washed in phosphate-buffered saline (PBS) for 2 × 1 h, and incubated for 2 h in PBST (0.3% Triton in PBS) containing 5% normal goat serum (NGS) followed by overnight incubation at 4°C in PBST containing 5% NGS with mouse monoclonal parvalbumin antibody (1 : 3000) and Cy5-conjugated avidin (1 : 500). Slices were washed in PBS and incubated for 2 h with an anti-mouse Cy3-conjugated secondary antibody (1 : 100) and washed in PBS (times). Stained cells were visualized using Zeiss LSM 410 confocal microscope.
Drugs or antibodies were obtained from the following sources: carbachol, m-CPBG, BAPTA, apamin, YM022, CNQX, Picrotoxin (Sigma–Aldrich), mouse monoclonal parvalbumin antibody (Sigma–Aldrich), Cy5-conjugated avidin, and anti-mouse Cy3-conjugated secondary antibody (Invitrogen).
5-HT3aR Activation Impairs Gamma Oscillations
To explore the role of 5-HT3aRs in hippocampal gamma oscillations, we first applied 5-HT3aR agonist m-CPBG (1 μm) and examined its effect on carbachol (25 μm)-induced gamma oscillations in the CA1 region of hippocampal slices from wide-type mice. Without 5-HT3aR activation, the integral power of oscillation (20–50 Hz) was 98.9 ± 17.2 μV2 (n = 13) with a peak frequency of 38.9 ± 0.4 Hz. m-CPBG reduced the integral power to 56.6 ± 8.2 μV2 (n = 8, P < 0.05; Fig. 1A,B,D) as well as peak power of oscillations (Supplementary Fig. 1A). However, m-CPBG had no effect on the peak frequency (38.1 ± 1.5 Hz, Fig. 1A,B,D) and half band width (Supplementary Fig. 1B). The autocorrelation of oscillations was suppressed, and the normalized autocorrelation function decay time (see Supplementary Methods) was reduced in the presence of m-CPBG (Fig. 1B and Supplementary Fig. 2). In addition, the lower concentration of m-CPBG (0.3 μm) also had effect, whereas the dose of 0.1 μm was ineffective (Supplementary Fig. 3). However, in the slices from 5-HT3aR knockout mice, m-CPBG changed neither the power of oscillation (control: 97.0 ± 9.7 μV2, n = 11; m-CPBG: 92.1 ± 9.7 μV2, n = 9, P > 0.05) nor the peak frequency (Fig. 1E), which confirmed the specificity of m-CPBG action.
As the 5-HT3aR-EGFP transgenic mouse line was used for this study, to rule out the possibility of interference by GFP knock-in on network oscillations and the effect of m-CPBG, we compared the power of carbachol-induced gamma oscillations and the inhibitory effect of m-CPBG (1 μm) on oscillations between slices from GFP+ and control animals and found that there were no differences (Supplementary Fig. 4). These results suggest that GFP knock-in does not change network property and the effect of m-CPBG on oscillations.
5-HT3aR Activation Suppresses and Desynchronizes 5-HT3aR-Expressing Interneuron Firing by Afterhyperpolarization Enhancement
To investigate the mechanisms underlying this effect, we first determined how 5-HT3aR activation affects firing of 5-HT3aR-expressing interneurons during oscillations on slices derived from the 5-HT3-GFP line (Gensat BAC transgenic, eGFP controlled by 5-HT3aR promoter), in which GFP-positive neurons expressed 5-HT3aR (as confirmed by the presence of 5-HT3aR mRNA) (Supplementary Fig. 5). In the hippocampus, 5-HT3aRs are almost exclusively expressed in CCK interneurons (Morales and Bloom 1997; Ferezou et al. 2002; Keimpema et al. 2012). These 5-HT3aR-expressing cells were PV-negative (Fig. 2A). m-CPBG reduced their firing rate from 13.3 ± 2.0 (n = 7) to 8.1 ± 1.3 Hz (n = 8, P < 0.05). Meanwhile, the r-value (magnitude of mean vector from circular statistics; see Materials and Methods), which quantifies synchrony of firings within oscillation cycle, was reduced from 0.23 ± 0.02 to 0.09 ± 0.02 (n = 8, P < 0.001), suggesting desynchronization of firings. However, the phase of mean vector (angle φ) of spikes in oscillation cycle was not altered by m-CPBG (control: 82.8 ± 14.9°; m-CPBG: 80.3 ± 18.8°, P > 0.05, Fig. 2B,C).
Given that somatodendritic 5-HT3aR activation depolarizes these interneurons in the hippocampus (Ferezou et al. 2002; Sudweeks et al. 2002), the reduction of firing rate was unexpected. To clarify the effect of m-CPBG, we monitored the membrane potential and firing frequency of the 5-HT3aR cells by whole-cell recordings. Following carbachol-induced membrane depolarization, m-CPBG further depolarized the membrane potential of 5-HT3aR cells from –41.0 ± 0.9 mV to –35.2 ± 1.0 mV (n = 7, P < 0.01). Under this additional depolarization, the spike frequency decreased from 13.4 ± 1.6 Hz to 6.9 ± 2.0 Hz (P < 0.05). In addition, AHP appeared to be enhanced (Fig. 2D,E).
Afterhyperpolarization has a significant role in spike-frequency accommodation (Savic et al. 2001; Tombaugh et al. 2005). The firing frequency of neurons decreases with enhanced AHP (Kandel and Spencer 1961). To assess AHP and spike-frequency accommodation more accurately by eliminating the interference by carbachol, we examined the effect of m-CPBG without carbachol. Under these conditions, m-CPBG induced 5.3 ± 1.1-mV depolarization of membrane potential (Supplementary Fig. 6A,B) and increased the amplitude of AHP of the 5-HT3aR cells to 193.0% of control level (from 1.9 ± 0.7 mV to 3.6 ± 0.5 mV, n = 10, P < 0.01) (Fig. 2F,G). Consequently, spike-frequency accommodation was enhanced by m-CPBG during prolonged depolarization without carbachol (Supplementary Fig. 7). Given the important role of AHP in spike-frequency accommodation, the increase in AHP could contribute to the reduction of firing frequency by m-CPBG, which is independent of carbachol.
Activation of 5-HT3 receptor can increase intracellular Ca2+ by allowing its influx directly or through voltage-gated Ca2+ channels opened by membrane depolarization (Ronde and Nichols 1998). Afterhyperpolarization is mediated largely by Ca2+-dependent K+ current (Storm 1990). Consistently, with Ca2+ chelator BAPTA (15 mm) in the internal solution, most of AHP was blocked (Supplementary Fig. 6C). Small conductance potassium (SK) channels, which are activated by elevated intracellular Ca2+, contribute to AHP that regulates neuron firings (Stocker et al. 1999; Bond et al. 2004; Villalobos et al. 2004; Disterhoft and Oh 2006). To examine the role of SK channels in m-CPBG-mediated change of AHP, we blocked SK channels by including 100 nm apamin in the bath solution. Apamin abolished the effect of m-CPBG on AHP (n = 10, P > 0.05) (Fig. 2G), suggesting that SK channels mediate this effect.
5-HT3aR Activation Reduces Inhibitory Inputs to PV Interneurons and Desynchronizes Their Firings
Gamma oscillations are entrained by PV interneurons that do not express 5-HT3aRs but receive GABAergic inhibitory synaptic projections from 5-HT3aR-expressing CCK cells (Karson et al. 2009). The decreased firings in 5-HT3aR interneurons may reduce inhibitory input to PV interneurons and therefore disinhibit these cells. To test this hypothesis, we examined how m-CPBG modulates spontaneous IPSCs and EPSCs (sIPSCs and sEPSCs) in PV cells during gamma oscillations in the presence of carbachol. The PV expression in interneurons was confirmed by infusion of biocytin and immunostaining for PV after recording. These cells were 5-HT3aR negative (Fig. 3A). Indeed, m-CPBG reduced sIPSCs frequency from 23.3 ± 2.7 to 18.9 ± 2.6 Hz (n = 8, P < 0.001) without affecting sIPSCs amplitude (control: 71.5 ± 9.0 pA; m-CPBG: 70.4 ± 8.5 pA, P > 0.05) in PV interneurons (Fig. 3B,C). In contrast, the frequency and amplitude of sEPSCs were not altered (frequency—control: 34.3 ± 4.2 Hz; m-CPBG: 33.2 ± 4.3 Hz, amplitude—control: 97.4 ± 14.8 pA; m-CPBG: 97.7 ± 14.4 pA, n = 7, P > 0.05, Fig. 3B,C). The results suggest that activation of 5-HT3aRs results in disinhibition in PV interneurons.
In line with these observations, m-CPBG increased the firing rate of PV interneurons from 24.5 ± 3.1 to 46.0 ± 6.1 Hz (n = 8, P < 0.01, Fig. 3D,E). However, the spikes were less phase-coupled to oscillation cycle in the presence of m-CPBG. The r-value of spikes was reduced from 0.41 ± 0.05 to 0.14 ± 0.02 (n = 8, P < 0.001), suggesting desynchronization in firings, whereas the angle φ remained unchanged (control: 58.3 ± 6.7°; m-CPBG: 58.2 ± 5.8°, P > 0.05, Fig. 3D,E).
The important role of synchrony of PV cells in oscillations was revealed in regression analysis showing that the synchrony of firings, r-values, of PV cells were positively correlated with oscillation power (linear regression, n = 8 and 15 for control and m-CPBG treatment, respectively. R2 = 0.7116; P < 0.01, Fig. 3F).
In addition to somatodendritic 5-HT3aRs, there are 5-HT3aRs expressed in axon terminals, which mediate presynaptic effects to regulate neurotransmitter release to its target neurons in the hippocampus (Turner et al. 2004). Exogenous CCK has been shown to directly depolarize PV cells and increases their firings via CCK-B receptor (Foldy et al. 2007). It is possible that activation of 5-HT3aRs in axon terminals of 5-HT3aR cells enhances their CCK release to increase PV cell firings. To determine whether this presynaptic mechanism was involved in 5-HT3aR-dependent inhibition of gamma oscillations, we examined effects of m-CPBG in the presence of CCK-B receptor antagonist YM022. However, YM022 (1 μm) did not prevent m-CPBG from reducing the power of gamma oscillations or from increasing and desynchronizing firing in PV cells (n = 7, Fig. 3G), suggesting that m-CPBG inhibits oscillations in a CCK-independent manner.
5-HT3aR Activation Desynchronizes Firings in Pyramidal Cells
One important consideration on the actions of 5-HT3aR-expressing CCK interneurons is that these cells also innervate PCs. However, unlike CCK→PV synapses, GABA release in CCK→PC synapses is tonically inhibited by endocannabinoids via cannabinoid type1 receptors (Losonczy et al. 2004). To examine the effect of m-CPBG on PCs, we compared the sIPSCs, sEPSCs, and firings in PCs during oscillations with or without m-CPBG. Intriguingly, m-CPBG did not change the amplitude (control: 62.1 ± 9.8 pA, m-CPBG: 65.5 ± 9.7, n = 7, P > 0.05) or frequency (control: 20.6 ± 1.7 Hz, m-CPBG: 19.4 ± 1.6 Hz, P > 0.05) in sIPSCs or sEPSCs (amplitude—control: 43.1 ± 1.0 pA; m-CPBG: 42.8 ± 1.0 pA, n = 7, P > 0.05, frequency—control: 10.6 ± 1.8 Hz; m-CPBG: 10.0 ± 1.7 Hz, P > 0.05, Fig. 4A,B). The firing rate in PCs was not significantly affected by m-CPBG (control: 6.5 ± 0.9 Hz; m-CPBG: 5.9 ± 0.7 Hz, n = 8, P > 0.05), but the r-value of firings was reduced (control: 0.36 ± 0.06; m-CPBG: 0.15 ± 0.02, n = 8, P < 0.01), suggesting desynchronization of firings in PCs, which coincides with impaired oscillations. The angle φ remained unchanged (control: 7.9 ± 9.1°; m-CPBG: 2.9 ± 10.9°, P > 0.05, Fig. 4C,D).
Previous studies on serotonergic modulation of hippocampal rhythmic oscillations showed that 5-HT3 receptor antagonist reduces the amplitude of high-frequency (200 Hz) network oscillations and increases the frequency of the hippocampal theta rhythm (Staubli and Xu 1995; Ponomarenko et al. 2003). 5-HT2c receptor activation suppresses hippocampal theta oscillations in free moving animals (Sörman et al. 2011). Activation of 5-HT6 receptor also modulates hippocampal theta oscillations (Ly et al. 2013). In addition, it has been suggested that 5-HT decreases the amplitude of prefrontal gamma rhythms through 5-HT1a receptor and 5-HT2a receptor in vivo (Puig et al. 2010). 5-HT, as well as fenfluramine, which releases 5-HT from serotonergic terminals, suppresses kainate-induced gamma oscillations in vitro (Wójtowicz et al. 2009). In this study, we provided direct evidence of the role of 5-HT3aRs in hippocampal gamma oscillations and investigated how 5-HT3aR-expressing interneurons modulate oscillations. Making use of 5-HT3aR-GFP transgenic mice, we recorded and analyzed the relation of firings of 5-HT3aR or PV interneurons and field gamma oscillations in hippocampal slices. As summarized in Figure 5, we demonstrated that reduced and desynchronized firings of 5-HT3aR-expressing interneurons via 5-HT3aR activation impaired gamma oscillation by disrupting PV interneuron firing synchrony.
Activation of 5-HT3aRs induces membrane depolarization (Zhou and Hablitz 1999; Ferezou et al. 2002), which is expected to increase the excitability of the cells. However, in this study, despite membrane depolarization induced by m-CPBG, firing frequency of 5-HT3aR cells was reduced. The reduced firings of 5-HT3aR cells can be explained by two mechanisms. Primarily, since 5-HT3aR-expressing CCK cells are non-fast-spiking interneurons, which have been reported to display spike-frequency accommodation during theta rhythm (Klausberger et al. 2005), one reason to account for the reduction in firing could be an enhancement of spike-frequency accommodation by increased mAHP from activation of SK channels following Ca2+ influx via the cation channels of 5-HT3a receptors. Consistently, by giving prolonged injections of depolarizing currents in 5-HT3aR cells, we found that m-CPBG reduced the number of spikes and enhanced spike-frequency accommodation, which is independent of carbachol or membrane potential depolarization. In addition to the above-mentioned mechanism, the second mechanism could be depolarization block resulting from cumulative depolarizing effect of carbachol and m-CPBG. The average membrane potential equilibrium value during the depolarization block (Veq), at which action potentials amplitude decreases to 0, lies in a range between −40 and −35 mV in neurons of hippocampal CA1 region (Bianchi et al. 2012). On the background depolarization induced by carbachol, m-CPBG further depolarized the membrane potential of 5-HT3aR cells to –35.2 ± 1.0 mV, which reached Veq and thus could result in depolarization block and decreased firings.
How could 5-HT3aRs activation disrupt the synchrony of firings of 5-HT3aR cells? One potential explanation resides in the diverse distributions of 5-HT3aRs in individual CCK cells. The difference in the number of receptors expressed in individual cells could lead to different levels of firing frequency accommodation and therefore phase shift among these cells when the receptors are activated, which could lead to desynchronized firings of these neurons.
As two major perisomatic inhibitory interneuron types, PV and CCK interneurons, provide parallel innervation to the perisomatic domains of PCs, regulating firing patterns of PCs with distinct properties (Whittington and Traub 2003; Freund and Katona 2007). Additionally, intricate interactions may occur between the interneurons. In the hippocampus, direct inhibitory synaptic connections from 5-HT3aR-expressing CCK to PV cells have been demonstrated and activation of CCK cells inhibits firings in synaptically coupled PV cells (Karson et al. 2009). We found that suppression of 5-HT3aR cell firing by 5-HT3aR activation reduced inhibitory inputs in PV cells. Moreover, the opposite changes in firing frequency of 5-HT3aR and PV cells suggest that 5-HT3aR cells inhibit PV cells under control conditions and that PV cells are thus disinhibited when the firing rate of 5-HT3aR cells is diminished in the presence of m-CPBG. These results are consistent with previous findings that identified CCK and PV cells are excited at different phase of gamma and theta rhythm in vivo (Klausberger et al. 2005; Tukker et al. 2007), which is probably due to delays introduced by their crossed inhibition (Karson et al. 2009).
Previous studies showed that increased synchrony of interneuron firing correlates with a higher amplitude of gamma oscillations (Tukker et al. 2007). Our results, which showed that desynchronized firing of PV interneurons reduced the amplitude of gamma oscillations, also support the positive correlation between synchrony of interneuron firing and amplitude of oscillations. However, the frequency of gamma rhythm was not changed by more frequent firing of PV cells in this effect. Therefore, one interesting implication of our findings is that only synchronous firings of interneurons contribute to network oscillations and determines the frequency, whereas the asynchronous firing could not interfere with the frequency of oscillations.
In the hippocampus, 5-HT3aRs are expressed more abundantly in postsynaptic somata and dendrites (Miquel et al. 2002). Additionally, 5-HT3aRs expressed in axon terminals could modulate neurotransmitter release of CCK cells, probably including CCK, so the role of CCK should also be considered. We showed that changes of oscillations and the firing rate of PV interneurons were not dependent on CCK release. Although it has been shown that exogenously applied CCK excites PV interneurons and enhances their firings (Foldy et al. 2007; Neu et al. 2007), thus far, it has not been found that activation of CCK interneurons triggers synaptic release of CCK. Together, these findings suggest that CCK may not act as an excitatory transmitter in the hippocampus (Karson et al. 2009) and plays a negligible role in regulation of gamma oscillation.
In addition to CCK→PV synapses, CCK neurons also project to PCs (Losonczy et al. 2004). However, it has been reported that CCK→PC synapses are tonically inhibited by endocannabinoids since high levels of cannabinoid type1 (CB1) receptor are expressed in presynaptic terminals of CCK→PC synapses (Katona et al. 1999; Losonczy et al. 2004; Neu et al. 2007; Klausberger and Somogyi 2008). Moreover, activation of muscarinic acetylcholine receptors (mAChR) by carbachol enhances endocannabinoid release and further suppresses these inhibitory inputs to PCs (Kim et al. 2002; Szabo et al. 2010). In contrast, in the CCK→PV synapses, CB1 receptor-mediated inhibition of GABA release and depolarization-induced suppression of inhibition have not been found (Karson et al. 2009). Due to this difference, it could be predicted that the effect of 5-HT3aRs during carbachol-induced gamma oscillations would be spatially restricted mainly in CCK→PV synapses but not in CCK→PC synapses, which had been suppressed by CB1 receptors. Given the increased inhibitory inputs from PV interneurons to PCs, one possible explanation for the intriguing observation of unchanged sIPSCs and firing rate in PCs may be that the increased inhibition is balanced by decreased inhibitory inputs from other interneuron class in the network due to enhanced inhibition from PV to these cells.
Pharmacological methods for inducing gamma oscillations in vitro rely on different mechanisms. While carbachol application mimics cholinergic input from basal forebrain and the resulting oscillations could be blocked by AMPA and GABAA receptor antagonists (Fisahn et al. 1998), application of kainate mimics increased glutamatergic inputs from other cortical areas and/or thalamus, resulting in oscillations that are insensitive to AMPA blockers (Fisahn et al. 2004). Given the diverse properties of gamma oscillations induced by different methods (Whittington et al. 1997; Fisahn et al. 1998, 2004; Bracci et al. 1999; Womelsdorf et al. 2007) and also the discrepancies between in vitro and in vivo data on gamma band power due to the partially severance of afferent inputs during the slicing process in in vitro studies (Gandal et al. 2012; Kocsis et al. 2013), to further understand the contribution of 5-HT3aRs in gamma oscillations and in neurological and psychiatric disorders, it is worth to investigate its role in gamma oscillations induced by other pharmacological methods or electrical stimulus and in animal models as well as patients. Nevertheless, despite the above-mentioned limitations, our study revealed a potential role for serotonergic modulation of 5-HT3aR interneurons and their crosstalk with PV cells for oscillatory activity. Our findings may facilitate the understanding of cellular mechanisms of gamma oscillations, especially the functions of CCK interneurons in generation of gamma rhythm.
This work was supported by the NIMH Intramural Research Program, NIH MH097826, and the National Natural Science Foundation of China (31070931).
We thank Drs R Wang, J Yu, and F Yang for constructive comments on the manuscript. Conflict of Interest: None declared.