Fundamental brain functions depend on a balance between excitation (E) and inhibition (I) that is highly adjusted to a 20–80% set point in layer 5 pyramidal neurons (L5PNs) of rat visual cortex. Dysregulations of both the E–I balance and the serotonergic system in neocortical networks lead to serious neuronal diseases including depression, schizophrenia, and epilepsy. However, no link between the activation of neuronal 5-hydroxytryptamine receptors (5-HTRs) and the cortical E–I balance has yet been reported. Here we used a combination of patch-clamp recordings of composite stimulus-locked responses in L5PN following local electrical stimulations in either layer 2/3 or 6, simultaneous measurement of excitatory and inhibitory conductance dynamics, together with selective pharmacological targeting and single-cell reverse transcriptase-polymerase chain reaction. We show that cortical serotonin shifts the E–I balance in favor of more E and we reveal fine and differential modulations of the E–I balance between 5-HTR subtypes, in relation to whether layer 2/3 or 6 was stimulated and in concordance with the specific expression pattern of these subtypes in pyramidal cells and deep interneurons. This first evidence for the functional segregation of 5-HTR subtypes sheds new light on their coherent functioning in polysynaptic sensory circuits.
The balance between excitation (E) and inhibition (I) results from the coordinated and highly regulated activities of recurrent excitatory and inhibitory cortical networks and governs numerous brain processes, including cortical integration of sensory information (Wehr and Zador 2003; Wilent and Contreras 2004; Marino et al. 2005; Priebe and Ferster 2005; Higley and Contreras 2006). Any disruption of the E–I balance has been involved in central pathologies such as epileptic seizures (Roberts 1984; Cobos et al. 2005), schizophrenia (Kehrer et al. 2008), and Alzheimer's disease (Small 2007). We previously showed that the somatic E–I balance is continuously adjusted to a 20–80% set point in layer 5 pyramidal neurons (L5PNs) of rat visual cortex (Le Roux et al. 2006) by homeostatic processes involving glutamate N-methyl D-aspartate (NMDA) (Le Roux et al. 2007) and γ-aminobutyric acid (GABA)A (Le Roux et al. 2008) receptors, and we revealed the endogenous cholinergic modulation of this E–I balance by nicotinic and muscarinic receptors (Lucas-Meunier et al. 2009). Neocortical L5PNs constitute a crucial population that elaborate cortical outputs following the integration of thalamocortical inputs, mainly relayed by synapses from layer 2/3 neurons onto their apical and basal dendrites (Thomson and Bannister 1998). Inhibitory cortical circuits contain a wide range of GABAergic interneuron classes that afford to monitor and balance E at specific domains of L5PNs: dendrites, soma, and axon hillock (Markram et al. 2004).
The rat visual cortex receives a massive input of extensively ramified 5-hydroxytryptamine (5-HT) fibers that derive from midbrain raphe nuclei (Dori et al. 1996) and target both excitatory and inhibitory neurons in all cortical layers (Dahlstrom and Fuxe 1964; Vu and Tork 1992). Due to a diffuse release from nonjunctional axonal varicosities, cortical serotonin can act as a neuromodulator faraway from its release sites, on 14 characterized receptor subtypes, whose majority is expressed on cortical pyramidal neurons and/or interneurons (Gu and Singer 1995; Bunin and Wightman 1998, 1999). In rat visual cortical slices, exogenous 5-HT application has opposite effects on spontaneous and evoked currents in pyramidal neurons (Zhou and Hablitz 1999), and 5-HT receptors’ (5-HTRs) activation by endogenous serotonin has been shown to decrease long-term potentiation in vitro (Edagawa et al. 2001). All 5-HTR subtypes are G protein–coupled receptors that are able to recruit multiple signaling modules, with the exception of the 5-HT3R which is a ligand-gated cation channel (Hoyer et al. 2002). This surprising versatility can be linked with most of neurobiological diseases resulting from perturbations or dysfunctions of the serotoninergic system: major depression, schizophrenia, anxiety, obsessive–compulsive disorder, drug abuse (Jones and Blackburn 2002), or epilepsy (Bagdy et al. 2007). Despite the numerous studies on 5-HTR signaling pathways, cortical distribution, and neuronal localization (Bockaert et al. 2006), their functional involvement on the modulation of the E–I balance has not been investigated.
In order to keep functional the dynamic interactions between excitatory and inhibitory neurons in the visual cortical network, no pharmacological blockade of E or I was used in this study. So considered, the method that we used (for review, see Monier et al. 2008) allows an optimal determination of the E–I balance in the soma of L5PNs following synaptic activations. The main finding of this study is that the E–I balance is increased, decreased, or unchanged following the selective blockade of 5-HT1A, 2A, 3, or 7Rs through a diversity of modulatory effects affecting the activity of both excitatory and inhibitory neurons in various cortical layers and input integration in L5PN.
Materials and Methods
Cortical 5-HT Depletion
Wistar rats were intraperitoneally injected with either physiological serum (0.9% isotonic NaCl, control group) or p-chlorophenylalanine (pCPA, 200 mg/kg) once daily for 4 consecutive days and sacrificed 24 h after the fourth injection in order to obtain the best depletion efficacy (see the comparative study of Kornum et al. 2006).
Slice Preparation and Electrophysiology
In accordance with the guidelines of the American Physiological Society, 19- to 25-day-old Wistar rats were decapitated and their brain quickly removed. A hemisection of the left hemisphere was removed, attached to the stage of a Vibratome tissue slicer (Campden Instrument, Loughborough, Leicestershire, UK), and immersed in ice-cold artificial cerebrospinal extracellular solution containing the following (in mM): 126 NaCl, 26 NaHCO3, 10 glucose, 2 CaCl2, 1.5 KCl, 1.5 MgSO4, and 1.25 KH2PO4 (pH 7.5, 310–330 mOsm). Parasagittal slices (250 μm) containing primary visual cortex were cut and transferred to an holding chamber filled with the extracellular solution, continuously bubbled with 95% O2/5% CO2, and maintained at 36 °C. Each slice was transferred to a recording chamber mounted on the X–Y translation stage of an upright microscope (Zeiss Axioskop 2 FS+) and perfused (∼2 mL/min) at room temperature (∼25 °C) with the extracellular solution. The borosilicate glass pipettes (of 3–5 MΩ resistance) contained the following intracellular solution (in mM): 140 K-gluconate, 10 4-2-hydroxyethyl-1-piperazine ethanesultanic acid, 4 ATP, 2 MgCl2, 0.4 GTP, and 0.5 ethylene glycol bis(2-aminoethyl ether)-N,N,N′N′-tetraacetic acid (pH 7.3 adjusted with KOH, 270–290 mOsm). Cells were visualized using video-enhanced differential interference contrast (DIC) optics and ×40 long working distance water-immersion lens. Stable whole-cell voltage-clamp and current-clamp recordings were obtained from L5PNs (identified from the shape of their soma and primary dendrites and from their current-induced firing profile to 1-s depolarizing steps ranging from −100 to 200 pA, see Fig. S3 for a typical L5PN spiking profile), were obtained using a Multiclamp 700A amplifier (Axon Instruments), filtered at 2 KHz by a low-pass Bessel filter, and sampled at 4 KHz using a Digidata 1322A acquisition board (Axon Instruments, Molecular Devices, Sunnyvale, CA) connected to a personal computer. Voltage data were corrected off-line for a measured liquid junction potential of −10 mV.
After capacitance neutralization, bridge balancing was done online under current clamp to make initial estimations of the access resistance (Rs). These values were checked and revised as necessary off-line by fitting the mean voltage response to a short hyperpolarizing current pulse (applied at rest) with the sum of 2 exponentials. The first one (fast decaying exponential) was used to fit the electrode response and extract Rs. The second one (slower decaying exponential) permitted to fit the membrane response and extract the membrane input resistance (see Table 2) and time constant (τm) (for review, see Monier et al. 2008). F/I curves were extracted from the current-induced firing profiles and fitted with a linear function whose mean slope allowed to estimate the intrinsic excitability of the L5PNs under our experimental conditions (see Table 2). Only cells with a resting potential more negative than −60 mV and with an access resistance lower than 25 MΩ were kept for further analysis. The access resistance was compensated off-line in voltage-clamp mode.
Electrical stimulations (1–10 μA, 0.2 ms, 0.05 Hz) were applied in 2 entrance sites of visual cortex: layer 2/3 and 6 using 1 MΩ impedance bipolar tungsten electrodes (TST33A10KT, WPI, Hertfordshire, UK) (see Fig. 1A). The stimulation intensity was adjusted in current clamp (see Fig. S1) to be strong enough to induce a subthreshold postsynaptic response due to coactivation of excitatory and inhibitory circuits but weak enough to avoid to recruit dominant nonlinear processes linked for instance to NMDA receptor activation (Le Roux et al. 2007). Thus, intensity level was always set to 2–3 times the amplitude of the stimulation necessary to induce a detectable response in current clamp. Under voltage clamp, 5–10 trials were repeated for each holding potential. A control recording was made after 15 min of patch-clamp equilibration at 5–7 holding potentials, and after 15 min of continuous drug perfusing (10 min for 5-HT perfusion), a “drug recording” was made in same conditions.
Continuous Estimation of the Synaptic Conductances in Voltage Clamp
Data were analyzed off-line with specialized software (Acquis1™ and Elphy™, Biologic UNIC–CNRS, Gif sur Yvette, France).
The analysis method, fully published in our previous papers (Le Roux et al. 2006, 2007, 2008) and extensively reviewed by Monier et al. (2008), is based on the continuous measurement of conductance dynamics during the full-time course of the stimulus-evoked synaptic response, as primarily described in vivo on cat cortex (Borg-Graham et al. 1998; Monier et al. 2003). This method received further validation on various experimental models (Shu et al. 2003; Wehr and Zador 2003, 2005; Haider et al. 2006; Le Roux et al. 2006, 2007, 2008). To estimate conductances, the neuron is considered as the point-conductance model of a single-compartment cell, described by the following general membrane equation:
Evoked synaptic currents were measured and averaged at a given holding potential. In I–V curves for every possible delay (t), the value of holding potential (Vh) was corrected (Vhc) from the ohmic drop due to the leakage current through the access resistance [Vhc(t) = Vh(t) − I(t) × Rs]. An average estimate of the input conductance waveform of the cell was calculated from the best linear fit (mean least square criterion) of the I–V curve for each delay (t) following the stimulation onset. Only cells showing a Pearson correlation coefficient for the I–V linear regression higher than 0.95 between −90 and −40 mV were considered to calculate the conductance change of the recorded pyramidal neuron from the slope of the linear regression.
The synaptically evoked global conductance term [gT(t)] was then measured by subtracting the resting conductance observed in the absence of stimulation (on a time window of 100 ms before the electrical stimulation) from the input total conductance. The synaptic reversal potential of the synaptic conductance [Esyn(t)] was taken as the voltage of the intersection between the I–V curve during the synaptic response and the I–V curve at rest. Assuming that the evoked somatic conductance change reflects the composite synaptic input reaching the soma, Esyn(t) characterizes the stimulation-locked dynamics of the balance between E and I. The global synaptic conductance [gT(t)] was further decomposed into 2 conductance components [gE(t) and gI(t)] corresponding to the activation of excitatory and inhibitory synapses, respectively, each associated with known and fixed reversal potentials. Indeed, we showed (see Supplementary data in Le Roux et al. 2006) that the I–V curve in the presence of excitatory transmission blockers (6-Cyano-7-nitroquinoxaline-2,3-dione and D-2-amino-5-phosphonopentanoate) is linear between −80 and +10 mV with a reversal potential equal to −80 mV. In the presence of bicuculline in order to block inhibitory inputs on the L5PN, the I–V curve for E is also linear between −80 and +10 mV with a reversal potential equal to 0 mV.
Accordingly, the reversal potentials used for the decomposition of the global synaptic conductance were set at 0 mV for excitatory (Eexc) and −80 mV for inhibitory conductance (Einh). In addition, these values for the reversal potentials are classically accepted and used in other studies (Wehr and Zador 2003, 2005). Einh corresponds to the reversal potential corresponding to GABAA (and not an intermediate value between GABAA and GABAB) because in the presence of QX314 in the pipette (which blocks K+ efflux, Nathan et al. 1990), no variation of the synaptic response was observed. Under our experimental conditions, Esyn(t) took any intermediate values between −80 and −40 mV (see Supplementary data, Fig. B in Le Roux et al. 2006) and thus includes between Eexc (0 mV) and Einh (−80 mV) and within the limits of our voltage excursion (−80 and −40 mV), in such a way that the mathematical conditions of the oversimplification used to calculate gI(t) and gE(t) were fulfilled.
Like all somatic recordings, our recordings cannot make rigorous estimates of synaptic events in the distal dendrites, and estimated conductances are ratio of the overall excitatory and inhibitory drive contained in the local network stimulated (Haider et al. 2006). However, our measurements are relative changes in conductance magnitude, which reflect the cumulative contributions of E and I arriving at proximal portions of the neuron. Nevertheless, the distortion of synaptic events by transient voltage-dependent channels and capacitance near to the recording site are minimized in voltage-clamp method (Borg-Graham et al. 1998).
For each component, excitatory and inhibitory, we calculated the conductance change as the mean averaged over a time window of 200 ms. The contribution of each component was expressed by the ratio of its integral value (intgE or intgI) to that of global conductance change (intgT).
Reconstitution of the Membrane Potential and Shunting I Estimation
The membrane potential trajectory (VrecT) that would have been observed in current clamp (see Fig. 2D in Lucas-Meunier et al. 2009) was reconstructed from the experimentally derived excitatory and inhibitory conductance profiles (obtained in voltage clamp) on the basis of the prediction given by the combination of the different synaptic activation sources:
Because our method gives gE and gI at the somatic level (i.e., after dendritic integration), we do not have an estimation of shunting I due to the activation of GABAA receptors (which opens chloride permeability) in conditions where the reversal potential of Cl− is close to the resting potential. Shunting effects act to reduce the membrane depolarizing effects of concurrent distal excitatory events (Kullmann et al. 2005; Mody 2005).
To estimate the shunting I previously reported by theoretical studies as a tonic-like GABAA conductance (Staley and Mody 1992; Mitchell and Silver 2003; Mody 2005), the following parameter, described as M factor (Koch et al. 1990), was calculated: M = (intVrecT − intVrecI)/(intVrecE) using the integrals of VrecT, VrecI, and VrecE. M reflects the reducing coefficient of E by shunting I at the somatodendritic level when the membrane potential is near the Cl− reversal potential.
The right hemisphere of each pCPA or isotonic NaCl–injected rat was fixed in a 4% paraformaldehyde/15% saccharose solution during 1 week at 4 °C, cut into 30-μm slices, and stored at 4 °C in a 0.1 M phosphate-buffered saline (PBS) and 0.02% sodium azid solution for further immunostaining controls of 5-HT depletion. Fixed brain slices of pCPA or isotonic NaCl–injected rats were permeabilized during 30 min (10 mM glycine and 0.4 mM triton), saturated during 60 min (2% bovine albumin serum and 3% sheep serum in PBS), and incubated (overnight at 4 °C) with a rabbit polyclonal antibody anti-5-HT plasma membrane transporter (1/750e, Calbiochem, La Jola, CA) and with a mouse monoclonal antibody anti-glutamic acid decarboxylase 67 (GAD67) (1/1000e, Chemicon International, Inc., Temecula, CA) for double-staining experiments. Bound antibodies were detected with specific secondary fluorescent antibodies from Molecular Probes (Cergy Pontoise, France) (incubated for 2 h at room temperature) conjugated to fluorescein isothiocyanate (5-HT vesicular transporter) or conjugated to Alexa568 for GAD67 detection. Controls and cross-controls were performed by omitting the primary antibody, and the rest of the procedure was similar. In all cases, no labeling was observed.
Single-Cell Reverse Transcriptase-Polymerase Chain Reaction
The mRNA was amplified from single L5PNs or from single layer 6 interneurons in primary visual cortex slices prepared from 19- to 21-day-old rats, as mentioned above. All L5PNs were identified under video-enhanced DIC microscopy, exhibited both the stereotyped soma shape (large and triangular) and thick proximal apical trunk (Larkman and Mason 1990; Mason and Larkman 1990), and showed the typical regular adaptation discharge pattern of pyramidal neurons as described by McCormick et al. (1985), and the membrane of these cells had an input resistance of 238.4 ± 44.6 MΩ (n = 28). This value is typical of L5PN populations in slice of visual cortex (Monier et al. 2008; Lucas-Meunier et al. 2009) or pyramidal neurons of prefrontal cortex slices (Cauli et al. 2000). All these criteria were considered in detail and allowed to ascertain that all cells of the L5PN pool were not regular spiking interneurons that exhibited higher input resistance and that mostly showed a fusiform-type morphology (Cauli et al. 2000). Thus, interneurons were identified in the layer 6 on the basis of both their soma shape (small and nonpyramidal or fusiform) and their specific current-clamp electrophysiological profile and had an input resistance of 406.5 ± 63.9 MΩ (n = 21). The cytoplasmic content was harvested essentially as previously described by Poea-Guyon et al. (2006) with video-enhanced DIC optical monitoring. Harvests were considered successful when cytoplasm and small organelles entering the pipette tip were observed while applying negative pressure and when the soma of the neuron shrinked. Polymerase chain reaction (PCR) primers were designed based on GenBank sequences for serotonin receptors and are indicated in the Table 1. The complete procedure is described in detail in the Supplementary methods.
|Subtype||Sequence||5′ Position||Product length|
|Subtype||Sequence||5′ Position||Product length|
|Control||Citalopram||pCPA||WAY 100635||MDL 11939||SB269070||Tropisetron|
|Input resistance (MΩ)||240.3 ± 13||235.5 ± 19||229.2 ± 21||244.1 ± 24||223.9 ± 15||236.4 ± 16||251.3 ± 25|
|F/I (AP/s/pA)||68.2 ± 36||65.9 ± 41||64.2 ± 38||60.8 ± 44||62.7 ± 42||72.3 ± 43||70.1 ± 39|
|Control||Citalopram||pCPA||WAY 100635||MDL 11939||SB269070||Tropisetron|
|Input resistance (MΩ)||240.3 ± 13||235.5 ± 19||229.2 ± 21||244.1 ± 24||223.9 ± 15||236.4 ± 16||251.3 ± 25|
|F/I (AP/s/pA)||68.2 ± 36||65.9 ± 41||64.2 ± 38||60.8 ± 44||62.7 ± 42||72.3 ± 43||70.1 ± 39|
Note: F/I, values correspond to the mean slope of frequency/intensity curves fitted with a linear function; AP, action potential.
Data reported are mean ± the standard error of the mean of n cells. Statistical significance was evaluated using the nonparametric Mann–Whitney U-test to compare unpaired conditions or the 2-tailed Student's t-test for other paired samples. In both cases, data were expressed as percentage of control values. Statistical significances are indicated in figure legends.
All chemicals used for electrophysiology experiments were obtained from Sigma–Aldrich (Lyon, France) except citalopram, a-Phenyl-1-(2-phenylethyl)-4-piperidinemethanol (MDL) 11939, GR113308, and SB269970 (Tocris Bioscience, Bristol, UK). Ritanserin, MDL 11939, and GR113808 were dissolved in dimethyl sulfoxide (DMSO) and diluted in the extracellular solution. The final concentration of DMSO in the slice was 1/10 000 (v/v).
Stable somatic voltage-clamp recordings of L5PN subthreshold postsynaptic responses evoked by layer 2/3 or 6 electrical stimulation were obtained in slices from visual cortex (Fig. 1A and Fig. S1), and the decomposition method (Monier et al. 2008) was applied. For each recording (see Fig. 1B for a typical control recording and decomposition), the total input conductance (gT) was first extracted and its decomposition allowed to further evaluate the relative contribution of evoked excitatory and inhibitory inputs reaching the soma of the recorded pyramidal neuron (see Materials and methods). Typical layer 2/3 (Fig. 1B) or 6 (not shown) electrical stimulation produces a fast excitatory conductance (gE) elicited before a long-lasting inhibitory conductance (gI) in the L5PN, as shown by the latency distributions for gT, gE, and gI (Fig. 1E, layer 2/3). Layer 6 latencies were shorter with a preferential distribution between 10 and 14 ms for gT (not shown). Quantification of these somatic conductances showed that the control stimulus-locked composite signal at the soma of L5PNs is composed in average of 20% of E and 80% of I whatever the stimulated layer was (control pool, n = 116), as already reported (Le Roux et al. 2006). Distribution of integer and peak values of gT, gE, and gI over the control population (Fig. 1C, layer 2/3 stimulation) showed that the control E–I ratio is not correlated to conductance amplitude and is independent of the stimulation intensity (Fig. 1D). Similar distributions were observed with layer 6 stimulation (not shown). The shape of the corresponding reconstructed voltage drive dynamic (Vrec trace, Fig. 1B) fitted well with the voltage response recorded in current clamp in the same cell (Vm trace, Fig. 1B) (see also Lucas-Meunier et al. 2009). This voltage drive dynamic determines the ability of L5PNs to produce a putative response that may occur when any additional and concomitant input reaches the soma (i.e., responsiveness threshold) and allows the quantification of shunting I (here estimated as M factor, see Materials and methods) acting at the proximal part of the apical dendrite. The membrane input resistance and intrinsic excitability of L5PN (estimated from the slope of F/I curves, Table 2) included in the present database were not changed under the different experimental conditions compared with the control pool.
Endogenous Activation of Whole 5-HTRs Increases the E–I Balance
The presence of an extensive network of 5-HT fibers in the rat primary visual cortex was first confirmed by immunostaining the serotonin plasma membrane transporter (SERT) in acute cortical slices (Zhou et al. 1996). As expected, many thin and ramified 5-HT fibers containing small varicosities were observed in all cortical layers and close to GABAergic interneurons (see Fig. S2 in the Supplementary material). Furthermore, these slices exhibited a massive endogenous serotonin immunoreactivity in cortical input (L2/3 and L4) and output (L5) layers and to a lesser extent in the layer 6 (data not shown). Bath application of the selective SERT inhibitor citalopram (10 μM, 20 min) was then used in order to determine whether an increased tone of serotonin in the slice (Luparini et al. 2004) could affect the E–I balance in L5PN. SERT blockade led to a decreased amplitude of gT, gE, and gI when compared with the control response (Fig. 2A, layer 2/3 stimulation). Quantification of 18 experiments showed a large decrease in inhibitory conductance (roughly −25%) compared with excitatory conductance (Fig. 2B), and this led to a decreased inhibitory drive of the representative reconstructed voltage trace (Vrec, Fig. 2A). Indeed, SERT blockade produced an increased ratio of E and I from 19.9–80.1% to 23.4–76.6% (Fig. 2C, P < 0.01). This modulatory effect appeared 20 min after the onset of bath perfusion (no modulation was observed after 10 min). Similar modulations that increased net excitatory drive in L5PNs were observed when layer 6 was stimulated (Fig. 2B,C). Finally, as a control experiment, the perfusion of citalopram on 5-HT–depleted slices (pCPA depletion, see below and Materials and methods) failed to produce any significant change of gT, gE, and gI (n = 9, Fig. 2D).
In order to confirm this outcome, the pharmacological depletion of the serotoninergic system using the tryptophan hydroxylase inhibitor pCPA (see Materials and methods) was carried out. The depletion efficacy in pCPA-injected animals used for electrophysiological recordings was systematically checked by the immunolabeling of 5-HT fibers in cortical slices. Contrary to control animals (NaCl injected), almost all 5-HT fibers were depleted in each pCPA-injected animal, thus validating the depletion protocol (see Fig. S2). Whereas no significant modification of the current responses and conductance (Fig. 3A, “control”) or of the E–I balance (Fig. 3C, “c”) was reported in slice obtained from NaCl-injected animals, pCPA slices exhibited a decreased E–I balance (13.6–86.4%) compared with the typical 20–80% control set point (Fig. 3C, n = 75, P < 0.001, unpaired U-test; see also Fig. S3). This E–I balance shift appeared to be due to an increase in I as illustrated by the prominent GABAergic component of the composite current response that led to a dramatic increase of gI proportion over gE (Fig. 3A, middle panel, layer 2/3 stimulation). The corresponding Vrec showed a reduced excitatory drive (Fig. 3A, middle pannel) that can support a decrease of the putative responsiveness threshold of L5PNs after 5-HT depletion. Results obtained with layer 6 stimulation revealed a similar shift of the E–I balance after 5-HT depletion (Fig. 3C). Next, we checked whether this E–I balance shift was truly due to lack of activation of all 5-HTR subtypes in the slice or to a change in cortical connectivity during the 4 days of pCPA treatment, by perfusing exogenous serotonin (100 μM, 10 min) on pCPA slices. 5-HT perfusion strongly decreased intgI (compared with intgE that weakly decreased, Fig. 3B) and led to a partial recovery of the control E–I balance to 16.6–83.4% (n = 21, P < 0.05, Fig. 3C) and of the voltage drive shape (Fig. 3A, right panel). A similar recovery was obtained when the layer 6 was stimulated (see Fig. 3B,C). Altogether, these results strongly suggest that a global activation of 5-HTRs in visual cortex by endogenous serotonin shifts the E–I balance in favor of E. Because 5-HTRs’ blockade in visual cortical slices has been shown to produce neuronal modulations depending on the endogenous 5-HT tone (see Edagawa et al. 2001), selective blockers were used to further dissect the specific modulatory function of four 5-HTR subtypes (1A, 2A, 3, and 7) in the serotoninergic tuning of the E–I balance.
5-HT1A Blockade Increases Excitatory and Inhibitory Input but Does Not Change the E–I Balance
The potent and selective antagonist N-[2-[4-(2-methoxyphenyl)-1-piperazinyl]ethyl]-N-(2-pryidyl)cyclohexanecarboxamide (WAY) 100635 (20 nM) was used to block the 5-HT1AR in the cortical network. Both inward and outward currents evoked in the L5PN by layer 2/3 stimulation and the corresponding total (gT), excitatory (gE), and inhibitory (gI) conductances were increased after 5-HT1AR blockade (Fig. 4A). Irrespective of the stimulation location, integers of total (intgT), excitatory (intgE), and inhibitory (intgI) conductances were similarly increased by, in average, roughly 35% (Fig. 4B, n = 16, P < 0.01). The E–I balance thus remained close to the 20–80% set point after 5-HT1AR blockade whatever the stimulated layer was (Fig. 4C), and no change in the L5PN voltage drive shape was observed (not shown).
5-HT2A Blockade Decreases or Increases the E–I Balance with Regard to Stimulation Location
Given the strong modulation of the E–I balance observed after the unselective blockade of whole 5-HT2Rs with ritanserin (see Fig. S4), we emphasized on the role of the 5-HT2A subtype using the selective MDL 11939 antagonist (400 nM). Surprisingly, a large increase of the inhibitory conductance and an unchanged excitatory one was observed with layer 2/3 stimulation (Fig. 5A, left), whereas they were both decreased with layer 6 stimulation (Fig. 5A, right) after 5-HT2AR blockade. The quantification of 15 experiments confirmed these observations (Fig. 5B) and revealed a decreased (15.4–84.6%, P < 0.01) or an increased (22.9–77.1%, P < 0.001) ratio of E and I compared with the control set point when stimulation was applied in layer 2/3 or 6, respectively (Fig. 5C). Because the M factor increased when layer 6 was stimulated (from 0.68 to 0.76, P < 0.05), the shunting I decreased when deep layer 5–targeting afferents were recruited.
5-HT7 Blockade Strengthens Inhibitory Drive in L5PNs
We then examined the role of 5-HT7R positively coupled to adenylyl cyclase. Blockade of 5-HT7Rs with 500 nM of the selective SB269970 antagonist led to an important increase of gT due to an increase of gE and gI (Fig. 6A). Indeed, intgI was strongly increased (+44 ± 9%, P < 0.001) compared with intgE (+18 ± 9%, P < 0.05) when layer 2/3 was stimulated (Fig. 6B), thus increasing the net inhibitory drive as illustrated by the reconstructed voltage responses (Fig. 6A) and the E–I balance changes (17.5–82.5% after SB269970 perfusion, Fig. 6C, P < 0.001). The M factor decreased from 0.67 to 0.57 (P < 0.05), indicating an increase in the shunting I. Similar modulations were obtained when layer 6 was stimulated (see Fig. 6B,C) except for the M factor that did not vary significantly.
L5PNs and Layer 6 Interneurons Show Distinct 5-HTR mRNA Profiles
In order to better understand the role of the 5-HT1A, 2A, and 7Rs in the E–I balance tuning, we conducted single-cell reverse transcriptase (RT)-PCR experiments on L5PNs and deep interneurons of the visual cortex (see Materials and method for the identification of the cells). Cells were harvested, and the different mRNA profiles were established by reverse transcription and 2-round PCR amplification (see Materials and methods). A representative L5PN, whose typical spiking profile is shown in Figure 7A, coexpresses 5-HT1A and 5-HT2ARs but does not express the 5-HT7R (Fig. 7B). In 28 individual L5PNs (Fig. 7C), a vast majority expressed the 1A or the 2A subtypes (79% or 71%, respectively) and 32% the 5-HT7 subtype. A coexpression of both 1A and 2A subtypes was found in 63% of L5PNs. Three neurons coexpressed the 3 subtype mRNAs. 5-HT2AR mRNA was also detected in single layer 6 pyramidal neurons (data not shown).
Two GABAergic populations are mostly expressed in the cortical layer 6: large basket cell and Martinotti cells. These interneurons target either L5PNs’ soma and proximal apical/basal dendrites (basket cells) or distal apical and perisomatic dendrites (Martinotti cells), and both exhibit irregular spiking (IS) or fast spiking (FS) electrophysiological profiles (Markram et al. 2004). When pooled together, 62% of the interneurons harvested in the layer 6 (IS and FS cells) expressed the 5-HT2A mRNA (13/21) and 43% expressed the 5HT7 mRNA (9/21). The 5-HT1A mRNA was rarely amplified (less than 10%). Individually, whereas IS cells (Fig. 7D) and STUT cells (stuttering, not shown) never coexpress 5-HT2A and 5-HT7 mRNAs (n = 15), FS cells (Fig. 7E) show a strict coexpression of both of these subtypes (n = 6). These specific patterns strongly suggest a differential tuning role of 5-HT on the activity of distinct inhibitory populations that are known (Markram et al. 2004; Silberberg 2008) to influence dendritic processing and synaptic input integration in specific postsynaptic domains of L5PN dendrites (see also Fig. 9).
5-HT3 Blockade Leads to an Afferent-Specific Increase of the E–I Balance
Finally, we investigated the role of the ionotropic 5-HT3R. When layer 2/3 was stimulated, 5-HT3R blockade with tropisetron (200 nM) elicited a weak decrease of intgT (Fig. 8A inset, −9 ± 2%, P < 0.05, n = 17), due to a surprising increase of intgE (+15 ± 6%, P < 0.05) and an expected decrease of intgI (−24 ± 7%, P < 0.01). This produced a strong shift of the E–I balance in favor of E (25.6–74.4% after tropisetron perfusion, Fig. 8B, P < 0.001) and thus increased the responsiveness threshold of L5PNs (see the increased excitatory drive of Vrec after 5-HT3R blockade, Fig. 8C). In these conditions, the shunting I was decreased (the M factor increased from 0.68 to 0.80, P < 0.01). Interestingly, layer 6 stimulation failed to produce any significant modification of excitatory and inhibitory conductances (Fig. 8A) nor any changes in the E–I balance (data not shown) after 5-HT3R blockade.
We show in this study that a global activation of 5-HTRs by exogenous 5-HT or by endogenously released 5-HT both increases the net excitatory drive in L5PNs, and we reveal fine and differential modulations of the E–I balance in relation with both 5-HTR subtypes and stimulation location.
5-HT1ARs Tune Excitatory Transmission but not the E–I Balance in Cortical Networks
The hyperpolarizing 5-HT1AR subtype is predominantly localized at the soma and axon hillock of pyramidal neurons (DeFelipe et al. 2001; Hoyer et al. 2002; Czyrak et al. 2003) and is widely pharmacologically targeted for depression, anxiety, and migraine (Lanfumey and Hamon 2004). In line with its controversial expression in cortical interneurons (Palchaudhuri and Flugge 2005), we revealed a very weak expression of the 5-HT1AR mRNA in deep L5PN-targeting interneurons (see Fig. 7). Surprisingly, while an expected increase in excitatory inputs was observed after 5-HT1AR blockade, inhibitory inputs were also increased in the same proportion (see Fig. 4C). This apparent discrepancy must be accounted for the specificity of the stimulated network. Indeed, because 90% of the I received by the recorded L5PN are disynaptic (Le Roux et al. 2008), the increase in inhibitory inputs is likely due to the upstream depolarization of excitatory networks targeting inhibitory interneurons that project on the recorded L5PN. This is a functional evidence for a lack of direct effect of the 5-HT1AR in cortical inhibitory networks recruited by electrical stimulation of layers 2/3 or 6 because otherwise the increase in inhibitory inputs should have been stronger than the excitatory ones (after WAY 100635 perfusion). Moreover, because similar effects were observed with both stimulation sites (layer 2/3 or 6, see Fig. 4C), 5-HT1ARs should be homogeneously distributed in superficial and deep layers of primary visual cortex as observed elsewhere in cingulate cortex (Czyrak et al. 2003). Regarding the recurrent axonal projections between excitatory and inhibitory neurons in cortical columns (Bannister 2005), we bring here new functional data suggesting that 5-HT1ARs can strongly break the output of excitatory cortical networks (Amargos-Bosch et al. 2004) without changing the E–I balance in L5PNs.
Bidirectional and Afferent-Specific Tuning of the E–I Balance by 5-HT2ARs
The 5-HT2AR is heavily expressed in L5PN soma and dendrites (especially in the proximal part, Bockaert et al. 2006), and its activation mediates E both in vitro and in vivo (Ali et al. 1999; Zhou and Hablitz 1999; Amargos-Bosch et al. 2004; Beique et al. 2004; Villalobos et al. 2005; Beique et al. 2007). Spontaneous inhibitory postsynaptic currents are also increased in pyramidal neurons following 5-HT2AR activation (Zhou and Hablitz 1999). Nonetheless, it still remains hazardous to predict consequences of 5-HT2AR activation on the E–I balance.
First, our results corroborate recent findings (Beique et al. 2007) showing that 5-HT2A–mediated E in L5PNs depends on the activation of deep L5PN-targeting pyramidal neurons (see, in Fig. 5B, the discrepancy of intgE modulation between layer 2/3 and 6 stimulation). Activity of this deep L5PN population was most probably reduced in the presence of MDL 11939. If so, then a slightest disynaptic activation of deep interneurons projecting on L5PNs should also occur in these conditions. Indeed, we report a marked decrease of the inhibitory conductance when layer 6 was stimulated (Fig. 5B). This result also suggests the presence of 5-HT2ARs on a subpopulation of deep interneurons whose activity has decreased following 5-HT2AR blockade (intgI is more decreased than intgE). In agreement with this, we found that 5-HT2A mRNA is expressed in many layer 6 interneurons which are thought to be basket cells and Martinotti interneurons targeting L5PNs’ soma and dendrites (see single-cell RT-PCR results), as previously proposed (Santana et al. 2004).
In contrast, the 5-HT2AR should only be weakly expressed in layer 2/3 bipolar and double-bouquet interneurons (Markram et al. 2004) because this subtype is specifically expressed in parvalbumin- and calbindin-positive cells but not in calretinin cells (Jakab and Goldman-Rakic 2000; de Almeida and Mengod 2007). Both of these interneuron classes provide the vast majority of inhibitory inputs received by the apical dendrite of L5PNs and originating from superficial cortical layers (Markram et al. 2004). The unexpected increase in I reported after 5-HT2AR blockade with layer 2/3 stimulation is thus consistent with a lack of 5-HT2AR–mediated change in inhibitory afferent inputs onto L5PNs’ apical dendrite (intgI would have decreased in the presence of MDL 11939 otherwise). Conversely, this latter result underlies a postsynaptic upregulation of inhibitory currents in L5PNs’ apical dendrite (we showed that more than 70% of single L5PNs express the 5-HT2AR mRNA) probably due to a decreased protein kinase C-dependent phosphorylation of dendritic GABAA receptors consecutive to 5-HT2AR blockade (Feng et al. 2001). To conclude, we report distinct and layer-dependent actions of the 5-HT2AR that strongly support a functional relevance of its specific distribution in a sensory circuit, and we propose a model for 5-HT1A–5-HT2A interaction in dendritic integration (see Fig. 9) on the basis of the former proposal by Amargos-Bosch et al. (2004).
5-HT7Rs Tune the E–I Balance Independently of the Stimulated Afferents
Up to now, very little information is known about the functional role of the 5-HT7R in cortical processes (Hedlund and Sutcliffe 2004). Here, we report that blockade of this subtype produced a decrease of the E–I balance, due to a predominant increase of the I received by the recorded L5PN compared with a small increase of E. We also showed that a large proportion of layer 6 interneurons (43%) and more than 30% of single L5PNs expressed the 5-HT7R mRNA (Fig. 7). Thus, due to the disynaptic organization of I in the stimulated network, the strong increase in I recorded in the L5PN after 5-HT7R blockade is likely due to a combination of effects affecting both excitatory and inhibitory networks targeting L5PNs. Contrary to the 5-HT2A and 5-HT3 subtypes that showed afferent-specific modulations, the modulatory effects mediated by the 5-HT7 subtype appear unspecific of the stimulated afferents as expected, given its broad expression in various cortical layers, in both pyramidal cells and interneurons (Vanhoenacker et al. 2000; Neumaier et al. 2001). To conclude, the 5-HT7R distribution in cortical networks appears functionally relevant in such a way that it should permit a broad modulation of the inhibitory drive received by the L5PN in response to synaptic activations in either superficial or deep cortical layers.
5-HT3Rs Provide a Strong and Afferent-Specific Way to Decrease the E–I Balance in L5PNs
It is now well accepted that 5-HT3Rs are preferentially expressed on interneurons (Puig et al. 2004; Palchaudhuri and Flugge 2005; Chameau and van Hooft 2006) and mainly restricted to cholecystokinin/calretinin/vasoactive intestinal peptide (VIP)-positive and somatostatin/parvalbumin-negative GABAergic cells (Morales and Bloom 1997; Ferezou et al. 2002). 5-HT3Rs should therefore be restricted to bipolar cells and double-bouquet cells in upper cortical layers (Markram et al. 2004). Here, we bring the first functional evidence for a specific distribution of 5-HT3Rs in superficial cortical layers (neither change in excitatory nor inhibitory conductances was observed after 5-HT3R blockade when layer 6 was stimulated, Fig. 8) that match with the distribution of most 5-HT3Rs in cortical layers 1, 2, and 3 (Puig et al. 2004). Thus, the decrease in I observed after 5-HT3R blockade in upper layers (layer 2/3 stimulation) is most probably due to a decreased activity of bipolar and double-bouquet interneurons. As for the surprising increase in excitatory inputs, it is linked to the decrease in GABAergic shunting I, which normally attenuates distal inputs received by the apical dendrite of L5PNs. Indeed, the M factor is increased after 5-HT3R blockade, and this should be mostly supported by the decreased activity of bipolar and double-bouquet interneurons which axons target the proximal part of L5PNs’ apical dendrite (Markram et al. 2004) where the shunting I is the most efficient. Finally, these data reveal that the critical localization of the 5-HT3R permits an efficient modulation of specific L5PN-targeting inhibitory networks, and this allows a fine-tuning of both E and I received by L5PN apical dendrite.
Functional Segregation of 5-HTRs and Cortical E–I Balance Fine-Tuning
Cortical inhibitory system must be highly regulated to match E regardless of the intensity and complexity of stimuli (Markram et al. 2004). Here, we revealed that serotonin and specific 5-HTR subtypes (5-HT1AR excepted) predominantly tune inhibitory networks’ activity and/or inhibitory input processing in L5PNs (modulations of the inhibitory drive of the stimulus-locked composite responses are always stronger compared with excitatory drive ones) and, via modulations of the shunting I, regulate excitatory input integration at the proximal part of L5PN apical dendrite (Semyanov et al. 2003, 2004), that is, before output editing at somatic level. Altogether, we highlight complex 5-HT modulations that resulted in a fine-tuning of the E–I balance around the 20–80% set point contrary to the partial blockade of GABAA receptors which enhances the E–I balance to a 40–60% ratio (Le Roux et al. 2008), a value which is discrepant with a normal cortical activity. Thus, we suggest that the functional segregation of various 5-HTR subtypes in excitatory circuits, in specific domains of L5PNs (dendrite, soma, or axon hillock), and in distinct populations of GABAergic interneurons allows the serotoninergic system to finely balance E and I while keeping the sensory responsiveness of L5PNs in a functional range (i.e., avoiding hyper- or hypoactivity) and so ensures a dynamic control of layer 5 output processing in cortical sensory networks. This also questions the specificity of serotoninergic projections from various rhythmic raphe nuclei (Molliver 1987) that can lead to a continuous activation of some 5-HTRs on specific neuronal populations and therefore allowing a refine control of L5PN outputs in response to complex stimuli in vivo (Puig et al., 2005), in keeping with this view, understanding how the concerted action of multiple 5-HTR subtypes step in active cortical circuits seems critical to elucidate the pleiotropic functions of serotonin in normal and pathological states.
We are grateful to Dr Michel Hamon for his helpful comments. Confocal microscopy imaging was performed on “The Imaging and Cell Biology facility” of the IFR87 (Gif sur Yvette, France). Conflict of Interest: None declared.