Tonic inhibition mediated by extrasynaptic γ-aminobutyric acid type A (GABAA) receptors is a powerful conductance that controls cell excitability. Throughout the CNS, tonic inhibition is expressed at varying degrees across different cell types. Despite a rich history of cortical interneuron diversity, little is known about tonic inhibition in the different classes of cells in the cerebral cortex. We therefore examined the cell-type specificity and functional significance of tonic inhibition in layer 4 of the mouse somatosensory barrel cortex. In situ hybridization and immunocytochemistry showed moderate δ-subunit expression across the barrel structures. Whole cell patch-clamp recordings additionally indicated that significant levels of tonic inhibition can be found across cell types, with differences in the magnitude of inhibition between cell types. To activate tonic currents, we used 4,5,6,7-tetrahydroisoxazolo[5,4-c]pyridin-3-ol (THIP, a superagonist at δ-subunit–containing GABAA receptors) at a concentration that did not affect synaptic decay kinetics. THIP produced greater shifts in baseline holding current in inhibitory cells (low-threshold spiking [LTS], 109 ± 17 pA; fast spiking [FS], 111 ± 15 pA) than in excitatory cells (39 ± 10 pA; P < 0.001). In addition to these differences across cell types, there was also variability within inhibitory cells. FS cells with faster action potentials had larger baseline shifts. Because FS cells are known mediators of feedforward inhibition, we tested whether THIP-induced tonic conductance selectively controls feedforward circuits. THIP application resulted in the abolishment of the inhibitory postsynaptic potential in thalamic-evoked disynaptic responses in a subset of excitatory neurons. These data suggest multiple feedforward circuits can be differentiated by the inhibitory control of the presynaptic inhibitory neuron.
The primary somatosensory cortex of whisker bearing animals is characterized by the presence of multicellular structures known as “barrels,” located in layer 4 (Woolsey and Van der Loos 1970). An individual barrel corresponds to a “principal whisker” on the contralateral snout and neurons within the corresponding barrel respond to movements of the principal whisker in a direction-selective manner (Simons and Carvell 1989). Inhibition is a crucial component in the whisker-evoked response. A unitary whisker deflection results in excitation, immediately followed by inhibition (Swadlow 1995), the anatomical basis of which is that a single thalamic afferent can synapse on both excitatory and inhibitory neurons in layer 4 (White and Rock 1981). This type of feedforward inhibition helps ensure responses to only the most optimal sensory-evoked stimuli by shortening the time window of activation (Swadlow 2003; Wilent and Contreras 2005). Feedforward inhibition in barrel cortex is mediated by a specific class of inhibitory neurons: the fast-spiking (FS) inhibitory cell (Beierlein et al. 2002, 2003; Gibson et al. 1999; Swadlow 1989; but see also Porter et al. 2001). In addition to FS cells, there are both inhibitory and excitatory neurons that fire at a lower frequency than FS cells, including inhibitory low-threshold spiking (LTS) cells and excitatory regular-spiking (RS) spiny stellates and star pyramids (Schubert et al. 2003; Sun et al. 2006). FS cells are broadly tuned and extremely sensitive to whisker activation (Bruno and Simons 2002; Swadlow and Gusev 2002). This is largely due to the fact that they receive strong converging inputs from thalamic afferents emanating from multiple direction-selective relay neurons in the thalamus (Jensen and Killackey 1987; Swadlow 2003; Swadlow and Gusev 2002; White and Rock 1981). Strong excitatory inputs to FS cells cause the release of high levels of γ-aminobutyric acid (GABA), and thus the act of whisking likely increases the concentration of ambient GABA within the principal barrel.
Ambient GABA is the source of a major form of inhibitory neurotransmission known as tonic inhibition. In contrast to the more well known phasic inhibition, which is mediated by synaptic GABA type A (GABAA) receptors (Barnard et al. 1998), tonic inhibition is a persistent form of inhibition associated with extrasynaptic receptors (Farrant and Nusser 2005). Tonic inhibition can act as a powerful shunt and modulate the input–output function, or gain, of a neuron (for a review, see Semyanov et al. 2004). In this way, regulation of tonic inhibition may provide a means of maintaining a functional range of output from a diversity of input. Therefore tonic inhibition may be especially important in layer 4 barrel cortex, where FS neurons are activated by numerous and divergent thalamic afferents of differing modalities (Swadlow 2003). A variety of subunits of the GABAA receptor have been found to be involved in tonic inhibition in different regions of the CNS, including the δ-subunit (Farrant and Nusser 2005). The δ-subunit is found peri- and extrasynaptically (Jia et al. 2005; Nusser et al. 1998; Sun et al. 2004; Wei et al. 2003). These δ-subunit receptors are modulated by endogenous neurosteroids, including allotetrahydrodeoxycorticosterone (THDOC) (Stell et al. 2003). Additionally, 4,5,6,7-tetrahydroisoxazolo[5,4-c]pyridin-3-ol (THIP) acts as a superagonist at these receptors, having an even greater efficacy than that of the natural ligand GABA (Brown et al. 2002; Drasbek and Jensen 2006; Maguire et al. 2005; Ortinski et al. 2006). In contrast, THIP is only a partial agonist at γ-containing receptors (which are found in the synapse), and low concentrations of THIP do not affect inhibitory postsynaptic current (IPSC) kinetics. Therefore THIP can be used to selectively activate tonic inhibition.
It is well known that GABA released from subsets of intracortical inhibitory neurons shapes the response properties of excitatory neurons in layer 4 barrel cortex. The synchronous release of GABA generated from thalamic excitation of inhibitory cells should have effects on ambient GABA and extrasynaptic GABAA receptors, yet little is known about tonic inhibition in barrel cortex. Therefore in the present study we investigated tonic inhibition in layer 4 interneurons and its impact on feedforward inhibition. Using whole cell patch-clamp recordings, we demonstrate that THIP-induced tonic inhibition controls the output of barrel cortex neurons and, moreover, controls some excitatory microcircuits by selectively inhibiting feedforward inhibition.
All experiments were carried out in accordance with approved procedures established by the Georgetown University Animal Care and Use Committee (Protocol #07-073). For all experiments, adult Black 6 mice (C57BL/6, The Jackson Laboratory) of either sex were used. For electrophysiology recordings, mice between postnatal day 21 (P21) and P42 were deeply anesthetized with CO2 and then decapitated by guillotine. For histology, adult mice were anesthetized with isoflurane and perfused transcardially with a mixture of 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer, pH 7.4.
Preparation of slices for electrophysiology
Immediately following decapitation, brains were removed, blocked, and placed for 2–3 min in an ice-cold and oxygenated sucrose slicing solution (in mM): 234 sucrose, 11 glucose, 24 NaHCO3, 2.5 KCl, 1.25 NaH2PO4·H2O, 10 MgSO4, and 0.5 CaCl2, gassed with 95% O2-5% CO2. Two slice orientations were used. Tangential slices containing layer 4 were prepared as previously described (Fleidervish et al. 1998; Krook-Magnuson and Huntsman 2007). Thalamocortical slices (Agmon and Connors 1991) were prepared by making a simultaneous 10° horizontal and 35° from midline cut by placing the brain ventral-side down in a constructed angle indicator. The cut brain was glued cut-side down on a vibratome (Leica) stage and immersed in cold sucrose-slicing solution. Once the surface of the brain was established, slices of the first 2,400 μm were discarded. The following three slices of 300 μm each were collected.
Slices were incubated for ≥1 h prior to recording in preheated (32°C), oxygen equilibrated, artificial cerebral spinal fluid (ACSF) (in mM): 126 NaCl, 26 NaHCO3, 10 glucose, 2.5 KCl, 1.25 NaH2PO4·H2O, 2 MgCl2·6H2O, and 2 CaCl2·2H2O; pH 7.4. Slices were visualized with a fixed staged, upright microscope (Nikon, E600 FN) equipped with a ×4 objective and a ×60 insulated objective, infrared (IR) illumination, Nomarski optics, and an IR-sensitive video camera (COHU).
Whole cell patch-clamp recordings from neurons located in layer 4 of the barrel cortex were performed at room temperature (21–23°C) with continuous perfusion (2 ml/min) of ACSF. Experiments were conducted at room temperature to increase slice health, which helps maintain circuitry and also reduces the number of animals used. A recent study found that although values are affected by recording temperature, correlations between properties are not (Ali et al. 2007). A solenoid-controlled vacuum transducer produced brief suction pulses (120 psi at 20–50 ms) that were applied to break into the cell and establish the whole cell configuration. To vary the equilibrium potential for chloride (ECl), two different intracellular pipette solutions were used. When testing excitability effects of activating tonic inhibition and for feedforward experiments, a more physiological chloride equilibrium potential was desired; the intracellular solution for these experiments (“physiological/low Cl−”) contained the following (in mM): 130 K-gluconate, 10 KCl, 10 HEPES, 10 EGTA, and 2 MgCl2; ECl ≃ −60 mV. In cases where rundown appeared to be a potential complication (e.g., frequency–current [f–I] plots), MgATP (4 mM) and NaGTP (0.3 mM) were included in the intracellular solution. All other experiments were performed using a high chloride intracellular solution (“high Cl−”) (in mM): 70 K-gluconate, 70 KCl, 2 NaCl, 10 HEPES, 4 EGTA; ECl ≃ −16 mV. Experiments using antagonist alone to measure baseline shift used a modified ACSF (in mM): 126 NaCl, 26 NaHCO3, 10 glucose, 2.5 KCl, 1.25 NaH2PO4·H2O, 2 MgCl2·6H2O, 2 CaCl2·2H2O, and 0.005 GABA; pH 7.4. Glass pipettes were pulled (nonfilament borosilicate glass; Garner Glass) in five stages with a Flaming/Brown Micropipette Puller (Model P-97, Sutter Instrument) to obtain electrodes with resistances between 2.5 and 3.5 MΩ when filled with intracellular solution. Except where noted, all drugs were applied through bath application. In experiments examining baseline shift and effects on spontaneous inhibitory postsynaptic currents (sIPSCs), glutamate receptor blockers 6,7-dinitroquinoxaline-2,3-dione (DNQX, 20 μM final; Tocris Bioscience, Ellisville, MO) and (±)-2-amino-5-phosphonopentanoic acid (APV, 100 μM final; Tocris Bioscience) were included in the ACSF to isolate GABAA-receptor–mediated inhibition. Miniature inhibitory postsynaptic currents (mIPSCs) were recorded in the presence of APV, DNQX, and tetrodotoxin (TTX, 1 μM final; Sigma). Tonic inhibition was selectively activated by application of 4,5,6,7-tetrahydroisoxazolo[5,4-c]pyridin-3-ol (THIP, 20 μM final; Tocris Bioscience). To confirm that the actions of THIP were mediated through GABAA channels, or to assess tonic inhibition levels without addition of THIP, the GABAA antagonist (−)-bicuculline methobromide (BMI, 100 μM, Tocris Bioscience) was used.
Phasic and tonic inhibition in cortical neurons was measured in a gap-free acquisition mode for continuous recording of spontaneous currents. For calculating THIP-induced baseline shifts, current levels required to hold the cell at −60 mV in voltage-clamp in control and in the presence of drug were determined by creating a histogram of current levels over a time window of 30–60 s, fitting this histogram with a Gaussian curve, and taking the peak of the curve (Krook-Magnuson and Huntsman 2007). We measured baseline noise of the holding current as threefold the root mean square of the baseline between unitary inhibitory events using the program Metatape (created by J. Huguenard). Frequency, amplitude, and decay of sIPSCs were analyzed off-line using Clampfit (v. 9.2, Axon Instruments, Union City, CA). Double-exponential fits of baseline-subtracted averaged unitary inhibitory events (that decay to baseline and are not on the rising phase of a previous event) were made with the offset forced to zero. All events were best fitted to two exponentials. The double-exponential function: f(t) = A1e−t/τD1 + A2e−t/τD1 was used to identify the decay (τD1 and τD2) and amplitude (A1 and A2) of averaged IPSCs. These fits were then used to determine the weighted time constant (τD,W) = (τD1A1 + τD2A2)/(A1 + A2).
Thalamocortical slices were prepared as described earlier. The stimulating electrode was positioned so that it contacted intact fibers projecting from the ventrobasal complex of the thalamus to layer 4 barrel cortex. Recordings were made from neurons in barrels receiving input from the targeted fibers. After determining cell type by the methods subsequently described, the response properties of barrel neurons on stimulation of thalamocortical fibers were examined. Recordings were made in the whole cell current-clamp configuration with current injected to bring the baseline resting potential of the cell near −50 mV. Current pulses via the stimulating electrode were delivered at a rate of one per 15 s using CPI software (C. Pisaturo, Stanford University), and varied in amplitude from 0.1 to 1.2 mA, depending on the cell in question. For disynaptic responses, optimal pulse amplitudes were defined as those that elicited a clear excitatory postsynaptic potential (EPSP) + inhibitory postsynaptic potential (IPSP), but not an action potential. After determining the response properties of a neuron under normal conditions, ASCF with 20 μM THIP was bath applied for 4 min, and the response of the cell reexamined. The same was done after 4 min of THIP washout.
Cells were categorized into three main groups based on their response to current injection in current-clamp mode: excitatory regular spiking (RS), low-threshold spiking (LTS), and fast spiking (FS) (Supplemental Fig. S1).1 Categorization of cell type was based on the methods used in Bacci et al. (2003) and Beierlein et al. (2003). Specifically, the membrane potential of each cell was adjusted to −60 mV. Hyperpolarizing and depolarizing current pulses of 600 ms were then applied in 12 consecutive sweeps to establish an “IV test” for each cell (Bacci et al. 2003). To distinguish excitatory cells from inhibitory cells with similar firing frequencies, we measured the difference of the first afterhyperpolarization potential (AHP) from the last AHP in a depolarizing train of action potentials at threshold. Based on the findings of Beierlein et al. (2003), cells with a negative difference were classified as excitatory (Supplemental Table S1 and Supplemental Fig. S1). In separate experiments using biocytin to fill cells, we found that all cells with spiny dendrites match this AHP difference criterion for RS cells (n = 17; data not shown). We have chosen to follow the terminology of Beierlein et al. (2003) and refer to inhibitory cells firing in the same range (<75 Hz) as that of excitatory cells as “LTS.” Some, but not all, of our LTS cells fire a rebound burst following a depolarizing current step (Supplemental Fig. S1). LTS cells are also referred to as burst-spiking nonpyramidal cells (BSNPs) and show overlap in morphology and biochemical markers with low-frequency inhibitory cells that do not show a rebound burst: the regular-spiking nonpyramidal cells (RSNPs) (Kawaguchi and Kubota 1997). We did not require a rebound burst in our classification of LTS cells and this group therefore likely includes cells that could be referred to as RSNPs. In addition, all LTS cells were low-frequency cells and had an Fmax steady-state (described in the following text) firing frequency <75 Hz (i.e., in the range of excitatory cells). Cells with higher firing frequencies were classified as inhibitory FS cells (all had an Fmax steady state >95 Hz). Although not used as a defining characteristic, all FS cells also displayed a more hyperpolarized first AHP compared with the last. FS cells also showed less adaptation, smaller spike widths, and deeper AHPs than those of LTS and RS cells. Although there is often some variability between studies in various measurements—which may arise from subtle differences in recording conditions or cell typing, layer- or area-specific differences in neuron types and subtypes, species or strain differences, or natural variability within cell types—our FS and LTS cells matched previous descriptions of these cell types in layer 4 somatosensory cortex (Beierlein et al. 2000; Gibson et al. 1999). An overview of cell properties for each cell group and comparisons between groups are provided in Supplemental Table S1.
FS cells are classically associated with parvalbumin (PV) staining (Cauli et al. 1997; Kawaguchi and Kubota 1997); however, a recent study using GFP mice labeling somatostatin positive (SOM+) cells found that a population of SOM+ (and PV−) cells can also fire action potentials at a high frequency (Ma et al. 2006). Additionally, Wang et al. (2002) reported that a population of adapting cells (i.e., non-FS cells—adaptation ratio is another measure used to distinguish FS from LTS/RSNP) are likely PV+, similar to (nonadapting) FS cells. In the present study, we classified FS cells by firing frequency; in other experiments, we confirmed that cells firing in this frequency range are commonly PV+ (9 PV+, 1 SOM+). We investigated variability within our broad cell-type categories of “FS” and “LTS” by using correlation analyses with cell parameters (e.g., adaptation ratio, Supplemental Table S2).
Statistical analysis and definition of terms
In cases where data were not normally distributed (i.e., failed the Shapiro–Wilk test for normalcy), the Mann–Whitney U or Wilcoxon matched-pairs signed-ranks (for unpaired and paired, respectively) tests were used. Otherwise, two-tailed t-tests were utilized. Pearson correlation analysis values are given in the text. Pearson coefficients, as well as the nonparametric Kendall's and Spearman's coefficients, are provided in Supplemental Table S2. These nonparametric coefficients provide an estimate of correlation that does not require an assumption of normal distribution. Statistical analysis was done using Microsoft Excel 2003, Origin 7, and SPSS 15.0 and 16.0. Values in the text are reported as means ± SE. When examining changes in input resistance, input resistance was calculated from small current injections of 600 ms. Otherwise, input resistance was measured using Clampfit. Cells can display a “sag” in response to hyperpolarizing current steps, where the initial response to the step is more hyperpolarized than the later “steady-state” portion of the response. Because the sag differs depending on the degree of hyperpolarization, the Sag–V slope has been used (Ma et al. 2006). The sag–V slope here was the slope of the line of sag versus the steady-state response (V, membrane potential), resulting from −300-, −200-, and −100-pA current injections. Fmax steady state (Ma et al. 2006) here is the inverse of the average of the last three interspike intervals (ISIs) of the train, when the cell is firing at maximum frequency (with an 800-pA current-injection ceiling). Fmax initial is the inverse of the first ISI. Adaptation ratio is the Fmax steady state divided by the Fmax initial. The frequency–input current (f–I) slope was calculated from a linear regression of input (pA) versus output (firing during the entire 600 ms) up to the maximum firing frequency. Rheobase (pA) is the intercept of this linear regression with the current axis (Ma et al. 2006). Threshold (mV) for an action potential was taken as the point of inflection for action potentials at the first current injection to evoke a first full sweep (FFS) of action potentials. Spike height (mV) is the average difference between the peak amplitude of spikes and this threshold. Spike width (ms) was measured at half-height of the action potential; values reported are the average spike width for all action potentials fired in the FFS. Rise and fall rates (V/s) are averages of the maximum rates measured in Clampfit for maximum firing frequency current injections. AHP (mV) values given in Supplemental Table S1 are the difference between mean threshold and the first AHP for the FFS. “AHP Diff” (mV) refers to the difference between the first AHP and the last AHP for the FFS; positive values indicate a more depolarized first AHP compared with the last AHP.
To investigate the relationship between firing patterns and morphology, biocytin was included in the intracellular recording solution in separate experiments. At the end of each recording, after conducting the cell-typing experiments described earlier, biocytin (1%) was injected into the cell by injecting four to five depolarizing current pulses (1 nA each). Photographs and notes were taken to locate the recorded cells' locations. Slices were removed from the recording chamber and fixed in 4% PFA in 0.1 M phosphate-buffered saline (PBS, pH 7.4) for ≥24 h. Prior to resectioning, the fixed slices were transferred to 30% sucrose in 4% PFA for ≥30 min for cryoprotection. Slices were cut to a thickness of 40–50 μm on a freezing microtome and then incubated in 0.6% H2O2 for 30 min. Slices were transferred into 50% ethanol twice, 10 min each. After two rinses of 0.1 M PBS, sections were incubated in Texas-red (or fluorescein) conjugated Avidin-D (Vector Laboratories) for 1 h at room temperature. Following two 15-min washes, sections were mounted and covered with Vectashield mounting medium. Pictures of biocytin-filled cells were captured using an Olympus Fluoview laser-scanning confocal microscope.
In situ hybridization
The procedures for subcloning complementary RNA (cRNA) probes for the δ-subunit were the same as described in previous studies of GABAA-receptor subunit expression (Huntsman et al. 1994). Primers for the δ-subunit were as follows: 5′-GCC CAC TTC AAT GCC GAC TAC-3′ for the forward primer and 5′-GAT GCA GAC ACC ATC GAC ATC-3′ for the reverse. Total RNA was extracted from fresh mouse tissue and converted to complementary DNA (cDNA) in a first-strand synthesis reaction and amplified in 35 cycles of amplification using polymerase chain reaction. cDNA products were isolated from a 2% agarose gel (Qiagen) and subcloned into the Srf I restriction site of the pCR-Script SK(+) cloning vector (Stratagene). Inserted vectors were sequenced at the Georgetown University DNA facility. Probes were transcribed from linearized vectors using T3 or T7 (depending on orientation) RNA polymerase (Stratagene) and purified with phenol followed by ammonium acetate and ethanol. Probes were analyzed with a scintillation counter for relative incorporation levels.
For tangential slices through the posteromedial barrel subfield (PMBSF) the brains were blocked at a 30° angle along the central sulcus and flattened between two glass microscope slides. The slides were then placed in a large petri dish filled with 4% PFA and 30% sucrose for cryoprotection for two nights at 4°C. The flattened brains were then frozen directly on the stage of the sliding microtome (Leica), sectioned at a thickness of 40 μm, and collected in cold 4% PFA for in situ hybridization histochemistry. Free-floating sections were pretreated with phosphate buffer (0.1 M) and a sodium chloride/sodium citrate buffer. The sections were transferred to a hybridization buffer containing: dextran sulfate, deionized formamide, sodium chloride sodium citrate (SSC) buffer, and Denhardt's solution (50×). After a 1-h prehybridization period, the cRNA probe with the incorporation of a radiolabeled nucleoside (1 × 106 cpm μl−1 of [α-33P-UTP]) was added to the solution and incubated overnight at 60°C. After the overnight hybridization, sections were treated with RNase and washed in descending concentrations of SSC buffer to remove any nonspecific labeling. The sections were then mounted on slides and exposed to β-max autoradiographic film (Amersham). After development, slides were defatted and dipped in Kodak NTB2 photographic emulsion. After a 10-day exposure the slides were developed in Kodak D19 developer and fixed with Kodak rapid fixer. Adjacent sections were reserved in 0.1 M phosphate buffer for cresyl violet (Nissl) and cytochrome oxidase (CO) staining to determine barrel boundaries and levels of metabolic activity within layer 4. For CO staining the sections were rinsed in two washes of phosphate buffer (0.1 M), then placed in a filtered solution containing the following and incubated at 37°C for 2 h: diaminobenzidine (20 mg DAB, Sigma), sucrose (1.6 mg), and cytochrome c (16 mg, Sigma). The sections were rinsed twice for 10 min each and mounted onto subbed slides. Nissl-stained sections were mounted onto subbed slides and dried overnight. The slides were rinsed in ascending concentrations of ethanol, transferred to chloroform, then rinsed in a series of descending ethanols, placed in a cresyl violet solution, rinsed in ascending concentrations of ethanol, placed in SafeClear II, and finally coverslipped with permount mounting media (Fisher). All slides were analyzed on an upright microscope (Nikon E600) affixed with ×10 and ×40 objectives, and film autoradiograms were analyzed using a stereoscopic zoom microscope (Nikon EMZ1500). Images were taken with a digital camera (DXM1200F) affixed to each microscope and collected on a designated computer (Dell) with specialized software (Nikon, ACT-1).
Following perfusion, brains were cryoprotected overnight in 30% sucrose in 4% PFA and sliced tangentially with a sliding microtome. Sections were collected in PBS. Free-floating sections were then incubated at 4°C with rabbit anti-GABAA receptor δ-subunit antibody (1:1,000, Novus Biologicals), diluted in 0.1 M PBS containing 10% normal goat serum, 0.5% Triton X-100, and 2% bovine serum albumin. After three washes in 0.1 M PBS, 20 min each, the sections were transferred to fluorescein-conjugated anti-rabbit secondary antibody (Vector Laboratories) for 1 h at room temperature. Following two washes of 15 min each, the sections were mounted and covered with Vectashield mounting medium. Pictures were taken using an Olympus Fluoview laser-scanning confocal microscope.
GABAA δ-subunit expression in layer 4 barrel cortex
We investigated the mRNA expression of the δ-subunit in layer 4 of the mouse barrel cortex, using a tangential slice capturing the posteromedial barrel subfield (PMBSF). The barrels in the PMBSF correspond to the rows (A–E) of the main mystacial whiskers and are commonly observed in tangential slices using cytochrome oxidase (Fig. 1A) and Nissl (Fig. 1B) staining methods (Land and Simons 1985; Woolsey and Van der Loos 1970). Figure 1C is the film autoradiogram of the section in Fig. 1B showing δ-subunit mRNA expression in the PMBSF. Similar to the rat brain (Pirker et al. 2000; Shivers et al. 1989), δ mRNA and protein expression in the mouse cerebral cortex is present at highest levels in layers 2 and 6 with slightly lower expression in layer 4. In tangential slices, we observed specific labeling in layer 4 corresponding to the barrel pattern (marked by asterisks in both Fig. 1B and Fig. 1C). However, there was no obvious indication of preferential δ-subunit expression in the different cellular subdivisions of barrel structures (e.g., hollow vs. walls). Higher magnification revealed silver clusters indicating δ-subunit mRNA expression largely over cell bodies (Fig. 1E). Immunocytochemistry experiments using an antibody directed at the GABAA δ-subunit confirms layer 4 expression (Fig. 1F).
THIP-activated tonic inhibition in layer 4 neurons
Using whole cell patch-clamp techniques, we measured the potential for extrasynaptic GABAA-receptor–mediated tonic inhibition in thalamocortical and tangential slices through the PMBSF in layer 4 by bath application of THIP. These data represent recordings from >130 layer 4 barrel cortex neurons in control conditions and following bath application of 20 μM THIP. Our initial experiments measured THIP-induced tonic current as changes in the amount of current required to hold the cell at −60 mV in voltage clamp (baseline shifts). These experiments were done with high chloride intracellular solution and in the presence of DNQX (20 μM) and APV (100 μM) using the tangential slice capturing layer 4 of the PMBSF (Fig. 2A). Cells were grouped into three categories: excitatory and both LTS and FS inhibitory cells. All cell categories showed statistically significant baseline shifts (excitatory, 38 ± 10 pA, n = 32; LTS, 109 ± 17 pA, n = 24; FS, 111 ± 15 pA, n = 14; P < 0.001 for each, Wilcoxon matched-pairs signed-ranks) (Fig. 2). Additionally, both groups of inhibitory cells showed statistically significantly larger baseline shifts than excitatory cells (P < 0.001, Mann–Whitney U test). Although there was variability in the degree of shift within cell-type categories (explored further in the following text), there were no apparent differences across the barrel structures (Supplemental Fig. S2). When collapsed across all cell types, there was a correlation with baseline shift and input resistance (R = −0.385, P < 0.005, Fig. 2F). This correlation is driven in part by the higher Rin and smaller shifts in excitatory cells, but is not solely due to these between-cell type differences (Supplemental Table S2).
In previous studies, the expression of GABAA-receptor–mediated tonic inhibition is illustrated as a shift in holding current in the presence of antagonist. As with the agonist THIP, all cell types showed a significant baseline shift (P < 0.05, paired t-test) in the presence of bicuculline (100 μM, microperfusion; see methods for recording conditions). Also paralleling experiments using THIP application, bicuculline produced significantly larger baseline shifts in inhibitory neurons than in excitatory cells (inhibitory: 45 ± 7.9 pA; excitatory: 18 ± 7.7 pA, P < 0.05, Mann–Whitney U test, Supplemental Fig. S3).
Within LTS cells, the adaptation ratio was correlated with baseline shift (R = −0.40, P ≤ 0.05) (Fig. 3). The distribution of baseline shift versus adaptation ratio was largely continuous and did not suggest discrete subtypes. As noted earlier, we separated our inhibitory cells by firing frequency into LTS and FS cells; however, this method also largely separated our inhibitory cells by adaptation ratio. Adaptation ratio is another measure used in categorization of inhibitory neurons (Porter et al. 2001). Our FS cells generally had an adaptation ratio >0.6, whereas LTS cells mostly had lower adaptation ratios (this separation can be seen in Fig. 3C.2). Because baseline shifts of FS and LTS cells were similar in average magnitude, there was therefore no correlation between adaptation ratio and baseline shift across all inhibitory cells (P > 0.3). Although excitatory cells fire within a frequency range similar to that of LTS cells and similarly have adaptation ratios <0.6, no correlation for adaptation ratio and baseline shift was found for RS cells (P > 0.9, Fig. 3C.3).
Because FS cells may be important for feedforward inhibition (Beierlein et al. 2003; Bruno and Simons 2002; Swadlow et al. 1998), we further examined this within-category variability. Within FS cells (Fig. 4), there were significant correlations between baseline shift and spike width (R = −0.71, P < 0.005), rise and fall rate (R = 0.55 and R = 0.60, respectively, P < 0.05 for both), maximum steady-state firing frequency (Fmax steady state, R = 0.56, P < 0.05), and initial input resistance (R = −0.56, P < 0.05, Supplemental Table S2). The strongest of these correlations was with spike width (Fig. 4C.1). Spike width was additionally correlated to these other measures (rise rate, R = −0.66; fall rate, R = −0.63; Fmax steady state, R = −0.83; input resistance, R = 0.69; P < 0.05). There was no significant correlation between series resistance and spike width for FS neurons (P = 0.25). The spike width of FS cells ranged from 0.48 to 1.08 ms (mean: 0.69 ± 0.05 ms; SD: 0.18 ms) (Supplemental Table S1). Unlike the correlation between adaptation ratio and baseline shift seen with LTS cells, the distribution of spike width versus baseline shift for FS cells appears less continuous. A k-means clustering analysis supported two clusters of FS cells based on spike width, with cluster means of 0.61 and 1.00 ms. Not surprisingly, these two groups differed significantly in degree of baseline shift (129.4 ± 15.4 vs. 43.7 ± 4.9 pA, P < 0.05, t-test). The larger-spike-width FS cells are distinct from LTS cells, showing not only greater firing frequency, but also less adaptation (large-spike-width FS cells adaptation ratio 0.7 ± 0.1 vs. LTS 0.38 ± 0.03; P < 0.01, t-test). It should be noted that with an analysis of the thin-spike-width cluster alone, a correlation remained between spike width and baseline shift (R = −0.66, P < 0.05).
Although spike-width values for FS showed some overlap with spike-width values for excitatory and LTS cells, the spike width for FS cells was significantly shorter than that of both excitatory and LTS cells (excitatory, 1.73 ± 0.03 ms; LTS, 1.47 ± 0.05; vs. FS P < 0.0001, for both excitatory and LTS; Supplemental Table S1). Additionally, the spike-width values for FS cells reported here match predictions for FS cells under these recording conditions (Ali et al. 2007). Although a correlation remained between spike width and THIP-induced baseline shift when collapsed across cell categories, no statistically significant correlation existed between spike width and baseline shift for either excitatory or LTS cells (excitatory, R = −0.04, P = 0.81; LTS, R = 0.02, P = 0.94; Fig. 4C).
Excitatory cells showed a correlation with Rin (R = −0.36), Fmax steady state (R = −0.37), Fmax initial (R = −0.38), f–I slope (R = −0.37), and AHP (R = 0.4; for all, P ≤ 0.05). Correlation coefficients, including nonparametric, of baseline shift versus cell parameters for each cell group and collapsed across all cells are provided in Supplemental Table S2.
Effects of THIP on spontaneous phasic events
Low concentrations of THIP are reported to act selectively, enhancing tonic inhibition while not affecting the kinetics of synaptic IPSCs (Drasbek and Jensen 2006). In addition to the baseline shifts reported earlier, we examined the changes in baseline noise and effects of 20 μM THIP on sIPSC frequency and decay (Fig. 5). THIP produced an increase in the baseline noise across all cell types [excitatory, 57 ± 15% increase; LTS 127 ± 21% increase (Fig. 5, A–D); FS 145 ± 14% increase; P < 0.001 for all, Wilcoxon signed-ranks test]. Conversely, THIP produced no significant increases in decay time for sIPSCs (excitatory, control: 8.94 ± 0.26 ms, THIP: 8.35 ± 0.30 ms; LTS, control: 9.79 ± 0.77 ms, THIP: 9.34 ± 0.54 ms; FS, control: 4.99 ± 0.57, THIP: 5.56 ± 0.48) (Fig. 5, E and F). Nor did 20 μM THIP affect the rise time, weighted tau, amplitude, half-width, or 90–10% decay of miniature inhibitory events (mIPSCs; P > 0.05 for all, n = 4; unpublished observations). However, in line with THIP inhibiting inhibitory cells, there were clear effects of THIP on the frequency of spontaneous IPSCs across all cell types (excitatory, control: 2.25 ± 0.26 Hz, THIP: 0.98 ± 0.14 Hz, P < 0.00001; LTS, control: 2.61 ± 0.27 Hz, THIP: 1.14 ± 0.14 Hz, P < 0.00001; FS, control: 3.72 ± 0.81 Hz, THIP: 1.34 ± 0.35 Hz, P < 0.005; paired t-test) (Fig. 5G). This occurred in every cell examined (total n = 46 cells). These effects washed out within a short time window (3–4 min). Because of the increased noise in the presences of THIP, we reanalyzed IPSC frequency using a threshold for IPSC detection well above the background noise in THIP (generally a threshold >3SD beyond the baseline was used) to ensure that the decrease in sIPSCs was not due to decreased detection. There remained a significant decrease in sIPSC frequency (P < 0.01, Wilcoxon signed-ranks test). To ensure that the effects of THIP were mediated through actions on GABAA receptors, separate experiments used the GABAA antagonist bicuculline (100 μM). In all cells tested, bicuculline blocked the THIP-induced effects on holding current (n = 4, 141 ± 46% of shift blocked, Fig. 5H).
THIP-induced tonic inhibition shunts cells
The THIP-induced reduction in action-potential–dependent spontaneous-event frequency indicates a shunt of inhibitory neurons. Therefore we tested the effects of THIP on input resistance and firing frequency. These experiments were done using a low, “physiological” chloride intracellular solution without APV or DNQX in the ACSF to provide more physiological recording conditions. Bath application of 20 μM THIP reversibly decreased the input resistance of cells (RS: pre-THIP 226 ± 24 MΩ, THIP 117 ± 14 MΩ, washout 213 ± 22 MΩ, n = 24, Fig. 6A; LTS: pre-THIP 287 ± 78 MΩ, THIP 148 ± 45 MΩ, washout 206 ± 66 MΩ, n = 5, Fig. 6B; FS: pre-THIP 102 ± 11 MΩ, THIP 72 ± 7 MΩ, washout 98 ± 10 MΩ, n = 12, Fig. 6C; P < 0.05 Mann–Whitney U test). With the application of THIP, a change in the f–I (frequency–input current) plot could also be seen for all cell types, such that more injected current was required to achieve pre-THIP firing frequency levels (Fig. 6). There was a shift in the input–output relationship indicative of a simple subtraction function, fitting with previous studies of tonic inhibition using current steps (Semyanov et al. 2004).
Selective control of thalamic-evoked potentials by THIP
To investigate the role of tonic inhibition on intact inhibitory circuits, a thalamocortical slice preparation was used (Fig. 7A). Current pulses were delivered to fibers projecting from the thalamus to layer 4, and the response properties of the barrel neurons receiving this input were examined (Fig. 7). For each cell, a total of 10 sweeps per condition were collected, at the rate of one sweep every 15 s, with drug wash-in times for sweep selection based on the baseline shift data in voltage-clamp recordings represented in Fig. 5. All traces were collected with cells adjusted to a membrane potential of −50 mV. Off-line analysis of response amplitudes were performed on traces offset to 0 mV in Clampfit. Figure 7 illustrates the response properties exhibited by excitatory layer 4 barrel neurons in response to stimulation of thalamic fibers. Despite the reduction in input resistance in excitatory cells seen following the application of THIP (Fig. 6A), there was no effect of THIP on evoked monosynaptic EPSPs (percentage change of EPSP amplitude 3.19 ± 8.81%, n = 5, Fig. 7B). This may suggest that the THIP-induced shunt in an intact circuit has little functional effect directly on the postsynaptic excitatory cell or on the presynaptic thalamocortical terminals producing the EPSP. In other excitatory cells, a disynaptic response was elicited, such that the EPSP of the disynaptic response was rapidly truncated by a large IPSP (7.2 ± 0.6 ms between maximum amplitudes of the excitatory and inhibitory phases, n = 14). This EPSP–IPSP delay corresponds to the classic thalamic-evoked feedforward network (Agmon and Connors 1992; Gil and Amitai 1996). Although it is difficult to accurately measure an EPSP in a disynaptic response (due to the truncation by the IPSP), we did not detect any major reductions of the evoked EPSP by THIP (Fig. 7E). In contrast, excitatory neurons in layer 4 barrel cortex could be differentiated with respect to the sensitivity of the IPSP. Bath application of THIP revealed two distinct groups of excitatory cells with thalamic-evoked disynaptic responses: one group where THIP eliminated the IPSP (n = 5, Fig. 7C) and another group where the EPSP–IPSP response remained intact (n = 9, Fig. 7D). In the former group where THIP eliminated the IPSP, the IPSP returned during the washout period (bottom trace, black). These data suggest that the subset of excitatory cells in which THIP eliminates the IPSP receive input from inhibitory interneurons that are subject to significant tonic inhibition (such as thin-spike FS neurons represented in Fig. 4). In contrast, excitatory cells that did not exhibit a THIP-sensitive response may receive input from inhibitory neurons with low THIP sensitivity (such as the thick-spike FS neurons represented in Fig. 4). Therefore two distinct feedforward microcircuits can be differentiated by the sensitivity of the feedforward inhibition to THIP.
Qualitative observations also revealed two subtypes of layer 4 excitatory cells based on action potential firing patterns. In all five excitatory neurons in which THIP eliminated the IPSP, the action potential firing pattern prior to the first full sweep of action potentials showed variability in the ISI, or “stuttered” (Fig. 7C, inset). In all, 8 excitatory cells showed this stuttering firing pattern (denoted RSst in Fig. 7, E and F). The remaining excitatory neurons (6 of 14 cells) exhibited the classical regular-spiking pattern at or near threshold (denoted as RS in Fig. 7; see Fig. 7D, inset for an example of this firing pattern). The ISI variability was significantly different between these two groups (classical RS: 0.82 ± 0.70 s2, n = 6; RS stuttering: 4.49 ± 2.19 s2, n = 8; P < 0.05, Mann–Whitney U test). These groups did not differ on other measures, including sag (−7.89 ± 1.39 vs. −10.5 ± 1.83 mV, P = 0.12, Mann–Whitney U test); Fmax steady state (33.8 ± 3.9 vs. 33.4 ± 2.9 Hz, P = 0.94, t-test); adaptation ratio (0.24 ± 0.02 vs. 0.23 ± 0.01, P = 0.61, t-test); spike width, either at threshold (1.51 ± 0.06 vs. 1.70 ± 0.10 ms, P = 0.06, t-test) or at steady state (1.85 ± 0.06 vs. 2.01 ± 0.05 ms, P = 0.08, t-test); input resistance (276 ± 33 vs. 307 ± 37 MΩ, P = 0.55, t-test); or resting membrane potential (−69.6 ± 1.9 vs. −66.0 ± 1.1 mV, P = 0.15, t-test). There was a significant relationship between the EPSP amplitude and excitatory cell type based on ISI variability. The “stuttering” excitatory cells exhibited a larger-amplitude EPSP in the thalamic-evoked disynaptic response (4.37 ± 1.26 mV, n = 8 vs. 0.88 ± 0.31 mV, n = 6, P = 0.0378, Mann–Whitney U test, Fig. 7E). In contrast, there were no differences in IPSP amplitudes (3.71 ± 0.90 mV, n = 8 vs. 5.74 ± 1.39 mV, n = 6, P = 0.227, two-tailed t-test, Fig. 7F). The IPSPs in classical RS excitatory cells showed significantly less sensitivity to THIP than the IPSPs in stuttering RS cells (P < 0.05, Wilcoxon signed-ranks test, Fig. 7F). This suggests that the inhibitory input to stuttering RS cells, either at the inhibitory cell's soma or terminal, is selectively inhibited by THIP. To ensure that the differences between these excitatory cell types was not due to differences in their own sensitivity to THIP (rather than sensitivity of the presynaptic inhibition itself), we reanalyzed the THIP-induced baseline shifts in excitatory cells presented in Fig. 2. There was no difference in baseline shifts between classical RS and stuttering RS cells (classical RS: 27.9 ± 9.2 pA; stuttering RS: 34.6 ± 13.4 pA, P = 0.68, t-test), confirming that the difference in IPSP sensitivity between these groups is likely not postsynaptic.
To further investigate the difference suggested by their differential firing patterns, we filled cells with biocytin while recording their firing properties. Excitatory cells displayed two morphologies consistent with 1) spiny stellates and 2) small star pyramids. In all cases where a recovered morphology was spiny stellate, the cell displayed a classical RS firing pattern (Fig. 7G), whereas in four of five cases where the recovered morphology was star pyramidal, the cell displayed a stuttering RS firing pattern (Fig. 7H). Star pyramid cells were characterized by having a thick proximal dendrite and a slightly triangular soma (Egger et al. 2008).
The major findings of the present study are the following: 1) GABAA-receptor δ-subunits are expressed in layer 4 of the barrel cortex; 2) neurons in layer 4 barrel cortex show a cell-type–specific THIP-induced tonic inhibitory current; 3) excitatory cells show significantly less induced current than inhibitory neurons; 4) within cell categories, there is variability of tonic currents; 5) THIP-induced currents shunt cells, decreasing membrane resistance and firing frequency; 6) THIP increases baseline noise, decreases sIPSC frequency, and has no effect on synaptic decay; and 7) in thalamocortical slices, the inhibitory component of the disynaptic feedforward response is inhibited by THIP in a subset of excitatory cells.
Mediators of tonic inhibition in layer 4 barrel cortex
Previous studies in rat have found the highest levels of δ-subunit expression in upper layer 2 and layer 6, with lower levels in layer 4 (Pirker et al. 2000; Shivers et al. 1989). To investigate this further, we used tangential slices that isolate layer 4 primary somatosensory cortex from mouse. In the present study, we showed higher levels of δ-subunit mRNA expression in layer 4 within the PMBSF and δ-subunit expression in layer 4 cell bodies in this area of cortex using immunocytochemistry. It has become increasingly clear that different GABAA-receptor subtypes mediate different forms of functional inhibition: phasic and tonic (Farrant and Nusser 2005). Receptors containing the δ-subunit are believed to be involved in tonic inhibition (Brickley et al. 1996; Nusser et al. 1998; Semyanov et al. 2004). GABAA receptors containing the δ-subunit are located peri- and extrasynaptically, have a high sensitivity to GABA, and show minimal desensitization (Nusser et al. 1998; Saxena and Macdonald 1994). However, δ-subunits are not the only mediators of tonic inhibition. It has recently been shown that the activity of THIP at δ-subunit–containing GABAA receptors may be due to the lack of γ-subunits in these receptors, rather than the presence of the δ-subunit per se (Storustovu and Ebert 2006). There is evidence for some extrasynaptic αβ (with no γ or δ) GABAA receptors in the hippocampus (Mortensen and Smart 2006). Therefore some of the THIP-induced tonic inhibition reported may be mediated through αβ receptors. Additionally, other GABAA receptors may mediate additional tonic inhibition not investigated here (McCartney et al. 2007). However, this tonic inhibition is bicuculline sensitive and our finding of greater tonic inhibition in inhibitory cells was replicated by blocking currents with bicuculline. Greater tonic inhibition in inhibitory cells parallels findings in the hippocampus (Semyanov et al. 2003). We found that THIP increased conductance and altered the input–output relationship, shunting the cell. We also observed the downstream effects of this inhibition because THIP decreased the frequency of sIPSCs across all cell types. This is similar to what others have shown in cortical pyramidal cells (Drasbek and Jensen 2006). Greater tonic inhibition in inhibitory cells further explains the finding that THIP can act to paradoxically increase network excitability in the barrel cortex (Krook-Magnuson and Huntsman 2007). By inhibiting inhibitory cells to a greater extent than excitatory cells, and thereby decreasing the inhibitory tone onto excitatory cells, THIP increases the excitability of layer 4 networks.
Variability of tonic inhibition within inhibitory neurons
Given the large diversity of interneurons within layer 4 of the barrel cortex, it is not entirely surprising that we additionally found variability of THIP-induced currents within the inhibitory neuron population. Adaptation ratio has been used to separate classes of inhibitory cells (Porter et al. 2001). In the present study, we found variability of THIP responsiveness within LTS neurons; there was a correlation between baseline shift and adaptation ratio, such that LTS cells showing the greatest degree of adaptation also showed the largest THIP-induced shifts in baseline holding current. Moreover, excitatory cells, which have similar firing frequencies and adaptation ratios, lacked this correlation. Further, this correlation with adaptation ratio and baseline shift cannot hold across all inhibitory cells because LTS cells show greater adaptation than that of nonadapting FS cells. This shows the specificity of the correlation between adaptation ratio and baseline shift to within the LTS cell population. Furthermore, within LTS cells there was a fairly continuous distribution of both the size of baseline shift and the degree of adaptation, making discrete subtypes, at least in relation to THIP sensitivity, unlikely.
In addition to diversity within LTS cells, we found diversity within the FS class of interneurons. Morphologically, FS cells show diverse morphologies such as chandelier or basket cells (Kawaguchi and Kubota 1998; Markram et al. 2004). Biochemically, FS cells are classically associated with parvalbumin, but a class of somatostatin-positive cells can also fire at high frequencies (Ma et al. 2006). As noted in a recent study of layer 4 interneurons (Ali et al. 2007), correlation analysis may be a way to examine cell-type diversity without an a priori thorough cell-type and subcell-type categorization. In the present study, the variability of THIP-induced tonic inhibition in FS cells correlated most strongly with spike width, with additional correlations with measures that affect spike width (rise rate, fall rate). FS cells with the thinnest spike widths showed the greatest sensitivity to THIP. Ali et al. (2007) also found spike width to be the most robust measure correlating with cell morphology and synaptic connection properties. These correlations held for different species, recording temperatures, and animal ages.
Functional significance of tonic inhibition in feedforward circuits
In thalamocortical slices, we recorded thalamic-evoked disynaptic responses in layer 4 neurons consisting of an initial EPSP followed immediately by a robust IPSP (EPSP–IPSP). Despite the THIP-induced decrease in input resistance, the thalamic-evoked EPSP in excitatory cells was largely unaffected. Interestingly, THIP was able to eliminate the evoked IPSP in some RS cells. This may indicate that excitatory terminals onto RS cells are more distal than those onto inhibitory neurons, especially considering the somal expression of the δ-subunit suggested by the immunocytochemistry in Fig. 1. A second observation of these data shows that THIP sensitivity differentiates layer 4 excitatory microcircuits by the selective inhibition of the presynaptic inhibitory input, in that the evoked disynaptic IPSP was THIP sensitive in a subset of excitatory neurons. Studies both in vitro and in vivo have shown that FS cells often mediate the IPSP in the postsynaptic excitatory cell (Beierlein et al. 2003; Gibson et al. 1999; Simons 1978; Swadlow 2003). Our data in Fig. 4 show that FS cells exhibit varying degrees of THIP-induced current. These data illustrate that tonic inhibition can affect neurons involved in feedforward inhibition and further suggest that different feedforward circuits can be characterized by how they are controlled by ambient GABA (specifically the sensitivity of the presynaptic inhibitory input).
Excitatory neurons in layer 4 barrel cortex were differentiated by the sensitivity of their inhibitory inputs to tonic inhibition. It is noteworthy that there also appear to be two subtypes of excitatory neurons receiving feedforward thalamic input that can be differentiated by their firing properties. Specifically, we found that all excitatory neurons receiving THIP-sensitive inhibitory inputs have a distinct stuttering firing pattern, whereas the majority of excitatory neurons receiving THIP-insensitive inhibitory inputs showed a more classic regular spiking firing pattern. Additionally, the former group seems to receive stronger excitatory input from thalamic fibers, as illustrated by the larger excitatory component of the feedforward EPSP–IPSP response, compared with the latter group (Fig. 7E). Together, these findings suggest the possibility that, within these microcircuits, there is a relationship between the inputs to the excitatory cell and the cell's firing properties. Previous studies have shown that firing properties can be used to identify distinct cortical excitatory cell subtypes (Agmon and Connors 1992; Mason and Larkman 1990). Further characterization of these subtypes of excitatory cells will be important. To begin to address this, we examined whether morphology and firing pattern are related for these cells. In the present study, all spiny stellate cells recovered after biocytin filling displayed a classical regular-spiking firing pattern, whereas four of five star pyramids exhibited a stuttering firing pattern.
Tonic inhibition increases conductance and thus reduces membrane resistance and the time constant, modulating the input–output or gain of the neuron. Modulating the gain of a neuron helps maintain an appropriate level of output in the face of a wide range of input. This may be particularly important for cells that receive a large amount of convergent excitatory input (Jensen and Killackey 1987). Thalamocortical input to FS cells shows both divergence and convergence and thus FS cells show little selectivity but high sensitivity to the whisker-evoked response (Simons 1978; Swadlow et al. 1998). These neurons receive a high level of input and tonic inhibition may provide a means for modulating their gain and maintaining their firing within an operational window. It has been shown in the thalamocortical slice preparation that repetitive stimulation of thalamocortical fibers at a frequency of 10 Hz increases the ratio of excitation to inhibition in excitatory cells in layer 4 barrel cortex, such that IPSC amplitude decreases to a greater extent than EPSC amplitude, and that this change in feedforward inhibition is brought about by changes in the number of FS cells recruited (Gabernet et al. 2005). Given that high-frequency activity increases ambient GABA levels, it stands to reason that tonic inhibition could, to some extent, be mediating this effect. Even during high rates of stimulation not all FS cells appear to be depressed (Gabernet et al. 2005), just as during high states of whisking a continued degree of inhibition is desirable. The variability of FS cells' sensitivity to tonic inhibition would allow them to play these differing roles within cortical networks.
This work was supported by National Institute of Neurological Disorders and Stroke Grant NS-053719 to M. M. Huntsman.
We thank Dr. John Huguenard for comments on the manuscript.
↵1 The online version of this article contains supplemental data.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2008 by the American Physiological Society