In this study, Wistar rats (Charles River, either sex) aged 18–33 postnatal days (P18–P33) were maintained on a 12/12-h light/dark cycle with lights on from 7 a.m. to 7 p.m. Rats were anaesthetized with isoflurane at a concentration <0.1% and decapitated between 10:30 a.m. and 11:30 a.m. The brain was quickly removed and placed in an ice-cold modified artificial cerebrospinal fluid (ACSF) containing a high Mg²⁺ and a low Ca²⁺ concentration (4 mM MgCl₂ and 1 mM CaCl₂), other components are same to that in the perfusion ACSF as described below, to reduce potentially excitotoxic synaptic transmission during slicing. In order to maintain adequate oxygenation and a physiological pH level, the solution was constantly bubbled with carbogen gas (95% O₂ and 5% CO₂). Thalamocortical slices (Feldmeyer et al., 1999; Qi et al., 2017) were cut at 350 μm thickness using a Leica VT1000S vibrating blade microtome and then transferred to an incubation chamber containing preparation solution for a recovery period of at least 30 min at room temperature before being transferred to the recording chamber. After cutting, slices from animals older than P21 were transferred to a holding chamber placed in a water bath at 35°C for 30 min and then, the water bath was allowed to gradually cool down to the room temperature.

During recordings, slices were continuously superfused (perfusion speed ∼5 ml/min) with ACSF containing (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 1 MgCl₂, 2 CaCl₂, 25 NaHCO₃, 25 D-glucose, 3 mho-inositol, 2 sodium pyruvate, and 0.4 ascorbic acid, bubbled with carbogen gas (95% O₂ and 5% CO₂) and maintained at 30–33°C. Patch pipettes (5–8 MΩ) were pulled from thick-wall borosilicate glass capillaries and filled with an internal solution containing (in mM): 135 K-gluconate, 4 KCl, 10 HEPES, 10 phosphocreatine, 4 Mg-ATP, and 0.3 GTP (pH 7.4 with KOH, 290–300 mOsm). Biocytin at a concentration of 5 mg/ml was added to the internal solution in order to stain patched neurons after recordings.

Slices and neurons were visualised using an upright microscope equipped with an infrared differential interference contrast (IR-DIC) optics. The barrels can be identified in layer 4 as dark stripes with light “hollows” at low magnification (4× objective) and were visible in 6–8 consecutive slices. Neurons located inside the barrels were randomly selected for recordings. When being visualised at high magnification (40× magnification), putative excitatory neurons have ovoid-shape somata without obvious apical dendrites and putative interneurons have enlarged oval somata. They could also be differentiated by their action potential (AP) firing patterns during recording and by their morphological appearances thereafter. Whole-cell patch clamp recordings were made using an EPC10 amplifier (HEKA, Lambrecht, Germany). Signals were sampled at 10 kHz, filtered at 2.9 kHz using Patchmaster software (HEKA), and later analysed off-line using Igor Pro software (Wavemetrics, United States).

Custom-written macros in Igor Pro 6 (WaveMetrics, Lake Oswego, OR, United States) were used to analyse the recorded electrophysiological signals. Passive and active AP firing properties were assessed by eliciting a series of 1 s current pulses under current clamp configuration. The series resistance and capacitance were carefully adjusted after breaking through the membrane into whole-cell mode and continuously compensated by 80% during recordings. Membrane potentials were not corrected for a junction potential. Neurons with a series resistance exceeding 40 MΩ or with a depolarized resting membrane potential (>−55 mV) after rupturing the cell membrane were excluded from analysis. The resting membrane potential (V_rest) was recorded immediately after establishing the whole-cell recording configuration. Other passive membrane properties such as the input resistance R_in, membrane time constant τ_m, voltage sag were measured from membrane potential (V_m) traces induced by a series of hyper- and depolarizing subthreshold current pulses. Single AP properties such as the AP threshold, amplitude, half-width, afterhyperpolarisation (AHP) amplitude were measured for the first spike elicited by a rheobase current step. Repetitive firing properties such as the maximum firing frequency, slope of frequency-current curve were measured. The description of most electrophysiological parameters for data analysis has been described previously (Emmenegger et al., 2018).

Acetylcholine (30 μM) was applied through the perfusion system. Atropine (ATRO, 200 nM), mecamylamine (MEC, 10 μM), tropicamide (TRO, 1 μM), pirenzepine (PIR, 0.5 μM), dihydro-ß-erythroidine (DHßE, 10 μM), TTX (0.5 μM) and the cocktail of synaptic blockers including CNQX (10 μM), D-AP5 (50 μM), gabazine (10 μM) were all bath-applied; drugs were purchased from Sigma-Aldrich or Tocris. During recordings, a 3 min stable baseline with a V_m fluctuation <1 mV was recorded before applying the drug via the perfusion system. The change in V_m was calculated as the difference between the maximum V_m deflection (positive or negative) after drug application and the baseline. To avoid a misclassification of the V_m change because of background V_m fluctuation, we set a threshold of ±0.5 mV so that a V_m change ≤0.5 mV during drug application is considered to be no response.

Slices were fixed after electrophysiological recordings with 4% paraformaldehyde in 100 mM phosphate buffered saline (PBS) for at least 24 h at 4°C. To recover the morphology of biocytin-filled neurons, slices were rinsed several times in 100 mM PBS and then treated with 1% H₂O₂ in PBS for about 20 min in order to reduce any endogenous peroxidase activity. Slices were rinsed repeatedly with PBS and then incubated in 1% avidin-biotinylated horseradish peroxidase (Vector ABC staining kit, Vector Lab. Inc., Burlingame, CA, United States) containing 0.1% Triton X-100 for 1 h at room temperature. The reaction was catalysed using 0.5 mg/ml 3,3-diaminobenzidine (DAB; Sigma-Aldrich, St. Louis, MO, United States) as a chromogen. Slices were then rinsed with 100 mM PBS, followed by slow dehydration with ethanol in increasing concentrations and finally in xylene for 2–4 h. After that, slices were embedded using Eukitt medium (Otto Kindler GmbH, Freiburg, Germany).

Computer-assisted morphological 3D reconstructions of neurons were made using the NEUROLUCIDA^® software (MicroBrightField, Williston, VT, United States) and Olympus BV61 microscopy at 1000× magnification (100× objective, 10× eyepiece). Neurons were selected for reconstruction based on the quality of biocytin labelling when background staining was minimal. The cell body, dendritic and axonal branches were reconstructed manually under constant visual inspection to detect thin and small collaterals. Cytoarchitectonic landmarks such as barrels in the primary somatosensory cortex and layer borders, pial surface and white matter were delineated during reconstructions at a low magnification (4× objective). The position of soma and layers were confirmed by superimposing the DIC images taken during the recording. Tissue shrinkage was corrected using correction factors of 1.1 in the x–y direction and 2.1 in the z direction (Marx et al., 2012).

For all data, the mean ± s.d. is given. Statistical comparisons among multiple groups were done using a Kruskal–Wallis test followed by a Dunn–Holland–Wolfe non-parametric multiple comparison test. Wilcoxon Mann–Whitney U test was performed to assess significant differences between individual groups. To assess the differences between two paired groups under different pharmacological conditions, Wilcoxon signed-rank test was performed. Correlation analysis was performed by calculating Pearson’s linear correlation coefficients. Statistical significance was set at p < 0.05, n indicates the number of neurons analysed. To prepare box plots for dataset with n > 10, the web application PlotsOfData was used¹ (Postma and Goedhart, 2019). In box plots, the interquartile range (IQR) is shown as a box, the range of values that are within 1.5*IQR are shown as whiskers and the median is represented by a horizontal line in the box.

Based on their electrophysiological characteristics, L4 neurons can be broadly classified as regular spiking (RS) excitatory neurons, FS and nFS inhibitory interneurons (Figure 1A). Three L4 neuron types can be easily differentiated by only three electrophysiological parameters, i.e., the maximum firing frequency, AP half-width and the AHP amplitude (Figures 1B,C). L4 RS neurons show a regular spiking firing pattern with a prominent spike frequency adaptation during a 1 s depolarising pulse (Figures 1B,C). In contrast, L4 FS interneurons show a high-frequency firing pattern without obvious spike frequency adaptation. L4 interneurons of the nFS type show heterogeneous firing patterns including adaptive spiking, irregular spiking, late spiking, etc.

In addition to their electrophysiological diversity, L4 neurons show highly distinct dendritic and in particular axonal morphologies (Figure 1D). L4 excitatory neurons fall into two main groups, spiny stellate neurons (SSNs) without an obvious apical dendrite and star pyramidal cells (SPCs) (Feldmeyer et al., 1999; Lubke et al., 2000; but see Staiger et al., 2004). Their axons originate from the soma or the initial part of one basal dendrite and project locally in layer 4 and to supra- and infragranular layers. Dendrites of L4 interneurons are aspiny or sparsely spiny and exhibit small to large multipolar, bipolar, or bitufted orientation patterns. Their axons project either locally in layer 4 and/or to supra- and/or infragranular layers in the vertical direction and/or to neighbouring columns in the horizontal direction. In previous studies, we have classified L4 FS interneurons as small basket cells (sBCs), basket cells (BCs), and translaminar cells (TLCs) (Koelbl et al., 2015) and L4 nFS interneurons as local-projecting (LP; non-Martinotti cell-like), supragranular-projecting (SP; Martinotti cell-like), neurogliaform (NGF), VIP+-like (VIP) and transcolumnar-projecting, interneurons (Emmenegger et al., 2018; Figure 1D).

We bath-applied 30 μM ACh while monitoring changes in the V_m of L4 neurons under current-clamp conditions. Of 44 L4 RS neurons, 42 showed a hyperpolarisation; only two showed no change (Figures 2A,D). On average, ACh-induced V_m change in L4 RS neurons was −2.8 ± 1.4 mV (n = 44) (Figure 3A). Of 33 L4 FS interneurons, 25 neurons showed a weak but significant depolarisation of the V_m, four a weak hyperpolarisation and another four no change (Figures 2B,E). On average, the ACh application resulted in a change in V_m in L4 FS interneurons was 0.9 ± 1.5 mV (n = 33) (Figure 3A). Of 31 L4 nFS interneurons, 29 neurons showed a strong depolarisation of the V_m and two a hyperpolarisation (Figures 2C,F). For the majority of L4 nFS interneurons (25 out of 31), the ACh-induced depolarisation was subthreshold. In a small fraction of L4 nFS interneurons (4 out of 31), ACh application evoked a suprathreshold depolarisation so that spontaneous AP firing was initiated. On average, the ACh-induced change in V_m in L4 nFS interneurons was 5.2 ± 5.8 mV (n = 31) (Figure 3A). Note that the ACh-induced V_m changes were fully reversible by bath application of control ACSF (Figures 2A–C). To examine whether these ACh-induced changes in V_m resulted from a direct effect on the neuronal excitability or were caused indirectly by altering the activity of local synaptic microcircuits, a cocktail of synaptic blockers comprising CNQX (10 μM), D-AP5 (50 μM), and gabazine (10 μM) was applied before ACh. There is no difference in the V_m change elicited by ACh in the absence and in the presence of synaptic blockers (Supplementary Figure 1). However, a clear decrease in background noise of V_m was observed in the presence of synaptic blockers (Supplementary Figure 1). A correlation analysis between the change in V_m and the age of the animal, V_rest and R_in demonstrated that there is no clear age-dependence of the ACh effect on V_m for any of the three L4 neuron types (Supplementary Figure 2A); a significant negative correlation was found between the V_m change and V_rest for L4 RS neurons (r = −0.53, p = 1.8 × 10^–4) and L4 nFS interneurons (r = −0.48, p = 5.9 × 10^–3) (Supplementary Figure 2B). Furthermore, for L4 nFS interneurons, a significant positive correlation was found between the ACh-induced change in V_m and R_in (r = 0.77, p = 1.6 × 10^–7; Supplementary Figure 2C), i.e., L4 nFS interneurons with higher R_in showed a larger depolarisation.

To evaluate the cell-type specificity of ACh-induced changes in V_m in more detail, we grouped the ACh response with respect to the L4 neuron subtype identified by the electrophysiological and morphological features described above. The two L4 RS excitatory neuron subtypes did not exhibit a significantly different ACh response (SSCs: −2.9 ± 1.4 mV, n = 25; SPNs: −2.8 ± 1.6 mV, n = 19; p = 0.85; Figure 3B). Similarly, no significant difference was found in the ACh-induced change in V_m among the three L4 FS interneuron subtypes (sBCs: 0.8 ± 1.0 mV, n = 17; BCs: 1.3 ± 2.3 mV, n = 8; TLCs: 0.5 ± 1.3 mV, n = 8; p = 0.42; Figure 3C). In contrast, in four subtypes of L4 nFS interneurons the ACh-induced V_m change in VIP interneurons was significantly larger than that in the other three subtypes (LPs: 3.0 ± 2.7 mV, n = 9; SPs: 3.8 ± 2.2 mV, n = 11; NGFs: 2.8 ± 0.5 mV, n = 4; VIPs: 14.3 ± 6.7 mV, n = 6; p = 1.5 × 10^–3) (Figure 3D). We did not record the ACh response of the transcolumnar-projecting L4 nFS interneuron due to their scarcity.

Acetylcholine not only modulates V_m but also induces changes in other intrinsic properties of L4 neurons. We have previously studied the effects of 100 μM ACh on the intrinsic properties of L4 excitatory neurons (Eggermann and Feldmeyer, 2009) and demonstrated that ACh reduces their excitability through a hyperpolarisation of V_m and a reduction in R_in. In this study, we focussed mainly on L4 interneurons. Low concentrations of ACh (30 μM) induced no significant change in the intrinsic properties of L4 FS interneurons except for the V_m (cf. Figures 2, 3). For example, no change was found for the AP half-width (Control: 0.26 ± 0.06 ms, n = 8; ACh: 0.26 ± 0.05 ms, n = 8; p = 0.31) and the AP amplitude (Control: 88.0 ± 12.6 mV, n = 8; ACh: 80.0 ± 8.9 mV, n = 8; p = 0.08) (Figure 4A and Supplementary Table 1). In contrast, apart from V_m changes (cf. Figures 2, 3) ACh also altered three other intrinsic electrophysiological properties of L4 nFS interneurons: the AP half-width was increased (Control: 0.44 ± 0.10 ms, n = 10; ACh: 0.48 ± 0.11 ms, n = 10; p = 0.04) and the AP amplitude was decreased (Control: 92.8 ± 10.9 mV, n = 10; ACh: 82.6 ± 11.4 mVs, n = 10; p = 2.0 × 10^–3) (Figure 4B). Furthermore, the rheobase current was significantly reduced by ACh (Control: 158.08 ± 75.5 pA, n = 10; ACh: 84.0 ± 113.1 pA, n = 10; p = 5.9 × 10^–3) (Supplementary Table 1). Thus, in contrast to L4 excitatory neurons, ACh enhanced the excitability of all recorded L4 nFS interneuron types.

One particular L4 nFS interneuron (Figure 4C) responded to ACh application with a V_m hyperpolarisation, in contrast to most L4 nFS interneurons. Furthermore, ACh changed its repetitive firing property (Figure 4C). Under control condition, this neuron showed a regular spiking firing pattern with a small spike-frequency adaptation, which in the presence of 30 μM ACh was transformed to an accelerating firing pattern together with a spike amplitude accommodation. In addition, AP firing persisted even after terminating current injection. Because firing pattern and V_m returned to normal after washout (Figure 4C), the marked alteration in the firing pattern cannot be the result of deteriorating recording conditions. Hence, already at low concentrations, ACh can dramatically change the electrophysiological behaviour of a subpopulation of L4 nFS interneurons.

To reveal the molecular mechanism of ACh-induced V_m changes in L4 neurons, slices were superfused with the general mAChR antagonist ATRO (200 nM) before application of ACh. A comparison of the ACh-induced change in V_m before and during co-application of ATRO showed that the mAChR antagonist completely blocked the response in all L4 RS excitatory neurons (Control: −3.8 ± 0.9 mV, n = 9; ATRO: −0.4 ± 0.5 mV, n = 9; p = 3.9 × 10^–3) (Figures 5A,D) and all L4 FS interneurons (Control: 1.5 ± 0.6 mV, n = 4; ATRO: 0.1 ± 0.3 mV, n = 4; p = 0.13) (Figures 5B,D). This suggests that the ACh-induced V_m changes in these L4 neuron types are exclusively mediated by mAChRs. In contrast, in L4 nFS interneurons, ATRO largely (but not completely) blocked the V_m change induced by 30 μM ACh (Control: 9.2 ± 7.2 mV, n = 11; ATRO: 3.7 ± 4.5 mV, n = 11; p = 9.8 × 10^–4) (Figures 5C,D). In the majority (8 out of 11) of L4 nFS interneurons, ATRO nearly completely blocked the ACh-induced V_m change while in the remainder (3 out of 11), a residual ACh-induced change in V_m still persisted after the co-application of ATRO. We tested whether this residual depolarisation was mediated by nAChRs (see below). To identify the mAChR type mediating the modulatory effect, TRO (1 μM), a specific M4 mAChR antagonist, and PIR (0.5 μM), a specific M1 mAChR antagonist, were applied before ACh. We found that TRO completely blocked the ACh-induced hyperpolarisation in L4 RS excitatory neurons (Supplementary Figure 3A) while PIR completely blocked the ACh-induced depolarisation in L4 FS interneurons (Supplementary Figure 3B). However, in L4 nFS interneurons, PIR blocked the ACh-induced depolarisation only partially (Supplementary Figure 3C).

The fact that a subset of L4 nFS interneurons showed a strong ACh-induced depolarisation (Figures 2C, 3A,D) that was not fully blocked by ATRO (Figures 5C,D) indicates that these interneurons may respond to ACh via both mAChRs and nAChRs. In those L4 nFS interneurons, in which ATRO blocked the ACh-induced depolarisation only incompletely, MEC (1 μM), a general nAChR antagonist, together with ATRO were applied before ACh. An example recording from a putative L4 VIP+ nFS interneuron (Figure 6A) is shown in Figure 6B. This neuron exhibits an irregular firing pattern, a bipolar dendritic structure, and a narrow translaminar axonal projection, all of which are characteristics typical of VIP+ interneurons (Porter et al., 1998; Pronneke et al., 2015; Emmenegger et al., 2018). ACh induced a strong depolarisation in this neuron and elicited spontaneous AP firing. Even in the presence of ATRO, the ACh-induced AP firing still exists. Only when ATRO and MEC were applied together, was the ACh-induced change in V_m blocked (Figure 6B). A nAChR-mediated depolarisation was observed not only in L4 VIP+ interneurons (n = 2) but also in one putative SST+ interneuron and one NGF cell (Figure 6C). However, only VIP+ interneurons showed such a strong nAChR-mediated depolarisation. In one recording from a L4 VIP+ interneuron, we found that the ATRO-resistant depolarisation was completely blocked by DHβE, a specific antagonist for α4β2-subunit containing nAChRs (Supplementary Figure 3C).

A detailed neuronal cell-type classification is necessary and critical for an in-depth understanding the modulatory effects of ACh. Traditionally, neurons are classified based on their morphological (dendritic and axonal) and electrophysiological (repetitive firing) properties (Petilla Interneuron Nomenclature Group et al., 2008; DeFelipe et al., 2013). With the development and sophistication of single-cell mRNA sequencing techniques, the molecular features of neurons add an additional layer of complexity to neuronal classification (Zeng and Sanes, 2017; Yuste et al., 2020). We have performed a series of studies to dissect the neuronal diversity in layer 4 of rat barrel cortex (Feldmeyer et al., 1999; Lubke et al., 2000; Koelbl et al., 2015; Emmenegger et al., 2018). In general, layer 4 comprises three neuronal cell classes showing distinct repetitive firing properties: regular spiking, fast spiking and non-fast spiking (adapting, irregular, late, etc.). Taking the morphological diversity also into account, L4 RS excitatory neurons have been classified into spiny stellate cells and star pyramidal neurons (Feldmeyer et al., 1999; Lubke et al., 2000) while L4 FS interneurons have been divided into cluster 3 (small basket cells), cluster 2 (basket cells), and translaminar-projecting FS interneurons (Koelbl et al., 2015). Most of these L4 FS interneurons are parvalbumin-positive (PV+) but calbindin-negative. L4 nFS interneurons, on the other hand, have been separated into five morpho-electrophysiological subtypes including transcolumnar-projecting interneurons with an adapting firing pattern, locally projecting with an adapting firing pattern (presumably non-Martinotti cells), supragranular-projecting with an adapting firing pattern with a Martinotti-cell appearance, VIP+ cell-like with an irregular firing pattern (VIP+-like) and neurogliaform cells (Emmenegger et al., 2018). The former three subtypes are somatostatin-positive while the latter two Prox1-positive. Our classification of L4 neurons is in line with several other groups focusing on the barrel cortex or primary visual cortex of rats and mice (Gibson et al., 1999; Porter et al., 2001; Beierlein et al., 2003; Ma et al., 2006; Scala et al., 2019).

Previously, it has been shown that the ACh concentration in the cerebrospinal fluid is in a low micromolar range, which fluctuates between 1 and 10 μM depending on the brain state (Himmelheber et al., 2000; Mattinson et al., 2011; Teles-Grilo Ruivo et al., 2017). Recently, accumulating evidence indicates that functional synaptic contacts are also established by cholinergic afferents in the neocortex. ACh is released into the synaptic cleft and its concentration can reach a very high concentration (>1 mM) (Turrini et al., 2001; Bennett et al., 2012; Hay et al., 2016; Obermayer et al., 2019). However, the exact extracellular concentration of ACh is still under investigation due to the species differences and difficulties arising from the rapid breakdown by acetylcholinesterase. ACh modulates the intrinsic neuronal properties through both mAChRs and nAChRs. mAChRs are G-protein coupled receptors the activation of which initiates a signalling cascade inside the neuron. In contrast, nAChRs form ligand-gated cation channels (Unwin, 2003; Dani, 2015). These two types of receptors work at different concentrations of ACh. mAChRs already show a high affinity to ACh at low concentrations (in the range of 1–100 μM) while nAChRs require a high concentration of ACh for maximal activity (in the mM range). Previous studies have used puff-application of 1–10 mM ACh to study the nicotinic effects of ACh on excitatory and inhibitory neurons in several cortical areas (Xiang et al., 1998; Gulledge and Stuart, 2005; Gulledge et al., 2007; Poorthuis et al., 2013a,b). Puff-application of agonists has a high spatiotemporal resolution and is therefore a suitable strategy to simulate phasic ACh release at cholinergic presynaptic terminals; it is also required to minimise the effects of nAChR desensitisation. In addition, at such high concentrations, ACh will not be hydrolysed (and hence inactivated) immediately so that it may persist at low concentrations in the perisynaptic space. Bath application of cholinergic agonists such as ACh and carbachol is a good approach to simulate the latter condition because the agonist concentration will be maintained at a constant level to allow the measurement of neuronal properties at equilibrium. Carbachol shares both the muscarinic and nicotinic actions of ACh but shows a slower binding, dissociation and desensitisation kinetics, in particular at nAChRs. In addition, carbachol is not degraded by acetylcholinesterase. Carbachol is not the natural agonist and the relative affinity of carbachol for mAChRs and nAChRs is likely to be different from that of ACh so that we may observe an activation of nAChRs with low concentrations of carbachol but not with ACh. In order to simulate the in situ action of cholinergic agonists, the natural agonist ACh has been used instead of carbachol in this study. It is likely that in the extracellular space, a neurotransmitter/neuromodulator is only present at a μM concentration because of its rapid diffusion from the synaptic release site (Borroto-Escuela et al., 2015). In addition, application of high concentrations of ACh will mask the effects mediated by mAChRs so that the application of low concentrations (∼μM) of ACh is a prerequisite to uncover their functional effects.

We have shown previously that 100 μM ACh persistently hyperpolarises L4 excitatory neurons and reduces their intrinsic excitability (Eggermann and Feldmeyer, 2009), an ACh effect markedly different from that observed in most of pyramidal cells except for L6A corticocortical neurons (Yang et al., 2020). In L2/3, L5 and corticothalamic L6A pyramidal cells, ACh induces a persistent depolarisation and therefore enhances the excitability.

The ACh-induced hyperpolarisation in L4 excitatory neurons is mediated exclusively by M4 mAChRs, a finding that is supported by another study using optogenetic activation of synaptic ACh release (Dasgupta et al., 2018). Here, using a lower concentration of ACh (30 μM), similar results were obtained. Note that, as discussed above, 30 μM is an ACh concentration closer to the physiological range than 100 μM. There was no clear difference between the ACh-induced hyperpolarisation in both L4 RS neuron subtypes suggesting that SSNs and SPCs express the same mAChR subtype at a similar density.

The effects of ACh on FS interneurons have been a long-standing matter of debate. Conflicting results have been published by different research groups. Puff-application of 5 mM ACh induced a transient hyperpolarisation that was mediated by mAChRs in rat neocortical L5 FS interneurons (Xiang et al., 1998). Recently, in the mouse visual cortex it has been shown that optogenetically stimulated ACh release led to an indirect inhibition in L2/3 FS interneurons via “facilitation” of the cholinergic responses in L2/3 somatostatin-positive interneurons (Chen et al., 2015). On the other hand it has been postulated that ACh does not affect the V_m of FS interneurons. In the rat frontal cortex, bath-application of carbachol (10 μM) had no effect on L2/3 FS interneurons (Kawaguchi, 1997). In a follow-up study the same group used focal application of ACh (100 μM or 5 mM for comparison with the study by Xiang et al., 1998) onto FS interneurons in rat visual and prefrontal cortex; the authors concluded that the focal application itself (i.e., a mechanical artefact but not the transient ACh exposure) caused the hyperpolarising response (Gulledge et al., 2007). Except for Xiang et al. (1998), most investigators have been unable to show a direct effect of ACh on FS interneurons (Muñoz and Rudy, 2014); however, they reported a presynaptic effect of ACh. In contrast to previous studies, we found a persistent ACh-induced V_m depolarisation in L4 FS interneurons, an effect that appeared in all three subtypes of L4 FS interneurons. To the best of our knowledge, this is the first time that a direct depolarising effect of ACh on FS interneurons has been demonstrated conclusively. In addition, we were able to show that this effect is mediated by M1 mAChRs. The ACh-induced depolarisation persisted in the presence of GABA and glutamate receptor antagonists so that indirect effects of ACh can be excluded. We were unable to investigate the effect of ACh on another subtype of FS interneurons, the chandelier or axo-axonic cells, which are very scarce if not absent in cortical layer 4 (Wang et al., 2019).

Layer 4 nFS interneurons are a heterogenous population with diverse firing patterns, dendritic/axonal morphologies and molecular expression patterns. To elucidate modulatory effects of ACh on L4 nFS interneurons, a clear separation into identifiable subtypes is required. The local- and supragranular-projecting subtypes of L4 nFS interneurons display an adapting firing pattern similar to that of somatostatin-positive interneurons. Indeed, immunocytochemistry revealed that both subtypes of L4 nFS interneurons are somatostatin-positive (Emmenegger et al., 2018). Here, we found that L4 SST+-like interneurons including both local projecting (non-Martinotti-like) and supragranular projecting (Martinotti-like) cells that responded to ACh with a strong depolarisation that is predominantly mediated by mAChRs. Consistent with our findings, in the mouse barrel cortex it has been shown that L4 SST+ interneurons were depolarised and fired spikes in response to bath-applied muscarine (3 μM) (Xu et al., 2013). In one L4 SST+-like interneuron, ACh induced a depolarisation mediated by both mAChRs and nAChRs suggesting that even at μM ACh concentrations activation of nAChRs may be also possible. In a previous study, it has been shown that ACh directly excites SST+ neurons via both mAChRs and nAChRs in layer 2/3 of mouse visual cortex (Chen et al., 2015). However, a very high concentration of ACh (10 mM) was puff-applied in that study, which is very different from the bath-application of ACh (30 μM) described here.

Excitation of VIP+ interneurons by nAChRs has been observed in several cortical areas of both rat and mouse (Porter et al., 1999; Férézou et al., 2007; Koukouli et al., 2017; Askew et al., 2019; Prönneke et al., 2020). In the rat motor cortex, local pressure application of ACh (100 μM) or the selective nAChR agonist DMPP (100–500 mM) depolarised VIP+ interneurons located in layer 3–5 and induced a discharge of action potentials (Porter et al., 1999). Pharmacological experiments suggested that the ACh effect was mediated by non-α7 nicotinic receptors containing α4β2 and α5 subunits (Koukouli et al., 2017). In another study from the same group, it has been shown that bath-application of nicotine (1 μM) also resulted in a strong depolarisation leading to a sustained action potential discharge in VIP+ interneurons (Férézou et al., 2007). Similarly, bath-application of nicotine (1 μM) in the mouse auditory cortex caused sustained AP discharge in VIP+ interneurons across the layers (Askew et al., 2019). In the mouse barrel cortex, bath-application of ACh (40 μM) efficiently depolarised L2/3 VIP+ interneurons and changed the firing pattern from bursting to tonic spiking in a subpopulation (Prönneke et al., 2020); however, the authors concluded that cholinergic modulation was mediated exclusively by nAChRs. All of the aforementioned studies emphasised the critical role of nAChRs in the cholinergic modulation of VIP+ interneurons but overlooked any direct involvement of mAChRs. However, our recordings from L4 VIP+ interneurons demonstrated that both AChR subtypes participate in the cholinergic modulation in a cooperative way because ATRO partially blocked the depolarisation or shortened the duration of repetitive AP firing induced by ACh; ATRO together with the nAChR antagonist MEC completely blocked the ACh effect. Similarly, in rat frontal cortex, VIP+ cells showed a sustained V_m depolarisation in response to bath-applied muscarine (3 μM) in the presence of TTX (Kawaguchi, 1997) indicating that mAChRs are expressed in these interneurons.

For NGF cells, the focus of attention is mostly cortical layer 1 where NGF cells are abundant (Christophe et al., 2002; Gulledge et al., 2007; Arroyo et al., 2012; Brombas et al., 2014). Puff-application of nicotinic agonists such as ACh, DMPP, choline onto L1 NGF cells or optogenetic stimulation of cholinergic fibres in layer 1 has revealed nicotinic excitation of NGF cells. Similar results have been shown for L2/3 5-HT_3aR+ NGF cells of mouse barrel cortex (Lee et al., 2010). Here, we found that in L4 NGF cells of the barrel cortex, low concentrations of ACh led to a mAChR-mediated sustained depolarisation. In one L4 NGF cell, a participation of nAChRs in this depolarisation was also found.

In addition to SST+, VIP+, and NGF cells, layer 4 comprises other nFS subtypes (Tasic et al., 2018). In one L4 nFS interneuron, we were able to show that, in contrast to most other L4 nFS interneurons, ACh application resulted in a hyperpolarisation of this neuron and dramatically changed its repetitive firing pattern during the suprathreshold current injection. The cholinergic response of this neuron together with its firing pattern and morphology, is reminiscent of a subset of CCK+ neurons in L2/3 of rat frontal cortex which exhibited a prominent hyperpolarisation in response to muscarine (3 μM) and had large somata and an extensive axonal arbour (Kawaguchi, 1997). Therefore, the nFS interneuron showing an ACh-induced hyperpolarisation described here could be a L4 CCK+ neuron. Similarly, some hippocampal CA1 CCK+ interneurons showed also an ACh-induced hyperpolarisation mediated by mAChRs (McQuiston and Madison, 1999b; Cea-del Rio et al., 2011). Furthermore, the dramatic change in firing pattern induced by ACh has also been demonstrated in hippocampal CA1 CCK+ interneurons (McQuiston and Madison, 1999a; Lawrence et al., 2006; Cea-del Rio et al., 2010, 2011). In these neurons, through the activation of M1 and M3 mAChRs the AHP was superimposed by an afterdepolarisation which is often sufficiently strong to evoke APs in the absence of further stimulation (McQuiston and Madison, 1999a; Cea-del Rio et al., 2011).

In the neocortex, ACh is continuously released into the extracellular space and its level changes dramatically during the animal’s diurnal cycle and different behavioural states (Teles-Grilo Ruivo et al., 2017). Furthermore, there is increasing evidence for changes in ACh receptor expression levels going hand in hand with the diurnal change in ACh drive (Hut and Van der Zee, 2011). Most of the intracortical ACh is not released at synaptic contacts but rather diffusely into the extracellular space through an extrasynaptic volume transmission (Fuxe and Borroto-Escuela, 2016). Under this condition, cholinergic modulation is spatiotemporally slower but broader, thereby tuning neuronal network function. Modulation of neurons and their synaptic interactions through mAChRs may induce neuronal oscillations and therefore change the information processing mode in L4 neuronal microcircuits. Specifically, cholinergic activation of L4 FS, PV+, and nFS, SST+ interneurons by low concentrations of ACh may lead to the generation of persistent activity such as the gamma rhythm, which has been demonstrated to enhance the cortical circuit performance (Bartos et al., 2007; Sohal et al., 2009; Veit et al., 2017). Indeed, it has been demonstrated that bath-application of carbachol (10 μM) to activate mAChRs and kainate (300 nM) to increase the tonic excitatory drive elicited persistent gamma frequency network oscillations in cortical layer 4 of mouse barrel cortex (Buhl et al., 1998). In addition, differential modulation of L4 excitatory and inhibitory neurons, i.e., the persistent hyperpolarisation of L4 excitatory neurons and depolarisation of most L4 inhibitory neurons, will change the excitation-inhibition balance towards inhibition and reduce the responsiveness of the L4 recurrent excitatory microcircuit. Therefore, our results support the hypothesis that ACh has a filtering action in the major recipient layer of the neocortex (Eggermann and Feldmeyer, 2009). Because neocortical layer 4 is uniquely positioned to gate thalamocortical input to the neocortex, cholinergic modulation of L4 neuronal microcircuits will affect the whole barrel cortex together with the related cortical areas (e.g., M1 and S2) and finally the animal behaviour (Eggermann et al., 2014; Meir et al., 2018). In addition, our finding that mAChRs ubiquitously but differentially modulate the activity of L4 excitatory and inhibitory neurons might open the door to more specific therapeutic strategies to treat cognitive dysfunction or psychiatric disorders linked to degeneration of the cholinergic system in diseases such as Alzheimer’s disease and schizophrenia (Marin, 2012; Hampel et al., 2018).