Nicotinic excitation in neocortex is mediated by low-affinity α7 receptors and by high-affinity α4β2 receptors. There is evidence that α7 receptors are synaptic, but it is unclear whether high-affinity receptors are activated by volume transmission or synaptic transmission. To address this issue, we characterized responses of excitatory layer 6 (L6) neurons to optogenetic release of acetylcholine (ACh) in cortical slices. L6 responses consisted in a slowly decaying α4β2 current and were devoid of α7 component. Evidence that these responses were mediated by synapses was 4-fold. 1) Channelrhodopsin-positive cholinergic varicosities made close appositions onto responsive neurons. 2) Inhibition of ACh degradation failed to alter onset kinetics and amplitude of currents. 3) Quasi-saturation of α4β2 receptors occurred upon ACh release. 4) Response kinetics were unchanged in low release probability conditions. Train stimulations increased amplitude and decay time of responses and these effects appeared to involve recruitment of extrasynaptic receptors. Finally, we found that the α5 subunit, known to be associated with α4β2 in L6, regulates short-term plasticity at L6 synapses. Our results are consistent with previous anatomical observations of widespread cholinergic synapses and suggest that a significant proportion of these small synapses operate via high-affinity nicotinic receptors.
At cholinergic synapses involving nicotinic receptors, as for example those of the neuromuscular junction, the autonomic ganglia or the motoneuron-to-Renshaw cell connection (Fatt and Katz 1951; Zhang et al. 1996; Lamotte d'Incamps and Ascher 2008), acetylcholine (ACh) activates mostly subsynaptic receptors situated in the postsynaptic density, and only secondarily extrasynaptic receptors. In the brain, however, it has been suggested that in many cases the effects of ACh do not involve true synapses and are mediated by “volume transmission,” in which the neurotransmitter is released far from the receptors (Descarries et al. 1997; Lendvai and Vizi 2008; Sarter et al. 2009; Arroyo et al. 2014).
Cholinergic fibers from the basal forebrain (BF) widely innervate the telencephalon and exhibit varicosities that contain the ACh-synthesizing enzyme (choline acetyltransferase [ChAT]) and its vesicular transporter (Arvidsson et al. 1997; Mechawar et al. 2000). The reported proportions of synapse-forming varicosities greatly vary in the literature (Umbriaco et al. 1994; Mrzljak et al. 1995; Smiley et al. 1997; Turrini et al. 2001). However, a recent study has revealed a high occurrence of cholinergic synapses with a postsynaptic density containing the trans-synaptic organizer neuroligin 2 (Takacs et al. 2013). The small size of these synapses with relatively modest postsynaptic density thickenings have presumably led to underestimating their frequency in previous reports.
Nicotinic receptor subunits assemble into ligand-gated channels of two subtypes. Homomers formed with the α7 subunit exhibit low affinity for ACh and rapid kinetics, whereas heteromeric “non-α7” receptors composed of α and β subunits are endowed with higher affinity and slower kinetics (Dani and Bertrand 2007; Albuquerque et al. 2009). The existence of fast nicotinic transmission involving α7 receptors in the neocortex and hippocampus is established by numerous reports (Roerig et al. 1997; Alkondon et al. 1998; Frazier et al. 1998; Chu et al. 2000; Bell et al. 2011; Bennett et al. 2012). The cortical expression of somatodendritic α7 receptors is primarily restricted to subtypes of GABAergic interneurons that represent a minority of cortical neurons. In contrast, non-α7 receptors are expressed both in interneurons and in large subpopulations of excitatory neurons (Porter et al. 1999; Christophe et al. 2002; Gulledge et al. 2007; Kassam et al. 2008; Bailey et al. 2010; Hay et al. 2014; Hedrick and Waters 2015). A recent study of nicotinic currents evoked by optogenetic stimulation of BF fibers in layer I (L1) interneurons indicated that α7 receptors mediate synaptic transmission, but proposed that activation of non-α7 receptors is triggered by volume transmission (Bennett et al. 2012). This implies that conventional nicotinic synapses are restricted to GABAergic interneurons, at odds with the widespread occurrence of cholinergic synapses reported in the neocortex (Takacs et al. 2013).
L6 excitatory neurons express α4β2 receptors containing the auxiliary α5 subunit, but lack somatodendritic α7 receptors (Salas et al. 2003; Kassam et al. 2008; Bailey et al. 2010; Hay et al. 2014). These neurons thus provide an excellent model system to probe for the existence of non-α7 nicotinic synapses in the brain. Here, we selectively expressed channelrhodopsin in BF neurons and analyzed light-evoked nicotinic currents in L6 excitatory neurons from mouse neocortical slices. Using anatomical reconstruction and functional analyses of responses to low- and high-frequency photostimulation, we provide evidence for non-α7 nicotinic synapses in L6 and identify distinct modes for the recruitment of synaptic and extrasynaptic non-α7 receptors.
Materials and Methods
Animals, Virus, and Surgery
All experiments were carried out in accordance with the guidelines published in the European Communities Council Directive of 24 November 1986 (86/609/EEC). Our animal Protocol has been approved by the local ethics committee (Ce5/2012/062). Transgenic ChAT-Cre male mice were from the Mutant Mouse Regional Resource Center (line GM60, Tg(Chat-cre)60Gsat) (Gong et al. 2007). The R26RYFP line was from Jackson laboratories (line 006 148). For viral injections, 21- to 25-day-old mice were deeply anesthetized using a mix of ketamine and xylazine (100 and 10 mg/kg body weight, respectively) delivered intraperitoneally. Mice were restrained in a stereotaxic frame (David Kopf instruments, Phymep, France) and the skull was exposed under aseptic conditions. A small burr hole was drilled at coordinates AP = 0.1 mm and ML = 3 mm from bregma. One to two microliter of viral solution was injected with a canula (36G70 [canula] and 26G50 [guide] from Phymep) in the right nucleus basalis at a depth of 4.55 mm with a 20° angle from the medial plane. The speed of injection was 100–200 nl/min. Two minutes after the end of the injection, the canula was slowly pulled out and the skull skin was sutured. Animals were then housed in a BL-2 animal facility for at least 3 weeks with free access to food and drink. The construct AAV2/1-EF1a-DIO-hChR2(H134R)-EYFP-WPRE-HGHpA (titer: 3.1011 gc/ml) was produced from Addgene plasmid #20298 at the vector core facility of Nantes University (UMR 1089 IRT1). Lentiviral construct LV-hsyn-ChR2(H134R)-YFP-WPRE (titer: 50 ng p24/ml) was produced from Addgene plasmid #20945 by the gene transfer network of Ecole des Neurosciences de Paris. Genotyping was done by PCR with the following primers: ChAT-Cre, forward, GATCGCTGCCAGGATATACG, reverse, CATCGCCATCTTCCAGCAG (573 bp); R26RYFP, forward, AAAGTCGCTCTGAGTTGTTAT, reverse, GGAGCGGGAGAAATGGATATG, mutant reverse, AAGACCGCGAAGAGTTTGTC (wild type: 600 bp, mutant: 320 bp); α5 nAChR, wild-type forward, GTGAAAGAGAACGACGTCCGC, wild-type reverse, GCCTCAGCCCCTGAATGGTAG, mutant forward CTTTTTGTCAAGACCGACCTGTCCG, mutant reverse, CTCGATGCGATGTTTCGCTTGGTG (wild type: 380 bp, mutant: 290 bp).
Four to six weeks after injection, ChAT-Cre mice were deeply anesthetized. Heparin (80 U) was directly injected in the left heart ventricle to prevent blood coagulation, then the mice were perfused transcardially with ice-cold sucrose artificial cerebrospinal fluid (ACSF) containing (in mM): sucrose, 30; glucose, 2.5; NaCl, 126; KCl, 2.5; NaHCO3, 26; NaH2PO4, 1.25; MgCl2, 3; and kynurenic acid, 3. Brains were quickly removed and 300 µm-thick coronal slices were cut in sucrose ACSF using a vibratome (VT1000S; Leica). Slices were transferred to a holding chamber filled with standard ACSF bubbled with 95% O2 and 5% CO2. The composition of the standard ACSF was (in mM): NaCl, 126; KCl, 2.5; NaHCO3, 26; NaH2PO4, 1.25; MgCl2, 1; CaCl2, 2; sucrose, 10; glucose, 10. Slices were allowed to recover 1 h at 37°C then they were placed at room temperature. Drugs and chemicals were obtained from Sigma (Saint Louis, MO, USA), except 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), 2-amino-5-phosphonovalerate (APV), and gabazine obtained from Tocris Cookson (Bristol, UK).
Individual slices were transferred to a recording chamber and continuously superfused at 2 mL/min with oxygenated ACSF at room temperature. Patch pipettes (3–5 MΩ) were pulled from borosilicate glass (Harvard Apparatus LTD, Kent, UK) on a micropipette puller (Model PP-83, Narishige, Tokyo, Japan) and were filled with an intracellular solution containing (in mM): K-gluconate, 144; CaCl2, 0.1; MgCl2, 2; HEPES, 10; EGTA, 0.5 and 0.3 mg/mL biocytine, pH = 7.3, 295 mOsm. The I–V relationships of light-induced currents were obtained using an internal solution containing (in mM): Cs-gluconate, 130; HEPES, 10; EGTA, 0.2; and MgCl2, 3; pH 7.2, 295 mOsm. Cortical neurons were visualized under infrared videomicroscopy with Nomarski optics. Whole-cell patch-clamp recordings were obtained from spatially and morphologically identified L1 and L6 neurons and were performed using a patch-clamp amplifier (Multiclamp 700B, Molecular Devices, Sunnyvale, CA, USA) connected to a Digidata 1440A interface board (Molecular Devices). Signals were filtered at 1–5 kHz, digitized at 20 kHz and collected using the data-acquisition software pClamp 10.2 (Molecular Devices). Signals were analyzed offline with Clampfit 10.2 software (Molecular Devices. Voltage values were corrected for liquid junction potential (+11 and +13 mV for the K-gluconate and Cs-gluconate intracellular solutions, respectively). Unless stated otherwise, voltage-clamp recordings were performed at −71 mV.
Light-evoked currents obtained in L1 and L6 neuron were fitted using a Levenberg–Marquardt algorithm with a linear combination of 3 (L1) and 2 (L6) decaying exponential functions of the form A × exp(−t/τ), where A is the amplitude and τ is the time constant. Paired-pulse ratios (PPRs) were calculated from peak current amplitude. Data for I–V curves were obtained by varying the membrane potential between −133 and +47 mV in steps of 10 or 20 mV. Currents were normalized with respect to the current at −73 mV. Current measurements at each potential were done in duplicate and averaged. The I–V relations were fitted by using a simplified Woodhull model (Woodhull 1973), assuming that the inward rectification of the nicotinic currents was entirely due to blockade by internal Mg2+ despite the contribution of other factors (Ifune and Steinbach 1990). Normalized current values were fitted using Igor Pro (Wavemetrics) according to the formula:
Transduced cholinergic fibers were first visualized with a CoolLed excitation system (Andover, UK) and an YFP filter set (YFP-2427B-000, Semrock, Rochester, NY, USA) using a 535-nm LED mounted on the epifluorescene port of the microscope. This allowed sufficient illumination for visualizing YFP fluorescence while minimizing direct ChR2 activation. Photoactivation of ChR2 was performed through a GFP filter set (Olympus, France) using a 465-nm blue LED driven directly by the data-acquisition software pClamp 10.2 (Molecular Devices) through the Digidata interface board. Blue light pulses (10 ms unless stated otherwise) were delivered though a ×40 immersion objective at 40–50 s intervals. In this configuration, the stimulated area on the slice was a disk of roughly 1 mm wide and the mean power density was 3 mW/mm2.
Four- to six-week-old mice were perfused transcardially using a 0.1 M phosphate buffer (PB) pH7.4 solution containing 4% paraformaldehyde followed by 2 h of postfixation at room temperature in the same fixative. Brains were cryoprotected using 10% sucrose/0.1 M PB solution, sliced to 40 µm thickness using a freezing microtome, and kept in phosphate-buffered saline (PBS) at 4°C for up to 3 weeks until used. Free-floating sections were blocked for 2 h at room temperature in a PBS/0.25% Triton X-100/0.2% gelatine solution (PBS-GT) supplemented with 10% normal donkey serum before being incubated overnight at 4°C with primary antibodies diluted in PBS-GT. Slices were washed with PBS-GT before being incubated for 2 h at room temperature with secondary antibodies diluted in PBS-GT. After extensive washing in PBS, slices were mounted on gelatin-coated slides in Prolong gold (Invitrogen). Antibodies were used at the following dilutions: goat anti-ChAT (1:300; Millipore), chicken anti-GFP (1:1000, Aves Labs), donkey anti-chicken alexafluor488 (1:400; Jackson Immunoresearch), donkey anti-rabbit alexafluor555 (1:500; Invitrogen).
For immunohistochemistry after electrophysiological recording, slices were fixed overnight at 4°C in 4% paraformaldehyde/ 0.1 M PB pH7.4 and then rinsed in PBS. Residual aldehyde was inactivated by incubation with 50 mM NH4Cl/PBS solution for 10 min at room temperature followed by several washes in PBS. Slices were blocked for 2 h in PBS-GT and then incubated with anti-GFP (same dilution as above) for 2 days at 4°C. After several washes in PBS-GT, slices were incubated overnight at 4°C with goat anti-chicken alexafluor 488 (1:500, Invitrogen) and Alexafluor 555-conjugated streptavidin (1:500, Invitrogen) diluted in PBS-GT. After extensive washing in PBS, slices were mounted on gelatin-coated slides in Fluoromount-G (Southernbiotech, Birmingham, AL, USA).
3D Reconstruction and Rendering
To examine the presence of appositions between ChR2-YFP+ fibers and biocytin-filled-responsive cells, fluorescence images were first obtained using a confocal laser scanning microscope (SP5, Leica) equipped with a ×40 oil immersion objective (NA 1.40) using 488- and 561-nm Ar lasers combined with a numerical zoom ×1.7. Dendritic branches were systematically inspected within a radius of 200 µm around the soma. Coordinates of varicosities forming apposition with dendritic branches were recorded and Euclidean distance from the soma was calculated. Images of appositions were then acquired and processed according to Heck et al. (2012). Briefly, z-stacks were obtained sequentially using a confocal laser scanning microscope (SP5, Leica) equipped a ×63 oil immersion objective (NA 1.40) using 488- and 561-nm Ar lasers. The pixel size was set to 60 nm. The pinhole aperture set to 1 Airy Unit and z-steps of 200 nm were used. Laser intensity and photomultiplier tube gain was set so that the image occupied the full dynamic range of the detector. The experimental point spread function (PSF) for deconvolution of each fluorescence image was obtained using 170-nm-diameter fluorescent latex beads (PS-Speck, Life Technologies, France). Bead images were obtained using strictly the same settings and mounting conditions as for imaging of biological samples. At least 8 imaged beads were registered and averaged in order to increase the signal-to-noise ratio (SNR). The PSF was then determined using Huygens software 3.6 (Scientific Volume Imaging, The Netherlands) and used for deconvolution using Maximum Likelihood Estimation algorithm with the same software. The deconvolution settings were used so background intensity was averaged from the voxels with lowest intensity and SNR was estimated to a value of 20. 3D rendering was performed using the 3D viewer plugin of the FIJI software (Schindelin et al. 2012).
Nonstationary Noise Analysis
To determine the properties of single channels involved in nAChR-mediated currents, nonstationary noise analysis was performed on fully discriminated monosynaptic events as described (Sigworth 1980; Rigo et al. 2003). EPSCs were recorded at −81 mV using K-gluconate intracellular solution. Only EPSCs occurring within 8 ms after the photostimulus were considered. Between 30 and 50 individual EPSCs were selected for the nonstationary noise analysis of each neuron. EPSCs were aligned on the photostimulus and analysis was performed on the decaying phase of EPSCs. The variance over time of the decay phase was determined using the Clampfit 10.2 software (Molecular Devices). Baseline noise variance was subtracted. A time-to-time plot of the variance at time t (σ2(t)) as a function of the averaged current amplitude at time t (IAmp(t)) was fitted by a parabolic curve (2 magnitude polynomial function) of the form:
Results are given as mean ± standard error of mean. Between-group comparisons were performed using paired Mann–Whitney nonparametric test. A P-value below 0.05 was considered statistically significant. For the analysis of the recovery from paired-pulse depression, a 2-way ANOVA test was performed. Unless stated otherwise, all traces in the figures are single-trial responses.
Selective Photoactivation of Cholinergic Neurons
The mouse line GM60 harbors a BAC-based transgene that expresses the Cre recombinase under control of the ChAT (the ACh-synthesizing enzyme) promoter (Gong et al. 2007). We first examined the ability of this transgene to drive Cre-dependent recombination specifically in cholinergic neurons when combined with the R26RYFP reporter allele (Srinivas et al. 2001) (Supplementary Fig. 1A). In the BF and striatum, YFP immunoreactivity was observed in 89% and 92% of ChAT+ neurons, respectively (n = 651 and 356) and no YFP signal was detected in ChAT− neurons (Supplementary Fig. 1A). The overlap between YFP and ChAT expression was smaller in medial septal nucleus (42% of ChAT+ neurons, n = 313) and even weaker in neocortex (26% of ChAT+ neurons, n = 121, Supplementary Fig. 1A), where a small proportion of YFP+ neurons was ChAT− (30% of YFP+ neurons, n = 46, data not shown).
In order to express ChR2 selectively in corticopetal cholinergic fibers, we performed unilateral stereotaxic injection of a Cre-dependent adeno-associated virus (AAV, (Sohal et al. 2009)) encoding a ChR2-YFP fusion protein in the BF of ChAT-Cre mice. Preliminary experiments where injection was performed at coordinates AP: −0.22, ML: 1.12, DV: 4.80 primarily yielded YFP+ fibers in the amygdala and ventral cortical regions (not shown). In contrast, injection at a more dorso-caudal site (AP: 0.1, ML: 1.4, DV: 4.27) yielded YFP+ fibers in dorso-lateral cortical regions throughout all cortical layers. Using this latter paradigm, 90% of YFP + BF neurons (n = 131) were ChAT+, and YFP+ fibers typically occupied a dorso-lateral quadrant of the neocortex with a rostro-caudal extension of ∼1.5 mm throughout all cortical layers (Supplementary Fig. 1B). We found that 92% of YFP+ axonal varicosities in the neocortex were ChAT+ (n = 487; Supplementary Fig. 1B).
We probed the functional expression of ChR2 using patch-clamp recordings of YFP+ BF neurons (n = 5) in acute brain slices. Whole field illumination through a ×40 objective with a 5-s light pulse at 10–50% of maximal light power (see Materials and Methods) gave rise to a large inward peak current followed by a smaller steady state current (Supplementary Fig. 1C). Trains of 1-ms light pulses (2 Hz for 10 s) reliably induced 2 Hz firing of the recorded cells (Supplementary Fig. 1C). These results are consistent with desensitization properties of ChR2 currents (Mattis et al. 2011) and the ability of ChR2 to drive action potential firing in cholinergic neurons (Kalmbach et al. 2012; Hedrick and Waters 2015). These results indicate that the present strategy enables reliable and selective photoactivation of cholinergic neurons and can be used to examine the effects of endogenous ACh release on cholinoceptive neurons.
Light-Evoked Responses of L6 Neurons Only Involve Heteromeric Nicotinic Receptors
We first examined whether ACh released by photoactivation of ChR2+ fibers in neocortical slices elicits in L6 excitatory neurons responses similar to those reported for L1 GABAergic neurons (Arroyo et al. 2012; Bennett et al. 2012). Recordings in L6 focused on excitatory pyramidal neurons, which characteristically exhibited regular spiking properties and spiny dendrites (n = 21 neurons processed for biocytin histochemistry, Fig. 1A) (Peters and Jones 1984; Connors and Gutnick 1990). Recordings were also performed on L1 neurons for comparison. L1 neurons exhibited short duration action potentials (range: 0.48–1.4 ms, n = 32) and smooth or sparsely spiny dendrites (n = 18 neurons processed for biocytin histochemistry, Fig. 1C), as previously described (Zhou and Hablitz 1996). Single-light pulses (duration: 1–50 ms) evoked inward currents in both L1 and L6 neurons (Fig. 1B–D). Unless otherwise stated, photostimuli were applied with an interval of 50 s, which yielded stable response amplitudes, avoiding the decrease of the response observed with shorter intervals (20 and 40 s, n = 7 and 6 neurons, respectively; see also Fig. 7). Light-evoked currents persisted in the presence of glutamatergic and GABAergic antagonists (CNQX, 10 µM; APV, 50 µM; gabazine, 10 µM) without notable change in amplitude and kinetics (n = 5 L1 and 5 L6 neurons, not shown), consistent with their cholinergic nature. All subsequent experiments were carried out in the continuous presence of glutamatergic and GABAergic antagonists.
Responses of L6 neurons exhibited a slow onset (τON = 32.9 ± 3.0 ms, n = 46, see Materials and Methods for definition of time constant) and a monophasic decay that could be fitted with a single exponential (τOFF = 289.6 ± 17.9 ms, Fig. 1B). In contrast, light-evoked currents in L1 neurons exhibited a fast onset (τON = 5.3 ± 0.6 ms, n = 30) and a biphasic shape. It is noteworthy, however, that reported onset kinetics of the slower component of L1 responses measured upon pharmacological blockade of the fast component (Arroyo et al. 2012; Bennett et al. 2012) is close to onset kinetics of L6 responses (see above). The relative amplitude of the two components of L1 responses we recorded was variable between neurons, consistent with earlier reports (Arroyo et al. 2012; Bennett et al. 2012). Indeed, among 57 L1 neurons recorded, 17 lacked a detectable slower component and in 11 other neurons the faster component was only noticeable from the abrupt onset of the response. In a subset of 29 neurons, the two components were clearly distinguishable and each could be fitted with a single exponential decay (τOFF,FAST = 6.5 ± 0.6 ms, τOFF,SLOW = 296.8 ± 29.6 ms, Fig. 1B). The mean amplitude and charge transfer of the slower component in L1 neurons (13.8 ± 2.1 pA and 5140 ± 1203 pC, n = 29 neurons) were similar to those of the responses in L6 (19.04 ± 2.0 pA and 5210 ± 765 pC, n = 25 neurons).
We next characterized the nicotinic receptors (nAChRs) underlying L1 and L6 currents. Bath application of atropine, a muscarinic antagonist (10 µM), had no significant effect on the amplitude of the slow and fast components (n = 8 L1, 110 ± 5 and 86 ± 8% of control, respectively P > 0.1; n = 8 L6 neurons, 110 ± 11% of control P > 0.1), ruling out the involvement of muscarinic receptors. Light-induced currents in L6 neurons (n = 6) were blocked by the non-α7 antagonist dihydro-β-erythroidine (DHβE, 2 µM, Fig. 2A) but were not significantly altered by the α7 nAChR antagonist methyllycaconitine (MLA, n = 6, P > 0.5, Fig. 2A). In L1 neurons (n = 8), the fast component was abolished by MLA but was unaffected by DHβE (Fig. 2B). These results indicate that homomeric nAChR containing α7 subunits mediate the fast current in L1 interneurons and that heteromeric nAChRs comprising α and β subunits underlie the slower current observed in L1 and L6 neurons. Based on expression studies in these cell types, these heteromeric nAChRs most likely comprise α4 and β2 subunits and will be denoted α4β2 in the rest of the manuscript, although α4 and β2 are co-expressed with the α5 subunit in L6 neurons (Christophe et al. 2002; Salas et al. 2003; Hay et al. 2014).
The I–V curve of the α4β2 current in L6 neurons was roughly linear at negative potentials, reversed at 0 mV and exhibited a strong inward rectification at positive potentials with a RI = 0.07 ± 0.06 (n = 9, Fig. 2B). Similar I–V relationships were obtained for α4β2 and α7 currents recorded in L1 neurons (RI = 0.05 ± 0.05 and 0.03 ± 0.01, respectively, n = 9 and 9, respectively; Fig. 2D). The I–V curves could be fitted with a Woodhull equation and gave similar values for K(0) and δ (see Methods section): 0.78 mM and 0.92 for the α7 component, 0.32 mM and 1 for the α4β2 component in L1 neurons, 1.37 mM and 1 for the α4β2 current in L6 neurons. It should be noted that, in these experiments, spermine had not been added to the internal solution and, therefore, the main intracellular blocker was likely to be Mg2+ (Ifune and Steinbach 1990). These data are consistent with those obtained on agonist-evoked currents (Couturier et al. 1990; Bertrand et al. 1993; Buisson et al. 1996; McQuiston and Madison 1999; Christophe et al. 2002) or at other nicotinic synapses (Zhang et al. 1996; Lamotte d'Incamps and Ascher 2008).
These results indicate that nicotinic transmission in L6 only involves heteromeric receptors and that responses in L6 are similar to the slow component of responses in L1.
Close Apposition of ChR2+ Cholinergic Varicosities onto Responsive Neurons
It has been proposed that in L1, light-induced α7 currents occur at conventional synapses whereas α4β2 currents result from release of ACh from multiple sites activating distant receptors (Arroyo et al. 2012; Bennett et al. 2012). Given the similarity of light-induced α4β2 currents in L1 and L6 and the absence of α7 currents in L6, this hypothesis would predict that cholinergic BF fibers do not form synaptic contacts onto L6 excitatory neurons. This prompted us to compare anatomical relationships between YFP-labeled cholinergic fibers and biocytin-filled neurons recorded in L1 and L6 using confocal microscopy. In 9 L6 neurons considered for this analysis, it was possible to highlight 5 ± 1 close appositions of YFP+ varicosities onto dendrites of each responsive neuron (Fig. 3A,B), suggestive of synaptic contacts. This was supported by the close proximity of biocytin and YFP signals in single confocal planes (Fig. 3B). Contacts were confirmed by inspection from different angles of view following 3D deconvolution and reconstruction of the surrounding volume (n = 4 contacts, 2 neurons; see Materials and Methods and Fig. 3C–E). Among 6 L1 neurons analyzed, 3 exhibited both α7 and α4β2 light-evoked currents (range of α7 amplitude: 19–72 pA, range of α4β2 amplitude: 8–12 pA) and 3 exhibited only α7-mediated currents (range of amplitude: 25–60 pA). As in L6 neurons, it was possible to highlight 6 ± 1 close appositions between YFP+ varicosities and each responsive L1 neuron (n = 6, Fig. 3F,G). Four of these contacts were confirmed using 3D reconstruction as above (n = 2 neurons, Fig. 3H–J). These putative synaptic contacts were on average closer to neuronal somata in L6 (27.8 ± 3.4 µm) than in L1 (42.8 ± 5.2 µm, P < 0.05, Fig. 3K). It is noteworthy that these YFP+ contacts represent only the ChR2-expressing fraction of all cholinergic varicosities making close apposition onto cortical neurons. These results indicate that BF fibers closely contact both L1 and L6 neurons and suggest that nicotinic transmission occurs synaptically in both L1 and L6.
Onset Kinetics and Amplitude of L6 Nicotinic Currents Are Not Affected by Acetylcholinesterase Inhibition
The effects of acetylcholinesterase (AChE) inhibition on nicotinic currents mediated by heteromeric receptors have been described at several nicotinic synapses and result in a prolongation of the decay of the synaptic currents. In contrast, the onset of the synaptic currents is little affected, consistent with the clustering of receptors close to the release site (Fatt and Katz 1951; Zhang et al. 1996; Lamotte d'Incamps et al. 2012).
We thus examined the effects of AChE inhibition on nicotinic currents evoked by single-light pulses in L6 neurons (n = 6). Application of the AChE inhibitor neostigmine (2 µM) increased fluctuation of baseline current (σ2 CONTROL = 5.4 ± 1.4 pA2 vs. σ2 NEO = 11.0 ± 3.1 pA2), suggestive of low levels of extracellular ACh in resting conditions as found at other nicotinic synapses (Katz and Miledi 1970; Lamotte d'Incamps et al. 2012). The decay kinetics of light-evoked currents were markedly slowed down by neostigmine, attesting to the effectiveness of AChE inhibition (τOFFCONTROL = 387 ± 64 ms vs. τOFFNEO = 4362 ± 781 ms, P < 0.05, Fig. 4). Conversely, neostigmine did not alter the onset time constants and latencies of L6 currents (τONCONTROL = 36 ± 10 ms vs. τONNEO = 36 ± 14 ms, P = 1; onset latency: control, 5.9 ± 0.7 ms, NEO, 6.8 ± 0.8 ms, P = 0.3; Fig. 4), and also failed to increase their amplitude (64.4 ± 26.4 pA in control vs. 58.2 ± 24.3 pA in neostigmine, P = 0.4). These results indicate that ACh released upon single stimuli activates the same pool of receptors regardless of ACh diffusion around the release site, as one would expect from the existence of postsynaptic clusters of α4β2 nAChRs in L6 neurons.
In L1 neurons (n = 8), the effects of neostigmine on the α4β2 current were not fully analyzed because its onset was obscured by the α7 component. We found that neostigmine slowed down the decay of α4β2 currents (τOFFCONTROL = 365 ± 66 ms vs. τOFFNEO = 6260 ± 641 ms, P < 0.05, Fig. 4) but did not significantly change their amplitude (13.1 ± 3.8 vs. 14.1 ± 6.0 pA, P = 0.9), similar to our observations in L6. No significant effect of neostigmine was observed on the α7 current (τONCONTROL = 5.7 ± 1.8 ms vs. τONNEO = 5.8 ± 1.8 ms, P = 0.8; onset latency: control, 2.9 ± 0.3 ms, NEO, 3.0 ± 0.4 ms, P = 0.4; τOFFCONTROL = 7.5 ± 1.5 ms vs. τOFFNEO = 5.9 ± 1.4 ms, P = 0.6; Fig. 4). The differential sensitivity of α7 and α4β2 current decays to AChE inhibition is in agreement with previous observations on L1 neurons (Bennett et al. 2012) and can be explained on the basis of the low affinity/fast desensitization of α7 receptors and the high affinity/slow desensitization of α4β2 receptors (Dani and Bertrand 2007; Albuquerque et al. 2009).
These results are consistent with the existence of postsynaptic clusters of nAChRs in both L1 and L6 neurons.
Near-Saturation of α4β2 Receptors During Nicotinic Transmission onto L6 Neurons
We noticed that α4β2 currents in L6 exhibit low trial-to-trial variability, as reported by Bennett et al. (2012) for L1 interneurons. Bennett et al. (2012) postulated that this low variability results from nonsynaptic activation of α4β2 receptors by volume transmission from multiple ACh release sites. However, another possible explanation is the near-saturation of postsynaptic α4β2 receptor clusters by ACh released from presynaptic terminals. We addressed this issue by determining the unitary properties of nAChR channels underlying L6 postsynaptic currents (n = 13 neurons), which involve only α4β2 receptors and exhibit monoexponential decay. This was performed using nonstationary noise analysis that infers properties of the single channels underlying a given current, from the fluctuations around the mean amplitude of that current (Sigworth 1980). In the sample L6 pyramidal cell recordings shown in Figure 5A,B, nicotinic current amplitudes ranged between 16 and 22 pA (mean amplitude = 18.8 pA). The mean variance of the baseline noise (2.38 pA2) was subtracted from the variance of the currents. The plot of the variance versus the mean amplitude was fitted with a parabolic equation, from which we calculated values of nAChR single-channel conductance, number of channels and their maximal probability of opening (Fig. 5B). The mean value of nAChR single-channel conductance was 13.9 ± 1.1 pS, consistent with values obtained on recombinant α4β2 receptors (Li and Steinbach 2010). The mean number of channels underlying L6 currents was 42 ± 8 and the maximal probability of opening of these channels was 0.94 ± 0.02. We observed a linear correlation between the number of channels and current amplitudes indicating that the maximal opening probability of nAChR channels was constant over a large range of current amplitudes (Fig. 5C). The high maximal probability of opening of α4β2 channels (0.94 ± 0.02) indicates quasi-saturation of these receptors upon ACh release in L6. These results are in favor of synaptic transmission since receptor saturation is difficult to reconcile with volume transmission.
The variability of the synaptic currents depends on both the probability of release and the probability of occupation of the postsynaptic receptors. Thus the low variance of the synaptic current cannot be only explained by saturation of nAChRs, but also implies a very high probability of ACh release. We thus examined the effect of decreasing the release probability on the properties of nicotinic currents in L6 pyramidal cells (n = 7). This was achieved by lowering Ca2+ from 2 to 1 mM and increasing Mg2+ from 1 to 3 mM in the perfusion solution (Del Castillo and Katz 1954; Silver et al. 1996). In these conditions, the failure rate of evoked nicotinic currents increased from 0% to 79.4 ± 2.6%, consistent with a decrease of ACh release probability (Fig. 5D,E). The mean amplitude of nicotinic currents (excluding failures) decreased from 8.9 ± 1.6 pA in control conditions to 2.5 ± 0.2 pA in low Ca2+/high Mg2+ solution, indicating that nicotinic currents in control conditions involved multiquantal ACh release. Importantly, neither onset nor decay kinetics were significantly altered in low Ca2+/high Mg2+ conditions (τONCONTROL = 26.5 ± 3.1 ms vs. τONLow Ca/High Mg = 26.8 ± 3.1 ms, P > 0.5; τOFFCONTROL = 205 ± 17 ms vs. τOFFLow Ca/High Mg = 183 ± 23 ms, P = 0.2, Fig. 5E). These results indicate that recruitment of nAChRs in low release probability conditions proceeds similarly as in high release probability conditions, consistent with the existence of nicotinic synapses involving α4β2 receptors in L6 neurons.
Recruitment of Extrasynaptic Nicotinic Receptors by Repetitive Stimulations
We next examined whether large amounts of ACh released upon repetitive stimulation can recruit additional, that is, extrasynaptic, receptors. In awake animals, BF cholinergic neurons undergo episodes of 10–20 Hz firing, which may underlie large cortical ACh transients observed during attention tasks (Lee et al. 2005; Parikh et al. 2007). We thus tested the effect of trains of light pulses (10 Hz for 5 s) on L1 and L6 neurons in the continuous presence of atropine (10 µM).
We first verified that ACh release persisted along trains of light pulses, as reported with trains of electrical stimulation (Buhler and Dunwiddie 2001; Stanchev and Sargent 2011), by monitoring fast α7 currents in L1 neurons exhibiting minimal α4β2 currents (n = 4). Indeed, fast currents occurred in response to individual light pulses throughout the train, indicative of ACh release at each light pulse (Fig. 6A). Their amplitude rapidly decreased at the beginning of the train, consistent with the short-term depression at this synapse (Bennett et al. 2012), and then remained roughly stable. No conspicuous change in α7 current kinetics was observed along the train (Fig. 6A).
Subsequent analyses of responses to train stimulation focus on α4β2 currents in L1 neurons exhibiting such currents (n = 9) and in L6 neurons (n = 6). In all cases, the maximal current (Amax) reached during the train was larger than the current elicited by the first pulse (A1; L1: Amax/A1 = 1.7 ± 0.2, P < 0.05; L6: Amax/A1 = 1.7 ± 0.2, P < 0.05, Fig. 6B,C). Hence, more receptors are recruited by trains than by single stimuli. Given the quasi-saturation of synaptic α4β2 receptors observed with single stimuli, this suggests that large amounts of released ACh are able to activate α4β2 extrasynaptic receptors in both L1 and L6 neurons. Decay kinetics of responses of L1 neurons to trains of stimuli (τOFFTRAIN = 608.6 ± 109.7 ms) were slower than those measured upon single stimuli (see above τOFFSINGLE = 296.8 ± 29.6 ms, P < 0.05), whereas only a modest change of decay kinetics was observed in L6 neurons (τOFFTRAIN = 342.7 ± 47.2 ms vs. τOFFSINGLE = 257.3 ± 21.6 ms, P = 0.13). We next tested the effect of the AChE inhibitor neostigmine (2 µM) on the response to repetitive photostimulation. Neostigmine did not increase responses to the first pulse (L1: A1NEO/A1CONTROL = 1.2 ± 0.2, P = 0.3; L6: A1NEO/A1CONTROL = 0.9 ± 0.1, P = 0.3, Fig. 6B,C), as observed for responses to single stimuli. In contrast, the maximal current amplitude was strongly increased by neostigmine in both L1 and L6 neurons (AmaxNEO/AmaxCONTROL = 3.1 ± 0.7, P < 0.05, and 1.8 ± 0.2, P < 0.05, respectively), in agreement with the recruitment of a larger pool of extrasynaptic nAChRs due to reduced ACh clearance. Decay time constants dramatically increased in both L1 and L6 neurons as compared with control conditions (L1: 13.3 ± 1.9 s, P < 0.05; L6: 18.7 ± 4.7 s, P < 0.05), consistent with the effect of neostigmine observed on responses to single photostimuli. These data indicate that large amounts of ACh evoked by high-frequency stimulation of BF fiber activate α4β2 extrasynaptic receptors in both L1 and L6 neurons.
The α5 nAChR Subunit Selectively Regulates Short-Term Plasticity of L6 Nicotinic Synapses
It is known that nicotinic transmission onto L1 neurons exhibits short-term depression (Bennett et al. 2012). Likewise, we found that interstimulus intervals (ISI) shorter than 50 s resulted in a reduction of current amplitude in L1 and L6 neurons (see above). We thus examined short-term plasticity of nicotinic synapses onto L6 neurons in the presence of atropine (10 µM). A two-pulse protocol (1 ms pulse, 0.5 s ISI), which reliably induces 2 Hz firing of cholinergic neurons (see Supplementary Fig. 1), resulted in a reduction of the second response amplitude (PPR2/1 = 0.47 ± 0.03, n = 5 neurons, Fig. 7A). These values are in agreement with the paired-pulse depression of nicotinic currents observed in L1 following optogenetic stimulation or in the thalamus following electrical stimulation of cholinergic fibers (Bennett et al. 2012; Sun et al. 2013). Increasing light pulse duration (up to 50 ms, 0.5 s ISI) in the same neurons did not significantly change the PPR2/1 (0.50 ± 0.03 for 10 ms pulse, 0.51 ± 0.04 for 50 ms pulse; P > 0.05 for comparison with 1 ms pulse, Fig. 7A), indicating that short-term depression was not influenced by ChR2 desensitization.
We next investigated the recovery from short-term depression of nicotinic currents in L6 neurons (n = 10) using a series of 7 light pulses applied with increasing ISI (from 1 to 60 s). Nicotinic currents depressed between the first and second stimulus (ISI = 1 s, 0.47 ± 0.03), as observed for the shorter ISI. The recovery of α4β2 currents in L6 neurons was slow and only reached PPR7/1 = 0.87 ± 0.05 for ISI = 60 s. The auxiliary α5 nAChR subunit is co-expressed with α4 and β2 subunits in L6 neurons (Salas et al. 2003; Bailey et al. 2010; Hay et al. 2014). The α5 subunit assembles with α4 and β2 subunits and modulates desensitization of heteromeric nAChRs (Ramirez-Latorre et al. 1996; Gerzanich et al. 1998; Kuryatov et al. 2008). To test the possible contribution of the α5 subunit to the recovery from short-term depression of nicotinic currents in L6 neurons, we expressed ChR2-YFP into the BF of α5-null mice ((Salas et al. 2003), see Materials and Methods) and recorded light-evoked nicotinic currents in the presence of glutamatergic, GABAergic, and muscarinic antagonists as described above. The amplitude of α7 currents in L1 neurons and of α4β2 currents in L6 neurons of α5−/− mice (63 ± 33 pA, n = 3, and 23 ± 3 pA, n = 24, respectively) were comparable with those measured above in α5+/+ animals. Furthermore, when examining the recovery profile of L6 nicotinic currents using the same stimulation paradigm as above, we found that short-term depression in α5−/− mice (for ISI = 1s, PPR2/1 = 0.48 ± 0.04, n = 10) was similar to that observed in α5+/+ animals (P > 0.95, Fig. 7). Nonetheless, L6 nicotinic currents recovered faster (P < 0.05) and more completely (PPR7/1 = 0.98 ± 0.04 for ISI = 60 s) in α5−/− than in α5+/+ mice (Fig. 7B,C). These results indicate that the α5 subunit participates in heteromeric receptors that mediate nicotinic transmission onto L6 neurons and contributes to the slow recovery from short-term depression observed at this synapse.
We characterized fast nicotinic transmission elicited by optogenetic release of ACh from BF fibers onto excitatory L6 cortical neurons. Our results indicate that low frequency stimulation results in point-to-point nicotinic transmission occurring synaptically via heteromeric receptors likely composed of α4, α5, and β2 subunits. Conversely, high-frequency stimulation results in a switch to tonic responses and in the recruitment of extrasynaptic receptors by spillover.
Cortical Nicotinic Transmission
We selectively expressed ChR2 in cholinergic neurons using the GM60 Cre mouse line, which proved as efficient for optogenetic release of ACh in the neocortex as the GM24 line (Arroyo et al. 2012; Bennett et al. 2012). Results obtained with the GFP reporter mouse line show that the GM60 line targets more efficiently and selectively cholinergic neurons in corticopetal BF nuclei and striatum than in hippocampopetal medial septal nucleus and neocortex. The BAC used to generate the GM60 Cre line contains the first intron of the ChAT gene, thus also the vesicular ACh transporter gene (Bejanin et al. 1994; Roghani et al. 1994; Gong et al. 2007). Nonetheless, currents we measured in non-cre mice with same genetic background (α5−/− line) provided no evidence for enhanced nicotinic transmission in the GM60 line, in contrast with the hypercholinergic phenotype reported for another BAC-based ChAT-ChR2-YFP transgenic mouse line (Kolisnyk et al. 2013). Light-induced currents were purely nicotinic without noticeable contribution of a glutamatergic component, consistent with the segregation of BF neurons into cholinergic, glutamatergic, and GABAergic types (Gritti et al. 1997). In contrast, ACh-glutamate co-transmission occurs onto Renshaw and interpeduncular nucleus neurons, whose input cholinergic neurons express vesicular glutamate transporters (Herzog et al. 2004; Lamotte d'Incamps and Ascher 2008; Ren et al. 2011). The present study and previous reports (Arroyo et al. 2012; Bennett et al. 2012) indicate that BF cholinergic projections widely mediate point-to-point nicotinic transmission onto diverse excitatory and inhibitory neuron types in the neocortex. Our current findings along with observations in the striatum (English et al. 2012), the thalamus (Sun et al. 2013), and the hippocampus (Bell et al. 2011; Grybko et al. 2011; Leao et al. 2012) suggest that the occurrence of point-to-point nicotinic transmission in the brain has been largely underestimated.
Nicotinic Transmission in L6 Is Mediated by Classical Synapses
We provide evidence that nicotinic transmission onto L6 excitatory neurons occurs at synapses bearing non-α7 heteromeric nAChRs likely composed of α4, α5, and β2 subunits. 1) We observed close appositions between each responsive neuron and ChR2+ BF varicosities. 2) Onset latencies and kinetics of nicotinic currents and their amplitudes were insensitive to AChE inhibition, in agreement with the existence of postsynaptic clusters of nAChRs close to ACh release sites. 3) We found evidence for the quasi-saturation of nAChRs upon ACh release. 4) L6 response kinetics was not altered in low release probability conditions.
The present and earlier (Arroyo et al. 2012; Bennett et al. 2012) observations are consistent with a model of cortical nicotinic synapse, in which high-affinity (α4β2) receptors are clustered in the postsynaptic density. As first shown at the neuromuscular junction (Anderson and Stevens 1973), the rate of ACh unbinding is a major factor determining the decay of the synaptic current. By analogy, we propose that slow ACh unbinding from high-affinity α4β2 receptors is responsible for the slow decay of α4β2 currents. We found that α4β2 currents exhibited low trial-to-trial variability. The low trial-to-trial variability of α4β2 currents implies saturation of the receptors, consistent with their high affinity and as deduced from noise analyses. It also implies a high ACh release probability, which is likely due to the use of ChR2 that dramatically enhances release probability at glutamatergic synapses when compared with electrical stimulation (Schoenenberger et al. 2011). The small amplitude of mean responses obtained in low release probability conditions (2.5 pA) suggests that individual nicotinic synapses in L6 are on average equipped with few postsynaptic nAChR channels. Indeed, given the unitary channel conductance, we measured at this synapse (13.9 pS), this current amplitude reflects the opening of just 3 channels. For comparison, hippocampal glutamatergic synapses are estimated to contain 30 NMDA receptor channels (Silver et al. 1992; Nusser et al. 1998), which share with non-α7 nAChRs a high-affinity/opening probability. It is noteworthy that the small number of nAChR channels per individual L6 nicotinic synapse is in agreement with the small size of cholinergic synapses, whose surface is a tenth of that of GABAergic synapses upon ultrastructural examination (Takacs et al. 2013).
The mixed α7 and α4β2 nicotinic synapse of L1 neurons is similar to those of the ciliary ganglion and the Renshaw cell (Zhang et al. 1996; Ullian et al. 1997; Lamotte d'Incamps and Ascher 2008), and bears resemblance to the glutamatergic synapse with low-affinity/fast desensitizing AMPA and high-affinity/slow desensitizing NMDA receptors. Several differences between α7 and α4β2 current properties likely result from the different receptor affinities or desensitization kinetics (Dani and Bertrand 2007). Indeed, the fast decay of the α7 current is readily explained by the fast desensitization and/or ACh unbinding rate of this receptor. Bennett et al. (2012) reported a high trial-to-trial variability of α7 currents, in contrast with α4β2 currents. This variability is attributable to partial occupation of α7 receptors due to their low affinity. The different latencies and onset kinetics of α7 and α4β2 currents may also be explained by biophysical properties of these receptors. Indeed, similar differences between AMPA and NMDA currents arise from the delay between glutamate binding and channel opening of NMDA receptors at the glutamatergic synapse (Hestrin et al. 1990; Dzubay and Jahr 1996). The present observation of pure α7 and α4β2 synapses in L1, despite functional expression of both receptor types in virtually all L1 neurons (Christophe et al. 2002), suggests that clustering of postsynaptic α7 and α4β2 nAChRs is differentially regulated. Our results further indicate that the auxiliary α5 subunit participates in postsynaptic α4β2 heteromers and modulates recovery from short-term depression at L6 nicotinic synapses. Hence, combinatorial expression of nAChR subunits endows cortical nicotinic synapses with diverse cell type-specific properties.
Train Stimulation Recruits Extrasynaptic Receptors and Elicits Tonic Responses
Train stimulation resulted in a tonic current that slowly developed during the train in L1 and L6 neurons. The tonic current involved extrasynaptic α4β2 receptors since its amplitude was larger than that of α4β2 synaptic current elicited by the first pulse and since decay kinetics of the tonic current were slower than those measured upon single stimuli, as expected from slower neurotransmitter diffusion in the extrasynaptic space than in the synaptic cleft (Nicholson and Phillips 1981). It is likely that the small size of cholinergic synapses (Takacs et al. 2013) facilitates spillover of ACh from the synaptic cleft, and thus the recruitment of extrasynaptic receptors. Decay kinetics of responses to train stimulation were faster in L6 than in L1 neurons. This difference was abolished upon AChE inhibition, suggesting that AChE activity is higher in L6 than in L1. AChE inhibition resulted in recruitment of a larger pool of extrasynaptic α4β2 receptors and drastically slowed down onset and decay kinetics. This suggests that activation of extrasynaptic α4β2 receptors in control conditions proceeds from diffusion of ACh at short distance from the synapse, presumably by spillover from the synaptic cleft, whereas AChE inhibition may unveil ACh release from varicosities distant from the postsynaptic neuron. It is noteworthy that our recordings were performed at room temperature and that spillover extent can be reduced at higher temperature, as demonstrated for glutamatergic transmission due to increased activity of glutamate transporters (Asztely et al. 1997). It is thus possible that, at physiological temperature, the increased AChE activity may limit extrasynaptic diffusion of ACh, and hence the recruitment of extrasynaptic α4β2 receptors.
Our results indicate that both point-to-point and tonic nicotinic transmission occur in the neocortex depending on the regime of BF neurons. Indeed, temporal summation of α4β2 currents presumably occurs in L1 and L6 when BF fibers discharge in excess of 1 Hz. Hence, pure nicotinic transmission that prevails in the neocortex presumably undergoes a switch from point-to-point to tonic mode accompanied with loss of temporal fidelity at moderate discharge frequencies of BF neurons. In contrast, ACh-glutamate co-transmission occurring at other synapses (Lamotte d'Incamps and Ascher 2008; Ren et al. 2011) may allow the co-existence of time-locked point-to-point and tonic transmission over a larger range of frequencies. Hence, the operating mode of nicotinic transmission is likely to depend critically on the brain region and behavioral state.
This work was supported by Centre National de la Recherche Scientifique, Université Pierre et Marie Curie-P6 and by grants from Ecole des Neurosciences de Paris (“Network for Viral Transfer”), Fondation pour la Recherche sur le Cerveau/Rotary Club de France, and Agence Nationale de la Recherche (“IHU Institut de Neurosciences Translationnelles de Paris,” ANR-10-IAIHU-06).
The authors thank Alain Chédotal, David DiGregorio, Thierry Gallopin, Pascal Legendre, Uwe Maskos, Stéphanie Pons, Martine Soudan, Kenneth Pelkey, and the IBPS Cell Imaging Facility for valuable help and discussion. Conflict of Interest: None declared.