Glutamatergic Nonpyramidal Neurons From Neocortical Layer VI and Their Comparison With Pyramidal and Spiny Stellate Neurons

Slice preparation

All experiments were carried out in accordance with the guidelines published in the European Communities Council Directive of 24 November 1986 (86/609/EEC). Young Wistar rats (18 ± 2 days old, range 14–21) were decapitated and brains were quickly removed and placed into cold (∼4°C) oxygenated artificial cerebrospinal fluid (ACSF) containing (in mM): 126 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 2 CaCl₂, 1 MgCl₂, 26 NaHCO₃, 20 glucose, 5 pyruvate, and 1 kynurenic acid (nonspecific glutamate receptor antagonist; Sigma, St. Louis, MO). Parasagittal 300-μm-thick sections of cerebral sensorimotor cortex were prepared as described previously (Cauli et al. 1997). Slices were collected and subsequently transferred to a holding chamber containing ACSF saturated with 95% O₂-5% CO₂ and held at room temperature. Slices were allowed to equilibrate for 1 h before recordings.

Whole cell recordings

Individual slices were then transferred to a recording chamber placed under a microscope (Olympus BX51WI). Slices were maintained immersed and continuously superfused at 1–2 ml/min with oxygenated ACSF at room temperature. Patch micropipettes (3–5 MΩ) were pulled from borosilicate glass capillaries (OD, 1.5 mm; ID, 0.86 mm; Harvard Apparatus, Les Ulis, France) on a Brown-Flaming micropipette puller (Model PP-83, Narishige, Tokyo, Japan). Electrodes were filled with 8 μl of internal solution containing (in mM): 144 K-gluconate, 3 MgCl₂, 0.5 EGTA, and 10 HEPES plus 2 mg/ml biocytin (Sigma). The pH was adjusted to 7.2 and osmolarity to 285/295 mOsm. Whole cell recordings were made from neocortical neurons selected under infrared video microscopy (Stuart et al. 1993) using a patch-clamp amplifier (Multiclamp 700B, Molecular Devices, Sunnyvale, CA) connected to a Digidata 1322A interface board (Molecular Devices). Signals were amplified and collected using the data-acquisition software pClamp 9.2 (Molecular Devices). Resting membrane potential was measured just after passing in whole cell configuration and only cells with a resting membrane potential more hyperpolarized than −50 mV were analyzed. Membrane potentials were not corrected for junction potential. Cells were maintained at a holding potential of −60 mV by continuous current injection and their firing behavior was tested by applying depolarizing current pulses. Signals were filtered at 5 kHz, digitized at 10 kHz, saved to a personal computer, and analyzed off-line with Clampfit 9.2 software (Molecular Devices).

Twelve electrophysiological parameters were determined for each neuron. Membrane resistance was measured by applying a hyperpolarizing current pulse (amplitude −50 pA, duration 800 ms). On injection of hyperpolarizing current pulses a “sag,” indicative of a hyperpolarization-activated cationic current (“I_h”), followed the initial hyperpolarization peak. The amplitude of the sag was quantified as a decrease in membrane resistance. Thus membrane resistance was measured when the sag conductance was active (R-sag) and inactive (R-hyp). R-sag was measured as the slope of the linear portion of a current–voltage (I–V) plot, where V was determined at the end of an 800-ms hyperpolarizing current pulse and R-hyp as the slope of the linear portion of an I–V plot, where V was determined as the maximal negative potential during the 800-ms hyperpolarizing pulse. The sag (% membrane resistance decrease) was calculated according to (R-hyp − R-sag)/R-hyp.

Analysis of the waveforms of the first two spikes was performed on action potential (AP) discharges elicited by pulses of depolarizing current in the 50- to 150-pA range. The amplitude of the first two APs (A1 and A2) was measured from the threshold to the peak of the spike. Their duration (D1 and D2) was measured at half-amplitude. The amplitude reduction and the duration increase were calculated according to (A1 − A2)/A1 and (D2 − D1)/D1, respectively. Amplitudes of the afterhyperpolarization (AHP) of the first and the second spikes were measured between the spike threshold and the peak of the AHP. The frequency adaptation parameters were also measured on discharges elicited by application of 800-ms depolarizing current pulses (sampling rate 10 kHz). The instantaneous discharge frequency was determined all along the discharge and plotted as a function of time at all stimulation intensities tested. The instantaneous discharge frequencies between the first two spikes (f_initial), 200 ms after the beginning of the discharge (f₂₀₀) and at the end of the stimulation (f_final), were then measured. Early and late accommodations were calculated according to (f_initial − f₂₀₀)/f_initial and (f₂₀₀ − f_final)/f₂₀₀, respectively. Statistical analysis of adaptation parameters (early and late adaptation) was performed on the values corresponding to an evoked discharge with an initial firing frequency of 50 Hz.

Statistical analyses

Unsupervised cluster analysis was used to classify sampled neocortical neurons without a priori knowledge of the number of groups (Cauli et al. 2000) by combining the 12 electrophysiological variables described earlier. After centering and reducing the data, cluster analysis was performed using squared Euclidean distances and Ward's linkage rules (Ward 1963). A Thorndike critical threshold procedure was used to suggest the number of different clusters in the data set (Thorndike 1953). Descriptive statistics and cluster analysis were calculated with Statistica v 6.0 (StatSoft France, Paris).

The comparison of occurrence of a given molecular marker between populations of VGluT+/GAD− neurons was made according to

(1)

where p_a, p_b, n_a, and n_b represent the percentage of occurrence (p) and the number of individuals (n) in two populations a and b; p is the percentage of occurrence in the overall population with q = 1 − p; |ɛ| was compared with a normal distribution for statistical significance (Fisher and Yates 1946).

Comparison of the mean numbers of GABAergic interneuron markers expressed per cell in different populations was performed using a Mann–Whitney U test.

Single-cell reverse transcription–polymerase chain reaction

At the end of the recording, the cell cytoplasm was aspirated into the recording pipette while maintaining a tight seal. Then, the pipette was removed delicately to allow outside-out patch formation. Next, the content of the pipette was expelled into a test tube and reverse transcription (RT) was performed in a final volume of 10 μl as described previously (Lambolez et al. 1992). Next, two steps of polymerase chain reaction (PCR) were performed essentially as described previously (Cauli et al. 1997). The cDNAs present in 10 μl of the RT reaction first were amplified simultaneously using the primer pairs described in Table 1 (for each primer pair the sense and antisense primers were positioned on two different exons). Taq polymerase (2.5 U; Qiagen, Hilden, Germany) and 20 pmol of each primer were added to the buffer supplied by the manufacturer (final volume, 100 μl), and 21 cycles (94°C for 30 s, 60°C for 30 s, and 72°C for 30 s) of PCR were run. Second rounds of PCR were performed using 2 μl of the first PCR product as template. In this second round, each cDNA was amplified individually with a second set of primer pair internal to the primer pair used in the first PCR (nested primers; see Table 1) and positioned on two different exons. Thirty-five PCR cycles were performed (as described earlier). Then 10 μl of each individual PCR were run on a 2% agarose gel, with ΦX174 digested by Hae III as a molecular weight marker and stained with ethidium bromide. The RT-PCR protocol was tested on 500 pg of total RNA purified from rat neocortex. All the transcripts were detected from 500 pg of neocortical RNA. Sizes of the PCR-generated fragments were as predicted by the mRNA sequences (see Table 1). A control for mRNA contamination from surrounding tissue was performed by placing a patch pipette into the slice without establishing a seal. Positive pressure was then interrupted and, following the removal of the pipette, its content was processed as described. No PCR product was obtained using this protocol (n = 20).

Intracellular labeling

Slices containing the recorded neurons filled with biocytin were fixed overnight in 4% paraformaldehyde in phosphate-buffered saline (PBS) at 4°C. The morphology of the recorded neurons was investigated by histochemical labeling of intracellular biocytin, with Streptavidin Alexafluor 568 (Invitrogen, Cergy Pontoise, France) or diaminobenzidine (DAB).

For Alexafluor 568 staining, slices were washed six times in 0.1 M phosphate buffer for 10 min, rinsed three times in 0.1 M PBS gelatin (0.2%) Triton (0.25%) for 10 min, and incubated with Streptavidin Alexafluor 568 for 4 h. Slices were washed four times in 0.1 M PBS for 10 min and mounted in Vectashield (Vector Laboratories, Burlingame, CA). Images of labeled cells were captured with a digital camera (Cool Snap Photometrics, Roper Scientific, Ottobrunn, Germany) mounted on a microscope (Leica DMR HC, Wetzlar, Germany) or reconstructed by confocal microscopy (Leica DMI Simil SP5) with excitation at 561 nm and emission at 566–700 nm.

For DAB staining, biocytin-filled neurons were labeled using the avidin-biotin-peroxidase complex (ABC) method (ABC Elite kit; Vector Laboratories). After blocking endogenous peroxidase with 3% H₂O₂ for 15–30 min, slices were rinsed five times in PBS for 10 min, permeabilized in 2% Triton in PBS for 1 h, and incubated in ABC for 2 h. Incubated slices were washed eight times in 0.1 M PBS for 10 min before immersion in a solution containing 0.05% of 3,3′-diaminobenzidine tetrahydrochloride (Sigma) in 0.1 M PBS and 0.02% H₂O₂. Visualization of DAB staining was monitored under the dissecting microscope and stopped by rinsing the slices five times in PBS for 10 min. The slices were then mounted in aqueous medium (Moviol) and images of the neurons were captured with a digital camera (Qicam from QImaging, Surrey, BC, Canada) mounted on a microscope (Leica DMR HC, Leica Microsystems, Rueil-Malmaison, France). Reconstructions and morphological analysis of representative neurons were made with the Neurolucida tracing system (Microbrightfield Europe, Magdeburg, Germany) attached to a Leica DMR HC microscope using a ×100 oil-immersion objective.

For correlated light- and electron microscopy, slices containing the recorded neurons filled with biocytin were fixed in 2.5% paraformaldehyde, 1.25% glutaraldehyde, and 15% (vol/vol) saturated picric acid in 0.1 M phosphate buffer (pH 7.4). Visualization of biocytin was performed as described (Tamás et al. 1997). Three-dimensional light microscopic reconstructions were carried out using Neurolucida with ×100 objective. Correlated light- and electron microscopy was performed as described earlier (Szabadics et al. 2001; Tamás et al. 1997).

Characterization of VGluT+/GAD+ nonpyramidal neurons from layer VI

Nonpyramidal neurons from neocortical layer VI (n = 114), identified under infrared videomicroscopy as lacking a prominent apical dendrite, were electrophysiologically characterized and subsequently analyzed by scPCR (see methods). The nonpyramidal morphology of the recorded neurons was confirmed by histochemical labeling of intracellular biocytin (see methods). Twelve electrophysiological parameters were determined for each neuron and the scPCR protocol was designed to detect the expression of mRNAs encoding the vesicular glutamate transporter (VGluT1) and γ-aminobutyric acid (GABA)–synthesizing enzymes (glutamic acid decarboxylase, GAD 65 and GAD 67), in addition to the following markers of GABAergic interneuron diversity (Cauli et al. 1997; Kubota et al. 1994): calcium binding proteins calretinin (CR), calbindin (CaB), and parvalbumin (PV); and neuropeptides vasoactive intestinal peptide (VIP), somatostatin (SOM), cholecystokinin (CCK), and neuropeptide Y (NPY). Expression of the Nurr1 mRNA was additionally tested in a subset of 24 neurons.

We found that the majority of layer VI nonpyramidal neurons (n = 71 of 114, 62%) were negative for GAD 65 and GAD 67 (GAD−), but VGluT1 positive (VGluT+) on scPCR analysis. The 43 remaining cells were GAD 65 and/or GAD 67 positive and were discarded from the present study. The nonpyramidal morphology was confirmed for 34 of the 71 VGluT+/GAD− neurons following biocytin labeling. All lacked the prominent apical dendrite typical of pyramidal neurons and their somata were localized in layer VI at a maximum of 350 μm from the white matter. Numerous dendritic spines were observed on all VGluT+/GAD− neurons analyzed with confocal microscopy (n = 8). This indicates that deep layer VI contains a high proportion of putative glutamatergic nonpyramidal neurons. Figure 1 displays two examples of the morphological, electrophysiological, and molecular analysis of VGluT+/GAD− nonpyramidal neurons from layer VI. Both cells, located in deep layer VI close to the white matter, lacked a prominent apical dendrite. Consistent with their identification as putative glutamatergic neurons, numerous spines were observed on the dendrites of both neurons. Analysis of AP firing patterns showed that both neurons exhibited long-duration APs, relatively small AHP, and decrease of their AP amplitude and frequency along the discharge. These electrophysiological properties define the broad class of regular-spiking neurons (RS), to which glutamatergic neurons from other layers also belong (Connors and Gutnick 1990; see also Figs. 5 and 6). Molecular analysis showed that both nonpyramidal neurons expressed the VGluT1 mRNA. None of the GABAergic interneuron markers was detected in these neurons except the NPY and CCK mRNAs. Thus the electrophysiological and molecular properties of these spiny neurons are consistent with their identification as putative glutamatergic neurons.

Of the 14 VGluT+/GAD− neurons additionally probed for Nurr1 expression, 5 (36%) expressed the Nurr1 mRNA, consistent with a previous report showing Nurr1 expression in half of layer VI putative glutamatergic neurons (Arimatsu et al. 2003). This suggests that both corticocortical ipsilateral (Nurr1-positive) and commissural (Nurr1-negative) projecting neurons were present in our sample of VGluT+/GAD− layer VI nonpyramidal neurons. Figure 2 displays examples of a Nurr1-positive and a Nurr1-negative neuron. Both cells lacked a prominent apical dendrite and numerous spines were observed on their dendrites. Both neurons exhibited an RS firing pattern with reduction of AP amplitude and frequency along the discharge. Finally, both cells expressed CCK mRNA, and one expressed NPY mRNA, in addition to VGluT1.

The glutamatergic nature of layer VI spiny nonpyramidal neurons was assessed by examining the morphology of local synaptic contacts made by four such neurons using electron microscopy (see example in Fig. 3). These neurons were located in deep layer VI close to the white matter and exhibited RS firing patterns. The axons made horizontal collaterals in layer VI (n = 2) and/or vertical collaterals running across all cortical layers (n = 3). All synapses (n = 21) examined from the four axons were asymmetrical, as visible in Fig. 3, where a labeled axon terminal of the recorded neuron makes a synapse with an unlabeled dendritic spine exhibiting a postsynaptic density. These results confirm that layer VI spiny nonpyramidal neurons are glutamatergic excitatory neurons.

The morphological diversity observed in our sample of VGluT+/GAD− layer VI nonpyramidal neurons is further illustrated in Fig. 4 and was tentatively assigned to morphological subgroups of spiny neurons previously described in layer VI (Prieto and Winer 1999; Tömböl 1984; Zhang and Deschênes 1997). Neurons shown in Figs. 1A, 2A, and 4A fell into the category of “tangential pyramidal” neurons due to the presence of a main dendrite extending obliquely and the general asymmetry of the dendritic tree. This tangential pyramidal morphology was observed in 13 of 34 biocytin-labeled neurons. Figure 4B shows an example of “inverted pyramidal” morphology also reported for layer VI spiny neurons, which we observed in a total of 3 neurons. In addition, 4 neurons exhibiting a vertically extended fusiform soma with two dendritic tufts emerging from its upper and lower poles (Fig. 4C) appeared similar to earlier described “fusiform vertical” spiny neurons. The excitatory nature of layer VI inverted pyramidal and fusiform vertical spiny neurons has been earlier assessed (Mercer et al. 2005). Finally, the neurons shown in Figs. 1B and 3 exhibiting horizontal fusiform somata, but markedly different dendritic trees, and the multipolar neurons shown in Figs. 2B and 4D provide examples of the 14 VGluT+/GAD− nonpyramidal neurons that we could not assign to morphological groups of layer VI spiny neurons earlier defined.

Analysis of AP firing patterns showed that, regardless of their morphological features, all VGluT+/GAD− nonpyramidal neurons were RS cells, exhibiting long-duration APs, relatively small AHPs and decreasing AP amplitude and frequency (Figs. 1–4 and Table 2). On injection of hyperpolarizing current pulses, a “sag” indicative of a hyperpolarization-activated cationic current (I_h), followed the initial hyperpolarization peak. The variability of sag amplitudes observed among layer VI VGluT+/GAD− nonpyramidal neurons (see Figs. 1–3) could not be correlated with cellular morphology. In some neurons, application of a hyperpolarizing current pulse was followed by a marked depolarizing rebound, or low-threshold spike (LTS), which triggered APs (see example in Fig. 4C). LTS was observed in 17% (n = 12 of 71) of layer VI VGluT+/GAD− nonpyramidal neurons, including all morphologically identified fusiform vertical neurons (n = 4). However, the presence of the LTS was not specific to this morphological type, in agreement with previous observations (van Brederode and Snyder 1992). These electrophysiological properties are consistent with those reported earlier for layer VI spiny nonpyramidal neurons (van Brederode and Snyder 1992). Mean values of 12 electrophysiological parameters obtained for the 71 VGluT+/GAD− layer VI nonpyramidal neurons are given in Table 2. Comparison between subgroups of morphologically identified layer VI nonpyramidal neurons (tangential pyramidal, n = 13; fusiform vertical, n = 4; and inverted pyramidal, n = 3) revealed significant differences in only 4 of these 12 electrophysiological parameters. Tangential neurons had larger (P ≤ 0.05) first- and second-spike amplitudes (94 ± 8 and 91.4 ± 9.1, respectively) than those of fusiform vertical (82.3 ± 1.7 and 76.3 ± 7.3, respectively) and inverted pyramidal neurons (82.4 ± 4.3 and 75.7 ± 4.3, respectively). Fusiform vertical neurons had greater (P ≤ 0.05) spike amplitude reduction than that of tangential neurons (73.3 ± 6.3 and 2.8 ± 3, respectively). Finally, early adaptation was greater in fusiform vertical neurons (68.7 ± 17.3) than that in inverted pyramidal neurons (48.7 ± 4.3, P ≤ 0.05) and in tangential neurons (46.9 ± 6.8, P ≤ 0.01).

Molecular analysis of VGluT+/GAD− nonpyramidal neurons revealed variable expression patterns of GABAergic interneuron marker mRNAs, with some cells coexpressing up to four markers (see example in Fig. 4C), whereas none of the markers was detected in other neurons (see example in Fig. 4D). In our sample of 71 VGluT+/GAD− layer VI nonpyramidal neurons, the mean occurrence of interneuron marker expression detected was 1.3 ± 1 per cell, with CCK showing the highest occurrence (61%), followed by NPY (23%), CR (20%), and SOM (8%). CaB, PV, and VIP were detected in only five cells, three cells, and one cell, respectively. These results are summarized in Fig. 7. The occasional occurrence of CaB and PV in putative glutamatergic nonpyramidal neurons from layer VI was previously described (van der Gucht et al. 2003).

Thus our combined morphological, electrophysiological, and molecular data support the existence of a large population of glutamatergic nonpyramidal neurons in layer VI exhibiting extensive morphological diversity.

Glutamatergic neurons from other layers

To compare layer VI glutamatergic nonpyramidal neurons with well-established subgroups of neocortical glutamatergic neurons, we next collected spiny stellate neurons and pyramidal cells.

Spiny stellate nonpyramidal neurons from layer IV (n = 10), identified under infrared videomicroscopy from the regular polygonal shape of their soma (Fig. 5), were electrophysiologically characterized and analyzed by scPCR. Morphological examination of 8 of these neurons following biocytin staining confirmed their identification as spiny stellate cells, with multiple primary dendrites arising from the cell body, the presence of numerous dendritic spines, and the absence of a prominent apical dendrite (Fig. 5). Consistent with earlier reports (Connors and Gutnick 1990; Schubert et al. 2003; Staiger et al. 2004), spiny stellate neurons exhibited RS firing patterns with long-duration APs, small AHPs and adaptation of AP amplitude and frequency along the discharge (Fig. 5 and Table 2). In addition, an LTS was frequently observed in our sample of stellate neurons (n = 8 of 10; see Fig. 5). Molecular analysis showed that all spiny stellate neurons were VGluT+/GAD− (Figs. 5 and 7), consistent with their established glutamatergic phenotype (Feldmeyer et al. 1999; Lübke et al. 2000; Schubert et al. 2003; Staiger et al. 2004). The mean occurrence of GABAergic interneuron marker detected per spiny stellate cell was low (0.7 ± 0.7), with four cells expressing none of these markers and one cell coexpressing a maximum of two markers. CCK and CR were detected twice, whereas VIP, CaB, and SOM were detected only once in our sample of 10 spiny stellate neurons (Fig. 7).

Pyramidal neurons from layer II/III (n = 28) and layer V (n = 33), identified under infrared videomicroscopy by their pyramidal-shaped soma and prominent apical dendrite, were electrophysiologically characterized and analyzed by scPCR. On morphological examination of 8 layer II/III and 12 layer V pyramidal neurons, all cells demonstrated a prominent apical dendrite ramifying in layer I and basal lateral dendrites (see examples in Fig. 6). All pyramidal neurons exhibited RS firing patterns with long-duration APs, small AHPs, and adaptation of AP amplitude and frequency (Fig. 6 and Table 2). None of the pyramidal neurons collected from either layer II/III or layer V exhibited an LTS. Pyramidal neurons were VGluT+/GAD− (Figs. 6 and 7) and the mean occurrence of GABAergic interneuron marker detected per cell was low in both layer II/III (1.0 ± 0.9) and layer V (0.9 ± 0.9) pyramidal neurons. We found that in addition to CCK, detected in 44% of layer II/III and 32% of layer V cells, CR, CaB, NPY, and SOM were detected in some pyramidal neurons (Figs. 6 and 7), in good agreement with previous reports (Burgunder and Young 3rd 1990; Cauli et al. 2000; Celio 1990; Gallopin et al. 2006; Hill et al. 2007; Kubota et al. 1994; Morino et al. 1994; Ong et al. 1994; Schiffmann and Vanderhaeghen 1991).

Comparison between subgroups of neocortical glutamatergic neurons

All neurons from each of the four glutamatergic subgroups (i.e., layer II/III pyramidal, layer IV stellate, layer V pyramidal, and layer VI spiny nonpyramidal neurons) were VGluT+/GAD− and showed a low mean occurrence of GABAergic interneuron marker per cell. Although layer VI nonpyramidal neurons expressed slightly more GABAergic interneuron markers per cell (1.2 ± 1) than stellate (0.7 ± 0.7), layer II/III (1.0 ± 0.9), and layer V (0.9 ± 0.9) pyramidal neurons, these differences were not statistically significant. The marker showing the highest occurrence in all subgroups was CCK (Fig. 7). The occurrence of CCK was significantly higher in layer VI nonpyramidal neurons (61%) than in stellate (20%, P ≤ 0.05) and layer V (32%, P ≤ 0.01) pyramidal neurons. The only other significant molecular difference between the subgroups was the more frequent occurrence of NPY in layer VI nonpyramidal neurons (23%) compared with layer II/III pyramidal neurons (3.7%, P ≤ 0.05). Thus the present subgroups of glutamatergic neurons exhibited only minor differences in GABAergic interneuron marker expression profiles.

Although all glutamatergic neurons collected in the present study exhibited RS firing patterns, neuronal subgroups exhibited significantly different mean electrophysiological properties (Table 2). Layer VI nonpyramidal neurons were characterized by a smaller sag, longer-duration APs, larger AHPs, and smaller early adaptation compared with the same parameters of all other subgroups. The long-duration APs and small sag of these neurons have been noted in a previous study of layer VI spiny neurons (van Brederode and Snyder 1992). Layer V pyramidal neurons exhibited a smaller input resistance, larger sag, larger AP amplitudes, and smaller late adaptation compared with identical parameters of all other subgroups. The mean AP amplitude reduction and duration increase were larger for stellate neurons than that for all other subgroups. Finally, mean values obtained for layer II/III pyramidal neurons generally corresponded to a midrange value in the data spread across the other three groups.

In spite of their distinctive mean properties, each neuron subgroup showed a large interindividual variability for all electrophysiological parameters considered. Because these subgroups were defined a priori based on morphological and/or layer specificity, we investigated whether an unsupervised cluster analysis (see methods) combining all electrophysiological parameters would segregate subgroups of glutamatergic neurons. Molecular parameters that demonstrated a very low occurrence and/or showed only minor differences between neuron subgroups were excluded from the analysis. The unsupervised cluster analysis segregated neurons into three clusters (Fig. 8), suggested by the Thorndike threshold (dotted line). Almost all layer V pyramidal neurons from our sample were assigned to cluster 1 (94%, n = 31 of 33). These layer V pyramidal neurons comprised 72% of cluster 1, which additionally included 9 layer II/III pyramidal, 1 stellate, and 1 layer VI neuron. Cluster 2 in contrast was heterogeneous, comprising not only the majority of layer II/III pyramidal (64%, n = 18 out 28) and stellate neurons (n = 8 of 10), but also part of our sample of layer VI neurons (24%, n = 17 of 71) and 2 layer V pyramidal cells. These findings suggest that subpopulations of stellate, layer II/III pyramidal, and layer VI nonpyramidal glutamatergic neurons share similar electrophysiological properties. Finally, cluster 3 was composed almost exclusively of layer VI excitatory cells (96%) and contained the majority of layer VI neurons in our sample (75%, n = 53 of 71). Thus this layer VI neuronal subpopulation exhibited a distinctive electrophysiological signature. Accordingly, layer VI neurons grouped in cluster 2 differed significantly from those in cluster 3 for most electrophysiological parameters (Table 3). The firing patterns of cluster 2 and of cluster 3 layer VI neurons resembled those of previously reported phasic–tonic and tonic cells, respectively (van Brederode and Snyder 1992). Phasic–tonic cells showed an abrupt decrease of AP amplitude and/or frequency at the beginning of the discharge, often associated with a spike doublet (see Fig. 4, A–C), whereas Figs. 1, 2, and 4D display examples of tonic cells showing little or smooth variations of AP and AHP amplitudes and of AP frequency. The distribution of layer VI neurons in two different clusters did not correlate with their somatodendritic morphologies (as noted previously by van Brederode and Snyder 1992) or projection specificity. Indeed, both clusters contained tangential, inverted pyramidal, and fusiform vertical layer VI glutamatergic neurons and, in cluster 3, corticocortical Nurr1-positive ipsilateral neurons were not segregated from Nurr1-negative commissural neurons.

Layer VI glutamatergic nonpyramidal neurons

The present results show that a large proportion of layer VI nonpyramidal neurons are glutamatergic, in contrast with the situation generally found in other layers. The presence of VGluT1 and the absence of GAD indicate that these neurons are excitatory and use glutamate as their main neurotransmitter. Indeed, examination at the electron microscopic level showed that these neurons formed asymmetrical excitatory synapses. Although a subset of GABAergic interneurons possesses dendritic spines (Kawaguchi and Kubota 1993; McCormick et al. 1985; Thomson et al. 1996), a high spine density is generally considered to be a good criterion to identify excitatory neurons (Connors and Gutnick 1990). The presence of numerous dendritic spines on all VGluT+/GAD− neurons examined in the present study shows that this criterion applies to morphological subgroups of spiny nonpyramidal neurons previously defined in layer VI (Prieto and Winer 1999; Tömböl 1984; Zhang and Deschênes 1997). All VGluT+/GAD− layer VI nonpyramidal neurons exhibited RS firing patterns. In the neocortex, RS firing patterns are observed not only for glutamatergic neurons, but also for several GABAergic interneuron subtypes (Cauli et al. 1997, 2000; Kawaguchi 1995). In contrast with pyramidal neurons that, based solely on their electrophysiological properties, are completely segregated from RS GABAergic interneurons following cluster analysis (Cauli et al. 2000), layer VI glutamatergic nonpyramidal neurons exhibited both pyramidal-like and interneuron-like AP firing features. Although the long AP duration of layer VI neurons was typical of glutamatergic neurons in the neocortex, their AHP amplitude and the adaptation of their AP amplitude and frequency were clearly in the range of those we previously reported for RS GABAergic interneurons under similar experimental conditions (Cauli et al. 1997, 2000; Gallopin et al. 2006). The present results thus expand the range of electrophysiological properties associated with glutamatergic neurons, which appear to overlap substantially those of GABAergic interneurons.

Anatomical and electrophysiological diversity of layer VI glutamatergic neurons

The diversity of somatodendritic morphologies of layer VI nonpyramidal glutamatergic neurons extends beyond the three morphological types we identified in our sample of biocytin-labeled nonpyramidal cells, which also comprised neurons we were unable to classify (n = 14 of 34, 41%). Layer VI further contains stellate and star pyramidal spiny neurons (Kaneko and Mizuno 1996; Tömböl 1984) whose glutamatergic nature is well established (Feldmeyer et al. 1999; Lübke et al. 2000; Schubert et al. 2003; Staiger et al. 2004), as well as classical pyramidal cells. However, in contrast with these latter morphological types that are also numerous in other cortical layers, the present glutamatergic nonpyramidal neurons seem to be preferentially localized in layer VI. This is reportedly the case for tangential pyramidal neurons (Tömböl 1984), whereas somatodendritic morphologies of inverted pyramidal and vertical fusiform cells generally correspond to GABAergic interneurons in other cortical layers. For instance, interneurons expressing nitric oxide synthase, which coexpress GAD in all cortical layers, often exhibit inverted pyramidal morphology (Gabbott and Bacon 1995; Gabbott et al. 1997). Vertical fusiform neurons (i.e., showing a vertical bipolar/bitufted dendritic arborization) have been recognized as inhibitory interneurons based on their symmetrical synapses (Somogyi and Cowey 1981). It is noteworthy, however, that the existence of bipolar/bitufted neurons forming asymmetrical synapses (Peters and Harriman 1988; Peters and Kimerer 1981) and of inverted pyramidal cells with high spine density (Miller 1988) has also been reported in layers II–V. Nonetheless, although glutamatergic nonpyramidal neurons appear to be present in all cortical layers, the present results suggest that layer VI contains a higher proportion of morphologically diverse glutamatergic nonpyramidal neurons than that of other layers.

Layer VI nonpyramidal spiny neurons form corticocortical projections but, in contrast with pyramidal neurons of the same layer, send little or no projection to the thalamus (Prieto and Winer 1999). Among the morphological types of layer VI spiny nonpyramidal neurons presently studied, all send both ipsilateral and commissural corticocortical projections (Prieto and Winer 1999). Nonetheless, ipsilateral and commissural projection neurons form distinct populations differentiated by the selective expression of Nurr1 in ipsilateral neurons (Arimatsu et al. 2003). The present observation that both Nurr1-positive and Nurr1-negative cells were collected in this study suggests that both populations of ipsilateral and commissural projection neurons are represented in our sample of VGluT+/GAD− layer VI nonpyramidal neurons.

Unsupervised analysis of electrophysiological diversity segregated layer VI glutamatergic nonpyramidal neurons into two clusters, which corresponded to earlier described phasic–tonic and tonic layer VI spiny (presumably glutamatergic) neurons (van Brederode and Snyder 1992). Consistent with this previous report, we found that electrophysiological classes did not correlate with somatodendritic morphological subgroups. Similarly, the distribution of layer VI neurons in two different clusters did not correlate with their ipsilateral or commissural projection specificity.

Subgroups of glutamatergic neocortical neurons

All subgroups of neocortical glutamatergic neurons were VGluT+/GAD−, which appears to be a reliable criterion for the identification of glutamatergic excitatory neurons in the neocortex. Comparison of layer VI nonpyramidal neurons with other subgroups indicates that, independent of their broad morphological diversity, neocortical glutamatergic neurons share a number of common properties.

At the molecular level, both pyramidal and nonpyramidal glutamatergic neurons showed a low occurrence of GABAergic interneuron markers. This demonstrates that these markers are preferentially associated with the GABAergic phenotype but not with the nonpyramidal morphology. Occasional expression of CCK, CaB, CR, NPY, and SOM, previously reported in pyramidal neurons (Burgunder and Young 3rd 1990; Celio 1990; Gallopin et al. 2006; Hill et al. 2007; Kubota et al. 1994; Morino et al. 1994; Ong et al. 1994; Schiffmann and Vanderhaeghen 1991), was also observed in our sample of nonpyramidal glutamatergic neurons. Among these markers, CCK showed the highest abundance, especially in layer VI where it was detected in 61% of the neurons, and a widespread expression in all subgroups of neocortical glutamatergic neurons. In contrast, NPY was found almost exclusively in layers V and VI glutamatergic neurons, indicating that its expression in upper neocortical layers (Hendry 1984; Kuljis and Rakic 1989a,b) is restricted to GABAergic interneurons.

At the electrophysiological level, all glutamatergic neurons collected exhibited RS firing patterns. Some glutamatergic neurons fire bursts of APs and form the class of intrinsically bursting neurons (Connors and Gutnick 1990) comprising subpopulations of pyramidal, stellate, and star pyramidal glutamatergic neurons (Connors and Gutnick 1990; Schubert et al. 2003; Staiger et al. 2004). Since this phenotype is uncovered in otherwise RS neurons following deep anesthesia (Christophe et al. 2005), it is likely that our sample of glutamatergic neurons contained a proportion of intrinsically bursting neurons, whose burst firing was hidden in our experimental conditions. Nonetheless, the four glutamatergic neuron subgroups of the present study exhibited differences in the electrophysiological properties considered and neurons within each subgroup showed substantial interindividual variability.

Analysis of the electrophysiological diversity of our entire sample of glutamatergic cells by unsupervised clustering aggregated neurons into three clusters showing morphological heterogeneity. Indeed, cluster 1 comprised one third of our layer II/III pyramidal cells sample in addition to layer V pyramidal neurons, cluster 2 contained layer II/III pyramidal neurons together with stellate and layer VI neurons, and cluster 3 contained morphologically diverse layer VI neurons. This indicates that electrophysiological similarities between neocortical glutamatergic neurons extend beyond the borders of subgroups defined by somatodendritic morphologies and layer position. A similar conclusion was drawn by Staiger et al. (2004) from comparison of layer IV star pyramidal, stellate, and pyramidal neurons. A well-documented correlate to electrophysiological heterogeneity within subgroups of glutamatergic neurons is the projection target (for layer V pyramidal cells, see reviews by Hattox and Nelson 2007; Molnar and Cheung 2006; and for layer VI neurons, see Brumberg et al. 2003; Kumar and Ohana 2008). Although layer II/III pyramidal and layer VI nonpyramidal neurons are major sources of both ipsilateral and commissural corticocortical projections (Code and Winer 1985; Innocenti 1986; Prieto and Winer 1999; Winguth and Winer 1986), spiny stellate neurons predominantly project locally (Lund 1984) and layer V pyramidal neurons form the main contingent of corticofugal projections (Molnar and Cheung 2006). Here, the grouping of layer II/III and layer V pyramidal neurons (cluster 1); of layer II/III pyramidal, stellate, and layer VI neurons (cluster 2); and of ipsilateral and contralateral layer VI neurons (cluster 3) did not correlate with projection target specificity. Thus our results suggest that electrophysiological similarities between neocortical glutamatergic neurons extend beyond anatomically defined subgroups.