To understand sensory representation in cortex, it is crucial to identify its constituent cellular components based on cell-type–specific criteria. With the identification of cell types, an important question can be addressed: to what degree does the cellular properties of neurons depend on cortical location? We tested this question using pyramidal neurons in layer 5 (L5) because of their role in providing major cortical output to subcortical targets. Recently developed transgenic mice with cell-type–specific enhanced green fluorescent protein labeling of neuronal subtypes allow reliable identification of 2 cortical cell types in L5 throughout the entire neocortex. A comprehensive investigation of anatomical and functional properties of these 2 cell types in visual and somatosensory cortex demonstrates that, with important exceptions, most properties appear to be cell-type–specific rather than dependent on cortical area. This result suggests that although cortical output neurons share a basic layout throughout the sensory cortex, fine differences in properties are tuned to the cortical area in which neurons reside.
The organization of the cortex into layers of different cell-types is a stereotypic attribute with functional consequences. For example, sensory responses in somatosensory cortex and visual cortex (VC) are layer and cell-type specific (Gilbert and Wiesel 1979; Armstrong-James et al. 1992; Brecht and Sakmann 2002; Martinez et al. 2005; de Kock et al. 2007) and a correlation between morphology and response property can be drawn. In somatosensory (vibrissal) cortex (BC), in rodents, the thick-tufted and the slender-tufted pyramidal neurons in layer 5 (L5) respond differently to tactile (whisker deflection) stimuli (Manns et al. 2004; de Kock et al. 2007). The functional difference between slender- and thick-tufted pyramidal neurons raises several questions. First, what are the defining characteristics of these neurons in terms of cell-type–specific properties in dendrite morphology, intrinsic excitability, axon projection target, gene expression, and properties of synaptic transmission? Second, are these properties unique for a cortical area or retained in different cortical areas? Classification of L5 pyramidal neurons has relied on a combination of soma–dendrite morphology, intrinsic electrophysiological properties, gene expression pattern, as well as projection targets (Hevner et al. 2003; Molnar and Cheung 2006; Morishima and Kawaguchi 2006; Nelson et al. 2006; Sugino et al. 2006; Le Be et al. 2007; Watakabe et al. 2007). A scheme that has emerged is the correlation between soma–dendrite morphology and projection targets (Hübener and Bolz 1988; Hübener et al. 1990; Larkman and Mason 1990; Mason and Larkman 1990; Hattox and Nelson 2007). Uncertainties, however, of cell identities remain when correlating parameters from individual studies. The identification of genetically labeled neurons throughout the neocortex enabled in the present work the combined study of many parameters and their dependency on the sensory modality.
We have characterized 2 populations of L5 pyramidal neurons genetically labeled with enhanced green fluorescent protein (EGFP) and compared each cell type in somatosensory cortex with the corresponding cell type in VC. We used 2 mouse lines for the 2 L5 cell types expressing EGFP under the control of either a promoter for a transcription factor (ets variant gene [etv] 1) or the promoter for a glycosyltransferase (glycosyltransferase 25 domain containing [glt] 2). Each is specifically expressed in a subpopulation of the slender- and thick-tufted pyramidal neurons, respectively.
We show that in both somatosensory cortex and VC, cell-type–specific properties define 2 populations of tufted L5 pyramidal neurons. Each subpopulation retains its basic properties between cortices. In both cortices, the 2 different populations can be clearly distinguished based on anatomical and functional parameters. However, when comparing a given cell-type between the 2 cortices, some parameters are characteristically tuned to the cortical area. These differences suggest that tactile and visual stimuli are processed by similar local neuronal circuits that, however, show variations depending on cortical region.
Materials and Methods
Experiments were performed on Glt25d2 bacterial artificial chromosome (BAC)-EGFP transgenic mice and Etv1 BAC-EGFP transgenic mice obtained from the GENSAT project. Information on the genes and expression data at the GENSAT Web site (http://www.gensat.org). See also Gong et al. (2007) and Doyle et al. (2008) for a brief description of the Etv1and Glt25d2 BAC-EGFP lines and other BAC lines. Experiments were performed on mice with both parents BAC-EGFP mice and mice with one parent BAC-EGFP and one wild-type C57Bl/6. Experiments were conducted in accordance with the German animal welfare guidelines.
Stereotaxic injections, using isofluorane anesthesia, were done at postnatal day 60–120. Cholera toxin subunit B (Molecular Probes) conjugated to Alexa 555 (pseudo color red) or Alexa 647 (pseudo color blue) for retrograde labeling was mixed with a solution containing adeno-associated viral particles (AAVs) encoding for dTomato (AAV-dTomato) for anterograde labeling. In total, 50- to 100-nL solution, containing tracers, was slowly injected into target sites (pontine nucleus, caudate putamen, and posterior medial thalamic nucleus [POm]) with the following coordinates (mm) relative to bregma, midline, and dura: caudate putamen: 0.8, 1.8, −3.6; pontine nucleus: −3.6, 0.55, −5.3; and POm: −2.0, 1.15, −2.9. To avoid inadvertent diffusion of tracer from the intended target area, the tracer injection volume was kept to a minimum. The resulting fraction of retrogradely labeled cells is, thus, unlikely to reflect all cortical cells projecting into the target. The relevant comparison is the relative fraction of retrogradely labeled etv- versus glt-pyramids from injection into different targets and the depth distribution of EGFP and retrogradely labeled cells.
Coronal slices containing primary somatosensory (barrel) cortex or VC were prepared by placing the anterior part of the brain on a 10-deg slope and then cutting perpendicular to the brain surface, discarding the tissue caudal to the cut. The brain was subsequently glued with the cut surface facing down, and 300-μm thick slices were cut. In a few experiments, parasagittal slices (300 μm) containing VC were also used. Slices were maintained at room temperature before recording. All experiments were performed close to physiological temperatures (32–34 °C).
Patch pipettes for whole-cell recordings were filled with (in mM) 105 K-gluconate, 30 KCl, 10 phosphocreatine-Na2, 10 N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (HEPES), 4 ATP-Mg, 0.3 GTP, with pH adjusted to 7.3 with KOH (osmolarity 300 mOsm). Biocytin (2 mg/ml; FLUKA) was included in the pipette to allow the morphology of the neurons to be reconstructed in the fixed tissue. For cell-attached stimulation, the pipette contained the following solution (in mM): 105 Na-gluconate, 30 NaCl, 10 HEPES, 10 phosphocreatine, 4 ATP-Mg, and 0.3 GTP (adjusted to pH 7.3 with NaOH). The extracellular solution contained (in mM): 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 25 glucose, 2 CaCl2, and 1 MgCl2, continuously bubbled with 95% O2 and 5% CO2. Membrane potential was not corrected for junction potential. Dual whole-cell recordings to investigate synaptic connectivity 1) in barrel cortex (BC) were made on 19 Etv1 BAC-EGFP transgenic mice with an average age of 22 days (range: 18–25) days and on 14 Glt25d2 BAC-EGFP transgenic mice, average age 25 days (range: 16–39 days) 2) in VC (including both primary and secondary VC) were performed on 7 Etv1 BAC-EGFP transgenic mice, average age 23 days (range: 19–25 days) and 2 Glt25d2 BAC-EGFP transgenic mice (age 21 and 22 days). One pipette was used to obtain whole-cell recordings from one cell and the second pipette was used to search for synaptically connected neurons. The search pipette was used for cell-attached stimulation or whole-cell recording. Only cell pairs located within 20 μm of one another in the z-plane were tested. Cell pairs with no synaptic connection in either direction were counted as 2 unconnected pairs, pairs with a unidirectional connection were counted as one connected pair and one unconnected pair, and pairs with a bidirectional connection were counted as 2 connected pairs. The horizontal distance (x-axis) between neurons was measured from the center of the soma, with the vertical y-axis aligned to the apical dendrite. Recordings were made with an Axoclamp2B amplifier (Axon Instruments, Union City, CA), filtered at 3 kHz and sampled at 10 kHz using custom written software (Igor; WaveMetrics, Lake Oswego, OR).
Slices were fixed in 100 mM PBS containing 4% paraformaldehyde and kept at 4 °C for 1–2 days. Slices were stained for cytochrome oxidase, to visualize barrels, and biocytin to visualize cell morphology. Biocytin-labeled neurons were visualized using the avidin:biotinylated horseradish peroxidase complex (Vectastain ABC kit, Vector Laboratories, CA). The location of the neurons within the BC could, thus, be confirmed, and neurons were reconstructed in 3D with the Neurolucida software (MicroBrightField, Colchester, VT) using an Olympus (BX 51) or Zeiss microscope with an oil immersion ×100/1.4 numerical aperture (NA) objective.
Quantification of Dendrite Morphology, Intrinsic Membrane Properties, and Cell Counting
Apical tuft dendrite was defined in BC as the dendritic profiles in the upper 10% of the pia–white matter distance and upper 15% of the pia–white matter distance in VC. This corresponds approximately to layer 1. Apical dendrite tuft width was the maximum horizontal width. Primary obliques were dendrites bifurcating from the main apical dendrite and located between the soma and the major apical dendrite bifurcation. “Nodes” are bifurcation or trifurcation points. Soma and dendrite parameters were calculated with Neurolucida Explorer (MicroBrightField). Fluorescently labeled cells were counted manually using the AMIRA (Visage Imaging, Germany version 4.0; Mercury Computer Systems) software. Threshold current was the minimum current to elicit at least one action potential at resting membrane potential. Frequency adaptation was calculated by measuring the average time interval between successive spikes elicited using three 1-s pulses with current in the range 200–400 pA, in steps of 20 pA. The average interspike time intervals (ISIs), excluding the 2 first ISIs, was normalized to the average of the third ISI and plotted as a function of the ISI number (ISI #). Subsequently, a linear regression line was fitted. The slope of the line is taken as the “adaptation index.” An adaptation index of “10,” thus, means that on average the time interval increases with 10 percentage points between successive spikes.
EGFP-labeled cells were identified in the living slice using two-photon microscopy. Two-photon excitation was made by a Ti:Sa-Laser (MIRA 900F; Coherent, Santa Clara, CA) set at 820–840 nm and pumped by a solid-state laser (Verdi 5W; Coherent). Scanning was performed using a resonant scanning unit (TCS-SP2RS; Leica Microsystems, Mannheim, Germany) mounted on an upright microscope (DMLFS; Leica Microsystems) equipped with a ×40 objective (HCX APO, ×40, NA 0.8; Leica). A dichroic mirror (560DCXR) split the fluorescence to one detector (bandpass filter HQ525/50M) for green fluorescence from EGFP-expressing L5 pyramidal neurons and to one detector (bandpass filter, HQ610/75M) recording red fluorescence from Alexa Fluor 594. Online superposition of fluorescence and infrared contrast-enhanced image allowed targeted recordings from fluorescently labeled neurons.
Two-tailed unpaired t-tests, and one-way analysis of variance (ANOVA) using Bonferroni posttests, were done with GraphPad Prism 4 (GraphPad Software, CA). Significant differences in the morphological data and the intrinsic properties were determined with one-way ANOVA and Bonferroni posttests unless otherwise stated. Data are reported as mean ± standard deviation (SD), unless stated otherwise.
Numerical Analysis of Cell Densities in a Barrel Column
The distribution of L5 etv-expressing pyramidal neurons (etv-pyramids) and L5 glt-expressing pyramidal neurons (glt-pyramids) was determined by manual cell counting of all EGFP-labeled cells in 2 barrel columns in one brain slice from each of 3 etv-mice and 2 glt-mice (age 25–28 days). No correction was done for tissue shrinkage. GAD67 and NeuN immunohistochemistry was used to estimate column and layer borders. A detailed description of the immunocytochemistry and analytical tools developed for the numerical analysis of cells in a barrel column will be presented elsewhere (Meyer HS, Wimmer VC, Schwarz D, Sakmann B, and Helmstaedter M, unpublished data).
Soma Depth Location and Dendrite Morphology of 2 Genotypically Defined L5 Pyramidal Neurons in Primary Somatosensory (Barrel) Cortex
Genetic labeling of cortical cells using the BAC method (Gong et al. 2002, 2003, 2007; Doyle et al. 2008) has generated several lines of transgenic mice with fluorescent labeling of potentially specific classes of neurons. To investigate subpopulations of L5 pyramidal neurons in different cortical areas, we have used 2 BAC transgenic mouse lines where expression of Glt25d2 (glt) or Etv1 (etv) was reported with EGFP expression. Expression of these 2 genes is shown to identify L5 pyramidal neurons in the primary somatosensory (barrel) cortex (Fig. 1A,B), VC (Fig. 3), and other neocortical areas (data not shown). For each mouse line, an average density map was calculated showing the 2D soma distribution in 2 neighboring barrel columns (Fig. 1C). L5 etv-pyramids were located predominately in L5A (Fig. 1C; median depth from pia 610 μm) and glt-pyramids in L5B (Fig. 1C; median depth from pia 690 μm). In the horizontal column axis, the highest density was located in the center of the barrel columns (Fig. 1C). In somatosensory cortex, etv- and glt-pyramids are, thus, expressed in 2 distinct sublayers. The peak cell densities are approximately 80 μm apart.
A profile in 1D representing the ratio of etv- and glt-pyramids to all excitatory neurons in a barrel column shows that at the peak cell density, the etv- and glt-pyramids represent approximately 55% of the excitatory neurons (Fig. 1D). The fraction of etv-pyramids and glt-pyramids to the total number of neurons was estimated to 29.0 ± 3.1% in L5A (etv) and 12.3 ± 3.5% in L5B (glt). The average neuron density in each layer was higher (Supplementary Table 1) than previously reported for mouse somatosensory cortex (White and Peters 1993).
In the acute brain slice preparation, EGFP-labeled etv- and glt-pyramids were filled and subsequently stained for biocytin, and their dendrite morphology was reconstructed and quantified. Figure 2 (and Supplementary Fig. 1) illustrates examples of reconstructed etv- (Fig. 2A) and glt-pyramids (Fig. 2B). In both populations, the apical dendrites extended to layer 1. Compared with the glt-pyramids, the etv-pyramids have a narrower apical dendrite tuft width (etv-pyramids: 157 ± 54 μm, n = 29; glt-pyramids 356 ± 84 μm, n = 19) and had fewer primary apical oblique dendrites (etv-pyramids: 3.7 ± 2.2 primary branches; glt-pyramids: 8.9 ± 2.6 primary branches). The apical dendrite diameter at the base was for glt-pyramids 4.3 ± 1.3 μm, which was wider (P < 0.0001, unpaired t-test) than for etv-pyramids (2.7 ± 1.2 μm). Furthermore, for glt-pyramids, the average total apical dendrite length (4990 ± 1052 μm) was longer than the average total basal dendrite length (2585 ± 1038 μm; Fig. 2C). In contrast, for etv-pyramids, the average total apical dendrite length (2363 ± 823 μm) was shorter (P = 0.0044, unpaired t-test) than the average total basal dendrite length (3006 ± 994 μm; Fig. 2C). The apical tuft dendrite is the most obvious discriminating morphological parameter for the 2 cell-types with etv-pyramids having a shorter total length, fewer nodes, and a smaller width compared with the glt-pyramids. For the basal dendrite, etv-pyramids have more nodes compared with glt-pyramids (Table 1). glt- and etv-pyramids are, thus, morphologically different cell types corresponding to thick-tufted and slender-tufted pyramidal neurons, respectively (Markram et al. 1997; Schubert et al. 2001, 2006; Larsen and Callaway 2006; de Kock et al. 2007).
Soma Depth Location and Dendrite Morphology of etv- and glt-Pyramids in VC
Specific EGFP labeling of 2 pyramid populations in L5 was also observed in VC (Fig. 3). In contrast to the somatosensory cortex, the somata of etv- and glt-pyramids are more intermingled in L5 (Compare Fig. 1 and Fig. 3). Intermingling of the soma position in L5 in VC was further confirmed by crossing etv-mice with glt-mice (Fig. 4). In the glt/etv-hybrid mice, the 2 cell types can be separated because glt-pyramids have EGFP fluorescence coupled to ribosomal proteins in the nucleolus (visible as a bright spot in the soma; Fig. 4B), whereas etv-pyramids have a homogenously labeled soma and brighter EGFP-labeled apical dendrites. In the glt/etv-hybrid mice, the layer separation in BC (Fig. 4A,D) and intermingling in VC was evident (Fig. 4C,E). This raises the possibility that synaptic inputs to the basal dendrites of these 2 cell types are coming from different sources in BC, whereas in VC, glt- and etv-pyramids share inputs (e.g., from the pulvinar).
Similar to BC for glt-pyramids (Fig. 3E), the total apical dendrite length (3867 ± 1288 μm, n = 15) was significantly longer than the total basal dendrite length (2640 ± 792), and for the etv-pyramids (Fig. 3E), the total apical dendrite length (2264 ± 647 μm, n = 21) was shorter than the total basal dendrite length (2730 ± 773. Compared with BC, the etv-pyramids in VC had on average the same number (4) of oblique dendrites (P = 0.39, unpaired t-test). glt-Pyramids in VC and BC both had on average 9 oblique dendrites. etv-Pyramids appear as slender-tufted (apical dendrite tuft width: 182 ± 50 μm; apical dendrite diameter: 2.7 ± 0.7 μm) and glt-pyramids as thick-tufted pyramidal neurons (apical dendrite tuft width: 290 ± 92 μm; apical dendrite diameter: 2.9 ± 1.1 μm). Similar to BC, the dendritic tuft is the most obvious discriminator between the 2 cell types: etv-pyramids have a shorter total length and narrower width compared with glt-pyramid tufts. In contrast to BC, the basal dendrites of glt- and etv-pyramids are indistinguishable for all parameters tested here (Table 1). Taken together, the morphological properties show that glt- and etv-pyramids are clearly 2 different cell types in both cortical areas.
Thalamocortical Innervation Fields
To determine the thalamic inputs to etv- and glt-pyramids in BC, we expressed dTomato via AAV injections into POm and ventral posterior medial thalamic nucleus (VPM) in order to map thalamocortical innervation fields. POm axons densely overlapped with basal dendrites and soma of etv-pyramids (Fig. 5A). Potentially also the apical tufts of both etv- and glt-pyramids were densely innervated in L1 (Fig. 5B). VPM axons were predominately located in layer 4 and lower L5B. They more sparsely overlapped with basal dendrites and somata of glt-pyramids (Fig. 5C). The preferential innervation of different populations of L5 pyramidal neurons by VPM and POm axons confirms that the different thalamic nuclei can be part of separate thalamocortical circuits (Ahissar and Kleinfeld, 2003; Bureau et al. 2006).
Projection Fields of Somatosensory L5 etv- and glt-Pyramids
The knowledge of cortical long-range projections to subcortical targets is relevant for elucidating the possible function of corticofugal projections in, for example, driving behaviors that depend on sensory input.
The subcortical targets of the 2 populations of pyramidal neurons in BC were determined by injections of retrograde tracers (in 5 animals) into striatum, POm, and pons (Fig. 6). Glt-pyramids projected to ipsilateral pons (Fig. 6A) and POm (data not shown) but not to striatum (Fig. 6A). The depth distribution of glt-pyramids overlapped with retrogradely labeled corticopontine cells but was well separated from corticostriatal cells (Fig. 6E). Glt-pyramids (9 of 9 cells) at the level of the white matter had only an ipsilaterally projecting main axon, thus, projecting to thalamus and pons and also likely to the spinal cord and trigeminal nucleus (Wise and Jones 1977). Etv-pyramids projected to ipsilateral striatum, and to some extent to contralateral striatum (1%, 2 of 156 cells), but not to pons (Fig. 6C) or POm (data not shown). In 5 etv-pyramids, the axon could be traced to the white matter, where 4 of 5 cells had a bifurcating main axon (Figs. 2 and 8), indicating both an ipsilateral and contralateral projection. The majority of etv-pyramids in BC, thus, appear to project both ipsilaterally and contralaterally. The retrograde labeling indicated, however, that a minority of the etv-pyramids project to contralateral striatum and, thus, the contralateral projection, presumably, targets mainly contralateral cortex. The depth distribution of etv-pyramids overlapped with retrogradely labeled corticostriatal cells but was largely separated from corticopontine cells (Fig. 6F); etv is, thus, inferred to be expressed in a subpopulation of L5A slender-tufted pyramidal neurons projecting to the ipsilateral striatum and contralateral cortex, with a minority projecting also to contralateral striatum and with no projections to pons. The target specificity of etv- and glt-pyramids is furthermore indicated by EGFP labeling in striatum in etv-mice and in pons in glt-mice (Supplementary Fig. 3).
Similarly, in VC, the glt-pyramids project to pons (Fig. 6B), but not striatum (Fig. 6B), and etv-pyramids project to striatum (Fig. 6D) but not to pons (Fig. 6D). Compared, however, with BC, there were relatively fewer retrogradely labeled etv-pyramids (Fig. 6C,D), and the depth distribution of etv-pyramids did not overlap to the same degree with retrogradely labeled corticostriatal cells (Fig. 6H). This difference could indicate that, in VC, the etv-pyramids are a relatively small subpopulation of the corticostriatal cells. We conclude that in VC and BC slender-tufted L5 etv-pyramids are corticostriatal and that the majority of thick-tufted L5 glt-pyramids are corticopontine and thalamic projection neurons.
In BC, the depth distribution of retrogradely labeled corticostriatal and corticopontine cells was well separated (Fig. 6E,F), whereas in VC, there was a larger overlap (Fig. 6G,H). This difference indicates that, at least in some parts of the VC, the corticostriatal pyramidal neurons are to a higher degree than in BC located at the same depth as corticopontine cells. This finding supports the observation that somata of etv- and glt-pyramids are less well separated in VC (Fig. 4). This is, thus, potentially an anatomical substrate for a less specific thalamocortical innervation of infragranular neurons.
The projection patterns of etv- and glt-pyramids are in agreement with those inferred from EGFP antibody staining (http://www.gensat.org). The layer localization of retrogradely labeled corticostriatal and corticopontine cells is similar to previously reported results (Wise and Jones 1977; Wiesendanger and Wiesendanger 1982; Akintunde and Buxton 1992; Jinno and Kosaka 2004; Hattox and Nelson 2007).
Intrinsic Membrane Properties of etv- and glt-Pyramids in BC and VC
Cell-type and area specificity of intrinsic membrane properties was then investigated. The resting membrane potential was −68 ± 1 mV for etv-pyramids in both BC and VC. glt Neurons had a resting membrane potential of −64 ± 1 mV in BC and −66 ± 1mV in VC. Thus, in contrast to VC, in BC, etv- and glt-pyramids have significantly different membrane potentials.
The input resistance (MOhm) for etv-pyramids was cortex specific with 86 ± 4 (n = 45) in BC and 110 ± 4 (n = 33) in VC (unpaired two-tailed t-test, P < 0.0001). Furthermore, input resistances of etv- and glt-pyramids were indistinguishable in VC but were different in BC (VC: etv 110 ± 4 MOhm and glt 116 ± 13 MOhm; BC: etv 86 ± 4 MOhm and glt 53 ± 3 MOhm). Threshold current was dependent on the cortical area for the glt-pyramids but not for etv-pyramids and was furthermore not significantly different between glt-pyramids and etv-pyramids in both areas (in pA) etv-pyramids 158 ± 9 in BC and 131 ± 12 in VC and 2) glt-pyramids 158 ± 11 in BC and 103 ± 5 in VC).
Spike frequency adaptation was quantified by measuring the average time interval between successive spikes to calculate an “adaptation index” (see Materials and Methods). Adaptation index was for: 1) etv-pyramids 23 ± 27 in BC and 17 ± 17 in VC. The relatively high adaptation index indicates that the interspike interval increased between successive spikes (Fig. 7A,B; Supplementary Fig. 4A,B) and 2) glt-pyramids 1 ± 2 in BC and 2 ± 2 in VC, indicating a close to constant interspike interval (Fig. 7C,D; Supplementary Fig. 4C,D); etv- and glt-pyramids can, thus, be referred to as adapting and nonadapting, respectively. Among the glt-pyramids, approximately 50% and 20% of the cells were intrinsically bursting in BC and VC, respectively (Fig. 7E,F; Supplementary Fig. 4E,F). glt-Pyramids responding with bursts of action potential (Fig. 7E,F; Supplementary Fig. 4E,F) were not included in the calculation of the adaptation index.
A comparison between etv- and glt-pyramids showed that etv-pyramids in BC were more hyperpolarized and had a higher input resistance than glt-pyramids. The adaptation index was higher for etv-pyramids in both areas. Threshold current was cortex specific only for glt-pyramids, but input resistance was cortex specific for both cell types. Intrinsic electrical properties are subject to both, region, and cell-type specificity: etv- and glt-pyramids have cell-type–specific properties in both sensory cortices, concomitantly showing a region specificity.
Monosynaptic Connections between etv-pyramids and glt-pyramids
Paired whole-cell recordings were made to investigate cell-type and area dependency of synaptic properties. A monosynaptic connection between etv-pyramids in BC (Fig. 8, Supplementary Fig. 5A) was found in 13 of 76 recorded pairs (17%; 9 unidirectional pairs and 2 bidirectional pairs). The average monosynaptic excitatory postsynaptic potential (EPSP) amplitude was 0.39 ± 0.26 mV (range: 0.09–0.82 mV, n = 13 pairs), with a coefficient of variation (CV) of 0.46 ± 0.11. For EPSP amplitudes larger than 0.23 mV, the failure rate was low (range: 0–3%, n = 6), but increased for smaller EPSP amplitudes (failure range: 8–28%, n = 3). The paired-pulse ratio at an interstimulus interval of 100 ms was 0.91 ± 0.19. The connection was depressing in the 10- to 40-Hz stimulation range (Fig. 8C,D).
Synaptic connectivity between glt-pyramids in BC (Supplementary Fig. 6) was less extensively investigated, but the data suggest that these cells are connected with a lower connection probability (6%; 2 connected pairs of 35) compared with etv-pyramids.
To compare synaptic properties within a pyramidal cell population in different cortical areas, we also made whole-cell recordings from pairs of etv-pyramids in the VC (Fig. 8E–G, Supplementary Fig. 5B). A monosynaptic connection was found in 6 of 33 (18%) recorded pairs. The average monosynaptic amplitude was 0.82 ± 0.94 mV (range: 0.08–2.3 mV, n = 6 pairs), with a CV of 0.43 ± 0.20. For EPSP amplitudes larger than 0.23 mV, the failure rate was low (range 0–3.8%, n = 4) but increased for smaller EPSP amplitudes (failure: 9 and 18%, n = 2). Similar to the BC, the etv–etv pyramid connection was depressing in the 10- to 40-Hz stimulation range (Fig. 8G). The paired-pulse ratio at an interstimulus interval of 100 ms was 0.80 ± 0.36. The similar synaptic properties and connectivity (chi-square test 0.89) for connected etv-pyramids in BC and VC suggest that within a population of cells synaptic properties may be retained across different cortical areas. Furthermore, corticostriatal etv-pyramids appeared more interconnected than corticopontine/thalamic glt-pyramids.
Comparison between Cortices
The majority of properties of etv- and glt-pyramids were similar in VC compared with the properties of the corresponding cell type in BC, although with some notable differences outlined below. A given cell type shows variations in dendrite geometry depending on the cortical region. Compared with BC, etv cells in VC have fewer basal dendrite nodes, but the same total length. The basal dendrite tree of glt-pyramids is similar in BC and VC, but in VC, the tuft dendrite is shorter and has fewer nodes (Table 1). The soma depth was also characteristic for the cortical region with a clear separation of glt- and etv-pyramids in BC and the tendency to share the layer in VC. Taken together, the anatomical data suggest that etv- and glt-pyramids are more similar to each other in VC and more distinct in BC. Differences between the cortical regions could also be observed for intrinsic electrical properties. Both cell types had a lower input resistance and needed larger currents to reach spiking threshold in BC. In summary, the differences for a given cell type in different areas suggests that a basic cell layout is fine tuned to the cortical area and by the sensory system, respectively.
Untangling the diversity of cortical cell types requires the combination of genetic targeting, quantification of cell anatomy and physiology, and eventually the molecular characterization of the different cell types (Luo et al. 2008). We have undertaken this approach to provide a detailed histological and physiological characterization of L5 neocortical cell types using BAC-transgenic mice. This allows evaluation of the extent of conservation of cellular properties in different cortical areas and the influence of the sensory modality on L5 neurons. In both VC and somatosensory cortex, the Etv1 line had selective EGFP expression in “slender-tufted” L5 pyramidal neurons, and the Glt25d2 line had selective EGFP-expression in “thick-tufted” L5 pyramidal neurons. Etv-pyramids in both somatosensory cortex and VC share characteristic properties compared with glt-pyramids, being more hyperpolarized, having a higher input resistance, an adapting spiking pattern, and being more densely interconnected. In BC, the slender-tufted corticostriatal etv-pyramids were predominately found in upper L5 and the thick-tufted corticopontine/thalamic glt-pyramids in lower L5. In contrast, in VC, the 2 populations were not well separated, suggesting a more common thalamic innervation of these cell types in VC but not in BC. A cell classification in L5 based on soma–dendrite morphology, intrinsic electrical excitability, properties of synaptic transmission and gene expression can, thus, differentiate between cells with different projection targets. The 2 genetically defined L5 pyramidal neuron types were found to have in general conserved properties in BC and VC, with the notable anatomical exceptions and, thus, a dependency on the sensory modality (touch or vision), in apical dendrite shape and layer organization.
Subpopulations of Pyramidal Neurons in L5
In BC, 16% of all neurons located in L5 were etv-pyramids and 9% were glt-pyramids. Within the layer, however, the local density of labeled pyramidal neurons was significantly higher (Fig. 1C,D). By means of retrograde labeling, the fraction of corticostriatal or corticothalamic cells in L5 of mouse somatosensory cortex has been estimated to 26% and 8%, respectively, of the total number of neurons (Hattox and Nelson 2007). Because fewer etv-pyramids are labeled than this estimated total fraction of corticostriatal cells, it is likely that Etv1 labels a subpopulation of corticostriatal (slender-tufted) pyramidal neurons. Because etv-pyramids have callosally projecting axons, they are likely to be part of the intratelencephalically projecting type of corticostriatal neurons (Wilson 1987; Reiner et al. 2003). The thalamic POm projection of glt-pyramids is, most likely, a collateral of the main axon projecting predominately to pons and POm, but also to other brainstem nuclei and the spinal cord (Veinante et al. 2000), suggesting an anatomical substrate for efference copy of motor instructions to POm.
In comparison to related work on the classification of pyramidal neurons in L5, the dendrite morphology and spike frequency adaptation of etv- and glt-pyramids are similar to those found in retrogradely labeled corticostriatal and corticothalamic pyramidal neurons, respectively (Hattox and Nelson 2007). Furthermore, potassium channel Kv3.1-expressing L5 pyramidal neurons (Akemann et al. 2004) which, with respect to their dendrite morphology and spike frequency adaptation, are similar to the etv-pyramids are described here, with one noticeably difference. The Kv3.1-expressing cells made local intracortical connections only and did not have axons projecting to subcortical targets. Taken together with our results, this comparison suggests that slender-tufted pyramidal neurons in L5 are a very heterogeneous group of neurons. It may consist of at least 2 subpopulations, corticocortical and corticofugal projecting neurons (Ivy and Killackey 1981; Hattox and Nelson 2007; Le Be et al. 2007). According to the anatomical classification by Tsiola et al. (2003), the L5 glt-pyramids in VC appear to fall in the class denoted “subgroup 1A” and the L5 etv-pyramids in the class denoted “subgroup 1B.” Pyramidal cells in L5 that are organized in cell clusters (Krieger et al. 2007) are similar to the glt-pyramids; these clustered pyramidal neurons were located in lower L5 and projected to POm and pons. The less significant bundling of apical dendrites originating from glt-pyramids suggests, however, at most a partial overlap between the cell types. Notably, the glt- and etv-pyramids project to pons/POm or striatum, respectively, but not to both areas. There were, however, non–EGFP-labeled cells in L5 that were double labeled by tracer injections into pons and striatum (data not shown; compare Levesque et al. 1996). A possible difference in the cell-type–specific properties of etv- and glt-pyramids in other cortices, notably the motor cortex (Miller et al. 2008), remains to be investigated. Genetically defined cell types, thus, enable a detailed investigation of epigenetic factors controlling the formation of local cortical circuits in different cortical areas.
Using Genetic Markers to Understand the Contribution of Different Cell Types to the Animal's Behavioral Repertoire
Genetic markers that define specific cell-types can be used to selectively manipulate cortical circuits. This will aid to understand the contribution of a particular pyramidal cell type in driving behaviors. Conditional silencing, for example, of pyramidal neurons in L5 could unravel their contribution to learning and execution of behaviors. The transcription factor Etv1 (also referred to as ER81) has been shown to participate in neurogenesis in the mouse olfactory bulb (Stenman et al. 2003) and circuit formation in the spinal cord (Arber et al. 2000). Taken together with the present results, this indicates that etv-dependent proteins can participate in the formation of neuronal circuits in different areas of the central nervous system.
The restricted expression of etv in L5 is consistent with previous studies (Xu et al. 2000; Sugitani et al. 2002; Beggs et al. 2003; Hasegawa et al. 2004; Watakabe et al. 2007) and with the in situ hybridization data documented in the Allen Brain Atlas (http://www.brain-map.org). However, expression differs in essential aspects from previous reports of dendrite morphology, functional properties, and projection targets of etv-expressing cells. Yoneshima et al. (2006) reported that ER81 was expressed in non-tufted cells, was equally distributed in L5, and was expressed in neurons projecting to the spinal cord or superior colliculus, whereas Sugino et al. (2006) report that ER81 mRNA was detected in non-adapting L5 and L6 cells. The reasons for these discrepancies remain unclear. We note that the restricted expression to L5 reported for the Etv1 EGFP-reporter mice used in the present study, and that reported for an independently generated Etv1 BAC Cre recombinase driver line (Gong et al. 2007) agrees well.
Cell-Type Specific Properties in Somatosensory and VC
In both somatosensory cortex and VC, etv-pyramids were distinct from glt-pyramids on the basis of the soma–dendrite morphology, intrinsic membrane properties, and projection target. Furthermore, the properties of synaptic transmission between etv-pyramids in the 2 cortices were similar, with a connectivity ratio of approximately 20%. The connectivity for glt-pyramids in BC was estimated, based only on a limited sample, to 6% which is similar to the 10% estimated for clustered thick-tufted pyramidal neurons in L5 (Krieger et al. 2007). In rat VC, the connectivity ratio for thick-tufted L5 pyramidal neurons has also been estimated to approximately 10% (Song et al. 2005). A similar difference in connection probability for L5 pyramidal neurons projecting to different targets has been observed in rat neocortex (Morishima and Kawaguchi 2006; Le Be et al. 2007).
Comparing etv and glt pyramidal neurons, we find that they are clearly distinguishable based on morphology, intrinsic, and synaptic properties and long-range circuitry. However, when comparing individual cell types in BC and VC, we find that the difference between cell types (etv vs. glt) is much more pronounced than the differences of these cell types between different cortical areas. The individual cell types are, thus, influenced by the cortical region they belong to, but the overall layout of etv- and glt-pyramids is retained. Our results, thus, raise the intriguing possibility that genes can be identified that characterize cell types which not only share morphology but are also incorporated in comparable local circuits with a stereotypic layout in different cortical areas. It can, thus, be concluded that for a genetically defined neuron population, cell-type–specific properties such as soma–dendrite morphology, intrinsic electrical excitability, projection target, and properties of synaptic transmission are not necessarily unique for a cortical area. They may be retained, with some noted modification, between sensory cortical areas, and possibly other neocortical areas.
We thank Randy Bruno and Gilad Silberberg for comments on the manuscript, Rolf Sprengel for AAV-dTomato, and Marlies Kaiser for excellent technical assistance. This work was supported by Åke Wibergs stiftelse to P.K. Author contributions: Conceived and designed the experiments: A.G., H.S.M., B.S., and P.K. Performed the experiments: A.G., H.S.M., and P.K. Analyzed the data: A.G., H.S.M., B.S., and P.K. Contributed reagents/materials/analysis tools: H.S.M., E.S., and N.H. Wrote the paper: A.G., B.S., and P.K. Conflict of Interest: None declared.