| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1 Division of Neurology Research, St. Elizabeth's Medical Center, Tufts University, Boston, MA 02135, USA
2 Brain and Mind Institute, EPFL, Lausanne, 1015, Switzerland
3 Department of Neurobiology, Weizmann Institute of Science, Rehovot 76100, Israel
4 Cold Spring Harbour Laboratory, 1 Bungtown Road, Cold Spring Harbour, NY 11724, USA
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
1G, Kv3.4, Kv4.2, Kv1.1 and HCN2. In summary, this study provides the first detailed analysis of the anatomical, electrophysiological and molecular properties of Martinotti cells located in different neocortical layers. It is proposed that MCs are crucial interneurones for feedback inhibition in and between neocortical layers and columns.
(Received 5 August 2004;
accepted after revision 24 August 2004;
first published online 26 August 2004)
Corresponding author H. Markram: Brain and Mind Institute, EPFL, Lausanne, 1015 Switzerland. Email: henry.markram{at}epfl.ch
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
While some interneurones largely restrict their axons mostly to the layer in which the soma is located (small basket cells, neurogliaform cells, chandelier cells), many interneurones also extend their axons across multiple layers (large basket cells, Martinotti, bitufted, double bouquet, and bipolar cells) (Fairen et al. 1984). Very little is known, however, about the selectivity of interlayer targeting. Most interneurone types also restrict their axons to the dimensions of a cortical column with two major exceptions; large basket cells (LBCs) and Martinotti cells (MCs) (Fairen et al. 1984; Wang et al. 2002). LBCs typically provide long horizontal axons that cross multiple columns in the same layer that the soma is located, while MCs send their axons up to layer I to form an axonal arborization that crosses multiple columns in layer I to contact the distal tuft dendrites of pyramidal cells. Because of this distal dendritic innervation, the nature of MC-mediated inhibition is expected to differ substantially from that of LBCs.
The Martinotti cell was first reported by Carlo Martinotti in 1889 (Martinotti, 1889) and first named by Ramon y Cajal (Ramon y Cajal, 1891). This cell type, first found in layer V, was defined according to its ascending axonal collaterals reaching layer I where they ramify to form a fanlike spread of axonal collaterals (Fairen et al. 1984), with axons bearing spine-like boutons (Marin-Padilla, 1984). Subsequent studies revealed that MCs are present ubiquitously in different layers and areas of the neocortex as well as different ages and in many species. These include: mouse somatosensory (Lorente de No, 1922) and visual cortices (Valverde, 1976); rat visual (1, 3 and 4 months old; Ruiz-Marcos & Valverde, 1970), cingulate (1–2 months; Vogt & Peters, 1981) and frontal cortices (Kawaguchi & Kubota, 1998); the visual cortex of rabbit (Shkol'nik-Yarros, 1971), alticola (Hedlich & Werner, 1990), microtus brandti (Hilbig et al. 1991) and guinea-pig (Hedlich & Werner, 1986); several neocortical areas of microtus agrestis, hamster, hedgehog, the dwarf bat, dog, cow and sheep (Ferrer et al. 1986a,b); several neocortical areas in the mature cat (Marin-Padilla, 1972) including in the visual cortex (O'Leary, 1941); in the adult monkey pre-frontal cortex (Gabbott & Bacon, 1996); and also in human somatosensory (Ramon y Cajal, 1911), motor (Marin-Padilla, 1970) and visual cortices (Luth et al. 1994). The ubiquitous existence of MCs in the cerebral cortex of different ages and species suggests a central role in information processing in the cortical column. Indeed, MCs have been proposed to be involved in memory formation and storage (Eccles, 1983) and in neurodegenerative diseases (Beal et al. 1988).
Most previous studies on MCs have been limited to descriptions of their anatomical properties, largely based on Golgi staining, which may render only some neuronal processes visible (Ramon y Cajal, 1911; Marin-Padilla, 1970; Ferrer et al. 1986a,b). At the electrophysiological level, MCs have been reported to display ‘regular’ or ‘burst’ firing patterns to depolarizing somatic current injections (Kawaguchi & Kubota, 1997), which have also been reported to be low threshold interneurone (Kawaguchi, 1995). A more recent study has reported regenerative calcium activity in MCs which may be important in the bursting type of MCs (Goldberg et al. 2004). At the biochemical level, somatostatin (SOM) is reportedly expressed in MCs (Wahle, 1993; Kawaguchi & Kubota, 1996), while other biochemical markers, including neuropeptide Y (NPY) (Beal et al. 1988; Kuljis & Rakic, 1989; Obst & Wahle, 1995) and calbindin (CB) have also tested positive in MCs (Conde et al. 1994; Kawaguchi & Kubota, 1996; Kawaguchi & Kubota, 1997).
A systematic multidimensional study of the detailed anatomical, physiological and molecular properties of MCs is, however, still lacking. In particular, at the anatomical level, the presence of MCs in different layers is not well established, the morphological similarities and differences between MCs in different layers is not known, the cross-layer and cross-columnar axon targeting principles of MCs has not been isolated, the dendritic expanse of different MCs has not been determined, and the different shapes of MC somata are not clear. At the electrical level, the number of different response types and their frequencies of occurrence, as well as the heterogeneity of detailed electrical parameters, have not been determined. The ion channel genes supporting MC electrical properties are also not known. At the biochemical level, the co-expression profiles for neuropeptide and calcium binding protein genes are not known.
In the present study, we employed whole-cell patch-clamp recordings to acquire electrophysiological properties, histochemical staining to obtain 3-D anatomical reconstructions and morphometric analyses, as well as aspiration of the cytoplasm to determine the mRNA expression profile of neurones using multiplex RT-PCR (Fig. 1). We report for the first time that (a) MCs are found in all layers II–VI, (b) they all display some common anatomical, electrophysiological and biochemical properties, (c) they display distinct layer-specific differences, and (d) infragranular MCs also target layer IV in addition to layer I. We also report for the first time, the ion channel genes expressed in MCs.
|
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
All experimental procedures were carried out according to the Swiss federation rules for animal experiments. Wistar rats (13- to 16-days-old) were rapidly decapitated and neocortical slices (sagittal; 300 µm thick) were sectioned on a vibratome (DSK, Microslicer, Japan) filled with iced extracellular solution (composition below). Optimal slices (2–3 per hemisphere), running parallel to apical dendrites of PCs, were selected for recording. Slices were incubated for 30 min at 34°C and then at room temperature (24–25°C) until transferred to the recording chamber (34 or 24–25°C). The extracellular solution contained (mM): 125 NaCl, 2.5 KCl, 25 glucose, 25 NaHCO3, 1.25 NaH2PO4, 2 CaCl2 and 1 MgCl2. Neurones in somatosensory cortex were identified using IR-DIC microscopy, with an upright microscope (Zeiss Axioplan, fitted with 40 x W/0.75 NA objective, Zeiss, Oberkochen, Germany). Recorded neurones were selected up to 120 µm below the surface of the slice and laterally separated by up to 150 µm.
Electrophysiological recording
Somatic whole-cell recordings (pipette resistance 6–12 M
when there was no cytoplasm harvesting or 1–3 M for recordings where the cytoplasm was harvested) were made and signals were amplified using either Axoclamp 2B or Axopatch 200B amplifiers (Axon Instruments). Neurones were submitted to a series of somatic current injection protocols, during whole-cell patch-clamp recordings, designed to capture their key active and passive electrical properties (Fig. 6). We focused on the shape of the first two action potentials (APs) generated just above threshold (AP Waveform), the change in the AP amplitude with time (AP Drop) and the membrane time constant for brief hyperpolarizing current pulses (Delta), the neuronal response to ramp current injection (AP Threshold), the sag, overshoot and rebound spike produced by different hyperpolarization current pulses of increasing magnitude (Sag), discharge responses to step current pulses of increasing magnitude (ID Rest), the subthreshold current–voltage relationship (I–V), the hyperpolarization after a burst of APs (s-AHP), the discharge responses to step current pulses around threshold (ID Thresh). A numerical breakdown of the electrical behaviour was obtained by measuring various aspects of the voltage responses to these stimulation protocols (Tables 4–
7). Recordings were sampled at intervals of 10–400 µs using Igor (Wavemetrics, Lake Oswego, OR, USA), digitized by either an ITC-16 or ITC-18 interface (Instrutech, Great Neck, NY, USA) and stored on the hard disk of a Macintosh computer for off-line analysis (Igor). Voltages were recorded with pipettes containing (mM): 100 potassium gluconate, 20 KCl, 4 ATP-Mg, 10 phosphocreatine, 0.3 GTP, 10 Hepes (pH 7.3, 310 mOsm, adjusted with sucrose) and 0.5% biocytin (Sigma). Neurones were filled with biocytin by diffusion during the 30–90 min recordings. Membrane potentials were not corrected for the junction potentials between pipette and bath solution (
9 mV).
|
|
|
|
|
Most signals were sampled at 4 kHz, except for single action potentials (up to 20 kHz) and low-pass pre-filtered by 1 kHz (4–8 pole Bessel). Intrinsic properties: input resistances were approximated by linear regression of voltage deflections (± 15 mV from resting potential, –70 ± 2 mV) in response to 2 s current steps of four to eight different amplitudes after reaching steady state (end 200 ms of a 1 s current pulse). Steady-state current–voltage relationships were linear for most interneurones, allowing for this analysis. Membrane time constants were determined by fitting the decay phases of depolarizing and hyperpolarizing pulses (1 ms duration; voltage deflections of < 10 mV) to an exponential function, or by fitting the rising phases of the voltage traces used for determining the input resistances to an exponential function. Single AP analysis was performed on the first AP elicited by near-threshold depolarizations. Peak values of the AP and the fast AHP (fAHP) were determined by averaging 5–10 values around the peak. Maximum rise and fall rates were obtained as peak values after differentiating the 1st or 2nd AP. Classification of discharge behaviours was done according to previous studies (Gupta et al. 2000; Toledo-Rodriguez et al. 2003). Neurones were classified according to both their initial and steady-state discharge responses to step current injections. The initial response could be an onset delay (‘d’), typically several hundred milliseconds before initiation, a burst onset (‘b’), in which a group of two to five high frequency APs preceded the steady-state response, or an onset that was neither a delay nor a burst, which is referred to as a classical onset response (‘c’). The steady-state discharge pattern was divided into accommodating (‘AC’), non-accommodating (‘NAC’) and irregular spiking (‘IS’) responses. Accommodating discharge was characterized by a monotonic decrease in discharge rate throughout the duration of the current step injection while non-accommodating cells displayed little (less than ±10%) fluctuation in their instantaneous discharge frequency throughout the response. The degree of accommodation was quantified according to the ratio between the initial five ISIs and the latest five ISIs in the discharge response. Irregular spiking discharge behaviour was characterized by irregular discharges of APs. Discharge behaviours were robust up to 4 x threshold current injections (30–300 pA, depending on the individual neuronal input resistance) and stable for different holding potentials (from –85 to –60 mV) as well as different temperatures (20–24 and 34°C).
Histological procedures
After recording, slices were fixed for 24 h in cold 0.1 M phosphate buffer (PB, pH 7.4) containing 2% paraformaldehyde, 1% glutaraldehyde and 0.3% saturated picric acid. Thereafter, slices were rinsed several times (10 min each) in PB. To block endogenous peroxidases, slices were transferred into phosphate buffer containing 3% H2O2 for 30 min. After five to six rinses in PB (10 min each), slices were incubated overnight at 4°C in biotinylated horseradish peroxidase conjugated to avidin according to the manufacturer's protocol (ABC-Elite, Vector Laboratories, Peterborough, UK): 2% A, 2% B and 1% Triton X-100. Following incubation, sections were washed several times in PB and developed with diaminobenzidine (DAB, 0.14%) under visual control using a bright-field microscope (Zeiss Axioskop) until all processes of the cells appeared clearly visible (usually after 2–4 min). The reaction was stopped by transferring the sections into PB. After washing in the same buffer, slices were mounted in aqueous mounting medium (IMMCO Diagnostics, Inc). In some cases, slices were re-sectioned into 100 µm thick sections before mounting. Staining of slices for EM examination was performed as described in Wang et al. (2002); the histochemical staining procedure was the same as mentioned above except for one step where the slices were quickly frozen with liquid N2 instead of using Triton X-100. After the histochemical staining, the slices in which the filled cells were well visualized were re-sectioned at 100 µm thickness. To enhance the staining contrast, slices were postfixed for 45 min in 1% phosphate-buffered osmium tetroxide (Merck) and counterstained in 1% uranyl acetate. After several rinses in PB, sections were flattened between two glass slides and dehydrated through an ascending series of ethanol in small glass vials for 15 min each. Following two 10 min washes in propylene oxide (Merck), slices were embedded in Epon resin overnight at room temperature (Durcupan, Fluka, Buchs, Switzerland); afterwards slices were flat-embedded in Epon resin between a coated glass slide and a cover slip. Subsequently, light microscopic observation and 3-D reconstruction of the cells were carried out. Later some areas rich with interneurone boutons were re-embedded for cutting serial ultrathin sections, which were finally examined with electron microscopy (EM).
3-D computer reconstruction
3-D neurone models were reconstructed from stained cells using the Neurolucida system (MicroBrightField Inc., USA) and a bright-field light microscope (Olympus). After the staining procedure, there is 25% shrinkage of the slice thickness and 10% anisotropic shrinkage along the X- and Y-axes. Only shrinkage of thickness was corrected.
Quantitative morphometry (morphological analysis)
Reconstructed neurones were quantitatively analysed with NeuroExplorer (MicroBrightField Inc., USA). An array of eight axonal, and six dendritic parameters, designated as the ‘morphology profile’ (m-profile), was obtained to quantitatively compare the axonal and dendritic arbours of the Martinotti cells. An example of a layer V MC is given in Fig. 2. The axonal parameters include (Fig. 2A and B): (1) axonal Sholl distance (ASD), defined as the number of axonal intersections as a function of distance from the soma. A series of Sholl circles with 20 µm stepped radii centred in the interneurone soma were delineated and the number of axonal intersections in each stepped region and their distances to the centre were calculated. The maximum radius of Sholl circles used for the ASD calculation was 1 mm; (2) axonal segment lengths (ASL), defined as the length of axonal segments between two branch points or between a branch point and an end point; (3) axonal branch order (ABO), the branching frequency of an interneurone axon tree; (4) bouton density (BD), calculated as the number of boutons per axon length; (5) maximum axonal branch angle (MABA), the maximum angle formed between the extending distal line of the parent axonal segment and child axonal segments. Although the MABAs vary from less than 10 deg to nearly 180 deg, most MABAs lie between 40 and 100 deg; (6) planar axonal branching angle (PABA), the angle formed between the extending distal line of the parent axonal segment and a child axonal segment; (7) total number of axonal segments (SEG); (8) total number of boutons per cell (BT). Dendritic parameters were obtained by applying the same criteria as for the axonal structure and designated (Fig. 2A and C): (1) dendritic Sholl distance (DSD); (2) dendritic segment length (DSL); (3) dendritic branch order (DBO); (4) maximum dendritic branch angle (MDBA); and (5) planar dendritic branch angle (PDBA), respectively. In addition, the average length of the dendritic tree (ALDT) was defined as the average length of a single dendrite including all its branches (dendritic tree).
|
In order to determine the postsynaptic targets of a filled MC, a region containing boutons only belonging to the studied MC was selected. The targets of all the boutons encountered in subsequent serial sections were examined under EM according to established criteria (Peters et al. 1991). Briefly, a postsynaptic target of a filled bouton was judged to belong to soma/dendrite/spine, based on its ultrastructural characteristics. Identification of the nature of postsynaptic dendritic shafts (pyramidal versus interneurone) was according to the previously published criteria (Peters et al. 1991). Briefly, dendritic shafts of a presumed PC have less cell organs presenting light cytoplasm and the dendritic shafts of a presumed interneurone have more cell organs presenting darker cytoplasm. When the plane of the section was not perpendicular to the junction of membranes, the synaptic cleft was revealed by tilting the section using the goniometer of the EM.
Cytoplasm harvesting and single-cell reverse transcription
This was performed as previously described (Cauli et al. 1997; Wang et al. 2002). In brief, recording pipettes were loaded with 5 µl of RNAse-free intracellular solution. At the end of the recording, cell cytoplasm was aspirated into the recording pipette under visual control by applying gentle negative pressure. Only cells in which the seal was intact throughout the recording, and whose nucleus was not harvested, were further processed. The electrode was then withdrawn from the cell to form an outside-out patch that prevented contamination as the pipette was removed. The tip of the pipette was broken and the contents of the pipette expelled into a test tube by applying positive pressure. mRNA was reverse transcribed using an oligo-dT primer (25 ng µl–1) and 100 Units of Superscript II reverse transcriptase (Gibco, BRL) in a final volume of 20 µl. After 50 min incubation at 42°C, the cDNA was frozen and stored at –20°C before further processing.
Multiplex PCR
Multiplex PCR conditions were optimized using total RNA purified from rat neocortex, so that a PCR product could be detected from 0.25–1 ng of total RNA without contamination caused by non-specific amplification. For the list of primer pairs included into the different multiplexes, the name and accession number of the genes amplified and the length of the PCR product see Table 1. Three different multiplex PCR reactions were performed for testing the expression of 30 mRNA species from each cell. The genes co-amplified in each of three multiplex–PCR reactions were Pool I (CB, PV, CR and GAPDH), Pool II (Kv1.1, Kv1.2, Kv1.6, Kv2.1, Kv2.2, Kv3.1, Kv3.2, Kv4.2, Kvß1, Kvß2, HCN1 and HCN2), Pool III (Kv1.4, Kv3.3, Kv3.4, Kv4.3, HCN3, HCN4, Ca1A, Ca1B, Ca1G, Ca1I, Caß1, Caß3, Caß4 and SK2). Pool 1 was already calibrated to give PCR products for each gene with even intensity (Wang et al. 2002). Pools II and III were calibrated to give PCR products for each gene with even intensity starting from 1 ng of brain total mRNA. During calibration, different combinations of genes were distributed between the two pools (II and III) and different primer pairs were tested until an even amplification of all genes in the pool was obtained.
|
Controls for the RT-PCR
For each PCR amplification, controls for contaminating artefacts were performed using sterile water instead of cDNA. A control for non-specific harvesting of surrounding tissue components was randomly employed by advancing pipettes into the slice and retrieving without seal formation and suction. Both types of controls gave negative results throughout the study. Amplification of genomic DNA could be excluded by the intron-overspanning location of many of the primers and by the fact that the cell nucleus was never harvested. Moreover controls in which the RT was omitted gave negative results.
Statistical analysis
Student's t tests were applied to compare between two groups of quantitative parameter values. Data are given as means ± standard deviation (S.D.) or means ± standard error (S.E.M.).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Martinotti cells were readily distinguished from other interneurones mainly according to their axonal arborization. The essential criteria were: (a) axonal projections with a horizontal spread in layer I that typically extended beyond a column radius (150 µm) (see Fig. 1A), and (b) axonal collaterals decorated with spiny boutons (Fig. 1C). Additionally, MCs often presented more spines on their dendrites than other types of interneurones at this stage of development. One hundred and eighty MCs were anatomically verified according to the two essential axonal criteria throughout layers II–VI (layer II/III, n = 62; layer IV, n = 52, layer V, n = 51, layer VI, n = 15). Of these neurones, 91 with high quality axonal staining (27 in layer II/III, 28 in layer IV, 29 in layer V and 7 in layer VI) and 77 with high quality dendritic staining (22 in layer II/III, 20 in layer IV, 26 in layer V and 9 in layer VI) were examined at the light microscopic (LM) level for cross-layer comparisons, and 24 (10 in layer II/III, 7 in layer IV, 6 in layer V and 1 in layer VI) were randomly selected and 3-D computer-reconstructed in order to perform a quantitative comparison of MCs located in different layers.
Morphological characteristics
Most MC somata were ovoid or spindle shaped (94%), whereas others could have a pyramidal, round or multipolar form. MC somata usually gave rise to vertically orientated bundles of two to four primary dendrites from opposite somatic poles (bitufted dendritic morphology, 89%), with one of the primary dendrites branching more frequently and descending to deeper layers (72%, Fig. 3). Typically, MC dendrites were beaded (96%) and bore spines with sparse to medium density (86%). MC dendrites often branched frequently giving rise to the most elaborate dendritic trees of all the interneurones at this stage of neocortical development (DBO in Table 2). The lateral expanse of the dendritic arborization was usually less than that of the basal dendritic arbour of PCs (300 µm of diameter) (Table 2), suggesting that MCs receive input signals from several layers, but within the dimensions of a cortical column.
|
|
A small fraction of MCs (5%) were consistently noticed to display much finer axons than most of the MCs (Table 2, AMC). A quantitative analysis revealed; (i) longer axonal segments, (ii) smaller and denser boutons (BD, P < 0.05), (iii) narrower axon branching angles (MABA, PABA, P < 0.05), and (iv) more axonal collaterals projecting towards deeper layers. Interestingly, they also seemed to present differences in the gene expression (see below). Such MCs were found in layer II/III (n = 6) and lower layers (n = 3).
Laminar specificity
Comparison between MCs located in different layers revealed layer specificity in morphologies of the dendritic and axonal arbours.
Layer II/III MCs.
These MCs usually had the most extensive dendritic arbours. Most MCs (91%) projected their dendrites (mainly with a primary prominent dendrite) down to deeper layers, and more than half projected their dendrites as far as infragranular layers (layers V and VI). Their axonal collaterals were distributed mainly in layer I. Most MCs (74%) formed a more dense axonal cluster in layer I while the remainder formed a more dense axonal cluster around their own somata (Fig. 3A1). The clusters in layer I appeared as a secondary peak 350–370 µm from the somata in the axon Sholl distance histogram (ASD, reflecting the overall axonal spread, Fig. 3A2). On average, more than half (56%) of the total boutons of a layer II/III MC were distributed in layer I (see Table 2). These data indicate that layer I is the major target for layer II/III MCs.
Layer IV MCs. The dendritic distribution of layer IV MCs tended to be localized more within layer IV (43% with dendrites confined to layer IV; Fig. 3B1). The remainder of the MCs projected a few dendritic branches into neighbouring layers, but rarely further. Their dendritic trees were smaller than those of layers II/III and V MCs, presenting significantly shorter average lengths of dendritic trees (ALDT, P < 0.05, Table 2), shorter vertically spreading dendritic arbours (P < 0.05, Table 2), and smaller dendritic spread distance (DSD, reflects the overall dendritic spread, P < 0.05, Table 2). Their axon distribution also tended to be mostly restricted to their layer. They typically formed a prominent local axonal cluster around their somata (89% of layer IV MCs), sending only a few collaterals up to layer I which formed a sparse arborization, with a few horizontally projecting axons. Only a single peak close to the soma is therefore present in the ASD histogram (Fig. 3B2). Only 18% of the total boutons of layer IV MCs are distributed in layer I, which was significantly lower than both layer II/III (P < 0.01) and layer V MCs (P < 0.05; Table 2). Layer IV MCs generally also tended to present lower axonal branch orders (ABO), and smaller total axonal lengths compared with layer II/III MCs (Table 2).
Layer V MCs.
The dendritic trees of layer V MCs were similar to those of layer II/III MCs. Seventy-three per cent distributed their dendrites mainly in the infragranular layers. We found that in addition to targeting layer I, the MCs also targeted layer IV. Most of these MCs formed larger axonal clusters in layer IV and in layer I than around their somata (94% of layer V MCs, Fig. 3C1). On average, most boutons were located in layer IV, but 36% were located in layer I (Table 2), suggesting that layer I is also a major target for layer V MCs. A typical layer V MC therefore presented two secondary peaks in the ASD histogram, one up to 520 µm away from the soma representing the axonal clusters formed in layer IV and another at 750–850 µm away representing the axonal clusters formed in layer I (ASD, Fig. 3C2). Consistent with their deep location, layer V MCs presented a greater number of axonal segments (SEG), larger total axonal lengths, overall axonal spread (ASD), higher branch orders (ABO) and more boutons and than MCs in layer II/III and IV (Table 2).
Layer VI MCs. These MCs were similar to those in layer V (Fig. 3D). Their dendrites mostly extended within infragranular layers (8/9). Like layer V MCs, they also formed axonal clusters in both layers IV and I, but more often formed denser axonal clusters around their somata (4/7) than found for layer V MCs (2/27).
In summary, these data reveal layer specificity of dendritic and especially axonal morphologies of MCs. Layer II/III MCs typically send their dendrites down to deep layers and their axons up to target layer I. Layer IV MCs tend to restrict both dendrites and axons to layer IV. Layer V and VI MCs span the dendrites in the deeper layers, and send their axons up to layers IV and I.
Synaptic targets of Martinotti cells under EM
To examine the synaptic targets of MCs, random EM examination of MC boutons was obtained from seven MCs. From 130 boutons (including 7 boutons with two synaptic sites) 137 synapses were identified, which were all found to be symmetrical synapses (Fig. 4). Of these synapses, 71% were formed onto dendritic shafts (Fig. 4A and B) (42% onto PC shafts, 10% onto interneurone shafts and 19% onto dendritic shafts whose origins could not be identified), 22% onto spines (Fig. 4C), and 7% onto somata (Fig. 4D).
|
Electrophysiology of Martinotti cells
Electrophysiological properties were obtained from the discharge responses to current injection in current-clamp mode. The electrophysiological properties were studied in detail for 127 out of the 180 MCs, which included 67 MCs recorded at 24–25°C (30 in layer II/III, 19 in layer IV, 14 in layer V and 4 in layer VI) and 60 MCs recorded at 34°C (13 in layer II/III, 10 in layer IV, 30 in layer V and 7 in layer VI).
Electrophysiological characteristics
MCs displayed diverse discharge responses to sustained somatic current injections. Responses were distinguished according to previously established criteria (see Methods; Gupta et al. 2000; Wang et al. 2002; Toledo-Rodriguez et al. 2003). The majority of the MCs displayed accommodating responses (AC, 90%) which is analogous to ‘regular spiking’ (Kawaguchi, 1995; Cauli et al. 1997), a small per cent displayed non-accommodating responses which is analogous to ‘fast spiking’ (NAC, 8%) and only a few MCs displayed irregular spiking responses (Kawaguchi & Kubota, 1993; Porter et al. 1999) (IS, 2%; see Table 3). The manner in which the neurones began their discharge (onset response) was also examined according to previously established criteria (see Methods) (Fig. 5). Of the ACs (n = 115), 84.5% displayed classical onset responses (c-AC) while 15.5% displayed burst onset responses (b-AC). All the NAC MCs (n = 10) displayed classical onset responses (c-NAC) while the two IS MCs started their response with a burst (b-IS) (Table 3). These responses were seen at 24–25 and 34°C, with the exception of the two IS responses, which were observed at 34°C.
|
|
In order to examine the electrophysiological properties of MCs in more detail, we applied a spectrum of stimulation protocols designed to capture key active and passive electrical properties (Fig. 6). The active properties studied included characteristics of the AP and AHP waveform, and measurements of the discharge behaviour such as the rate of accommodation or irregular firing. The passive membrane properties studied included: I–V relationships, membrane time constants at different potentials, and membrane voltage decay following a brief hyperpolarizing current pulse. In order to standardize these profiles across all neurones, the amplitude of stimulation was normalized according to the minimal step current required to reach AP threshold. Subsequent analysis yielded 58 values that represent key active and passive properties of the neurone (Tables 4 and 6).
This analysis was separate and independent from the response classification of MCs mentioned above and was aimed at finding those electrical parameters that are the most uniform and those that are the most heterogeneous among MCs. The coefficient of variation of an electrical parameter across neurones was used as an index of heterogeneity. In general, parameters measuring similar electrical features had similar coefficients of variation (CV) (Table 5 and Fig. 7). The most uniform parameters (absolute CV < 0.1) across MCs were the threshold for AP generation and resting membrane potential. This is significant in view of the fact that interneurones, in general, can display wide variations in resting potentials (around 10 mV) and in action potential thresholds (around 20 mV). The action potential waveform was also highly uniform (first and second AP amplitude, total and half-width duration, rise and fall time).
|
Laminar specificity
The different response types of MCs were not evenly distributed throughout the cortical layers (Table 3). The accommodating MCs were found in all cortical layers, while NACs were found only in upper layers (layer II/III, 66% and IV, 34%) and bursting MCs were mostly found in layer V. Consistent with the different response types, layer IV MCs also presented with the lowest AP thresholds while layer VI presented with the highest (Table 4). Interestingly, opposite to the layer trend seen in PCs, upper layer MCs tended to display greater rectification in the peak and steady-state current–voltage curves (typically referred to as sag, due to Ih currents), than deeper layer MCs.
Molecular properties of Martinotti cells
To obtain the molecular expression profiles for the calcium binding proteins (CB, PV and CR) and the neuropeptides (NPY, VIP, SOM and CCK), single-cell RT-PCR was performed for 63 MCs located throughout layers II–VI (25 in layer II/III, 19 in layer IV, 14 in layer V, 5 in layer VI). Martinotti cells were found to be 100% SOM positive, PV negative and VIP negative (n = 63, Table 7). MCs are therefore distinct from basket cells, chandelier cells, some bipolar cells and bitufted cells in terms of their main patterns of expressing calcium binding proteins and neuropeptides (Fig. 8C).
|
1A, Ca1B, Ca1G and Ca1I) as well as their auxiliary subunits (Caß1, Caß3 and Caß4) (see Table 1 for the list of primer pairs, the name and accession number of the genes amplified and the length of the PCR product). Molecular heterogeneity
In more than half of the MCs, SOM was the only biochemical marker expressed (Table 7). The remaining MCs co-expressed SOM with CB, CR, NPY or CCK. The majority of the co-expressions were with two biochemical markers and only three cases of co-expression with the three biochemical markers, SOM, CB and NPY, and 1 case with SOM, CB and CCK. There seemed to be some expression trends in MCs with different discharge responses. CCK was expressed only in the c-AC MCs (n = 6/6), while NPY was detected only in MCs with classical onset discharges (n = 10, 8 c-ACs, 2 c-NACs). NPY was also expressed in 2 of the 3 NACs examined. CB on the other hand, was expressed by MCs belonging to all three electrophysiological subtypes (c-AC, 9; b-AC, 2; c-NAC, 2). CB was also detected in all four of the rare slender axon MCs examined, while it was rarely expressed in the other MCs (15%, 6/35).
There was not sufficient data to make any statistically valid statements of layer specificity, but there seems to be a tendency for the lack of CCK expression in layer V MCs, lack of CB expression in layer VI MCs and presence of CR expression in layer IV (Table 7).
Ion channel expression
Frequency of expression is a poor index to describe the expression profile of neurones because of the uncertainties of harvesting and amplifying a particular mRNA, but positive expression is very informative and trends may be used as guidelines for future studies. MCs were screened for the expression of 26 K+ and Ca2+ channel alpha and beta subunit genes including: (a) the voltage-activated K+ channels (Kv1.1, Kv1.2, Kv1.4, Kv1.6, Kv2.1, Kv2.2, Kv3.1, Kv3.2, Kv3.3, Kv3.4, Kv4.2 and Kv4.3) and their auxiliary subunits (Kvß1 and Kvß2); (b) the K+/Na+-permeable hyperpolarization-activated ion channels (HCN1, HCN2, HCN3 and HCN4); (c) the Ca2+-activated K+ channel (SK2); (d) the voltage-activated calcium channels (Ca1A, Ca1B, Ca1G and Ca1I) and their auxiliary subunits (Caß1, Caß3 and Caß4, Fig. 8D). We further examined only cells that expressed a minimum of 5 out of the 34 genes, including GAPDH, and only MCs that displayed c-AC responses, since this was the major response type (n
= 24).
The ion channels with the highest detected expression (from highest to lowest) were Caß1, Kv3.3, HCN4, Caß4, Kv3.2, Kv3.1, Kv2.1, HCN3, Ca1G, Kv3.4, Kv4.2 and HCN2. SK2 expression was never detected in MCs (it has been detected in other types of interneurones using the same technique, Toledo-Rodriguez et al. 2004). It is interesting to note that Kv3.1 and Kv3.2 were also expressed in MCs, even though these ion channels have been related with fast spiking behaviour. The expression of these channels by themselves is therefore not sufficient to produce high frequency discharge or the co-expression of other genes in MCs prevents the development of fast firing behaviour.
MCs often display a depolarization following the AP (ADP) and we therefore examined the correlation between the presence of the ADP and the frequency of expression of some of the ion channel genes. From the 24 MCs whose ion channel expression was investigated, 8 presented ADPs (2.9 ± 2.7 mV; measured as mV after potential following the first AP generated in the AP threshold stimulus protocol) and the remainder showed no after potential at all. Of the four Na+/K+ ion channels tested, HCN2 was expressed more frequently in MCs with a prominent ADP (50% with ADP versus 25% without ADP). Conversely, HCN3 expression was more frequent in MCs without ADPs (25% with ADP versus 50% without ADP). From the high voltage Ca2+ channels, Ca1A was more expressed in MCs with ADPs (38% with ADP versus 19% without ADP). In contrast, some of the low voltage Ca2+ channels showed no or a mild expression preference (Ca1I 13% with ADP versus 6% without ADP; Ca1G 38% in both cases). Finally, A-type K+ channels such as Kv4.2 were more frequently detected in MCs without ADPs (13% with ADP versus 38% without ADP). Kv2.1 was also more frequently detected in MCs without ADPs (13% with ADP versus 56% without ADP). While this study shows which genes can be expressed in MCs, the number of MCs in this analysis (n
= 24) is too low for rigorous statistical validation of comparisons and the descriptions should be viewed only as guidelines for future studies.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The present study provides the first detailed anatomical, electrophysiological and molecular analysis of Martinotti cells in layers II to VI of rat somatosensory cortex. The study yielded features common to MCs in all layers, layer-specific morphological and electrophysiological properties, as well as the multidomain, multilayer and cross-columnar targeting nature of MCs. MCs are mostly bitufted and accommodating interneurones always express SOM but never express PV or VIP and target layer I. Morphologically, the main dendritic axis is down towards deeper layers while the axonal focus is up toward layer I. MCs are specialized to target multiple layers including the same layer where the soma is located, layer IV and layer I. Electrophysiologically, MCs are shown to display different discharge responses in different layers with a tendency for accommodating MCs to be located in all layers, non-accommodating MCs to be located in upper layers and bursting MCs in deeper layers. Molecularly, the study confirms previous works showing that SOM is commonly expressed in MCs and extends these findings not only by demonstrating that SOM expression as an essential property of all MCs in all layers, but also that the lack of PV and VIP is an essential part of their molecular profile. We further find that MCs can express CR, NPY and CCK, which may depend on the layer in which the MC is located. Finally, we show for the first time, using the single-cell multiplex RT-PCR approach, the expression of ion channel genes in MCs (from highest to lowest); Caß1, Kv3.3, HCN4, Caß4, Kv3.2, Kv3.1, Kv2.1, HCN3, Ca1G, Kv3.4, Kv4.2, Kv1.1 and HCN2; are most often detected in c-AC MCs. Additionally, HCN2 and the high threshold calcium channels seem to be more expressed in c-AC MCs that display ADPs. The data also shows that MCs express Kv3.1, which is a powerful delayed rectifier commonly expressed in fast spiking interneurones. Low voltage-activated calcium channels, as well as ion channel genes supporting A-type currents, can be expressed by MCs. In conclusion, this study provides deeper insight into the morphological, electrophysiological and molecular design of MCs and reveals the anatomical substrate for a crucial role in multiple inhibitory feedback loops within and between cortical columns.
Morphology
Martinotti cells constitute roughly 16.5% of the interneurones in somatosensory cortex of young rats (authors' unpublished estimates). The proximal dendritic arbour is bipolar or bitufted in nearly 90% of cases with an ovoid or spindle-like soma and can only be approximately identified using IR-DIC microscopy. MCs can, however, be identified easily after staining, by two unique axonal features – the axonal arborization in layer I and spiny boutons. The axonal arborization in layer I can extend to most of the neighbouring and even to more distal columns as far as 2 mm away. MCs are also unique in terms of their dendritic morphology – they display the most elaborate dendritic arbours of all interneurones (authors' unpublished data) and they bear spines more often on their dendrites at this stage of development. Since interneurones are essentially spine free after maturation (Fairen et al. 1984), this perhaps suggests that MCs complete their maturation slower than other interneurones.
Electrophysiology and ion channel expression
Previous reports indicated that MCs can respond with ‘regular’ and ‘burst’ firing in response to depolarizing somatic current injections and that the threshold for discharging responses is low (referred to as low threshold spiking, LTS) (Kawaguchi, 1995; Kawaguchi & Kubota, 1997; see also Goldberg et al. 2004). In this study, we found three main response types with 90% accommodating (regular spiking), 8% non-accommodating and 2% irregular spiking responses. The bursting response type was found in a subset of the accommodating and all of the irregular spiking MCs, making up a total of about 17% of all MCs. These data are therefore consistent with findings in the adult rat and further show that some MCs can display fast spiking responses and provide estimates of the frequencies of occurrences of the different response types.
The spike train accommodation found in most MCs superficially appears similar to regular spiking in PCs, but the responses are different in many respects. The most important difference is that, while accommodation in PCs is caused by AHPs due to Ca2+-activated K+ outward conductances, MCs lack AHPs, as well as the SK2 ion channel that produces Ca2+-activated K+ currents; instead MCs display ADPs more often. Further studies are required to determine which ion channels contribute towards MC accommodation.
Many MCs, especially deep layer MCs, display bursting behaviour. Prominent after-depolarizations, which facilitate bursting, are commonly found in accommodating MCs and not in non-accommodating MCs. The ADP is believed to result from the combined activity of the hyperpolarization-activated Na+/K+ channels (HCN1–HCN4). Indeed, the hyperpolarization-activated Na+/K+ channel (HCN2) seems to be more expressed in MCs presenting ADPs, but the ADP is not the only cause of bursting in MCs since some non-bursting MCs can also display ADPs. The MCs, however, seem to lack the T-type voltage-activated Ca2+ channel Ca1I which may be required to boost the depolarization to trigger a burst.
Layer-specific electrophysiology
This study revealed that MCs can respond differently to depolarizing input. Some of these differences seem to depend on the layer in which the interneurone is located. We found that the AP threshold is lowest for layer IV MCs and highest for layer VI MCs. This may pre-dispose to earlier activation of layer IV MCs at the onset of processing in the microcircuit and later stage activation of layer VI MCs at a more advanced stage of microcircuit processing. Indeed, several in vivo studies have shown that putative inhibitory neurones in layer IV respond most vigorously at low thresholds and with short latencies (Simons, 1978; Yamamoto et al. 1988; Swadlow, 1989). Layer VI MCs were actually found to display the highest thresholds of all interneurones (data not shown) suggesting that they permit the most elaborate integration. This may be important to allow specific thalamic input to escalate to a critical level via the positive feedback loop between layer VI PCs and thalamic neurones, before layer VI MCs would break the positive feedback loop.
The various electrophysiological subtypes, while much less frequent, may also be important to allow MCs to perform multiple tasks. For example, most bursting MCs were located in layer V where many intrinsic bursting PCs are also commonly found. These bursting PCs are the main neurones providing the output to subcortical regions (see De la Pena & Geijo-Barrientos, 1996). Bursting could make signal transmission more intense and reliable and the higher fraction of bursting MCs in layer V may therefore be important for resetting the cortical column after moments of intense output.
Biochemical properties
Previous immunohistochemical studies have reported that MCs belong to the group of SOM-positive interneurones (Wahle, 1993; Kawaguchi & Kubota, 1997; Kawaguchi & Kubota, 1998). We confirm this finding at the mRNA level and further show that SOM is expressed by all MCs and in all layers regardless of their anatomical and electrophysiological differences. SOM is a neuropeptide that acts as a co-released inhibitory neurotransmitter by opening different types of potassium channels (Tallent & Siggins, 1997). SOM also activates the delay rectifier channels in cultured neocortical neurones. In the solitary nucleus SOM inhibition occurs through hyperpolarization and augmentation of a non-inactivating voltage-dependent outward current blocked by muscarinic agonists (IM). SOM also augments the IM current in CA1 hippocampal neurones, inhibits calcium currents in rat neocortical neurones, and reduces N-type currents in dissociated hippocampal PCs. The potential regulatory activity of SOM is even more complex when one takes into account that the SOM gene gives rise to two main biologically active products, SOM 14 and SOM 28. Both of these are believed to act as neurotransmitters showing opposite effects on K+ currents in rat neocortical neurones. A single MC can therefore exert a complex modulatory effect on the microcircuit (for references see Schweitzer et al. 1998).
The overall significance of such modulation is not known, but changes in SOM immunoreactivity and in morphology of SOM-positive neurones in the neocortex is a well recognized neurochemical indicator in late stage Alzheimer's disease (Armstrong et al. 1985). Whether this means that malfunction of MCs are involved in Alzheimer's disease is not clear because other morphological classes of interneurones (albeit less common) also express SOM. At the protein level, SOM was found to be expressed in wide arbour cells (Kawaguchi & Kubota, 1998). At the mRNA level, it was found to be expressed in small fractions of the small and nest basket cells, but hardly ever in large basket cells (Wang et al. 2002; Toledo-Rodriguez et al. 2003). SOM expression has also been detected in some bipolar, double bouquet and bitufted cells (Toledo-Rodriguez et al. 2003) (and authors' unpublished results).
All MCs lack the expression of PV and VIP. This is not due to false negatives, since the same technique reliably detected the expression of PV and VIP in other interneurone types (Wang et al. 2002; Toledo-Rodriguez et al. 2003, 2004). The finding is also in agreement with previous studies showing that at the protein level PV, VIP and SOM expression are found in different interneurones in layers II/III and IV of the rat frontal cortex (Kawaguchi & Kondo, 2002). SOM mRNA can, however, be co-expressed with VIP mRNA in bipolar, double-bouquet and small basket cells (Wang et al. 2002; Toledo-Rodriguez et al. 2003, 2004) (and authors' unpublished results), and regular spiking or irregular spiking cells (Cauli et al. 1997). At the mRNA level, a small percentage of interneurones co-express PV with SOM in large and nest basket cells (Wang et al. 2002), and fast spiking and regular spiking neurones (Cauli et al. 1997). A small basket cell also expressing SOM was found to co-express PV and VIP (Wang et al. 2002).
In addition to SOM, MCs also less frequently express other neuropeptides and calcium binding proteins, including CB, CR, NPY and CCK. These results confirm and extend previous findings (NPY expression (Beal et al. 1988; Kuljis & Rakic, 1989; Obst & Wahle, 1995); CB expression (Conde et al. 1994; Kawaguchi & Kubota, 1996, 1997; DeFelipe, 1997). These co-expressions could reflect molecular and functional heterogeneity of the MCs, but there is not sufficient data to make conclusive statements. Some trends are, however, worthy of further study. For example, the most common co-expression patterns were SOM–CB, SOM–NPY and SOM–CCK. CR expression was only detected in layer IV MCs, and CB was absent in layer IV MCs, which is consistent with the lack of burst-type MCs in layer IV. While the numbers are insufficient for a conclusive statement, it could be that CCK expression is absent from layer V MCs.
A multi-domain targeting interneurone
Interneurones differ in terms of their anatomical, electrophysiological and gene expression properties, and multiple attempts have been made to classify interneurones according to one or several of these properties. One scheme classifies interneurones according to the domain of the postsynaptic neurones targeted (Somogyi et al. 1982, 1998; Thomson & Deuchars, 1997; Toledo-Rodriguez et al. 2003). According to this scheme, there are four main classes of interneurones: those that inhibit the axon initial segment (e.g. axo-axonic or chandelier cells), those that inhibit the somatic and peri-somatic region (e.g. large, small and nest basket cells), those that inhibit the dendrites (e.g. bitufted cells, double bouquet cells, neurogliaform cells and bipolar cells), and those that target the most distal dendrites.
MCs target distal dendrites and therefore fall into the 4th class of distal dendritic targeting cells. However, MCs in all layers also arborize locally around their somata and EM analysis further revealed that while MCs innervate primarily dendrites of PCs, they can even innervate somata of PCs and dendritic shafts of interneurones, sharing some features of the 2nd and 3rd classes. This innervation overlaps with that of basket cells and suggests that MCs can also exert a classical fast local inhibitory feedback, more related to gain control than dendritic integration (Bernander et al. 1994; Mainen & Sejnowski, 1996). MCs can also innervate somata in layer I, which presumably belong to the rare layer I interneurones. These data therefore suggest that MCs have at least four different targeting strategies: (1) distal tuft dendrites of PCs in layer I, (2) apical and basal dendrites of PCs and interneurones, (3) somata of PCs, and (4) somata of interneurones in layer I.
A multi-layer and cross-columnar targeting interneurone
Layer I is an almost cell-free layer, which is scattered with a few rare cells such as the Cajal-Retzius cells. The somatic innervations found in layer I are presumably also onto these rare interneurones. Indeed, GABAA responses have been recorded in Cajal-Retzius cells (Radnikow et al. 2002). One possible function of such an innervation could be to modulate Cajal-Retzius cells, which releases reelin to guide PC positioning within the cortical column (Goh et al. 2002), suggesting that MCs may play an important role in development of the column.
The innervation of PC dendrites in layer I from neurones located in deeper layers is a specialized feature of MCs. These tufted dendrites have been shown to be essentially electrically separate compartments of the neurone (Yuste et al. 1994; Larkum et al. 1999, 2001) and the effect on dendritic integration is highly complex (Bernander et al. 1991; Larkum et al. 2001; Rhodes & Llinas, 2001). Layer I is a layer where the dendrites from PCs in all layers converge and inhibitory control of these tuft<
