NMDA receptor (NMDAR) activation requires concurrent membrane depolarization, and glutamatergic synapses lacking AMPA receptors (AMPARs) are often considered “silent” in the absence of another source of membrane depolarization. During the second postnatal week, NMDA currents can be enhanced in rat auditory cortex through activation of the α7 nicotinic acetylcholine receptor (α7nAChR). Electrophysiological results support a mainly presynaptic role for α7nAChR at these synapses. However, immunocytochemical evidence that α7nAChR is prevalent at postsynaptic sites of glutamatergic synapses in hippocampus and neocortex, along with emerging electrophysiological evidence for postsynaptic nicotinic currents in neocortex and hippocampus, has prompted speculation that α7nAChR allows for activation of NMDAR postsynaptically at synapses lacking AMPAR. Here we used dual immunolabeling and electron microscopy to examine the distribution of α7nAChR relative to AMPAR (GluR1, GluR2, and GluR3 subunits combined) at excitatory synapses in somatosensory cortex of adult and 1-week-old rats. α7nAChR occurred discretely over most of the thick postsynaptic densities in all cortical layers of both age groups. AMPAR immunoreactivity was also detectable at most synapses; its distribution was independent of that of α7nAChR. In both age groups, approximately one-quarter of asymmetrical synapses were α7nAChR positive and AMPAR negative. The variability of postsynaptic α7nAChR labeling density was greater at postnatal day (PD) 7 than in adulthood, and PD 7 neuropil contained a subset of small AMPA receptor-negative synapses with a high density of α7nAChR immunoreactivity. These observations support the idea that acetylcholine receptors can aid in activating glutamatergic synapses and work together with AMPA receptors to mediate postsynaptic excitation throughout life.
- sensory cortex
- receptive field properties
- α7 nicotinic acetylcholine receptor
- AMPA receptor
- NMDA receptor
- synaptic plasticity
- early postnatal development
- postsynaptic density
- electron microscopy
Plasticity of receptive field properties in sensory cortex depends on experience-dependent modification of glutamatergic pathways (Kleinschmidt et al., 1987; Fox et al., 1996; Rasmusson, 2000) and cholinergic modulation of these pathways (Bear and Singer, 1986; Sillito and Murphy, 1987; Müller and Singer, 1989; Baskerville et al., 1997; Kilgard and Merzenich, 1998). Although developmental studies have focused on muscarinic acetylcholine receptors, recent knowledge regarding central nicotinic receptors, particularly the α7 nicotinic acetylcholine receptor subunit (α7nAChR), indicates that these too are important for synaptic plasticity. Specifically, α7nAChR subunits (1) form homo-oligomeric channels with greater Ca2+permeability than NMDA receptor (NMDAR) (McGehee, 1999), (2) are prevalent during the first postnatal week, before onset of the critical period for experience-dependent developmental plasticity (Fuchs, 1989;Bina et al., 1995; Broide et al., 1995), and (3) enhance NMDA currents specifically at synapses of neonatal cortex lacking AMPA receptor (AMPAR) activity (Aramakis and Metherate, 1998; Aramakis et al., 2000).
Despite its prevalence, the role of α7nAChR in modulating excitatory transmission and long-term synaptic changes is unclear. Much of the physiological effect of nicotine apparently results from presynaptic modulation of glutamatergic pathways, i.e., enhanced glutamate release (McGehee et al., 1995; Role and Berg, 1996; Jones et al., 1999). However, electrophysiological evidence exists for direct activation of postsynaptic nAChR in hippocampus (Alkondon et al., 1998; Frazier et al., 1998a,b; Hefft et al., 1999) and neocortex (Roerig et al., 1997;Chu et al., 2000). Also, previous immunocytochemical results showed a primarily somatodendritic distribution of α7nAChR within neocortical and hippocampal pyramidal neurons (Dominguez del Toro et al., 1994).
Past EM immunocytochemical work by us and others agrees with these immunocytochemical findings: α7nAChR is concentrated at postsynaptic sites of asymmetric synapses in cortex (Lubin et al., 1999) and hippocampus (Fabian-Fine et al., 2001). These synapses are presumably excitatory (Gray, 1959; Altschuler et al., 1984; Somogyi et al., 1986;Aoki et al., 1991). Although this suggests that some of the direct or indirect effects of nicotine occur via postsynaptic glutamate receptors, morphology alone cannot reveal the glutamate receptor subtypes at these α7nAChR-immunopositive synapses. Immunocytochemical studies of hippocampus indicate that NMDARs are present in most asymmetric synapses throughout postnatal development, whereas AMPARs are acquired selectively (Nusser et al., 1998; Kharazia and Weinberg, 1999; Liao et al., 1999; Petralia et al., 1999; Shi et al., 1999). These results complement physiological evidence from hippocampus (Isaac et al., 1995; Liao et al., 1995; Durand et al., 1996) and neocortex (Wu et al., 1996; Isaac et al., 1997; Rumpel et al., 1998) that many NMDAR-containing synapses lack AMPAR function early in development. These synapses are often considered physiologically “silent” without the local depolarization required for NMDAR activity (Feldman et al., 1999). Might α7nAChR contribute to postsynaptic activation of these synapses? To address this question, we examined the distribution of α7nAChR relative to AMPARs at asymmetric synapses in neonatal and adult rat somatosensory cortex. EM immunocytochemistry was used to resolve the presynaptic versus postsynaptic location of α7nAChR and to determine whether the presence of α7nAChR at nascent synapses precedes the arrival of AMPARs.
MATERIALS AND METHODS
Tissue preparation. Adult (200–300 gm) and postnatal day (PD) 7 Sprague Dawley rats were deeply anesthetized with sodium pentobarbital (50 mg/kg body weight) and then perfused transcardially with a mixture of 1% glutaraldehyde and 4% paraformaldehyde in 0.1m phosphate buffer (PB), pH 7.4. Brains were postfixed by immersion at room temperature (RT) for 1 hr (for adults) or overnight (for PD 7) in the same fixative. The prolonged postfixation for neonatal brains was included to compensate for less complete perfusion attributable to their relatively undeveloped vasculature. Variations in the extent of fixation did not have a measurable effect on subsequent quantitative immunolabeling results within each age group. Coronal sections, 40 μm thick for adults, 60–80 μm for PD 7, were made with a vibrating microtome (Leica Microsystems), and fixation was terminated by treatment with 1% sodium borohydride in 0.1 m PB. Sections were stored in PBS, pH 7.4 containing 0.02% sodium azide at 4°C for up to 2 months before use in experiments, except for representative sections that were mounted on gelatin-coated glass slides and Nissl stained according to standard histological methods.
Pre-embedding immunocytochemistry for AMPA receptor. Pre-embedding immunocytochemistry using horseradish peroxidase (HRP)–diaminobenzidine (DAB) for visualization was chosen to ensure the most sensitive detection of AMPAR subunits. Coronal sections containing somatosensory (S1) cortex (Paxinos and Watson, 1998) were subjected to 6–10 rapid freeze–thaw cycles using 20% DMSO as a cryoprotectant to enhance antibody penetration (Wouterlood and Jorritsma-Byham, 1993). Sections were rinsed several times in PBS, incubated for 1 hr at RT in 1% bovine serum albumin (BSA)/PBS to block unspecific binding sites, and then incubated for 1–2 d at RT in a mixture of AMPAR subunit antibodies [rabbit αGluR1, 1:450, and rabbit αGluR2/3, 1:320; both purchased from Chemicon (Temecula, CA)] diluted in 1% BSA/PBS. Sections were rinsed three times with PBS, incubated for 1 hr at RT in biotinylated anti-rabbit IgG (Vector Laboratories, Burlingame, CA), and then rinsed again three times with PBS. Immunoreactivity was visualized using HRP–DAB (ABC Elite kit, Vector Laboratories). Sections were then processed immediately for electron microscopy (see below), apart from one set that was mounted on glass slides and used for light microscopy.
Tissue processing for electron microscopy. After AMPAR immunocytochemistry, sections were immersed for 1 hr at RT with 1% OsO4 in 0.1 m PB. Although many EM studies using postembedding immunocytochemistry have used osmium-free processing to preserve optimal immunoreactivity, these methods do not allow detection of HRP–DAB reaction product, and they also yield poorer ultrastructural preservation than is obtained with processing using osmium. In our laboratory, osmium treatment was compatible with optimal subsequent detection of α7nAChR by postembedding immunocytochemistry (see the following section). After osmium treatment, sections were dehydrated in stepwise concentrations of ethanol up to 70% and immersed for 1–3 d at 4°C in 4% uranyl acetate in 70% ethanol. Sections were then dehydrated and embedded in EMBED 812 between Aclar sheets (Electron Microscopy Sciences, Fort Washington, PA). Sections were then drawn under a microscope equipped with a drawing tube, taking note of prominent features (e.g., large blood vessels, pial surface, areal density of pyramidal neurons, and border of the corpus callosum) for use in later identification of cortical layers under the electron microscope. Portions of each section spanning all layers of S1 cortex were capsule embedded, and 70–90 nm sections were cut on an MT-7 ultramicrotome (RMC, Tucson, AZ) and collected on Formvar-coated grids. Grids were then subjected to postembedding α7nAChR labeling (see below) or viewed directly using a JEOL 1200 XL electron microscope (JEOL Inc., Tokyo, Japan).
Postembedding immunocytochemistry for α7nAChR. α7nAChR labeling was performed by the postembedding gold (PEG) procedure of Erisir et al. (1997) modified from that of Phend et al. (1992). Conditions were adjusted to gain optimal detection of α7nAChR by PEG while maintaining high-quality ultrastructural preservation (particularly of membranes) and preservation of HRP–DAB reaction product corresponding to AMPAR subunit immunoreactivity. Mouse monoclonal antibody 306, directed against a predicted cytoplasmic domain of α7nAChR (Schoepfer et al., 1990; McLane et al., 1992), was obtained from Research Biochemicals International (Natick, MA) and used at a dilution of 1:250. This antibody was used previously in a light microscopic study of α7nAChR distribution in adult rat brain (Dominguez del Toro et al., 1994) and more recently in an EM study of adult rat hippocampus CA1 (Fabian-Fine et al., 2001). In the latter instance, the specificity of synaptic labeling with monoclonal antibody (mAb) 306 was confirmed by the use of a polyclonal antibody directed against a region containing the same epitope; mitochondrial labeling seen with mAb 306 was shown to be a lower-specificity reaction with a species of 48 kDa, smaller than the native molecular mass of the α7nAChR subunit. Incubation time in primary antibody was 2.5 hr; the inclusion of 0.1% Triton X-100 in the incubation mixture to remove osmium and enhance antibody penetration allowed us to avoid etching steps that cause severe deterioration of ultrastructure (Phend et al., 1992). Longer incubation times (up to 18 hr) were avoided, because this did not improve the extent or specificity of α7nAChR labeling and caused detectable loss of HRP–DAB reaction product. Gold-conjugated anti-mouse secondary antibody (10 nm) was obtained from Ted Pella, Inc. (Redding, CA) and used at 1:40 dilution. Samples in which primary antibody was omitted showed virtually no labeling. After PEG, grids were postfixed for 10 min in 2% glutaraldehyde and counterstained for 20–30 sec with Reynold's lead citrate to produce acceptable levels of contrast in the section without reducing the detection of HRP–DAB reaction product. Grids were then air dried and viewed under the electron microscope.
Quantitative analysis. In samples from each age group, cortical layers were identified under the electron microscope by comparing prominent features with drawings of the flat embedded tissue (see Tissue processing for electron microscopy above) and by comparison with Nissl-stained sections. In each region, neuropil was sampled only near the tissue–resin interface where penetration of pre-embedding immunoreagents was thorough, as assessed by the overall extent of HRP–DAB labeling. Random fields were photographed at 20,000× magnification, excluding those in which cell bodies, blood vessels, or empty space occupied more than half the field. Approximately 365 μm2 of neuropil was analyzed in each cortical layer in each animal.
Micrographs were printed at a final magnification of 50,000×. Asymmetric synapses were identified, and postsynaptic elements were classified as dendritic spines or dendritic shafts, by established morphological criteria (Peters et al., 1991). Both axospinous and axon-shaft synapses were included in the analysis; most of the identifiable synapses were axospinous (see Results). A fraction of the postsynaptic profiles were not clearly identifiable as spine or shaft, particularly in the neonatal group; these profiles were included in the analysis as long as they did not clearly belong to another category (e.g., axosomatic or axoaxonal synapses). Symmetric synapses were less numerous and rarely immunolabeled. Therefore these were excluded from the quantitative analysis.
Synapses were assessed for AMPAR and α7nAChR labeling on the basis of examination of the photographs printed for comparable density and contrast; for details of the criteria used to assign degrees of labeling in particular figures, see the figure legends. Postsynaptic labeling for AMPARs was assessed by the local presence of electron-dense floccular HRP–DAB product directly over the postsynaptic density (PSD). When present, HRP–DAB product was uniformly distributed over the PSD.
Assessment of labeling by 10 nm gold particles representing α7nAChR staining took into account the distance between the gold particle and the antigen binding site; i.e., particles falling on the PSD were considered part of the postsynaptic membrane, whereas those on the synaptic cleft were assumed to belong to either the presynaptic or postsynaptic membrane with equal probability (except for data presented in Fig. 2 A, where labeling over the cleft was categorized separately). Active zone length was determined using a measuring pen (Precision Technology Devices, Skokie, IL) calibrated to the scale bar imprinted on each micrograph. To simplify analysis, perforated synapses (those with discontinuous active zones) were excluded. Variations between data sets were tested for statistical significance by ANOVA followed by unpaired two-tailed ttests (Microsoft EXCEL software); specific details are provided in Results and figure legends.
Subcellular distribution of α7nAChR in adult cortex
In adult cortex, 10 nm gold particles corresponding to α7nAChR immunoreactivity were concentrated over thick PSDs of asymmetric (typically axospinous) synapses throughout superficial, middle, and deep cortical layers (Fig.1 A,Sp1 , Sp2 ,B,C). In cases where PSDs were separated across two active zones or perforated within a single active zone, gold particles clustered specifically over PSDs and were almost absent along nonjunctional synaptic plasma membranes (data not shown). Thick PSDs with high and low densities of immunogold occurred immediately adjacent to one another (Fig.1 B). More than two-thirds of all asymmetric postsynaptic profiles were immmunoreactive for α7nAChR, and this value did not vary significantly across layers or between animals (Table 1, Adult, Total). Asymmetric synapses also showed labeling at the presynaptic membrane and in the adjacent cytoplasm (Fig. 1 B,C).
Labeling was distributed more sparsely over nonsynaptic portions of plasma membranes of dendritic spines and within the cytoplasm of dendritic spines (Fig. 1 C,arrowhead) and shafts (data not shown). In contrast, a subcellular compartment other than the PSD that showed a high density of α7nAChR was the rough endoplasmic reticulum (Fig.1 D), which was highly immunoreactive in both pyramidal and nonpyramidal cells. The predominance of immunogold labeling over PSDs, as indicated by these examples, was confirmed by quantitative analysis of the distribution pattern of colloidal gold particles within the vicinity of 420 synapses sampled across varying laminar depths of adult cortices (Fig.2). This analysis showed that nearly 50% of the synaptic gold particles aggregated directly on the PSD (Fig.2 A). Moreover, analysis of gold particle distribution along the length of PSDs revealed that α7nAChR immunoreactivity is distributed evenly, with a slightly higher concentration toward the center of the active zone captured within single ultrathin sections (Fig. 2 B, left panel).
Subcellular distribution of α7nAChR in PD 7 cortex
α7nAChR labeling in PD 7 cortex was qualitatively similar to labeling in adult cortex, with extensive labeling of dendritic elements exhibiting asymmetric synapses (Figs. 2,3 A–E) and on the endoplasmic reticulum of neuronal perikarya (Fig.3 F). Where fortuitous planes of section revealed extensions of synaptic specializations away from the postsynaptic membrane or accentuated curvatures of the PSD (Fig. 3 B,black arrow), immunogold particles followed these irregular PSD profiles.
Occasionally, gold particles clustered near plasma membranes lacking PSDs. These sites appeared to be nascent postsynaptic specializations, based on the alignment of the plasma membrane with that of the axon terminal (Fig. 3 E, profile N). Axonal labeling at or near the presynaptic membrane was also seen (note Fig.3 D in particular) but was not as prevalent as postsynaptic labeling.
The overall incidence of α7nAChR immunoreactivity at asymmetric synapses in neonates was comparable to that seen in adults in that slightly more than two-thirds of all synapses were labeled, and immunoreactivity likewise did not vary significantly across layers (Table 1, PD 7, Total). Quantitative analysis of colloidal gold particles across presynaptic to postsynaptic distances from the cleft and along the lengths of PSDs of 270 representative synapses revealed a striking resemblance to the pattern seen for adult synapses (Fig.2 A,B).
Of the α7nAChR-positive asymmetric synapses in PD 7 cortex, 66.2% were axospinous, and they most likely represent synapses onto excitatory (pyramidal or spiny stellate) neurons; 10.3% were axon-shaft, and the remainder of postsynaptic profiles was not clearly identifiable (Table 2, PD 7). These figures closely reflect the overall distribution of synapses (Table 2, PD 7, Total, α7nAChR positive plus α7nAChR negative). The proportion of α7nAChR-positive synapses that are axospinous is slightly higher in adulthood than in PD 7 cortex (Table 2, Adult and PD 7, α7(+)). The difference, which is not statistically significant, parallels an overall increase in the proportion of synapses (labeled plus unlabeled) that are axospinous (Table 2, Adult and PD 7, Total). This trend probably reflects the greater differentiation of adult cortex.
AMPA receptor distribution by light microscopy
The presence of α7nAChR at the majority of postsynaptic sites of asymmetric, presumably excitatory synapses in neonatal as well as adult cortex [see also Lubin et al. (1999)], suggested that a subset of the silent synapses (i.e., those lacking AMPAR) reported previously might indeed be subject to postsynaptic nicotinic stimulation. We examined the distribution of AMPAR in relation to α7nAChR using a combination of previously characterized antibodies against peptides corresponding to the C termini of GluR1 and GluR2/3 subunits to get an inclusive measure of the presence of AMPAR.
First, light microscopy was performed to confirm that GluR1/2/3 immunocytochemistry was adequate using tissue fixation conditions compatible with α7nAChR immunolabeling. Light microscopy of AMPAR-stained adult cortex, visualized by HRP–DAB (Fig.4 B), showed somatodendritic labeling of pyramidal and nonpyramidal neurons throughout the cortex, with large pyramidal cell bodies in layer V being especially prominent (Fig. 4 B, thick bracket). Staining was relatively intense in layer I, because of nearly complete labeling of the neuropil, and relatively light in layer IV (Fig. 4 B, thin bracket) and in the superficial part of layer VI. This laminar distribution pattern agreed with our expectations, on the basis of earlier reports on the laminar distribution of GluR1 and GluR2/3 examined separately (Petralia and Wenthold, 1992; Kharazia et al., 1996; Arai et al., 1997; Brennan et al., 1997; Petralia et al., 1997).
Labeling in PD 7 cortex showed less differentiation across layers (Fig.4 D), corresponding to the relatively undeveloped laminar pattern seen by Nissl staining (Fig. 4 C). However, a dark band of immunoreactivity was appreciable in layer I (Fig.4 D,E), and strongly stained neuronal cell bodies were present in layers II/III (Fig.4 E), with less labeling in layer IV (Fig.4 D, thin bracket). Darkly labeled cell bodies were already apparent in layer V (Fig. 4 F) and in the subplate zone (Fig. 4 G) (Herrmann, 1996; Brennan et al., 1997).
Subcellular distribution of AMPA receptors in adult cortex by electron microscopy
EM examination (Fig. 5) confirmed the somatodendritic distribution of AMPARs. Patches of AMPAR labeling were seen within dendritic shafts, along the plasma membrane or over microtubules (Fig. 5 C). HRP–DAB product occurred within dendritic spines (Fig. 5 A), and relatively less occurred in the portion of the spine head away from the PSD [Fig. 5 A,arrowhead, C, to the left andright of the spine apparatus (SA)]. In cases where spine necks were labeled (Fig. 5 C, curved arrow), labeling was also associated with the membrane of the dendritic shaft from which the spine emerged.
The intensity of PSD labeling varied considerably (Fig. 5 A, compare open arrow, curved arrow, andsmaller solid arrow). In some instances, synapses with obvious AMPAR immunoreactivity, and those below the level of detectability (referred to as AMPAR negative), could be seen in close proximity within the same membrane-bound compartment (Fig.5 B), showing that the diffusibility of HRP–DAB product did not lead to unspecific labeling of all PSDs. Most of the asymmetric synapses had detectable postsynaptic AMPAR labeling in all cortical layers examined (Table 1, AMPAR++, AMPAR+), with the proportion diminishing to 52% in the deepest layer (Table 1, Adult, Total). In agreement with previous reports (Petralia and Wenthold, 1992), presynaptic AMPAR labeling was rarely seen.
Localization of α7nAChR relative to AMPA receptors at asymmetric synapses of adult cortex
There was no significant correlation between the postsynaptic distribution of AMPAR and α7nAChR; i.e., the incidence of α7nAChR labeling at intensely and moderately AMPAR-positive synapses ranged from two-thirds to three-quarters across the layers but was not significantly different from its incidence at AMPAR-negative synapses (Table 1). Individual synapses varied widely in their α7nAChR and AMPAR content. For example, in Figure 5 A, the most strongly AMPAR-positive synapse (open arrow) is also α7nAChR positive (inset, circle), whereas the two other synapses, with less AMPAR immunoreactivity, are α7nAChR negative (straight and curved arrow, respectively). Compare this with Figure 1 B, where an intensely AMPAR-positive, α7nAChR-negative synapse (curved arrow) is seen near an AMPAR-negative synapse with abundant α7nAChR labeling (arrowheads).
Distribution of AMPA receptor relative to α7nAChR at synapses in PD 7 cortex
The relative distribution of α7nAChR and AMPAR in neonates resembled that of adult cortex (Fig. 6; compare with Fig. 5). The overall incidence of postsynaptic AMPAR labeling was comparable to that seen in adults; i.e., approximately two-thirds across all layers (Table 1). However, the intensity of AMPAR labeling in PD 7 synapses never approached that of the most intensely labeled synapses in the adult (Fig. 6 A, compare with Fig. 5 A). Colocalization of AMPAR and α7nAChR was seen postsynaptically at PSDs of asymmetric synapses (Fig.6 A) and at extrasynaptic sites (Fig.6 B,C). As in adults, the incidence of α7 labeling was not significantly correlated with the degree of AMPAR labeling (Table 1). Specifically, neither the laminar position nor AMPAR immunoreactivity was a good predictor of α7nAChR immunoreactivity. One notable point is that the synapses with intense AMPAR immunoreactivity were α7nAChR immunoreactive almost without exception. However, a substantial fraction of all synapses (22% in the adult group, 25% in PD 7, combined across all layers) were AMPAR negative but α7nAChR positive and might therefore be candidate sites for cholinergic modulation of NMDAR-containing, AMPAR-lacking silent synapses.
Quantitative analysis of postsynaptic α7nAChR immunoreactivity
A count of 10 nm gold particles corresponding to α7nAChR labeling over discrete labeled synapses confirmed our qualitative observation that the density of α7nAChR labeling was higher in neonatal than in adult synapses (Fig. 7). To simplify the analysis, we included only orthogonally sectioned synapses and used PSD length, rather than PSD area, as a measure of the size of each profile. Considerable variability in the density of labeling (Fig. 7, y-axis) and lengths of PSDs (Fig. 7,x-axis) was noted among both adult and PD 7 populations of synapses. A weak positive correlation was seen between synapse length and the number of 10 nm gold particles per synapse, regardless of the degree of AMPAR labeling present: AMPAR++, AMPAR+, and AMPAR+/− (Fig.7, compare panels). Across the three categories of AMPAR labeling, the density of α7nAChR labeling was consistently greater for PD 7 synapses than for adult synapses (Fig. 7, compare the slopes of the linear regression plots). This difference between PD 7 and adult was statistically significant within each category of AMPAR labeling (p < 0.05; two-tailed t test) and likewise within each cortical layer examined (superficial, middle, and deep, with all categories of AMPAR immunoreactivity combined; data not shown). To assess the consistency of results between experimental animals, pairwise t tests were performed among animals of the same age group and within the same cortical layer. Of 18 possible comparisons, 17 yielded no significant difference between animals (p > 0.5), whereas one pair (in PD 7, deep layers) was significantly different (p = 0.003). We did not detect significant layer-dependent differences in the density of α7nAChR labeling, and data from all cortical layers are combined in Figure 7 and subsequent figures.
We examined whether the variation in density of α7nAChR labeling was the result of random variations in the degree of labeling or whether it might reflect an underlying heterogeneity of α7nAChR density. Previous EM immunocytochemical studies have modeled the detection of antigen in ultrathin sections as a Poisson process (Kharazia and Weinberg, 1999; Racca et al., 2000). In Figure8, the density of α7nAChR labeling at discrete synapses is seen to differ in two ways from a Poisson distribution around a single mean. (1) In both adult and PD 7 data sets, there is a disproportionately large number of postsynaptic profiles without gold particles [Fig. 8, leftmost bin in each histogram; compare black bar (experimental) withgray bar (Poisson)], and this population is larger in the PD 7 group than in the adult group. (2) The PD 7 group contains a statistically distinct subpopulation of 21 synaptic profiles with a relatively high density of α7nAChR (50 or more particles per micrometer); these profiles constitute 8.9% of the total in the PD 7 data set versus 2.4% in the adult data set.
The data for PD 7 are presented as a scatter plot in Figure9, subdivided according to the degree of AMPAR labeling at each synapse. The plot shows that the group of profiles with 50 or more gold particles per micrometer consisted mostly of those with relatively short active zones (250 nm or less) and without intense AMPAR labeling. Eleven of these synapses belonged to the superficial layers, four belonged to the middle layers, and six belonged to the deep layers. Fourteen of the synapses were judged to be axospinous, whereas one (in layer VI) was axon-shaft, and six could not be classified as one or the other. Of these densely α7nAChR-labeled synapses, 14 were found in animal one, 3 were found in animal two, and 4 were found in animal three.
Prevalence of postsynaptic α7nAChR in both adult and neonatal somatosensory cortex suggests a widespread role in excitatory transmission
A large body of work documents the influence of cholinergic activity on experience-dependent plasticity in cortical receptive field properties (Sillito and Kemp, 1983; Kilgard and Merzenich, 1998). For example, acetylcholine can switch the flow of excitatory activity from columnar to intercolumnar connections through the muscarinic receptors (Xiang et al., 1998) while also boosting the activity of thalamic (glutamatergic) afferents through nicotinic receptors (Gil et al., 1997).
To the best of our knowledge, ours is the first study of the ontogeny of α7nAChR at the EM level. α7nAChR was found at most postsynaptic densities of synapses identified as glutamatergic by the presence of AMPAR (GluR1, GluR2, and GluR3) subunits and by morphology in both adults and neonates. This implies that the contribution made by α7nAChR in excitatory synaptic transmission in neocortex may be much greater than was recognized previously. We observed dually α7nAChR-positive and AMPAR-positive synapses not confined to the thalamo-recipient zone (layer IV) but also present in the supragranular and infragranular layers. Moreover, most of these synapses were axospinous. Axospinous junctions are unlikely to be on interneurons, the class previously identified as the main site of postsynaptic nicotinic activity (Alkondon et al., 1998; Frazier et al., 1998a,b) but instead are on excitatory neurons, i.e., pyramidal or stellate cells. This conclusion is strengthened by similar findings reported in adult guinea pig prefrontal cortex (Lubin et al., 1999) and in adult rat hippocampal area CA1 (Fabian-Fine et al., 2001). In the latter case, α7nAChR was detected postsynaptically at almost all asymmetric synapses examined. This study used the same antibody as the present one (mAb 306 from Sigma/RBI), but in conjunction with a different EM processing protocol (freeze-substitution followed by embedding in Lowicryl). Despite the differences in EM processing techniques, both sets of data concur in demonstrating strong postsynaptic α7nAChR immunoreactivity.
Our results and those of Fabian-Fine et al. (2001) also concur in visualizing substantial α7nAChR immunoreactivity at extrasynaptic sites within dendritic spines (e.g., in association with the spine apparatus) and on the endoplasmic reticulum within neuronal cell bodies, indicating a large reserve pool and rapid, possibly activity-dependent turnover and trafficking of α7nAChR. Our results extend previous light microscopic immunocytochemical data on α7nAChR localization to neuronal cell bodies and dendrites (Dominguez del Toro et al., 1994) and radioligand binding data obtained with α-bungarotoxin (Barrantes et al., 1995).
The activity of postsynaptic nicotinic receptors has been well documented in the peripheral nervous system (cf. Jacob and Berg, 1983;Zhang et al., 1994), and the existence of α7nAChR-mediated EPSPs was recently reported in hypothalamus (Hatton and Yang, 2002). In hindsight, it is surprising that much of the past electrophysiological evidence for postsynaptic nicotinic effects in forebrain was confined to hippocampal GABAergic (aspiny) interneurons (cf. Alkondon et al., 1998; Frazier et al., 1998a,b). Evidence for postsynaptic nicotinic activity in excitatory neurons of hippocampus (Hefft et al., 1999) and neocortex (Roerig et al., 1997; Chu et al., 2000) is comparatively sparse. A possible explanation for why cortical glutamatergic neurons have not shown more prominent α7nAChR currents by somatic recordings is that α7nAChR might act through metabolic pathways triggered by Ca2+ influx (for review, see Edwards, 1995) rather than through membrane depolarization. Alternatively, perhaps the depolarizing action of α7nAChR is confined to spines, especially on the distal, fine branches of dendrites (e.g., in layers I and II), making these currents relatively more difficult to detect in whole-cell recordings. In addition, decreases in extracellular Ca2+ during periods of high neuronal activity (Vernino et al., 1992; Amador and Dani, 1995), the relatively low agonist affinity and high desensitization rate of α7nAChR (Albuquerque et al., 1997), and the progressive glial ensheathment of excitatory synapses in maturing animals (Ling and Leblond, 1973; Parnavelas et al., 1983) could have obscured the biophysical detection of α7nAChR.
Many of the α7nAChR-enriched synapses at PD 7 are AMPA receptor-negative profiles with short active zones
Comparison with a theoretical Poisson distribution [see similar analyses by Kharazia and Weinberg (1999), Petralia et al. (1999),Takumi et al. (1999), and Racca et al. (2000)] indicates that a fraction of all asymmetric synapses was truly α7nAChR negative in both adults and neonates, whereas most synapses showed a moderate density of α7nAChR immunoreactivity. A smaller fraction, in neonates only, possessed a distinctly higher density of α7nAChR. The majority of these α7nAChR-enriched profiles appeared to represent synapses with relatively small active zones. The alternative possibility, namely that these profiles represented sections through the periphery of larger synapses, with α7nAChR concentrated toward the edge of the PSD (i.e., in a ring), was ruled out because we did not see this distribution of α7nAChR either in individual PSDs viewed en face or in a random sample of larger PSDs viewed in cross section (Fig.2 B).
On the basis of the observation that synaptic AMPAR content increases steeply with active zone length in adult S1 cortex, other authors have proposed that synapses with the shortest active zones may be essentially AMPAR negative (Kharazia and Weinberg, 1999), and potentiation of these synapses may involve the activity-dependent insertion of AMPAR (Liao et al., 1995; Shi et al., 1999). In the present study, many of the α7nAChR-enriched synapses with short active zones in the neonatal group were not detectably immunoreactive for GluR1, GluR2, or GluR3 and may indeed be functionally AMPAR negative. Such α7nAChR-enriched, AMPAR-negative synapses may rely on α7nAChR-mediated depolarization to relieve the voltage-dependent Mg2+ blockade of NMDAR. Perhaps this is the same population of synapses shown by Aramakis and colleagues (Aramakis and Metherate, 1998; Aramakis et al., 2000) to be silent (AMPAR lacking) immature synapses in which NMDA current is enhanced by an α7nAChR- or NMDAR-mediated mechanism. These results pointed clearly to a presynaptic effect (enhancement of glutamate release), whereas ours show a mostly postsynaptic concentration of α7nAChR and a lesser presynaptic concentration. This difference might be explained by a mechanism in which a postsynaptic α7nAChR-mediated influx of Ca2+ leads to the activation of the Ca2+/calmodulin-dependent neuronal nitric oxide synthase (NOS); the product, nitric oxide, is a retrograde messenger that could enhance presynaptic glutamate release (Cudeiro and Rivadulla, 1999; for review, see Contestabile, 2000). Our earlier findings on the subcellular localization of NOS support this idea, because NOS was found within NMDAR-positive spines forming asymmetric axospinous junctions (Aoki et al., 1997, 1998).
Possible interactions between α7nAChR and postsynaptic scaffolding proteins
The clustering of α7nAChR at PSDs of asymmetric synapses implies that α7nAChR is retained at the PSD through interaction with other PSD components. Studies of the neuromuscular junction have identified an extracellular protein, agrin (McMahan, 1990; Sanes and Lichtman, 1999), which is essential for nAChR clustering at the motor endplate via the transmembrane kinase MuSK (Glass et al., 1996) and the cytoplasmic protein rapsyn (Apel et al., 1997). Mice deficient in agrin showed no abnormalities in glutamatergic and GABAergic synapse formation (Serpinskaya et al., 1999). Likewise, rapsyn deficiency abolished neuromuscular but not neuronal nAChR clustering (Apel et al., 1997; Feng et al., 1998). Therefore, α7nAChR may interact with known or yet unknown scaffolding proteins of glutamatergic CNS synapses. An interaction was reported recently (Huh and Fuhrer, 2001) between α7nAChR and PICK1, a PDZ domain-containing cytoplasmic protein that promotes clustering of AMPARs (Dev et al., 1999; Xia et al., 1999). We found, however, that α7nAChR distribution at asymmetric synapses is independent of AMPAR distribution. This suggests that α7nAChR and AMPAR at synapses are not bound to a common postsynaptic scaffolding protein. α7nAChR and AMPAR may interact through PICK1 or other interacting proteins at extrasynaptic sites, e.g., during intracellular assembly and targeting.
AMPAR immunoreactivity in neonatal versus adult cortex
The fact that the most intense AMPAR immunoreactivity was seen only in adult cortex may reflect the progressive acquisition of AMPAR subunits over postnatal development (Petralia et al., 1999). In both adult and neonatal cortices, we saw wide variations in AMPAR immunoreactivity among thick PSDs, lending support to the idea that AMPAR content is an important variable in determining the strength of individual synapses.
Taken together with the existing literature, our results indicate that the role of α7nAChR in somatosensory cortex, and perhaps in all cortical areas, is pervasive. Such a role is in keeping with the widespread distribution of cholinergic afferents through the cortex and elsewhere in the brain (Woolf, 1991) and their frequent triadic relationship with glutamatergic terminals and dendritic spines (Aoki and Kabak, 1992). Mice lacking α7nAChR do not display obvious anatomical or behavioral abnormalities (Orr-Urtreger et al., 1997;Paylor et al., 1998), so it is likely that α7nAChR is necessary for refining, rather than establishing, sensory receptive field properties and other types of cortical organization dependent on precise synaptic connectivity. The widely known effects of nicotine on cognitive and attentional processes suggest that the role of α7nAChR will emerge as these high-order processes become better understood.
This research was supported by National Institutes of Health (NIH) Grants R01 NS41091 and R01 NEI 13145-01 to C.A., and NIH P30 EY13079 core grant, National Institute of Mental Health Training Grant 5T32MH195424-10, and Office of Naval Research Solicitation No. 99-019 to the Center for Neural Science. We thank Claudia Farb, Mona Lubin, and Veeravan Mahadomrongkul for discussion and technical assistance.
The work described in this paper was presented at the Society for Neuroscience 31st annual meeting, San Diego, CA, November 10–15, 2001 (abstract 362.15).
Correspondence should be addressed to Chiye Aoki, New York University Center for Neural Science, 4 Washington Place, Room 809, New York, NY 10003. E-mail:.