Abstract
Subunit proteins that comprise functional AMPA receptors were localized by immunocytochemical methods in the adult macaque primary visual cortex (V1). GluR1, GluR2/3/4c, and GluR4 immunoreactivity consisted of rich plexuses of punctate profiles scattered throughout the neuropil, in radial arrays, and outlining the membrane of somata and proximal dendrites. Cytoplasmic immunoreactivity was limited. GluR2/3/4c immunostaining was more prominent along the somata surface and exhibited greater levels of cytoplasmic immunoreactivity than GluR1 and GluR4 immunostaining. The density of AMPA subunit immunoreactive elements also varied across layers and compartments of macaque V1. Immunoreactivity for GluR1, GluR2/3/4c, and GluR4 was densest in three bands that corresponded to layers IVA, IVC, and VI. Immunostaining for each subunit was also unevenly distributed within many of the layers. In layers II–III, patches of intense immunostaining coincided with cytochrome oxidase (CO)-rich blobs. In layer IVA, intense subunit staining formed a conspicuous honeycomb pattern. In layer IVC, subunit staining formed a radial lattice. GluR2/3/4c subunit immunostaining was also preferentially distributed within the CO-rich blobs of layers V–VI. These findings demonstrate that AMPA subunit immunoreactivity is densely concentrated in layers and compartments receiving direct geniculocortical innervation. This distribution, which differs from that of excitatory synapses, suggests that the density of AMPA receptors is unevenly distributed at synaptic and possibly extrasynaptic sites within macaque visual circuits.
Primary visual cortical circuits use excitatory, chemically mediated synaptic signals for transmitting and generating the wide variety of functional properties required for sensory analysis (Streit, 1984; Tsumoto, 1990; Carder and Hendry, 1994). Geniculocortical afferents form three parallel, but largely separate, routes through which information from the retina is passed via the dorsal lateral geniculate nucleus to the primary visual cortex (V1) (Blasdel and Lund, 1983; Freund et al., 1989; Hendry and Yoshioka, 1994). Excitatory activity arriving through each of the thalamic pathways is integrated with the activity of a series of layer-specific cortical connections that progressively transform and relay visual information. The majority of these intrinsic circuits are also excitatory, accounting for 75–85% of the total synaptic connections any single neuron receives (Beaulieu et al., 1992). Thus, input to cortical cells is primarily excitatory and comes not only from a patterned thalamic input, but from a massive convergent input from other excitatory cortical cells.
Clearly, each excitatory input is likely to have a different role in shaping the response properties of a neuron. How a neuron integrates incoming signals will depend not only on the cellular geometry and intrinsic membrane properties of the cell, but also on the shape and size of the synaptic current. The excitatory neurotransmitter glutamate can act through a variety of postsynaptic receptors, including the cation-specific ion channels; AMPA, NMDA, and kainate receptors; and G-protein-coupled metabotropic glutamate receptors (Collingridge and Lester, 1989; Monaghan et al., 1989; Schoepp et al., 1990). Multiple classes of excitatory receptors can be activated at a given synapse with significant variations in the contribution of each class (Huettner and Baughman, 1988; Bekkers and Stevens, 1989; Jones and Baughman, 1991; Stern et al., 1992). Even at synapses using the same receptor class, the physiological properties are not necessarily equivalent (Colquhoun et al., 1992; Stern et al., 1992; Hestrin, 1993; Livsey et al., 1993; Koh et al., 1995). The recent identification of families of genes coding for glutamate receptor subunit proteins has revealed that each receptor family is composed of subunits that can assemble in various combinations to form ligand-gated channels with unique pharmacological, kinetic, and gating characteristics (Hollmann et al., 1989, 1991; Keinänen et al., 1990; Nakanishi et al., 1990). Variations in the receptor classes and subtypes expressed, their densities, and their placement are likely to influence how a neuron integrates a particular incoming signal.
As an initial step in understanding the relationship between glutamate receptor expression and synaptic communication in different parts of the visual cortical circuit, the present study examined the distribution of AMPA receptors in macaque V1. AMPA receptors are endowed with kinetic properties that allow them to transmit fast repetitive signals and are therefore used at most glutamatergic synapses in the vertebrate CNS (Collingridge and Lester, 1989). Using AMPA subunit immunoreactivity as a marker, one potentially specific for AMPA receptors, the present study reports that rich plexuses of punctate profiles immunoreactive for each of the subunits were unevenly distributed across layer and compartments of macaque V1. Implications of this subunit organization are discussed.
MATERIALS AND METHODS
Ten normal adult monkeys (Macaca fascicularis) were used in this study. All animals were killed with an overdose of Nembutal, followed by perfusion through the heart with 2–4% paraformaldehyde and 0–0.1% glutaraldehyde in 0.1 mphosphate buffer. The occipital lobes from each monkey were cut into sagittal blocks that included V1. Blocks were sunk in 20–30% sucrose/phosphate buffer solution at 4°C.
Sections were cut serially on a sliding microtome. Some blocks of occipital cortex were cut sagittally. Others were flattened while freezing and were cut parallel to the opercular surface of the occipital lobe. Sections of varying thickness were cut and processed (alternating 15 and 30 μm or 20 and 40 μm). The thicker sections in each series were reacted histochemically for cytochrome oxidase (CO) (Wong-Riley, 1979) or stained with thionin. Thinner sections were processed immunocytochemically for AMPA-selective subunits with commercially available rabbit antisera directed against synthetic peptides corresponding to C-terminal sequences of GluR1 (KMSHSSGMPLGATGL), GluR2/3/4c (KQNFATYKEGYNVGIESVKI), and GluR4 (KHTGTAIRQSSGLAVIASDLP) (Chemicon, Temecula, CA, and Upstate Biotechnology, Lake Placid, NY). Because these antibodies have been raised against peptides near the carboxyl termini, they recognize both the flip and flop versions of the GluR subunits. The characterization and immunocytochemical application of these antibodies has been reported by Wenthold et al. (1992), Petralia and Wenthold (1992), andMartin et al. (1993a,b). A wide range of dilutions were tested for each. Sections were preincubated in 0–0.05% Triton X-100 and 3–10% normal serum in 0.1 m phosphate buffer (dilution buffer). In some cases, 1% nonfat dried milk was added to the dilution buffer. After 3–5 hr, sections were transferred to a solution containing the primary antibody and 3–10% normal serum in 0.1 mphosphate buffer and incubated for 48–72 hr at 4°C. Subsequently, the sections were processed by the avidin–biotin–peroxidase method (Vector Laboratories, Burlingame, CA; Sigma ImmunoChemicals, St. Louis, MO) or the peroxidase anti-peroxidase method (DAKO, Carpinteria, CA) and reacted in 3,3′-diaminobenzidine tetrahydrochloride and hydrogen peroxidase. Some immunostained sections were osmicated. Sections in which the normal rabbit serum (1:1000) was substituted for the primary antisera or the primary antisera was omitted served as controls. No signal was produced with these procedures.
RESULTS
Laminar distribution
Immunoreactivity for the GluR1, GluR2/3/4c, and GluR4 subunits was present throughout the thickness of macaque V1. The intensity of immunostaining varied across layers, forming three bands that contained high levels of immunoreactivity, three bands that formed intermediate levels of immunoreactivity, and one band that contained a low level of immunoreactivity (Fig. 1). When compared with the pattern of CO (Fig. 1A) in adjacent sections, the dense bands were determined to include layers IVA, IVC, and VI; the intermediate bands II–III, IVB, and V; and the low band layer I.
Immunoreactivity for each of the subunits was located predominantly in the neuropil and consisted of a rich plexus of punctate profiles (Fig.2). Most of the immunostained profiles were small circular elements, less than 1 μm in diameter, but some were larger and irregularly shaped. Immunoreactive elements were scattered throughout the neuropil (Fig. 2) but were also frequently found in radial clusters, along the length of apical dendrites (Fig.2D,E), and surrounding the somata and proximal processes of neurons (Fig. 2A–C). Some cells displayed cytoplasmic immunoreactivity (Fig. 2C). GluR2/3/4c immunostaining associated with the cell surface and/or cytoplasm was always greater than that found with GluR1 or GluR4 immunostaining.
The density of different immunostained elements varied throughout the layers of V1. In supragranular layers, GluR1 and GluR2/3/4c immunostaining consisted of scattered fine and coarse puncta, and dense immunostaining surrounding the somata of neurons. GluR2/3/4c immunostaining differed from that of GluR1 staining in that coarse puncta were more evident and a greater number of somata displayed immunoreactive puncta along their surface (Fig.3A,B). Immunoreactivity for GluR4 consisted of fine puncta scattered throughout the neuropil, and along the surface of a limited number of pyramidal neurons and apical dendrites (Fig.3C). In layer IVC, immunostaining for GluR1, GluR2/3/4c, and GluR4 consisted of a radial lattice in which immunostained puncta formed a meshwork of small radial clusters separated by unstained regions (Fig. 4). In general, immunostaining associated with the cell surface and cytoplasm was less prominent in this layer. Layer V subunit immunostaining was associated mainly with puncta scattered throughout the neuropil, whereas layer VI subunit immunostaining was characterized by high densities of cell surface and cytoplasmic immunoreactivity, as well as scatter puncta.
The density of immunostained punctate profiles was decreased in tissue fixed with glutaraldehyde, whereas cytoplasmic immunostaining was significantly increased. All specific immunoreactivity was eliminated by either omitting the primary antibody or substituting normal rabbit serum for the primary antibody.
Compartmental distribution
In tangential sections through layers II–III, intense immunoreactivity for GluR1, GluR2/3/4c, and GluR4 subunits was distributed in periodic patches that measured 150 × 250 μm with a center-to-center spacing of 400–600 μm (Fig. 5). Comparison of the immunostaining for each subunit with the histochemical stain CO (Fig. 5A) showed that immunostained patches corresponded to CO-rich blobs. Within CO blob regions, scattered immunoreactive punctate profiles, as well as puncta surrounding somata, appeared to be larger and/or more intensely stained than the profiles in the surrounding interblob regions. The resolution was such that it could not be determined whether an increase in the overall density of punctate profiles also contributed to the increase in immunostaining.
Periodic patches of immunostaining for GluR2/3/4c subunits corresponding to stripes of CO-rich blobs were also evident in layers V and VI (Fig. 6). Immunoreactive puncta and cytoplasmic staining appeared to be larger and/or more intense in CO-rich blob regions than profiles found in the surrounding interblob regions.
In layer IVA, GluR1, GluR2/3/4c, and GluR4 immunostaining formed a lattice consisting of intensely stained walls surrounding lacunae of less intense immunostaining (Figs. 7A,8). The CO staining in an adjacent section revealed a similar lattice, in which intensely stained walls surrounded weakly stained lacunae (Fig. 7B). Direct comparison by aligning the same radial blood vessels confirmed that the CO-stained and subunit-immunostained walls occupied the same regions. Puncta within the walls were often larger (GluR1 and GluR2/3/4c) and more intensely stained (GluR1, GluR2/3/4c, and GluR4) than puncta within the lacunae. Cells either exhibiting cytoplasmic GluR2/3/4c immunoreactivity or outlined with puncta immunoreactive for GluR1 and GluR2/3/4c were prominent in the lacunae but not apparent in the walls of the honeycomb.
DISCUSSION
Specificity of subunit immunostaining
An important issue for interpreting the patterns of AMPA subunit immunoreactivity is the identity of the immunostained elements. Several lines of evidence suggest that the extremely rich immunostaining of fine punctate profiles observed throughout the neuropil represents synaptic and extrasynaptic AMPA receptors. Glutamate receptor subunits have been shown to form aggregates that are enriched in synaptic membranes (Wenthold et al., 1990, 1992; Rogers et al., 1991; Blackstone et al., 1992; Hullebroeck and Hampson, 1992). These subunits cluster at specific postsynaptic sites (Craig et al., 1993) that correspond to hot spots of functional AMPA receptor activity (Bekkers and Stevens, 1989;Jones and Baughman, 1991). Extrasynaptic receptors have also been identified using high-resolution subunit localization and outside-out patch-clamp recordings of glutamate-gated channels in the soma and dendrites of both neurons and glia (Jones and Baughman, 1991; Hestrin, 1993; Martin et al., 1993a,b; Molnár et al., 1993; Baude et al., 1994). The present findings now demonstrate anatomical localization of the subunits that form AMPA receptors. Immunostaining consists of discrete puncta scattered throughout the neuropil, as well as along the surface of somata. Although the precise subcellular localization of AMPA subunits requires further characterization, immunoelectron microscopy using the same or similar antibodies has confirmed that subunit immunolabeling is present in postsynaptic densities of rat (Petralia and Wenthold, 1992; Martin et al., 1993b; Molnár et al., 1993; Baude et al., 1994; Phend et al., 1995) and primate (Martin et al., 1993a; our unpublished observations). Taken together, these data strongly suggest that the immunocytochemical methods used in this study recognize preferentially AMPA subunits that are assembled into receptors.
These results contrast with previous light microscopic studies using the same antibodies in which subunit immunostaining was preferentially localized to intracellular compartments of pyramidal cell somata and proximal apical dendrites (Petralia and Wenthold, 1992;Martin et al., 1993a,b; Vickers et al., 1993; Conti et al., 1994). The intracellular accumulation of reaction product could be related to various stages of synthesis, transport, assembly, and degradation of receptor subunits, but not to the number or location of actually expressed receptor molecules. GABAA receptor subunit studies demonstrate that the level of intracellular immunoreactivity may not be well correlated with levels of immunoreactivity on the plasma membrane (Somogyi, 1989). Technical issues may explain the differences observed in subunit protein localization. Although the presence of glutaraldehyde, high percentage of paraformaldehyde, and lengthy perfusion times may preserve more labile cytoplasmic pools, they can also produce excessive cross-linking, resulting in the failure of the antiserum to reach antigenic sites (Griffiths, 1993). The limited punctate immunolabeling reported in previous studies could be attributed to steric hindrance of epitopes associated with synaptic membrane. Increased levels of cytoplasmic immunostaining and decreased levels of immunostained puncta were also observed in glutaraldehyde-processed tissue in the current study.
AMPA subunit organization in macaque V1
A striking result was the marked variability in the density of AMPA subunit immunoreactivity across layers and compartments of macaque V1. Assuming that intensity of immunostaining is directly related to subunit density leads to the prediction that AMPA subunit immunostaining would follow the distribution of asymmetric or excitatory synapses in macaque V1. This was not the case, however, as immunostaining was more intense in the CO-rich blobs of layers II–III than in surrounding interblobs, even though these two compartments contain the same density of asymmetric synapses (Beaulieu et al., 1992). Immunostaining was more intense in layers IVC and VI than other cortical layers, even though these two layers contain significantly less asymmetrical synapses (Beaulieu et al., 1992). These qualitative differences in the present study cannot be explained by limited subunit recognition, because the antibodies used recognize the majority of AMPA subunits known to be expressed in the CNS (Hollmann and Heinemann, 1994). One explanation for the current results is that the number of AMPA receptors differs at individual synapses. This hypothesis is supported by anatomical and physiological evidence demonstrating that the density of subunit immunostaining (Nusser et al., 1994; Siegel et al., 1994) and the ratio of AMPA/kainate to NMDA receptors (Huettner and Baughman, 1988; Stern et al., 1992; Thomson and Deuchars, 1994) varies at different synapses.
The distribution of AMPA subunits observed in these experiments strongly suggests that thalamocortical synapses may preferentially express a high density of AMPA receptors. The intense immunostained puncta in layer IVA forms a pattern identical to that observed with CO staining (Livingstone and Hubel, 1982; Horton, 1984), to the distribution of geniculocortical axon terminals (Hendrickson et al., 1978; Livingstone and Hubel, 1982; Blasdel and Lund, 1983; Itaya et al., 1984), and to the organization of horizontally aligned, thin dendrites that surround cones of neurons (Peters and Sethares, 1991). Similarly, the intense immunostained puncta in layer IVC forms a pattern of alternating intensely stained stripes and poorly stained stripes that can be related to geniculocortical afferent terminations (Hubel and Wiesel, 1972; Hendrickson et al., 1978; DeFelipe and Jones, 1991), clusters of dendrites (Peters and Sethares, 1991), and glutamate immunoreactive processes (Carder and Hendry, 1994). This hypothesis is also supported by a large body of literature indicating that excitatory neurotransmission at thalamocortical synapses is primarily mediated by AMPA receptors (Hagihara et al., 1988; Shirokawa et al., 1989; Fox et al., 1989, 1992; Nishigori et al., 1990; Gil and Amitai, 1996). Despite the fact that geniculocortical afferent terminals may represent less than 20% of the asymmetrical synapses in layer IVC (Garey and Powell, 1971; Tigges and Tigges, 1979; Peters et al., 1994), physiological evidence indicates that thalamic input plays an essential role in driving cortical cells (Tanaka, 1983; Ferster et al., 1996). These data suggest that the strength of a connection may depend on diverse mechanisms, such as the large size of thalamic synapses and increased concentration of AMPA receptors, rather than the number of synapses formed by a pathway. Preliminary electron microscopic evidence indicating that thalamocortical synapses express a higher density of AMPA subunits than corticocortical synapses provides support for this idea (Weinberg and Kharazia, 1996) (R. Weinberg, personal communication).
Other synaptic and extrasynaptic receptors also might contribute to the observed staining pattern. The relative contribution of AMPA and NMDA receptors at the majority of excitatory synapses in layer IVC is unknown. Excitatory synapses onto GABA-immunopositive postsynaptic neuronal elements are mediated primarily by AMPA receptors (Thomson and Deuchars, 1994). Although inhibitory interneurons are preferentially distributed in geniculocortical recipient layers, excitatory synapses onto inhibitory interneurons are not (Beaulieu et al., 1992). Although there is considerable evidence that subunit expression is related to cell type (Vickers et al., 1993; Bochet et al., 1994; Jonas et al., 1994; Geiger et al., 1995), the pattern of subunit immunostaining observed in the present study does not correspond to the numerical densities of neurons, astrocytes, oligodendrocytes, and microcytes (O’Kusky and Colonnier, 1982). However, there is widespread evidence that receptor expression is influenced by local environmental cues (Gall et al., 1990; Pasqualotto et al., 1993; Audinat et al., 1994;Bessho et al., 1994; DeFelipe et al., 1994; Kamphuis et al., 1994;Kraus et al., 1994; Perez-Velazquez and Zhang, 1994; Ehlers et al., 1995). Thus, synaptic and/or extrasynaptic expression of AMPA receptors may also be influenced by the milieu of particular layers and compartments, rather than being solely determined by cell type.
Differences in AMPA subunit distributions
Although results from the present study are in agreement with physiological studies demonstrating the predominant use of AMPA receptors in layers receiving geniculocortical afferents, radioligand binding studies report low levels of binding in these same layers (Shaw and Cynader, 1986). Variations in AMPA subunit composition might be important for explaining this inconsistency. Homeric AMPA receptors are activated by kainate and to a lesser extent by AMPA, yet exhibit high-affinity [3H]AMPA but not [3H]kainate binding sites (Boulter et al., 1990; Keinänen et al., 1990;Hollmann et al., 1991). Heteromeric AMPA receptors generally show larger current amplitudes but lower potencies than homomeric receptors (Boulter et al., 1990, Nakanishi et al., 1990). Subunit interaction may exacerbate these discrepancies between the pharmacological properties and the physiological response (Boulter et al., 1990; Nakanishi et al., 1990; Sakimura et al., 1990). The technical difficulty of measuring tritiated ligand binding at the very low affinity levels that are sufficient to produce electrophysiological responses in heteromeric receptors, along with the greater sensitivity of immunocytochemical methods, may partly explain the low levels of receptor binding and high levels of subunit immunostaining in layer IVC.
Subtle differences in the laminar, tangential, and subcellular localization of the receptor subunits suggest that both synaptic and extrasynaptic AMPA receptors found in V1 represent heterogeneous populations. It is assumed that where these differences exist the physiological and pharmacological properties of the AMPA receptors will vary. Studies correlating the functional properties of AMPA receptors with the specific mRNA at the single-cell level indicate that expression of GluR2 determines the formation of receptors with relatively slow gating properties and high calcium permeability, whereas the GluR1 and GluR4 subunits promote assembly of more rapidly gated receptors with low calcium permeability (Bochet et al., 1994;Jonas et al., 1994; Geiger et al., 1995). From the results of the present study, it could be argued that cells in the CO blobs of layers V–VI express greater densities of GluR2/3/4c than cells in the interblobs and are therefore more likely to express slow AMPA receptors with little calcium permeability. Even in layers and compartments in which each of the subunits is densely expressed, differences in AMPA receptors are still likely to exist. For example, GluR1-immunostained puncta in the walls of the IVA lattice are densely distributed and homogeneously expressed compared with the nonuniform distribution of large, intensely GluR2/3/4c-immunostained puncta in this same compartment. These data suggest that a large portion of the AMPA receptors in the walls of the IVA lattice is composed of the GluR1 subunit, whereas a smaller subset is composed of predominantly the GluR2/3/4c subunit. GluR2/3/4c immunostaining is also more prominent along the cell surface of pyramidal neurons than GluR1 and GluR4 immunostaining, suggesting that the functional properties of AMPA receptors in the postsynaptic membrane could differ from those in the somatic membrane patches.
The physiological importance of AMPA receptor differences remains elusive. First, the repercussions of certain channel properties are unknown. For example, it remains to be shown whether the additional route of synaptically mediated Ca2+ entry may be linked to an increase in synaptic efficacy. Alternatively, Ca2+ entry through AMPA receptors could activate Ca2+-dependent K+ channels or could lead to the inactivation of NMDA receptors (Medina et al., 1994). Second, although differences in the relative proportions of distinct subunits across the cortical layers indicate unique pharmacological, kinetic, and gating characteristics of AMPA receptors in each layer, the physiological relevance of the expression of specific AMPA receptor subtypes is dependent on its cellular and subcellular localization. Thus, activation of a particular receptor subtype will lead to different functional consequences depending on where the receptor is operating.
Footnotes
I thank Priya Swamy and Dana Yoo for technical assistance, Dr. Stewart Hendry for providing the monkey material, and Drs. David Calkins, Stewart Hendry, Lucia Galli-Resta, and Michael Steinmetz for valuable criticism of this manuscript.
Correspondence should be addressed to Renee K. Carder, Department of Neurology MC 2030, University of Chicago, 5841 South Maryland Avenue, Chicago, IL 60637.