GIRK1 immunoreactivity is present predominantly in dendrites, dendritic spines, and somata in the CA1 region of the hippocampus

Abstract

Electron microscopic analysis of the CA1 region of the rat hippocampus revealed that specific immunoreactivity (IR) for a G protein-gated, inwardly rectifying potassium channel (GIRK1) was present exclusively in neurons and predominantly located in spiny dendrites of pyramidal cells. Within stratum lacunosum-moleculare and the superficial stratum radiatum, GIRK1-IR was often present immediately adjacent to asymmetric (excitatory-type) postsynaptic densities in dendritic spines. The subcellular localization of GIRK1-IR in the Golgi apparatus of pyramidal cell somata and in the plasma membrane of dendrites and dendritic spines confirms the hypothesis that GIRK1 is synthesized by pyramidal cells and transported to the more distal dendritic processes. G protein-coupled receptor activation of a dendritic potassium conductance would attenuate the propagation of excitatory synaptic inputs and thereby produce postsynaptic inhibition. Thus, these results show that the GIRK family of channels joins the list of voltage-sensitive channels now known to be expressed in dendritic spines.

Potassium channels have dominant control of the membrane potential of mammalian neurons (1). Modulation of potassium channel activity, therefore, can profoundly influence neuronal excitability via changes in the resting membrane potential, action potential duration, firing rate, and neurotransmitter release and by indirectly controlling voltage-gated channels. Potassium channels can be modulated indirectly by a variety of soluble second messengers (for reviews see refs. 1 and 2) as well as directly by G protein α (3) and βγ (4–8) subunits. Many different neurotransmitters, including acetylcholine, adenosine, opioids, γ-aminobutyric acid (GABA), norepinephrine, somatostatin, serotonin, and dopamine act through their respective receptors and G proteins to activate G protein-coupled potassium channels (see refs. 9 and 10).

A family of G protein-activated, inwardly rectifying potassium channels (GIRKs) has been recently cloned (11–13) and shown to form functional, heteromultimeric K⁺ channels (7, 14–16). One member of this family, GIRK1, has been previously localized throughout the brain using both in situ hybridization (17) and light microscopic immunocytochemistry (18, 19). We chose the hippocampus for further anatomical localization, because extensive anatomical and physiological studies previously performed in this brain region make it a useful model for studying GIRK-mediated functions. Although high levels of GIRK1 mRNA were found in the principal cells of the hippocampal formation (i.e., the pyramidal and granule cells; refs. 17 and 20), light microscopic studies showed that the somata of these neurons contained weak (19) or no significant (18) GIRK1-immunoreactivity (IR). In contrast, higher levels of diffuse GIRK1-IR were seen in fields that contain the dendrites of pyramidal and granule cells: stratum (str.) radiatum, str. oriens, and particularly str. lacunosum-moleculare of the hippocampus and str. moleculare of the dentate gyrus. Together, these in situ and immunocytochemical data strongly suggest that GIRK1-IR may be located in the dendrites of pyramidal and granule cells (18, 19). Alternatively, this labeling may be attributable to the presence of GIRK1 in terminal fields of afferents to the hippocampal formation (e.g., from the entorhinal cortex and thalamus). The latter suggestion was supported by immunolabeling of hippocampal cultures, which failed to demonstrate the localization of GIRK1-IR in pyramidal cell dendrites (18) and also by a recent study showing decreases in GIRK1-IR following lesions of the thalamus (19). Therefore, the aim of the present study was to determine the subcellular distribution of GIRK1-IR in the CA1 region of the rat hippocampus using electron microscopy with a previously characterized anti-peptide antibody directed against the carboxyl-terminal tail of GIRK1 (18). A higher-resolution anatomical localization of the channel should provide important clues to neuronal functioning by defining the cell types expressing the channel and the subcellular localization of the protein within the neuron.

MATERIALS AND METHODS

GIRK1 Antibody.

We used a previously characterized (18), affinity-purified rabbit antibody (KG-6, no. 2708) raised against the synthetic peptide sequence (TRMEGNLPAKLRKMNSK) corresponding to residues 482–498 of the cloned rat GIRK1 channel, at the predicted intracellular carboxyl terminus. Although this antibody does not cross react with any other known protein (18), the staining is conservatively described as GIRK1-IR.

Immunocytochemistry.

Adult (300–400 g; n = 8) male Sprague–Dawley rats were deeply anesthetized with sodium pentobarbital and perfused through the ascending aorta sequentially with the following: (i) 10–20 ml of heparin-saline, (ii) 3.75% acrolein and 2% paraformaldehyde in 0.1 M phosphate buffer (PB; pH 7.4), and (iii) 2% paraformaldehyde in PB. The brains were removed from the skull, cut into 5- to 6-mm-thick blocks, and postfixed in the last fixative for 30 min. Coronal sections (40 μm thick) were cut through the hippocampal formation on a vibrating microtome (Vibratome; Technical Products International, St. Louis), collected in PB, and pretreated with 1% sodium borohydride to reduce reactive aldehydes (21).

Sections were immunocytochemically labeled using the immunoperoxidase technique as previously described in detail (22). Briefly, for light microscopy, tissue from three rats was incubated as follows: (i) for 12–18 hr at room temperature in affinity-purified anti-GIRK1 antibody (0.2–0.3 μg/ml) diluted in 0.1 M Tris-saline buffer (TS; pH 7.6) containing 0.1% BSA and 0.25% Triton X-100; (ii) biotinylated goat anti-rabbit IgG (Vector Laboratories) diluted 1:400 in TS with 0.1% BSA; and (iii) avidin–biotin–peroxidase complex (Vectastain Elite ABC kit, Vector Laboratories) at twice the recommended dilution in TS. Labeling was made visible by incubating sections in PB containing 3,3′-diaminobenzidine (DAB) and hydrogen peroxide.

For electron microscopy, sections from five animals were processed using the immunoperoxidase technique as described above, except in place of 0.25% Triton X-100, either 0.035% Triton X-100 was substituted or tissue was briefly freeze-thawed to enhance antibody penetration (21). Other sections from three rats were processed with the immunogold procedure as described by Chan et al. (23). For this procedure, which is less sensitive than immunoperoxidase but allows more precise subcellular localization of the immunoreactive sites (23), the sections were incubated in a higher concentration of anti-GIRK1 (0.8 μg/ml). The sections were transferred to 0.01 M phosphate-buffered saline (PBS; pH 7.4), then incubated in PBS containing 0.1% gelatin and 0.8% BSA for 30 min. Sections were incubated 2 hr in goat anti-rabbit IgG conjugated to 1-nm gold particles (Amersham) diluted 1:50 in PBS–gelatin–BSA. The sections were rinsed in PBS, incubated 10 min in 2% glutaraldehyde, transferred to 0.2 M sodium citrate buffer (pH 7.4), and gold particles were silver-intensified (Inten-SEM kit; Amersham) for 6–8 min.

Immunocytochemical controls included: (i) omitting the primary antibody, (ii) substituting preimmune sera or preimmune IgG (both at 0.2 μg/ml) for the primary antibody, and (iii) preadsorbing the primary antibody with the antigenic peptide (10 μM) for 24 hr at 4°C before incubating the tissue. For light microscopy, immunolabeled sections were mounted onto gelatin-coated glass slides, dried, dehydrated through graded ethanols and xylenes, and coverslipped in DPX mounting medium (Aldrich). The hippocampal formation was examined and photographed on a Nikon Microphot microscope using brightfield and differential interference contrast optics. For electron microscopy, labeled sections were fixed in 2% osmium tetroxide for 1 hr and embedded in EMbed 812 (Electron Microscopy Sciences, Fort Washington, PA) as described previously (24). Immunolabeled regions of the CA1 region were taken from six immunoperoxidase-labeled and six immunogold-labeled sections and glued onto Epon blocks, and ultrathin sections (≈50 nm) were collected on copper grids. Ultrathin sections were counterstained with uranyl acetate and lead citrate (25), then examined and photographed on a Philips 201 electron microscope (Philips Electronic Instruments, Mahwah, NJ).

To determine whether gold particles were targeted to certain regions along the plasmalemma of dendritic spines, fields in str. radiatum and lacunosum-moleculare of 21 ultrathin sections taken from two GIRK-immunogold labeled vibratome sections were examined, and all labeled (two or more gold particles) spines forming synapses were photographed (n = 35). The spine plasmalemma was divided into three compartments: postsynaptic density (psd), perisynaptic region (defined as the 100 nm of membrane adjacent to the edge of the psd), and extrasynaptic membrane. The membrane length and number of membrane-associated gold particles in each compartment was measured, and the density of gold particles per micrometer of membrane was calculated for each spine profile. The densities in each compartment were compared using ANOVA and are expressed in the text as mean ± SEM.

RESULTS

As previously reported for this (18) and another GIRK1 antibody (19), light microscopic analysis revealed that GIRK1- IR was heterogeneously distributed in the hippocampal formation (Fig. 1). In the CA1 region of hippocampus, str. lacunosum-moleculare had the most intense labeling, followed in intensity by the adjacent (“superficial”) portion of str. radiatum and by str. pyramidale, while str. oriens and the deep portion of str. radiatum had very light to no labeling. When examined at higher magnification, GIRK1-IR within str. lacunosum-moleculare and radiatum appeared diffuse, and in str. pyramidale it outlined but did not fill the cell bodies (Fig. 1B). Very rarely, somata outside of str. pyramidale were also outlined (data not shown). The CA3 region contained a similar pattern of GIRK1-IR, except that labeling in str. oriens was denser than that in CA1 (Fig. 1A). In the dentate gyrus, GIRK1-IR was also laminated; diffuse labeling was intense in the outer half of str. moleculare, moderate in str. granulosum, and sparse in the hilus (Fig. 1A). In str. granulosum, many granule cells were outlined, and very few (on average, less than one per 40-μm section) non-granule cells were labeled in the hilus or str. moleculare. The rare occurrence of nonprincipal cell labeling contrasts with the more frequent occurrence of such labeling observed with this antibody in a previous study (18) and is presumably due to the different fixation and processing conditions used.

Figure 1.

Light microscopic distribution of GIRK1 IR in the hippocampal formation. (A) GIRK1-IR is present in some strata within the dentate gyrus (DG) and hippocampal regions CA1 and CA3. In the CA1 region of the hippocampus, GIRK1-IR is most dense in str. lacunosum-moleculare (l-m), followed by the adjacent portion of str. radiatum (r). Lighter labeling is present in str. pyramidale (p), and very light labeling is present in deep str. radiatum and in str. oriens (o). In the CA3 region, the pattern of GIRK1-IR is similar to that in CA1, except that the labeling is much denser in str. oriens. Within the dentate gyrus, GIRK1-IR is strongest in the outer portions of str. moleculare (m), moderate in str. granulosum (g), and light in the hilus (h). (Bar = 450 μm.) (B) High magnification of the CA1 region, showing the diffuse appearance of GIRK1 labeling in str. lacunosum-moleculare, radiatum, and oriens and the outlining of some pyramidal cell somata (arrows) in str. pyramidale. (Bar = 15 μm.)

Electron microscopic analysis of the CA1 region of the hippocampus revealed that GIRK1-IR was detected exclusively in neurons. The majority of labeled profiles were dendrites (Figs. 2 and 3 B and C); out of 324 nonsomatic profiles photographed in str. lacunosum-moleculare, radiatum, and oriens, 316 were dendrites, 8 were axons or axon terminals, and none were glia. In addition, a few labeled somata were observed in str. pyramidale. The immunolabeling was specific as determined by comparison with control sections for which the primary antibody had either been preadsorbed with cognate peptide, omitted, or substituted with preimmune IgG. Dendritic GIRK1-IR was present both in shafts (Figs. 2C and 3 B and C) and in spines (Fig. 2 A and B); in fact, dendritic spines were often labeled even when the parent dendrite contained little or no detectable GIRK1 labeling (Fig. 2A). Labeled spines were observed more frequently than labeled shafts: counts of GIRK1-peroxidase labeled dendritic profiles in randomly selected fields in str. lacunosum-moleculare and radiatum revealed that 72% (144/199) were dendritic spines and 28% (55/199) were dendritic shafts.

Figure 2.

Subcellular localization of GIRK1 IR in dendrites and dendritic spines in the CA1 region of hippocampus. (A) A dendritic spine in str. radiatum contains GIRK1–immunogold labeling (arrows). The labeled spine forms an asymmetric synapse (curved arrow) with an unlabeled axon terminal (uT). (B) A dendritic spine head with GIRK1–immunoperoxidase labeling (arrows) forms a synapse (curved arrow) with an unlabeled axon terminal (uT) in superficial str. radiatum. An adjacent unlabeled spine (uS) synapses with the same unlabeled terminal. (C) A dendrite labeled with GIRK1–immunoperoxidase in str. lacunosum-moleculare. The labeled dendrite forms a synapse (curved arrow), which is partially obscured by peroxidase reaction product, with an unlabeled terminal (uT). (Bar = 0.5 μm.)

Figure 3.

Subcellular distribution of GIRK1-IR in somata and dendrites in the CA1 region of hippocampus. (A) GIRK1–immunogold labeling (arrows) in a pyramidal cell soma is localized to the Golgi apparatus (G), small vesicular structures, and, less frequently, the plasma membrane (arrowhead). A gold particle is present in the nucleus (N), but this labeling was nonspecific. (B) In superficial str. radiatum, two dendritic shafts (D1 and D2) contain GIRK1–immunogold labeling (arrows). This labeling is associated primarily with the plasma membrane and to a lesser extent with intracellular organelles. Dendrite D1 possesses a thin, unlabeled simple spine (S) and receives a symmetric synapse on the shaft (curved arrow) from an unlabeled terminal (uT). (C) Dendrites in str. lacunosum-moleculare contain GIRK1–immunogold labeling (arrows) associated with the plasma membrane, intracellular vesicles, and a postsynaptic density (curved arrow). (Bar = 0.5 μm.)

GIRK1-labeled dendrites (both spines and shafts) were most commonly observed in str. lacunosum-moleculare (Figs. 2C and 3C), followed by the adjacent superficial portion of str. radiatum (Figs. 2 A and B and 3B). A few dendrites in deep str. radiatum and in str. oriens contained GIRK1 IR, but most profiles in these fields lacked detectable labeling as predicted by the light microscopy.

In all subfields, the vast majority of GIRK1-labeled profiles resembled those of pyramidal cells. For dendrites, in deep str. radiatum (Fig. 3B) the shafts had moderate to large diameters (0.7 to >1.0 μm), received symmetric but not asymmetric synapses, and extended simple spines which received asymmetric synapses. In superficial str. radiatum (Fig. 2 A and B), dendritic shafts were smaller in diameter (0.4–0.9 μm), received symmetric synapses, and possessed somewhat complex spines, which received asymmetric synapses. In str. lacunosum-moleculare (Figs. 2C and 3C), still smaller (usually <0.6 μm) dendritic shafts received symmetric and occasionally asymmetric synapses and extended spines similar to those in deep str. radiatum. In str. pyramidale, GIRK1-IR was present in some pyramidal cell somata (Fig. 3A), which had characteristically smooth nuclei and therefore could be distinguished from nonpyramidal neurons, which have deeply invaginated nuclei. In addition to the numerous pyramidal cell-like profiles, some labeled dendritic profiles were observed that did not posess distinguishing characteristics, and two GIRK1-labeled dendrites with the characteristics of nonpyramidal cells (i.e., nonspiny dendritic shafts with many asymmetric synapses) were observed out of all profiles examined. Interestingly, only a subpopulation of dendritic shafts, spines, or pyramidal cell somata were labeled (Figs. 2B and 3 B and C). In some cases, very similar profiles, such as adjacent pyramidal cell somata, two spines extending from the same labeled dendrite, or two spines contacted by a single axon terminal (Fig. 2B), were differentially labeled.

GIRK1 labeling was commonly located at the plasma membrane and smooth endoplasmic reticulum (Figs. 2 and 3) in dendrites and somata. In pyramidal cell somata, labeling was also found over the Golgi apparatus (Fig. 3A). When immunogold was used as a label, some gold particles were seen in the nucleus (Fig. 3A); however, the nuclear labeling was shown to be nonspecific by its persistence in control sections and absence in tissue processed with the immunoperoxidase technique.

In dendritic spines and thin shafts, GIRK-IR was present with striking frequency near psd (Figs. 2 and 3C). Although some could not be classified due to the presence of immunolabel, many labeled synapses were of the asymmetric type previously associated with excitatory neurotransmission (26). In contrast, GIRK1-IR was not detected at or near the psd of symmetric (inhibitory-type) synapses on large labeled dendritic shafts in str. radiatum (Fig. 3B). To determine whether this apparent targeting of GIRK-IR to the vicinity of asymmetric synapses was statistically significant, GIRK1–immunogold-labeled spines with clear psd and membrane-associated gold particles were analyzed (see Methods). The labeling densities (mean number of gold particles per micrometer of plasmalemma) in each region were: 0.12 ± 0.12 at the psd, 2.13 ± 0.5 at the perisynaptic membrane, and 1.17 ± 0.20 at the extrasynaptic membrane. The densities in each group were significantly different by ANOVA (P < 0.0005). The extrasynaptic membrane contributed 68% of the total plasmalemma measured and contained 71% of the gold particles. In contrast, the perisynaptic membrane was 13% of the total plasmalemma length but contained 26% of the gold particles, and the psd made up 18% of membrane length but contained only 3% of the gold particles. These data suggest that GIRK-IR is enriched in the perisynaptic membrane, is present but not enriched in the extrasynaptic membrane, and is excluded from the psd. However, the exclusion of immunogold labeling from the psd contrasts with the heavy labeling of psd in GIRK1–peroxidase-labeled spines (Fig. 2 B and C). This discrepancy in psd labeling is consistent with reports by others that labeling at the psd varies greatly with different immunocytochemical markers (27, 28) and may be due to diminished access of the particulate gold and/or to diffusion of DAB from antigen-containing sites in the perisynaptic membrane.

DISCUSSION

The subcellular localization of GIRK1-IR in the Golgi apparatus of pyramidal cell somata and in the plasma membrane of pyramidal cell-like dendrites and dendritic spines confirms the hypothesis that GIRK1 is synthesized by pyramidal cells and transported to the more distal dendritic processes. The rare occurrence of GIRK1-IR in axon terminals, particularly in str. lacunosum-moleculare, argues against the suggestion that GIRK1 is present in the terminals of the thalamocortical projection to the hippocampus. Ponce et al. (19) presented this suggestion because kainate lesions of the thalamus decreased GIRK1-IR in str. lacunosum-moleculare. An alternative explanation, which reconciles the present observations with those of Ponce et al., is that the loss of afferent input could alter pyramidal cell morphology. Lesions of perforant path afferent input to dentate granule cells have been shown to cause a decrease in the number of spines in granule cell dendrites (29). Similar postsynaptic remodeling may occur in CA1 pyramidal cells following lesions of afferent thalamocortical input. Since GIRK1-IR was prominently localized to dendritic spines, a partial loss of these spines would account for the decreased GIRK1-IR observed in thalamus-lesioned animals.

The detection of GIRK1-IR in a subset of dendritic shafts and spines suggests that the GIRK1 protein is differentially distributed within dendrites. Such an uneven distribution could result from the different influences of diverse afferent inputs. The innervation of pyramidal cell dendrites includes afferents from the entorhinal cortex, reuniens nucleus of the thalamus, CA3 region of the hippocampus, locus coeruleus, and pyriform cortex (30). Alternatively, an uneven distribution of GIRK1 protein could result from different levels of activity within individual spines. Another possibility is that GIRK1 is present in all dendritic spines of a given neuron, but was not detected with the methods used. Although penetration of reagents could account for nonuniform labeling of structures, this factor was minimized by only analyzing labeling adjacent to the tissue–Epon interface. Ultimately, analysis of serial sections from a higher number of animals will be necessary to confirm that GIRK1 is indeed localized to just a subpopulation of dendritic spines.

GIRK1 forms functional heteromultimeric channels with GIRK2 (7, 15), GIRK3 (7), and GIRK4/CIR (14–16) in Xenopus laevis oocyte expression systems. In situ hybridization has revealed that GIRK1, GIRK2, and GIRK3 are all highly expressed in CA1 pyramidal cells (20), suggesting that GIRK1 forms functional heteromultimeric K⁺ channels in these neurons. The dendritic localization of GIRK1 in CA1 pyramidal cells suggests that a GIRK-mediated K⁺ conductance could reduce the propagation of synaptic excitation from the distal dendrites to the pyramidal cell soma. Dendrites possess active conductances, which can either amplify or attenuate excitatory synaptic input. Amplification is thought to occur through activation of Na⁺, Ca²⁺, or NMDA glutamate receptor conductances. Attenuation occurs through hyperpolarization or shunting of excitatory currents due to decreased membrane resistance (31). Hyperpolarization would increase the relative threshold for voltage-gated Na⁺ and Ca²⁺ channel activation, decrease action potential duration, and help maintain the voltage-sensitive Mg²⁺ block of NMDA receptor channels. A GIRK-mediated K⁺ conductance is likely to attenuate synaptic currents through membrane hyperpolarization or shunting. In support of a GIRK-mediated hyperpolarization, activation of the 5HT_1A serotonin receptor, the GABA_B receptor or the A₁ adenosine receptor elicits a hyperpolarizing response that is mediated by an increased G protein-activated, Ba²⁺- and voltage-sensitive potassium conductance in CA1 pyramidal cells (32–39). At least one of these receptors (5HT_1A) functionally couples to GIRK1 when the two proteins are coexpressed in Xenopus oocytes (13). Therefore, neurotransmitters that activate G protein-coupled receptors may potentially attenuate the propagation of excitatory synaptic currents from the dendrites to the soma in CA1 pyramidal cells through an increased GIRK-mediated K⁺ conductance. The functional coupling of G protein-coupled receptors to GIRK in dendritic spines is supported by the present observation that GIRK1-IR is targeted to the perisynaptic region, which has also been reported to concentrate G protein-linked receptors (27, 28).

In addition to postsynaptic actions of GABA_B receptors described above, activation of presynaptic GABA_B receptors decreases both GABAergic (40, 41) and glutamatergic (42–44) transmission to CA1 pyramidal cells. Given the relative lack of GIRK1-IR in presynaptic terminals in the present study, presynaptic GABA_B receptors may not be coupled to GIRK1-containing K⁺ channels in the CA1 region of the hippocampus. This hypothesis is consistent with a recent report suggesting that the actions of presynaptic GABA_B receptors on GABA release in CA1 are mediated through Ca²⁺ channels (40).

The presence of GIRK1-IR in dendritic spines, and particularly in the vicinity of asymmetric synapses, suggests that a GIRK-mediated K⁺ conductance could postsynaptically modulate excitatory transmission at individual synapses. It has recently been shown that functional voltage-sensitive Ca²⁺ channels are present in the dendritic spines of CA1 pyramidal cells and that action potentials invade the spine head (45). Furthermore, theoretical modeling suggests that activation of voltage-sensitive Na⁺ and Ca²⁺ channels in the spine head may amplify synaptic efficacy (46). Our results showing that another set of channels, the GIRK family, are also localized at or near the spine head suggests that these G protein-activated channels may similarly participate in the fine control of information processing as influenced by local synaptic transmitters.

In summary, we have shown that GIRK1 is localized to dendritic shafts and spines. These findings suggest that activation of a GIRK-mediated K⁺ conductance could alter the propagation of synaptic currents from distal dendrites to somata. Activation of a GIRK-mediated K⁺ conductance in dendritic spines could also modulate postsynaptic responses at individual excitatory synapses. Therefore, neurotransmitters that activate G protein-coupled receptors may potentially modulate synaptic input to CA1 pyramidal cells via postsynaptic activation of a GIRK-mediated K⁺ conductance on either the dendritic shaft or individual dendritic spines.