Distinct axo-somato-dendritic distributions of three potassium channels in CA1 hippocampal pyramidal cells

Potassium channels comprise the most diverse family of ion channels and play critical roles in a large variety of physiological and pathological processes. In addition to their molecular diversity, variations in their distributions and densities on the axo-somato-dendritic surface of neurons are key parameters in determining their functional impact. Despite extensive electrophysiological and anatomical investigations, the exact location and densities of most K+ channels in small subcellular compartments are still unknown. Here we aimed at providing a quantitative surface map of two delayed-rectifier (Kv1.1 and Kv2.1) and one G-protein-gated inwardly rectifying (Kir3.2) K+ channel subunits on hippocampal CA1 pyramidal cells (PCs). Freeze-fracture replica immunogold labelling was employed to determine the relative densities of these K+ channel subunits in 18 axo-somato-dendritic compartments. Significant densities of the Kv1.1 subunit were detected on axon initial segments (AISs) and axon terminals, with an approximately eight-fold lower density in the latter compartment. The Kv2.1 subunit was found in somatic, proximal dendritic and AIS plasma membranes at approximately the same densities. This subunit has a non-uniform plasma membrane distribution; Kv2.1 clusters are frequently adjacent to, but never overlap with, GABAergic synapses. A quasi-linear increase in the Kir3.2 subunit density along the dendrites of PCs was detected, showing no significant difference between apical dendritic shafts, oblique dendrites or dendritic spines at the same distance from the soma. Our results demonstrate that each subunit has a unique cell-surface distribution pattern, and predict their differential involvement in synaptic integration and output generation at distinct subcellular compartments.


Introduction
Potassium channels comprise the most diverse group of ion channels, with more than 100 subunits being identified and classified into different channel families based on their molecular and biophysical properties (Hille, 2001;Gutman et al., 2003). It is well known that different neuron types express distinct sets of K + channel subunits, endowing them with characteristic electrophysiological properties. Recently, it also became clear that differential distributions of distinct K + channel subunits in various subcellular compartments add another layer of complexity to their functional impact. Hippocampal and cortical pyramidal cells (PC) express a wide variety of K + channel subunits (http://www.brain-map.org), which might reside in distinct axo-somato-dendritic compartments. Indeed, electrophysiological experiments have identified K + currents in PC dendrites, where they regulate the integration of synaptic inputs, control action potential (AP) back-propagation and dendritic electrogenesis (Hoffman et al., 1997;Migliore et al., 1999;Cai et al., 2004;Losonczy et al., 2008). They have also been found in axons, where they set the threshold and sculpt the shape of the APs in addition to regulating repetitive firing properties of PCs (Kole et al., 2007;Shu et al., 2007;Rasband, 2010;Bender & Trussell, 2012;Kole & Stuart, 2012). Unfortunately, small subcellular compartments such as oblique dendrites, dendritic tufts, dendritic spines, nodes of Ranvier and axon terminals are inaccessible for patch-pipette recordings, rendering their potassium currents enigmatic. Immunohistochemistry, using subunit specific antibodies, offers an alternative and powerful mean of identifying K + channel subtypes in small subcellular compartments (Rhodes & Trimmer, 2006;Fritschy, 2008;Lorincz & Nusser, 2008b). However, ion channels at low densities could remain undetectable due to the limited sensitivity and resolution of traditional light microscopic (LM) immunolocalisation methods. Electron microscopic (EM) SDS-digested freeze-fracture replica immunolabelling technique (SDS-FRL;Fujimoto, 1995;Rash et al., 1998;Masugi-Tokita & Shigemoto, 2007) allows the detection of ion channels and receptors with a labelling efficiency close to 100% (Tanaka et al., 2005;Lorincz & Nusser, 2010). With such a sensitive and high-resolution method, we have previously determined the relative densities of Kv4.2 and Nav1.6 subunits in a large number of axo-somatodendritic compartments of CA1 PCs (Lorincz & Nusser, 2010;Kerti et al., 2012), creating the first step towards the generation of a quantitative molecular map of the neuronal surface. Here we aim at extending our quantitative surface map of ion channels by investigating the relative densities of two delayed-rectifier (Kv1.1 and Kv2.1) and one G-protein-gated inwardly rectifying (Kir3.2) K + channel subunits.

Materials and methods
All animal procedures were approved by the Institute of Experimental Medicine Protection of Research Subjects Committee as well as the Food-Safety and Animal-Health Office of the Pest District Government Bureau, which is in line with the European Union regulation of animal experimentations.
Testing the specificity of the immunoreactions Specificity of the immunoreactions for Kv1.1 and Kv2.1 subunits was tested by using two antibodies raised against different nonoverlapping epitopes of the respective proteins, which revealed identical labelling patterns in the CA1 region. In addition, the labelling pattern for the Kv1.1 in the CA1 area was identical to that published by Lorincz & Nusser (2008a) with the mouse anti-Kv1.1 antibody; the specificity of that immunoreaction was verified in Kv1.1 À/À mice. The labelling pattern revealed by the Kv2.1 antibodies was also consistent with published data (Trimmer, 1991;Maletic-Savatic et al., 1995;Du et al., 1998;Lim et al., 2000;Misonou et al., 2004;Rhodes & Trimmer, 2006;Sarmiere et al., 2008). In the present study, we used the same anti-Kir3.2 antibody as the one used by Kulik et al. (2006) and Koyrakh et al. (2005) and obtained a very similar labelling pattern; the specificity of the reactions was verified in Kir3.2 À/À mice by Kulik et al. (2006) and Koyrakh et al. (2005). The specificity of immunolabelling for the Kir3.1 and Kir3.3 subunits with these antibodies was proven in the cerebellum by Aguado et al. (2008) using Kir3.1 À/À and Kir3.3 À/À mice, respectively and the specificity of the Kir3.3 labelling in the hippocampus is presented in the web site of the company. Our immunolabelling both in the cerebellum and the hippocampus is identical to the ones shown in the website and in Aguado et al. (2008).
Replicas were then incubated for 2 h in TBS containing 5% BSA and goat anti-rabbit IgGs coupled to 10-or 15-nm gold particles (1 : 100; British Biocell International Ltd., Cardiff, UK), or goat anti-mouse IgGs coupled to 10-or 15-nm gold particles (1 : 50 or 1 : 100; British Biocell). In most double-labelling reactions, a mixture of the two primary then a mixture of the two secondary antibodies was applied. However, we also performed sequential double-labelling reactions (e.g. for Kv1.1 and pan-NF as well as Kv1.1 and SNAP-25; see Fig. 2B and G, respectively) in which the anti-Kv1.1 primary antibody incubation was followed by the anti-rabbit secondary antibody incubation. After this the replicas were incubated with the anti-pan-NF or anti-SNAP-25 primary antibodies and then with the anti-mouse secondary antibodies. Finally, replicas were rinsed in TBS and distilled water before being picked up on copper parallel bar grids. Specimens were analysed with a transmission electron microscope (JEM-1011; JEOL Ltd., Tokyo, Japan).

Quantification of the density of immunogold particles
Quantitative analysis of immunogold labelling for the Kv1.1, Kv2.1 or Kir3.2 subunits was performed on CA1 PC somata, 11 different dendritic compartments, AISs and axon terminals in five CA1 sublayers (n = 3 rats for each subunit; see also Kerti et al., 2012). Briefly, the SR was divided into proximal (0-120 lm), middle (120-240 lm) and distal (240-360 lm) parts based on the distance from stratum pyramidale (SP). The main apical dendrites, oblique dendrites, spines and axon terminals were grouped according to this criterion. Oblique dendrites were identified based on their small diameter and the presence of at least one emerging spine from the dendritic shaft. Structures were only considered to be spines if they emerged from a dendritic shaft. Axon terminals were identified either (i) based on the presence of an active zone (AZ) facing a postsynaptic density (PSD) on the opposing exoplasmic-face (E-face) of a spine or dendrite; or (ii) based on the presence of synaptic vesicles on their cross-fractured portions; or (iii) the presence of a large number of gold particles labelling SNAP-25. Images of AISs of CA1 PCs were taken in SP and SO. To quantify the Kv2.1 and Kir3.2 subunits in the AISs, Kv1.1 subunit (n = 3 rats) and pan-NF (n = 3 rats) were used as molecular markers. In all experiments, the quantified ion channels were visualised with 10-nm gold-conjugated IgGs. All antibodies in this study recognised intracellular epitopes on their target proteins and consequently were visualised by gold particles on the protoplasmic-face (P-face). Nonspecific background labelling was measured on E-face structures surrounding the measured P-faces, as described previously (Lorincz & Nusser, 2010). Images of identified profiles were taken with a Cantega G2 camera (Olympus Soft Imaging Solutions GmbH, M€ unster, Germany) at 10 000-15 0009 magnification. Gold particle counting and area measurements were performed with iTEM software (Olympus Soft Imaging Solutions). Gold particle densities are presented as mean AE SD between animals. Statistical comparisons were performed with STATISTICA software (Scientific Computing, Rockaway, NJ, USA).

Results
Axonal location of the Kv1.1 subunit in hippocampal CA1 PCs First, we investigated the distribution of the Kv1.1 subunit in the CA1 area of the hippocampus using LM immunofluorescent localisations with two antibodies directed against different, nonoverlapping parts of the Kv1.1 protein (see Materials and Methods) and found identical labelling ( Fig. 1A-D). At low magnifications, an intense punctate neuropil labelling was seen in the SO and SR in agreement with published data (Veh et al., 1995;Rhodes et al., 1997;Monaghan et al., 2001;Lorincz & Nusser, 2008a), corresponding to either presynaptic terminals (Monaghan et al., 2001) or dendritic spines. At higher magnifications, AISs ( Fig. 1E-G) and the juxta-paranodal region of myelinated axons were also observed. In order to unequivocally identify the origin of the punctate neuropil labelling of the SO and SR, and to assess the densities of the Kv1.1 subunit in 18 axo-somato-dendritic compartments of CA1 PCs, we turned to SDS-FRL.
We started by investigating whether the AISs, which were the most intensely immunolabelled profiles in our immunofluorescent reactions, contained a high density of gold particles in our replicas. In the SP and SO, several elongated structures contained a high density of immunogold particles labelling the Kv1.1 subunit ( Fig. 2A-D). Only P-face profiles were intensely labelled, consistent with the intracellular location of the epitope recognised by this antibody (aa. 478-492). These structures were then molecularly identified as AISs by the high density of pan-NF labelling ( Fig 2B and D). In AISs, gold particles consistently avoided the PSD of axoaxonic GABAergic synapses identified as dense intramembrane particle (IMP) clusters (Fig. 2D). In the alveus, strongly Kv1.1 subunit-immunoreactive profiles were found surrounded by cross-fractured myelin sheets (Fig. 2E). These structures are likely to correspond to the juxtaparanodal region of myelinated axons that are strongly labelled in the immunofluorescent reactions. Next we assessed the origin of the neuropil labelling of the SO and SR. Small P-face membrane profiles containing an AZ and facing a PSD on the opposing spine or dendritic shaft membrane were consistently labelled (Fig. 2F). Doublelabelling experiments for the Kv1.1 subunit and SNAP-25, a member of the SNARE protein complex restricted to axon terminals (Hagiwara et al., 2005), confirmed that these profiles were axon terminals (Fig. 2G). These axon terminals contained a moderate number of gold particles, without any enrichment in the AZs. In P-face somatic and dendritic membranes (main apical and small oblique dendrites), gold particles were not more numerous than in the surrounding E-face membranes, which we consider as background labelling. After this qualitative assessment of the reactions, we calculated the densities of the Kv1.1 subunit in 18 distinct compartments by counting gold particles on P-face membranes and divided these counts by the measured membrane areas ( Table 2). The background labelling was determined on the surrounding E-face plasma membranes and was found to be 0.3 AE 0.1 gold/lm 2 (mean AE SD; n = 3 rats). The gold particle density values were not significantly higher (ANOVA, P < 0.001; Dunnett's post hoc test, P = 0.999; n = 3 rats) than background in somata, apical dendrites, tuft dendrites in the SLM, oblique dendrites and dendritic spines. In contrast, gold particle densities on axon terminals were significantly above background (ANOVA, P < 0.001; Dunnett's post hoc test, P < 0.05; n = 3 rats; Fig. 2H) in SO, proximal and middle SR. In distal SR and SLM gold particle densities on axon terminals were very similar, but the difference from background did not reach significance (ANOVA, P < 0.001; Dunnett's post hoc test, P = 0.07; n = 3 rats). These densities on axon terminals were seven-to eightfold lower (ratios calculated after background subtraction; ANOVA, P < 0.001; Dunnett's post hoc test, P < 0.001; n = 3 rats) than that found in AISs.
Immunoreactivity for the Kv2.1 subunit is non-uniform in hippocampal CA1 PCs For immunofluorescent localisation of the Kv2.1 subunit in the CA1 area, two antibodies recognising non-overlapping epitopes of the protein were used to ensure that our immunosignal was the consequence of a specific antigen-antibody interaction. When the immunoreactions for the Kv2.1 subunit were analysed at low magnification, strong immunolabelling of the SP and proximal part of the SR was observed (Fig. 3A-D), in line with the results of previous work (Trimmer, 1991;Maletic-Savatic et al., 1995;Du et al., 1998;Lim et al., 2000;Misonou et al., 2004;Rhodes & Trimmer, 2006). Interneurons in all CA1 layers also showed somato-dendritic labelling ( Fig. 3A and D). High-magnification confocal microscopic images revealed plasma membrane-like labelling of somata, proximal apical and basal dendrites (Fig. 3E). Double-labelling experiments for Kv2.1 and Nav1.6 ( Fig. 3E-K) revealed a clustered Kv2.1 labelling of AISs, confirming the results of Sarmiere et al. (2008). Interestingly, high-magnification single confocal images indicated that these Kv2.1-positive clusters did not overlap with the Nav1.6 containing parts of AISs (Fig. 3K).
Using SDS-FRL, we detected immunogold labelling for the Kv2.1 subunit on P-face membranes, as expected from an antibody that recognises an intracellular epitope (aa. 837-853). Strong Kv2.1 immunoreactivity of PC somata and proximal dendrites was observed (Fig. 4A-F). The labelling consisted of scattered and clustered gold particles in the plasma membrane ( Fig. 4B and F). Some of the Kv2.1-positive gold clusters were located over loose patches of IMPs, which might be GABAergic perisomatic synapses. To molecularly identify GABAergic synapses, we performed doublelabelling experiments for Kv2.1 and NL-2, a specific marker of inhibitory synapses (Fig. 4C and D). These reactions revealed that the Kv2.1 subunit was absent from NL-2-containing areas, but occasionally the Kv2.1-and NL-2-positive clusters were close to each other. The non-uniform distribution of gold particles labelling for the Kv2.1 subunit was also characteristic for AISs (molecularly identified with either Nav1.6 or Kv1.1; Fig. 4I-L), where Kv2.1 clusters never overlapped with presumed axoaxonic synapses ( Fig. 4J and L) and also showed a segregation from the Nav1.6 ( Fig. 4I and J) and Kv1.1 (Fig 4K and L) labelling. Distal apical dendrites ( Fig. 4G and H), oblique dendrites, dendritic spines and axon terminals rarely contained any gold particles.
Next we performed a quantitative comparison of the Kv2.1 densities in 18 axo-somato-dendritic compartments of CA1 PCs ( Table 2). The densities of immunogold particles for the Kv2.1 subunit in the somato-dendritic compartments and axon terminals were calculated from single-labelling experiments. Somata and proximal apical dendrites contained high densities of gold particles, which were significantly (ANOVA, P < 0.001; Dunnett's post hoc test, P < 0.001; n = 3 rats; Fig. 5A) higher than background. The densities of the Kv2.1 subunit in apical dendrites in the middle and distal SR, SLM tuft dendrites, oblique dendrites, dendritic spines, and axon terminals were not significantly different from the nonspecific background labelling (ANOVA, P < 0.001; Dunnett's post hoc test, P > 0.26; n = 3 rats). The density of the Kv2.1 subunit in AISs was calculated from double-labelling experiments with the Kv1.1 subunit. The strength of the Kv2.1 labelling of somata [11.4 AE 3.8 gold particles per lm 2 (gold/lm 2 )] in these double-labelling experiments was very similar to that found in single-labelling reactions (P = 0.66, unpaired Student's t-test). AISs contained, on average, 11.5 AE 1.8 gold/lm 2 , which did not differ significantly from the gold particle content of somata (P = 0.97, unpaired Student's t-test), but was significantly above the background (ANOVA, P < 0.01; Dunnett's post hoc test, P < 0.01; n = 3 rats; Fig. 5B).

Distance-dependent increase in Kir3.2 subunits in CA1 PC dendrites
In the CA1 area, the Kir3.1, Kir3.2 and Kir3.3 subunits showed a qualitatively similar pattern of labelling (Fig. 6). The SO, SP and the proximal part of the SR were weakly labelled and the labelling intensity increased gradually towards the distal SR and SLM. Out of the three subunits, the overall intensity was highest for the Kir3.2 and weakest for the Kir3.3 subunit. Our LM observations are in line with those of previous reports (Ponce et al., 1996;Koyrakh et al., 2005;Kulik et al., 2006;Fernandez-Alacid et al., 2011). Because the Kir3.2 subunit forms homomeric channels and seems to be an essential component of heteromeric Kir3 channels (Liao et al., 1996), we performed SDS-FRL of this subunit and addressed whether and how its dendritic density changes as a function of distance from the soma. High-resolution EM images revealed only very few gold particles for the Kir3.2 subunit on the P-face membranes (epitope 374-414 is cytoplasmic) of somata and proximal apical dendrites ( Fig. 7A-C). In comparison, more gold particles were observed on apical dendrites in distal SR and SLM (Fig. 7D-G). In SR, spiny oblique dendrites and dendritic spines were also moderately labelled (Fig. 7H and I).
Following the quantitative analysis of our reactions, we found that the densities of gold particles in proximal, middle and distal apical dendrites and dendritic tufts in SLM showed an apparently linear increase as a function of distance from the soma ( Fig. 7J; Table 2). The densities of gold particles in oblique dendrites and dendritic spines showed a similar distance-dependent increase. Of these, apical dendrites and oblique dendrites in distal SR and den-dritic tufts and spines in SLM showed significantly (ANOVA, P < 0.001 with Dunnett's post hoc test, P < 0.01; n = 3 rats; Fig. 7J) higher densities than background. To determine the density of gold particles in the AISs, we carried out double-labelling experiments with the AIS marker pan-NF and the Kir3.2 subunit. In these reactions, the density of gold particles labelling the Kir3.2 subunit in dendritic tufts in the SLM (14.0 AE 1.0 gold/lm 2 ) was even stronger than in single-labelling reactions with a similarly low background, showing that the quality of double-labelling reactions was even better than that of the single labelling. Despite of this, the density of gold particles in the AISs was not significantly different from background (ANOVA, P < 0.001; Dunnett's post hoc test, P = 0.95; n = 3 rats; Fig. 7J inset). Finally, we investigated the density of the Kir3.2 subunit in presynaptic nerve terminals in all layers of the CA1 area. The axon terminals were identified based on their structural features (see Materials and Methods). The densities of gold particles labelling for the Kir3.2 subunit on the axon terminals were not significantly (ANOVA, P < 0.001; Dunnett's post hoc test, P > 0.05; n = 3 rats) different from the background labelling ( Fig. 7J), demonstrating that with the sensitivity of this method and this antibody, presynaptic Kir3.2 channels remained undetectable.

Discussion
In the present study we extend the current knowledge on the distribution of the Kv1.1, Kv2.1 and Kir3.2 subunits, and advance our quantitative molecular map of the surface of CA1 PCs by determining their densities in 18 well-defined subcellular compartments. The highly sensitive, high-resolution quantitative EM immunogold method SDS-FRL allowed us to study the distribution of these K + channels in small subcellular compartments usually unresolvable and inaccessible with traditional LM immunohistochemical methods. Our results reveal that these three K + channel subunits take up different subcellular locations in CA1 PCs, resulting in unique, subunit-specific labelling patterns. The Kv1.1 subunit is only present in significant amounts in axons, particularly at the AIS, which has an approximately eight-fold higher density than axon terminals. In con-trast, the Kv2.1 subunit is detected at similar densities in AISs, somata and proximal dendrites, but not elsewhere. The density of gold particles labelling the Kir3.2 subunit is different again, showing a quasi-linear increase from the soma towards the distal dendrites, with no significant difference between the main apical dendrites, oblique dendrites and dendritic spines at approximately the same distance from the soma. We also reveal that within the AIS the Kv1.1/Nav1.6 and Kv2.1 subunits are segregated into distinct subdomains, all separate from GABAergic synapses, demonstrating that the surface of the AIS is molecularly more complex than previously anticipated.
The Kv1.1 subunit is restricted to axonal compartments of CA1 PCs Using SDS-FRL, we detected the highest gold particle density for the Kv1.1 subunit in AISs of CA1 PCs, in accordance with our immunofluorescent reactions and previous LM reports. Immunogold particles labelling the Kv2.1 subunit are accumulated over a loose IMP cluster, whereas some nearby dense IMP cluster remains unlabelled. (C and D) Low (C) and high (D) magnification images of the P-face of a CA1 PC soma; 10-nm gold particles labelling the Kv2.1 subunit are associated with an IMP cluster distinct from the adjacent IMP cluster identified as an inhibitory synapse by the enrichment of NL-2 (15-nm gold). (E) Low-magnification image of the P-face of a thick apical dendrite in the proximal SR (prox. SR) labelled for the Kv2.1 subunit. (F) Enlarged view of the boxed region shown in E. Immunogold particles labelling the Kv2.1 subunit are concentrated over two IMP clusters, but scattered labelling is also present. (G and H) A distal apical dendrite contains very few gold particles labelling the Kv2.1 subunit. (I and J) Low-magnification (I) and high-magnification (J) images of the P-face of an AIS co-labelled for the Kv2.1 (10-nm gold) and Nav1.6 (15-nm gold) subunits. The regions immunolabelled for the Kv2.1 and Nav1.6 subunits seem mutually exclusive. (K and L) Low-magnification (K) and high-magnification (L) images of an AIS co-labelled for the Kv2.1 (10-nm gold) and Kv1.1 (15-nm gold) subunits. Gold particles labelling the Kv2.1 subunit are concentrated over an IMP cluster, but excluded from the PSD of a putative axoaxonic synapse. dist., distal; Sp, spine. Scale bars, 500 nm (A, C, E, G, I and K); 100 nm (B, D, F, H, J and L).
Immunofluorescent labelling of Kv1.1 revealed an intense, homogeneous neuropil labelling in SO and SR that could originate from axonal or dendritic profiles. Although Kv1 channels are generally considered 'axonal' channels, associating the neuropil labelling with axons without experimental evidence might be premature because Kv1 subunits could also be present in somato-dendritic compartments as shown in, e.g., the ventral cochlear nucleus (Oertel et al., 2008). Therefore, we used the highly sensitive SDS-FRL method to assess the origin of the neuropil labelling. We could not detect any Kv1.1 subunit in the somato-dendritic compartments of CA1 PCs, but found immunogold particles in axon terminals. It is important to note that glutamatergic axon terminals in SR mainly originate from CA3 PCs and those in the SLM from layer 3 PCs of the entorhinal cortex. In SO, there is a mixture of local collaterals of CA1 PCs and Schaffer collaterals of CA3 PCs. Thus our result regarding the axonal labelling of SR and SLM does not directly prove that this ion channel is present in the same density in CA1 PCs. However, the similar gold densities in axon terminals of SO and SR suggests that the Kv1.1 density of local axon collaterals of CA1 PCs is similar to that of Schaffer collaterals. Nevertheless, within SNAP-25positive axon terminals, the Kv1.1 subunit is apparently homogeneously distributed without any preferential accumulation within or nearby the AZs. The density of Kv1.1 in axon terminals was approximately eight-fold lower than in AISs. This eight-fold difference in the density might not be the only difference in these two axonal compartments, because a potential differential distribution of associated b subunits (McIntosh et al., 1997) that alter the kinetic properties of Kv1 channels could add another layer of complexity. In addition, the subunit composition of heteromeric Kv1 channels might also be different in these two axonal compartments. Homomeric Kv1.1 channels are retained in endoplasmic reticulum , suggesting the presence of heteromeric functional channels in the plasma membranes. In CA1 PCs, Kv1.1, Kv1.2 and Kv1.4 are the most abundantly expressed Kv1 subunits (http://brain-map.org). In AISs (Lorincz & Nusser, 2008a) and juxta-paranodal regions of myelinated axons (Wang et al., 1993), immunolabelling for the Kv1.1 and Kv1.2 subunits showed a very similar distribution pattern, suggesting that they might form heteromeric channels in these compartments. The neuropil of the CA1 area is strongly labelled for the Kv1.1 and Kv1.4, but not the Kv1.2, subunit (Rhodes et al., 1997;Lorincz & Nusser, 2008a), indicating that in presynaptic glutamatergic terminals the Kv1.1 subunit might be co-assembled with the Kv1.4 subunit, similar to other axon terminals containing fast-inactivating, dendrotoxin-sensitive outward currents (Geiger & Jonas, 2000).
Axo-somato-dendritic distribution of the Kv2.1 subunit The Kv2.1 subunit is the most ubiquitously expressed Kv subunit in the brain. It is responsible for mediating the majority of the delayedrectifier K + current in various types of neurons and it is thought to shape AP trains during high frequency firing (Murakoshi & Trimmer, 1999;Du et al., 2000;Malin & Nerbonne, 2002). In agreement with previous LM results (Trimmer, 1991;Maletic-Savatic et al., 1995;Du et al., 1998;Lim et al., 2000;Misonou et al., 2004;Rhodes & Trimmer, 2006;Sarmiere et al., 2008), we detected the highest density of gold particles labelling the Kv2.1 subunit in AISs, somatic and proximal dendritic plasma membranes. These compartments contained the Kv2.1 subunit at similar densities. Kv2.1 subunits have been shown to form clusters in neuronal plasma membranes (Misonou et al., 2004(Misonou et al., , 2005. Altered neuronal activity induces calcineurin-mediated dephosphorylation and dispersion of the clustered channels, which is paralleled by changes in their biophysical properties (Misonou et al., 2004(Misonou et al., , 2005Mulholland et al., 2008). In our immunogold reactions, both clustered and scattered Kv2.1 subunits were present in the axo-somato-dendritic plasma membranes, suggesting the presence of Kv2.1 channels in differential phosphorylated states. Due to its high sensitivity, SDS-FRL allows the detection of proteins in compartments where they are expressed at low densities. We were able to detect the scattered Kv2.1 subunits in AISs, somata and proximal dendrites, where their average density is moderate (~10 gold/lm 2 ) but their out-of-cluster density is very low. However, we were still unable to detect this subunit in other dendritic compartments (e.g. oblique dendrites and spines). Pre-embedding immunogold experiments localised the Kv2.1 subunit adjacent to inhibitory PSDs (Du et al., 1998). We were not able to detect gold particles inside the NL-2-containing GABAergic PSDs, but Kv2.1 clusters were often formed within less than a micrometre distance from perisomatic GABAergic synapses. Somato-dendritic Kv2.1 clusters have been found over subsurface cysternal organelles and their role in Ca 2+ signalling has been proposed (Du et al., 1998), but whether and how they affect neighbouring Kv2.1 channel function remain to be determined.
Kv2.1 and Kv1.1 subunits are segregated within the AIS of CA1 PCs An intriguing finding of our present work is that the Kv1.1 subunits and the clustered Kv2.1 subunits occupy discrete, non-overlapping subdomains within the AISs of CA1 PCs. The AIS is a highly specialised subcellular compartment, accumulating voltage-gated ion channels at high densities, especially Nav channels, thereby ensuring the lowest threshold for AP generation (Rasband, 2010;Kole & Stuart, 2012). Various mechanisms control the expression of ion channels in the AIS. Ankyrin-G clusters Nav as well as Kv7 channels in the AIS via direct binding (Kordeli et al., 1995;Garrido et al., 2003;Pan et al., 2006). Expression of the Kv1 channels in the AIS depends on different mechanisms involving PSD-93 (Ogawa et al., 2008), Caspr2 (Inda et al., 2006) and recently the contribution of the cytoskeletal linker protein 4.1B has also been proposed (Duflocq et al., 2011). Interestingly, although anchored by different proteins in the AISs, the Kv1.1 and Kv1.2 subunits occupy the same plasma membrane areas as the Nav1.6 subunit (Figs 1 and 2; see also Lorincz & Nusser, 2008a). We previously found nonhomogeneous labelling for Ankyrin-G, NF and the Nav1.6 subunit within the AIS of CA1 PCs, where discrete patches remained unlabelled, and subsequently we identified that these 'empty' patches contain the PSDs of axoaxonic synapses (Lorincz & Nusser, 2010). Similarly, here we found that small clusters of the Kv2.1 subunits avoided plasma membrane areas rich in the Nav1.6 ( Fig. 4I and J) and Kv1.1 (Fig. 4K and L) subunits. However, our reactions also demonstrated that these Kv2.1 clusters are also distinct from GABAergic axoaxonic synapses. They were not only morphologically different, but often separated by Nav1.6-and Kv1.1-rich membrane areas. AISs are also polarised along the proximodistal axis, having high density of Nav1.1 and Nav1.2 subunits in the proximal regions and Nav1.6, Kv1.1 and Kv1.2 subunits in the more distal part (Lorincz & Nusser, 2008a, 2010. The distribution of the Kv2.1 subunit does not seem to follow such a tendency, as clusters are uniformly distributed along the entire length of the AIS of CA1 PCs. We hypothesise that these compartments are not necessarily electrically distinct, but these channels might require differential regulation and association with intracellular signalling cascades, which could underlie their sub-micrometre segregation.
The density of Kir3.2 increases as a function of distance from soma in dendritic shafts and spines of CA1 PCs An enhanced spontaneous activity of Kir3 channels has been reported (Chen & Johnston, 2005) at resting potential in the apical dendrites of CA1 PCs as compared to the somata, which might be the result of increased channel numbers or increased open probability. Our LM reactions revealed similar labelling pattern for the Kir3.1, Kir3.2 and Kir3.3 subunits in the CA1 area in agreement with previous studies (Ponce et al., 1996;Koyrakh et al., 2005;Kulik et al., 2006;Fernandez-Alacid et al., 2011). The SO, SP and the proximal part of the SR was weakly labelled and the labelling intensity gradually increased towards the distal SR and SLM. EM immunogold studies in CA1 PCs revealed the presence of Kir3.1 (Drake et al., 1997;Fernandez-Alacid et al., 2011), Kir3.2 (Koyrakh et al., 2005;Kulik et al., 2006;Fernandez-Alacid et al., 2011) andKir3.3 (Fernandez-Alacid et al., 2011) in dendritic shafts and spines, but their relative densities in different axo-somato-dendritic compartmens are still unknown. Because the Kir3.2 subunit is an essential component of heteromeric Kir channels (Liao et al., 1996), we performed the high-resolution EM localisation of this subunit. The density of gold particles labelling the Kir3.2 subunit showed a quasi-linear increase from the soma towards the distal dendrites, with no significant difference between the main apical dendrites, oblique dendrites or dendritic spines at approximately the same distance from the soma. It is known that, in dendrites, constitutively active Kir3 channels contribute to the resting membrane potential and decrease excitability (Luscher et al., 1997), but understanding how integration of synaptic inputs in dendrites is affected by this non-uniform 'leak' conductance will require detailed computational and experimental investigations. Gold particles labelling the Kir3.1 and Kir3.2 subunits have been occasionally found in presynaptic axon terminals at low numbers (Drake et al., 1997;Koyrakh et al., 2005;Kulik et al., 2006); we also observed a few gold particles labelling the Kir3.2 subunit in excitatory axon terminals, but the density was not significantly different from the background labelling obtained on surrounding E-face membranes.

Unique distribution patterns of ion channels in CA1 PCs
Hippocampal CA1 PCs have been the subject of many functional, structural and molecular investigations. There is a large body of experimental data revealing the presence of voltage-and ligandgated ionic currents in different subcellular compartments, but less is known about the density and distribution of the underlying channel subunits. To reveal the distribution and densities of molecularly and often functionally distinct subunits, probably the best method is high resolution, quantitative immunolocalisation using subunitspecific antibodies. Electrophysiological recordings detected a similar increase in the density of the hyperpolarisation-activated mixed cation current (I h ) and A-type K + current (I A ) in the main apical dendrites of CA1 PCs (Hoffman et al., 1997;Magee, 1998), suggesting a very similar distribution of the underlying HCN1/2 and Kv4.2 subunits. However, EM immunogold studies revealed that the subcellular distribution of the HCN1 and Kv4.2 subunits are A B Fig. 5. Densities of gold particles labelling the Kv2.1 subunit in different subcellular compartments of CA1 PCs. (A) Bar graphs show the Kv2.1 subunit densities (mean AE SD) in different axo-somato-dendritic compartments. Significant densities of gold particles labelling the Kv2.1 subunit (*) are found on the somata and proximal apical dendrites (ApDendr) of CA1 PCs (ANOVA, P < 0.001; Dunnett's post hoc test, P < 0.001; n = 3 rats). (B) Gold particle density for the Kv2.1 subunit on the AISs of CA1 PCs was significantly different from the background (BG; ANOVA, P < 0.01; Dunnett's post hoc test, P < 0.01; n = 3 rats). These data are obtained from double-labelling experiments for the Kv2.1 and Kv1.1 subunits. Gold particle densities for the Kv2.1 subunit on PC somata was similar in single-and double-labelling experiments (cf. A and B). AxTerm, axon terminal; OblDendr, oblique dendrite; Tuft Dendr, tuft dendrite; prox., proximal; mid., middle; dist., distal. Bar graphs demonstrate the densities (mean AE SD) of gold particles in different subcellular compartments of CA1 PCs. *Values that are significantly different from the background labelling (ANOVA, P < 0.001; Dunnett's post hoc test, P < 0.01; n = 3 rats). The inset shows that gold particle density in the AIS was not significantly different from the background labelling (ANOVA, P < 0.001; Dunnett's post hoc test, P = 0.95; n = 3 rats). The AISs were identified based on their immunopositivity for pan-NF. BG, background; mid., middle; Sp, spine; AxTerm, axon terminal. Scale bars, 1 lm (A), 500 nm (D and F), 250 nm (H and I), 100 nm (B, C, E and G). markedly different. Consistent with the electrophysiological findings, a distance-dependent increase in the density of the HCN1 subunit was found in distal dendritic shafts (Lorincz et al., 2002). In contrast, the Kv4.2 subunit showed only a moderate (70%) increase in the density in the SR (Kerti et al., 2012). Neither the HCN1 nor the Kv4.2 subunit was detected in the axonal compartments of CA1 PCs. SDS-FRL localisation of the Nav1.6 subunit revealed a completely different distribution pattern. The Nav1.6 subunit was identified in both somato-dendritic and axonal compartments. The highest density was found in nodes of Ranvier and in AISs, and only a low density was detected in somata and proximal apical dendrites. In the main apical dendrites, the density of Nav1.6 subunit decreased as a function of distance from the soma and dendritic spines were always immunonegative. Among the ion channels studied so far in CA1 PCs, including the Kv1.1, Kv2.1 and Kir3.2 subunits presented here and the Nav1.6 and Kv4.2 published previously, we found that each ion channel subunit has its own unique distribution pattern on the surface of a given neuron type. It remains to be seen whether this is a general rule or eventually we will find two distinct subunits with identical distribution pattern.