Coronal slices of the brainstem of mice (age: 3- to 6-weeks) were cut with a vibroslicer (Leica VT1000) and were transferred to a storage chamber to be incubated in ACSF containing (in mM): 124 NaCl, 5 KCl, 1.25 Na₂HPO₄, 2.5 MgSO₄, 2 CaCl₂, 26 NaHCO₃ and 20 D-glucose, bubbled with 95% O₂ and 5% CO₂ (all chemicals were purchased from Sigma-Aldrich). After an incubation period of at least 1 h a slice was placed in the recording chamber through which ACSF (34–35°C) flowed at 6–7 ml/min. Whole-cell patch-clamp recordings were performed using an EPC-10 amplifier (HEKA Electronics, Germany). The patch pipettes were filled with intracellular solution containing (in mM): 120 K-gluconate, 9 KCl, 10 KOH, 3.48 MgCl₂, 4 NaCl, 10 HEPES, 4 Na₂ATP, 0.4 Na₃GTP, and 17.5 sucrose. Neurobiotin tracer (final concentration: 1%, Vector Labs) was added to the intracellular solution. The pH of the intracellular solution was adjusted to 7.25. Input resistance (Ri) was measured by injection of hyperpolarizing test currents (200 pA; 100 ms) and was calculated from the voltage transient toward the end of current injection. Neurons were excluded if the input resistance varied by >15%. Data was analyzed in Clampfit 9.2 (Axon Instruments, Foster City, CA, USA). The neurons were held in whole-cell configuration for 10 min and then the electrode was gently retracted allowing the cell membrane to form a gigaseal again. The electrode was removed leaving the neuron in place and the slice was then moved to the recovery chamber in order to allow the tracer to diffuse for 260 min. At the end of the diffusion period the slice was immersed in a paraformaldehyde (PFA; 4%) solution for 1 h and then transferred to a phosphate buffer (PB 0.1 M) before being processed for immuno-histochemistry.

The tracer transfer method was executed under four different conditions. In control condition, brain slices were maintained before, during and after the tracer injection exclusively in normal ACSF. In all the other conditions, brain slices were pre-incubated (at least 1 h before the tracer injection), injected and recovered in ACSF containing either Forskolin (10 μM), or KN93 (10 μM), or H-89 (1 μM). A total of 47 IO neurons were successfully held in whole cell configuration for 10 min and filled with Neurobiotin. Resting mV, input resistance and membrane capacitance were measured as described in our previous works (Khosrovani et al., 2007). AP were evoked in current clamp with five 300 ms steps of increasing amplitude (100 pA) every 5 s. Maximal firing frequency was calculated as the inverse of the inter-spike interval between the first and the second AP evoked with the injection of 500 pA. Current step protocol entails the injection of depolarizing currents, which are known to down-modulate the GJs (Haas et al., 2011). To limit the influence of current injections, we applied the above mentioned protocol only four times to all the recorded neurons in all conditions.

Tracer transfer of Neurobiotin was visualized with immunohistochemistry (IHC) on brainstem slices in which electrophysiological experiments were successfully performed. Additionally, Cx36 GJs were visualized in these slices using IHC. The procedure was performed at room temperature and pH of all solutions was 7.4. Slices were washed twice with PB (0.1 M) and the tissue was incubated in normal horse serum (10%) and Triton-X100 (0.5%) for 2 h. In order to identify Cx36 GJs, free floating slices were labeled with goat polyclonal anti-Cx36 antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA; 1 ug/ml) for 96 h at 4°C. After incubation with primary antibody, IO slices were first rinsed four times for 10 min in PB (0.1 M) then incubated for 4 h with donkey anti-goat Cy3 conjugated antibody (Jackson Immunoresearch, West Grove, PA, USA; 1:200), and Alexa Fluor 488 conjugated Streptavidin (Invitrogen, 1:200); the latter incubation procedure is to attach a fluorescent signal to the Neurobiotin tracer. Incubation with secondary antibodies was followed by four wash steps of 10 min with PB (0.1 M). Finally, slices were mounted on coverslips, embedded in Vectashield mounting medium and stored in dark at 4°C until further use.

To determine the size of the tracer-coupled neuronal network and the distribution of Cx36 GJs, confocal images were obtained using a Zeiss LSM700 confocal laser scanning microscope operated by standard Zen 2009 software. First, a z-stack that included all visible tracer-coupled IO neurons surrounding the primary filled neuron was obtained with a 20× objective (see Figure 1D); planes were separated by optimal z-scaling automatically determined by the operation software. Confocal images were opened in Zen 2009 Light edition and tracer-coupled neurons were manually quantified by two researchers, who had no knowledge of the pharmacological treatment of the slices.

To study the localization of Cx36 puncta on dendrites of the primary Neurobiotin-injected IO neurons, two channel images were made using a 63× objective and 2× digital zoom; one image in which the Neurobiotin-filled dendrites were excited at 488 nm and one image in which the Cx36 GJs were excited at 555 nm. For each primary cell, three z-stack scans of proximal and three of distal dendritic trees were made. Dendritic branches up to 50 μm from the soma were considered proximal, while further than 70 μm were considered distal. Three proximal and three distal segments lying in the same focal plane were analyzed in every neuron. Cx36 puncta were quantified using Fiji software (ImageJ: image processing software), the two channel image was first separated in two images of one channel (an image of the dendritic tree and an image of the Cx36 GJ puncta). The length of dendritic tree was measured in three dimensions with the simple neurite tracer function preprogrammed in Fiji software. Cx36 puncta were detected using the 3D object counter function. Center of mass of the puncta was determined using threshold setting of 50 and size between 10 and 500 voxels (corresponding with a puncta size range of ~0.3–1.2 μm). Puncta corresponding to these criteria were selected and considered part of Cx36 GJs. Subsequently, images with the dendritic trees and the selected Cx36 puncta represented by center of masses objects were merged into a one channel image (Figures 2B–D). Center of masses that co-localized with measured dendritic tree were then manually counted with the use of the count cells function of Fiji software. Two researchers, who had no knowledge of the pharmacological treatment of the slice, counted the number Cx36 puncta for each dendritic tree. An average of the amount of Cx36 puncta per 100 μm dendrite was calculated for both proximal and distal dendritic trees of each primary tracer-filled IO neuron.

To determine the density of Cx36 puncta in slices of αCaMKII−/−, βCaMKII−/− mutants and their wild-type (WT) littermates, Cx36 protein were labeled with goat polyclonal anti-Cx36 antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA) as describe above. Volumetric samples (10.6 10³ μm³) were taken from these slices by making 16 z-stack scans (excited at 555 nm; 63× objective; 2× digital zoom) of 51.4 by 51.4 μm that were separated by 0.25 μm. In these samples, all Cx36 puncta were detected and quantified by using Fiji software and the procedure described above. From each mouse three samples were taken from the IO and one background sample was taken just outside the IO. The density of puncta from the background sample was subtracted from the IO samples.

Adult αCamkII−/−, βCamKII−/− mutant mice and WT littermates were anesthetized with an overdose of Nembutal and transcardially perfused with 4% paraformaldehyde and 1% glutaraldehyde in 0.12 M cacodylate buffer; the brainstem containing the IO was processed for electron microscopy. The sections of 100 μm were cut on a vibratome, osmicated in OsO₄, stained en bloc in tannic acid and uranyl acetate, dehydrated in dimethoxypropane, and embedded in Araldite. Subsequently, various olivary subnuclei were identified in semi-thin sections, pyramids were made, and ultrathin sections were cut accordingly on a Reichert ultratome, counterstained with uranyl acetate and lead citrate, and examined with the use of a Philips CM-100 electron microscope. The sections of the various tissue blocks were screened per surface area, and GJs were identified and quantified by a researcher who had no knowledge about the origin of the tissue. GJs were identified by their morphology; a membrane plaque with a small interneuronal pace (Figures 3A,B, black arrowheads) with electron-dense deposits on both sides (Figure 3B, white arrowheads).

Data from the αCamkII WT littermates and βCamKII WT littermates were merged and is indicated as the WT littermate group. The results are expressed as average ± SEM (unless noted otherwise). Statistical significance has been assessed by using one-way ANOVA, post hoc Tukey’s HSD and Bonferroni tests and Dunnett’s test (SPSS software). Given the consistency between the two approaches, only Tukey’s HSD test’s results have been reported.

Generally, the tracer-coupled network appears as a group of neuronal somata stained with different intensity (Figure 1A). To explore the modulation of Cx36 GJ in neuronal network formation in the IO, the size of the tracer-coupled network was studied under the above mentioned four different conditions. Our data revealed a significant interaction between the treatment and the size of the tracer-coupled network (one-way ANOVA, F_(3,43) = 4.2, p < 0.05). In untreated slices, Neurobiotin distribution showed that the primary stained neuron is on average tracer-coupled to 15.3 ± 3.2 neurons (n = 12; Figures 1A,E). In the group of slices that were treated with Forskolin, a significant reduction in the number of coupled neurons was shown with an average of 4.9 ± 1.5 tracer-coupled neurons (n = 12; Figures 1B,E; Tukey’s HSD post hoc test, p < 0.05). This result indicates that an increased PKA activity down-regulates the coupling between IO neurons. Alternatively, the possibility that the inhibition of PKA activity increases the number of coupled neurons was explored by incubating slices with H-89. In those slices, the number of tracer-coupled neurons was not significantly different from the number of tracer-coupled neurons observed in the untreated slices (n = 12; Figures 1D,E; Tukey’s HSD post hoc test, p = 0.15). On average the primary cell was tracer-coupled to 7.9 ± 2.9 neurons in H-89 treated slices. Finally, CaMKII as Cx36 GJ modulator was investigated by blocking its activity with KN93 (Vest et al., 2007). This block of CaMKII activity significantly reduced the number of tracer-coupled neurons when compared to the number of tracer-coupled neurons that were measured in untreated slices (n = 11; Figures 1C,E; Tukey’s HSD post hoc test, p < 0.05). On average the primary cell was tracer-coupled to 4.9 ± 1.7 neurons in KN93 treated slices. Overall, our results show that both the PKA and CaMKII pathway can modulate the size of the coupled network.

The activation of PKA and the inhibition of CaMKII both caused a reduction in number of tracer-coupled neurons. This reduction can either be due to a reduced opening probability of GJ channels or by a reduced availability/presence of GJ channels at electrical synaptic sites. In order to elucidate the mechanism that is responsible for this reduction of tracer-coupling, Cx36 puncta on the proximal and distal dendrites of primary stained IO neurons were visualized and quantified. The primary neuron of coupled network is the whole-cell recorded neuron that is directly loaded with Neurobiotin tracer, and for this reason it always presents clear visible dendritic arborizations (Figure 2A). The length of proximal and distal dendritic branches of each primary neuron were measured (Figure 2A white boxes, Figure 2B) and the number of Cx36 puncta expressed on those dendritic fragments were counted (Figures 2C,D, see “Materials and Methods” section). For both the proximal and distal dendrites, a significant interaction was found between the treatment and the number of Cx36 puncta (proximal: one-way ANOVA, F_(3,35) = 3.37, p < 0.05; distal: one-way ANOVA, F_(3,35) = 3.48, p < 0.05). On proximal dendrites, the number of Cx36 puncta is 14.9 ± 2.9/100 μm (n = 15) in untreated slices and it is slightly but not significantly reduced in the slices treated either with Forskolin or H-89 (11.7 ± 3.1/100 μm, n = 7, p = 0.83 and 11.2 ± 1.7/100 μm, n = 10, p = 0.71 respectively, Tukey HSD post hoc test). However, when slices were incubated with KN93, the number of Cx36 GJ puncta was significantly reduced compared to the untreated condition (Tukey HSD post hoc test, p < 0.01). On average the proximal dendrites expressed 2.8 ± 0.9 Cx36 puncta/100 μm in KN93 treated slices (n = 7, Figure 2E). Comparable results were obtained when counting Cx36 puncta on distal dendrites, although absolute values are slightly lower for the counting of distal dendrites; the number of Cx36 puncta is 10.5 ± 2.2/100 μm dendrite in untreated slices (n = 15), 8.8 ± 1.9/100 μm dendrite when treated with Forskolin (n = 7) and 9.1 ± 1.7/100 μm dendrite when treated with H-89 (n = 10). The number of Cx36 GJs puncta/100 μm dendrite is slightly but not significantly different between the untreated condition and Forskolin condition (Tukey HSD post hoc test, p = 0.9) and between the untreated condition and H-89 condition (Tukey HSD post hoc test, p = 0.9). Also on the distal dendrites treatment with KN93 significantly reduces the expression of Cx36 compared to control conditions (Tukey HSD post hoc test, p < 0.01). On average the distal dendrites expressed 1.3 ± 0.5 Cx36 puncta/100 μm in KN93 treated slices (n = 7, Figure 2F). These results show that activation of PKA down-regulates tracer coupling not by removing the Cx36 GJs from the membrane and therefore it is most likely that the activation PKA reduces the opening probability of GJs. The inhibition of CaMKII activity does reduces the number of Cx36 puncta on the proximal and distal dendrites of olivary neurons and therefore activation of CaMKII is involved in the incorporation of Cx36 GJs at the site of electrical synapse. Despite the fact that activation of PKA and inhibition of CaMKII both reduce the number of tracer-coupled olivary neurons, the current results indicate that they use a different mechanism.

If CaMKII is required for the incorporation of Cx36 GJs at the site of the electrical synapse, as indicated by our pharmacological study, then genetic removal of the CaMKII protein should have a detrimental effect on the formation of the electrical synapse. Therefore, we studied the amount of GJs in αCaMKII and βCaMKII knock-out mice at an immunohistochemical as well as at the ultrastructural level. For the immunohistochemical analysis, volumetric samples (10.6 10³ μm³) were taken from IOs of αCaMKII−/−, βCaMKII−/− mutants and wild type littermates (control). Using the immunohistochemical approach a density of 28.8 ± 5.4 Cx36 puncta/1000 μm³ was found in the WT mice (n = 6), whereas lower densities were found in the αCaMKII−/− mutant mice (23.8 ± 3.8 Cx36 puncta/1000 μm³, n = 6) and in the βCaMKII−/− mutant mice (5.7 ± 2.0 Cx36 puncta/1000 μm³, n = 9; one-way ANOVA, F_(2,18) = 13.7, p < 0.01; Figure 3E). The density of Cx36 puncta in the IOs of βCaMKII−/− mutant mice were significantly smaller than in WT mice (Tukey HSD post hoc test, p < 0.01) and in αCaMKII−/− mutant mice (Tukey HSD post hoc test, p < 0.01; Figure 3E).

Ultrastructural analyses of GJs in WT mice show characteristics consistent with earlier studies (De Zeeuw et al., 2003; Van Der Giessen et al., 2006). WT GJs are present in the IO with a density of 87 ± 4 GJs/mm² (n = 5, Figure 3F1), consist of a plaque with an average diameter of 312 ± 24 nm (n = 18, Figure 3F2) and an interneuronal space of 3.8 ± 0.2 nm thickness (n = 18, Figure 3F3); they show electron-dense deposits at both sides of the membrane and have attachment plaques surrounding the plaque with gap-junction channels (Figures 3A,B). For both the GJ density and interneuronal space, a significant interaction was found between the three CaMKII groups (GJ density: one-way ANOVA, F_(2,12) = 282, p < 0.01; GJ interneuronal space: one-way ANOVA, F_(2,26) = 8.1, p < 0.05). No significant interaction was found between the three groups and the diameter of GJs (one-way ANOVA, F_(2,26) = 2.2, p = 0.13). In αCaMKII−/− mutant mice, we observed GJs with normal characteristics as mentioned previously. The density of GJs in αCaMKII−/− mutant mice, 96 ± 9 GJs/mm² (n = 4), was not significantly different from WT mice (Tukey HSD post hoc test, p = 0.25). Whereas in βCaMKII−/− mutant mice, the density of GJs in the olivary subnuclei was significantly smaller than in WT and αCaMKII−/− mice (1 ± 1 GJs/mm², n = 6; Tukey HSD post hoc test, p < 0.01). Despite the lack of GJs in βCaMKII−/− mutant mice, the morphological characteristics of all glomeruli were comparable to WT (Figure 3C1). The vast majority of glomeruli exhibited two dendritic spines in the core of glomeruli surrounded by several excitatory and inhibitory terminals nonetheless centrally located GJs were missing. Moreover, the extraglomerular neuropil including dendritic lamellar bodies and glial GJs appeared normal. In the entire area of βCaMKII−/− IO that we scanned, we found only one GJ which had a normal interneuronal space (3.9 nm) compared to WT, however with a smaller plaque diameter (135 nm; Figure 3F2). We also observed one adhesion-plaque structure, which was not observed in previous studies (Figure 3C2). The GJ interneuronal space was significantly wider in αCaMKII−/− mutant mice (5.9 ± 0.6 nm, n = 10; Tukey HSD post hoc test, p < 0.01; Figure 3D) than in WT (3.8 ± 0.2 nm, n = 18) and in βCaMKII−/− mutant mice (3.9 nm, n = 1).

We conclude from these data that absence of βCaMKII leads to a strong reduction of GJs in all subnuclei of the IO. Despite the lack of βCaMKII, general ultrastructural morphology of olivary glomeruli is unchanged.