|Home | About | Journals | Submit | Contact Us | Français|
In contrast to chemical transmission, few proteins have been shown associated with gap junction-mediated electrical synapses. Mixed (electrical and glutamatergic) synaptic terminals on the teleost Mauthner cell known as “Club endings” constitute because of their unusual large size and presence of connexin 35 (Cx35), ortholog of the widespread mammalian Cx36, a valuable model for the study of electrical transmission. Remarkably, both components of their mixed synaptic response undergo activity-dependent potentiation. Changes in electrical transmission result from interactions with co-localized glutamatergic synapses, the activity of which leads to the activation of Ca++/calmodulin-dependent kinase II (CaM-KII), required for the induction of changes in both forms of transmission. However, the distribution of this kinase and potential localization to electrical synapses remains undetermined. Taking advantage of the unparalleled experimental accessibility of Club endings, we explored the presence and intraterminal distribution of CaM-KII within these terminals. Here we show: 1) unlike other proteins, both CaM-KII labeling and distribution were highly variable between contiguous contacts, and 2) CaM-KII was not restricted to the periphery of the terminals, where glutamatergic synapses are located, but also was present at the center where gap junctions predominate. Accordingly, double-immunolabeling indicated that Cx35 and CaM-KII were co-localized and biochemical analysis showed that these proteins associate. Because CaM-KII characteristically undergoes activity-dependent translocation, the observed variability of labeling likely reflects physiological differences between electrical synapses of contiguous Club endings, which remarkably co-exist with differing degrees of conductance. Taken together, our results indicate that CaM-KII should be considered a component of electrical synapses although its association is non-obligatory and likely driven by activity.
As a result of their unusually large size (~10 μm), a group of contacts on the lateral dendrite of the teleost Mauthner (M−) cells has provided since their discovery (Bartelmez, 1915) the opportunity of examining diverse structural features of vertebrate synapses. These “striking synapses” are terminations of auditory afferents originating in the anterior part of the sacculus, and are known as “Large Myelinated Club endings” (Club endings; CEs) (Bartelmez, 1915). Studies on CEs provided early evidence for the lack of protoplasmatic continuity between pre- and postsynaptic elements (Bartelmez and Hoerr, 1933) and, with the advent of electron microscopy, with early evidence for the structural basis of electrical transmission: the gap junction (Robertson et al., 1963). Ultrastructural (Tuttle, et al., 1987) and confocal microscopy (Pereda et al., 2003) studies show that while gap junctions (up to ~200) are distributed throughout the surface of the contact, chemical synapses are predominantly restricted to the periphery (Tuttle, et al., 1987). Thus, the large size of these terminals and differential segregation of the structural components for both forms of transmission makes them amenable for exploring the presence and detailed sub-cellular distribution of various proteins associated to these modalities of synaptic transmission.
The physiological features of CEs can also be easily explored, allowing the uncommon correlation between synaptic structure and function (for review see Pereda et al., 2004). Consistent with the co-existence of gap junctions and chemical synapses at these terminals (“mixed synapses”), electrical stimulation of saccular afferents evokes a mixed, electrical and chemical, synaptic potential in the lateral dendrite of the M-cell (Furshpan, 1964; Lin and Faber, 1988a). While chemical transmission is mediated by glutamate (Wolszon et al., 1997), gap junctions contain connexin 35 (Cx35) (Pereda et al., 2003) the fish ortholog of the neuronal connexin 36 (Cx36), which is widely expressed throughout mammalian brain (Condorelli et al., 1998). Strikingly, both components of the mixed synaptic response undergo activity-dependent potentiation of their respective strengths (Yang et al., 1991; Pereda and Faber, 1996). Thus, changes are not restricted to chemical synapses but also involve the regulation of gap junction-mediated electrical synapses. The mechanism underlying modifications in electrical transmission reflects a functional interaction with neighboring co-localized glutamatergic synapses, where activity leads to the activation of the multifunctional Ca++/calmodulin-dependent kinase II (CaM-KII), necessary for the induction of changes in both forms of transmission (Pereda et al., 1998). Yet, the distribution and relationship of CaM-KII to chemical and electrical synapses within CEs remains undetermined.
CaM-KII has been implicated in mechanisms of activity-dependent plasticity in chemical synapses (Fink and Meyer, 2002; Schulman 2004; Merill et al., 2005; Wayman et al., 2008). This kinase is an essential and abundant component of glutamatergic postsynaptic densities (PSDs) (Kennedy, 2000) where it associates with other proteins, such as NMDA receptors (Yang and Schulman, 1999; Strack et al., 2000). Recent data indicate that this kinase can also molecularly interact with Cx36 (Alev et al., 2008). Together with the well-established functional role of CaM-KII in CEs (Pereda et al., 1998), the high homology of Cx36 with its fish ortholog suggests a potential association of this kinase with Cx35 at these terminals. Moreover, the dynamic spatial properties of CaM-KII which are based on its ability to translocate to active synapses (Shen and Meyer, 1999; Merill et al., 2005; Rose et al., 2009; Lee et al., 2009), including in fish (Gleason in al., 2003), suggest that its intraterminal distribution could reflect its functional role in these terminals.
To investigate the distribution of CaM-KII in CEs, we took advantage of the unique experimental accessibility of the M-cell system for imaging single synaptic terminals using confocal microscopy. Here we show that, the distribution of CaM-KII was highly variable between contiguous CEs. Furthermore, CaM-KII was found not only in the periphery of the terminals, where glutamatergic PSDs are located, but also at the center where gap junctions predominate. Double-immunolabeling and biochemical approaches showed that Cx35 and CaM-KII associate, indicating that these proteins functionally interact at electrical synapses in CEs. We speculate the observed heterogeneinity of labeling may reflect differences in functional states between adjacent CEs.
Goldfish (Carassius auratus) 2–5” long were perfused with phosphate buffered saline (1X PBS) at pH 7.4 for 10–15 min followed by cold 4% paraformaldehyde in 0.1 M phosphate buffer for 10–15 min. Brains were dissected out and kept overnight at 4°C in 4% paraformaldehyde in 0.1 M phosphate buffer. Brains were sectioned (40–50 μm) with a TPI microtome (Vibratome, Technical Products International). The sections were rinsed at room temperature three times for 10 min each with 1X PBS, blocked and permeabilized for 45 min to 1 h at room temperature in PBStr (1X PBS, 0.3–0.4% Triton X-100, pH 7.4) plus 10% normal goat serum (NGS). Sections were incubated overnight at 4°C using a moving platform with rabbit polyclonal G301 anti-αCaM-KII (1:1000) and/or mouse monoclonal anti-Cx35/36 (Chemicon; 1:250/500). A previously characterized teleost anti-NR1 antibody directed against the NR1 subunit of the electric fish Apteronotus (Berman et al., 2001; gift of Drs. L. Maler and R. Dunn) was used as a marker for glutamatergic synapses. Then, sections were rinsed in PBStr four times for 10 min each wash and incubated for 1–2 h at room temperature with Alexa Fluor 594-conjugated goat anti-rabbit and Alexa Fluor 488-conjugated goat anti-mouse (Molecular Probes; 1/500) or Texas Red goat anti-rabbit (Jackson ImmunoResearch; 1/500) or FITC goat anti-mouse (Chemicon; 1/250) secondary antibodies. Finally, sections were rinsed with PBStr three times for 10 min each wash and then 10 min with 50 mM phosphate buffer, pH 7.4, and mounted on slides in a propyl gallate-based antifading solution to reduce photobleaching. Control sections were routinely incubated with secondary antibodies in the absence of primary antibodies.
Sections were imaged with an Olympus BX61WI confocal microscope with a mortised fixed stage with 20X air, 40X apo/340 water, and 60X oil objective lenses. FLUOVIEW FV500 software was used for data acquisition. XY images were scanned in the z axis at 0.5–0.8 μm intervals for 3D reconstruction. Z-plane sections and Z-plane stacks were analyzed using Image J software (National Institutes of Health). For presentation purposes, some images were processed with Adobe Photoshop (Adobe Systems) and Canvas X (ACD Systems). For the analysis of CaM-KII labeling, 8 bit images of individual CEs were background-subtracted and thresholded using Image J to include signals at least 1.5- to 2-fold greater than the background signal. Regions of interest corresponding to the surface area of individual CEs were identified using differential interference contrast (DIC) microscopy and/or Cx35 labeling during double-labeling experiments. To investigate its distribution at CEs, CaM-KII labeling at two concentric, peripheral and central, areas of each CE (considered for this purpose as an ellipse; Fig.4D) were quantified using Image J (Fig.4D). These areas represented 36% and 49% of the total surface of the terminal, respectively. A transition ring between these two regions (15%) was not taken into account (Fig.4D) to better contrast the distribution of labeling between these areas. The rationale behind this analysis was that while CaM-KII is expected to be present in the periphery of the terminals where the majority of the glutamatergic PSDs are located, its localization in the center would be suggestive of its association to gap junctions, which predominate in this area. To illustrate the variability in the distribution of CaM-KII labeling within the population of CEs we used a ratio between labeling at central and peripheral areas. This index was defined as: “Periphery-Center Index” = Labeling Center − Labeling Periphery / Labeling Center + Labeling Periphery, where “Labeling Center” = Σ (labeling Intensity above threshold at Center * number of pixels) / Center area, and “Labeling Periphery” = Σ (labeling Intensity above threshold at Periphery * number of pixels) / Periphery area. Indices approximating −1 indicate prevalence of labeling at the periphery of the CE, whereas those approximating +1 indicate prevalence of labeling at the center. Finally, indices approximating 0 are indicative of a comparable labeling between center and periphery. For presentation purposes, some images were processed using Adobe Photoshop (Adobe Systems, San Jose, CA) and Canvas X (ACD Systems). For colocalization analysis, images of individual CEs were background subtracted and thresholded to include signals that were at least 2–3 fold greater than the scatter labeling at the dendrite. Regions of interest corresponding to individual CEs were identified in Metamorph using transmitted light images and Cx35 labeling. Co-localization was measured as the percentage of the area labeled for one channel that was also labeled for the second channel and the converse.
Fresh tissue samples of ~90 mg and containing both M-cells, as well other smaller reticulo- and vestibulospinal neurons, were obtained from goldfish hindbrain ventral to the cerebellum and between rostral and caudal margins of the cerebellar peduncles. Samples were rapidly stored at −80°C until homogenization. At 4°C, IP buffer (20 mM TRIS-HCl, pH 8.0, 140 mM NaCl, 1% Triton X-100, 10% glycerol, 1 mM EGTA, 1.5 mM MgCl2, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride (PMSF), 5 μg/ml of the protease inhibitors leupeptin, pepstatin A, and aprotinin) were added to the thawed samples for homogenization. Homogenates were sonicated at 4°C 3 times for 10 s each, 35% duty cycle with a W-225R sonicator (Misonix, Inc) and centrifuged for 10 min at 14,000 × g at 4°C, and the supernatant was collected for further analysis (Western blotting and co-immunoprecipitation). Protein was quantified using BCATM Protein Assay Kit (Pierce). Aliquots of supernatants (2 mg protein) were pre-cleared for 1 h at 4°C using 30 μl protein A-coated agarose beads (Santa Cruz BioTech, Santa Cruz, CA, USA) and centrifuged at 14,000 g for 10 min at 4°C. Supernatants were incubated with 2 μg of G301 rabbit anti-αCaM-KII antibody overnight at 4°C on a moving platform, followed by 1 h incubation with 30 μl protein A-coated agarose beads. The mix was then centrifuged as above for 10 min at 4°C. The pellets were washed five times with 500 μl-1 ml of wash buffer (20 mM TRIS-HCl, pH 8.0, 150 mM NaCl, 0.5% NP-40) and incubated at 60°C for 2 min in SDS-PAGE loading buffer containing 5–10% β-mercaptoethanol. Samples were resolved on 10% or 12% SDS-PAGE gels. Proteins were transferred at 30V overnight to nitrocellulose membranes (Schleicher and Schuell, Germany) in standard tris-glycine transfer buffer (pH 8.3) containing 0.5% sodium dodecylsulphate. In order to confirm transfer efficiency, membranes were stained with Ponceau-S (Sigma, USA). Blots were blocked for 1–3 h at room temperature with 4–5% nonfat dry milk in TSTw (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Tween 20), rinsed briefly in TSTw, and then incubated overnight at 4°C using a moving platform with mouse anti-Cx35/36 (2 μg/ml) or mouse anti-ZO-1 (2–3 μg/ml) antibodies. Antibodies were diluted in TSTw containing 1% nonfat dry milk. Membranes were washed with TSTw 3 times for 10 min., incubated for 1 h with horseradish peroxidase-conjugated goat anti-mouse IgG (Santa Cruz biotechnology, USA, diluted 1:7500), washed with TSTw 4 times for 10 min and once for 10 min with 20 mM TRIS-HCL, pH 8.0. Proteins were then resolved by chemiluminescence ECL (Amersham Biosciences, USA). Control samples were precipitated in the absence of CaM-KII antibody.
Surgical and recording techniques were similar to those described previously (Smith and Pereda, 2003; Curti and Pereda, 2004). Briefly, individual VIIIth nerve afferents were sequentially penetrated outside the brain (electrodes contained 2.5 M KCl; 35–45 MΩ) while a second electrode (5 M KAc, 4–12 MΩ) was kept inserted in the M-cell lateral dendrite, 350–400 μm from the axon cap of the cell. Afferents with electrical synapses on the M-cell were identified by the presence of electrotonic coupling potentials when the M-cell antidromic spike was evoked by stimulating the spinal cord (Furshpan, 1964; Lin and Faber, 1988a; Smith and Pereda, 2003; Curti and Pereda, 2004). The resting potential of the afferents and the M-cell averaged 71 mV (±05 mV S.E.M., n=140) and 78.7 mV (±2.5 mV S.E.M., n=95), respectively. Only, afferents with resting potentials of at least −60 mV were used for the analysis. For dye coupling evaluation, the M-cell was recorded intracellularly with electrodes containing a 4% solution of Neurobiotin (MW: 322.8) in 2.5 M KCl and this solution was iontophoretically-injected (400 ms pulses of 50 nA for 20 min), visualized with Avidin-conjugated DAB and examined under transmitted light microscopy. To avoid unintentional extracellular leakage of Neurobiotin around the CEs, the injections were performed in the soma, which is about 400 μm from the CEs terminal field. After injections, waiting periods of 1 to 2 h were used to allow diffusion of the dye (for details see Smith and Pereda, 2003). Coupling coefficients for individual CEs were estimated as the ratio between the amplitude of the electrical component (or coupling potential) of the unitary synaptic potential and the amplitude of the presynaptic spike (coupling coefficient = coupling potential / presynaptic spike). This approximation is a convenient experimental advantage of the M-cell because: 1) the presynaptic action potential is an active signal of constant amplitude between Club endings afferents, and 2) the fast time constant of the M-cell dendrite, estimated in ~400 μs (Fukami et al., 1965), does not disproportionably attenuate fast voltage transients allowing reliable measurements of coupling. An average value of 88 mV (±1.09 mV S.E.M.; n=150) was used for the presynaptic spike amplitude, as the amplitude observed at the recording site at each might not be representative of its amplitude at the contact. The spike recorded at the site of depolarization regenerates in subsequent nodes and the presynaptic terminal (Lin and Faber 1988a; Smith and Pereda, 2003; Curti et al., 2008).
The amino acid sequence of the murine CaM-KIIα (NP_803126.1) regulatory site (Rosenberg et al., 2005) was aligned to its fish counterpart (zebrafish CaM-KIIα; UnitProKB:Q32PV2) and amino acid sequences of αCaM-KII binding and phosphorylation sites of mammalian Cx36 (gi|8928062) (Alev et al., 2008) were aligned to both perch Cx35 (gi|3420235) and Cx34.7 (gi|3420237) using EMBOS (EMBL-EBI, http://www.ebi.ac.uk/Tools/emboss/align/). In all cases, amino acid homology was expressed in terms of percentage of similarity, which takes into account identity as well conserved amino acid substitutions. Group-based Prediction System Version 2.1 (GPS v2.1) was used for prediction of αCaM-KII phosphorylation sites in mammalian Cx36 and perch Cx35 and Cx34.7, using a high threshold with a cutoff of 2.917. High threshold analysis has been validated by large scale predictions of mammalian phosphorylation sites (GPS 2.0; Xue et al., 2008). The cutoff values used in this analysis offered 98.5% accuracy, 100% sensitivity, 98.48% of specificity and 0.7117 Mathew correlation coefficient (MCC) and are were set on a calculated False Positive Rate (FPR) (GPS 2.0; Xue et al., 2008). The prediction of serine 293 as phosphorylation target site in Cx36 required a lower threshold (medium), which still offers high accuracy and sensitivity (90.71% accuracy, 100% sensitivity, 90.56% of specificity and 0.4673 MCC).
The CEs represent the most recognizable input to the M-cells and originate from a population of about 90 saccular afferents that segregate to the distal portion of the lateral dendrite (Fig. 1A) (Bartelmez, 1915). Because of their unusually large size and high expression of Cx35, CEs can be unequivocally identified using antibodies that recognize Cx35 (Fig. 1B) (Pereda et al., 2003; Flores et al., 2008). As illustrated in Fig. 1B, the use of an anti-Cx35 monoclonal antibody (Cx35/36,Chemicon) characteristically yields intense punctate staining at the contact areas between CEs and M-cells (Fig. 1B–D). The number of anti-Cx35 fluorescent puncta observed at individual CEs is consistent with a previous ultrastructural demonstration of 63–200 closely spaced gap junction plaques at individual CEs (Tuttle et al. 1983), suggesting that each punctum represents an individual plaque (Flores et al., 2008). The correspondence of this immunolabeling to CEs was confirmed in double–labeling experiments in which the saccular afferents were also labeled (Fig. 1B–D), where Cx35 labeling was found at the contact area of each afferent and the M-cell dendrite. These contacts were also easily identifiable on the surface of the lateral dendrite using differential interference contrast (DIC) microscopy (inset Fig. 1E). Additionally, this labeling pattern was confirmed using other anti-connexin antibodies (Pereda et al., 2003) and anti-zonula occludens-1 (ZO-1) antibodies (see below), a protein which exhibits a high-degree of co-localization with Cx35 (Flores et al., 2008).
In order to investigate the presence and distribution of CaM-KII at CEs, we performed immunofluorescence labeling using an antibody that exhibits relative selectivity for the αsubunit of rat brain CaM-KII (G-301, directed against a sequence in the autoregulatory domain of the rat brain α-subunit) and that we previously showed recognizes goldfish CaM-KII in western blots (Pereda et al., 1998). As previously reported, this antibody showed intense labeling of the M-cell (Fig.2A–B; Fig. 3B–C), which is consistent with the fact that CaM-KII interacts with proteins associated with cytoskeleton (Sahyou at al, 1985; Ohta et al., 1986). This distinct pattern constituted an internal control for the specificity of CaM-KII labeling. At the surface of the dendrite, this antibody labeled ovoidal structures that were reminiscent, because of the size and location, of CEs. The observed labeling matched that of CEs in double-labeling experiments using a Cx35 antibody (Fig.2C–D), and when visualizing these terminals with the help of DIC (not shown). Thus, our results indicate that CaM-KII can be detected at the areas of contacts between CEs and the M-cell.
Interestingly, in contrast to the predictable regularity of the labeling observed using the Cx35 antibody (Figs. 1B, ,3A),3A), the amount of labeling with the CaM-KII antibody was not constant but highly variable between contiguous CEs (Fig. 3B–C; Fig. 4C). Such variability of labeling was observed in all analyses of CaM-KII immunorectivity (n=5 fish). Furthermore, detailed examination of individual CEs using high-resolution confocal microscopy revealed that the spatial distribution of CaM-KII within the contact areas was also highly variable. While gap junctions are known to be distributed throughout the entire surface of the contact (Fig. 4A), glutamatergic synapses are primarily distributed in the periphery (Fig.4B), where PSDs are located (Tuttle et al., 1986). Such differential segregation makes CEs ideal for examining the relative distribution of diverse synaptic proteins to these co-existent modalities of transmission (see cartoon in Fig. 1A). We found that while CaM-KII labeling was predominant in the periphery, which is consistent with localization to glutamatergic PSDs (Fig. 2B), it was also unexpectedly found at the center of the contacts and, in some cases, distributed throughout their entire surface (Fig. 4C). In order to document and quantify the difference in the spatial distribution of CaM-KII, we expressed this variability as a ratio between the amount of labeling detected at the periphery and that detected at the center of the terminals (Fig. 4D). As illustrated in the summary graph (Fig. 4D), the estimated “Periphery/Center Index” (see Methods for details) was highly variable between CEs, indicating that the intraterminal distribution of CaM-KII is not a consistent feature at these contacts. Confirming such variability in its spatial distribution, the amount of CaM-KII labeling in both the total area and at the center of the contact were highly variable between CEs (Supp. Fig. 1A and 1B respectively). Furthermore, these two measurements were highly correlated (Supp. Fig. 1C), indicating that the amount detected at the center substantially contributes to the total amount of CaM-KII in CEs.
The unexpectedly variable spatial distribution of CaM-KII within CE terminals suggested that it could be associated with proteins other than those located at PSDs. In particular, the presence of CaM-KII labeling in the center of the terminals suggests that CaM-KII could also associate with proteins at Cx35-containing gap junction plaques, which are highly predominant in this region of the contact (Tuttle et al., 1986). To investigate this possibility we looked for co-localization of Cx35 and CaM-KII in double labeling experiments (Fig. 5). We found a clear localization of CaM-KII at many Cx35-labeled puncta (Fig. 5A–B). The degree of co-localization of Cx35 with CaM-KII was lower in comparison to previous studies of co-localization of this connexin with the scaffold protein ZO-1 (see below), and it was highly variable between adjacent CEs. Averaged over individual endings, 66.7% (± 32.5, S.D.) of the area of Cx35 immunolabeling also showed CaM-KII labeling and, conversely, 70.1 % (± 18.6, S.D.) of the area of CaM-KII labeling also showed Cx35 labeling (n=28). As already mentioned above, the degree of colocalization was highly variable, ranging from 6.4% to 81% and 15.9% to 79% for Cx35 over CaM-KII and CaM-KII over Cx35, respectively. Strikingly, the variability of co-localization was found in immediately adjacent Cx35-labeled puncta, within the same region of a single contact (Fig. 5B). Because each punctum is thought to represent an individual gap junction plaque, such differential labeling of adjacent plaques indicates that the co-localization of Cx35 and CaM-KII might not be obligatory in nature and reflective of their functional interaction. Thus, its distribution to the center of the terminals might represent the extent of its functional association with Cx35 at gap junction plaques in CEs.
The co-localization of Cx35 and CaM-KII together with the high homology of this connexin with its mammalian ortholog Cx36 (O’Brien et al., 1998), which has been shown recently to interact with CaM-KII (Alev et al., 2008), suggested that these two proteins might also associate and interact in goldfish brain. To directly explore the possible association of Cx35 with CaM-KII, we first performed in-silico analysis of mammalian and teleost αCaM-KII sequences as well as of mouse Cx36 and perch Cx35 (see methods). Alignments of the mammalian and teleost sequences of the αCaM-KII regulatory region revealed that they are highly homologous (100% similarity in the pseudotarget and pseudosubstrate regions) (Supp. Fig. 2A), and that core amino acids for binding of the αCaM-KII autoregulatory region to αCaM-KII catalytic domain are conserved [Only one gene for αCaM-KII has been identified in teleost; ZFIN ZDB-GENE-051113–72]. Furthermore, sequence alignment of the described Cx36 cytoplasmic loop (CL) binding (pseudotarget) and carboxy terminus (CT) binding (pseudosubstrate) binding regions (Alev et al., 2008) showed that these areas were highly conserved in Cx35, with 86% and 92% similarity for the CT and CL regions, respectively (Fig. 5C and Supp. Fig. 2B). Similar results were obtained for Cx34.7, another perch Cx36 homolog (O’Brien et al., 1998), which showed 76.2% and 84% of similarity for the CT and CL regions, respectively (Supp. Fig. 2B), suggesting that CaM-KII association mechanisms are highly conserved and likely represent an important regulatory feature of neuronal connexins.
To investigate whether Cx35 and αCaM-KII associate in goldfish brain we performed co-immunoprecipitation studies using the G-301 antibody. This analysis was performed in samples obtained from a small area of the goldfish hindbrain containing the colossal M-cells, along with some other smaller reticulospinal and vestibulospinal neurons (see Supp.Fig.3 and Methods). Thus, this sample should be representative of the M-cell contents. αCaM-KII was immunoprecipitated from these hindbrain homogenates and immunoblots were probed with the Cx35 monoclonal antibody (Cx35/36, Chemicon). Immunoprecipitation of αCaM-KII resulted in the detection of a Cx35 band at the predicted molecular weight (Fig. 5D, lanes 2 and 3), which were comparable to migration profiles previously reported (Flores et al. 2008). While western blots detected a small presence of Cx35 (Fig. 5D, lane 1) in the input, more prominent bands were seen with the immunoprecipitation from two different samples (Fig. 5D, lanes 2 and 3), indicating that αCaM-KII more efficiently pulled-down this connexin as a result of their association. Interestingly, multiple bands were identified with the Cx35 antibody, suggesting the existence of multiple phosphorylation states of this connexin (Fig. 5D, lanes 2 and 3).
Recently, four αCaM-KII phosphorylation sites were described for Cx36 (Alev et al., 2008). In order to investigate the presence of putative phosphorylation sites forαCaM-KII in Cx35 we performed in-silico analysis using Group-based Prediction System version 2.1 (GPSv.2.1) software (see methods) in mouse Cx36, and perch Cx35 and Cx34.7 sequences. This analysis confirmed the presence of residues S110, T111, S277 and S315 in Cx36 as αCaM-KII targets. Five sites, S110, T111 both in the intracellular loop, and S128, S276 and S298 in the carboxy terminus were identified for Cx35 (Supp. Fig.4), where the last two correspond to sites S293 and S315 of Cx36, respectively. Residues S110 and S276 (S110 and S293 of Cx36) (Urschel et al., 2006) were previously described as PKA phosphorylation sites (Kothmann et al., 2007; O’Brien, 1998), suggesting that they are targets shared by both kinases. S298 (equivalent to S315 in Cx36) on the other hand seems to be exclusive for αCaM-KII. Only three putative sites were identified for Cx34.7: S110, S277 and S300, which correspond to residues S110, S276 and S298 of Cx35, and S110, S293 and S315 of Cx36, respectively. Thus, our in-silico analysis is consistent with the possibility of multiple phosphorylation states for Cx35.
Taken together, our in-silico and biochemical results suggest that CaM-KII associates with Cx35 in goldfish brain. Several proteins have been also shown to interact with connexins indicating the existence of protein complexes associated to gap junction channels (Herve et al., 2004). We have recently reported that ZO-1 extensively co-localizes and associates with Cx35 at CEs (Flores et al., 2008). In contrast to the less abundant and highly variable labeling of CaM-KII, ZO-1 was shown to extensively co-localize with Cx35 (85.53%, ± 14.45 S.D., of Cx35 colocalizes with ZO-1 labeling and 86.06%, ± 15.45 S.D., of ZO-1 colocalizes with Cx35, n=116; Flores et al., 2008) (Fig. 6A) and to directly interact with the last four aminoacids of its carboxy-terminus (Flores et al., 2008), suggesting that this scaffold protein might play a structural role in gap junction plaques at these terminals. Thus, we hypothesized that if CaM-KII directly interacts with Cx35, immunoprecipitation studies with the G-301 antibody should be also able to detect ZO-1. For this purpose, we performed co-immunoprecipitation studies in which CaM-KII was immunoprecipitated from hindbrain homogenates. In this case, immunoblots were probed with a monoclonal ZO-1 antibody, also used for immunolabeling of ZO-1. Immunoprecipitation of CaM-KII resulted in the detection of a ZO-1 band at the predicted molecular weight (Fig. 6B). These results suggest that CaM-KII associates with Cx35, and together with ZO-1, is likely to be part of the same macromolecular complex where channel-forming, scaffold and regulatory proteins coexist.
While both sequence alignments and immunoprecipitation studies suggest that Cx35 and αCaM-KII interact at CEs, co-localization studies suggest that this association might not be obligatory, as variable labeling was found even between plaques within the same terminal. Because CaM-KII is known to translocate and aggregate to target sites following activity (Shen and Meyer, 1999; Merill et al., 2005; Rose et al., 2009; Gleason in al., 2003), its intraterminal distribution could be indicative of the degree of potentiation at individual CEs. Accordingly, the variability of CaM-KII labeling between immediately adjacent CEs was reminiscent of the variability of unitary electrical components of the mixed (electrical and chemical; Fig. 7A) synaptic potential evoked by stimulation of a single CE, which differ dramatically in amplitude (Smith and Pereda, 2003). To further demonstrate the extent and consistency of this phenomenon we compared the strength of electrical coupling between individual CE afferents and the M-cell lateral dendrite between different fish. For this purpose, we obtained multiple simultaneous pre- and postsynaptic recordings sequentially from individual CE afferents and the same M-cell lateral dendrite. As shown in Figure 7A where seven consecutive single terminals recordings are illustrated, the amplitude of the electrical components differed dramatically. [Characteristically, most CE single terminal recordings lack a chemical component (Lin and Faber, 1988b) and transmitter release is enhanced by co-activation of multiple terminals (see Pereda et al., 2004)] This same variability was observed for five different fish in which unitary potentials were recorded in each M-cell dendrite during stimulation of 6–8 terminals (Fig.7C). Given that these single terminals originate from almost 90 large saccular afferents (Bodian, 1937), this sample represents approximately 8–10% of the total population of CEs in each lateral dendrite. In contrast to previous studies (Smith and Pereda, 2003), we expressed this variability in terms of coupling coefficients of orthodromic coupling produced by presynaptic action potentials (Fig. 7C), an estimate which is greatly facilitated by the fast time constant of the lateral dendrite of the M-cell (Fukami et al., 1965) (see methods). The properties of the presynaptic action potentials are highly constant across the population of CEs (Curti et al., 2008). Furthermore, because they were all recorded in the same postsynaptic cell (which makes corrections for differences in input resistance unnecessary), these coupling coefficients are highly correlated with the gap junctional conductance of individual terminals. The differences in conductance were confirmed by tracer coupling to the afferents after injection of the M-cell with Neurobiotin, where unlabeled afferents and afferents showing different degrees of staining were observed in the same dendritic area (Figure 7B). This variability of coupling was also observed for the whole population of recorded afferents. Interestingly, both the spatial distribution of CaM-KII (expressed as Periphery / Center Index; Mean = −0.11 ± 0.41 S.D., n = 43) (Figure 7D) and the coupling coefficients of individual CEs (Mean = 0.008 ± 0.003 S.D., n = 92) (Fig. 7E) showed wide variability and comparable distributions. The similarity opens the possibility that differences in the spatial distribution of CaM-KII could represent differences in the functional state of electrical synapses at these terminals, where the amount and localization of CaM-KII to gap junctions would be indicative of the degree of potentiation. Unfortunately, detecting changes in the distribution of CaM-KII following the application of the high-frequency stimulating protocols leading to synaptic potentiation (Yang et al., 1990; Pereda and Faber, 1996; Pereda et al., 1998) was impractical and difficult to interpret due to: 1) the inability of identifying the contacts in the dendrite which correspond to the sub-population of afferents stimulated by the extracellular electrode in anatomical sections and 2) to the fact that within the population of stimulated afferents, although most synapses potentiate, others were shown to depress as a result of the same stimulating protocol (see Smith and Pereda, 2003, figure 4), indicating that both depression and potentiation co-exist in the population of stimulated afferents.
“These endings are large enough to be useful for the elucidation of many details and also for the experimental studies on which we are engaged at present.”
As pointed out by Bartelmez and Hoerr seventy-six years ago (Bartelmez and Hoerr, 1933), the unusual large size of the CEs makes them a valuable model to test specific questions regarding vertebrate synaptic transmission. The present analysis was only possible at these terminals, which offered an unparalleled access for exploring the distribution of CaM-KII within a single contact and the possibility to correlate it with well-established pharmacological and physiological properties. CEs have recently become a model for the study of interactions between chemical and electrical synapses (Pereda et al., 1998; Smith and Pereda, 2003; Pereda et al., 2004). We previously showed that within these mixed contacts the activity of glutamatergic synapses not only leads to their own potentiation but also of nearby electrical synapses (Smith and Pereda, 2003). Our results indicate that a likely major molecular candidate for this interaction is the enzyme CaM-KII. Further, our analysis of the intraterminal distribution of this kinase and its association to Cx35 indicates that CaM-KII should be considered not only a major component of glutamatergic PSDs but also of gap junction-mediated electrical synapses.
A characteristic property of CaM-KII is its ability to translocate to active synaptic sites (Shen and Meyer, 1999; Merill et al., 2005; Rose et al., 2009; Gleason in al., 2003). It has been proposed that the level of CaM-KII associated with glutamatergic PSDs at any given time may be a combination of 1) transiently associated kinase, and 2) a population of this kinase that is retained after translocation and that might encode the history of correlated neuronal activity (Schulman, 2004). Translocation to contiguous synapses has been shown to occur (Rose et al., 2009; Merill et al., 2005), including to GABAergic synapses (Madsen and Caroll, personal communication), suggesting that gap junctions can also be the target of translocation/re-distribution of CaM-KII from nearby glutamatergic synapses. In contrast to glutamatergic synapses, where accumulation of this kinase at PSDs can be difficult sometimes to detect (Dosemeci et al., 2001; Merill et al., 2005), the large number and intraterminal distribution of gap junctions at CEs (predominant at the center of the contact) made it easier to detect their association with CaM-KII. Based on the association to Cx35 that we report here (see below), the distribution of CaM-KII might similarly reflect the history of synaptic activation of individual CEs. That is, the distribution of CaM-KII labeling within the population of CEs is consistent with the variability in coupling between adjacent contacts, previously shown to result from terminal-specific activity-dependent modifications initiated by glutamatergic synapses at each of these terminals (Smith and Pereda, 2003), and they both have comparable distributions. While this conclusion will require of further confirmation, there is experimental evidence that supports this possibility: 1) activated CaM-KII leads to enhancement of electrical and chemical transmission of CEs (Pereda et al., 1998), 2) the activity of CaM-KII is required for activity-dependent potentiation of CEs (Pereda et al., 1998), and finally, 3) changes in glutamatergic transmission parallel changes in electrical coupling in CEs and, accordingly, the amount of CaM-KII at individual spines of hippocampal CA1 neurons was shown to be positively correlated with the strength of its glutamatergic synapses (Asrican et. al., 2007).
Furthermore, our in-silico analysis and biochemical studies, in combination with our previous data (Pereda et al., 1998), suggest that association of CaM-KII with Cx35 leads to its activation, ultimately resulting in potentiation of electrical transmission in CEs. According to current models (Bayer et al., 2001; Schulman, 2004) the binding of CaM-KII to Cx36 (Alev et al., 2008) and likely to Cx35 (based on the sequence homology and Co-IP we report here), will lock this kinase in a persistently active state, even after dissociation of calmodulin. More specifically, the αCaM-KII autoinhibitory domain is considered a gate consisting of a pseudosubstrate segment, that blocks binding of substrates to the “S site”, as well as a pseudotarget segment that blocks access to its target proteins (“T site”) (Bayer et al., 2001; Hudmon and Schulman, 2002; Schulman, 2004). Binding of Ca++/calmodulin opens the gate and allows substrates to bind (kinase activation), thereby enabling interactions with some targets such as to the NR2B subunit of the NMDA receptor (Bayer et al., 2001), Drosophila Eag potassium channel (Sun et al., 2004) and Cx36 (Alev et al., 2008) all of which contain similar “pseudosubstrate” and “pseudotarget” binding regions. This mechanism would be equivalent to the classic effect of autophosphorylation of Thr286, positioned like a “wedge” in the inhibitory gate, further displacing it (Singla et al., 2001; Schulman, 2004) and enabling interactions with additional targeting proteins (Bayer et al., 2001; Schulman, 2004). As mentioned above, our previous work showed that intradendritic injections of a constitutively active form of αCaM-KII in the M-cell led to a persistent potentiation of the electrical component of the mixed synaptic potential evoked by stimulation of CEs and CaM-KII activation was found to be required for the induction of activity-dependent potentiation of this synaptic potential (Pereda et al. 1998). Taken together, these evidence suggest that the activity of CaM-KII indeed leads to potentiation of electrical transmission in CEs, and differences in the spatial distribution of this kinase could thus represent differences in the functional state of electrical synapses at these terminals, where the amount and localization of CaM-KII to gap junctions would be indicative of the degree of potentiation.
Our current experimental approaches made impractical to detect changes in the distribution of CaM-KII following activity-dependent protocols. This question could be more easily addressed in future experiments using zebrafish embryos, where it is possible to track the translocation and/or redistribution of fluorescently-tagged kinase to synaptic sites (Gleason, 2003). Nonetheless, these results provide, for the first time, evidence of the distribution and association of CaM-KII to identifiable electrical synapses in-vivo, where it is possible to correlate the presence of this kinase with its known functional properties.
The in-silico analysis revealed the existence of multiple phosphorylation sites for αCaM-KII in Cx35. These predictions were consistent with those recently reported for Cx36, (Alev et al., 2008), which were confirmed by biochemical approaches, providing a high degree of confidence on this analysis. By extending our analysis to an additional fish homolog of Cx36, Cx34.7, we found that Cx36, Cx35, and Cx34.7 shared most of the identified residues. The analysis also showed that residues S315 / S298 / S300 in Cx36, Cx35 and Cx34.7, respectively, constitute exclusive phosphorylation sites for CaM-KII that are not shared with other kinases. On the other hand, residues S110 and S293 in Cx36, S110 and S276 in Cx35, and S110 and S277 in Cx34.7, are phosphorylation sites shared with cAMP-dependent protein kinase A (O’Brien et al., 1998; Mitropoulou and Bruzzone, 2003; Ouyang et al., 2005; Urschel et al., 2006), a kinase also shown to promote enhancement of electrical transmission at these terminals (Pereda et al., 1994; Cachope et al., 2007). Interestingly, our IP experiments (restricted to a small portion of the hindbrain that mainly contains the M-cells) showed multiple bands for Cx35 pulled down by CaM-KII. Although final proof would require dephosphorylation of samples with protein phosphatase, this observation is consistent with the possibility that Cx35 coexists at multiple states of phosphorylation. We did not investigate this possibility further, which will be the focus of future research efforts. In summary, our sequence alignment shows that CaM-KII binding regions and phosphorylation sites are highly similar between Cx36 and its fish homologs, indicating that essential regulatory mechanisms are highly conserved and likely to pertain to all Cx35- and Cx36-mediated electrical synapses. These mechanisms could include, as shown for glutamate receptors, direct phosphorylation of the channels themselves (Derkach et al., 1999) or the promotion of their insertion in the plasma membrane (Shi et al., 1999; Liao at al., 2001; Shi et al., 2001).
In contrast to the extensive information on proteins associated with chemical synapses (Kim and Sheng, 2004) involved in channel insertion, anchoring or removal from the plasma membrane (Lin et al., 2004; Malinow et al., 2000; Migaud et al., 1998; Mori et al., 1998; Shi et al., 1999; Sprengel et al., 1998; Steigerwald et al., 2000), very few proteins have been shown to associate to the neuronal gap junctions. It is currently accepted that gap junctions are not only comprised of the defining intercellular channel proteins, but also associated scaffold and regulatory proteins (Herve et al., 2004). Neuronal gap junctions in particular, characteristically exhibit a PSD-like structure in EM, described as a “semi-dense cytoplasmatic matrix” (Sotelo and Korn, 1978), and which likely represents (similar to glutamatergic PSDs) proteins associated with these intercellular channels. Cx36 (Condorelli et al., 1998), the most prevalent neuronal connexin that is widely distributed throughout the mammalian brain (Condorelli et al., 2000), was shown to interact with two relevant proteins: CaM-KII (Alev et al., 2008) and ZO-1 (Li et al., 2004). ZO-1 is a member of the MAGUK family of proteins which was reported to interact with several other connexins (Itoh et al., 1999). Our recent work indicates that ZO-1 also interacts with Cx35 in CEs (Flores et al., 2008). Because its co-localization with Cx35 is extensive (Flores et al., 2008), ZO-1 is likely to play a structural role and the properties of the ZO-1/Cx35 association suggest the existence of a dynamic relation between these two proteins, possibly including a role of ZO-1 in regulating gap junctional conductance at these highly modifiable electrical synapses (Flores et al., 2008). Our present results show that CaM-KII also associates with Cx35 at CEs. This study extends the observations of Alev and colleagues (Alev et al., 2008) showing that CaM-KII associates with neuronal connexins in anatomically identifiable, native, electrical synapses. More importantly, our data also indicates that CaM-KII is likely a non-obligatory component of electrical synapses. That is, its presence is highly variable, even between gap junction plaques within the same contact, and likely is linked to differences in the degree of potentiation between individual terminals (see above). In comparison with ZO-1, which would play a more permanent or structural role, the presence of CaM-KII seems to be non-obligatory and possibly regulated by neural activity (Fig. 8). This conclusion is consistent with the dynamic translocation properties of CaM-KII and whose presence is thought to represent the history of synaptic activation. Thus, our results indicate that ZO-1 and CaM-KII probably belong to the same macromolecular complex, and together with other identified proteins [association to zonula occludens proteins ZO-2 (Ciolofan et al., 2006) and ZO-3 (Li et al., 2009) and calmodulin (Burr et al., 2005) were also reported] as well as those that might be identified in the future, are likely components of the “semi-dense cytoplasmatic matrix”, associated with intercellular channels in neuronal gap junctions.
In summary, CaM-KII is not only a major component of glutamatergic postsynaptic densities but is also an important and dynamic component of gap junction-mediated electrical synapses in CEs. Given the widespread distribution of Cx35- and Cx36-mediated electrical synapses, our data suggest that the properties observed at these identifiable gap junctions might also apply to electrical synapses elsewhere, including those in mammalian brain.
This research was supported by National Institutes of Health grants DC03186 and NS0552827 (to A.E. Pereda). We thank Carl Castillo for help with immunochemistry and Joseph Zavilovitz for help with the confocal microscope during initial experiments. We also thank Costa Dobrenis and the Morphology Core of the Rose F. Kennedy Center for outstanding support. Finally, we thank David Spray, Rolf Dermietzel and Theresa Szabo for useful discussions and comments on the manuscript.