Tubulin-Dependent Transport of Connexin-36 Potentiates the Size and Strength of Electrical Synapses

Connexin-36 (Cx36) electrical synapses strengthen transmission in a calcium/calmodulin (CaM)/calmodulin-dependent kinase II (CaMKII)-dependent manner similar to a mechanism whereby the N-methyl-D-aspartate (NMDA) receptor subunit NR2B facilitates chemical transmission. Since NR2B–microtubule interactions recruit receptors to the cell membrane during plasticity, we hypothesized an analogous modality for Cx36. We determined that Cx36 binding to tubulin at the carboxy-terminal domain was distinct from Cx43 and NR2B by binding a motif overlapping with the CaM and CaMKII binding motifs. Dual patch-clamp recordings demonstrated that pharmacological interference of the cytoskeleton and deleting the binding motif at the Cx36 carboxyl-terminal (CT) reversibly abolished Cx36 plasticity. Mechanistic details of trafficking to the gap-junction plaque (GJP) were probed pharmacologically and through mutational analysis, all of which affected GJP size and formation between cell pairs. Lys279, Ile280, and Lys281 positions were particularly critical. This study demonstrates that tubulin-dependent transport of Cx36 potentiates synaptic strength by delivering channels to GJPs, reinforcing the role of protein transport at chemical and electrical synapses to fine-tune communication between neurons.


Introduction
Synaptic plasticity at glutamatergic chemical synapses involves highly orchestrated molecular and morphological processes that fine-tune communication between neurons. Short-to long-term functional and morphological adaptations require dynamic trafficking and turnover of channel and receptor proteins. While the molecular machinery involved in chemical synaptic plasticity and the respective signaling network has been extensively investigated, fewer details are known about the plasticity of electrical synapses. In vivo, modifiable responses of electrical synapses have been found in both lower and higher vertebrates [1,2]. In vitro, a phenomenon coined "run-up" was found to rapidly and profoundly strengthen gap-junctional conductance exclusively at Cx36 gap-junction plaques (GJPs); an observation later was shown to occur in a Ca 2+ /calmodulin-dependent kinase II (CaMKII)-dependent manner [3][4][5]. The Cx36 plasticity phenomenon parallels the interaction

Electrophysiology
Neuro-2a cell pairs expressing EGFP-tagged wild-type Cx36 or Cx36∆279-292 were used for dual whole-cell patch clamp recording. In brief, Neuro-2a cells were perfused in a bathing solution supplemented with 100 µM Colchicine for 1, 3, or 24 h and the run-up was induced as previously described [5]. Whole-cell patch clamp recordings were performed as previously described [5,26] over a time course up to 70 min at room temperature.

Fluorescent Imaging and Gap-Junction Plaque Measurements
Transfected Neuro-2a cells were fixed with 4% formalin for 15 min at room temperature. Cells were subsequently rinsed with PBS and distilled water prior to mounting with Fluoroshield™ with DAPI (Sigma, Oakville, ON, Canada) mounting medium. Mounted slides were visualized using a Zeiss LSM 700 confocal microscope controlled by the ZEN 2010 software program (Oberkochen, Germany) at room temperature. Cells were imaged using the Plan-Apochromat 63x/1.40 Oil DIC M27 or the EC Plan-Neofluar 40x/1.3 Oil M27 oil immersion lenses. Images were generated at a high resolution of up to 2048 × 2048 pixels and an average of 4 scanning repetitions in the single plane or each plane of a Z-stack. The single or multi-track mode was selected accordingly for the compilation of localization studies. The gap-junction plaque area (µm 2 ) was measured using the ImageJ (NIH, Bethesda, MD, USA) free-hand tool. The percentage of Cx36 gap-junction-paired cells versus the total paired Cx36-expressing cells was measured as previously described [27]. Mander's overlap coefficient was used to quantify the overlap between corresponding fluorescent signals derived from Cx36 and cell organelles in co-transfected cells. Mander's overlap coefficient was calculated using the ZEN 2010 program. Co-localization was accepted at a Mander's coefficient > 0.5; this threshold indicated that the occurrence of true co-localization surpassed the probability of chance.

Fluorescence Recovery After Photobleaching (FRAP)
For the quantitative analysis of connexon trafficking and gap-junction regeneration, FRAP was performed 48 h post-transfection using the Zeiss LSM 700 confocal microscopy in combination with Zeiss 63X (Plan-Apochromat, DIC M27 1.4) oil immersion lens and the Zen 2010 software. The microscope was equipped with an incubation chamber to maintain cells at 37 • C in DMEM without phenol red. Since previous reports [20,28] have described connexon replenishment occurring at the lateral ends of the gap-junction plaque, this was selected as the region of interest (ROI). ROIs were manually drawn at the non-junctional plasma membrane and lateral ends of the gap-junction plaque using the rectangle tool. ROIs were bleached using the 405-nm laser line at 70% emission strength with a single event of 20 iterations. Prior to recording, confocal images were optimized along the focal plane yielding in the largest GJP surface area. The mobile fraction (%) and intensity of fluorescence (RFU) were recorded in vivo for up to 200 s per ROI, and fluorescence values were collected at 1-s intervals. Observations were recorded with a pinhole of 1.0 airy units (AU) with a scan area of 31.1 µm × 31.1µm, zoom level between 1-3, and resolution of 512 × 512 pixels. Consistent parameters were used across all experiments. Experiments with considerable fluctuations in fluorescence intensity or ROI frame-shifts were not included in the analyses. Background subtraction was carried out with an ROI of the extracellular space to improve the signal-to-noise ratio. Loss of fluorescence was corrected next by expressing the previous values as a fraction of an unbleached control ROI. FRAP recordings were then normalized by expressing the corrected fluoresces as a fraction of the averaged pre-bleached values. The mobile fraction (M f = (F max − F 0 )(1 − e −kt )) and half-time of recovery (T 1/2 = ln 2/k) were calculated by the mono-exponential association equation in GraphPad: F t = F 0 + (F max − F 0 )(1 − e −kt ), where F t is the fluorescence after background subtraction at time t; F 0 and F max are the fluorescence values immediately after and at the end of recovery following the bleaching event, respectively; and k is the first-order rate constant for recovery. Repopulation figures and X-Y scatterplots were done using the R software. Sample sizes ranged based on the efficiency of gap-junction plaque formation.

Total Internal Reflection Fluorescence (TIRF).
For all experiments, Neuro-2a cells were plated onto 35-mm MatTek dishes and co-transfected with pShuttle-mCherry-Tubulin and the pEGFP-Cx36 wild-type or mutants as previously outlined. Time-lapsed TIRF microscopy was performed using the Zeiss Observer.Z1 spinning-disk microscope in combination with the Zeiss 100X (Plan-Apochromat, DIC, M27, 1.46) oil immersion lens, Photometrics Evolve™512 camera, and the Zen 2 (2014) software. The microscope was equipped with an incubation chamber to maintain the temperature at 37 • C and CO 2 levels at 5%. Cells were imagined in DMEM without phenol red. Images were acquired at a resolution of 512 × 512 pixels in 20-s intervals for a total duration of 400 s. Images were processed in Imaris (Zurich, Switzerland) by tracking single particles expressing EGFP within the TIRF field.

Molecular Modeling
A helical segment spanning amino acids 47-62 of the RB3 protein stathmin-like-domain X-ray structure [29] was substituted with a sequence from the Cx36 carboxy tail region (277-292) using PyMOL 1.8.6 and further refined with the FlexPepDock module of Rosetta 3.9 [30]

Statistics
Beeswarm boxplots were generated in GraphPad Prism 6 (GraphPad Software, La Jolla, CA, USA). Plaque repopulation graphs and mobile fraction vs. half-time of recovery XY scatterplots were created in R-statistical program (The R Foundation, Aukland, New Zealand). All data are represented as the mean ± SEM. Unless otherwise stated, statistical analysis was carried out using SPSS (IBM Corporation; Armonk, NY, USA) and significance was determined by the Kruskal-Wallis test. For the FRAP data, statistical analysis was carried out using GraphPad and statistical significance was determined by the Kruskal-Wallis test followed by a Dunn's multiple comparison test. p < 0.05 was considered significant. XY plots of the mobile fraction and half-time of recovery represented the mean ± CI. All comparisons were made to the wild-type (WT) control unless otherwise stated. Sample sizes and p values are provided throughout the manuscript and figure captions.

Cx36 Interaction with the Tubulin Cytoskeleton is Required for Electrical Plasticity
The interaction between tubulin at the carboxy-terminal (CT) of Cx36 was initially identified by mass spectrometry using both affinity-purified samples and immunoprecipitated protein complexes from murine brain lysates [19]. This result was independently confirmed by a BioID screen using a full-length rat Cx36 protein as bait in the Neuro-2a mouse neuroblastoma cell line ( Figure 1A). Within the Cx36-CT, a potential tubulin site was identified by sequence comparison with the analogous tubulin binding sequence in connexin-43 (Cx43) [20,21]. This sequence overlapped with the previously reported binding sites for calmodulin (CaM) [22,31] and the calcium/calmodulin-dependent protein kinase II (CaMKII) [4], all of which demonstrated high conservation across Cx36 orthologs belonging to the major vertebrate Chordata subphylum ( Figure 1B). Since the predicted tubulin-binding site and the established CaMKII/CaM-binding sites are situated within a fourteen amino-acid segment, any mechanistic model must consider competition among these proteins. This signature motif was absent in other connexin isoforms ( Figure S1).

Figure 1.
Tubulin interaction at the Cx36 carboxyl terminus potentiates electrical plasticity. (A) Fulllength Cx36 protein was expressed as a fusion protein with the BirA* ligase to perform a BioID assay using Neuro-2a cell lysate. Lysates were subjected to western blot analysis before (−) and after (+) the addition of streptavidin-conjugated beads. Immunodetection identified tubulin as an interaction candidate of Cx36. (B) Sequence alignment of the tubulin-binding motif confirmed in Cx43 elucidated the potential Cx36 interface, localized to the Cx36 carboxyl terminus (CT). The putative binding motif of Cx36, shaded in blue, demonstrated robust conservation across orthologs belonging to the major vertebrate Chordata subphylum. The † symbol indicates the selective use of the Cx35b protein corresponding to the gjd2b gene. Conserved residues within the Cx36 CT are indicated by the asterisks. Both CaM-and CaMKII-binding sites appeared to share some overlap with the tubulin- (A) Full-length Cx36 protein was expressed as a fusion protein with the BirA* ligase to perform a BioID assay using Neuro-2a cell lysate. Lysates were subjected to western blot analysis before (−) and after (+) the addition of streptavidin-conjugated beads. Immunodetection identified tubulin as an interaction candidate of Cx36. (B) Sequence alignment of the tubulin-binding motif confirmed in Cx43 elucidated the potential Cx36 interface, localized to the Cx36 carboxyl terminus (CT). The putative binding motif of Cx36, shaded in blue, demonstrated robust conservation across orthologs belonging to the major vertebrate Chordata subphylum. The † symbol indicates the selective use of the Cx35b protein corresponding to the gjd2b gene. Conserved residues within the Cx36 CT are indicated by the asterisks. Both CaM-and CaMKII-binding sites appeared to share some overlap with the tubulin-binding motif, potentially indicating competitive or cohesive binding mechanisms. The CaMKII-binding motif is outlined in blue, and the CaM motif is outlined in red. (C,D) Neuro-2a cells cultured in a bathing solution supplemented with 100 µM Colchicine for 1 h (n = 5) or 24 h (n = 6) demonstrated a significant reduction in gap-junctional (GJ) conductance relative to the untreated wild-type (n = 6). Data are the mean ± SEM; **** p < 0.0001, Kruskal-Wallis test; each sample was compared to the untreated wild-type (WT). (E,F) Intracellular perfusion with monomeric tubulin (50 µM; n = 5) through the recording electrode was equally sufficient to block an increase in junctional currents in comparison to the untreated WT (n = 4). Data are the mean ± SEM; **** p < 0.0001, Mann-Whitney test.
In previous studies [3][4][5], we have proposed that microtubule-dependent trafficking of Cx36 serves as a potential modality in achieving run-up plasticity by facilitating activity-dependent transport of Cx36 to gap-junction plaques. Here, the predicted role of Cx36 in run-up plasticity was demonstrated by dual whole-cell patch clamp experiments measuring electrical coupling dynamics in pairs of Neuro-2a cells expressing EGFP-tagged Cx36 constructs. Targeted ablation of the putative tubulin-binding motif [3,5] was used to investigate the involvement of this motif to run-up plasticity. In Neuro-2a cell pairs with a gap-junction plaque (GJP) present at the juxtamembrane, gap-junctional (GJ) conductance increased by a factor of about 10 during a ten-minute recording; by contrast, in cells expressing the Cx36∆279-292 mutant, this run-up was severely blunted ( Figure S2). This result showed that GJPs formed by Cx36 without the shared binding region for tubulin and CaM/CaMKII had reduced potential for plasticity.
To determine the subcellular localization of the tubulin-binding deficient mutant, Cx36Δ279-292-EGFP or the Cx36 wild-type control were co-expressed with mCherry-tagged tubulin or the  Figure 2B, examples of Cx36 gap-junction plaques are indicated by circles, and in Figure 2C, examples of annular junctions formed in Cx36∆279-292-expressing cells are indicated by circles. Scale bar: 10 µm, nuclear DAPI staining in blue. Quantification of gap-junction plaque (GJP) size (D) and incidences of GJP-pairings (E) demonstrated a significant reduction in connexon recruitment in cells expressing the tubulin-binding deficient mutant Cx36∆279-292, indicative of trafficking impairments. Data are mean ± SEM; for GJP area, sample sizes (n) were 28 and 26 for Cx36 WT and ∆279-292 proteins, respectively. For GJP-paired cells, n = 30, where 10 images were collected in 3 independent experiments. * p < 0.05, *** p < 0.0001, Kruskal-Wallis test. (F) Co-localization of Cx36 vesicles to organelles was quantified by Mander's overlap coefficient to decipher its intracellular localization. Relative to the wild-type, expression of the tubulin-binding deficient protein resulted in a reduction in ER, Golgi, and tubulin co-localization; caveolin co-localization remained unaffected. Data are mean ± SEM; sample sizes (n) of WT Cx36 were n = 30 for the ER, n = 30 for the Golgi, n = 34 for tubulin, and n = 40 for Caveolin-1. Sample sizes of Cx36∆279-292 were n = 35, n = 15, n = 36, and n = 33 for colocalization of the ER, Golgi, tubulin, and Caveolin-1 proteins, respectively. * p < 0.05, *** p < 0.0001, Kruskal-Wallis test; each sample was compared to its corresponding wild-type values. (G) Cx36 vesicular diameter of cells co-transfected with caveolin-1 demonstrated an increase in size with mutant expression over the wild-type. Data are mean ± SEM; sample size (n) of WT Cx36 was n = 50 vesicles. ** p < 0.001, Kruskal-Wallis test; data were collected from 3 independent experiments.
How were Cx36 connexons delivered to GJPs in the absence of tubulin binding? In general, connexons can replenish GJPs by two distinct pathways. Targeted direct delivery of connexons to the GJP relies on a complex of cytoskeletal and adherence-junction proteins [32]. Alternatively, connexons are routed to the plasma membrane and subsequently recruited into the GJP at the lateral ends [28], [33], [34]. This process occurs despite the ability of microtubules to anchor directly at the GJP [21]. How were Cx36 connexons delivered to GJPs in the absence of tubulin binding? In general, connexons can replenish GJPs by two distinct pathways. Targeted direct delivery of connexons to the GJP relies on a complex of cytoskeletal and adherence-junction proteins [32]. Alternatively, connexons are routed to the plasma membrane and subsequently recruited into the GJP at the lateral ends [28,33,34]. This process occurs despite the ability of microtubules to anchor directly at the GJP [21].
To answer this question, fluorescence recovery after photobleaching (FRAP) was used to determine recovery kinetics of EGFP-tagged Cx36 wild-type or ∆279-292 proteins imaged at 1.0-s intervals preand post-bleach iterations. For visualization and quantification of connexon recovery, the fluorescence intensity of each region of interest (ROI) during the post-bleach recovery phase was corrected for background noise and normalized against the pre-bleached fluorescence intensity values. After bleaching, we observed that the plasma membrane (PM) of wild-type Cx36 recovered more efficiently than at the GJP (mobile fraction (M f ) in %, GJP: 25.6 ± 0.6, n = 28; PM: 69.6 ± 3.6 n = 12, p < 0.01 × 10 −2 ; half time of recovery (T 1/2 ), GJP: 11.5s; PM: 23.5s) ( Figure 4A,D). The higher M f seen at the PM suggested that, under control conditions, Cx36 is initially delivered to the plasma membrane and subsequently routed to the plaque, where it becomes stabilized, supporting the mechanisms previously described [28,34,35]. In Cx36∆279-292-expressing cells, the M f was again higher at the PM than at the GJP (GJP: 50 ± 0.5%, n = 22; PM: 97.1 ± 1.9%, n = 10, p < 0.01 × 10 −2 ), suggesting that no changes were made to the route of delivery. Cx36∆279-292 expression also led to a T 1/2 that was faster at the PM and GJP (GJP: 0.9s; PM: 7.3s) than the WT ( Figure 4B,D). We conclude that the interaction with the tubulin cytoskeleton aided in the temporal aspects of Cx36 incorporation from the PM into the GJP in addition to quantity of delivery, which was verified by GJP size and formation.   To answer this question, fluorescence recovery after photobleaching (FRAP) was used to determine recovery kinetics of EGFP-tagged Cx36 wild-type or Δ279-292 proteins imaged at 1.0-s intervals pre-and post-bleach iterations. For visualization and quantification of connexon recovery, the fluorescence intensity of each region of interest (ROI) during the post-bleach recovery phase was corrected for background noise and normalized against the pre-bleached fluorescence intensity values. After bleaching, we observed that the plasma membrane (PM) of wild-type Cx36 recovered more efficiently than at the GJP (mobile fraction (Mf) in %, GJP: 25.6 ± 0.6, n = 28; PM: 69.6 ± 3.6 n = 12, p < 0.01 × 10 −2 ; half time of recovery (T1/2), GJP: 11.5s; PM: 23.5s) ( Figure 4A,D). The higher Mf seen at the PM suggested that, under control conditions, Cx36 is initially delivered to the plasma membrane and subsequently routed to the plaque, where it becomes stabilized, supporting the mechanisms previously described [28], [34], [35]. In Cx36Δ279-292-expressing cells, the Mf was again higher at the PM than at the GJP (GJP: 50 ± 0.5%, n = 22; PM: 97.1 ± 1.9%, n = 10, p < 0.01 × 10 −2 ;), suggesting that no changes were made to the route of delivery. Cx36Δ279-292 expression also led to a T1/2 that was faster at the PM and GJP (GJP: 0.9s; PM: 7.3s) than the WT ( Figure 4B,D). We conclude that the interaction with the tubulin cytoskeleton aided in the temporal aspects of Cx36 incorporation from the PM into the GJP in addition to quantity of delivery, which was verified by GJP size and formation.  . Cx36 connexons are initially delivered to the plasma membrane via the tubulin cytoskeleton and diffuse laterally to the GJP. Fluorescence Recovery After Photobleaching (FRAP) microscopy was used to bleach the non-junctional plasma membrane (PM) and gap-junction plaque (GJP) regions of interest (ROI). The average fluorescence intensity of each ROI during the recovery phase was normalized to the pre-bleached intensity values and plotted to quantify the relative recovery of fluorescence. In cells expressing Cx36 wild-type (A) and ∆279-292 (B), the plasma membrane (PM) recovered more than at the GJP. Recovery curves are the mean ± SEM. Sample sizes (n) of WT Cx36 were n = 28 for the GJP and n = 12 for the PM. For Cx36∆279-292, the sample sizes were n = 22 for the GJP and n = 10 for the PM. **** p < 0.0001, Kruskal-Wallis test. (C) An overlap of the GJP curves for the WT (in Figure 4A) and ∆279-292 (in Figure 4B). (D) Plot of the mobile fraction (M f ) against the half-time of recovery (T 1/2 ) extrapolated from the Cx36 WT (in Figure 4A) and ∆279-292 (in Figure B) recovery curves. The mobility, as indicated by M f , of both WT and ∆279-292 GJPs were lower than its respective PM ROI, suggesting that the GJP was more stable. Mutant GJPs were significantly more mobile than the WT control. Additionally, T 1/2 for ∆279-292 was faster at the GJP and PM in comparison to the wild-type, indicating diffusion was favored over a controlled trafficking mechanism. Data are the mean ± CI.
In-depth comparisons of FRAP at the GJP revealed that the mutant GJPs were significantly more mobile (p < 0.01 × 10 −2 ). This loss of stability was also accompanied by a large reduction in the T 1/2 , reflecting dynamics that were consistent with diffusion outcomes (Figure 4C,D). Overall, the results advocate for a mechanism in which tubulin interactions aid in the temporal aspects of GJP coalescence, providing a means of regulating the "when" and "how much" of connexon delivery.
confirmed that the interaction between Cx36 and the tubulin cytoskeleton was not interdependent on the interactions or stability of the actin cytoskeleton.
At the GJP, cytoskeletal interference resulted in a T 1/2 that was uniformly faster across all treatment groups (WT: 11.5 s, n = 28, Colchicine: 1.3 s, n = 26; Paclitaxel: 0.6 s, n = 24; Cytochalasin D: 0.8 s, n = 22). Here, the incorporation of Cx36 into the GJP after actin-or tubulin-cytoskeletal disruption was consistent with diffusion dynamics (Figure 7). Similar to results shown earlier (Figure 4), disruption of the tubulin cytoskeleton with Colchicine increased M f at the GJP (WT: 25.6 ± 0.6%, Colchicine: 46.4 ± 0.4%, p < 0.01 × 10 −2 ). No significant difference in recovery was found with Paclitaxel treatment (34.7 ± 0.3%, p > 0.99), likely attributable to the direct stabilization of the GJP via tubulin stabilization. Although the GJP remained stable under paclitaxel treatment, it is important to highlight that transport into the GJP was impaired as shown previously (Figures 5 and 6). In this regard, tubulin acts separately, first as a conduit for Cx36 transport and secondly as a GJP stabilizer in the Neuro-2a cell model. Regardless, Cx36 mobility was not completely abolished following tubulin disruption, inferring that the actin cytoskeleton may provide a secondary transport modality. In support, Cytochalasin D treatment significantly increased GJP mobility (55.4 ± 0.4%, p < 0.01 × 10 −2 ) (Figure 7), confirming that the actin cytoskeleton contributes to Cx36 delivery and stabilization at the GJP. 4), disruption of the tubulin cytoskeleton with Colchicine increased Mf at the GJP (WT: 25.6 ± 0.6%, Colchicine: 46.4 ± 0.4%, p < 0.01 × 10 −2 ). No significant difference in recovery was found with Paclitaxel treatment (34.7 ± 0.3%, p > 0.99), likely attributable to the direct stabilization of the GJP via tubulin stabilization. Although the GJP remained stable under paclitaxel treatment, it is important to highlight that transport into the GJP was impaired as shown previously (Figures 5,6). In this regard, tubulin acts separately, first as a conduit for Cx36 transport and secondly as a GJP stabilizer in the Neuro-2a cell model. Regardless, Cx36 mobility was not completely abolished following tubulin disruption, inferring that the actin cytoskeleton may provide a secondary transport modality. In support, Cytochalasin D treatment significantly increased GJP mobility (55.4 ± 0.4%, p < 0.01 × 10 −2 ) (Figure 7), confirming that the actin cytoskeleton contributes to Cx36 delivery and stabilization at the GJP. treatment demonstrated a significant reduction. While the T1/2 remained lower than the untreated wild-type, inhibition of the actin cytoskeleton with Cytochalasin D (CytoD) resulted in an increase in the mobile fraction, indicating that GJPs became more fluid and suggesting that actin may play a role in fine-tuning Cx36 incorporation at the GJP. Recovery curves are the mean ± SEM and mean ± CI for the XY plots of mobile fraction against half-time of recovery. Sample sizes (n) were n = 28 for the WT, n = 26 for Colchicine, n = 24 for Paclitaxel, and n = 22 for Cytochalasin D treated GJPs. **** p < 0.0001, Kruskal-Wallis test; all samples were compared to the wild-type control.

Characterization of the Tubulin-Binding Motif
A previous report demonstrated that a 26 amino-acid peptide of Cx43 mediates binding to αand β-tubulin isoforms. This region adopts a helical conformation upon binding to tubulin and is regulated by phosphorylation [38]. Here, we explored whether Cx36 might have a similar binding mechanism using structural modeling of the Cx36 carboxyl-terminal domain. The calmodulin-

Discussion
This research uses a neuroblastoma model of synaptic plasticity to explore the interactions between Cx36 and tubulin. By sequence similarity to Cx43 [20], a tubulin-binding motif was identified in the carboxy-terminal tail of Cx36. In experimental support, the previously reported binding of tubulin with the carboxyl-terminal tail of Cx36 [19] was confirmed independently here using a BioID assay. The Cx36-binding motif is highly conserved and unique among connexin isoforms as it shares its interaction surface amongst tubulin-, CaM-, and CaMKII-binding partners. Typically, CaM-binding motifs in other connexins are located in different positions in the amino-terminus (Cx32), cytoplasmic loop (Cx43, Cx44, and Cx50), or nonoverlapping region within the CT domain (Cx32) [12,[41][42][43][44][45].
Binding of CaMKII to connexins other than Cx36 has not been reported.
The Cx36-tubulin-binding motif differs from other microtubule-binding motifs found in proline-rich regions of several microtubule-associated proteins (MAPs) or the motor proteins kinesin and dynein. For example, the microtubule-binding motif in both motor proteins has a P-X 6 -E-X 4 -L core consensus sequence, surrounded by several conserved polar, hydrophobic, and charged amino acids [46,47]. Similarly, the tubulin-binding motif of Cx36 is enriched in hydrophobic (isoleucine, leucine, arginine, and valine) and electrically charged basic (arginine and lysine) amino acids. However, the absence of a proline-rich region appears to be a significant and unique property of the Cx36-tubulin motif.
Three-dimensional structural modeling of the Cx36-tubulin-binding site to the engineered stathmin-like domain (SLD) predicted similarities between Cx36 and SLD. The Cx36 carboxyl terminus likely adopts a helical structure upon binding with tubulin similar to previous reports on Cx43-tubulin interactions [38]. The previously resolved NMR structure of the Cx36-CaM-binding complex suggests that a helical structure is formed as early as interaction with calcium-activated CaM occurs in the ER-Golgi complex. Siu et al. have also shown that CaM binding strongly relies on access to W277 and intracellular calcium [22]. The new experimental evidence supports roles of individual amino acids downstream of W277, with I280 and the adjacent amino acids K279 and K281 critical in the Cx36-tubulin-binding mechanism as demonstrated by co-localization studies and consistent with the structural model. Noteworthily, phosphorylation of S293 [48][49][50] downstream of the overlapping tubulin/CaM/CaMKII-binding site decreases GJ communication, suggesting that competitive and cooperative mechanisms exist side-by-side in a protein domain encompassing less than 20 amino acids [51]. Consistent with other studies [52][53][54][55][56][57], we anticipate that salt bridges formed by the lysine residues contribute to protein stability and specificity, driving molecular recognition and catalysis between protein interfaces. We also found that site-directed alanine mutagenesis produced variability in protein expression levels that were not previously observed with the expression of Cx36∆279-292. Since large amino-acid deletions in the loop, helix, and terminal positions are generally more tolerated than substitutions [58,59], we concluded that the Cx36 tertiary structure and functionality were likely compromised, leading to expression variability. Again, we believe this advocated for the significance of positional non-covalent interactions to sustain both the protein itself and protein-protein interfaces.
The Cx36∆279-292 mutant exhibited less co-localization to the ER-Golgi complex but more co-localization with caveolin-1 vesicles. Although the ER stress response was not explored in this study, equal protein expression of wild-type and mutant proteins as well as lack of signs of protein degradation suggest that involvement of the endoplasmic reticulum-associated degradation (ERAD) pathway is minor. We propose a mechanism in which the mutant protein is prematurely released from the ER-Golgi complex prior to its packaging into transport vesicles. Since we have demonstrated that Cx36∆279-292 is still transport competent, a different mechanism of transport may be favored. Interaction with actin is a possibility. The increase in the association between Cx36∆279-292 and caveolin-1-positive vesicles could be an indication of alternative transport pathways.
The pharmacological manipulation of the microtubule network and genetic manipulation of the tubulin-binding motif confirmed that the trafficking of Cx36-carrying vesicles and the formation of GJPs were critically dependent on an intact and functionally dynamic cytoskeleton. The processes observed appear to follow the generally accepted chain of events in which connexons packaged in vesicles emerge from the Golgi reach the cell membrane via microtubules at multiple nonrandom insertion sites [28,33,34]. After that, hemichannels form clusters of gap junctions known as plaques at cell-cell borders. In addition to the modest similarities in the binding motif, the aforementioned transport mechanism parallels the reports for Cx43 [28,33].
Lauf et al. previously described the directed transport of Cx43 to the plasma membrane, demonstrating mobility of approximately 70%, before incorporation into GJPs. At GJPs, Cx43 has been described as exhibiting either low or high mobility states at the lateral ends. Categorization of the mobility state was found to be independent of GJP size but, rather, to be determined by the C-terminal domain influencing channel density, protein interaction candidates, or both [60]. Our studies have confirmed unique differences between Cx43 and Cx36 kinetics both at the plasma membrane and GJP. In this research, we report mobility of wild-type Cx36 as approximately 70% at the plasma membrane, comparable to Lauf et al., and as 25% at the GJP. We suspect that mobility in the PM contributes to the fine-tuning of Cx36 incorporation, especially under activity-dependent circumstances. In a related study [50], FRAP microscopy of GJPs formed by Cx36-HaloTag fusion proteins was studied in HeLa cells. Here, the initial M f was reported to be 56% and subsequently decreased to 41% after sequential bleaching. Wang et al. also found half-time of recovery to be 1.55 ± 0.22 s [50]. In this study, the carboxyl terminus of Cx36 was fused with the GFP protein tag, increasing the molecular mass by 27 kDa. We anticipate that the lower mobile fraction and longer half-time of recovery differences observed between our wild-type protein and the Cx36-HaloTag can be attributed to the large C-terminal GFP tag, decreasing the mobility potential of Cx36. Nevertheless, we consistently reported considerably shorter half-times of recoveries upon manipulation of the tubulin cytoskeleton or interaction thereof. Since a faster T 1/2 is typically reflective of weaker binding mechanisms, incorporation of Cx36 at the GJP under these circumstances was likely mediated by diffusion rather than a controlled transport mechanism.
Theoretical and practical considerations suggest that other transport mechanisms must exist alongside tubulin-dependent transport. In neurons, microtubules participate in axonal vesicular transport and tubulin entry into dendritic spines is activity-dependent or associated with development [61]. Lateral diffusion in the membrane as mechanisms for transport of Cx36 into dendritic spines cannot explain temporal aspects of the plasticity of electrical synapses. Here, experimental evidence showed that treatment conditions targeting tubulin or the tubulin-binding motif reduced the GJP area and formation but that the formation of GJPs at the juxtamembrane was not wholly abolished. We propose a second, actin-dependent transport and GJP formation process as suggested by Lynn et al. [62], similar to actin-dependent transport of receptor and channel proteins into synaptic spines [63][64][65][66]. Binding of Cx36 to actin directly or indirectly through adaptor proteins has not been addressed in this study. Although previous reports indicated that interaction with the actin cytoskeleton was involved in the turnover and invagination endocytosis of gap junctions [67], a more recent study on Cx30 proposed that actin is involved in the anchoring of connexons and short distance transport rather than in the facilitation of GJ internalization [27]. Follow-up studies resolving details of this second transport mechanism will help to understand how molecular and functional asymmetry is created at a vertebrate electrical synapse.
Wayakanon et al. investigated the consequences of GJ transport and internalization of various CT truncated Cx43 mutants. Deletion of amino acids 235-242, which corresponds to the tubulin-binding motif, resulted in the absence of GJPs [68]. In contrast, our results demonstrated the retention of GJP formation and communication followed deletion of the tubulin-binding motif in Cx36, albeit at reduced efficiencies. Similarly, deletion of tubulin-binding motif in the N2RB subunit of altered NMDA receptors reduced but not abolished transport to the synapse [69]. Further, cell type-based differences will affect transport and internalization of both electrical and chemical modalities. We propose that neurons may have developed specialized transport mechanisms supporting on-demand plasticity. Our results point to actin as a secondary conduit for Cx36 transport, similar to NMDA interactions with actin filaments [70].
Further investigations would be required to determine whether actin-and tubulin-binding efficiencies are the basis for a tug of war between two competing or cooperating pathways.
What is the physiological relevance of Cx36 interaction with microtubules? Cx36 has one of the lowest voltage sensitivities (half-inactivation voltage ±75 mV) and single-channel conductance (10-15 pS) among other connexin isoforms [71]. The detection of Cx36 GJP in the nervous system had been elusive for a long time due to the small size [72,73]. Arguably, the small GJP size and Cx36-specific gating properties suggest that a precise electrical coupling can be achieved through the tight regulation of the number of channels present in the gap-junction plaque [71]. This mechanism was attributed to tubulin-dependent trafficking and was complemented by the phosphorylation via CaM/CaMKII [4] and protein kinase A (PKA) [50,51,[74][75][76]. We acknowledge that such a refined transport mechanism could serve both to enhance neuronal plasticity and/or to reduce excitotoxicity and subsequent neuronal cell death, both of which are influenced by Cx36 GJ coupling. Our results point to the significance of the Cx36-tubulin interaction; tubulin is a fundamental component in the modulation of neuronal synapses via channel abundance and stability.
Our investigation provides insight on a critical step towards molecular and functional asymmetry at a vertebrate electrical synapse. Tubulin-dependent transport connects two major bookends of the life cycle of Cx36. Before interaction with tubulin, direct interaction with CaM occurs primarily at the ER/Golgi complex [22]. Vesicular transport to the GJP involves tubulin and most likely another transport system accounting for the structural and functional specialization of pre-and postsynaptic compartments of soma-somatic, dendro-dendritic, dendro-somatic, and axon-dendritic contact sites in mature neurons where Cx36 is expressed [23,24]. Outcomes of this research extend to previously recognized similarities of both NMDA receptors and Cx36 interaction with CaM/CaMKII. Similar to Cx36, NMDA receptor channel subunit CTs bind tubulin dimers or soluble forms of tubulin. The efficient modulation of microtubule dynamics by the NR1 and NR2 cytoplasmic domains suggests an interaction of the receptor and the subsynaptic cytoskeletal network that may play a role during morphological and functional adaptations, as observed during synaptogenesis and in adult CNS plasticity. Importantly, results have shown that tubulin-dependent trafficking of Cx36 is part of the molecular machinery potentiating electrical synaptic strength. In this regard, the activity-dependent modulation of the cytoskeleton controls the formation and plasticity of electrical synapses.