Neurobeachin controls the asymmetric subcellular distribution of electrical synapse proteins

Summary

The subcellular positioning of synapses and their specialized molecular compositions form the fundamental basis of neural circuits. Like chemical synapses, electrical synapses are constructed from an assortment of adhesion, scaffolding, and regulatory molecules, yet little is known about how these molecules localize to specific neuronal compartments. Here we investigated the relationship between the autism- and epilepsy-associated gene Neurobeachin, the neuronal gap junction channel-forming Connexins, and the electrical synapse scaffold ZO1. Using the zebrafish Mauthner circuit we found Neurobeachin localizes to the electrical synapse independently of ZO1 and Connexins. By contrast, we show Neurobeachin is required postsynaptically for the robust localization of ZO1 and Connexins. We demonstrate Neurobeachin binds ZO1 but not Connexins. Finally, we find Neurobeachin is required to restrict electrical postsynaptic proteins to dendrites, but not electrical presynaptic proteins to axons. Together, the results reveal an expanded understanding of electrical synapse molecular complexity and the hierarchical interactions required to build neuronal gap junctions. Further, these findings provide novel insight into the mechanisms by which neurons compartmentalize the localization of electrical synapse proteins and provide a cell biological mechanism for the subcellular specificity of electrical synapse formation and function.

Graphical Abstract

eTOC_Blurb

Martin et al. show that the autism- and epilepsy-linked Neurobeachin functions in electrical synapse formation to compartmentalize postsynaptic proteins to neuronal dendrites. The findings reveal a mechanism by which neurons selectively assembly electrical synapses with subcellular specificity.

Introduction

Neuronal synapses can be separated into two main classes, chemical and electrical, each of which is required for neural circuit development and function^1,2. Chemical synapses are well-studied structures with molecularly diverse asymmetrical junctions, which are known to rely upon highly specialized proteins that orchestrate their localization, assembly, and function^3,4 By contrast, electrical synaptic transmission is achieved via neuronal gap junction channels, which in vertebrates are created between neurons by apposed hemichannels of Connexin hexamers that allow for the bidirectional flow of ions and other small molecules⁵. In addition to channels, electrical synapses contain a diverse proteome including cell adhesion, scaffold, and regulatory molecules, yet the mechanisms required to construct neuronal gap junctions remain poorly understood^6-8. Electrical synapses form between all neuronal compartments with functional effects on circuit activity dependent upon each unique coupling arrangement. For example, dendro-dendritic electrical synapses allow for lateral excitation via the spread of local synaptic potentials between neighboring neurons⁹. Alternatively, axo-axonic electrical synapses can promote strong and highly synchronized firing, and such coupling may contribute to fast ripples such as those aberrantly present in patients with epilepsy^1,10-13. Thus, at a cell biological level, the mechanisms specifying electrical synapse subcellular specifity are critical for appropriate neural circuit formation and function.

While neurons control the subcellular locations of neuronal gap junctions, the mechanisms by which neurons compartmentalize electrical synapse proteins are not well understood. Recent work in Caenorhabditis elegans, which use Innexins to form gap junction channels, identified a cAMP-dependent signaling pathway regulating the trafficking of Innexins to distal synaptic sites¹⁴. The findings support the notion of regulatory mechanisms controlling the trafficking of electrical synapse proteins to distinct subcellular locations. Yet because Innexins are evolutionarily distinct from Connexins¹⁵, it is not clear if vertebrates share related pathways. For chemical synapses, the mechanisms controlling the polarized distribution of synaptic proteins to axons and dendrites are well studied and require the coordination of motor proteins, the cytoskeleton, and numerous adaptors and scaffolds to ensure the fidelity of compartmentalized delivery^16-19. Whether there are distinct, or shared, mechanisms to guide the polarized transport of electrical synapse proteins remains unknown.

The formation of electrical and chemical synapses are generally thought to be cell biologically and biochemically distinct processes. Yet in recent years molecules have emerged that suggest links between the development and function of these structures^1,2,8,20. One of these proteins, Neurobeachin, was found to be required for both electrical and chemical synapse development^20-22. Neurobeachin is a large (~330 kDa) protein that is highly conserved in vertebrates and contains various protein-binding domains including a BEACH domain, AKAP domain, and multiple WD40 domains²¹. More than 20 de novo variants of Neurobeachin have been identified in patients with autism, intellectual disability, and epilepsy^23-26. Previous work suggests Neurobeachin regulates the localization of chemical synapse proteins including glutamate receptors and the scaffolds SAP102 and PSD95^21,27-30. Yet its mechanistic role in electrical synapse development is unknown.

Here we explore the contributions of Neurobeachin to electrical synapse formation and examine its biochemical and cell biological functions. We used the stereotyped and identifiable synaptic contacts of the Mauthner cell of larval zebrafish as they are exquisitely accessible to in vivo cell biological analysis. We show that Neurobeachin is localized to the electrical synapse and functions postsynaptically to promote the robust localization of the electrical synapse scaffold Zonula Occludens 1 (ZO1) and neuronal Connexins. We find that Neurobeachin interacts with ZO1 but not the Connexin proteins, supporting a hierarchical model of electrical synapse formation in which Neurobeachin recruits ZO1 which in turn recruits Connexin proteins. Finally, we find that Neurobeachin functions to restrict postsynaptic electrical synapse proteins to dendritic synapses, such that in its absence postsynaptic proteins are inappropriately localized and stabilized at presynaptic locations in the axon. The results presented provide new insight into the molecular complexity and hierarchical interactions that build electrical synapses. Moreoever, the results reveal a mechanism for the asymmetric synaptic localization of electrical synapse components providing a basis for the subcellular specialization of neuronal gap junctions.

Results

Neurobeachin is required for the localization of electrical synapse proteins

To investigate how Neurobeachin regulates electrical synapse development we used the zebrafish Mauthner cell circuit, which uses a combination of electrical and chemical synapses to control a fast escape response^31-33. This circuit is composed of two Mauthner cells per fish that receive sensory input from several modalities, including the auditory system, and send output onto motorneurons and interneurons in the spinal cord, including Commissural Local (CoLo) interneurons (Figure 1A,B). Electrical synapses are prominent throughout the circuit, including at mixed electrical/glutamatergic inputs from auditory afferents onto the lateral Mauthner dendrite, called Club Ending (CE) synapses (Figure 1B)³⁴. Additionally, the Mauthner cell uses electrical synapses between its axon and CoLos, called M/CoLo synapses (Figure 1B). These M/CoLo synapses repeat down the length of the spinal cord, with CoLo being present as a single cell in each hemisegement of the ~30 segments of the trunk. Mauthner and CoLo cell morphology can be observed using the transgenic line Et(Tol-056:GFP)³⁵. The CE and M/CoLo electrical synapses are molecularly asymmetric composed of presynaptic Connexin 35.5 hemichannels (Cx35.5 – encoded by the gjd2a gene) that pair with postsynaptic Cx34.1 hemichannels (encoded by the gjd1a gene)³⁶. In addition, the electrical synapse scaffold ZO1b (encoded by the tjp1b gene) localizes exclusively postsynaptically where it binds Cx34.1 (Figure 1B)^37-38. At the CE synapses, the Mauthner cell is the postsynaptic partner and uses ZO1b and Cx34.1 to build electrical synapses, while at the M/CoLo synapses, the Mauthner cell is presynaptic and uses Cx35.5 (Figure 1B). The asymmetry of Mauthner cell electrical synapses highlights the cell biological requirement of the neuron to deliver unique electrical synaptic proteins to distinct neuronal compartments.

Figure 1: Neurobeachin is required for the localization of electrical synapse proteins.

(A) Dorsal view of the larval zebrafish illustrating location of the two Mauthner cells (green), anterior left. Boxed region indicates the region shown in B. (B) Simplified illustration of the Mauthner circuit, anterior up, highlighting the Club Ending (CE) and M/CoLo synapses and the known protein localization of electrical synapse components Cx35.5 (orange), Cx34.1 (cyan), and ZO1b (magenta). The center dashed-circle diagram depicts the standard synapse organization for each Mauthner synapse which includes gap junction hemichannels composed of Cx35.5 in the presynapse, and Cx34.1 in the postsynapse where it binds scaffolding protein ZO1b. Arrows in Mauthner soma indicate the compartmentalization of the indicated electrical synapse proteins. (C-H) Confocal images of Mauthner electrical synapses in 5 days post fertilization Et(Tol-056:GFP) transgenic wildtype (WT) or nbeaa−/− mutants. Panels show staining for GFP labeling the Mauthner cell (green), Cx35.5 (orange), Cx34.1 (cyan), V5-ZO1b (magenta) at (C-E) CE synapses and (F-H) M/CoLo synapses. Dashed circles identify locations of Mauthner and CoLo contact. Images in this figure and other figures are maximum intensity projections of ~12μm at CE and ~20μm at M/CoLo synapses. (I,J) Quantification of indicated electrical synapse protein fluorescence intensities, normalized to WT, at (I) CE synapses and (J) M/CoLo synapses in WT and nbeaa^−/− mutant animals. Mean ± SEM are shown, all comparisons were made between genotyped WT and mutant siblings. For I, n = 5, 5, 5, 5, 10, and 12 fish for each bar shown, left to right. Two Mauthner cells were analyzed per fish and averaged to represent a single value for each fish. For graph in J, n = 6, 6, 6, 6, 10, and 10 fish, for each bar shown, left to right. 20-24 M/Colo synapses were analyzed and averaged for each fish. ** indicates p < 0.01, *** indicates p < 0.001, **** indicates p < 0.0001 by unpaired t-test. Scale bars are as indicated. See also Figure S1.

We first examined the role of Neurobeachin (encoded by the nbeaa gene) in localizing the presynaptic Cx35.5 and the postsynaptic Cx34.1 and ZO1b to Mauthner cell electrical synapses (Figure 1C-J). We previously found that Neurobeachin is required for Connexin localization to Mauthner cell electrical synapses²⁰. However, the molecular identity of the Connexins (Cx35.5/gjd2a and Cx34.1/gjd1a), the ZO1b/tjp1b scaffold, and the molecularly asymmetric distribution of these proteins at the synapse, were all unknown at the time. Therefore, we examined how the loss of Neurobeachin impacted the localization of all three of these electrical synapse components by comparing 5 days post fertilization (dpf) wildtype and nbeaa^fh364/fh364 mutants (hereafter called nbeaa−/−²⁰). First, we tested Cx35.5 and Cx34.1 localization at CE (Figure 1C,D,I) and M/CoLo synapses (Figure 1F,G,J). In nbeaa−/− mutants, we observed similar decreases of both pre- and postsynaptic Connexin proteins at electrical synapses. Next, using a transgenic line that marks endogenous ZO1b with a V5 epitope (tjp1b^b1406, hereafter called Pt(V5-ZO1b)³⁸), we examined 5 dpf wildtype and nbeaa−/− mutant fish for ZO1b at CE (Figure 1E,I) and M/CoLo synapses (Figure 1H,J). We observe a similar reduction of ZO1b at Mauthner dendritic and axonal electrical synapses in nbeaa−/− mutant fish. While nbeaa−/− mutant Mauthner cell dendrites have reduced complexity²⁰, we observed no gross morphological defects in the primary lateral or ventral dendrites that emerge from the soma, nor in the axons that course down the spinal cord (Figure S1A), suggesting that overall neuronal development and compartmentalization remains intact. Further, we observe that CE and M/CoLo synapses are still present in their stereotypical locations (Figure 1C-H), yet the amount of Connexins and ZO1 is significantly reduced (Figure 1I,J), suggesting that appropriate interneuronal contacts are made in mutants, yet there is a defect in robustly localizing synaptic proteins. The diminished synaptic localization is not due to decreased overall Connexin and ZO1 protein abundance, as similar protein levels are present in wildtype and nbeaa−/− mutants (Figure S1B-D). Together these results support the notion that nbeaa−/− mutants have defects in the localization of electrical synapse proteins. We conclude that loss of Neurobeachin function affects localization of pre- and postsynaptic Connexins and the postsynaptic scaffold ZO1.

Neurobeachin localizes to electrical synapses independent of ZO1 and Connexins

In rat cerebellar neurons, Neurobeachin localizes to the trans side of the Golgi, on tubulovesicular vesicles in dendrites, and at the postsynapse of glutamatergic chemical synapses²¹. Thus, we next examined Neurobeachin localization using the Mauthner cell circuit. At 5 dpf, we observe Neurobeachin localization in the Mauthner cell soma, yet within this compartment it showed little colocalization with electrical synapse proteins (Figure S2A,B). To examine Neurobeachin’s somatic distribution, we stained for GM130/Golga2, a cis-Golgi apparatus marker, and observed the Mauthner cell of wildtype and nbeaa−/− mutants. Golgi staining within the Mauthner soma was found throughout the compartment but asymmetrically biased towards the axon; this distribution was not altered in nbeaa−/− mutants (Figure S2C,D). In addition, we noticed a high degree of close proximity localization between Neurobeachin and GM130 as expected for colocalized proteins. This pattern is similar to what has been observed in rodents^21,29.

We next examined Neurobeachin localization at Mauthner electrical synapses. At the CE synapses on Mauthner’s lateral dendrite, we see Neurobeachin localization at distinct foci adjacent to the electrical synapse proteins ZO1b and Cx34.1, as well as a diffuse staining colocalized across the extent of the observed gap junction staining (Figure 2A). This pattern of discrete localization surrounding and at times overlapping the Connexin/ZO1 staining is reminiscent of what has been described for Neurobeachin with chemical postsynaptic scaffolds in mammalian cultured neurons²⁹. We also observe Neurobeachin staining at M/CoLo synapses, with staining appearing adjacent to ZO1 and Connexins, and we never observe staining within the prominent Mauthner axon between each synapse (Figure 2B). We note that at each Mauthner associated synapse (CE or M/CoLo) the pre- or postsynaptic localization of Neurobeachin cannot be resolved due to the limits of light microscopy — we address where the protein functions below. Neurobeachin staining is eliminated in nbeaa−/− mutants supporting the specificity of the antibody (Figure 2C-E, Figure S2B). We conclude that Neurobeachin can localize to electrical synapses, yet does so in a pattern that is distinct from the ZO1 scaffold and neural Connexins.

Figure 2: Neurobeachin localization to electrical synapses is independent of ZO1b and Connexins.

(A-D) Confocal images of Mauthner electrical synapses in 5 days post fertilization Et(Tol-056:GFP) transgenic wildtype (WT, A,B) and nbeaa−/− (C,D) zebrafish. Panels show staining for GFP labeling the Mauthner cell (green), Cx34.1 (cyan), ZO1 (magenta), and Neurobeachin (yellow) at (A,C) CE synapses, and (B,D) M/CoLo synapses. Dashed circles identify locations of Mauthner and CoLo contact. B^I, B^II and D^I, D^II represent two independent M/CoLo synapses for each genotype. (E) Quantification of Neurobeachin fluorescence intensities at CE synapses comparing wildtype (WT, n=5) and nbeaa−/− mutants (n=6). ** indicates p < 0.01 by Mann Whitney. (F-G) Neurobeachin localization at Mauthner CE synapses in 5 dpf Et(Tol-056:GFP) transgenic zebrafish from (F) wildtype (WT, top)_and ZO1b−/− mutant animals (bottom), and (G) wildtype (WT, top) and Cx34.1−/− mutant animals (bottom) with markers and scale bars as indicated. Quantification of Neurobeachin fluorescence intensities at CE synapses in (F) compares wildtype (WT, n=5) and ZO1b−/− mutants (n=5). ns indicicates not significant by t-test. Quantification of Neurobeachin fluorescence intensities at CE synapses in (G) compares wildtype (WT, n=5) and Cx34.1−/− mutants (n=5). ns indicicates not significant by t-test.For all quantification, two Mauthner cells were analyzed per fish and averaged to represent a single value for each fish. Mean ± SEM are shown and values are normalized to WT sibling controls for each experiment. See also Figure S2.

Next we examined whether Neurobeachin localization to the electrical synapse was dependent upon the function of other electrical synapse proteins. The intracellular scaffold ZO1b is required for Connexin localization, but ZO1b largely localizes to electrical synapses independently of the Connexins³⁸. We therefore examined Neurobeachin localization in ZO1b and Connexin mutants. We examined ZO1b/tjp1b^{b1370/b1370 37} and Cx34.1/gjd1ab^{fh436/fh436 36} mutants and found that Neurobeachin localizes to electrical synapses in a manner similar to their wildtype siblings (Figure 2F-G). We conclude that Neurobeachin localizes to the electrical synapse independent of ZO1b and Connexins.

Neurobeachin binds the electrical synapse scaffold ZO1b but not the Connexins

Neurobeachin is thought to orchestrate the delivery of neurotransmitter receptors to the glutamatergic chemical synapse through binding with chemical synapse scaffolds such as membrane associated guanylate kinase (MAGUK) family members PSD95 and SAP102^30,39. Given that the electrical synapse scaffold ZO1b is also a member of the MAGUK family, we next examined Neurobeachin/ZO1b interactions. We attempted to clone full-length nbeaa sequences from fish, but repeated and varied attempts prevented the recovery of the entire coding sequence. We were able to generate N- and C-terminal fragments of Neurobeachin, each tagged with mVenus (Figure 3A), and used heterologous expression in HEK293T cells to test for interactions with previously cloned zebrafish ZO1b and neural Connexins³⁸. We first co-transfected ZO1b with mVenus-tagged Neurobeachin fragments and performed coimmunoprecipitation (coIP) assays. We found that the C-terminal fragment of Neurobeachin pulls down ZO1 (Figure 3B-C). We next performed the reciprocal coIP experiment and co-transfected Flag-tagged-ZO1b with the mVenus-tagged Neurobeachin fragments and found that Flag-tagged ZO1b robustly pulls down the C-terminal fragment of Neurobeachin (Figure 3D,E). We also observe a faint, yet consistent, band in both IP experiments for ZO1b and N-terminal Neurobeachin fragment interactions (Figure 3C-E). Finally, we tested whether Neurobeachin could bind the pre- or postsynaptic Connexins. We co-transfected each Connexin with N-terminal and C-terminal Neurobeachin and find no evidence of pull down with Neurobeachin in immunoprecipitates (Figure 3F-I). By contrast, we find robust interaction between ZO1b and both Connexins, as previously shown³⁸. We conclude that the C-terminal fragment of Neurobeachin can bind the postsynaptic electrical synapse scaffold ZO1b in vitro.

Figure 3: Neurobeachin binds the postsynaptic electrical synapse scaffold ZO1b, but not the neural Connexins, in vitro.

(A) Schematic of Neurobeachin protein indicating domains included in mVenus-tagged N- and C-terminal fragments. (B) HEK293T/17 cells were transfected with plasmids to express ZO1b and either mVenus (lane 1), mVenus-tagged N-terminal-Neurobeachin (lane 2), or mVenus-tagged C-terminal-Neurobeachin (lane 3). Lysates were immunoprecipitated with anti-GFP antibody, which recognizes mVenus, and analyzed by immunoblot for the presence of ZO1b using ZO1 antibody (upper), or mVenus-containing proteins using GFP antibody (middle). Total extracts (bottom) were blotted for ZO1 to demonstrate expression and antibody recognition of expressed proteins. (C) Quantification of three representative blots in B, normalized to the strongest band. (D) HEK293T/17 cells were transfected with plasmids to express Flag-ZO1b and either mVenus (lane 1), mVenus-tagged N-terminal-Neurobeachin (lane 2), or mVenus-tagged C-terminal-Neurobeachin (lane 3). Lysates were immunoprecipitated with anti-Flag antibody and analyzed by immunoblot for the presence of mVenus-containing proteins using GFP antibody (upper), or ZO1b using ZO1 antibody (middle). Total extracts (bottom) were blotted for GFP to demonstrate expression and antibody recognition of expressed proteins. (E) Quantification of three representative blots in D, normalized to the strongest band. (F-I) HEK293T/17 cells were transfected with plasmids to express Cx35.5 (F) or Cx34.1 (H) and either mVenus-tagged N-terminal Neurobeachin (lane 1), mVenus-tagged C-terminal-Neurobeachin (lane 2), or mVenus-tagged ZO1b (lane 3). Lysates were immunoprecipitated with anti-GFP antibody and analyzed by immunoblot (F) for the presence of Cx35.5 using a Cx35.5 specific antibody or (H) Cx34.1 using a Cx34.1 specific antibody (upper), mVenus-containing proteins using GFP antibody (middle). Total extracts (bottom) were blotted for (F) Cx35.5 or (H) Cx34.1 to demonstrate expression and antibody recognition of expressed proteins. (G) Quantification of three representative blots in F, normalized to the strongest band. (I) Quantification of three representative blots in H, normalized to the strongest band. Mean ± SEM are shown and n = 3 for all blots.

Neurobeachin is required postsynaptically for electrical synapse protein localization

Observing that Neurobeachin binds ZO1b, and given that ZO1b is a postsynaptic scaffold³⁸, we next wanted to examine whether Neurobeachin functioned postsynaptically in vivo. To achieve this, we generated chimeric animals via blastula transplantation⁴⁰ to create larvae with Mauthner circuit neurons derived from distinct genotypes (Figure 4A). We first examined the localization of ZO1b within the Mauthner cell by using donor embryos that were Et(Tol-056:GFP); Pt(V5-ZO1b) double transgenics, with the former transgene labeling Mauthner and CoLo cells with GFP, and the latter tagging endogenous ZO1b protein with a V5 epitope tag. The host embryos were non-transgenic wildtypes while the Et(Tol-056:GFP); Pt(V5-ZO1b) donor embryos were either homozygous wildype at the nbeaa locus (nbeaa+/+) or homozygous mutant (nbeaa−/−). This allows us to examine the effects of the loss of Neurobeachin from the Mauthner cell. We focused on chimeric animals in which individual Mauthner cells were derived from the donor (GFP+) and found V5-ZO1b present at CE synapses in wildtype control transplants (nbeaa+/+ to WT), while in nbeaa−/− mutant Mauthner cells in a wildtype host (nbeaa−/− to WT), V5-ZO1b staining was reduced (Figure 4B).

Figure 4: Neurobeachin is required postsynaptically for the proper dendritic compartmentalization of ZO1b and Cx34.1.

(A) Diagram illustrating 5 days post fertilization chimera in which a Mauthner cell is derived from the GFP+ transgenic donor (green) while the other neurons of the circuit are derived from the GFP-host (gray). Boxed outline 1: Location of CE synapses shown by representative confocal images in B and C. Boxed outline 2: Location of M/CoLo synapses shown by representative confocal images in D and E. (B-E) Confocal images of GFP+ Mauthner cells in chimeric animals. (B,D) Chimeras generated by transplanting cells from Et(Tol-056:GFP); Pt(V5-ZO1b) double transgenic donors into a wildtype (WT) host. (C,E) Chimeras were generated by transplanting cells from Tg(Cx34.1-GFP^AE37) transgenic donors into a WT host. Left images show transplants from a nbeaa+/+ animal into a WT host while right images show transplants from a nbeaa−/− mutant into a WT host. Mauthner CE synapses (B,C) and M/CoLo synapses (D,E) are identified at stereotypical positions in the hindbrain and spinal cord of the GFP+ Mauthner cell, respectively. Panels show staining for GFP labeling the Mauthner cell (green in B,D) or Cx34.1-oxGFP (cyan in C,E), V5-ZO1b (magenta), and Cx35.5 (orange). Outline of Mauthner cell lateral dendrite shown by green dashed line in (B). M/CoLo sites of contact shown by white dashed circles in (D,E). White arrowheads denote aberrant presynaptic V5-ZO1b or Cx34.1 staining in transplanted nbeaa−/− Mauthner cells. (F-K) Quantification of synaptic protein localization at M/CoLo synapses in control nbeaa+/+ to WT (WT) and in nbeaa−/− to WT (nbeaa−/−) transplants for V5-ZO1b (F-H) or Cx34.1-GFP (I-K). Fluorescence intensity is quantified and graphed as fold change comparing protein localization associated with the transplanted Mauthner cell to that of the host synapses within the same animal. (F,I) Each dot represents the average fold change for all synapses in an animal. (G,J) Each indicator represents the fold change for anterior segments (first eight segments imaged) and posterior segments (next eight segments imaged) with a line connecting values for an individual animal. (H,K) Each dot represents the average fold change for anterior or posterior segment synapses in an animal. Mean ± SEM are shown. For (F-H) wt, n = 5, nbeaa−/−, n = 5. For (I-K) wt, n = 5, nbeaa−/−, n = 5. In (F,I) ** indicates p < 0.01 by unpaired t-test. In (H,K) * indicates p < 0.05, ** indicates p < 0.01, and *** indicates p < 0.001 by ANOVA. ant = anterior synapses, pos = posterior synapses. Scale bars are as indicated. See also Figure S3.

In the next set of transplants, we examined the effect of removing Neurobeachin from Mauthner on Connexin localization. For the transplant donor cells, we used the Tg(Cx34.1-GFP^AE37) transgenic line that labels Cx34.1 with GFP. At the dendritic CEs of nbeaa−/− mutant Mauthner neurons, we observe that the localization of postsynaptic Cx34.1-GFP and presynaptic Cx35.5 (labeled by antibody) are both reduced as compared to wildtype transplants (Figure 4C). These data show that the loss of Neurobeachin from the transplanted postsynaptic neuron affects the localization of the dendritically localized ZO1b and Cx34.1. In addition, the removal of Neurobeachin from Mauthner affects presynaptic Cx35.5 localization in the neighboring auditory afferent neuron. We conclude Neurobeachin is required cell autonomously in the postsynaptic neuron to localize ZO1b and Cx34.1, and is also required cell non-automously for presynaptic Cx35.5 localization.

Neurobeachin is required to restrict postsynaptic electrical synapse proteins to the dendritic compartment

We next took advantage of the fact that the Mauthner cell is both a postsynaptic partner at the CEs as well as the presynaptic partner at the M/CoLo electrical synapses. This configuration, in conjunction with the chimeric transplant experiments, allows us to visualize the compartmentalized localization of components of the electrical synapse (Figure 1A). For these experiments, as above, we focused on chimeric animals in which individual Mauthner cells were derived from the GFP+ donor. In control transplant experiments from Et(Tol-056:GFP); Pt(V5-ZO1b) to non-transgenic wildtypes (nbeaa+/+ to WT), the V5-tagged ZO1b is exclusively localized at CE synapses and is not present at M/CoLo synapses (Figure 4B,D, left panels). Thus, within the Mauthner cell, ZO1b is compartmentalized exclusively postsynaptically to the dendritic synapses and excluded from localizing presynaptically within the axon. Strikingly, in transplanted nbeaa−/− mutant Mauthner cells (nbeaa−/− to WT), we observed localization of V5-ZO1b presynaptically at M/CoLo synapses (Figure 4D, right panels, quantified in 4F). The Mauthner cell axon extends the length of the spinal cord and it makes en passant electrical synapses with CoLo neurons found in each hemi-segment. The mislocalized V5-ZO1b appears strongly at M/CoLo synapses found in the anterior half of the spinal cord, and diminishes at synapses found posteriorly (Figure 4G,H). We conclude that Neurobeachin is required to restrict ZO1b localization to the dendritic compartment and prevent it from localizing at axonal electrical synapses.

We next questioned whether the loss of Neurobeachin from the Mauthner cell would affect Connexins. In control Tg(Cx34.1-GFP^AE37) transplants to non-transgenic wildtypes (nbeaa+/+ to WT), where Mauthner cells were derived from the donor, Cx34.1-GFP is exclusively found compartmentalized to the CE synapses and is excluded from M/CoLo synapses (Figure 4C,E, left panels). By contrast, we observe that transplanted nbeaa−/− mutant Mauthner cells (nbeaa−/− to WT) mislocalize Cx34.1 presynaptically at axonal M/CoLo synapses (Figure 4E, right panels, quantified in 4I). In addition, analogous to our V5-ZO1b transplant experiments, this mislocalization of Cx34.1 appears strongly at anterior M/CoLo synapses and diminishes posteriorly (Figure 4J,K). In the transplants in which Mauthner cells are nbeaa−/− mutant, and the CoLo neurons are derived from the wildtype host (i.e. they are GFP- and therefore nbea+/+), we observed that the Cx35.5 staining at M/CoLo synapses is not affected (Figure 4D,E, right panels). This result suggests that Neurobeachin does not function presynaptically to localize Connexins. To test this directly, we examined whether presynaptic Cx35.5 would mislocalize to dendritic postsynaptic CE synapses in nbeaa−/− mutants (Figure S3A-D). We used CRISPR to V5-tag endogenous Cx35.5 protein and examined injected mosaic F0 animals for the localization of Cx35.5-V5 in wildtype and nbeaa−/− mutants. In wildtype, we identified injected animals in which all M/CoLo synapses have V5 staining, as expected for the presynaptic usage of Cx35.5 in the Mauthner axon; in such cases, no Cx35.5-V5 staining was observed at CE synapses (Figure S3A,B). In nbeaa−/− mutants, we found animals with the same axonal distribution of Cx35.5-V5 present at all M/CoLo synapses (Figure S3D), however the amount localized was reduced as expected in nbeaa−/− mutants (compare to Figure 1F). Critically, in nbeaa−/− mutants, no Cx35.5-V5 staining is observed at CE synapses (Figure S3C). We conclude that Neurobeachin is not required for axonal Cx35.5 compartmentalization, but is required to restrict Cx34.1 to the dendrite.

Taken together, we find that the selectively removing Neurobeachin from the Mauthner cell leads to a reduction of ZO1b and Cx34.1 at the dendritic CEs, and that these proteins are mislocalized to the axonal M/CoLo presynapses. By contrast, presynaptic Cx35.5 in the Mauthner cell is unaffected by the removal of Neurobeachin. Thus, we conclude that Neurobeachin functions specifically to constrain dendritically localized synaptic components to their appropriate compartment, thereby supporting the subcellular specificity of electrical synapse formation (Figure 5).

Figure 5: Model of Neurobeachin function in electrical synapse formation.

Diagram showing proposed Neurobeachin (yellow) roles in regulating the hierarchical assembly and compartmentalization of Mauthner neuron electrical synapse proteins.

Discussion

The experiments described here investigating Neurobeachin provide new biochemical and cell biological insight into the development of electrical synapse molecular organization. First, the findings expand the understanding of the hierarchical construction of the electrical synapse beyond the simplistic view of electrical synapses composed solely of Connexin gap junction channels. We find that Neurobeachin localizes to the electrical synapse independent of ZO1b or Connexins, yet Neurobeachin is necessary for these proteins to robustly localize to the developing synapse. Furthermore, we have identified a biochemical connection between Neurobeachin and ZO1b, and previous work found that ZO1b in turn binds Cx34.1³⁸. Thus, our emerging model is that Neurobeachin sits at the top of an electrical synapse hierarchy, directing ZO1b to the synapse, which in turn localizes Connexin to build the neuronal gap junction. This new model expands the molecular ‘borders’ of an electrical synapse and suggests that neuronal gap junction channels are the output of complex molecular interactions, including Neurobeachin, that regulate electrical synapse formation, analogous to chemical synaptogenesis. We further show that Neurobeachin functions postsynaptically in building the electrical synapse, and so why are presynaptic Cx35.5 proteins affected when Neurobeachin is mutant in all cells of the animal? We have observed that the removal of Cx34.1 from only the postsynaptic neuron results in a decrease in presynaptic Cx35.5 localization in the adjacent, coupled neuron³⁶. Further, removing ZO1b from only the postsynaptic neuron cell autonomously reduces the localization of postsynaptic Cx34.1, and non-cell autonoumously affects presynaptic Cx35.5 localization³⁸. Thus, we hypothesize that eliminating Neurobeachin function in all cells results first in the reduction of postsynaptic ZO1b/Cx34.1 localization, which then in turn results in the reduction of Cx35.5 presynaptically. This model is supported by the transplant experiments in which Neurobeachin function is specifically removed from the Mauthner neuron. These results add further support to the notion that electrical synapses, like chemical synapses, are molecularly complex synaptic compartments, that can have asymmetric biochemical compositions with molecular functions that have both cell autonomous and non-cell autonomous effects on building the neuronal gap junction channels.

While clear evidence of electrical synapse subcellular specificity exists in invertebrate and vertebrate neural circuits^9-13,41, the mechanisms to achieve this are largely unknown. The results presented here establish that Neurobeachin is required for the compartmentalized synaptic localization of postsynaptic, but not presynaptic, electrical synapse proteins. These findings identify Neurobeachin as the first molecular scaffold to regulate the subcellular specificity of vertebrate electrical synapses. Given that Neurobeachin localizes on vesicular structures near the Golgi, in the dendrites, and also at electrical synapses, it is tantalizing to suggest that it directs synaptic components, such as a vesicular-associated ZO1b/Cx34.1 co-transported complex, to the postsynaptic site. We note that in nbeaa−/− mutant transplants the mislocalized postsynaptic proteins are most strongly concentrated at axonal synapses near the cell soma, and diminish at more distal synapstic sites. We suspect that in nbeaa−/− mutant cell axons, ZO1b/Cx34.1 that is found mislocalized to the axon may be in limiting quantities insufficient to localize to all M/CoLo synapses down the length of the spinal cord. Alternatively, these normally dendritically restricted proteins may not competently engage with the axonal transport machinery. We are currently unable to observe electrical synapse proteins while they are being trafficked within dendrites and axons. Instead, at this time, we can only view these proteins once they have accumulated at synaptic contacts. Therefore, we cannot exclude the possibility that in wildtype neurons, electrical synapse proteins are trafficked into both compartments and postsynaptic proteins are then stabilized at their appropriate synaptic contacts by interactions with Neurobeachin. Future technical improvements in detection and imaging will be required to assess the trafficking pathways taken by these, and other, electrical synapse proteins. Despite this, our results clearly demonstrate that Neurobeachin is required to restrict postsynaptic ZO1b and Cx34.1 to their normal dendritic locations, but it does not have a role in localizing presynaptic Cx35.5 to its axonal location. Taken together, we propose a model in which Neurobeachin is necessary for the trafficking and/or stabilization of postsynaptic proteins to the electrical synapse (Figure 5).

How might Neurobeachin direct electrical synapse proteins to specific compartments? Neurobeachin contains an A-Kinase Anchoring Protein (AKAP) domain which could promote the formation of a synaptic PKA signaling complex to phosphorylate electrical synapse proteins upon cAMP stimulation^21,42. Phosphorylation can alter gap junction processes such as Connexin trafficking, assembly, degradation, and channel function⁴³. Additionally, recent findings in C. elegans reveal that cAMP signaling is necessary for the trafficking of Innexins to specific subcellular regions within neurons¹⁴. Thus, it is enticing to speculate that Neurobeachin may be the cell biological link between cAMP signaling and the compartmentalized trafficking of electrical synapse components, and such experiments await further testing in both vertebrate and invertebrate neurons. Overall, these emerging pieces of evidence imply that vertebrate and invertebrate neurons, despite differences in the Connexin and Innexin building blocks of gap junctions, may share mechanisms in building and compartmentalizing electrical synapses within neural circuits.

Why would a neuron compartmentalize electrical synapse components? The localization of electrical synapses to distinct subcellular compartments has critical impacts on neuronal function, such as lateral excitation via dendro-dendritic electrical synapses⁹ or strong synchronization via axo-axonic electrical synapses^10-13. These configurations require the compartmentalized delivery or capture of electrical synapse components at distinct subcellular sites. In the Mauthner cell, such compartmentalization creates molecularly asymmetric electrical synapses, with unique pre- and postsynaptic components^36-37. Molecular asymmetry is an important feature of electrical synapses which can determine functional properties, for instance it is often associated with electrical rectification^44-45. In nbeaa−/− neurons, postsynaptic proteins are inappropriately localized presynaptically, co-mingling with the normal presynaptic proteins at the axonal presynapse. The resulting synapse is therefore more symmetric, with ZO1 and Cx34.1 localized on both sides of the synapse, which could alter the functional properties of the circuit. Thus, Neurobeachin represents a cell biological mechanism by which neurons can direct electrical synapse components to appropriate locations, building unique synaptic structures at subcellular localizations to articulate specific circuit functions.

While here we have focused on the electrical synapse, it is intriguing to note that Neurobeachin is necessary for both electrical and chemical synapse development^20-22. Here we show that Neurobeachin binds to the electrical synapse scaffold ZO1, while previous work showed that it can bind to chemical synapses scaffolds including PSD95 and SAP102^30,39. Indeed, electrical and chemical synapse formation are well recognized to be intimately coordinated during development^1-2,8. Neurobeachin therefore presents itself as an intriguing biochemical bridge uniting synapse formation and invites an essential question of whether or not this protein uses similar or distinct mechanisms to regulate these molecularly diverse structures. Previous work shows that Neurobeachin co-localizes with chemical synapse scaffolding molecules like SAP102 along dendrites as well as at the synapse³⁹. Upon Neurobeachin loss, chemical synapse receptors do not localize to the cell surface, but instead accumulate within the cell soma²⁹. Yet whether Neurobeachin functions to restrict postsynaptic chemical synapse proteins to the dendrite, as we have shown here for electrical synapse proteins, remains to be examined. Additionally, Neurobeachin mutations have been identified in patients with autism, intellectual disability, and epilepsy, with more than 20 variants occurring broadly across the recognized domains of the protein including throughout the C-terminal domain^23-26, and our results here found this region is required to bind the electrical synapse scaffold ZO1b. Thus, understanding the structure/function relationship of how Neurobeachin coordinates electrical and chemical synapse formation will reveal insight into these neurodevelopmental disorders and provide potential therapeutic targets. Altogether, this work illuminates a biochemical and cell biological mechanism at the nexus of balancing electrical and chemical synaptogenesis and opens a window for future investigations underlying these fundamental processes.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Adam C. Miller (acmiller@uoregon.edu).

Materials availability

Plasmids generated in this study are available upon request.

Data and code availability

• Microscopy and other data reported in this paper will be shared by the lead contact upon request.

• No original code is reported in this study.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Zebrafish

Zebrafish (Danio rerio) were bred and maintained with approval from the Institutional Animal Care and Use Committee (IACUC AUP 21-42) within the University of Oregon fish facility. Fish were kept at 28.5°C on a 14 hr on and 10 hr off light cycle and the developmental timepoints used originated from standard developmental staging³¹. Animals were housed in groups according to genotype with a limit of 30 animals per tank. All fish for this project originated from the ABC background and most fish had the enhancer trap transgene Et(Tol-056:GFP) in the background, which labels the Mauthner and CoLo neurons with GFP³⁵, unless otherwise noted. Mutant lines were genotyped for all experiments. At this stage of development, zebrafish sex is not yet determined⁴⁶.

Generation of Cx34.1-oxGFP fish

To generate the fish in which Cx34.1 is tagged with a fluorescent protein, Cx34.1 was tagged with oxGFP, which is optimized for use in oxidizing environments and biological membranes⁴⁷. A Cx34.1 coding region expression plasmid was made using zebrafish genomic BAC clone CH211-87N9. The fluorescent protein coding region was inserted 21 amino acids before end of the C-terminus at a location that was previously found not to disrupt regulation of coupling by Connexin phosphorylation and not to block the C-terminal PDZ-domain interaction motif⁴⁸. Promoter elements of the Connexin gene were derived from the same BAC clone. The transgene construct also includes a mCherry cardiac reporter for ease of identification in injected embryos. The transgenic fish line (Cx34.1oxGFP^AE37) was generated at the Zebrafish Core of the Albert Einstein College of Medicine using the Tol2 transposon system. The localization of Cx34.1-oxGFP was found at the expected electrical synapses throughout the brain, and particularly at CE and M/CoLo synapses of the Mauthner circuit as shown in this manuscript, using double-labeling with anti-Cx34.1 and anti-GFP. Moreover, electrical transmission at Mauthner CE synapses was found to be indistinguishable from that of WT fish during electrophysiological recordings (data not shown).

Cas9-mediated genome engineering of Cx35.5-V5

To generate fish at which endogenous Cx35.5 is tagged with V5 epitope tag, a single guide RNA (sgRNA) targeting exon 2 of the endogenous gjd2a coding sequence was designed using the CRISPRscan algorithm⁴⁹ and synthesized as previously described⁵⁰. The sgRNA was generated using the T7 megascript kit (ThermoFisher, AMB13345). The target site, including PAM, is: 5’-CCCAATTTCGGCCGCACTCAGTC-3’. The gjd2a-V5 single stranded donor oligo (ssODN) was designed to repair into the endogenous gjd2a locus 42 bp prior to the stop codon and was synthesized complementary to the coding strand. This location was previously found not to disrupt Connexin localization and C-terminal PDZ-domain interactions⁴⁸. The ssODN contained a 5’ 40 bp and 3’ 50 bp homology arms flanking a 5x glycine linker, with a V5 epitope tag in the center. The sequence of this repair construct is: 5’-CCGCAGGAAGTCCATCTATGAGATCAGGAACAAGGATTTGGGAGGTGGAGGCGGAGGTAAGCCTATCCCTAACCCTCTCCTCGGTCTCGATTCTACGGGAGGTGGAGGCGGACCGCGGATGAGTATGCCCAACTTTGGACGAACCCAGTCCAGTGACTCGGC-3’. Upon correct repair, the inserted sequence was designed to disrupt the endogenous recognition site of the CRISPR target sequence by creating four coding-silent mutations to prevent further double stranded breaks after repair. Injection mixes were prepared in a pH 7.5 buffer solution of 300 mM KCl and 4 mM HEPES and contained a final concentration of 200 pg/nL ssODN, 200 pg/nL gRNA, and 1600 pg/nL Cas9 protein (IDT, 1081058). Injection mixes were incubated at 37 °C for 5 min immediately prior to injection to promote formation of the Cas9 and sgRNA complex. Finally, 1 nL of solution was injected into embryos at the one-cell stage. F0 injected animals were used for analysis and signal verified by coimmunostaining with V5 and Cx35.5 antibody labeling.

Cell culture

HEK293T/17 verified cells were purchased from ATCC (CRL-11268; STR profile, amelogenin: X). Cells were passaged and maintained in Dulbecco's Modified Eagle's Medium (DMEM, ATCC) plus 10% fetal bovine serum (FBS, Gibco) at 37°C in a humidified incubator in the presence of 5% CO₂. Low passage aliquots were cryopreserved in liquid nitrogen according to manufacturer’s instructions. Cells from each thawed cryovial were monitored for mycoplasma contamination using the Universal Mycoplasma Detection Kit (ATCC, 30–1012K).

METHOD DETAILS

Blastula cell transplantation

Cell transplantation was performed at the high stage approximately 3.3 hr into zebrafish development using standard techniques⁴⁰. Embryos were chemically de-chorionated with protease type XIV (Sigma Aldrich, 9036-06-0) prior to transplantation. Cells were transplanted using a 50 μm wide glass capillary needle attached to an oil hydraulic. For nbeaa+/+ or nbeaa−/− ; Et(Tol-056:GFP); Pt(V5-ZO1b) to wildtype transplants, and nbeaa+/+ or nbeaa−/−; Tg(Cx34.1-GFP^AE37) to wildtype transplants, cells from genotyped donor animals were transplanted into non-transgenic ABC hosts. Approximately 20 cells were deposited ~10–15 cell diameters away from the margin, with a single donor embryo supplying cells to 1-3 hosts. At 5 dpf, larvae were fixed in 2% trichloroacetic acid (TCA) in PBS⁵¹, donor animals genotyped, and host animals were sorted based on the genotype of the donor and processed for immunohistochemistry. As more than a single cell is transplanted to the host embryo, using confocal microscopy, immunostained larvae were examined for GFP+ Mauthner and GFP+ CoLo cells in the chimeras. Animals with successful GFP+ Mauthner cell transplants were imaged and analyzed, with any synapses resulting at the junction of both a GFP+ Mauthner cell and a GFP+ CoLo cell disregarded.

Immunohistochemistry and confocal imaging

Anesthetized, 5 days post fertilization (dpf) larvae were fixed for 4 hr in 2% TCA in PBS⁵¹. Fixed tissue was washed in PBS + 0.5% Triton X-100, followed by standard blocking and antibody incubations. Primary antibody mixes included combinations of the following: rabbit anti-Cx35.5 (Fred Hutch Antibody Technology Facility, clone 12H5, 1:200), mouse IgG2A anti-Cx34.1 (Fred Hutch Antibody Technology Facility, clone 5C10A, 1:200), mouse IgG1 anti-ZO1 (Invitrogen, 33–9100, 1:350), mouse IgG2a anti-V5 peptide (Invitrogen, R960-25, 1:500), rabbit anti-Neurobeachin (Invitrogen, PA5-58903, 1:100), mouse anti-GM130, BD Biosciences, 610823, 1:100), mouse anti-RMO44** Neurofilament-M (Invitrogen, 13-0500, 1:50), and chicken anti-GFP (abcam, ab13970, 1:500). All secondary antibodies were raised in goat (Invitrogen, conjugated with Alexa-405,–488, –555, or –633 fluorophores, 1:500). Tissue was then cleared stepwise in a 25%, 50%, 75% glycerol series, dissected, and mounted in ProLong Gold antifade reagent (ThermoFisher, P36930). Images were acquired on a Leica SP8 Confocal using a 405-diode laser and a white light laser set to 499, 553/554/557 (consistent within experiments), and 631 nm, depending on the fluorescent dye imaged. Each laser line’s data was collected sequentially using custom detection filters based on the dye. Quantitative images of the Club Endings (CEs) and M/CoLo transplants were collected using a 63x, 1.40 numerical aperture (NA) oil immersion lens, and other images of M/Colo synapses were collected using a 40x, 1.20 NA water immersion lens. For each set of images, the optimal optical section thickness was used as calculated by the Leica software based on the pinhole, emission wavelengths and NA of the lens. Within each experiment where fluorescence intensity was to be quantified, all animals (including 5 or more wildtype controls) were stained together with the same antibody mix, processed at the same time, and all confocal settings (laser power, scan speed, gain, offset, objective, and zoom) were identical. Multiple animals per genotype were analyzed to account for biological variation. To account for technical variation, fluorescence intensity values for each region of each animal were an average across multiple synapses.

Analysis of confocal imaging

For fluorescence intensity quantitation, confocal images were processed and analyzed using FiJi software⁵². To quantify staining at M/Colo synapses, a standard region of interest (ROI) surrounding each M/CoLo site of contact was drawn, and the mean fluorescence intensity was measured. For the quantification of staining at the club endings, confocal z-stacks of the Mauthner soma and lateral dendrite were cropped to 200 x 200 pixels centered around the lateral dendritic bifurcation. A FIJIscript was generated to clear the outside of the Mauthner cell based on GFP staining, and a standard threshold was set within each experiment to remove background staining. The image was then transformed into a max intensity projection, synapses thresholded to WT, and the integrated density of each stain within the club ending synapses was extracted. Standard deviations and errors were computed using Prism (GraphPad) or Excel (Microsoft) software. Figure images were created using FiJi, Photoshop (Adobe), and Illustrator (Adobe). Statistical analyses were performed using Prism (GraphPad). For all experiments, unless otherwise listed in the figure legend, values were normalized to wildtype control animals, and n represents the number of fish used.

Cell transfection and co-immunoprecipitation

Full-length Cx34.1, Cx35.5, and ZO1b were cloned into the pCMV expression vectors as previously described³⁸. A Flag-tag was added to the N-terminus of the ZO1b construct to facilitate Flag-ZO1b pull down experiments. N-terminal and C-terminal Neurobeachin constructs were cloned into the pCMV expression vector with an mVenus tag. Low passage HEK293T/17 cells were seeded 24 hr prior to transfection (1 × 10⁶ cells/well of a six-well dish), and the indicated plasmids were co-transfected using 2ug DNA and Lipofectamine 3000 (Invitrogen) following the manufacturer’s instructions. Cells were collected 24 hr post-transfection and lysed in 500 mL solubilization buffer (50 mM Tris [pH7.4], 100 mM NaCl, 5 mM EDTA, 1.5 mM MgCl2, 1 mM DTT and 1% Triton X-100) plus a protease inhibitor cocktail (Pierce). Lysates were centrifuged at 20,000 x g for 30 min at 4°C, and equal amounts of extract were immunoprecipitated with 0.5 ug rabbit anti-GFP (Abcam, Ab290) or 0.5 ug mouse anti-Flag (Sigma, F3165) overnight with rocking at 4°C. Immunocomplexes were captured with 40 μl prewashed Protein A/G agarose beads for 1 hr with rocking at 4°C. Beads were washed three times with lysis buffer, and bound proteins were boiled for 5 min in the presence of LDS-PAGE loading dye containing 200 mM DTT. Samples were resolved by SDS-PAGE using a 4–20% or 8-16% gradient gel and analyzed by Western blot using the following primary antibodies: chick anti-GFP (Abcam Ab13970), rabbit anti-Cx34.1 3A4-conjugated-680LT, and rabbit anti-Cx35.5 12H5-conjugated-680LT, mouse IgG1 anti-ZO1 (Invitrogen, 33–9100, 1:1000). Compatible near-infrared secondary antibodies were used for visualization with the Odyssey system (LI-COR). Quantification was performed using Image Lab (Bio-Rad Laboratories, Inc.). To account for variable expression, values for immunopreciptated proteins were divided by the values detected in extracts, and values reported in the graphs are ratios created of co-immunopreciptated proteins to the immunopreciptated proteins. Resulting values were normalized to the strongest IP band⁵³. Statistical analyses were performed using Prism (GraphPad).

Immunoprecipitation and western blotting of fish brain homogenates

Brains from wildtype, nbeaa−/−, Cx34.1−/−, and Cx35.5−/− euthanized adult fish (4–15 months old) were removed, snap frozen in liquid nitrogen and stored at −80°C until use. To test comparable levels of connexins, two brains of each condition were homogenized in 250uL HSE buffer (20 mM Hepes [pH7.5], 150 mM NaCl, 5 mM EDTA, 5 mM EGTA, and 1 mM DTT) plus a protease inhibitor cocktail using a glass homogenizer. Detergent was added to the homogenate (final 2% octyl ß-Dglucopyranoside, Anatrace) and solubilized overnight with rocking at 4°C. Solubilized homogenate was cleared by centrifugation at 20,000 x g for 30 min at 4°C and Bradford-equalized amounts of extract were immunoprecipitated with 1.5 ug mouse anti-Cx34.1 or mouse anti-Cx35.5 overnight with rocking at 4°C (For details regarding the specificity of antibodies see Miller et al., 2017). Immuno-complexes were captured with 40 μl prewashed Protein A/G agarose beads for 1 hr with rocking at 4°C. Beads were washed three times with lysis buffer, and bound proteins were boiled for 5 min in the presence of LDS-PAGE loading dye containing 200 mM DTT. Samples were resolved by SDS-PAGE using a 4–20% gradient gel and analyzed by Western blot using rabbit anti-Cx34.1 3A4-conjugated-680LT or rabbit antiCx35.5 12H5-conjugated-680LT. Quantification was performed using Image Lab. For the ZO1 assay, three brains of each condition (nbeaa+/+ and nbeaa−/−) were homogenized in 300uL of HSE buffer (20 mM Hepes [pH7.5], 150 mM NaCl, 5 mM EDTA, 5 mM EGTA, and 1 mM DTT) plus a protease inhibitor cocktail using a glass homogenizer. Detergent was added to the homogenate (final 2% octyl ß-Dglucopyranoside, Anatrace) and solubilized overnight with rocking at 4°C. Solubilized homogenate was cleared by centrifugation at 20,000 x g for 30 min at 4°C. Samples were boiled for 5 min in the presence of LDS-PAGE loading dye containing 200 mM DTT. Proteins were examined by western analysis using mouse anti-ZO1 (1:1000), and rabbit anti-beta tubulin (abcam, ab6046, 1:10,000). Compatible near-infrared secondary antibodies were used for visualization. Quantification was performed using Image Lab (Bio-Rad Laboratories, Inc.). Resulting values were normalized to the wildtype values⁵³. Statistical analyses were performed using Prism (GraphPad).

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification and statistical details including analysis methods and materials used are included above in the Methods details as well as in the figure legends for each figure.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
rabbit anti-Cx35.5	Fred Hutch Antibody Technology Facility	clone 12H5
mouse IgG2A anti-Cx34.1	Fred Hutch Antibody Technology Facility	clone 5C10A
mouse IgG1 anti-ZO1	Invitrogen	33–9100
mouse IgG2a anti-V5 peptide	Invitrogen	R960-25
rabbit anti-Neurobeachin	Invitrogen	PA5-58903
mouse anti-GM130	BD Biosciences	610823
mouse anti-RMO44** Neurofilament-M	Invitrogen	13-0500
chicken anti-GFP	abcam	ab13970
goat anti- Alexa405/488/555/633	Invitrogen
rabbit anti-GFP	abcam	ab290
mouse anti-Flag	Sigma	F3165
rabbit anti-beta tubulin	abcam	ab6046
Bacterial and virus strains
Biological samples
Chemicals, peptides, and recombinant proteins
Cas9 protein	IDT	1081058
protease type XIV	Sigma Aldrich	9036-06-0
ProLong Gold antifade reagent	ThermoFisher	P36930
Critical commercial assays
Universal Mycoplasma Detection Kit	ATCC	30–1012K
T7 megascript kit	ThermoFisher	AMB13345
Deposited data
Experimental models: Cell lines
HEK293T/17; STR profile, amelogenin: X	ATCC	CRL-11268
Experimental models: Organisms/strains
Et(Tol-056:GFP) zebrafish	Satou et al. 200935	zf206Et
tjp1bΔ16bp zebrafish	Marsh et al., 201737	b1370
gjd1aΔ8bp zebrafish	Miller et al., 201736	fh436
gjd2aΔ5bp zebrafish	Miller et al., 201736	fh437
Pt(V5-ZO1b) zebrafish	Lasseigne et al., 202138	b1406
nbeaaQ906* zebrafish	Miller et al., 201520	fh364
Tg(Cx34.1-oxGFP) zebrafish	This publication.	AE37
Oligonucleotides
Cx35.5 target site: 5’-CCCAATTTCGGCCGCACTCAGTC-3	This publication.	n/a
Cx35.5 repair oligo: 5’-CCGCAGGAAGTCCATCTATGAGATCAGGAACAAGGATTTGGGAGGTGGAGGCGGAGGTAAGCCTATCCCTAACCCTCTCCTCGGTCTCGATTCTACGGGAGGTGGAGGCGGACCGCGGATGAGTATGCCCAACTTTGGACGAACCCAGTCCAGTGACTCGGC-3’	This publication.	n/a
Recombinant DNA
Plasmid: pCMV:Cx34.1	Lasseigne et al., 202138	n/a
Plasmid: pCMV:Cx35.5	Lasseigne et al., 202138	n/a
Plasmid: pCMV:ZO1b	Lasseigne et al., 202138	n/a
Plasmid: pCMV:Flag-ZO1b	This paper.	n/a
Plasmid: pCMV: mVenus N-term Nbea	This paper.	n/a
Plasmid: pCMV: mVenus C-term Nbea	This paper.	n/a
Plasmid: pCMV: mVenus	This paper.	n/a
Plasmid: pCMV: mVenus ZO1b	This paper.	n/a
Software and algorithms
FIJI		Fiji.sc
Graphpad Prism		Graphpad.com
ImageLab		Bio-rad.com
Other

Highlights.

• Neurobeachin instructs hierarchical electrical synapse assembly.

• Electrical synapse proteins require Neurobeachin for robust synaptic localization.

• Neurobeachin localizes to electrical synapses and interacts with the scaffold ZO1.

• Neurobeachin restricts postsynaptic electrical synapse proteins to the dendrite.