E3 ubiquitin ligases LNX1 and LNX2 localize at neuronal gap junctions formed by connexin36 in rodent brain and molecularly interact with connexin36

Electrical synapses in the mammalian central nervous system (CNS) are increasingly recognized as highly complex structures for mediation of neuronal communication, both with respect to their capacity for dynamic short‐ and long‐term modification in efficacy of synaptic transmission and their multimolecular regulatory and structural components. These two characteristics are inextricably linked, such that understanding of mechanisms that contribute to electrical synaptic plasticity requires knowledge of the molecular composition of electrical synapses and the functions of proteins associated with these synapses. Here, we provide evidence that the key component of gap junctions that form the majority of electrical synapses in the mammalian CNS, namely connexin36 (Cx36), directly interacts with the related E3 ubiquitin ligase proteins Ligand of NUMB protein X1 (LNX1) and Ligand of NUMB protein X2 (LNX2). This is based on immunofluorescence colocalization of LNX1 and LNX2 with Cx36‐containing gap junctions in adult mouse brain versus lack of such coassociation in LNX null mice, coimmunoprecipitation of LNX proteins with Cx36, and pull‐down of Cx36 with the second PDZ domain of LNX1 and LNX2. Furthermore, cotransfection of cultured cells with Cx36 and E3 ubiquitin ligase‐competent LNX1 and LNX2 isoforms led to loss of Cx36‐containing gap junctions between cells, whereas these junctions persisted following transfection with isoforms of these proteins that lack ligase activity. Our results suggest that a LNX protein mediates ubiquitination of Cx36 at neuronal gap junctions, with consequent Cx36 internalization, and may thereby contribute to intracellular mechanisms that govern the recently identified modifiability of synaptic transmission at electrical synapses.

Gap junction‐forming connexins have a remarkably fast turnover time, with a half‐life of between 2–4 hr, and several connexins have been reported to undergo degradation via E3 ubiquitin ligase‐mediated mechanisms. We found that electrical synapses formed by gap junctions composed of Cx36 in rat and mouse brain are associated with the E3 ubiquitin ligase proteins LNX1 and LNX2, and that these ligases molecularly interact with Cx36, promoting its degradation and potentially contributing to its rapid rate of turnover.

Keywords: connexin degradation; connexin trafficking; electrical synapses; gap junctions; ubiquitination

• Abbreviations
• co‐IP coimmunoprecipitation
• CNS central nervous system
• Cx36 connexin36
• eCFP enhanced cyan fluorescent protein
• eGFP enhanced green fluorescent protein
• IP immunoprecipitation
• LNX1 ligand of NUMB protein X1
• LNX2 ligand of NUMB protein X2
• LNX1/2 LNX1 and LNX2
• MUPP1 multi‐PDZ domain protein1
• PBS 50 mm sodium phosphate buffer, pH 7.4, containing 0.9% NaCl
• PDZ PSD‐95/SAP90, DlgA, ZO‐1
• PVDF polyvinylidene difluoride
• RIPA buffer 50 mM Tris pH 8.0, 150 mM NaCl, 1.0% IGEPAL CA‐630, 0.5% sodium deoxycholate
• SDS‐PAGE sodium dodecylsulphate polyacrylamide gel electrophoresis
• TBS 50 mM Tris–HCl, pH 7.4, containing 1.5% sodium chloride
• TBSTr TBS containing 0.3% Triton X‐100
• TBS‐Tw 20 mM Tris pH 7.4, 150 mM NaCl, 0.2% Tween‐20 and 5% skim milk powder
• trCx36 truncated Cx36
• ZO zonula occludens

Gap junctions formed between apposing cells by various members of the family of 20 connexin proteins mediate gap junctional intercellular communication, where junctional channels allow cell‐to‐cell passage of ions and small molecules (Goodenough & Paul, [24]). Such junctions between neurons in the central nervous system (CNS) provide for direct electrical neuronal communication and are the morphological basis of electrical synapses (Bennett, [4]; Bennett & Zukin, [5]). In the mammalian CNS, neuronal gap junctions formed by connexin36 (Cx36) are widely distributed (Condorelli, Belluardo, Trovato‐Salinaro, & Mudo, [10]; Nagy, Pereda, & Rash, [58]), and have critically important functions in establishing synaptic circuitry during development (Miller & Pereda, [53]) and in facilitating synchronous neuronal activity in mature brain (Connors, [11]; Hormuzdi, Filippov, Mitropoulou, Monyer, & Bruzzone, [30]). It has recently been recognized that electrical synapses are highly dynamic, with reports indicating that neuronal communication mediated by these synapses is subject to modification by a variety of processes. These include regulation of electrical coupling by various neurotransmitters (Nagy et al., [58]), by the state of neuronal activity (Haas, Zavala, & Landisman, [27]), by connexin phosphorylation (Kothmann, Massey, & O'Brien, [36]; O'Brien, [60]; Urschel et al., [74]), and by ion‐mediated gating of Cx36‐containing channels (Palacios‐Prado et al., [62], [61]). Importantly, such regulation has a functional impact on the behaviour of neuronal networks interconnected by electrical synapses in neural circuitry (Bloomfield & Volgyi, [8]; Curti & O'Brien, [12]; Haas, Greenwald, & Pereda, [26]; Pereda, [63]; Pereda et al., [64]).

The exact mechanisms underlying modulation of the strength of electrical coupling are not fully understood, but one possibility is via regulated trafficking of Cx36 into and out of gap junctions via exocytosis and endocytosis of Cx36 channels (Pereda, [63]; Pereda et al., [64]). Such trafficking has been documented at vertebrate electrical synapses, and interference with this process rapidly altered electrical coupling (Flores et al., [18]), which is consistent with the high turnover rate of Cx36, estimated to be ~1–3 hr (Flores et al., [18]; Wang, Lin, Mitchell, Ram, & O'Brien, [75]) and is similar to that of other connexins (Berthoud, Minoque, Laing, & Beyer, [6]; Evans & Martin, [16]). Such trafficking presumably occurs via processes governing Cx36 removal and disposal, which for most connexins are poorly understood, with the exception of connexin43 (Cx43) that undergoes ubiquitination and degradation by the proteasome pathway (Berthoud et al., [6]; Herve, Derangeon, Sarrouilhe, Giepmans, & Bourmeyster, [29]; Kjenseth, Fykerud, Rivedal, & Leithe, [35]). The attachment of ubiquitin to target proteins is a common cellular sorting mechanism for determination of protein fate along the endocytotic and exocytotic pathways (Deshaies & Joazeiro, [14]). Ubiquitination involves the covalent linkage of ubiquitin to lysine residues of target proteins. This process occurs via a stepwise enzyme cascade that requires three distinct enzyme participants; an ubiquitin‐activating enzyme (E1), an ubiquitin‐conjugating enzyme (E2), and an ubiquitin ligase (E3). The E3 ligase in this cascade determines target‐protein specificity, and over 600 E3 ligases have been identified and distinguished by characteristics of their catalytic domain, which include the HECT, RING, and U‐box family of proteins (Bhoj & Chen, [7]). Many RING domain E3 ligases contain additional domains required for their specific interaction with target proteins, including PDZ (PSD‐95, DlgA, ZO‐1) domains that mediate interaction with the PDZ ligand of their target proteins (Ye & Zhang, [78]).

We previously reported that Cx36 contains a PDZ domain interaction motif, consisting of a SAYV amino acid sequence at its carboxy‐terminus, that mediates its interaction with the PDZ domains of various proteins, including the first PDZ domains of zonula occludens (ZO) proteins ZO‐I, ZO‐2, ZO‐3, the 10th PDZ domain of multi‐PDZ domain protein1 (MUPP1), and the single PDZ domain in AF6 (aka, afadin) (Li, Lu, & Nagy, [47]; Li, Lynn, & Nagy, [48]; Li, Olson, Lu, & Nagy, [50]; Li, Olson, Lu, Kamasawa et al., [49]). This carboxy‐terminus SAYV PDZ ligand is contained in other proteins (e.g., claudin‐1 and CD8α), where it mediates protein–protein interaction for ubiquitination by specific E3‐ubiquitin ligases, namely, Ligand of NUMB protein X1 (LNX1) and Ligand of NUMB protein X2 (LNX2) (D'Agostino et al., [13]; Takahashi et al., [72]). This raised the possibility of Cx36 interaction with LNX1 and/or LNX2. These two related proteins are RING domain E2‐dependent E3 ubiquitin ligases whose PDZ domains are required for mediation of interactions with proteins targeted for ubiquitination. They share identical domain structure comprised of one amino‐terminal RING domain and four PDZ domains (Rice, Northcutt, & Kurschner, [68]). Each of the four PDZ domains of LNX1 and LNX2 interact with specific classes of PDZ ligands (Lenihan, Saha, & Young, [45]; Wolting et al., [76]). LNX2 is expressed in many tissues, including adult brain (Lenihan, Saha, Mansfield, & Young, [44]; Rice et al., [68]). More complex is LNX1, where three protein isoforms have been identified; a full‐length form containing the RING finger catalytic domain (LNX1p80), an alternatively spliced form devoid of this domain and thus lacking catalytic activity (LNX1p70), and a novel recently characterized LNX1p62 brain isoform also lacking the catalytic domain (Dho et al., [15]; Lenihan et al., [44]; Xie et al., [77]). In brain, LNX1p80 appears to be absent, whereas catalytically inactive LNX1p70 is expressed mainly in neurons and in a small subset of oligodendrocytes (Dho et al., 1998; Lenihan et al., [44]; Rice et al., [68]).

In order to explore means whereby Cx36 deployment, trafficking, and degradation could contribute to the plasticity of electrical synapses, we investigated Cx36 relationships with LNX1 and LNX2. We examined immunofluorescence localization of these proteins in combination with immunolabelling for Cx36, and used molecular and biochemical approaches to assess Cx36/LNX interaction.

Animals were utilized according to approved protocols by the Central Animal Care Committee of the University of Manitoba and by the Hospital for Sick Children Animal Care and Use Committee, with minimization of the numbers animals used. A total of four Sprague–Dawley rats and eight wild‐type (WT) C57BL/6 mice used in this study were obtained from Central Animal Care Services at the University of Manitoba. In addition, the two C57BL/6 LNX1 knockout (KO) mice used were obtained from the Toronto Centre for Phenogenomics. Brains from these animals were prepared and taken for immunofluorescence labelling and immunoblotting as described below. The LNX1 gene‐ablated mutant mouse strain was generated in the laboratory of Dr. Mark Henkemeyer (University of Texas Southwestern) as described (Liu et al.,[51]). Briefly, LNX1 was targeted by homologous recombination in the R1 ES cell line, germline transmission was obtained by aggregation chimeras, and the LNX1 KO mice obtained were backcrossed to C57BL/6 mice for multiple generations and were bred and maintained under institute guidelines at the Toronto Centre for Phenogenomics.

Antibodies used included a mouse monoclonal anti‐Cx36 (Cat. No. 39‐4200) and a rabbit polyclonal anti‐Cx36 (Cat. No. 51‐6300), which were obtained from ThermoFisher (ThermoFisher, Rockford, IL, USA) and used at a concentration of 2–3 μg/ml. In addition, rabbit polyclonal anti‐LNX1 and LNX2 antibodies were obtained originally from Invitrogen (Carlsbad, CA, USA), before this provider was acquired by Life Technologies and then ThermoFisher. Two anti‐LNX1 antibodies were generated against peptide sequences corresponding to amino acids 200–213 and 247–259 in LNX1, designated ab200 and ab247, and one anti‐LNX2 antibody was generated against a peptide corresponding to amino acid sequence 442–458 in LNX2, designated ab442. Additional rabbit polyclonal anti‐LNX1 antibody, used for immunoblotting of LNX1 from mouse tissues and for co‐IP of Cx36 with LNX1 from mouse brain, were generated against GST‐fusion proteins containing LNX1 amino acids (aa) 8–210 and aa365–478, designated ab8 and ab365 respectively (Dho et al., [15]). All antibodies against LNX were used at a dilution of 1:250. Mouse monoclonal anti‐His 6X‐tag (Cat. No. R930) and a monoclonal rabbit anti‐eGFP (enhanced green fluorescent protein) (Cat No. G10362) were obtained from ThermoFisher and both were used at a concentration of 2 μg/ml, a mouse monoclonal anti‐FLAG (M2) antibody (Cat. No. F1804) was obtained from Millipore Sigma (Sigma‐Aldrich Corp., St. Louis, MO, USA) and used at a dilution of 1:1000. Mouse monoclonal anti‐HA tag (F‐7) antibody (Cat. No. 7392) was obtained from Santa Cruz (Santa Cruz Biotechnology, Dallas TX, USA) and used at a concentration of 2 μg/ml.

Procedures used here for preparation of tissue for immunofluorescence have been previously described in detail for immunolabelling of Cx36 in combination with other proteins (Li et al., [48]; Lynn, Li, & Nagy, [52]; Rubio & Nagy, [69]). Briefly, animals were first euthanized with an intraperitoneal injection of equithesin (3 ml/kg) and then transcardially perfused sequentially with a prefixative solution followed by a fixative solution containing buffered 1% formaldehyde/0.2% picric acid. Cryostat sections were collected at a thickness of 15 μm and immunolabelled with monoclonal anti‐Cx36. For double immunofluorescence labelling, sections were simultaneously incubated with two primary antibodies (a mouse monoclonal and a rabbit polyclonal: Cx36 + LNX1; Cx36 + LNX2) in a dilution buffer containing 50 mM Tris–HCl, pH 7.4, 1.5% sodium chloride and 0.3% Triton X‐100 (TBSTr) and supplemented with 5% normal goat or normal donkey serum. Following incubation for 24 hr at 4°C, the sections were washed for 1 hr in TBSTr, and then incubated for 1.5 hr at room temperature simultaneously with appropriate combinations of secondary antibodies diluted in TBSTr with normal goat or donkey serum. Secondary antibodies included Cy3‐conjugated goat anti‐mouse IgG diluted 1:500 (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) and Alexa Flour 488‐conjugated goat anti‐rabbit IgG diluted 1:1000 (Molecular Probes, Eugene, OR, USA). After secondary antibody incubations, sections were washed in TBSTr for 20 min, then in 50 mM Tris–HCl buffer, pH 7.4 for 30 min, covered with antifade medium, and coverslipped. Control procedures included omission of one of the primary antibodies with inclusion of each of the secondary antibodies to establish the absence of inappropriate cross‐reactions between primary and secondary antibodies.

Wide‐field and confocal immunofluorescence microscopy were conducted on a Zeiss Axioskop2 fluorescence microscope (Carl Zeiss Canada, Toronto, ON, Canada) and on a Zeiss LSM 710 confocal microscope using ZEN software (Olympus Canada, Inc., Markham, ON, Canada) respectively. Figure plates were assembled using Adobe Photoshop CS (Adobe Systems, San Jose, CA, USA) and CorelDraw Graphics (Corel Corporation, Ottawa, ON, Canada). Confocal images of double labelling of Cx36 with either LNX1 or LNX2 in brain sections were acquired using an oil immersion 40x objective lens with numerical aperture of 1.4. All of the confocal images presented here were captured at a zoom factor ranging from 3.3 to 5.5 and a resolution of 1024 × 1024 pixels, and show single scans or maximum intensity projection z‐stacks of 2–7 optical scans, with a step size of 0.43 μm (defined as optimum for a 40x objective), giving projection thicknesses ranging from 1.72 to 3.44 μm, and are displayed without deconvolution. Laser illumination intensities and photomultiplier tube gains settings were adjusted to well below saturation of the most intense labelling observed to allow capture of the full range of labelling intensities and to minimize halation which can artificially increase the size of immunofluorescent areas of interest. As we observed minimal fading of fluorescence with the laser setting chosen, the laser dwell time was increased to a moderate level to provide for better image quality. A total of 56 single and z‐stack images were collected for a total of 122 images when the z‐stack optical scans are considered separately, representing collectively all the brain regions examined. Before compression of the optical scans into a single image, each of the scans in a z‐stack series was examined separately to exclude possible artifactual overlap of labelling for Cx36 with label for LNX that may occur from labels that were actually separated in the z‐dimension. In addition, the image stacks were displayed in a 3‐dimensional configuration, allowing 360° image rotation in the x, y, and z dimensions to further confirm the label for Cx36 and LNX remained associated at all angles of rotation.

HeLa cells (HeLa‐WT), HeLa cells stably transfected with Cx36 (HeLa‐Cx36), Neuro‐2A (N2A) and HEK293T cells were grown in Dulbecco's Modified Eagle's medium supplemented with 10% foetal bovine serum and maintained in a tissue culture incubator at 37°C with 5% CO2. Transient transfections were conducted using Lipofectamine LTX transfection reagent according to manufacturer's instructions (Invitrogen). Expression plasmids used for transfections included previously described LNX1p80‐FLAG, LNX1p70, and LNX2‐FLAG (Nie et al., [59]). Additionally, transfection with mutant expression plasmids included LNX1p80C48A‐FLAG and LNX2C51A‐FLAG that each contains a functionally negative mutation within the zinc coordination site of the N‐terminal RING domain of LNX1p80 and LNX2, which is required for ubiquitination competence (Nie et al., [59]). A HA‐ubiquitin plasmid in pcDNA3.1 was a gift from Dr. E. Yeh (Addgene plasmid # 18712; Kamitani, Kito, Nguyen, & Yeh, [33]). A full‐length Cx36 plasmid and a truncated Cx36Δ4 plasmid that lacks coding sequence for the last four C‐terminal amino acids in Cx36 (trCx36), which constitute its PDZ domain interaction motif, were previously described (Li, Olson, Lu, Kamasawa et al., [49]; Li, Olson, Lu, & Nagy, [50]). C‐terminal tagging of Cx36 prevents its interaction with proteins that bind the Cx36 C‐terminus. Thus, in the construct we used here (provided by Dr. S. G. Hormuzdi, University of Dundee), enhanced cyan fluorescent protein (eCFP) is inserted 15 amino acids upstream of the C‐terminal PDZ ligand in otherwise full‐length Cx36, freeing this ligand for interaction with the PDZ domains of Cx36‐interacting partners. This construct was previously shown to form functionally communicating gap junctions (Helbig et al., [28]). Following cell transfection (18–20 hr), cultures of cells grown on glass coverslips were washed in 50 mM sodium phosphate buffer, pH 7.4, containing 0.9% NaCl (PBS), fixed for 10 min in 2% formaldehyde in PBS, rinsed, and stored at 4°C. Immunofluorescence labelling of cells was conducted as described above for tissue sections.

Animals were euthanized with an intraperitoneal injection of equithesin (3 mg/kg), brains dissected, homogenized in 14 volumes of immunoprecipitation (IP) buffer (20 mM Tris‐HCl, pH 8.0, 140 mM NaCl, 1% Triton X‐100, 10% glycerol, 1 mM EGTA, 1.5 mM MgCl2, 1 mM dithiothreitol, 1 mM phenylmethylsulphonyl fluoride) supplemented with a protease inhibitor cocktail (Millipore Sigma), and then briefly sonicated. For preparation of cell lysates, cultures were rinsed with PBS, lysed in ice‐cold IP buffer (Li et al., [48]), briefly sonicated, centrifuged, and resulting supernatants used for IP, pull‐down assays, and directly for western blot analysis. The protein concentration in all samples was determined using a kit (Bio‐Rad Laboratories, Hercules CA, USA). For immunoblotting, samples in loading buffer were separated by sodium dodecylsulphate polyacrylamide gel electrophoresis (SDS‐PAGE), and proteins transferred to polyvinylidene difluoride membranes. Following transfer, membranes were blocked for 60 min in TBS‐Tw (20 mM Tris pH 7.4, 150 mM NaCl, and 0.2% Tween‐20) containing 5% skim milk powder and then incubated overnight at 4°C with primary antibody in TBS‐Tw containing 1% skim milk. After washing in TBS‐Tw, blots were incubated with horse radish peroxidase‐conjugated secondary antibodies and reactive bands revealed by chemiluminescence (Fisher Scientific).

Coimmunoprecipitation (co‐IP) of proteins from brain extracts was conducted as we have previously described (Li et al., [48]; Li, Olson, Lu, Kamasawa et al., [49]). In brief, clarified brain homogenates in IP buffer were precleared with protein‐A agarose beads and centrifuged. Supernatants were incubated overnight at 4°C with anti‐LNX1 ab200, followed by 1 hr incubation with protein‐A coated agarose beads. After centrifugation, beads were washed extensively with washing buffer (20 mM Tris‐HCl, pH 8.0, 150 mM NaCl and 0.5% NP‐40) and then incubated at 60°C for 20 min in SDS‐PAGE loading buffer containing 10% β‐mercaptoethanol. Samples were taken for western blotting as described above. Control samples were taken through the IP procedure with exclusion of primary antibody. For IP of ubiquitinated Cx36‐eCFP, cells cotransfected for 16–18 hr with Cx36‐eCFP, HA‐ubiquitin and with either LNX2‐FLAG, LNX2C51A‐FLAG, or FLAG‐vector were lysed in radioimmunoprecipitation buffer (RIPA; 50 mM Tris pH 8.0, 150 mM NaCl, 1.0% IGEPAL CA‐630, 0.5% sodium deoxycholate; modified from Takahashi et al., [72]) containing 2% SDS, protease inhibitors and 10 mM N‐ethylmaleimide. Samples were then briefly sonicated and clarified by centrifugation at 15,000 g for 10 min. In sample aliquots containing 200 μg total protein, the 2% SDS was brought down to 0.1% SDS using diluent RIPA buffer that lacked SDS, and samples were then precleared with protein‐G magnetic beads (Dynabeads, ThermoFisher) for 1 hr at 4°C. The precleared lysates were incubated overnight at 4°C with either 2 μg of anti‐eGFP or 2 μg of normal rabbit IgG (Santa Cruz) and then incubated for 1.5 hr with 50 μl of protein‐G magnetic beads. After extensive washing in RIPA buffer, beads were diluted with SDS‐PAGE loading buffer supplemented with 10% β‐mercaptoethanol and boiled for 10 min. Following separation of beads from IP material with magnetic force, IP material was electrophoresed by SDS‐PAGE using 10% gels made using a TGX Stain‐Free FastCast Acrylamide kit (Bio‐Rad), transferred to PVDF membranes using the semi‐dry Trans‐Blot‐Turbo system according to manufacturer's instructions (Bio‐Rad) and imaged with a Chemidoc MP imaging system (Bio‐Rad). Membranes were probed with antibodies using procedures described above. IP experiments showing accumulation of ubiquitinated Cx36‐eCFP in the presence of LNX2 and reduced accumulation in presence of LNX2C51A‐FLAG were repeated three times with similar results as shown in Figure 9a. Experiments showing effects of LNX2‐FLAG or LNX2C51A‐FLAG on Cx36‐eCFP detection or chloroquine and NH4Cl‐mediated recovery of Cx36‐eCFP detection in the presence of LNX2‐FLAG were repeated three times with similar results as shown in Figure 10.

For pull‐down assays, a bacterially expressed GST‐LNX1 fusion protein plasmid was used to investigate molecular interaction between Cx36 and LNX1. Fusion protein was prepared, and pull‐downs using this fusion protein and HeLa‐Cx36 cell lysates were performed as we have previously described (Li et al., [48]; Li, Olson, Lu, Kamasawa et al., [49]). To characterize interactions between Cx36 and the PDZ domains of LNX1 and LNX2, bacterial expression plasmids containing the separate His‐tagged PDZ domains of LNX1 or LNX2 in the pET30A vector were utilized (a gift from Dr. R. Hall, Emory University School of Medicine, Atlanta GA, USA; Fam et al., [17]). Plasmids were introduced into competent e‐coli BL21 (DE3) cells and fusion protein expression was induced with 1 mM isopropyl‐β‐d‐thiogalactopyranoside. Following bacterial cell lysis and clarification of the extract, His‐tagged fusion proteins were purified from the supernatant using Ni‐NTA resins according to manufacturer's instructions (Novagen, Madison, WI, USA) and used in pull‐down assays as described elsewhere (Li et al., [48]). Briefly, nickel beads linked to fusion proteins containing His‐tagged PDZ domains 1–4 of LNX1 or PDZ domains 1, 2, and 4 of LNX2 were incubated overnight with lysates of HeLa or N2A cells transfected with plasmid encoding either Cx36 or trCx36 lacking its four amino acid c‐terminal PDZ ligand. The beads were washed in PBS buffer containing 1% Triton X‐100 and proteins were eluted from beads by incubation for 15 min at 60°C with SDS‐PAGE sample buffer. The resulting pull‐down samples were then analysed by western blotting. Pull‐down assays as described above for the PDZ domains of LNX1 and LNX2 were repeated three times with similar results as shown in Figures 6 and 7.

To confirm loss of LNX1 proteins in gene‐targeted mice, tissue homogenates from adult LNX1 null and WT controls were taken for immunoblotting and immunoprecipitation. Tissues were collected and homogenized in radioimmunoprecipitation buffer. Total protein concentration of homogenates was determined, 50 μg of protein was separated by SDS‐PAGE, immunoblotted as described above and blots probed with anti‐LNX1 ab8 antibody. Immunoprecipitation of LNX1 from tissue homogenates was conducted using anti‐LNX1 ab365 antibody or normal rabbit IgG. Immunoprecipitates were immunoblotted and probed with anti‐LNX1 ab8.

Immunofluorescence labelling of LNX1 in combination with labelling for Cx36 was examined in the inferior olive of mouse and rat, and in the retinal inner plexiform layer of mouse, two areas where neuronal gap junctions containing Cx36 are highly concentrated (Kamasawa et al., [32]; Li, Olson, Lu, Kamasawa et al., [49]). As previously described, immunolabelling for Cx36 appears exclusively punctate throughout the CNS, and ultrastructural studies involving freeze‐fracture replica immunogold labelling (FRIL) have established the localization at neuronal gap junctions (Nagy, Dudek, & Rash, [55]; Rash, Olson, Davidson, et al., [65]; Rash, Olson, Pouliot, et al., [66]). Because intracellular Cx36 in neurons in vivo remains undetectable with currently available anti‐Cx36 antibodies for reasons as yet undetermined (Bautista & Nagy, [3]; Nagy, Bautista, Blakley, & Rash, [54]; Rubio & Nagy, [69]), immunofluorescence localization of this connexin is considered to reveal sites in vivo at which it is contained in gap junctions (Nagy, Pereda, & Rash, [57]; Nagy et al., [58]). The typical punctate appearance of Cx36 labelling at neuronal gap junctions is shown in the mouse inferior olive in Figure 1a1, where we refer to this labelling as Cx36‐puncta. Immunofluorescence labelling of LNX1 was also largely punctate in this structure, hence referred to as LNX1‐puncta, and had a distribution very similar to that of Cx36 (Figure 1a2). Double immunofluorescence labelling of Cx36 in combination with labelling of LNX1 showed that in overlay images many Cx36‐puncta were colocalized with LNX1‐puncta (Figure 1a3), indicating localization of LNX1 to neuronal gap junctions. Conversely, not all LNX1‐puncta were associated with labelling for Cx36. Where colocalization occurred, labelling for LNX1 was often confined entirely within an areas occupied by individual Cx36‐puncta, and sometimes LNX1 was curiously confined to small subregions of those puncta.

A control for specificity of LNX1 detection and LNX1/Cx36 colocalization was provided by the use of two LNX1 antibodies directed against different, nonoverlapping sequences within this protein. Labelling patterns obtained in the inferior olive of mouse with anti‐LNX1 ab247 (Figure 1b) was similar to that obtained with anti‐LNX1 ab200 (Figure 1c), and both gave similar colocalization with Cx36 (Figure 1b,c). Examination of labelling in genetically ablated LNX1‐deficient mice (LNX1 KO mice) was used as a further control for localization of LNX1 at Cx36‐containing gap junctions. In the inferior olive of these KO mice, labelling with both anti‐LNX1 ab247 and ab200 was totally absent at Cx36‐puncta (Figure 1d,e). However, both antibodies produced residual punctate labelling, which appear to represent cross‐reaction with another as yet unidentified protein, perhaps a closely related member of the family of LNX proteins (Katoh & Katoh, [34]). Counts of immunopositive puncta in 10–14 fields double‐labelled for Cx36 and LNX1 revealed that 50 ± 3.2% and 42 ± 3.2% (mean ± SEM of the number of field examined) of Cx36‐puncta were colocalized with LNX1‐puncta labelled with anti‐LNX1 ab200 and ab247, respectively, in sections of inferior olive from WT mice, while this was reduced to 2 ± 1.2% and 1 ± 0.4% Cx36‐puncta colocalization with LNX1‐puncta using these antibodies for labelling in sections from LNX1 KO mice. Immunofluorescence labelling of Cx36 in combination with immunolabelling of LNX1 in the inner plexiform layer of mouse retina (Figure 2a) and in the inferior olive of rat (Figure 2b) gave similar results as in the inferior olive of mouse. In both of these CNS regions, many Cx36‐puncta displayed colocalization with LNX1 and, conversely, some LNX1‐puncta in both areas showed an absence of association with Cx36‐puncta.

We next examined double immunofluorescence labelling for Cx36 in conjunction with labelling for LNX2 in the inferior olive of mouse. Throughout this structure, where Cx36‐puncta are abundantly found (Figure 3a1), labelling of LNX2 had a closely matching distribution, and was also evident exclusively as immunoreactive puncta (Figure 3a2), but the density of LNX2‐puncta was somewhat less than that of Cx36‐puncta (Figure 3a3). Consistent with this observation and in contrast to results concerning LNX1, it appeared that the vast majority of LNX2‐puncta were colocalized with Cx36 but, conversely, not all Cx36‐puncta were associated with labelling for LNX2 (Figure 3b,c). The degree of colocalization of Cx36 with LNX2 at individual puncta exhibited a variety of patterns. Where colocalization occurred, labelling for LNX2 either entirely encompassed Cx36‐puncta and extended beyond the area occupied by Cx36‐puncta, or was confined entirely within Cx36‐puncta, occupying large to very small spatial subregions of the area circumscribed by Cx36‐puncta (Figure 3b,c). Also frequently observed was partial overlap of labelling for Cx36 and LNX2 at individual puncta (Figure 3b,c).

Recognition of LNX1 with the rabbit polyclonal anti‐LNX1 ab200 antibody we generated was demonstrated by immunoblotting of HeLa cell lysates derived from transiently transfected cells expressing LNX1p70 or LNX1p80‐FLAG. In these lysates, bands were detected migrating at ~70 kDa and ~80 kDa (Figure 4a, lanes 1 and 2), which corresponded to apparent molecular weights of LNX1p70 and LNX1p80‐FLAG respectively. These bands were absent in nontransfected WT cells (HeLa‐WT; Figure 4a, lane 3). Probing the same lysates with anti‐FLAG revealed detection of the LNX1p80‐FLAG, corresponding to this LNX1 isoform recognized by anti‐LNX1, but not the non‐FLAG tagged LNX1p70 isoform (Figure 4a, lanes 4 and 5, respectively). Comparison of results with anti‐LNX1 ab200 versus ab247 after immunoblotting of these same lysates is shown in Figure 4b, which indicated that each of these antibodies have the capacity to detect both the full‐length isoform LNX1p80 and the shorter isoform LNX1p70. Immunoblotting of LNX1 and LNX2 in homogenates of brain is not shown due to difficulties in detection of these LNX isoforms in brain extracts, as also reported by others (Lenihan, Saha, Heimer‐McGinn, et al., [43]).

The molecular association of LNX1 with Cx36 was initially examined by co‐IP procedures. Homogenates of mouse brain taken for IP with anti‐LNX1 showed detection of Cx36 in IP material probed with anti‐Cx36 (Figure 4c, lane 3). The migration profile of the band detected in the IP material corresponded to that of Cx36 in lysates of HeLa cells stably expressing Cx36 (HeLa‐Cx36) (Figure 4c, lane 2), and this band was absent after IP with omission of anti‐LNX1 (Figure 4c, lane 4). Specificity of the band identified as Cx36 is indicated by lack of its detection in a control lane loaded with lysate from nontransfected HeLa‐WT cells (Figure 4, lane 1). In an initial pull‐down experiment, we examined the interaction of Cx36 with bacterially expressed LNX1‐GST fusion protein. Immunoblots of lysates of HeLa‐Cx36 cells taken for pull‐down with LNX1‐GST and probed with anti‐Cx36 showed detection of a band (Figure 4d, lane 2) that comigrated with Cx36 seen in input lysate from HeLa‐Cx36 cells (Figure 4d, lane 1). Further effort to examine interaction of LNX proteins with Cx36 was focused not using brain tissue, but rather transfection of cells in culture.

To confirm that LNX1 gene‐targeted mice were deficient for LNX1, tissue homogenates from these mice and their WT counterparts were taken for immunoblotting and immunoprecipitation using LNX1 antibodies raised against distinct regions of LNX1. As expected, anti‐LNX1 ab8 antibody detected LNX1p80 in WT tissues, but not in tissues from LNX1 KO mice (Figure 5a). Lack of detection of LNX1p80 in lanes loaded with homogenate from WT small intestine is likely attributable to an insufficient amount of protein loading as indicated by only a small fraction of actin detected in lanes loaded with this tissue as compared to lanes loaded with other samples. Nonetheless, results for kidney, liver, caecum, and colon provided evidence that LNX1 p80 protein was present in WT tissues and absent in the tissues of LNX1 KO mice. In blots of material obtained after IP with anti‐LNX1 ab365 and probed with anti‐LNX1 ab8 (Figure 5b), the brain‐specific LNX1p70 isoform was detected in IP material from WT brain, while the p80 isoform was not detected (Figure 5b, lane 1). In contrast, IP material from peripheral tissue homogenates (heart, colon, and kidney) showed the presence of the p80 isoform and absence of the p70 isoform (Figure 5b, lanes 3, 5, 7). After IP from tissues of LNX1 KO mice, there was an absence of LNX1 detection (Figure 5b, lanes 2, 4, 6, 8), confirming ablation of the LNX1 in these mice and providing confidence in the identity of LNX1 bands seen in blots of IP material from WT tissues. Positive control IPs showed detection of LNX1p70 (Figure 5b, lane 9) and LNX1p80 (Figure 5b, lane 10) in IP material from lysates of LNX1p70‐ or LNX1p80‐transfected HEK293T cells respectively. A band migrating slightly faster than the LNX1p70 isoform seen only in IP material from WT brain, and which like the p70 isoform was absent after IP from brain of LNX1 KO mice, may represent the p62 isoform of LNX1 that was reported to be uniquely expressed in brain (Lenihan et al., [44]). Intense bands running at an even lower molecular weight that were detected in WT colon and kidney (Figure 5, lanes 5 and 7, respectively) and that were absent in LNX1 KO mice may represent degradation products of LNX1.

To more comprehensively examine molecular interactions between LNX proteins and Cx36 in vitro, bacterial fusion proteins expressed from recombinant plasmids containing His‐tagged PDZ domains of LNX1 and LNX2 were used for pull‐down assays. Pull‐downs involved incubation of the His‐PDZ domains with lysates of HeLa or N2A cells that had been transfected with either full‐length Cx36 or truncated Cx36 (trCx36) lacking the C‐terminal four amino acids that constitute the PDZ ligand. Immunoblots loaded with positive control input lysates and probed with anti‐Cx36 showed detection of Cx36 and trCx36 in material used for pull‐downs with LNX1 PDZ domains (Figure 6a, lane 1; and 6C, lanes 2,3) or for pull‐downs with LNX2 PDZ domains (Figure 7, lanes 1, 5). The Cx36 band was absent in HeLa‐WT cells (Figure 6a, lane 2 and 6C, lane 1) as well as in N2A‐WT cells (not shown). Full‐length Cx36 was detected after pull‐down with the second PDZ domain of both LNX1 (Figure 6a, lane 4) and LNX2 (Figure 7a, lane 3), but was absent in pull‐downs using the PDZ1, PDZ3, and PDZ4 domains of LNX1 (Figure 6a, lanes 3, 5 and 6, respectively) or the PDZ1 and PDZ4 domains of LNX2 (Figure 7a, lanes 2 and 4, respectively). This molecular interaction between Cx36 and the PDZ2 domains of the two LNX proteins was abrogated when pull‐downs were conducted with trCx36 and either the PDZ2 domain of LNX1 (Figure 6c, lane 5) or the PDZ2 domain of LNX2 (Figure 7a, lane 6), indicating requirement of the Cx36 PDZ ligand for interaction of Cx36 with LNX1 or LNX2. Blots probed for pull‐down material were stripped and reprobed with monoclonal anti‐His (Figure 6b, lanes 7–12; Figure 6d, lanes 6 and 7, which corresponded to lanes 4 and 5 in Figure 6c; and Figure 7b, lanes 7–12). These blots probed with anti‐His showed approximately equivalent amounts of His‐PDZ domain fusion proteins were used in the pull‐down assays.

To examine the effect of LNX expression on Cx36 localization in cultured cells, N2A cells were cotransfected with Cx36‐eCFP (Figure 8) and with either LNX1p80‐FLAG, LNX1p70 that lacks E3 ligase activity, LNX2‐FLAG, or with the functional mutant form of LNX2 (LNX2C51A‐FLAG) that has a mutated zinc coordination RING domain essential for ligase activity. When transfected alone into cells, the intrinsic fluorescence of Cx36‐eCFP was typically distributed at both intercellular gap junctions and to punctate intracellular structures (Figure 8a). Cells transfected with Cx36‐eCFP (Figure 8b1) and immunolabelled with anti‐Cx36 (Figure 8b2), showed colocalization of eCFP fluorescence and Cx36 immunofluorescence (Figure 8b3), providing confidence in specificity of Cx36 detection in cultured cells. Cells cotransfected with Cx36‐eCFP and LNX1p70 (Figure 8c1) and simultaneously probed with anti‐Cx36 (Figure 8c2) and anti‐LNX1 (Figure 8c3) displayed colocalization of eCFP fluorescence, immunolabelling of Cx36, and immunolabelling of LNX1p70 at large gap junctions located at intercellular contacts and at intracellular punctate structures (Figure 8c4). In contrast, cells cotransfected with Cx36‐eCFP (Figure 8d1) and LNX1p80‐FLAG (Figure 8d3), and probed with anti‐Cx36 (Figure 8d2) showed a loss of signals for eCFP and Cx36 at cell–cell contacts, with retention of only a fractional amount of intracellular Cx36‐eCFP fluorescence. Cells expressing LNX1p80‐FLAG (Figure 8e) taken for double‐labelling with anti‐FLAG (Figure 8e1) and anti‐LNX1 (Figure 8e2) showed a similar distribution of intracellular immunofluorescence for the FLAG tag and LNX1, indicating fidelity for LNX1p80‐FLAG detection and confirming the diffuse nature of LNX1p80 immunoreactivity in these cells.

In cells expressing Cx36‐eCFP (Figure 8f1) and LNX2‐FLAG (Figure 8f2), and probed with anti‐FLAG, there was an absence of Cx36‐eCFP at cell–cell contacts suggesting lack of gap junctions, while intracellular Cx36‐eCFP was retained, but at diminished levels. To determine if absence of Cx36‐eCFP‐containing gap junctions was dependent on the E3 ligase activity of LNX2, N2A cells were cotransfected with Cx36‐eCFP and with mutant LNX2C51A‐FLAG. In these cells, Cx36‐eCFP fluorescence was maintained at cell–cell contacts presumably reflecting gap junctions, and appeared at intense levels intracellularly (Figure 8g1), and FLAG immunoreactivity for LNX2C51A‐FLAG was seen colocalized with eCFP (Figure 8g2).

To determine if Cx36 is targeted for ubiquitination by LNX2 and if so whether ubiquitination of Cx36‐eCFP by LNX2 required a functional RING domain of LNX2, lysates of cells expressing Cx36‐eCFP and HA‐ubiquitin together with either LNX2‐FLAG, LNX2C51A‐FLAG, or FLAG vector were taken for IP of Cx36‐eCFP, immunoblotted, and probed with anti‐HA for detection of HA‐ubiquitin conjugated to Cx36‐eCFP in the IP material (Figure 9).

The results indicated increased HA‐ubiquitinated material present in IP material gathered from cells expressing LNX2‐FLAG (Figure 9a, lane 2) versus those expressing the empty FLAG‐vector (Figure 9a, lane 1), and the increase was not seen in Cx36‐eCFP IP material from cells expressing the LNX2C51A‐FLAG mutant that lacks a functional RING domain (Figure 9a, lane 3). These results suggest that Cx36‐eCFP is subject to ubiquitination mediated by LNX2, which is consistent with findings by others in studies of various targets of LNX E3 ligases, including claudins, KIF7 (kinesin‐like protein), ERC2 (ERC protein2) and SRGAP2 (SLIT‐ROBO Rho GTPase‐activating protein 2), where ubiquitination by LNX was shown to require a functional RING domain (Lenihan et al., [44]; Takahashi et al., [72]). The lack of HA detection in control IPs conducted with normal rabbit IgG and lysates of the same LNX2‐FLAG–transfected cells used in lane 2 (Figure 9a, lane 4) indicated that HA detection in IPs performed with anti‐eGFP antibody was not due to nonspecific binding of HA‐ubiquitinated proteins to Protein G‐conjugated magnetic beads or to IgG. This latter observation suggests that our detection of baseline levels of HA‐ubiquitinated Cx36‐eCFP in IPs of FLAG‐vector cotransfected cells was potentially due to either a small degree of nonspecific IP of proteins with anti‐eGFP or to the activity of an as yet unidentified endogenous E3 ubiquitin ligase on Cx36‐eCFP in N2A cells. This latter possibility is consistent with observations concerning other proteins that are acted upon by multiple E3 ubiquitin ligases, including p53 and Cx43 (Basheer et al., [1]; Chen, Kristensen, Foster, & Naus, [9]; Fykerud et al., [22]; Satija, Bhardwaj, & Das, [70]). However, we cannot at present exclude the possibility that LNX2 interacts with and mediates the ubiquitination of additional proteins that directly or indirectly associate with Cx36 and are localized at neuronal gap junctions.

We next determined if lysosomal inhibition with chloroquine or ammonium chloride could rescue the loss of Cx36‐eCFP in cells overexpressing LNX2. This was examined by immunoblotting of lysates from cells cotransfected with Cx36‐eCFP (Figure 10a, lanes 1–7) and either LNX2‐FLAG, LNX2C51A‐FLAG, or empty FLAG vector, followed by probing of immunoblots with anti‐Cx36 ab51‐6300. In cells transfected with Cx36‐eCFP and cotransfected with FLAG‐vector, anti‐Cx36 recognized a band migrating ~60 kDa, close to the calculated molecular weight of Cx36‐eCFP (Figure 10a, lane 1), which was absent in cells transfected with FLAG‐vector only (Figure 10a, lane 8). The Cx36‐eCFP band was only weakly detected in cells expressing LNX2‐FLAG (Figure 10a, lane 2), whereas the robust expression of Cx36‐eCFP was retained in cells expressing the LNX2C51A mutant (Figure 10a, lane 3). When cells expressing Cx36‐eCFP and LNX2 were treated with chloroquine (Figure 10a, lanes 4–6; 10, 50, 100 mM, respectively) or 25 mM ammonium chloride (Figure 10a, lane 7), the intense band for Cx36‐eCFP was recovered. After stripping, the immunoblot shown in Figure 10a was reprobed with anti‐FLAG, which showed roughly equal abundance of LNX2‐FLAG in all lanes loaded with lysates from cells cotransfected with Cx36‐eCFP and LNX2‐FLAG (Figure 10b, lanes 10, 12–15), as well as in lysate from cells cotransfected with Cx36‐eCFP and LNX2C51A‐FLAG (Figure 10b, lane 11), and its absence in cells transfected with Cx36‐eCFP and FLAG‐vector (Figure 10b, lane 9), or FLAG‐vector only (Figure 10b, lane 16). These results indicate that the ultimate fate of Cx36‐eCFP in cells overexpressing LNX2‐FLAG was its targeting to the lysosomal compartment for degradation.

The present results add Cx36 to the list of proteins that interact with LNX1 and/or LNX2 (LNX1/2) and that may serve as targets for ubiquitination by one or the other of these E3 ubiquitin ligases. The cellular distribution and subcellular localization of LNX1/2 proteins in the CNS has not as yet been subject to comprehensive documentation, owing perhaps to their apparently low levels of expression in neural tissue (Lenihan et al., [44]), and difficulties in their detection with currently available anti‐LNX antibodies. LNX1 and LNX2 mRNAs have, however, been localized by in situ hybridization primarily in neurons in brain (Lenihan et al., [44]; Rice et al., [68]), and the present immunofluroescence results indicate that at least some of those neurons correspond to neuronal populations that express Cx36. The localization of LNX1/2 specifically at sites of Cx36‐puncta in the brain regions examined indicates with reasonable certainty that these E3 ligases in part reside at Cx36‐containing gap junctions that form electrical synapses, because (as noted earlier) immunofluorescence labelling of Cx36 with the antibodies and immunolabelling protocols used here appears to be restricted exclusively to cellular sites at which this connexin forms gap junctions (Nagy et al., [55], [57], [58]; Rash, Yasumura, Dudek, & Nagy, [67]; Rash, Olson, Davidson, et al., [65]; Rash, Olson, Pouliot, et al., [66]). Although anti‐LNX1 detects both p70 and p80 as shown here in transfected cells, it likely recognized LNX1p70 at these junctions given the absence of LNX1p80 expression in brain (Lenihan et al., [44]; Rice et al., [68]). We examined LNX1/2 association with Cx36 in only a few limited brain regions that harbour the highest density of Cx36‐containing gap junctions, namely the retina and inferior olivary nucleus. However, these junctions are known to be widely distributed in brain and spinal cord (Bautista, McCrea, & Nagy, [2]; Condorelli et al., [10]; Nagy et al., [58]). Thus, it remains to be determined whether LNX1/2 proteins are similarly associated with Cx36 in other regions of the CNS.

We have previously reported the immunofluorescence localization of several other proteins at Cx36‐containing gap junctions in adult rat and mouse brain, including ZO‐1, AF6 (aka, afadin), cingulin, and MUPP1 (Li et al., [48]; Li, Olson, Lu, Kamasawa et al., [49]; Lynn et al., [52]). Labelling for some of these proteins (e.g., ZO‐1, cingulin) often showed total overlap of labelling with Cx36 at individual Cx36‐puncta. Here, a remarkable heterogeneity was seen in the degree of overlap between labelling for LNX proteins and Cx36 at Cx36‐puncta. In cases where labelling for LNX1/2 extended beyond the bounds of Cx36‐puncta, LNX1/2 outside of, but contiguous with, these puncta could be associated with plasma membrane structures immediately adjacent to Cx36‐puncta (i.e., adherens junctions, Nagy & Lynn, [56]). Indeed, protein components of other types of cell–cell junctions, including E‐cadherin at adherens junctions and claudin‐1 plus occludin at tight junctions, are targets for ubiquitination by RING‐ and Hect‐type E3 ubiquitin ligases (Fujita et al., [20]; Takahashi et al., [72]; Traweger et al., [73]), raising the possibility that one or more of the many components of adherens junctions that flank Cx36‐containng gap junctions may represent binding partners of LNX1/2. Labelling of LNX1/2 entirely confined within and occupying variable tiny to large portions of Cx36‐puncta suggest that actions of LNX1/2 may be operative within subdomains of Cx36‐containing gap junctions. If LNX1/2 contributes to processes governing turnover of Cx36, then these subdomains may correspond to sites within Cx36‐containing gap junction plaques where a portion of this connexin may undergo removal by internalization, as has been reported to occur at gap junctions composed of other connexins (Flores et al., [18]; Gaietta et al., [23]; Laird, [37]; Lauf et al., [38]). It is noteworthy in this context that ultrastructural studies have demonstrated endocytic annular profiles to be associated also with subdomains of Cx36‐containing gap junctions between neurons in mammalian brain, specifically parvalbumin‐expressing neurons in the striatum, and these profiles were interpreted to represent portions of gap junctions undergoing internalization (Fukuda, [21]), which interestingly may correspond to subdomains of Cx36‐puncta found to harbour LNX1/2. In view of these considerations, it may be of interest to explore the characteristics of Cx36 turnover, morphological features, and deployment of Cx36‐containing gap junctions in the CNS of LNX1/2 KO mice.

Studies on the turnover, recycling and degradation of gap junctions have focused mainly on Cx43, which undergoes processing by both the lysosomal and proteasomal pathways (Leithe, [39]), with Cx43 degradation regulated by both its phosphorylation and by ubiquitination via the E3 ubiquitin ligase Nedd4, which colocalized with Cx43 at gap junctions (Leithe, Brech, & Rivedal, [40]; Leithe & Rivedal, [41], [42]; Leykauf et al., [46]). These findings suggested that Cx43 was ubiquitinated at gap junctions within the plasma membrane, which signalled Cx43 for ultimate proteasomal degradation. Subsequently, Cx43 was found to undergo ubiquitination by other E3 ubiquitin ligases (Basheer et al., [1]; Chen et al., [9]; Fykerud et al., [22]), and additional connexins were reported to be ubiquitinated, including connexin26, connexin32, and connexin40 (Leithe, [39]). We now add Cx36 to this list, which expands the growing number of substrates and/or binding partners of LNX1/2 proteins that were identified by biochemical or proteomic methods and by strategies that combined biochemical approaches with bioinformatic analysis (Guo et al., [25]; Song et al., [71]; Wolting et al., [76]).

Our finding that the classical C‐terminal four amino acid PDZ ligand of Cx36 (i.e., SAYV), which is required for interaction of Cx36 with PDZ domains in AF6, MUPP1, and of ZO‐1 (Li et al., [48]; Li, Olson, Lu, Kamasawa et al., [49]; Li et al., [47]), mediates Cx36 association with the PDZ2 domains of both LNX1 and LNX2 is consistent with the highly homologous sequence of these two LNX proteins, particularly in regions that define their substrate specificity (Flynn, Saha, & Young, [19]; Rice et al., [68]). Based on our results from transfected cells in culture, it is likely that association of LNX1/2 with electrical synapses in brain arises by virtue of Cx36 PDZ ligand interaction with the PDZ2 domains in LNX1/2. This was supported by colocalization of Cx36‐containing gap junctions with both the catalytically inactive LNX1p70 isoform and LNX2 isoform bearing an E3‐ubiquitin ligase loss‐of‐function mutation in transfected cells. These results suggested LNX1/2 targeting to Cx36 at the plasma membrane, similar to such targeting of the E3 ubiquitin ligase Nedd4 to gap junctions composed of Cx43 (Leithe, [39]). Colocalization of Cx36 with LNX1p80 or LNX2 in transfected cells was not observed due to the loss of Cx36 at cell–cell contacts, which was interpreted to indicate a potential ubiquitin‐related mechanism for degradation of Cx36 by these E3 ligase‐competent isoforms of LNX1/2, as supported by the observed retention of Cx36‐containing gap junctions in cells expressing the LNX2 loss of function mutant. Consistent with this idea was that ubiquitination of Cx36‐eCFP in the presence of LNX2 was not observed in cells transfected with the LNX2C51A mutant. Furthermore, the observed loss of Cx36‐eCFP in cells expressing LNX2 was not seen in cells expressing mutant LNX2 lacking a functional RING domain. In addition, the loss of Cx36‐eCFP detection seen in the presence of LNX2 was abrogated by treatment of LNX2‐expressing cells with inhibitors of lysosomal degradation, suggesting that one mechanism for ubiquitinated Cx36 degradation is via the lysosomal pathway, similar to processing of Cx43 after its Nedd1‐mediated ubiquitination or to the ubiquitination and degradation of claudins in MDCK cells overexpressing LNX1, and recovery of claudin detection in these cells by treatment with lysosome inhibitors (Leithe, [39]; Takahashi et al., [72]).

Results concerning LNX1p80, although consistent with those involving LNX2, are perhaps less physiologically relevant given the absence of LNX1p80 expression in adult brain as seen here and reported by others (Lenihan et al., [44]; Rice et al., [68]). In this regard, it is noteworthy that both LNX1 and LNX2 contain, in addition to four unique PDZ domains, classical C‐terminal PDZ ligand motifs, and LNX1 was shown to interact with LNX2 and has a similar cell‐type expression pattern and regional distribution as LNX2 (Lenihan et al., [44]; Rice et al., [68]). It has been suggested that the observed self‐binding of the PDZ1 domains of LNX1/2 with their own C‐terminal PDZ ligands may serve to inhibit interaction with target PDZ ligands, providing a mechanism for self‐regulating their actions on target proteins (Rice et al., [68]). Our results showing the presence of both LNX1/2 at Cx36‐containing neuronal gap junctions, together with our findings and reports that only LNX1 protein isoforms that lack E3 ligase activity are expressed in brain (Lenihan et al., [44]; Rice et al., [68]), suggest that LNX1 at electrical synapses may confer stability on these synapses via PDZ‐mediated regulatory interactions with LNX2 that inhibit the E3 ligase activity of LNX2 on Cx36. Alternatively, the association of E3 ligase‐incompetent LNX1p70 with Cx36 may moderate, in a competitive manner, the availability of Cx36 for interaction with ligase‐competent LNX2.

Sequence analyses have revealed that the LNX1/2 genes are ancestrally related to the MUPP1 that contains 13 PDZ domains, the last four (PDZ 10‐13) of which share similar splice junctions and primary sequence homology with PDZ1‐4 domains of LNX1/2 (Flynn et al., [19]). Computed potential PDZ interactions (Hui & Bader, [31]) revealed that almost all proteins predicted to interact with the PDZ1 or PDZ2 domains of LNX1/2 do interact with either PDZ10 or PDZ11 of MUPP1 (Flynn et al., [19]). It was shown empirically that interactors with MUPP1 PDZ10‐13, including the synaptic proteins SynGap1 (synaptic GTPase activating protein) and HTR2C (Serotonin receptor 2C subunit), also interacted with the PDZ1 and two domains of LNX1/2 (Flynn et al., [19]). These predictive/empirical findings are supported by our previous report on the interaction of Cx36 with the 10th PDZ domain of MUPP1 (Li et al., [48]) taken together with the present demonstration of Cx36 interaction with the PDZ2 domains of LNX1/2.

Mechanisms underlying the dynamic nature and reported plasticity of electrical synapses formed by Cx36‐containing neuronal gap junctions may be diverse, and the biochemical correlates of these mechanisms may be commensurate with an equally diverse composition of structural and regulatory proteins associated with these synapses. Thus, in addition to various previously reported cellular processes that could potentially influence the efficacy of electrical synaptic transmission (Nagy et al., [58]; Pereda, [63]; Pereda et al., [64]), the actions of LNX1/2 at electrical synapses composed of Cx36 could in part provide the basis for the steady‐state turnover of Cx36 at gap junctions. These LNX isoforms may also govern rapid, moment‐to‐moment trafficking of Cx36‐containing channels at these junctions, thereby regulating the availability of functional channels for mediation of electrical coupling.

This work was supported by grants from the Canadian Institutes of Health Research and the Canadian Natural Sciences and Engineering Research Council to J.I. Nagy and from both of these agencies to C.J. McGlade. We thank Dr. R. Hall for bacterial expression plasmids containing the His‐tagged PDZ domains of LNX1 and LNX2, and B. McLean for excellent technical assistance.

The authors declare no competing financial interests.

The raw data will not be deposited in a public repository due to the complexity of its assembly, but it can be obtained upon request to the corresponding author.

B.D.L. conducted the molecular, IP, pull‐down, and immunofluorescence studies involving cell cultures; X.L. conducted IP and pull‐down of Cx36 with LNX1 using brain tissues, J.I.N. was responsible for the immunofluorescence work involving brain sections; S.G.H. contributed the Cx36‐eCFP for transfection approaches and design of the work involving this construct; C.J.M. and E.K.G. generated the LNX1 C57BL/6 null mice, LNX1 and LNX2 expression constructs, the full‐length LNX1‐GST fusion protein construct, and LNX1 antibodies used for immunoprecipitation; and B.D.L., X.L. and J.I.N. participated in data analysis, figure compilation, and wrote the manuscript, with final critical evaluation of data interpretation, intellectual content, and manuscript approval provided by C.J.M. and S.G.H.

GRAPH: supinfo/ejn14198-sup-0001-reviewer_Comments.docx