Amyloid-β regulates gap junction protein Connexin 43 trafficking in cultured primary astrocytes

Altered expression and function of astroglial gap junction protein Connexin 43 (Cx43) has increasingly been associated to neurotoxicity in Alzheimer disease (AD). While earlier studies have examined the effect of increased amyloid-β (Aβ) on Cx43 expression and function leading to neuronal damage, underlying mechanisms by which Aβ modulates Cx43 in astrocytes remain elusive. Here, using mouse primary astrocyte cultures, we have examined the cellular processes by which Aβ can alter Cx43 gap junctions. We show that Aβ 25-35 impairs functional gap junction coupling yet increases hemichannel activity. Interestingly, Aβ 25-35 increased the intracellular pool of Cx43 with a parallel decrease in gap junction assembly at the surface. Intracellular Cx43 was found to be partly retained in the endoplasmic reticulum-associated cell compartments. However, forward trafficking of the newly synthesized Cx43 that already reached the Golgi was not affected in Aβ 25-35 exposed astrocytes. Supporting this, treatment with 4-phenylbutyrate, a well-known chemical chaperone that improves trafficking of several transmembrane proteins, restored Aβ-induced impaired gap junction coupling between astrocytes. We further show that interruption of Cx43 endocytosis in Aβ 25-35 exposed astrocytes resulted in their retention at the cell surface in the form of functional gap junctions indicating that Aβ 25-35 causes rapid internalization of Cx43 gap junctions. Additionally, in silico molecular docking suggests that Aβ can bind favorably to Cx43. Our study thus provides novel insights into the cellular mechanisms by which Aβ modulates Cx43 function in astrocytes, the basic understanding of which is vital for the development of alternative therapeutic strategy targeting connexin channels in AD.


Introduction
Connexin 43 (Cx43) is a predominant gap junction protein that underlies astroglial networks essential for maintaining brain homeostasis (1)(2)(3)(4)(5)(6). Gap junctions are arrays of a few tens to thousands of cell-to-cell channels, termed as gap junction channels that allow direct intercellular exchange of small molecules such as ions (Ca +2 , K + , Na + ), amino acids (glutamate), second messengers (ATP, cAMP, IP 3 ) and metabolites (glutathione, glucose) up to approximately 1kD in size (7,8). A single gap junction channel comprises of two opposing channels, called hemichannels or connexons, formed by oligomerization of six connexin proteins.
Under normal conditions, hemichannels exhibit a low probability of opening and have been proposed to mediate physiologic release of gliotransmitters in the extracellular medium (9,10).
Accumulating evidence suggests that altered functioning of gap junctions and hemichannels could be related to the onset and progression of homeostatic imbalances observed in various neurodegenerative diseases (11,12). In fact, loss of astrocytic gap junctions in mice has been shown to cause profound neurological phenotypes including widespread dysmyelination and hippocampal CA1 vacuolation (13). In this context, several recent studies have revealed an association between astroglial connexins and the neurodegenerative phenotype observed in Alzheimer's disease (AD), the most prevalent cause of dementia largely affecting the elderly population in our society (14,15). Accumulation of amyloid-β (Aβ) peptides and reactive astrogliosis are among the pathological features that characterise all AD brains with Aβ peptides widely believed to drive AD neuropathology (16,17). Although Aβ neurotoxicity involves activation of NMDA receptors, sustained elevations of Ca 2+ , mitochondrial dysfunction, oxidative stress and immune activation (18), complete understanding of the underlying mechanisms associated with AD pathology remain elusive till date.
Interestingly, several recent studies indicate that astroglial Cx43 channel functions, both gap junctions as well as hemichannels, are modified in AD. Studies in post-mortem human AD brains and mouse models of AD have shown increased immunoreactivity of Cx43 around Aβ plaques (19)(20)(21). In cultured rat astrocytes, Aβ  treatment was reported to impair gap junction communication (22). However, the extent of gap junction coupling in astrocytic networks in brain slices of AD mouse models was found to be either maintained or reduced depending on the brain area and age of the animal (21,23). Recently, it was shown that treatment with Aβ [25][26][27][28][29][30][31][32][33][34][35] increased Cx43 hemichannel activity in astrocytes releasing ATP and glutamate subsequently leading to neuronal damage (24). Supporting this, inhibition of glial hemichannels prevented the inflammatory profile evoked by Aβ in astrocytes and reduced neuronal damage in hippocampal slices exposed to Aβ as well as in AD mouse model (25,26). Further, astroglial targeted connexin43 knock-out in a mouse model of AD lessened neuronal damages and improved cognitive function (21,27). Altogether, these earlier studies suggest that Aβ alters Cx43 function in astrocytes leading to neurotoxicity and that Cx43 could be a potential therapeutic target for AD. However, the underlying mechanisms by which Aβ modulates Cx43 function in astrocytes remain largely unknown.

Aβ increases intracellular Cx43 expression without altering its steady state levels in mouse astrocytes
To determine if the altered functioning of gap junctions and hemichannels was because of a change in steady-state Cx43 levels, we examined Cx43 mRNA and total protein levels. Our quantitative PCR analysis showed no significant alterations in Cx43 mRNA levels in mouse primary astrocytes treated with 10μM Aβ 25-35 for 24, 48 and 72h compared to control astrocytes ( Figure 2A). Consistent with the mRNA levels, Cx43 total protein expression also did not reveal any significant alterations in mouse astrocyte cultures subjected to 10μM Aβ 25-35 for 24, 48 and 72h compared to control astrocytes ( Figure 2B). Antibody directed against Cx43 detected three bands at around 43 kDa, representing a faster migrating non-phosphorylated form (P0) and two slower migrating phosphorylated forms (P1, P2) (31). No significant difference was noted in the steady-state levels of the phospho-bands in Aβ 25-35 treated versus control astrocytes. Thus, Cx43 steady-state levels were not found to be significantly altered in mouse primary astrocyte cultures treated with Aβ [25][26][27][28][29][30][31][32][33][34][35] . To examine if the observed changes in functional gap junction communication and hemichannel activity was because of changes in cellular distribution of Cx43, we determined cellular expression pattern of Cx43 by immnolabelling studies ( Figure  2C,D). While control astrocytes showed mostly an organized distribution with lined up Cx43 puncta typically representing the cell-cell junctions ( Figure 2C), mouse astrocytes treated with 10μM Aβ [25][26][27][28][29][30][31][32][33][34][35] for 72h showed Cx43 labelling scattered throughout the cell cytoplasm ( Figure 2D). Interestingly, in many places Cx43 labelling was also found to be accumulated near the perinuclear region suggesting an increase in intracellular Cx43 levels ( Figure 2D, arrowheads). Additionally, astrocytes showed heterogeneous Cx43 labelling with punctate structures of varied sizes. While larger puncta might correspond to gap junction plaques at the cell-cell interface, smaller puncta might represent intracellular Cx43 to be sorted to the plasma membrane or small junctions internalized and targeted for degradation. The quantification of Cx43 punctate areas revealed significant difference between control and Aβ 25-35 treated astrocytes with Aβ 25-35 treated cultures showing a lower frequency of large Cx43 puncta >1µm 2 in size ( Figure 2E).

Effect of Aβ on Cx43 gap junction assembly in mouse primary astrocytes
It is well established that gap junction plaques are resistant to 1% Triton X-100 at 4°C while monomers and lower oligomers of Cx43 are mostly solubilised under such conditions (32,33). In order to understand Cx43 assembly into gap junctions, membrane enriched fractions from control and Aβ 25-35 treated mouse astrocytes were subjected to 1% Triton X-100 at 4°C in a detergent solubility assay, separated into Triton X-100 soluble and insoluble fractions and analysed by immunoblotting for Cx43 antibody ( Figure 3A). Our results show that Cx43 levels are decreased in the Triton X-100 insoluble fractions with a concomitant increase in Triton X-100 soluble fractions of Aβ 25-35treated cultures compared to control astrocytes by guest on November 6, 2020 http://www.jbc.org/ Downloaded from consistent with reduced functional gap junction communication in Aβ 25-35 -treated astrocytes ( Fig  3A). Further upon in situ Triton X-100 extractions, control astrocytes retained Cx43 puncta mostly at the cell-cell junctions, representing gap junctions ( Figure 3B, arrowheads). Interestingly, in situ Triton X-100 extraction in Aβ 25-35 -treated astrocytes confirmed that the majority of the Triton X-100 insoluble Cx43 puncta corresponded to the intracellular pool ( Figure 3C, arrows), with occasional plasma membrane labelling ( Figure  3C). Thus, our results indicate that mouse primary astrocytes exposed to 10μM Aβ 25-35 for 72h exhibit reduced gap junction labelling with a concomitant increase in intracellular Cx43 localisation that may correspond to either internalized gap junctions or small aggregates of Cx43 that fail to traffic to the plasma membrane.

Aβ treatment causes Cx43 retention in ER/ERGIC of mouse primary astrocytes
To further understand where the intracellular Cx43 pool may be localized in Aβ treated astrocytes, we performed double immunolabelling with anti-Cx43 and well established markers of the endoplasmic reticulum (ER, calnexin; Figure 4A-I) or ER Golgi Intermediate Compartments (ERGIC, β-COP; Figure 5A-I) followed by confocal microscopy. Consistent with our previous findings, control astrocytes showed distinct organisation of Cx43 puncta mostly away from the nucleus ( Figure 4A-C and Figure 5A-C; small arrows) without much co-localization with either calnexin ( Figure 4A-C) or β-COP ( Figure  5A-C). In contrast, mouse primary astrocytes subjected to 10μM Aβ 25-35 for 72h showed a reticular staining pattern in the perinuclear region that co-localized mostly with calnexin ( Figure 4D-F, arrowheads) and also with β-COP ( Figure 5D-F, arrowheads). The number of points showing colocalization of Cx43 puncta with either Calnexin ( Figure 4G-I) or β-COP ( Figure 5G-I) was found to be significantly higher in Aβ treated astrocytes compared to control cultures. Thus, our results show that a portion of the intracellular Cx43 was retained in the ER/ERGIC in mouse primary astrocytes exposed to 10μM Aβ [25][26][27][28][29][30][31][32][33][34][35] for 72h compared to control astrocytes.

4-Phenylbutyrate restores functional gap junction communication in Aβ treated mouse primary astrocytes
Given that Aβ 25-35 treatment increased localisation of Cx43 in the ER/ERGIC cellular compartments, we reasoned that Aβ might impair proper folding and oligomerization of Cx43 which is a pre-requisite for trafficking of Cx43 to the cell surface. In this context, 4phenylbutyrate (4-PBA), a well-known histone deacetylase inhibitor, has been reported to restore proper trafficking of several mutant and misfolded transmembrane proteins including connexins to the cell surface (34)(35)(36). Hence, we subjected control and Aβ 25-35 treated (10μM, 72h) mouse astrocyte cultures to 5mM of 4-PBA for 24h. Our immunolabelling studies with anti-Cx43 antibody ( Figure 6A-D) showed a more organised distribution of Cx43 as short lines of Cx43 puncta typically representing gap junctions following 4-PBA treatment in Aβ 25-35 treated astrocytes ( Figure 6D). Astrocytes exposed to Aβ 25-35 alone exhibited Cx43 labelling scattered throughout the cytoplasm as shown earlier ( Figure 6C). Supporting this, lucifer yellow dye transfer assays, clearly showed that 4-PBA restored functional gap junction communication in Aβ 25-35 treated astrocyte cultures similar to control astrocytes ( Figure 6E-I). Thus, our results indicate that 4-PBA restores Cx43 expression at the cell surface typically in gap junctions with a parallel increase in functional gap junction coupling in Aβ treated astrocytes.
Aβ does not affect cell surface trafficking of the newly synthesized Cx43 reaching the Golgi complex Impaired gap junction communication could be either due to delay in trafficking of newly synthesized Cx43 to the plasma membrane or increased internalization of the Cx43 gap junctions from the cell surface. To understand if the forward trafficking of the newly synthesized Cx43 channels from the Golgi to the plasma membrane could be affected by Aβ, we subjected control and Aβ 25-35 treated astrocytes to a brefaldin A (BFA) forward trafficking assay. Treatment with BFA induces a fast reversible inhibition of COPI-mediated transport from the ER to the Golgi (37). Control and Aβ treated astrocytes were treated with 5μg/ml of BFA for 8h, and spatial distribution of Cx43 was analysed by immunofluorescent labelling of Cx43 ( Figure 7A-F). When control astrocytes were treated with BFA for 8h, pre-existing gap junction plaques were not identifiable at the cell surface ( Figure 7A). Most of the Cx43 labeling was found accumulated in the perinuclear region both in control and Aβ 25-35 treated astrocyte cultures upon BFA treatment ( Figure 7A,D). As expected, upon removal of BFA, transport of newly synthesized Cx43 was restored in control astrocyte cultures ( Figure 7B,C). One hour after BFA removal control astrocyte cultures still showed a lot of Cx43 labelling scattered throughout the cytoplasm and in the perinuclear region ( Figure 7B). Three hours post BFA removal control astrocytes started showing Cx43 puncta organised into short lines typical of gap junctions indicating trafficking to the cell surface ( Figure 7C). Similarly, in Aβ 25-35 treated astrocyte cultures 3h after BFA removal Cx43 immunolabelling exhibited distinct punctate labelling away from the nucleus. These results indicated that further trafficking to the cell surface of the newly synthesized Cx43 that could reach the Golgi complex might not be affected by Aβ ( Figure 7F).

Aβ increases internalization of Cx43 gap junctions from the cell surface
The short half-life of Cx43 makes functional gap junction communication between the coupled cells a highly dynamic process. The balance between the synthesis and degradation rate of Cx43 is an important way to control the level of functional gap junctions present at the cell surface. Cx43 internalization has been reported to be dependent on the dynamin GTPase activity (38,39), which can be specifically blocked by the non-competitive endocytosis inhibitor, dynasore. First, to determine if functional gap junction communication can be improved in Aβ treated mouse astrocyte cultures by blocking endocytosis, we treated control and Aβ [25][26][27][28][29][30][31][32][33][34][35] treated cultures with dynasore for 3h ( Figure  8A-E). Control cultures did not show any significant alterations in lucifer yellow dye transfer mediated through the gap junctions following dynasore treatment. Interestingly, dynasore intervention in Aβ 25-35 treated astrocytes demonstrated a marked improvement in lucifer yellow dye transfer from the scrape loading line indicating that Aβ causes faster internalization of Cx43 from the cell surface ( Figure  8A-E). Supporting this, our immunofluorescent labelling studies with anti-Cx43 antibody ( Figure 8F-I) showed the presence of lined up Cx43 puncta typically representing the gap junctions in Aβ 25-35 treated astrocytes in presence of Dynasore ( Figure 8I). Aβ [25][26][27][28][29][30][31][32][33][34][35] treated astrocytes in absence of dynasore exhibited Cx43 immunostaining scattered throughout the cell cytoplasm as shown earlier ( Figure 8H). Moreover, the cytoplasmic pool of Cx43 appeared to be less in Aβ 25-35 treated astrocytes exposed to dynasore ( Figure 8I) compared to cultures without dynasore ( Figure  8H,I) confirming those to be internalized gap junctions.
To further determine if the impaired gap junction communication in Aβ treated astrocytes was because of increased internalization rate of Cx43 rather than failed delivery to the plasma membrane, we first subjected the control and Aβ 25-35 treated astrocytes to BFA forward trafficking assay followed by dynasore intervention during the 3h BFA washout ( Figure  9). As expected, in the control astrocyte cultures functional gap junction communication was reestablished upon restoring the secretory pathway with BFA washout for 3h ( Figure 9A-E). The presence of dynasore in the culture medium during the 3h BFA washout showed an additional increasing trend without reaching statistical significance in the dye travel distance compared to the BFA washout only cultures without dynasore intervention. In contrast, the presence of dynasore in the cell culture medium during 3h BFA washout in the Aβ 25-35 treated astrocytes significantly improved dye transfer mediated through the functional Cx43 gap junctions ( Figure 9F-J). Taken together, these data suggest that Aβ increases endocytosis of Cx43 gap junctions thereby impairing functional gap junction communication between the mouse primary cultured astrocytes.

Discussion
In this study we investigated the underlying cellular mechanisms by which Aβ re-models the predominant astroglial gap junction protein Cx43 function in mouse primary cultured astrocytes. Our results indicate that exposure to 10μM of Aβ 25-35 for 72h functionally uncoupled gap junction and hemichannel activity without altering Cx43 steady-state levels. Interestingly, Aβ 25-35 treatment caused an increase in intracellular Cx43 expression with a concomitant decrease in gap junction labelling. A major portion of this intracellular Cx43 expression was found to be resistant to Triton X-100 extraction indicating the presence of Cx43 aggregates and/or internalized gap junctions in line with reduced gap junctional coupling. Additionally, we found that a pool of the cytoplasmic Cx43 expression localised to the ER/ERGIC compartments. Supporting this treatment with 4-PBA, that is known to improve surface trafficking of several transmembrane proteins, restored Cx43 expression in gap junction plaques and functional gap junctional coupling in mouse primary astrocytes exposed to Aβ [25][26][27][28][29][30][31][32][33][34][35] . Our data provide further evidence that the reduction in gap junction communication in Aβ 25-35 treated astrocytes was also caused by increased internalization of Cx43 from the cell surface via endocytosis and not due to the impaired delivery of Cx43 from the Golgi to the cell surface. Furthermore, using molecular dynamic simulations we showed that Aβ might bind Cx43, thus mediating the observed effects. Thus, this study provides new insights into the cellular mechanisms underlying altered Cx43 function in astrocytes exposed to Aβ peptides.
Earlier studies have suggested that both reactive astrogliosis and increase in Aβ levels, the two key neuropathological features that characterise all AD brains, are associated with altered astrocytic connexin channel functions subsequently leading to neuronal damage (21)(22)(23)(24)27). Accumulating evidence indicates that gap junctions and hemichannels are oppositely regulated in several neuropathological conditions (40). In line with this inverse correlation, we observed that Aβ 25-35 significantly reduced astrocyte-astrocyte gap junction communication as examined by intercellular lucifer yellow dye transfer and increased hemichannel activity measured by ethidium bromide dye uptake. Cx43 mRNA and total protein levels including the pattern of immunoreactive bands did not show any significant difference between Aβ 25-35 treated and control astrocyte cultures. Thus, the modulations in dye transfer and dye uptake induced by Aβ [25][26][27][28][29][30][31][32][33][34][35] were not due to changes in Cx43 levels or its altered phosphorylation state detectable by shifts in Cx43 electrophoretic mobility. Interestingly, our immunolabelling studies showed a clear distinction in the Cx43 cellular distribution in Aβ 25-35 treated mouse astrocytes compared to control cultures. While we showed that expression pattern of Cx43 puncta characteristic of gap junction plaques appeared to be reduced in Aβ 25-35 treated astrocytes, a simultaneous increase in the intracellular Cx43 labelling explained the reduced astroglial gap junction coupling. Moreover, many of these Cx43 puncta were found to be resistant to Triton X-100 extraction at 4°C, suggesting intracellular Cx43 aggregates and/or internalized gap junctions that do not contribute to functional gap junction coupling.
The life cycle of Cx43 involves posttranslational insertion of Cx43 monomers into the ER membrane followed by their oligomerization into hexameric hemichannels or connexons in the Golgi. Preassembled hemichannels are translocated to the cell surface, which then dock with connexons from an apposing cell and undergo channel clustering to form morphologically identifiable gap junction plaques (41). Cx43 is unique in that it does not oligomerize in the ER unlike most multimeric membrane proteins. Instead, it exists as monomers in the ER membrane where Cx43 is stabilized by atleast one or more protein chaperones that prevent its premature oligomerization in the ER (36,42). Premature oligomerization in ER can cause Cx43 aggregation leading towards its degradation by the proteasomal pathway thereby causing reduced gap junction coupling. Our immunolabelling studies showed an increased co-localization of Cx43 with the ER marker calnexin and ERGIC marker β-COP in Aβ [25][26][27][28][29][30][31][32][33][34][35] treated astrocytes compared to control astrocytes pointing towards a reduced ER function in proper trafficking of Cx43 along the secretory pathway. Supporting this, our results indicate that treatment with 4-PBA can play an important role in improving gap junction coupling in Aβ 25-35 treated astrocytes. 4-PBA, a well-established histone deacetylase inhibitor, improves the trafficking of several transmembrane proteins, including Cx43, an effect that is mediated through the modulation of ER associated chaperones like ERp29 (34)(35)(36). Thus, it is possible that 4-PBA alters expression of one or more proteins of the Cx43 quality control pathway that regulates proper folding and trafficking of Cx43 to the cell surface, thereby restoring the impaired gap junction communication between astrocytes following Aβ 25-35 treatment. Interestingly, several recent studies have reported that administration of 4-PBA in established mouse models of AD exerted neuroprotective effects and reversed AD associated phenotypes including cognitive deficits (43)(44)(45). Thus, 4-PBA, a drug already approved for clinical use in treating urea cycle disorders (46) has been proposed to provide a novel approach for the treatment of AD. However, the underlying molecular mechanisms by which 4-PBA offers neuroprotection in AD is not fully understood and needs further investigation.
Depending on the cell type studied, Cx43 has a very short half-life ranging from 1.5-5 h (47)(48)(49). This makes gap junctions highly dynamic plasma membrane domains with fast turnover rates. Thus the amount of gap junctions and hemichannels present on the cell surface is largely determined by a balance between the formation and removal of such channels from the plasma membrane. Removal of gap junctions from the plasma membrane involves internalization of the entire junctional complex in the form of double membrane annular junctions that are targeted to lysosomes for degradation (50)(51)(52)(53). Among the many processes involved in the removal of gap junctions, the clathrin machinery has been reported to play a major role in gap junction internalization. Notably, clathrin has been detected on Cx43 GJs by immunofluorescence and silencing of clathrin and the adaptor proteins, AP-2 and Dab2, or the GTPase dynamin caused cells to harbor fewer annular gap junctions (38,39). Interestingly, our results using a dynamin blocker, dynasore clearly showed that gap junction communication between the astrocytes measured by lucifer yellow dye transfer was significantly restored in cultures exposed to Aβ [25][26][27][28][29][30][31][32][33][34][35] . Supporting this, Cx43 immuno-labelling showed elevated gap junction labelling in Aβ 25-35 treated astrocytes in presence of dynasore compared to those without dynasore intervention. This is further in line with our observations from in situ Triton X-100 solubility assays, showing Triton X-100 resistant Cx43 cytoplasmic labelling in Aβ 25-35 treated astrocytes, indicating that these could be internalised Cx43 gap junctions. In addition, we have used BFA to separate the processes of formation and removal of gap junctions from the cell surface. Our results using dye transfer and immunolabelling studies showed that delivery of newly synthesized Cx43 from the Golgi to the cell surface was not affected in presence of Aβ [25][26][27][28][29][30][31][32][33][34][35]

Primary astrocyte culture
Primary astrocyte cultures were prepared from postnatal day 0-1 mouse pup brains as previously described with minor modifications (55). Briefly, following removal of the olfactory bulbs, cerebellum and meninges, brain tissue was dissociated in Ca +2 and Mg +2 free HBSS with 0.25% Trypsin and 50μg/ml DNase I at 37°C for 30mins with shaking. Trypsin was neutralized by adding 0.25% Fetal bovine serum (FBS) and followed by a wash at 300xg for 10mins. The pellet was resuspended in HBSS and the cell suspension was passed through a 70μm cell strainer. Cells were seeded onto T25 flasks in astrocyte growth medium containing DMEM supplemented with 10% FBS and 1% penicillin/streptomycin. Media was replaced 24h after plating to remove all non-adherent cells and every third day thereafter. Primary astrocytes reached about 90% confluency at 8-9 days in vitro following which addition of fresh media was stopped for next 10 days to allow differential adhesion of astrocytes and microglia. Subsequently, culture flasks were shaken vigorously in an orbital incubator shaker at 200 rpm for 45mins at 37°C to dislodge the loosely adherent microglia growing on top of the strongly adherent astrocyte monolayer. The media was discarded and astrocytes were collected following trypsinization and plated for experiments. The enriched cultures from three independent culture batches were quantified for GFAP-positive astrocytes by flow cytometry (described later), which showed that cultures contained 97.45 ± 0.41% (mean±SD) of GFAPpositive astrocytes ( Figure S2)

Flow cytometry
Confluent astrocyte cultures were trypsinized and washed twice in flow buffer (PBS containing Ca +2 /Mg +2 and 2% FBS). Following centrifugation at 300xg for 5mins at 4°C, cells were resuspended in flow buffer and approximately 1x10 6 cells were added to each 5ml polystyrene round bottom FACs tubes. Cells were fixed in 100μl BD CytoFix Fixation buffer for 15mins at room temperature (RT). Following a wash with flow buffer, cells were incubated with rabbit anti-GFAP antibody (1:50) diluted in BD Perm/Wash Buffer for 30mins at RT. Subsequently, cells were washed thrice in in BD Perm/Wash Buffer and labelled with goat anti-rabbit fluorescein isothiocyanate (FITC)conjugated secondary antibody (1:50) for 30 min at RT. After a final wash, cells were resuspended in 500μl of flow buffer and subjected to flow cytometry using a BD FACS verse flow cytometer. Ten thousand to 50,000 events were collected for each tube assayed and the data was analyzed using BD FACSuite analysis software (55). The experiments were performed from three independent cultures.

RNA isolation and Real-time PCR
RNA was isolated using RNeasy RNA extraction kit following manufacturer's instructions (Qiagen). RNA (0.5-1μg) was treated with DNaseI and reverse transcribed using High Capacity cDNA Reverse Transcription kit following manufacturer's protocol. Quantitative real-time PCR was performed with DyNAmo Color Flash SYBR Green master mix using QuantStudio 3 Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific) with the following cycling conditions: initial denaturation at 95°C for 7 min followed by 40 cycles at 95°C for 10 s and 60°C for 30s. Primer sequences used were: Connexin- 43 (forward, 5'-CCCTTCACGCGATCCTTA-3' and reverse, 5'-TCATGCTGGTGGTGTCCTTG-3') and β-actin (forward, 5'-GTGACGTTGACATCCGTAAAGA-3' and reverse, 5'-GCCGGACTCATCGTACTCC-3'). Experiments were performed from four independent cultures and each sample was assayed in duplicate and mRNA normalized to β-actin mRNA. Relative gene expressions were calculated using the comparative C T method (2 -^C T ) (56).

Western blotting
Confluent monolayers of astrocytes were homogenized in ice-cold radioimmunoprecipitation (RIPA) lysis buffer (50mM Tris base, 150mM NaCl, 0.1% SDS, 1% Triton-X-100, 0.5% sodium deoxycholate; pH 7.4) containing protease and phosphatase inhibitor cocktails. Protein concentrations were determined using BCA protein assay kit following manufacturer's instructions. Protein samples (5μg) were resolved on 12% SDSpolyacrylamide gel and transferred to by guest on November 6, 2020 http://www.jbc.org/ Downloaded from polyvinylidene difluoride membranes. Membranes were blocked with 10% skimmed milk for 1h at RT and incubated overnight at 4°C with anti-Cx43 antibody. Following three washes in tris-buffered saline containing 0.1% Tween-20 (TBST), membranes were incubated with horseradish peroxidase-conjugated secondary antibodies and immunoreactive proteins were visualized using SuperSignal TM West Pico PLUS enhanced chemiluminescence kit. All blots were re-probed with anti-β-actin antibody to monitor equal protein loading and densitometry was performed using NIH ImageJ gel analysis tool (57). Experiments were performed from six independent cultures and each sample was loaded in duplicate.

Immunofluorescence and Confocal microscopy
Cells grown on glass coverslips were fixed with 4% paraformaldehyde for 15mins at RT and washed three times with phosphate buffered saline (PBS). Fixed cells were permeabilized with 0.25% TritonX-100 in PBS for 15mins at RT and blocked with PBS containing 0.1% Triton X-100 and 1% Bovine serum albumin (BSA) for 1h at RT. Cells were incubated overnight at 4°C with anti-Cx43 antibody combined with either anti-GFAP, anti-calnexin or anti-β-COP antibodies (at dilutions listed in Table 1) diluted in PBS with 0.1% Triton X-100 and 0.25% BSA and subsequently labeled with Alexa Fluor 488-and/or 568-conjugated secondary antibodies for for 2h at RT. Coverslips were washed thrice with PBS and mounted in ProLong Gold anti-fade reagent. Immunostained cells were visualized and imaged using a Olympus IX-81 inverted epifluorescence microscope. For colocalization studies, imaging of double-immunolabelled cells were performed using a Leica SP8 confocal laser scanning microscope (Leica Microsystems GmbH, Wetzlar, Germany) equipped with a HC PL APO 63x oil-immersion objective lens using 488 nm and 552 nm laser lines. Images were acquired and processed using LasX software (Leica Microsystems). Image analysis was done using ImageJ software (57). The determination of colocalization of Cx43 and the ER marker, Calnexin or the ERGIC marker, β-COP was carried out by using ImageJ Colocalization Threshold plug-in. Pearson's coefficient for colocalization was calculated that compares a pair of images from different fluorescent channels. Five to ten fields were analysed for each group from two independent experiments.

Triton X-100 solubilization assay
Cx43 assembly into gap junctions was analysed using Triton X-100 detergent solubility assay (33). Primary astrocytes were harvested in icecold PBS containing protease and phosphatase inhibitor cocktails. Cells were homogenized using a ball bearing cell homogenizer. Protein concentrations were determined using BCA protein assay kit and equal amounts of protein for Aβ [25][26][27][28][29][30][31][32][33][34][35] treated and control groups were centrifuged at 100,000xg for 1h at 4°C using a Beckman Optima Max ultracentrifuge (Beckman Coulter Inc., Brea, CA) to obtain a membraneenriched pellet. For detergent solubilization, the membrane enriched pellet was resuspended in PBS containing 1% Triton X-100 and incubated at 4°C for 30mins. The samples were centrifuged at 100,000xg for 30mins at 4°C and separated into Triton X-100 soluble supernatant and Triton X-100 insoluble pellet. The pellet was resuspended in lysis buffer containing 1% SDS and solubilized by sonication (Triton X-100 insoluble fraction). Equal volumes of soluble and insoluble fractions were resolved on 12% SDS-PAGE gel and analyzed for Cx43 levels by immunoblotting. Experiments were performed from four independent cultures.
For detergent extraction in situ, confluent astrocyte monolayers grown on PLL-coated coverslips were rinsed with PBS and incubated in ice-cold PBS containing 1% Triton-X-100 for 30mins at 4°C with gentle shaking. The control cells were treated similarly without 1% Triton X-100. The coverslips were fixed with 4% paraformaldehyde for 15mins at RT and immunolabelled with anti-Cx43 antibody as described earlier. Experiments were performed from three independent cultures.

Dye Transfer assay
Gap junction mediated intercellular coupling was evaluated by transfer of lucifer yellow dye in a scrape loading assay as described previously (55). Briefly, completely confluent monolayer cultures of control and Aβ 25-35 treated astrocytes grown on 35 mm culture dishes were scrape loaded with lucifer yellow dye and incubated at RT for 1min. The cells were then rinsed quickly three times with PBS and incubated at 37ºC for an additional 8mins in the growth media to allow the loaded dye to transfer to adjoining cells. Fluorescent images were captured using an Olympus® IX-81 epifluorescence microscope with a UV filter. The distance traveled by the dye in different treatments was measured from the scrape line using ImageJ software (57). Five to ten fields were analysed for each group from three to four independent experiments.

Dye uptake assay
Dye uptake assay using Ethidium bromide (EtBr) was done to study the hemichannel activity (24). The cells grown to approximately 70% confluency were washed with Locke's solution (154mM NaCl, 5.4mM KCl, 2.3mM CaCl 2 , 5mM HEPES buffer; pH 7.4) after removing media and were treated with Locke's solution containing 5μM EtBr for 10mins at 37°C. Subsequently, the cells were fixed in 4% PFA and preceded with normal immunofluorescence protocol as described above to probe with anti-GFAP antibody. The cells were mounted with mounting medium without DAPI (Vectashield, Vector Laboratories Inc., Burlingam, CA). Cells were imaged using Olympus® IX-81 epifluorescence microscope and the intensity of the nuclei was quantified using ImageJ software (57). Five to ten fields were analysed for each group from three independent experiments.

Homology Modeling
The primary sequence of mouse Cx43 was obtained from the Uniprot Knowledgebase (accession number P23242). Cx43 is 382 amino acids long and shares homology with various connexins ( Table 2). Since no single template structure shares very high sequence homology with the whole Cx43, multiple templates ( Table  2) were used along with the MODELLER package (58) for the Cx43 model building. We have generated 20 models and the model with the lowest DOPE score (59) was further used for docking studies.

Docking and Molecular Dynamics Simulations
Being an intrinsically disordered peptide, the conformation of Aβ is subject to its surrounding environment and peptide length (60). Previously reported monomeric structure of Aβ 1-42 (61)(62)(63), was used in docking studies with Cx43. The Aβ 25-35 segment was extracted from the fulllength monomer Aβ, as the conformation of this segment in fully aqueous environment remains experimentally unreported. Both structures were consequently used for blind docking with Cx43 using the ZDOCK server (64). The transmembrane helices of Cx43 were excluded for docking interaction.
The best ranked docked poses of Cx43 with either Aβ 1-42 monomer or Aβ 25-35 monomer were further used as the initial structure for performing implicit solvent molecular dynamics simulations. All simulations were carried out in AMBER16 (65) program suite. The Cx43 and Aβ were prepared with AMBER ff14SB force field (66) and the effect of water was implemented using the generalized Born model (67). A 1000 steps of steepest descent energy minimization was followed by another 1000 steps of conjugate-gradient minimization without any constraint. We have performed 200 ps of equilibration where the system temperature was raised from 0 K to 310 K using a Langevin thermostat. This was followed by a 100 ns production run in NVT condition where coordinates were saved every 5 ps. Throughout the equilibration and production run the four transmembrane helices of the Cx43 were constrained by applying a harmonic force of 1 kcal/mol to main-chain atoms. During the simulation, the infinite cutoff was used to calculate the non-bonded interaction energy, and SHAKE (68) was used to restrain the bonds involving hydrogen atoms. Interaction energy of the last 50 ns of data, were calculated using the NAMD Energy plugin (69).

Statistical analysis
All data are presented as mean±SD and each data is represented as a point in the scatter plot. Statistical significance was determined by oneway ANOVA followed by Newman-Keuls posthoc analysis for multiple comparisons or by guest on November 6, 2020 http://www.jbc.org/ Downloaded from unpaired two-tailed Student's t-test for single comparison with a significance threshold of p < 0.05. All analyses were performed using GraphPad Prism version 5.0 software (GraphPad Software, Inc., La Jolla, CA).