Dileucine-like motifs in the C-terminal tail of connexin32 control its endocytosis and assembly into gap junctions

Gap junctions (GJ s) are bicellular structures that are formed by the conglomeration of several cell–cell channels at cell-to-cell contact sites (Loewenstein, 1981; Sosinsky and Nicholson, 2005). The direct exchange of small molecules (≤1500 Da) between the cytoplasmic interiors of contiguous cells through these channels maintains cytoplasmic continuity through buffering of nutrients and sharing of second messengers (Goodenough and Paul, 2009; Loewenstein, 1981). Gap junctions are abundant in the polarized and well-differentiated epithelial cells of exocrine glands, and sharing of second messengers through gap junctional channels fulfills a homeostatic role by permitting the synchronous response of cells to various physiological stimuli (Bosco et al., 2011; Macara et al., 2014; Meda, 2017; Potolicchio et al., 2012). The gap junctional channels are formed of proteins called connexins (Cxs), which are encoded by a family of 21 distinct genes in humans (Beyer and Berthoud, 2009). While some Cxs are expressed ubiquitously, the expression of other Cxs is tissue specific, which suggests that gap junctional channels composed of each type of Cx play specific roles. Credence to this notion has come from the tissue-specific knock out of Cx genes (Dobrowolski and Willecke, 2009) and the identification of mutations in these genes in human genetic diseases, such as occulodentodigital dysplasia, palmoplanta keratoderma, karatitis-icthyosis-deafness syndrome and deafness (Dobrowolski and Willecke, 2009; Laird, 2006, 2010; Xu and Nicholson, 2013). Defects in the trafficking and assembly of Cxs into GJs often ensue during the progression of a wide variety of carcinomas (Aasen et al., 2016; Naus and Laird, 2010). Moreover, growth factors, tumor promoters, oncogenes and tumor suppressors modulate the trafficking and assembly of Cxs into GJs by multiple mechanisms (Leithe, 2016; Leithe et al., 2012). Hence, investigation into the molecular mechanisms involved in controlling the trafficking and assembly of Cxs into GJs in cancer cells may enhance our understanding about the pathogenesis of several types of cancers.

A gap junctional channel is formed by the collaborative efforts of two cells. The biogenesis of a GJ – often called a GJ plaque – takes place in a stepwise manner. Six Cxs first oligomerize to form a hexamer called a connexon, which, upon reaching the cell surface, docks with the connexon on a contiguous cell to form a gap junctional channel. Several channels then cluster at one particular spot to give rise to a gap junctional plaque (Laird, 2006; Thévenin et al., 2013). The half-life of both assembled (plaque-associated) and unassembled (non-plaque-associated) Cxs ranges from 2 to 5 h, rendering GJ plaques highly dynamic bicellular structures (Falk et al., 2009; Gaietta et al., 2002; Jordan et al., 2001; Laird, 2006; Piehl et al., 2007). Evidence to date supports the notion that GJs are degraded via the endocytosis of either the entire plaque or senescent cell–cell channels from the center of the plaque as double-membrane vesicles (Falk et al., 2009, 2016; Gaietta et al., 2002; Jordan et al., 2001; Nickel et al., 2008; Piehl et al., 2007). Connexons and GJs have been shown to be internalized by both clathrin- and non-clathrin-mediated pathway (Gumpert et al., 2008; Lin et al., 2003; Piehl et al., 2007; Schubert et al., 2002). However, it is not clear how the known endocytic mechanisms – utilized for the internalization of monomeric transmembrane proteins (Traub, 2009; Traub and Bonifacino, 2013) – are adapted to internalize oligomeric connexons or gap junctional plaques, which are conglomerations of cell–cell channels comprising 50–10,000 units. Furthermore, GJs have also been found to be degraded by autophagy upon internalization (Bejiarno et al., 2012; Fong et al., 2012; Hesketh et al., 2010; Lichtenstein et al., 2011).

A single cell often expresses multiple Cxs, which are localized on different membrane domains, where they are likely subjected to distinct regulatory cues (Guerrier et al., 1995; Laird, 2006). Connexin32 (Cx32; also known as GJB1), a 32 kDa Cx, is abundantly expressed in the polarized cells of exocrine glands, such as the prostate and pancreas, as well as in other well-differentiated cell types, such as hepatocytes and oligodendrocytes (Bosco et al., 2011; Laird, 2006). Our earlier studies showed that the ability of Cx32 and Cx43 (also known as GJA1) to assemble into GJs was impaired in prostate cancer cell lines and prostate tumors (Govindarajan et al., 2002; Habermann et al., 2002; Mehta et al., 1999). We also showed that the cytoplasmic tail of Cx32 was not required to initiate the formation of GJs but was essential for the growth and stability of the plaque (Katoch et al., 2015). These findings suggest that the cytoplasmic tail of Cx32 harbors important regulatory motifs that dictate the stability of Cx32 and/or its recruitment and incorporation into gap junctional plaques. Previous studies have also shown that a tyrosine-based sorting motif in the cytoplasmic tail of Cx43 controls its assembly into GJs by regulating endocytosis via the clathrin-mediated pathway in human pancreatic cancer cells (Johnson et al., 2013), as well as in several other cell lines (Fong et al., 2013; Gumpert et al., 2008; Piehl et al., 2007; Thomas et al., 2003). Therefore, we searched for the endocytic motifs in the cytoplasmic tail of Cx32. We report here that the tail of Cx32 harbors tyrosine-based and dileucine-based motifs, which regulate the trafficking, basolateral sorting and clathrin-mediated endocytosis of transmembrane proteins (Traub, 2009; Traub and Bonifacino, 2013). Here, we focused on examining the role of dileucine-based motifs in regulating the trafficking and endocytosis of Cx32. We provide evidence that these motifs have no effect on the intracellular sorting of Cx32 but govern its ability to form GJs at the cell surface by regulating its endocytosis via the clathrin-mediated pathway in cell lines derived from human prostate and pancreatic tumors.

We first explored the endocytic itinerary of Cx32 in the human pancreatic cancer cell line BxPC3 and the prostate cancer cell line LNCaP used in our previous studies (Johnson et al., 2013; Katoch et al., 2015). Cx32 is not expressed in LNCaP and BxPC3 cells (Johnson et al., 2013; Katoch et al., 2015; Mehta et al., 1999). We freshly introduced Cx32 retrovirally in both cell lines and obtained pooled populations of Cx32-expressing BxPC3 and LNCaP cells, hereafter referred to as Bx-32 and LN-32 cells. In both Bx-32 and LN-32 cells, the expression of Cx32 was robust and the protein formed GJs (Fig. S1A,B). Immunocytochemical analysis, by high-resolution structured illumination microscopy (SIM), showed that Cx32 colocalized discernibly with clathrin and α-adaptin, a component of adaptor protein complex 2 (AP-2), but not with caveolin-1, a marker for the non-clathrin-dependent pathway (Mayor and Pagano, 2007) (Fig. 1A). Moreover, no colocalization with the early endosomal antigen 1 (EEA1), which is a marker for the early endocytic vesicles (Mills et al., 1998; Mu et al., 1995; Stenmark et al., 1996), was observed (Fig. S1C). We previously showed that Cx43 did not colocalize with clathrin and EEA1, and yet was internalized into Rab5-positive vesicles (Rab5 has two forms known as Rab5a and Rab5b) in contacting BxPC3 cells (Johnson et al., 2013). These findings – combined with the fact that Rab5 is involved in clathrin-mediated endocytosis (Bucci et al., 1992; Zerial and McBride, 2001) – prompted us to examine whether Rab5 was involved in the endocytosis of Cx32. Therefore, we transiently expressed constitutively active (dominant) mutant of Rab5 tagged with EGFP (Rab5DA–GFP) in Bx-32 and LN-32 cells. Expression of constitutively active Rab5 causes the formation of giant endocytic vesicles, which allows unambiguous detection of proteins in Rab5-positve vesicles (Roberts et al., 1999; Stenmark et al., 1994). Connexin32 was robustly endocytosed into Rab5-positive vesicles (Fig. 1B). These findings suggest that Cx32 is endocytosed by the clathrin-mediated pathway into Rab5-positive vesicles in LN-32 and Bx-32 cells.

A motif search revealed that the cytoplasmic tails of Cx32 harbored two tyrosine-based [YxxØ]-type and two dileucine-based [DE]xxxL[LI]-type motifs (a single letter denotes an amino acid; x is any amino acid and Ø is a bulky hydrophobic amino acid). These motifs mediate the basolateral sorting and endocytosis of transmembrane proteins (Bonifacino, 2014; Traub and Bonifacino, 2013). As shown in Fig. 2A, the tyrosine-based motifs in Cx32 are YTLL and YLII with tyrosine at residue 7 and 211, respectively, whereas the dileucine-based motifs are EVVYLI with a leucine-isoleucine pair at residues 212 and 213 and EINKLL with a leucine–leucine pair at position 251 and 252 (Thévenin et al., 2013). Moreover, we fortuitously discovered a non-canonical motif LKDILR when we randomly chose to mutate leucine and arginine at position 263 and 264 non-specifically to serve as a control (Fig. 2A). To explore the role of these motifs in regulating the trafficking, sorting and endocytosis of Cx32, we constructed mutants where critical residues involving pairs of leucine-leucine, leucine-isoleucine and leucine-arginine were mutated into alanine residues (Fig. 2B), which rendered these motifs non-functional (Bonifacino and Traub, 2003). The mutant L212A/I213A is abbreviated as Cx32-LI, mutant L251A/L252A as Cx32-LL, and mutant L263A/R264A is abbreviated as Cx32-LR (Fig. 2B). The dileucine motifs EVVYLI and EINKLL will be referred to as LI and LL, respectively, whereas the putative motif LKDILR is denoted LR.

We retrovirally expressed wild-type Cx32 (Cx32-WT), Cx32-LI, Cx32-LL and Cx32-LR in BxPC3 and LNCaP cells and examined the formation of GJs immunocytochemically. The results showed that, compared to what was seen with Cx32-WT, expression of mutants Cx32-LL and Cx32-LR led to the formation of larger GJs. In contrast, mutant Cx32-LI-expressing cells rarely formed GJs and remained predominantly intracellular as vesicular puncta (Fig. 3A). Western blot analysis showed robust expression of Cx32-WT and mutants Cx32-LL, Cx32-LR and Cx32-LI in both cell types (Fig. 3B). To corroborate the immunocytochemical data, we measured surface areas of GJs at several cell–cell interfaces (see Materials and Methods). We used E-cadherin to delineate cell–cell interfaces. The mean surface area of GJs composed of Cx32-LL and Cx32-LR was nearly 2-fold larger than that of Cx32-WT GJs. In contrast, although GJ-like puncta at cell–cell interfaces were rarely seen with Cx32-LI, their mean surface area was 2- to 3-fold smaller than Cx32-WT puncta (Fig. 3C). Moreover, there was a concomitant 2- to 3-fold decrease in the number of GJs per interface in cells expressing Cx32-LL or Cx32-LR (Fig. 3D). Furthermore, Cx32-LL and Cx32-LR generally formed one to three large GJs at one cell–cell interface and several small GJs at the remaining interfaces, whereas Cx32-WT formed GJs of varying sizes (Fig. 3A). To examine whether motifs LL and LR cooperate with each other in enhancing GJ formation, we constructed a quadruple mutant Cx32-LL-LR in which leucine residues at position 251, 252 and 263 and arginine at position 264 were mutated into alanine residues (Fig. 2B). When expressed retrovirally in BxPC3 and LNCaP cells, the quadruple mutant Cx32-LL-LR was also robustly expressed (Fig. 3B, right blots) and formed large GJs as with Cx32-LL and Cx32-LR (Fig. 3A, right panels) but the mean surface area of GJs composed of Cx32-LL-LR was not significantly different from those with Cx32-LL or Cx32-LR alone (Fig. 3C).

To substantiate immunocytochemical data, we examined the GJ-forming ability of Cx32-WT, Cx32-LI, Cx32-LL and Cx32-LR biochemically by means of a detergent-solubility assay and upon in situ extraction with 1% Triton X-100 (TX100) as well as functionally. We found that GJs composed of Cx32-WT, Cx32-LL and Cx32-LR were detergent insoluble, while Cx32-LI puncta – both intracellular and at cell–cell interfaces – were detergent-soluble (Fig. S1D). Analyses of total as well as TX100-soluble and -insoluble fractions from cells expressing Cx32-WT or mutants by western blotting corroborated the immunocytochemical data (Fig. S1E). As determined by quantifying the amount of total, TX100-soluble and -insoluble fractions from four separate experiments, 35–45% of Cx32-WT, 70–75% of Cx32-LL and 55–65% of Cx32-LR was insoluble and only 5–10% soluble. In contrast, 50–75% of Cx32-LI was soluble and only 5–10% insoluble (Fig. S1F).

To explore the function of GJs, we examined the permeability of GJs composed of Cx32-WT, Cx32-LI or Cx32-LL by microinjecting GJ-permeable fluorescent tracers and by scrape-loading of Lucifer Yellow (El-Fouly et al., 1987). Consistent with the immunocytochemical data, we observed that LNCaP cells expressing Cx32-LI communicated poorly compared to those expressing Cx32-WT. However, we did not observe any significant difference in the junctional transfer of Alexa Fluor 488 and Alexa Fluor 594 between LNCaP cells expressing Cx32-WT and Cx32-LL (Fig. S2A). We failed to substantiate the data obtained by microinjecting LNCaP cells with various GJ permeable tracers in the scrape-loading assay because these cells detached from the substrate upon scrape loading. Intriguingly, measurement of GJ permeability as determined by the scrape-loading assay in BxPC3 cells expressing Cx32-WT, Cx32-LI, Cx32-LL or Cx32-LR also showed no consistent, discernible difference (Fig. S2B). There may be several explanations for these intriguing findings. One plausible explanation is that BxPC3 cells express Cx43 endogenously, which, although is inefficiently assembled into GJs (Johnson et al., 2013), may be sufficient to allow the passage of Lucifer Yellow even in cells expressing Cx32-LI. On the other hand, LNCaP cells are Cx-null (Mehta et al., 1996, 1999). Taken together, the results in Fig. 3 suggest that the GJ-forming ability of Cx32 is regulated differently by the LI and the LL or LR motifs. Rendering the LL or LR motif nonfunctional enhances, whereas rendering the LI motif nonfunctional inhibits, GJ formation. Moreover, the LL and LR motifs do not cooperate with each in enhancing the formation of GJs.

Besides regulating endocytosis, the dileucine motifs also govern the trafficking of transmembrane proteins from trans-Golgi network (TGN) to the cell surface, retrograde transport from endosomes to the TGN, as well as transport from the TGN to the cell surface (Bonifacino and Traub, 2003; Traub and Bonifacino, 2013). We used a cell surface biotinylation assay to determine whether differential trafficking and degradation from the cell surface might account for the difference in the ability of Cx32-WT, Cx32-LI, Cx32-LL and Cx32-LR to form GJs. As rationalized previously (VanSlyke and Musil, 2000), only Cx32 in hemichannels is expected to be biotinylated due to the small space of 2–4 nm at the GJ plaque between contiguous cells and the large size of biotin (2.24 nm). The results showed that Cx32-WT, Cx32-LL and Cx32-LR were robustly biotinylated whereas Cx32-LI was not (Fig. 4A). There was no difference in the biotinylation of E-cadherin, a cell surface protein, which was used as a positive control. We next examined the kinetics of degradation of Cx32-WT and the mutants at various times after cell surface biotinylation to determine whether the degradation of mutants from the cell surface was affected. Compared to Cx32-WT, mutants Cx32-LL and Cx32-LR degraded with slower kinetics (Fig. 4B). As shown in Fig. 4C, we found that 50% of Cx32-WT degraded within 2–3 h while degradation of Cx32-LL and Cx32-LR was profoundly inhibited. Taken together, these data suggest that Cx32-WT, Cx32-LL and Cx32-LR traffic normally to the cell surface, whereas Cx32-LI either traffics poorly to the cell surface or mis-sorts to other vesicular compartments. Moreover, these data rule out the involvement of the LL and LR motifs in regulating the transport of Cx32 from the TGN to the cell surface or between different endosomal compartments. Furthermore, these data hint that motifs LL and LR regulate the level of Cx32 at the cell surface, possibly by promoting endocytosis.

To examine the involvement of motifs LL and LR in regulating the endocytosis of Cx32, we undertook the following experimental approaches. First, we knocked down the heavy chain of clathrin by using siRNA in cells expressing Cx32-WT, Cx32-LL or Cx32-LR. Knockdown of clathrin has been shown to inhibit endocytosis of GJs composed of Cx43 (Gumpert et al., 2008). As shown in Fig. S3A, we achieved an efficient knockdown with 25 nM of siRNA. Quantitative determination of knockdown efficiency in three independent experiments showed 80±12% (mean±s.e.m.) knockdown. Knockdown of clathrin enhanced the formation of GJs composed of Cx32-WT (Fig. S3B, left panels) but had no effect on the formation of GJs composed of Cx32-LL (Fig. S3B, right panels). Second, we transiently expressed constitutively active Rab5DA–GFP in LNCaP cells expressing Cx32-WT, Cx32-LL or Cx32-LR. Compared to Cx32-WT, mutants Cx32-LL and Cx32-LR were less frequently internalized in Rab5-positive vesicles (Fig. 5A). For example, Rab5-positive vesicles containing Cx32-WT were 3-fold more abundant than those containing Cx32-LL or Cx32-LR (Fig. S3C). Moreover, Cx32-WT was more robustly internalized into Rab5-positive vesicles compared to Cx32-LL or Cx32-LR as assessed by counting the number of Cx32-positive puncta (Fig. S3D). Furthermore, Cx32-LL or Cx32-LR neither colocalized with clathrin and AP-2 nor with caveolin-1 (data not shown). These data suggest that Cx32-WT is endocytosed by the clathrin-mediated pathway, whereas Cx32-LL and Cx32-LR are not significantly endocytosed by this pathway.

In a third approach, we expressed Cx32-LL and Cx32-LR in another human pancreatic cancer cell line Capan-1 [in which retrovirally expressed Cx32-WT fails to form GJs because of endocytosis prior to GJ formation (Johnson et al., 2013)] to examine whether their expression would inhibit the endocytosis of Cx32-WT. To test this notion, we first engineered Myc-tagged Cx32-WT (Cx32-WT–Myc), Cx32-LL (Cx32-LL–Myc) and Cx32-LR (Cx32-LR–Myc) constructs (Fig. S3E). Transient expression of Cx32-LL–Myc or Cx32-LR–Myc in Cx32-WT-expressing Capan-1 cells resulted in the formation of GJs whereas expression of Cx32-WT–Myc had no effect (Fig. S3F). Gap junctions were formed in all cells expressing Cx32-LL–Myc (n=47) or Cx32-LR–Myc (n=42) but not in cells expressing Cx32-WT–Myc (n=39) in two independent experiments. As assessed by expressing Cx32-WT–Myc, Cx32-LL–Myc or Cx32-LR–Myc retrovirally in LNCaP and BxPC3, the addition of Myc tag had no effect on the ability of the tagged proteins to assemble into GJs as Cx32-LL–Myc and Cx32-LR–Myc formed larger GJs than Cx32-WT–Myc (Fig. S3G).

To examine whether the dileucine motifs LL and LR govern GJ formation directly by regulating the endocytosis of Cx32, we expressed Cx32-WT, Cx32-LL and Cx32-LR in Capan-1 cells retrovirally. As is evident from immunocytochemical analysis, Cx32-LL and Cx32-LR formed GJs whereas Cx32-WT did not (Fig. 5B). To test whether Cx32-WT is endocytosed via the LL or LR motifs by the clathrin-mediated pathway, we transiently expressed an mEGFP-tagged µ2 (a subunit of the AP-2 complex; μ2–GFP) in HEK293T cells, which do not express Cx32, along with Cx32-WT, Cx32-LI, Cx32-LL, Cx32-LR or Cx32-LL-LR. Results showed that µ2–GFP co-immunoprecipitated with Cx32-WT, Cx32-LI, Cx32-LL or Cx32-LR but not with Cx32-LL-LR (Fig. 5C). These data suggest that both the LL and LR motifs interact with AP-2 complex directly or indirectly, and mediate the endocytosis of Cx32, and rendering these motifs nonfunctional inhibits endocytosis. Collectively, the data shown in Fig. 5 and Fig. S3 document that the motifs LL and LR regulate the endocytosis of Cx32, whereas motif LI does not, and that inhibition of endocytosis induces the formation of GJs.

The dileucine motifs govern not only the endocytosis of transmembrane proteins but also intracellular sorting (Bonifacino, 2014; Bonifacino and Traub, 2003; Traub and Bonifacino, 2013). Because Cx32-LI largely remained in the cytoplasm as discrete vesicular puncta (Fig. 3) and was biotinylated poorly (Fig. 4A), we investigated whether it was missorted. Hence, we examined the secretory itinerary of Cx32-WT and Cx32-LI. We used the following markers for various subcellular compartments: calreticulin, an endoplasmic reticulum (ER)-resident protein (Sonnichsen et al., 1994); ERGIC53, an ER–Golgi intermediate compartment marker; β-cop (also known as COPB1) and Sec31A, markers for COPI- and COPII-containing vesicles, respectively (Brandizzi and Barlowe, 2013; Hauri et al., 2000); GM130 (also known as GOLGA2), a cis-Golgi marker (Nakamura et al., 1995); and TGN46 (also known as TGOLN2), a marker for the TGN (De Matteis and Luini, 2008). Neither Cx32-WT nor Cx32-LI colocalized discernibly with any of the secretory markers under steady-state conditions both in LNCaP and BxPC3 cells (Fig. S4A,B). One plausible explanation might be the rapid transit of Cx32 along the secretory pathway. Hence, we used pharmacological approaches to define their secretory itinerary. We used brefeldin A to block trafficking between the ER and Golgi (Klausner et al., 1992) and monensin to block trafficking between the TGN and cell surface (Tartakoff, 1983). We observed a significant colocalization of Cx32-WT or Cx32-LI with TGN46 upon monensin treatment (Fig. 6A). Measurement of Pearson's colocalization coefficient corroborated the immunocytochemical data (Fig. 6B). Neither Cx32-WT nor Cx32-LI colocalized discernibly with calreticulin and GM130 upon treatment with monensin and brefeldin A, respectively (Fig. S5A,B). For these experiments, retention of E-cadherin in the ER, cis-Golgi, and TGN upon treatment with brefeldin and monensin served as a positive control (Fig. S5C).

Previous studies showed that Cx43 was directly sorted from early secretory compartments to lysosomes for degradation (Qin et al., 2003). Hence, we tested whether Cx32-LI was preferentially sorted to lysosomes. To test this notion, we first examined the colocalization of Lamp1 with Cx32-WT, Cx32-LI, Cx32-LL or Cx32-LR immunocytochemically under steady-state conditions as well as upon treatment with leupeptin, which inhibits the degradation of proteins in the lysosomes. Our results showed that discernible colocalization was observed only upon treatment with leupeptin, and there was no difference in the extent of colocalization between Cx32-WT and mutants with Lamp1 (Fig. S6A,B). Next, we examined the degradation of Cx32-WT and mutants by western blot analysis (Fig. S6C). The results showed that leupeptin inhibited the degradation of Cx32-WT, Cx32-LI, Cx32-LL and Cx32-LR equally, suggesting that Cx32-LI was not preferentially degraded in the lysosomes. Taken together, the data in Fig. 6 and Figs S4, S5 and S6 suggest that Cx32-LI traffics normally to the TGN and is not preferentially sorted to lysosomes.

Since Cx32-LI did not mis-sort along the secretory pathway, we examined whether it was rapidly endocytosed by the clathrin-mediated pathway into Rab5-positive vesicles. Immunocytochemical analysis showed that Cx32-LI colocalized discernibly with clathrin in both LNCaP and BxPC3 cells (Fig. 7A). Moreover, we observed more Cx32-LI in Rab5-positive vesicles compared to Cx32-WT upon expressing Rab5DA–GFP in LNCaP cells (Fig. 7B). The mean number of Cx32-LI puncta per Rab5-positive vesicle was 4- to 5-fold higher than the mean number of Cx32-WT puncta (Fig. 7C). These data suggest that Cx32-LI is endocytosed more robustly by the clathrin-mediated pathway into Rab5-positive vesicles. Therefore, we examined whether inhibiting endocytosis would enhance the assembly of Cx32-LI into GJs. To test this notion, we subjected cells to K⁺ depletion or hypertonic sucrose. Both treatments inhibit clathrin-mediated endocytosis (Hansen et al., 1993; Heuser and Anderson, 1989; Larkin et al., 1986). Alexa Fluor 488-conjugated epidermal growth factor (EGF) was used as a positive control for endocytosis. Depletion of K⁺ or treating cells with hypertonic sucrose for 2 h induced the assembly of Cx32-LI into GJs, as assessed immunocytochemically through the appearance of GJ like puncta at cell–cell interfaces, with a concomitant 2–3-fold decrease in the number of intracellular puncta (Fig. 7D,F,G). As expected, Alexa Fluor 488-conjugated EGF was internalized in control cells but not in cells subjected to K⁺ depletion or hypertonic sucrose (Fig. 7D). Next, we used cell surface biotinylation assay to corroborate the immunocytochemical data. Hypertonic sucrose treatment or K⁺ depletion increased the cell surface expression of Cx32-LI profoundly. The biotinylation of E-cadherin was not discernibly affected in cells expressing Cx32-LI (Fig. 7E). Quantification of the data from three independent experiments showed that K⁺ depletion or hypertonic sucrose treatment increased cell surface expression of Cx32-LI by 2–5-fold (data not shown). These findings suggest that Cx32-LI traffics normally to the cell surface but is endocytosed prior to GJ formation by the clathrin-mediated pathway.

Our previous studies showed that expression of endocytosis-defective Cx43 mutant restored the defective GJ assembly of endogenously expressed Cx43 in BxPC3 cells (Johnson et al., 2013). Moreover, restoration of the defect in assembly occurred through inhibition of endocytosis upon formation of heteromers between an endogenously expressed Cx43 and an endocytosis-defective Cx43 mutant (Johnson et al., 2013). We used a similar experimental approach to determine whether expression of endocytosis-defective Cx32-LL and Cx32-LR would inhibit the endocytosis of Cx32-LI. Hence, we determined whether expression of Cx32-LL or Cx32-LR in trans would inhibit the endocytosis of Cx32-LI. To test this notion, we engineered Myc-tagged Cx32-LI (Cx32-LI–Myc) and GFP-tagged Cx32-LI (Cx32-LI–GFP), Cx32-LL (Cx32-LL–GFP) and Cx32-LR (Cx32-LR–GFP) constructs (Fig. S7A). Transient expression of Cx32-LI–Myc along with Cx32-LL–GFP or Cx32-LR–GFP, but not with Cx32-LI–GFP, induced the assembly of Cx32-LI–Myc into GJs in HEK293T cells. Transient expression of Cx32-WT–GFP induced the assembly of Cx32-LI–Myc partially and only in cells where it was expressed in large excess of Cx32-LI–Myc (Fig. S7B, top right panel). Moreover, Cx32-LI–Myc colocalized nearly completely with Cx32-LI–GFP, Cx32-WT–GFP, Cx32-LL–GFP or Cx32-LR–GFP (Fig. S7B). We used HEK293T cells for these studies because of the poor transfection efficiency of BxPC3 and LNCaP cells. By expressing GFP-tagged Cx32-WT or various mutants in BxPC3 cells, we found that addition of GFP had no effect on the ability of various mutants to assemble into larger GJs compared to Cx32-WT (data not shown), as was observed with Myc-tagged Cx32-WT and mutants (see Fig. S3G). To test whether Cx32-LI–Myc formed heteromers with GFP-tagged Cx32-WT, Cx32-LL or Cx32-LR, we transiently co-expressed Cx32-LI–Myc along with Cx32-WT–GFP, Cx32-LL–GFP or Cx32-LR–GFP in HEK293T cells. The results showed that Cx32-LI–Myc and GFP-tagged Cx32-WT, Cx32-LL or Cx32-LR co-immunoprecipitated each other (Fig. S7C). Taken together, these data suggest that Cx32-LI–Myc likely forms heteromeric connexons with GFP-tagged Cx32-WT, Cx32-LL or Cx32-LR, and that motifs LL and LR promote its endocytosis. Inhibiting endocytosis permits Cx32-LI to form GJs.

In the next series of experiments, we directly examined whether Cx32-LI is endocytosed through the LL or the LR motif. Hence, we created a quadruple mutant Cx32-LI-LL in which the leucine and isoleucine in the LI motif and both leucine residues in LL motif were mutated into alanine residues. Similarly, we also created the quadruple mutant Cx32-LI-LR in which leucine and isoleucine in the LI motif and leucine and arginine residues in the LR motif were mutated into alanine residues (Fig. 8A). These quadruple mutants as well as Cx32-LI and Cx32-WT were retrovirally expressed in LNCaP and BxPC3 cells in parallel, and their expression was assessed by western blotting (Fig. 8B) and their ability to form GJs was assessed immunocytochemically (Fig. 8C). The results showed that rendering the LL or LR motif alone nonfunctional overrode the effect of motif LI in inducing the endocytosis of Cx32, conferring upon it the ability to form GJs. As assessed by the decrease in the number of intracellular puncta, motif LL appeared to be more effective in inducing the endocytosis of Cx32 than motif LR (Fig. 8D). To test whether motifs LL and LR collaborate in inducing the endocytosis of Cx32, we created a mutant Cx32-LI-LL-LR in which all leucine residues as well as isoleucine and arginine were mutated into alanine residues (Fig. 8A). We expressed Cx32-LI-LL-LR and Cx32-WT retrovirally in LNCaP and BxPC3 cells and examined GJ formation and expression. We found that the Cx32-LI-LL-LR formed the largest GJs (Fig. 8C). The mean area of GJs formed of Cx32-LI-LL-LR was 1.5-fold larger than those formed by Cx32-WT, Cx32-LI-LL and Cx32-LI-LR, but was not larger than those formed of Cx32-LL and Cx32-LR alone (Fig. 8E). Taken together, the data shown in Fig. 8 and Fig. S7 document that motifs LL and LR regulate the endocytosis of Cx32-LI.

We discovered one tyrosine-based YxxØ-type and two dileucine-based [DE]xxxL[LI]-type motifs in the C-terminal tail of Cx32 (Thévenin et al., 2013). Tyrosine- and dileucine-based motifs mediate not only the endocytosis but also the basolateral sorting of transmembrane proteins (Bonifacino and Traub, 2003; Traub and Bonifacino, 2013). Here, we explored the role of the two dileucine-based motifs in regulating the trafficking, endocytosis and assembly of Cx32 into GJs in human pancreatic and prostate cancer cell lines. One motif, designated as LI, was located near the juxtamembrane domain in the C-terminal tail of Cx32 whereas the other motif designated as LL was located distally (Fig. 2). Moreover, we fortuitously discovered an additional putative motif, designated here as LR, when we randomly mutated leucine and arginine at position 263 and 264 to serve as a control for examining the role of the LI and LL motifs in regulating the intracellular sorting and assembly of Cx32 into GJs (Fig. 2). Our results showed that rendering the LL or LR motif nonfunctional enhanced the formation of GJs profoundly whereas rendering the LI motif nonfunctional had the opposite effect (Fig. 3). Moreover, these motifs played no discernible role in governing the intracellular sorting of Cx32 or its transport from the TGN to the cell surface. Furthermore, rendering the LL or LR motif nonfunctional controlled the formation of GJs by inhibiting the endocytosis of Cx32 through the clathrin-mediated pathway into Rab5-positive vesicles, whereas rendering the LI motif nonfunctional regulated GJ formation by augmenting its endocytosis via the LL and LR motifs.

The dileucine mutant Cx32-LL and mutant Cx32-LR formed larger GJs than did Cx32-WT (Fig. 3A,C), with a concomitant decrease in the number of GJs per cell–cell interface (Fig. 3D). The cell surface biotinylation data suggested that Cx32-LL and Cx32-LR trafficked normally to the cell surface (Fig. 4A) but degraded with a slower kinetics (Fig. 4B,C). These findings rule out the role of LL and LR motifs in controlling the trafficking of Cx32 from TGN to cell surface, and implicate their role in regulating its endocytosis and/or degradation at the cell surface to affect GJ formation. How GJs nucleate remains poorly understood (Falk et al., 2016; Thévenin et al., 2013). However, it is well-established that, after nucleation, the subsequent growth of a GJ plaque occurs upon recruitment of connexons to its periphery by diffusion (Gaietta et al., 2002; Lauf et al., 2002; Simek et al., 2009; Thomas et al., 2005) or upon direct delivery of connexons to its vicinity (Shaw et al., 2007). Therefore one plausible explanation for the formation of large GJs by Cx32-LL and Cx32-LR might be that they are less efficiently endocytosed – both as connexons from non-junctional membranes and as cell–cell channels from the center of the plaque – which allows them to diffuse to the plaque periphery and incorporate into plaque to increase its size. Three independent lines of evidence support this line of thought. First, compared to Cx32-WT, which colocalized discernibly with clathrin and was robustly internalized into Rab5-positive vesicles (Fig. 1A,B), mutants Cx32-LL and Cx-LR did not, and were rarely seen in Rab5-positive vesicles (Fig. 5A; Fig. S3C,D). Second, knockdown of clathrin in cells expressing Cx32-WT increased the size of GJs whereas knockdown had no discernible effect on the size of GJs composed of Cx32-LL or Cx32-LR (Fig. S3B). Previous studies have also shown that knockdown of clathrin reduces endocytosis of Cx43 (Gumpert et al., 2008). Third, transient expression of Cx32-LL or Cx32-LR in Capan-1 cells expressing Cx32-WT, in which it fails to form GJs due to its endocytosis (Johnson et al., 2013), induced the formation of GJs (Fig. S3E–G). Finally, when expressed retrovirally in Capan-1 cells, Cx32-LL or Cx32-LR formed GJs (Fig. 5B). These findings suggest that the LL or LR motif govern the endocytosis of Cx32 by the clathrin-mediated pathway, and inhibiting endocytosis is sufficient to induce the formation of GJs.

Previous studies showed that two distinct tyrosine-based motifs controlled the assembly of Cx43 into GJs cooperatively (Fong et al., 2013). Our results with Cx32 show that rendering the LL or LR motif nonfunctional enhanced the assembly of Cx32 into GJs equally, and there seemed to be no cooperativity between the LL and the LR motifs. In contrast, rendering the LI motif non-functional inhibited the formation of GJs profoundly (Fig. 3A,C). Moreover, our results with quadruple mutants Cx32-LI-LL and Cx32-LI-LR showed that mutating the LL or LR motif alone was sufficient to override the effect of LI motif in inhibiting assembly of Cx32 into GJs (Fig. 8). Furthermore, the mutant Cx32-LI-LL-LR, in which all motifs are rendered nonfunctional, formed the largest GJs (Fig. 8E), suggesting that both the LL and the LR motifs are likely used cooperatively to regulate endocytosis when the LI motif is rendered non-functional. Thus, our data implicate a complex crosstalk among the LI, LL and LR motifs in regulating the endocytosis of Cx32 and its subsequent ability to form GJs, and imply that mechanisms must exist to selectively activate or inactivate these motifs. Our ongoing studies are exploring the mechanisms involved in activating or inactivating these motifs in response to various physiological stimuli in LNCaP and BxPC3 cells.

The most surprising findings regarding the role of motif LI in controlling the trafficking and endocytosis of Cx32 was the failure of Cx32-LI to assemble into GJs and its intracellular accumulation as discrete, detergent-soluble vesicular puncta (Fig. 3A; Fig. S1). Moreover, Cx32-LI was also biotinylated poorly (Fig. 4A). The dileucine-based [DE]xxxL[LI]-type motifs not only interact with the AP-2 complex to govern endocytosis but also with the AP-1, AP-3 and AP-4 complexes to regulate sorting from the TGN to cell surface, retrograde transport from endosomes to the TGN, and sorting from the TGN to lysosomes (Apodaca et al., 2012; Bonifacino and Traub, 2003; Traub and Bonifacino, 2013). Previous studies have demonstrated direct sorting of Cx43 from the early secretory compartment to lysosomes (Qin et al., 2003) and showed multiple routes by which different Cxs trafficked to cell surface (George et al., 1999; Martin et al., 2001). However, our results showed that Cx32-LI trafficked normally to the TGN (Fig. 6A) and was trapped neither in secretory vesicles (Figs S4 and S5) nor was preferentially sorted to lysosomes (Fig. S6). Moreover, inhibiting endocytosis through K⁺ depletion or hypertonic sucrose increased its amenability to cell surface biotinylation and induced its assembly into GJs (Fig. 7D,E). These findings suggest that the dileucine-based LI motif in Cx32 neither acts as an endocytic motif nor controls its trafficking from ER to lysosomes, from the TGN to lysosomes and from the TGN to the cell surface.

The analysis of LI motif revealed it to be a hybrid motif in which the tyrosine-based YLII motif, consensus sequence YxxØ, overlapped with the dileucine-based motif EVVYLI, consensus sequence [DE]xxxL[LI] (see Fig. 2B). Both types of motifs mediate not only sorting of proteins between different endosomal compartments but also trafficking to the cell surface as well as their endocytosis in a contextual and cell-type specific manner (Bonifacino, 2014; Bonifacino and Traub, 2003; Carvajal-Gonzalez et al., 2012; Jain et al., 2015; Traub and Bonifacino, 2013). Specifically, the YxxØ-type motifs are recognized differentially at more than one site in the cell and mediate rapid internalization from the cell surface into endosomes as well as intracellular sorting to lysosomes and lysosome-related organelles (Marks et al., 1997, 1996; Traub and Bonifacino, 2013). Therefore, it is possible that rendering the LI motif non-functional activates the tyrosine-based YLII motif which acts in concert with the LL or LR motif to induce endocytosis of Cx32 by the clathrin-mediated pathway into Rab5-positive vesicles (Bonifacino and Traub, 2003). Several lines of evidence support this interpretation. First, Cx32-LI colocalized significantly with clathrin (Fig. 7A) and was internalized robustly into Rab5-positive endosomes compared to Cx32-WT (Fig. 7B,C). Second, co-expression of Cx32-LI–Myc along with Cx32-WT–GFP, Cx32-LL–GFP or Cx32-LR–GFP in HEK293T cells resulted in the assembly of Cx32-LI–Myc into GJs whereas co-expression of Cx32-LI-GFP had no effect (Fig. S7). Third, rendering the LL or LR motif nonfunctional in Cx32-LI, as in mutants Cx32-LI-LL and Cx32-LI-LR, restored GJ formation (Fig. 8C, compare column 2 with columns 3 and 4). Fourth, rendering both the LL and LR motifs nonfunctional, as in mutant Cx32-LI-LL-LR, not only enhanced the formation of GJs profoundly (Fig. 8C, column 5) but also reduced intracellular accumulation (Fig. 8D), suggesting that this mutant was barely endocytosed. Fifth, when expressed retrovirally in Capan-1 cells, in which Cx32-WT does not form GJs, Cx32-LL and Cx32-LR were assembled into distinct GJs (Fig. 5B). Finally, only Cx32-LL-LR failed to interact directly or indirectly with µ2-GFP (Fig. 5C).

It is noteworthy that the cytoplasmic tail of aquaporin 4 also harbors both the tyrosine-based and the dileucine-based sorting signals. The tyrosine-based motif regulates endocytosis of aquaporin 4 whereas the dileucine-based motif regulates its trafficking from endosomes to lysosomes (Madrid et al., 2001). Multiple dileucine-like motifs are utilized by vesicular glutamate transporter 1 for faster and multiple recycling pathways (Foss et al., 2013). In addition, clathrin-mediated endocytosis of activated epidermal growth factors is regulated by multiple mechanisms that are both redundant and cooperative (Goh et al., 2010). Similarly, the juxtamembrane domain of classical cadherins harbors multiple endocytic motifs, which are used to modulate cell–cell adhesion and adherens junction formation by regulating their trafficking, basolateral sorting and endocytosis in a contextual and cell-type specific manner (Cadwell et al., 2016; West and Harris, 2016). It is also possible that the LI motif regulates the trafficking and basolateral sorting of Cx32 from the TGN to the cell surface only in polarized and well-differentiated cells.

In essence, our studies have shown that multiple motifs govern the endocytosis of Cx32 and its ability to form GJs. The two canonical dileucine-based motifs LI and LL play no discernible role in regulating intracellular trafficking of Cx32 or its transport from the TGN to the cell surface but regulate its endocytosis. The motif LL is the endocytic motif whereas motif LI is not. Our studies have also identified a non-canonical motif LR fortuitously, and have shown it to be an endocytic motif. The assembly of Cx32 into GJs is profoundly enhanced upon rendering the LL or LR motif nonfunctional whereas rendering the LI motif non-functional facilitates endocytosis via the LL or LR motif. Our studies have shown that these three motifs play distinct as well as overlapping role in regulating the endocytosis of Cx32. How these motifs are utilized to regulate endocytosis of Cx32 to induce the formation of GJs in response to hormones and other physiological stimuli in various cell types, including polarized cells, remains to be explored.

Human pancreatic cancer cell line, BxPC3 (ATCC CRL-1687) and prostate cancer cell line, LNCaP (ATCC CRL 1740), were grown in RPMI 1640 (GIBCO) containing 5% fetal bovine serum (Sigma-Aldrich) in an atmosphere containing 5% CO₂ at 37°C as described previously (Johnson et al., 2013; Katoch et al., 2015; Mitra et al., 2006). EcoPack and PTi67, the two retroviral packaging cell lines, were also obtained from ATCC and were grown as described previously (Mehta et al., 1999; Mitra et al., 2006). BxPC3 and LNCaP cells were infected with various recombinant retroviruses and pooled polyclonal cultures from ∼2000 colonies were grown and maintained in RPMI containing G418 (200 µg/ml) (see next section).

The source of the retroviral vector LXSN has been described in our earlier studies (Chakraborty et al., 2010; Mitra et al., 2006). Wild-type rat Cx32 and its various mutants were cloned into pcDNA3.1 and pLXSN by using PCR cloning and standard recombinant DNA protocols. Site-directed mutagenesis was used to generate mutants with the QuikChange kit (Stratagene) according to manufacturer's instructions. To generate wild-type Cx32 and its various mutants tagged in-frame with monomeric GFP, we used plasmid pAcGFP-N1 (Clontech) and employed the strategies described in our previous studies (Johnson et al., 2013; Katoch et al., 2015). To generate wild-type Cx32 and its various mutants tagged in-frame with Myc, the coding sequence for the Myc tag was incorporated in the reverse primer for Cx32 in the PCR-based cloning. Recombinant DNA constructs were verified by DNA sequencing (ACGT Inc). Rab5 and its constitutively active mutant Rab5-Q67L tagged with GFP, designated here as Rab5DA–GFP, were gifts from Dr Steve Caplan (Department of Biochemistry and Molecular Biology, University of Nebraska Medical Center, Omaha, NE). Recombinant retroviruses were produced in EcoPack and PTi67 packaging cell lines. BxPC3 and LNCaP cells were multiply (two to four times) infected with various recombinant retroviruses produced from PTi67 cells and selected in G418 (400 μg/ml) as described previously (Chakraborty et al., 2010; Govindarajan et al., 2010; Mehta et al., 1999). Pooled cultures from ∼2000 colonies obtained from two to four dishes were expanded, frozen and maintained in selection medium containing G418 (200 µg/ml). Pooled polyclonal cultures were used within three to five passages for immunocytochemical and biochemical analyses.

A mouse hybridoma M12.13 secreting monoclonal antibody against rat Cx32 was a gift from Dr Dan Goodenough (Harvard University, Cambridge, MA). This monoclonal antibody recognizes residues 111–125 of rat, mouse and human Cx32. In some experiments, we also used two rabbit polyclonal antibodies raised against residues 106–124 and 262–279 of the C-terminal tail of Cx32 (Sigma; C-3595 and C-3470). Other rabbit polyclonal antibodies used in this study are as follows: anti-EEA1 (PA1-063) and anti-β-COP (PAI-061) from Affinity BioReagents, mouse monoclonal anti-GFP (11814460001, Roche) and rabbit polyclonal anti-GFP (2956S, Cell Signalling); anti-GM130 (610822, BD Transduction Laboratory); anti-TGN46 (ab5095, Abcam); anti-ERGIC53/P58 (E-1031) and anti-β-catenin (C2206) from Sigma. Other mouse monoclonal antibodies used in this study are as follows: anti-Sec31A (612350), anti-caveolin-1 (610058) and anti-EEA-1 (610457) from BD Transduction Laboratories; anti-α-adaptin (3B5) and anti-Lamp1 (H4A3) from Developmental Hybridoma; anti-c-Myc (MMS-150P, Covance); anti-E-cadherin, anti-β-catenin and anti-α-catenin (gifts from Dr Keith R. Johnson, University of Nebraska Medical Center, Omaha, NE). The dilutions for the primary antibodies used are shown in Table S1. For immunostaining, 3×10⁵ BxPC3 and 3×10⁵ LNCaP cells were seeded on glass coverslips in six-well clusters and allowed to grow for 72 h, after which they were fixed with 2% paraformaldehyde and immunostained as described previously (Chakraborty et al., 2010; Govindarajan et al., 2010; Johnson et al., 2013; Katoch et al., 2015). Anti-rabbit-IgG and anti-mouse-IgG secondary antibodies, conjugated to Alexa Fluor 488 or Alexa Fluor 594 (Invitrogen), were used as appropriate. The antibodies used in this study have been used by us for the past several years. Some of these antibodies have been produced by us and have been routinely authenticated by other investigators. Other antibodies used in this study were obtained from established vendors and the protein recognized by each of the purchased antibody was verified using cell lysates from the tissue or cell line that expresses the protein abundantly. We also used the negative controls from cell lines that have been previously characterized not to express the protein of interest.

Fluorescent images of immunostained cells were acquired with a Leica DMRIE microscope (Leica Microsystems, Wetzler, Germany) using a 63× oil objective (NA 1.35). This microscope is also equipped with a Hamamatsu ORCA-ER2 CCD camera (Hamamatsu City, Japan). Colocalization was measured in z-stacked images taken 0.3 µm apart using the commercial image analysis program Volocity 6.0.1 (Improvision, Lexington, MA) as described previously (Chakraborty et al., 2010; Govindarajan et al., 2010; Johnson et al., 2013; Katoch et al., 2015). Saturation of the detector, which would alter background adjustment, was prevented by minimizing exposure of each fluorescently tagged antibody during acquisition. A background-corrected Pearson's correlation coefficient was used to determine fluorescence colocalization as described previously (Barlow et al., 2010). For SIM, we used an ELYRA PS.1 Super Resolution System (Carl Zeiss) with a 63× oil objective (NA 1.4). Images were captured with a z-plane thickness of 90 nm and processed with the Zen Blue software (Carl Zeiss). The processed images were imported into Volocity, cropped as appropriate, and exported as TIFF files.

The surface area of a GJ plaque was measured as follows. After acquiring images of immunostained cells, serial z-sections (0.3 µm) were collected and subjected to iterative volume deconvolution by using Volocity image-processing software (Improvision). The de-convolved images of single optical sections were used for measuring GJ size. Each distinct fluorescent punctum at the cell–cell contact site, delineated by E-cadherin immunostaining, was considered as one GJ plaque. The area of each GJ plaque was calculated by drawing a region of interest (ROI) around each punctum using the free-hand ROI tool of the ‘Measurement’ module of Volocity. The area of each GJ is represented as the ‘pixel count’ where one pixel corresponded to 0.01 μm². In each captured image, three to five puncta were randomly chosen for the measurement of area and 5–15 images were used. We used the ‘Extended focus’ display option, which merges all captured z-planes, for determining the number of GJ plaques per cell–cell interface. The number of distinct visible puncta along the cell–cell interfaces of two adjoining cells were then counted. We chose only distinct interfaces at random to avoid ambiguity in localization of the puncta. Typically 60–70 interfaces were measured from three independent experiments.

BxPC3 (3×10⁶) and LNCaP (2×10⁶) cells, seeded in replicate 10 cm dishes in 10 ml of complete medium, were grown for 72 h. Cell lysis, detergent solubility in 1% Triton X-100 (TX100) and western blot analysis were performed as described previously (Chakraborty et al., 2010; Govindarajan et al., 2010; Katoch et al., 2015; Mitra et al., 2006). Normalization was based on equal cell number for the analysis of detergent-soluble and -insoluble fractions as determined by SDS-PAGE analysis of cell lysates.

For the cell surface biotinylation, BxPC3 (5×10⁵) and LNCaP (4×10⁵) cells expressing Cx32-WT and various mutants were seeded in 6 cm dishes in replicates and grown to 70–80% confluence. The biotinylation reaction was carried out at 4°C for 1 h with freshly prepared EZ-LinkSulfo-NHS-SS Biotin reagent (Pierce) at 0.5 mg/ml in phosphate-buffered saline (PBS) supplemented with 1 mM CaCl₂ and 1 mM MgCl₂. The affinity precipitation of biotinylated proteins was from 200 μg of total protein using 100 µl of streptavidin-agarose beads (Pierce) on a rotator overnight at 4°C. SDS-PAGE followed by western blotting was used to resolve the streptavidin-bound biotinylated proteins after elution as described previously (Chakraborty et al., 2010; Govindarajan et al., 2010, 2002). The kinetics of degradation of cell-surface-associated Cx32-WT and mutants was determined as follows. After biotinylation, washing and quenching biotin, biotinylated proteins were chased at 37°C for various times before affinity precipitation with streptavidin as described previously (Govindarajan et al., 2010; Johnson et al., 2013; Katoch et al., 2015). The protein concentration was determined by using the BCA reagent (Pierce).

For co-immunoprecipitation experiments, HEK293T cells were grown in 6-cm dishes to 80% confluence and transfected with 5 µg of each plasmid for co-transfection studies. At 24 h post transfection, the cells were harvested and lysed in a non-denaturing Tris-NP40 lysis buffer (10 mM Tris-HCl pH 8.0, 0.5% NP-40 and 1 mM EDTA) supplemented with 1 mM PMSF, 2 mM Na₃VO₄, 1 mM NaF and a protease inhibitor cocktail. Following lysis and protein estimation, 500 µg of total protein was incubated with antibodies and mixed overnight at 4°C. Next day, anti-mouse- (Sigma) or anti-rabbit-IgG affinity gel (MP Biomedicals) was washed with Tris-NP40 buffer once. The cell extract–antibody mix was incubated with IgG affinity gel and mixed at 4°C for 2 h. The immune complexes were washed four times with wash buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl and 0.25% Tween 20). After the final wash, the beads were suspended in 2× Laemmli sample buffer and incubated at 37°C for 45 min to elute bound proteins. Proteins were then analyzed by SDS-PAGE and western blotting.

BxPC3 (4×10⁵) and LNCaP (5×10⁵) cells were seeded on glass cover slips in six-well clusters 24 h prior to transfection. Cells were transfected with various plasmids in duplicate with XtremeGENE (Roche Diagnostics) according to the manufacturer's instructions. A total of 2 μg of plasmid DNA per well was used for transfection. When cells were to be co-transfected with two plasmids, 1 µg of each plasmid was used. Expression was analyzed at 24 h post transfection after fixing and immunostaining cells with the desired antibodies as described previously (Govindarajan et al., 2010; Johnson et al., 2013; Katoch et al., 2015; Mitra et al., 2006). For transfection of HEK293T for immunoprecipitation studies, replicate 6 cm dishes were transfected with 2.5 µg of two plasmids, and cells were lysed and immunoprecipitated as described above.

For hypertonic sucrose and K⁺ depletion treatments, 3×10⁵ BxPC3 cells expressing Cx32-WT and Cx32-LI were seeded on glass cover slips in six-well clusters and allowed to grow to confluence for 72 h. Cells were incubated for 1 h at 37°C in starvation medium (RPMI, 0.5% BSA and 30 mM HEPES, pH 7.5). Cells to be subjected to K⁺ depletion were pre-treated as follows: cells were rinsed three times with K⁺-free buffer (20 mM HEPES, pH, 140 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 5.5 mM D-glucose), followed by incubation for 5 min in hypotonic K⁺-free buffer (K⁺-free buffer diluted 1:1 with water), washed again three times with K⁺-free buffer, and incubated in it for 15 min at 37°C. Coverslips were then transferred to chilled binding medium (starvation medium containing Alexa Fluor 488-conjugated with EGF at 5 µg/ml) and incubated for 1 h at 4°C. Cells were then washed three times with chilled PBS. Cells to be subjected to hypertonic sucrose treatments were incubated in hypertonic sucrose medium (0.45 M sucrose) for 2 h, and cells to be subjected to K⁺ depletion were incubated in warm K⁺-free buffer for 2 h. Control cells were incubated in starvation medium for 2 h. At the end of the treatments, cells were fixed and immunostained for Cx32 as described above.

Stock solutions of various reagents were prepared as follows: leupeptin at 100 mM in water; monensin at 10 mM in ethanol (Calbiochem); brefeldin at 10 mM in DMSO (BIOMOL). These solutions were stored at −20°C in small aliquots. All solutions were appropriately diluted in the cell culture medium at the time of treatment. Controls were vehicle treated.

On-TARGETplus SMARTpool siRNAs with a target sequence against the heavy chain of clathrin (hCLTC) were purchased from Dharmacon (Lafayette, CO). For knockdown studies, LNCaP cells were seeded at a density of 2.5×10⁵ cells in each well of a 12-well cluster. After 24 h, cells were transfected with hCLTC siRNAs using Dharmafect reagent 2 as per the manufacturer's instructions and incubated further for 72 h with transfection reagents in the medium. After removing medium containing transfection reagents, cells were further incubated for additional 24 h in complete RPMI. Cells were fixed and immunostained for Cx32 as described above (see immunostaining).

Gap junctional communication was assayed by microinjecting fluorescent tracers and scrape-loading. The following fluorescent tracers were used for microinjection: Alexa Fluor 488 (molecular mass 570 Da; A-10436) and Alexa Fluor 594 (molecular mass 760 Da; A-10438). Alexa Fluor dyes were obtained as hydrazide sodium salts (Molecular Probes). Stock solutions of all Alexa Fluor dyes were prepared in water at 10 mM. These fluorescent tracers were microinjected into test cells with the Eppendorf InjectMan and FemtoJet microinjection systems (models 5271 and 5242, Brinkmann Instrument, Inc., Westbury, NY) and mounted on Leica DMIRE2 microscope as described previously (Chakraborty et al., 2010; Govindarajan et al., 2010, 2002; Johnson et al., 2013; Katoch et al., 2015; Mehta et al., 1986). Junctional transfer of the fluorescent tracer was quantified by scoring the number of fluorescent cells (excluding the injected one) at 5–10 min after microinjection. Cells were scrape-loaded as described previously (Govindarajan et al., 2010b, 2002). Briefly, BxPC3 cells expressing Cx32-WT or various mutants were seeded in 35 mm dishes at a density of 5×10⁴ cells/well and allowed to grow to confluence. Cell culture medium from freshly confluent cells was removed and replaced with 1 ml of medium containing rhodamine-conjugated fluorescent dextrans (10 kDa, 1 mg/ml; fixable) and Lucifer Yellow (0.25%). Cells were scrape-loaded using a sterile scalpel by making two longitudinal scratches, and were incubated for 1 min at room temperature. Cells were washed quickly two to three times with warm cell culture medium without serum and returned to the incubator at 37°C for 5 min, after which medium was removed, cells were washed twice with PBS and fixed with 3.7% buffered formalin at room temperature. The auto-fluorescence of cells was quenched with 0.1 M glycine for 5 min and, after washing once with PBS, images of the scrape-loaded cells were captured as described above.

Comparisons between two groups were made by using a Student's unpaired two-tailed t-test. Data are expressed as mean±s.e.m. Values of P<0.05 were considered to be statistically significant.

We thank Linda Kelsey for her invaluable technical help. We thank Dr Souvik Chakraborty for helpful discussion and suggestions. We also thank Drs Steve Caplan, Naava Naslavsky and Keith Johnson for helpful suggestions and sharing of antibodies and plasmids. Our special thanks to Dr Matthias Falk and his laboratory members for critically reading this manuscript and helpful comments. We thank Janice A. Taylor and James R. Talaska of the Advanced Microscopy Core Facility at the University of Nebraska Medical Center for providing assistance with super resolution microscopy. We gratefully acknowledge support from the University of Nebraska Medical Center in the form of a graduate fellowship to Anuttoma Ray.