Cx43-Gap Junctions Accumulate at the Cytotoxic Immunological Synapse Enabling Cytotoxic T Lymphocyte Melanoma Cell Killing

Upon tumor antigen recognition, cytotoxic T lymphocytes (CTLs) and target cells form specialized supramolecular structures, called cytotoxic immunological synapses, which are required for polarized delivery of cytotoxic granules. In previous reports, we described the accumulation of connexin 43 (Cx43)-formed gap junctions (GJs) at natural killer (NK) cell–tumor cell cytotoxic immunological synapse. In this report, we demonstrate the functional role of Cx43-GJs at the cytotoxic immunological synapse established between CTLs and melanoma cells during cytotoxicity. Using confocal microscopy, we evaluated Cx43 polarization to the contact site between CTLs isolated from pMEL-1 mice and B16F10 melanoma cells. We knocked down Cx43 expression in B16F10 cells and evaluated its role in the formation of functional GJs and the cytotoxic activity of CTLs, by calcein transfer and granzyme B activity assays, respectively. We found that Cx43 localizes at CTL/B16F10 intercellular contact sites via an antigen-dependent process. We also found that pMEL-1 CTLs but not wild-type naïve CD8+ T cells established functional GJs with B16F10 cells. Interestingly, we observed that Cx43-GJs were required for an efficient granzyme B activity in target B16F10 cells. Using an HLA-A2-restricted/MART-1-specific CD8+ T-cell clone, we confirmed these observations in human cells. Our results suggest that Cx43-channels are relevant components of cytotoxic immunological synapses and potentiate CTL-mediated tumor cell killing.


Introduction
Gap junctions (GJs) are clusters of intercellular channels found at the plasma membrane that allow direct communication between the cytoplasm of adjacenT cells. Each GJ channel is formed by two hexameric hemichannels, also known as connexons, one provided by each of the two contacting cells. In turn, each hemichannel is composed of six transmembrane proteins called connexins (Cx) [1]. Once functional GJs are established, these channels allow the exchange of small molecules (up to ≈1.4 nm) between the cytoplasm of adjacenT cells, including adenosine triphosphate (ATP), cyclic adenosine monophosphate (cAMP), cyclic guanosine monophosphate-adenosine monophosphate (cGAMP), inositol triphosphate (IP 3 ), Ca 2+ , glutamate, microRNAs, and small peptides [2]. The Cx gene family is Previously, we and others have demonstrated that Cx43 channels accumulate at the immunological synapses during DC-T cell and DC-NK cell interactions [15][16][17] and at the NK cell-tumor cell cytotoxic immunological synapse [17,18], allowing DC-mediated T-and NK-cell activation and NK cell-mediated target tumor cell killing, respectively. Given that the cytotoxic immunological synapses formed by NK cells and CTLs share some structural features [19], in this report we aimed to examine whether Cx43 channels have a role in CTL-target tumor cell immunological synapses. CTL cytotoxic immunological synapses occur when the T-cell receptor (TCR) recognizes its cognate antigenic peptide associated with major histocompatibility complex (MHC) class I molecules [19]. In order to test the role of Cx43 in functionally active cytotoxic immunological synapses, we used the murine model pMEL-1 (C57BL6 genetic background), which is characterized by having CD8 + T cells that express a transgenic Vβ13 chain in its TCR that specifically recognizes the peptide gp100 [25][26][27][28][29][30][31][32][33] (XXXRNQDWL) of the gp100 protein bound to H-2 D b molecules. The gp100 protein is considered an MAA, and it is expressed in melanocytic lineage cells and overexpressed in different melanoma cell lines, including the syngeneic B16F10 melanoma cells [20,21]. First, we differentiated pMEL-1 CTLs in vitro from pMEL-1 mice splenocytes by incubating them with interleukin (IL)-2 and the hgp100 [25][26][27][28][29][30][31][32][33] peptide for six days. This procedure allowed the generation of fully differentiated pMEL-1 effector CTLs, characterized by the high expression of activation and memory markers CD25 and CD44 on the CD8 + Vβ13 + T-cell population ( Figure S1A, Supplementary Materials). The cellular distribution of Cx43 was then assessed by confocal microscopy in pMEL-1 CTL/B16F10 cell conjugates. Cx43 was found to preferentially accumulate in the interface between pMEL-1 CTLs and B16F10 cells, which, as indicated by the accumulation of F-actin [19], would correspond to stable cytotoxic immunological synapses ( Figure 1A). When naïve CD8 + T cells isolated from splenocytes of wild-type C57BL6 mice ( Figure S1B, Supplementary Materials) were co-cultured with B16F10 cells, functional cytotoxic immunological synapses were not formed, as expected, and therefore Cx43 channels are not observed at the cell-to-cell interface ( Figure 1B). In addition, Cx43 was not detected in the cell interface of cell conjugates of pMEL-1 CTLs and syngeneic bladder carcinoma MB49 cells, which do not express the MAA gp100 and therefore do not trigger the formation of functional cytotoxic immunological synapses with gp100 25-33 -specific CD8 + T cells ( Figure 1C). Of note, the distribution of Cx43 and F-actin was observed uniformly expressed on the cell surface in unconjugated pMEL-1 CTLs or naïve CD8 + T cells ( Figure 1D,E), and both kinds of T cells showed similar Cx43 expression levels ( Figure 1D,E and Figure S1C, Supplementary Materials). These results indicate that Cx43 polarizes to the cytotoxic immunological synapses of murine CTLs and target tumor cells in an antigen-dependent manner ( Figure 1F). 33-specific CD8 + T cells ( Figure 1C). Of note, the distribution of Cx43 and F-actin was observed uniformly expressed on the cell surface in unconjugated pMEL-1 CTLs or naïve CD8 + T cells ( Figure  1D,E), and both kinds of T cells showed similar Cx43 expression levels ( Figure 1D,E and Figure S1C, Supplementary Materials). These results indicate that Cx43 polarizes to the cytotoxic immunological synapses of murine CTLs and target tumor cells in an antigen-dependent manner ( Figure 1F). (F) Quantification of Cx43 accumulation at the cytotoxic IS as ratio of Cx43 fluorescence intensity in the cell-cell contact site (red arrows) versus the opposite site (white arrows). *** p < 0.001, **** p < 0.0001, versus CTL/B16F10 conjugates; ns, non-significant (one-way ANOVA, Tukey's multiple comparison test); n = approximately 60 cell conjugates by condition, two independent experiments.

pMEL-1 CTLs Form Functional Cx43-GJ-Mediated Intercellular Communications with B16F10 Melanoma Cells
To determine whether pMEL-1 CTLs and target B16F10 melanoma cells can communicate with each other through Cx43 channels upon cytotoxic immunological synapse formation, we performed calcein transfer assays by flow cytometry analysis, as described before [17]. In contrast to MB49 cells, which did not induce Cx43 polarization to the contact site with pMEL-1 CTLs, B16F10 melanoma cells did acquire calcein from pMEL-1 CTLs after 30 min of co-culture (Figure 2A), concomitant with the Cx43 polarization to the cell-to-cell contact site. When we forced the recognition of MB49 cells by pMEL-1 CTLs through the pre-incubation of target tumor cells with the antigenic peptide hgp100 [25][26][27][28][29][30][31][32][33] , CTLs effectively transferred calcein to the MB49 tumor cells ( Figure 2B), indicating that the cell coupling between CTLs and target tumor cells is an antigen-dependent process. In order to test if the cell coupling between pMEL-1 CTLs and B16F10 cells is a Cx43-dependent mechanism, we knocked down the expression of Cx43 in B16F10 melanoma cells using specific anti-Cx43 siRNAs (siCx43). Our results showed that the knocking-down efficiency of Cx43 in these cells was approximately 70%, as compared with Cx43 expression observed in parental (non-transfected B16F10 cells) or B16F10 cells transfected with control-scrambled siRNAs (siScr) ( Figure 2C). In concordance with the localization of Cx43 at the intercellular contact site, pMEL-1 CTLs but not wild-type naïve CD8 + T cells transferred calcein to B16F10 parental cells, and this cell coupling was partially but significatively decreased when Cx43 was silenced in the target tumor cells ( Figure 2D,E). Overall, our results suggest that upon CTL cytotoxic immunological synapse establishment, Cx43 polarizes to the synapse allowing the effector/target cell coupling via Cx43-GJ channels.

Cx43 is Required for GrzmB-Mediated Cytotoxicity of pMEL-1 CTLs Against B16F10 Melanoma Cells
Cx43-GJ channels have been described as mediators of intercellular communications between NK cells and target tumor cells during NK-cell cytotoxic immunological synapse formation. Recent reports have shown that Cx43-channel inhibition during NK-cell immunological synapse  ) with calcein-AM pre-loaded pMEL-1 CTLs, at a 1:5 ratio. Calcein transfer from effector to target tumor cells was assessed by flow cytometry. The bar graph shows Violet BMQC + calcein + cells. (B) Calcein transfer from pMEL-1 CTLs to target tumor cells was evaluated as described before, after 30 min of co-culture. MB49 cells were pre-loaded or not with hgp100 25-33 peptide before co-culturing with pMEL-1 CTLs. The bar graph shows Violet BMQC + calcein + cells as a percentage of the maximum calcein transfer. (C) B16F10 cells were transfected with siRNAs against Cx43 (siCx43) or control-scrambled siRNAs (siScr). The expression of Cx43 and actin was assessed three days after transfection by Western blot in transfected or non-transfected (parental) B16F10 cells, and Cx43/actin ratios were quantified by ImageJ software. The bar graph at the bottom shows the average of Cx43 expression depicted as Cx43/actin ratio relative to parental untransfected cells (n = 5 independent experiments). (D) Representative dot plots showing the strategy for Cx43-GJ communication measuring. Target (parental, siScr-or siCx43-transfected B16F10) cells were pre-loaded with the CellTracker Violet BMQC and co-cultured for 30 min with calcein-AM pre-loaded effector cells (naïve CD8 + T cells or pMEL-1 CTLs), at a 1:5 ratio. Calcein transfer from effector to target tumor cells was assessed by flow cytometry. The numbers in the dot plots represent the percentage of Violet BMQC + calcein + cells. (E) The bar graph shows the cell coupling factor calculated as (%Violet BMQC + calcein + cells × Calcein MFI of Calcein + cells)/100. * p < 0.05; ** p < 0.01; *** p < 0.001; ns = non-significant (one-way ANOVA, Tukey's multiple comparison test). A.U., arbitrary units.

Cx43 is Required for GrzmB-Mediated Cytotoxicity of pMEL-1 CTLs against B16F10 Melanoma Cells
Cx43-GJ channels have been described as mediators of intercellular communications between NK cells and target tumor cells during NK-cell cytotoxic immunological synapse formation. Recent reports have shown that Cx43-channel inhibition during NK-cell immunological synapse establishment decreases GrzmB activity in target tumor cells and NK-cell anti-tumor cytotoxic capacity without affecting the degranulation process [17,18]. As expected, we detected significant levels of GrzmB activity in B16F10 cells, but not in MB49 cells, after 1 or 2 h of co-culture with pMEL-1 CTLs ( Figure 3A). Similarly to the results of Cx43 localization at the cytotoxic immunological synapse and the Cx43-mediated intercellular communication, pMEL-1 CTLs but not wild-type naïve CD8 + T cells were able to induce GrzmB activity on B16F10 parental cells, which was significatively decreased when Cx43 was silenced in the target tumor cells ( Figure 3B,C). establishment decreases GrzmB activity in target tumor cells and NK-cell anti-tumor cytotoxic capacity without affecting the degranulation process [17,18]. As expected, we detected significant levels of GrzmB activity in B16F10 cells, but not in MB49 cells, after 1 or 2 hours of co-culture with pMEL-1 CTLs ( Figure 3A). Similarly to the results of Cx43 localization at the cytotoxic immunological synapse and the Cx43-mediated intercellular communication, pMEL-1 CTLs but not wild-type naïve CD8 + T cells were able to induce GrzmB activity on B16F10 parental cells, which was significatively decreased when Cx43 was silenced in the target tumor cells ( Figure 3B,C).

Cx43-GJ Intercellular Communications are Required for Optimal GrzmB-Mediated Cytotoxicity of Human CTLs Against Melanoma Cells
The contribution of Cx43 channels on the CTL cytotoxic immunological synapse was further investigated using a human cell model. As effector human CTLs, we used an HLA-A2-restricted/MART-1-specific CD8 + T-cell clone (CdL43-1), obtained from a melanoma patient [10]. As positive or negative target tumor cells, we used an HLA-A2 + /MART-1 + human melanoma cell line (Mel1) and a myelogenous leukemia cell line K562 (HLA − ), respectively. We found that, as with murine CTLs, Cx43 polarizes to the cell-cell contact zone of CdL43-1 CTL/Mel1 conjugates, whereas it distributes homogeneously in CdL43-1 CTL/K562 conjugates. This was associated with effective cytotoxic immunological synapse formations, as indicated by F-actin remodeling ( Figure  4A-C). As expected, CdL43-1 CTL did not form stable conjugates with the HLA − K562 cells, as indicated by the low percentage of cell conjugates observed under microscopy ( Figure 4B). The polarization of Cx43 at the cytotoxic immunological synapse of human CTLs also correlates with

Cx43-GJ Intercellular Communications Are Required for Optimal GrzmB-Mediated Cytotoxicity of Human CTLs against Melanoma Cells
The contribution of Cx43 channels on the CTL cytotoxic immunological synapse was further investigated using a human cell model.
As effector human CTLs, we used an HLA-A2-restricted/MART-1-specific CD8 + T-cell clone (CdL43-1), obtained from a melanoma patient [10]. As positive or negative target tumor cells, we used an HLA-A2 + /MART-1 + human melanoma cell line (Mel1) and a myelogenous leukemia cell line K562 (HLA − ), respectively. We found that, as with murine CTLs, Cx43 polarizes to the cell-cell contact zone of CdL43-1 CTL/Mel1 conjugates, whereas it distributes homogeneously in CdL43-1 CTL/K562 conjugates. This was associated with As expected, CdL43-1 CTL did not form stable conjugates with the HLA − K562 cells, as indicated by the low percentage of cell conjugates observed under microscopy ( Figure 4B). The polarization of Cx43 at the cytotoxic immunological synapse of human CTLs also correlates with increased Cx43-mediated cell coupling with target melanoma cells, as indicated by calcein transfer assays and the inhibition of Cx43 channels by the Cx43-mimetic peptide gap27 ( Figure 4D). Similarly to the results of Cx43 localization at the cytotoxic immunological synapse and Cx43-mediated intercellular communication, CdL43-1 CTLs were able to induce GrzmB activity and target tumor cell killing on Mel1 cells but not in K562 cells, which was significatively decreased when Cx43 channels were blocked with gap27 mimetic peptide ( Figure 4E,F).
increased Cx43-mediated cell coupling with target melanoma cells, as indicated by calcein transfer assays and the inhibition of Cx43 channels by the Cx43-mimetic peptide gap27 ( Figure 4D). Similarly to the results of Cx43 localization at the cytotoxic immunological synapse and Cx43-mediated intercellular communication, CdL43-1 CTLs were able to induce GrzmB activity and target tumor cell killing on Mel1 cells but not in K562 cells, which was significatively decreased when Cx43 channels were blocked with gap27 mimetic peptide ( Figure 4E,F).  (C) Quantification of Cx43 accumulation at the cytotoxic IS as ratio of Cx43 fluorescence intensity in the cell-cell contact site (red arrows) versus at the opposite site (white arrows). (D) Target (K562 or Mel1) tumor cells were pre-loaded with the CellTracker Violet BMQC and co-cultured for 1 hour with calcein-AM pre-loaded CdL43-1 CTL effector cells at 1:3 ratios, in the presence of Cx43 mimetic inhibitory (gap27) or control (Scr) peptides (300 µM). Calcein transfer from effector to target tumor cells was assessed by flow cytometry. The bar graph shows the cell coupling factor calculated as (%Violet BMQC + calcein + cells × Calcein MFI of Calcein + cells)/100. (E) Mel1 or K562 cells were pre-stained with TFL4 (CellTracker) and NFL1 (viability marker) and co-cultured for 2 h with CdL43-CTLs, at a 1:3 ratio, in the presence of a permeable fluorogenic substrate for GrzmB (GranToxiLux; GrzmB activity), and gap27 or Scr peptides. GrzmB activity was evaluated on TFL4 + NFL1 − target tumor cells by flow cytometry. The bar graph shows the GrzmB activity in target tumor cells (TFL4 + NFL1 − GranToxiLux + ) as a percentage of the maximum (target tumor cell: Mel1; peptide: Scr). (F) CdL43-1 CTLs were co-cultured with Mel1 or K562 cells at different effector:target cell ratios in the presence of gap27 or Scr peptides. Cytotoxicity was assessed by conventional 51 Cr release assays. The results are plotted as a percentage of specific lysis. All the results are from three independent experiments. ** p < 0.01; **** p < 0.0001 (one-way ANOVA, Tukey's multiple comparison test, or two-tailed Student's t-test (C)).

Discussion
The establishment of anti-tumor immune responses requires strongly coordinated and regulated communications between immune cells and between immune cells and target tumor cells. In this sense, both DC-T cell synapses formed during DC-mediated T-cell activation and CTL or NK cell/target tumor cell synapses, established during the former's effector phases, are supramolecular structures that regulate fundamental intercellular communication mechanisms required for an effective control of tumor cells by the immune system. The mechanisms allowing the establishment of both types of immunological synapses share some molecular features, including the rearrangement of the cytoskeleton, receptors, and adhesion molecules in the intercellular contact zone. In the cytotoxic immunological synapses, those molecular rearrangements ensure the polarized release of perforin-and granzyme-containing cytotoxic granules towards the targeT cells, which finally triggers apoptosis [19,22,23]. Previously, the participation of Cx43 channels as functional structures of immunological synapses formed among DCs and CD4 + T cells, between DCs and NK cells, and among NK cells and target tumor cells has been described [15][16][17][18]. These channels were shown to be necessary for DC-mediated activation of both CD4 + T cells and NK cells, as well as for the efficient NK cell-mediated tumor cell killing. Taking into account the fact that the immunological synapses formed by CTLs and NK cells have functional and structural similarities [19,24,25], we aimed to determine in this report whether Cx43-GJ channels are also involved in the CTL/target tumor cell cytotoxic immunological synapses. To this end, we used CTLs isolated from pMEL-1 mice, which specifically recognize the epitope XXXRNQDWL for the MAA gp100 endogenously expressed by the syngeneic B16F10 melanoma cells, and a human CTL clone specific for the HLA-A2-restricted MART-1 [27][28][29][30][31][32][33][34][35] antigen, an MAA expressed in the HLA-A2 + Mel1 melanoma cell line.
Upon antigen recognition via TCR/cognate antigenic peptide-MHC I complex (pMHC I) interactions, a rearrangement of the actin cytoskeleton occurs at the proximity of the plasma membrane of T cells, leading to the formation of an actin ring structure along the site of the intercellular contact, which is indicative of an efficient cytotoxic immunological synapse formation [19]. We observed a structure resembling actin rings in pMEL-1 CD8 + CTLs upon antigen recognition of target B16F10 melanoma cells (Figure 1) and, only upon this antigen-specific driven process, Cx43 polarizes to the cell-to-cell contact sites, indicating that Cx43 accumulates in the cytotoxic immunological synapses of CTLs. It is known that once Cx proteins are translated, they indirectly bind to the cytoskeleton through interactions with motor proteins, such as myosin in the case of the actin cytoskeleton and kinesins in the case of the microtubule cytoskeleton. Through these interactions, Cxs traffic towards the cell plasma membrane for the formation of hemichannels and GJs [26]. In the present report, we observed that the polarization of Cx43 to the cell-to-cell contact site is concomitant with the F-actin accumulation, exclusively in the antigen-specific condition, suggesting that Cx43 could be mobilized towards the cytotoxic immunological synapses through actin filaments. However, this point needs to be further addressed. Additionally, a previous report indicates that the recruitment of T cell Cx43 to the contact area of CD3-and CD28-coated beads is dependent on the actin cytoskeleton but not on microtubules [15]. Furthermore, it has been reported that Cx43 interacts with the zonula occluden (ZO) proteins ZO-1 and ZO-2 [27]. The interaction of Cx43 with ZO-1 regulates the rate of undocked connexon aggregation into GJs [28]. In addition, ZO-1 and ZO-2 bind to the actin cytoskeleton in the cellular plasma membrane to form tight junctions [29,30] and the participation of ZO-2 in the immunological synapse established among T cells and APCs has recently been described [31]. These data allow us to suggest that Cx43 may interact with ZO-1 and/or ZO-2 for being located in the actin rings at the distal supramolecular activation cluster (dSMAC) of CTL cytotoxic immunological synapses. Of note, we previously demonstrated that Cx43 accumulates preferentially in the dSMAC of immunological synapses established between T cells and DCs [15].
Our results indicate that the antigen-driven polarization of Cx43 at the CTL cytotoxic immunological synapse is associated with an increase in CTL/target tumor cell coupling, measured as Cx43-dependent calcein transfer (Figure 2). Impairment on Cx43-mediated cell coupling, generated by Cx43 knockdown in B16F10 cells, was associated with diminished GrzmB activity in the target tumor cells (Figure 3). This suggests that through a yet uncharacterized mechanism, Cx43-mediated intercellular communications between CTLs and target tumor cells are necessary for an efficient cell death-leading GrzmB activity in target tumor cells, in both murine and human cells (Figures 4 and 5). We speculate that two possible complementary mechanisms may explain the role of Cx43 channels in the activity of CTL cytotoxic immunological synapses. (i) Cx43 channels could act as a form of intercellular adhesion along with other adhesion molecules present in the cytotoxic immunological synapses, as LFA-1 integrin-mediated binding to ICAM-1 on targeT cells [32]. In this context, it has been reported that Cx43 regulates cell adhesion in neurons, glioma cells, B cells, and monocytes [33][34][35]. More recently, it has been observed that chimeric antigen receptor T (CAR-T) cells form potent cytotoxic immunological synapses less reliant on LFA-1 to form stable conjugates than TCR immune synapses [36]. Interestingly, the authors observed that CAR-T cell synapses, whose signaling was stronger and more rapid than in TCR immunological synapses, were more enriched in Cx43, suggesting that Cx43 could be essential for the formation of effective cytotoxic immunological synapses even in the presence of disorganized LFA-1 adhesion rings [36]. (ii) Cx43 channels could allow the transfer of small signaling molecules from CTLs to target tumor cells, which contribute finally to efficient GrzmB activity in the former cell. A signaling molecule that fulfills this requirement is Ca 2+ , whose influx in targeT cells was reported as necessary for NK cell-and CTL-mediated GrzmB-induced apoptosis in targeT cells [37][38][39]. In those reports, it has been shown that perforin and GrzmB are endocytosed in a Ca 2+ -dependent manner in large endosomes, called gigantosomes, within the target cell near the synapse. The authors proposed that the disruption of this gigantosome leads to the release of GrzmB into the cytosol. Therefore, a rise of intracellular Ca 2+ concentrations in targeT cells is required for the efficient CTL and NK cell killing of their targets. It has been proven that pores formed by perforin allow this Ca 2+ influx in targeT cells [37][38][39]. In addition, we have previously shown that the inhibition of Cx43 channels in NK cell/K562 conjugates decreases NK cell-mediated Ca 2+ influx in the target tumor cells [17]. These data suggest that Cx43 channels could collaborate with perforin pores in inducing a Ca 2+ influx and thus play a role in the gigantosome formation and/or contribute to their content release, leading to an efficient GrzmB activity in tumor cells killed by CTLs. Further studies are needed to conclusively demonstrate this hypothesis.  This is the first study, to our knowledge, demonstrating that Cx43-GJ formation is important for CTL-mediated target tumor cell lysis. Our data indicate that reduced Cx43 expression may be a valuable mechanism for the immune evasion of malignant or pathogen-infected cells. Indeed, it is known that many tumor cells downregulate Cx43 expression [6]. Therefore, examining the expression levels of Cx43 in tumors may be an important strategy to design appropriate immune therapeutic treatments dependent on CTL activity.

Mice
Wild-type C57BL6 and transgenic pMEL-1 (C57BL6 background) mice [20] were bred at Universidad de Chile. For all experiments, mice between 8 and 12 weeks of age were bred in specific pathogen-free conditions. All animal experiments were performed in accordance with institutional guidelines for animal care and were approved by the Ethical Review Committees at the Faculty of Medicine of the Universidad de Chile, ethical number: CBA FMUCH 0825 (approval date: 2 October 2015).
Naïve CD8 + T cells were purified from total splenocytes of twelve-week-old C57BL6 wild-type mice by negative selection method using the Naive CD8α + T Cell Isolation Kit (MACS, Miltenyi Biotec), according to the manufacturer s instructions. The phenotype of these cells was analyzed by flow cytometry.

Microscopy
Confocal microscopy of fixed target tumor cell/T cell co-cultures was performed as described previously [18]. Target tumor cells were pre-loaded with 10 µg/mL AlexaFluor 647-conjugated wheat germ agglutinin (WGA; Invitrogen, Paisley, UK) according to the manufacturer s instructions. Then, target tumor and effector cells were co-cultured for 30 min at 37 • C and 5% CO 2 on poly-l-lysine-coated slides, at a 2:1 effector:target ratio, using 6 × 10 5 effector cells per slide in 100 µL of complete culture medium. Cell conjugates were gently washed with PBS twice and fixed with 4% paraformaldehyde for 15 min. After gentle washing with PBS, the cells were incubated in ammonium chloride (50 mM) for 10 min. Then, the cells were permeabilized for 10 min (0.5% Triton X-100 and 0.5% FBS) and blocked with 0.5% bovine serum albumin (BSA) for 15 min. Cells were stained with a 1:500 dilution of FITC-coupled monoclonal antibody anti-Cx43 (D-7; Santa Cruz Biotechnology, Dallas, TX, USA) for 1.25 h at room temperature, and then stained with 1:1000 dilution of Hoechst 33342 (Invitrogen) and 16.5 µM rhodamine-phalloidin (R415, Invitrogen) for 15 min. Cells were analyzed with a C2+ confocal microscope (1000×, Nikon Instruments Inc., Melville, NY, USA) and a Spinning Disk Olympus BX61WI microscope (400×, Center Valley, PA, USA). The recruitment of Cx43 to the cell-to-cell contact site (cytotoxic immunological synapses) was quantified using ImageJ software (version number 1.48v, National Institutes of Health, Bethesda, MD, USA) as previously described [18].

Transfections
Knockdowns of Cx43 were obtained by transfection of B16F10 cells with esiRNA against Cx43 (siCx43; Mission, Sigma-Aldrich, St. Louis, MO, USA), a pool of independent target-specific siRNAs against Cx43. A siRNA of scrambled sequence (siRNA Scrambled; Ambion, Thermo Fisher Scientific, Austin, TX, USA; siScr) without target in the murine genome was used as a negative control. Transfections were performed using the Silencer siRNA Transfection Kit (Ambion, Thermo Fisher Scientific) according to the manufacturer's instructions. Briefly, a transfection complex was generated mixing the siPORT amine transfection agent with the transfection medium OPTIMEM™ (Gibco). Then, the transfection complex was incubated for 10 min with 30 nM of siCx43 or siScr. B16F10 cells were incubated with siRNA-containing transfection complex for 6 h. After this period, the transfection complex was replaced with complete cell-culture medium. After 72 h, cells were harvested and used in different experiments.

Calcein Transfer Assay
GJ-mediated cell coupling was measured using calcein transfer assay as previously described [17]. For this study, pMEL-1 CTLs, wild-type naïve CD8 + T cells, or human CdL43-1 CTLs were loaded with calcein-acetoxymethyl (AM) (0.7 µM; Invitrogen) for 30 min at 37 • C according to the manufacturer's instructions. The membrane-permeable calcein-AM is hydrolyzed by intracellular nonspecific esterase, and the resulting green fluorescent hydrophilic calcein is then trapped inside the cells. B16F10 melanoma cells, MB49 urothelial carcinoma cells, and K562 lymphoma or Mel1 melanoma cells were loaded with the CellTracker Violet BMQC for 10 min (10 µM; Invitrogen), and the reaction was stopped by adding FBS for 1 minute. GJs are permeable to calcein but not to Violet BMQC. MB49 cells were pre-loaded or not with 1 µM of hgp100 [25][26][27][28][29][30][31][32][33] peptide (for 30 min) before co-culturing them with pMEL-1 CTLs. Violet BMQC-stained targeT cells were co-cultured with calcein-stained CTLs (2 × 10 5 cells) for different times and at 1:5 or 1:3 ratios. In the case of human cells, control (Scramble; Scr) or inhibitory (gap 27) peptides were added 15 min prior to the co-cultures and incubated at 37 • C, 5% CO 2 . After co-culture, cells were collected and calcein transfer was analyzed by flow cytometry. When indicated, a cell coupling factor that takes into account the percentage of cells acquiring calcein and the level of calcein acquired by acceptor cancer cells was calculated as Cell coupling factor = (%Violet BMQC + calcein + cells × Calcein MFI of Calcein + cells)/100.