Pathogenic Connexin-31 Forms Constitutively Active Hemichannels to Promote Necrotic Cell Death

Mutations in Connexin-31 (Cx31) are associated with multiple human diseases including erythrokeratodermia variabilis (EKV). The molecular action of Cx31 pathogenic mutants remains largely elusive. We report here that expression of EKV pathogenic mutant Cx31R42P induces cell death with necrotic characteristics. Inhibition of hemichannel activity by a connexin hemichannel inhibitor or high extracellular calcium suppresses Cx31R42P-induced cell death. Expression of Cx31R42P induces ER stress resulting in reactive oxygen species (ROS) production, in turn, to regulate gating of Cx31R42P hemichannels and Cx31R42P induced cell death. Moreover, Cx31R42P hemichannels play an important role in mediating ATP release from the cell. In contrast, no hemichannel activity was detected with cells expressing wildtype Cx31. Together, the results suggest that Cx31R42P forms constitutively active hemichannels to promote necrotic cell death. The Cx31R42P active hemichannels are likely resulted by an ER stress mediated ROS overproduction. The study identifies a mechanism of EKV pathogenesis induced by a Cx31 mutant and provides a new avenue for potential treatment strategy of the disease.

To investigate functional mechanism of pathogenic Cx31 mutants, we established cell lines stably expressing wildtype Cx31 and EKV associated mutation R42P. Cx31R42P stable cell lines expressing mutant protein at neglective level when cultured at 37uC. However, the amount of mutant proteins increased and cell death phenotype was observed when cells were grown at 26uC. The potential mechanism for Cx31R42P promoting cell death is that the mutant protein induces ER stress resulting in overproduction of reactive oxygen species (ROS). In turn, excessive ROS promote Cx31R42P hemichannels opening, leading to cell death.

Expression of Cx31R42P induces necrotic cell death
We have recently found that EKV pathogenic Cx31 mutants are temperature sensitive mutants. In cells, mutant proteins are rapidly degraded at 37uC while they become stable and form functional gap junctions at 26uC (unpublished data). To further investigate the pathophysilogical function of EKV pathogenic mutants, we stably expressed Cx31WT and Cx31R42P in Hela cells. At 37uC, exogenous Cx31WT formed gap junction plaques between adjacent cells ( Figure 1B). These plaques were further elongated at 26uC ( Figure 1E). In contrast, Cx31R42P-EGFP was hardly detectable at 37uC ( Figure 1C). However, the mutant protein not only showed increased detection but also formed gap junction plaque-like structures when cells were cultured at 26uC. Meanwhile, cytoplasmic aggregates of Cx31R42P were also detected ( Figure 1F, G). Similar observation was made in cells expressing myc-tagged Cx31 variants ( Figure S1).
After 96 h of incubation at 26uC, a large portion of Cx31R42Pexpressing cells exhibited condensed small nuclei (27.4666.45%, N = 3; Figure 1F, H, I; Figure 2A; Figure S1F-I). Cells with the characterized small nuclei were positive with pepidium iodine (PI) staining ( Figure S2). The LDH release from Cx31R42P cells was also significantly increased than that from Cx31WT cells and control Hela cells ( Figure 2C). In contrast, few Cx31WTexpressing cells or control Hela cells were found to have small nuclei when they were cultured at either 26uC or 37uC (Figure 1 and 2A). Neither small nuclei phenotypes nor increased LDH release was found when Cx31R42P cells were cultured at 37uC (Figure 2A, C). Consistent with the observation, transient expression of Cx31R42P at 37uC resulted in similar small nuclei phenotype ( Figure S3). The results suggest that expression of Cx31R42P cells induces cell death. Transmission electron microscopy (TEM) analysis revealed that Cx31R42P-induced small nuclei cells had morphological features of necrosis including translucent cytoplasm, swelling of organelles, disruption of the plasma membrane, dilatation of the nuclear membrane and condensation of chromatin into small, irregular patches [23] ( Figure 2F). Moreover, the percentage of nuclei with typical apoptotic morphology was similar among cells expressing Cx31 variants and control Hela cells ( Figure 2B), indicating that Cx31R42P unlikely induces apoptosis. Pan-caspase inhibitor z-VAD had little effect on Cx31R42P induced cell death ( Figure 2D). Thus, overexpression of Cx31R42P likely results in necrotic cell death.

Cx31R42P but not Cx31WT forms constitutively active hemichannels
To elucidate the molecular mechanism of EKV Cx31mutant induced cell death, we examined activity of gap junctions and hemichannels formed by Cx31R42P. Gap junctions formed by Cx31 variants effectively mediate dye transfer between cells ( Figure S4). Therefore, the gap junction activity of Cx31R42P unlikely promotes cell death. FFA, a well-characterized connexin hemichannel blocker, significantly inhibited Cx31R42P induced cell death as evidenced by abolishing cells with small nuclei ( Figure 3A) and significantly reducing LDH release ( Figure 3B). FFA had little effect on Cx31WT-expressing cells (data not shown). The results suggest that Cx31R42P forms constitutively active hemichannels at cell surface and that hemichannel activity plays important roles in Cx31R42P induced cell death.
Extracellular Ca 2+ (Ca 2+ o ) regulates gating of connexin hemichannels [3,[24][25]. Connexin hemichannels are at low open probability at physiological Ca 2+ o concentration and can further be inhibited with high Ca 2+ o concentration [26][27]. We next examined whether Cx31R42P induced cell death is regulated by Ca 2+ o . High Ca 2+ o was achieved by adding 3.2 mM calcium chloride to the culture medium to increase final Ca 2+ o concentration to 5 mM. Consistent with a notion that Cx31R42P forms constitutive active hemichannels, Cx31R42P-induced cell death with small nuclei was completely inhibited by high Ca 2+ o incubation ( Figure 3C). The LDH release from cells expressing Cx31R42P was also significantly reduced ( Figure 3D). In contrast, the percentage of typical apoptotic cell death was unaffected at high Ca 2+ o (data not shown). Also, expression level of Cx31R42P remains similar in physiological and high Ca 2+ o condition ( Figure  S5), indicating that the effect of high Ca 2+ o is not resulted by reduced expression of mutant protein.
Furthermore, we incubated Cx31R42P cells in high Ca 2+ o condition at 26uC for 4 days followed by changing into fresh medium with or without additional calcium. LDH release was measured at different time points after medium change. Significant increase in LDH release was observed in cells stably expressing Cx31R42P ( Figure 3E) but not in cells stably expressing Cx31WT ( Figure 3F) after cells were incubated in medium with physical Ca 2+ o condition for 2 h. These results suggest that Cx31R42P induced cell death is Ca 2+ o concentration dependent. Together, these results suggest that Cx31R42P forms constitutively active hemichennels to promote cell death.

Cx31R42P induces ROS production via ER stress
A previous study indicates that EKV-pathogenic Cx31 mutants induce ER stress [21]. Consistent with the observation, expression of Cx31R42P, but not the Cx31WT, upregulated the ER stress marker Bip ( Figure 4A). Treatment of salubrinal, an ER stress inhibitor [28], abolished cell death with small nuclei in Cx31R42P cells ( Figure 4B). LDH release was also markedly decreased (10.2761.57% v.s. 40.2162.90%; N = 3) ( Figure 4C). In contrast, cells with typical apoptotic nuclei remained unaffected (data not shown). The expression level of Cx31R42P was unchanged with the treatment ( Figure S5). The results suggest that Cx31R42P induced ER stress plays important roles in regulating Cx31R42P cell death.
Overproduction of ROS contributes to the execution of necrosis [23,[29][30]. To investigate the mechanism of Cx31R42P induced cell death, we detected intracellular ROS production with 29,79-Dichlorofluorescin diacetate (DCFH-DA) staining followed by FACS analysis. Results showed that levels of ROS were similar among cells expressing Cx31 variants at 37uC ( Figure 4D). At 26uC, the relative level of ROS in cells stably expressing Cx31R42P (310.20613.63%, N = 3) was significantly higher than that in mock transfected cells (100%) and cells stably expressing Cx31WT (86.9863.311%, N = 3) ( Figure 4E). Interestingly, ROS production in Cx31R42P treated with ER stess inhibitor salubrinal was reduced to 66.6665.094% comparing to that treated with solvent ( Figure 4F). The results suggest that the overproduction of ROS in cells expressing Cx31R42P is, at least partially, caused by ER stress.

ROS scavenger protects Cx31R42P-induced cell death
To determine whether ROS production contributes to Cx31R42P-induced cell death, we treated cells expressing Cx31R42P with ROS scavenger butylated hydroxyanisole (BHA). After four days of treatment, cell death were evaluated by DAPI staining and LDH release analysis. Results showed that the percentage of Cx31R42P cells with small nuclei was reduced to baseline ( Figure 5A). Likewise, LDH release was significantly decreased in cells treated with BHA compared with those treated with solvent (7.4560.53% v.s. 24.8660.40%, N = 3) ( Figure 5B). In contrast, cells with typical apoptotic nuclei appeared unaffected (data not shown) and the expression level of Cx31R42P remained unaffected ( Figure S5). The results suggest that Cx31R42P induces necrotic cell death via a ROS mediated pathway.

Gating of Cx31R42P hemichannels is regulated by the ROS scavenger
To examine the gating of Cx31R42P hemichannels, we assayed their uptake of dye with molecular weight less than 1 KD from the culture medium. Cells were cultured in high Ca 2+ o containing medium at 26uC for 4 days followed by in fresh medium containing 0.  Figure 6B3-4; D). As a control, KN-62, a P2X7 ATP receptor inhibitor, showed no effect on LY uptake by the Cx31R42P cells ( Figure 6B5-6; D). Interestingly, ROS scavenger BHA completely suppressed LY uptake by the Cx31R42P cells ( Figure 6B7-8; D). The results suggest that ROS plays an important role in maintaining activity of Cx31R42P hemichannels.
Cx31R42P hemichannels regulate ATP release from cells ATP has been recognized as an important autocrine and paracrine signaling molecule. In skin, ATP regulates keratinocyte differentiation and increase proliferation [31][32]. Connexin hemichannels mediates ATP release under various physiological and pathological conditions [2,[33][34]. To investigate whether  Figure 7A). In contrast, the rate of ATP release from Cx31R42P cells was significantly higher incubated in medium with physiological Ca 2+ than that incubated in medium with high Ca 2+ (1.3960.17% v.s. 0.0460.37610 22 %, N = 3). Nevertheless, LDH level was not affected, suggesting that increased ATP release from Cx31R42P cells is not resulted by the plasma membrane leakage ( Figure 7B). Furthermore, ATP release from Cx31R42P cells was completely blocked by hemichannel blocker FFA, ROS scavengers BHA and BHT ( Figure 7C). Consistent with the result that ER stress inhibitor, salubrinal, reduced ROS production, it also reduced ATP release though Cx31R42P hemichannels ( Figure 7D). The results suggest that Cx31R42P hemichannels mediate ATP release under physical Ca 2+ o .
Cx31R42P is considered mostly unfolded protein. Its unstable nature at 37uC and formation of intracellular aggregates in 26uC further support this notion. Both EKV and hearing loss pathogenic Cx31 mutants are reported to induce ER stress [19,21]. Thus, it is not surprise that expression of Cx31R42P causes ER stress, therefore, ROS overproduction [35][36][37]. Accordingly, inhibition of ER stress suppresses ROS production in cells expressing Cx31R42P. To our surprise, ROS scavenger treatment not only prevents Cx31R42P-induced cell death but also inhibit the activity of Cx31R42P hemichannels. Thus, the results suggest that the excessive ROS likely results in open of Cx31R42P hemichannels. This creates a feed-back mechanism to further enhance the deleterious of Cx31R42P mutant protein. It remains unclear how the ROS will regulates the gating of Cx31R42P hemichannels. Nevertheless, we cannot exclude the direct involvement of Cx31R42P induced ROS in executing cell death.
EKV is a rare hereditary skin disease pathologically characterized by hyperkaretostic perikaretosis [17,38]. Both abnormal differentiation and hyperproliferation of karetinocytes are detected in the lesion. It remains unclear how the mutant induced cell death to contribute the disease initiation and progression. One potential mechanism is that cell death promotes inflammation that produces cell growth-prone cytokines. Consistent with this hypothesis, inflammation is often detected in the disease tissues of EKV patients [17,[38][39][40].

Materials and Methods
Materials. Hela cells were obtained from ATCC and maintained as suggested by the provider. The EGFP antibody was purchased from Clontech. Myc-tag antibody was purchased from Cell Signaling. All secondary antibodies were purchased from Jackson ImmunoResearch. Salubrinal and z-VAD were purchased from Santa Cruz and Promega respectively. All other reagents were from Sigma.
EGFP-tagged and myc-tagged cDNAs encoding Cx31 variants and Hela cells stably expressing Cx31 variants were generated as described previously [19,41]. The stable cell lines were maintained in 200 mg/ml G418. To induce Cx31R42P expression, cells were seeded at 10% confluence, cultured at 37uC for four days followed by 26uC for four days.
Quantification of cell death. Cells were fixed in 4% paraformaldehyde for 10 min followed by being penetrated with 0.1% tritonX-100, and then the nuclei were stained with 1 mg/ml DAPI. The pictures were randomly taken with a fluorescence microscopy (Leica DMI 3000B). Over 500 cells were counted for each condition. Each experiment was repeated for three times.
LDH release detection. Cell culture medium was collected and centrifuged at 50006g for 10 min to remove the detached dead cells and cell debris. Cells were lysed in a lysis buffer (0.1% triton X-100 in PBS) for 10 min at 4uC. LDH in the medium and cell lysates were measured using CytoTox 96H Non-Radioactive Cytotoxicity Assay (Promega) according to manufacture's instructions. The rate of LDH release was given by: %LDH release = medium A492/total A492 = medium A492/medium A492+ cell lysate A492. Each experiment was repeated for three times.
TEM. Cells were trypsinized, pelleted by centrifugation, fixed in 2.5% glutaric dialdehyde and postfix treated with 2% osmium tetroxide. The samples were dehydrated with a graded series of acetones, infiltrated in acetone/resin (1:1) and embedded in Epon. Fifty nanometer sections were collected and double stained with uranyl acetate and lead nitrate. The samples were examined with a transmission electron microscope (JEOL-1230) at 80 Kv accelerating voltage.
ROS detection. Cells were digested with collagenase IV (Gibco), pelleted and suspended in the medium containing 20 mM DCFH-DA. After 30 min of incubation, cells were centrifuged at 20006g for 10 min, resuspended in a fresh medium and subjected to FACS analysis (Beckman-Coulter MoFLo XDP cell sorter). The results were analyzed with FlowJo 7.6 software for the mean fluorescence intensity (MFI). The data presented are the values relative to those obtained in Hela cells or solvent-treated control. The experiments were repeated for three times.
Dye uptake experiments. The dyes (0.5 mg/ml lucifer yellow and 10 mg/ml Dextran-FITC) were loaded to cells in the fresh medium with or without additional 3.2 mM calcium chloride. Fifteen minutes later, dye was removed. Cells were washed three times with medium containing additional 3.2 mM calcium chloride. Pictures were randomly taken using fluorescence microscopy Leica DMI 3000B. Over 500 cells were analyzed for each condition. The experiments were repeated for three times.
Detection of ATP release. Cell culture medium was collected and centrifuged at 50006g for 5 min to remove the detached dead cells and cell debris. ATP in the medium and cells was measured using a CellTiter-GloH Luminescent Cell Viability Assay (Promega) according to manufacture's instructions. The rate of ATP release was given by: %ATP release = medium RLU/total RLU = medium RLU/medium RLU+ cells RLU.
Statistic analysis. Statistic analyses were performed using Prism 5 software (GraphPad). The data is presented as mean 6 SEM. Two-tailed Student's t test was used to determine the significance of difference between 2 groups. One-way ANOVA with Tukey's and Dunnett's Multiple Comparison Test were used to assess the difference among three cell lines and between treatments groups against their respective controls, respectively. Two-way ANOVA with Bonferroni post-test was used to analyze the data of multiple-group treatments.

Supporting Information
Figure S1 Intracellular distribution of Cx31 variants in stable cell lines. Hela cells stably expressing myc-tagged Cx31WT, Cx31R42P and their control line (Hela) were grown either at 37uC or at 26uC. Cells were immunostained with antimyc (red) antibody and nuclei were stained with DAPI (blue). High magnification pictures of Cx31R42P-expressing cells are also shown to exemplify the gap junction plaque-like structures and small condensed nuclei (bottom panel). Note that Cx31WT-myc form gap junction plaques between adjacent cells incubated at 37uC (B) or at 26uC (E). Cx31R42P-myc is barely detectable at 37uC (C). It is accumulated predominantly in cytoplasm (F, G) and can form gap junctions (F and G, green arrow) at 26uC. Small