Reciprocal positive regulation between Cx26 and PI3K/Akt pathway confers acquired gefitinib resistance in NSCLC cells via GJIC-independent induction of EMT

Gefitinib efficiency in non-small-cell lung cancer (NSCLC) therapy is limited due to development of drug resistance. The molecular mechanisms of gefitinib resistance remain still unclear. In this study, we first found that connexin 26 (Cx26) is the predominant Cx isoform expressed in various NSCLC cell lines. Then, two gefitinib-resistant (GR) NSCLC cell lines, HCC827 GR and PC9 GR, from their parental cells were established. In these GR cells, the results showed that gefitinib resistance correlated with changes in cellular EMT phenotypes and upregulation of Cx26. Cx26 was detected to be accumulated in the cytoplasm and failed to establish functional gap-junctional intercellular communication (GJIC) either in GR cells or their parental cells. Ectopic expression of GJIC-deficient chimeric Cx26 was sufficient to induce EMT and gefitinib insensitivity in HCC827 and PC9 cells, while knockdown of Cx26 reversed EMT and gefitinib resistance in their GR cells both in vitro and in vivo. Furthermore, Cx26 overexpression could activate PI3K/Akt signaling in these cells. Cx26-mediated EMT and gefitinib resistance were significantly blocked by inhibition of PI3K/Akt pathway. Specifically, inhibition of the constitutive activation of PI3K/Akt pathway substantially suppressed Cx26 expression, and Cx26 was confirmed to functionally interplay with PI3K/Akt signaling to promote EMT and gefitinib resistance in NSCLC cells. In conclusion, the reciprocal positive regulation between Cx26 and PI3K/Akt signaling contributes to acquired gefitinib resistance in NSCLC cells by promoting EMT via a GJIC-independent manner.

Lung cancer, of which non-small-cell lung cancer (NSCLC) is the most common form, remains the leading cause of cancerrelated deaths worldwide. 1 Currently, gefitinib, as the first epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI), is one of the most accepted therapies against NSCLC carrying EGFR mutations. However, almost all NSCLC patients who initially respond well to EGFR-TKIs eventually develop acquired resistance. 2 Development of effective therapeutic interventions to overcome gefitinib resistance is an urgent need.
Epithelial-mesenchymal transition (EMT), during which cancer cells lose epithelial markers such as E-cadherin but gain mesenchymal markers such as vimentin, is known to be deeply involved in cancer progression and chemotherapy resistance. Specially in NSCLC, EMT plays pivotal roles in the acquired resistance to EGFR-TKIs such as gefitinib. 3,4 For example, restoring E-cadherin expression or silencing EMT regulator Slug increases gefitinib sensitivity in NSCLC cells with a mesenchymal phenotype. 5,6 Accumulating evidences indicate that constitutively activation of the phosphoinositide 3-kinase (PI3K)/Akt signaling is a central feature of EMT in many cancers including NSCLC. 7,8 However, the exact mechanism for the acquired gefitinib resistance of NSCLC remains unclear.
Connexins (Cxs) are a family of transmembrane proteins, which compose the intercellular gap junctions between the neighboring cells. 9 Gap junctions directly connect the cytoplasms of adjacent cells, thereby mediating direct exchange of signaling molecules smaller than 1 kDa, such as ions, small metabolites, and second messengers. This process is termed gap-junctional intercellular communication (GJIC). Cx expression and/or GJIC are frequently reduced or loss in malignant cell lines and cancers, while restoration of Cx expression and/ or GJIC retarded tumor growth and increased cytotoxicities of chemotherapeutics such as cisplatin and docetaxel. [10][11][12][13] Therefore, Cxs have long been deemed tumor suppressors. However, increasing new observations were apparently contradicting the 'dogma' and became clear that Cxs and GJIC also contribute to cancer progression and chemoresistance. For example, Cx32 expression was detected in breast cancer and significantly increased in lymph node metastases compared with primary tumors, suggesting Cx32 may be a sign of more malignant phenotype of breast cancer. 14 Besides, cytoplasmic accumulation of Cx32 exerted favorable effects for hepatocellular carcinoma (HCC) progression including invasion and metastasis by Cx linked, but GJICindependent mechanism. 15 Recently, Gielen et al. 16 reported that increasing the level of Cx43 confers temozolomide resistance in human glioma cells whereas knockdown of Cx43 sensitizes them to temozolomide treatment via both GJIC-dependent and -independent mechanisms.
Up to now, there are~21 isoforms of Cxs that distribute in almost all human organs in tissue-specific patterns. 17 Cx26, one of the most common isoforms of Cxs, is predominantly expressed in lung tissue. 18,19 Despite Cx26 has been considered as a potential tumor suppressor or chemotherapy sensitizer in some types of tumors, 20,21 Ito et al. 22 found that Cx26 helps lung squamous cell carcinoma (SCC, one histological type of NSCLC), acquire aggressive phenotypes, lymph node metastasis, and poor prognosis, indicating that a potential role of Cx26 on the malignant development of SCC. However, the roles of Cx26 and its derived GJIC in the development of gefitinib resistance in NSCLC have not been explored.
In this study, to clarify the potential role of Cx26 and its derived GJIC in gefitinib resistance in NSCLC, we first surveyed the expression of four major Cxs in different gefitinib-sensitive NSCLC cell lines and found a positive correlation between high level of Cx26 and gefitinib insensitivity in NSCLC cells. Such an association was further confirmed in established gefitinib-resistant (GR) HCC827 and PC9 cell lines both in vitro and in vivo. Importantly, we find a positive mutual regulation between Cx26 and PI3K/Akt pathway, which confers acquired gefitinib resistance in NSCLC cells by GJIC-independent induction of EMT.

Results
Cx26 upregulation is correlated with gefitinib insensitivity of human NSCLC cells. We first performed RT-PCR and western blotting to determine Cx expression phenotype in different gefitinib-sensitive NSCLC cell lines. As shown in Figures 1a-c, four major Cx isoforms, Cx26, Cx32, Cx31.1, and Cx43, were differentially expressed in four NSCLC cell lines (HCC827, PC9, A549, and H1299). In particular, Cx26 was the predominant Cx isoform expressed in various Figure 1 Increased Cx26 is positively correlated with gefitinib resistance in NSCLC cells. (a) Differential expression of Cx26, Cx31.1, Cx32, and Cx43 in different gefitinibsensitive NSCLC cell lines was determined by RT-PCR. (b and c) High level of Cx26 in gefitinib-insensitive A549 and H1299 cells than that in gefitinib-sensitive HCC827 and PC9 cells was detected by RT-PCR and western blotting. GAPDH or β-actin was used as internal loading control NSCLC cell lines. Moreover, the level of Cx26 was markedly higher in gefitinib-insensitive NSCLC cell lines (A549 and H1299) than that in gefitinib-sensitive NSCLC cell lines (HCC827 and PC9). These results suggest that Cx26 may be positively correlated with gefitinib insensitivity in NSCLC cells.
GR HCC827 and PC9 cell lines were established with acquired EMT characteristics and elevated Cx26 expression. To explore the mechanism of acquired gefitinib resistance of NSCLC, we generated two acquired GR NSCLC cell lines, HCC827 GR and PC9 GR, from their parental cells by continuous exposure to gefitinib starting at 0.001 μM and increasing in a stepwise manner to 1 μM. As shown in Figure 2a, gefitinib showed less cytotoxicity in established HCC827 GR and PC9 GR cells than that in their parental cells with IC50 of 12.17 ± 0.18 μM versus 0.06 ± 0.11 μM and 16.51 ± 0.17 μM versus 0.25 ± 0.07 μM, respectively.
Moreover, HCC827 GR and PC9 GR cells exhibited scattered, elongated, and mesenchymal-like morphology, while their parental HCC827 and PC9 cells showed rounded shape, typical of epithelial cobblestone appearance ( Figure 2b). Consistently, the expression of epithelial marker E-cadherin was greatly reduced, whereas the level of mesenchymal marker vimentin and slug was significantly elevated in HCC827 GR and PC9 GR cells (Figure 2c). A key feature of cancer cells undergoing EMT is enhanced migratory and invasive potential. As shown in Figure 2d, mobility and invasive capability of HCC827 GR and PC9 GR cells were significantly increased by 2.6-or 3.0-fold and 2.0-or 2.4-fold compared with their parental cells, respectively. Moreover, the levels of Cx26 were increased in HCC827 GR and PC9 GR cells (Figure 2e). These results suggest a potential role of Cx26 in the acquisition of EMT and acquired gefitinib resistance of NSCLC cells.
Cx26 induces acquired gefitinib resistance in NSCLC cells via GJIC-independent manner. Cxs have long been believed to regulate tumor development during carcinogenesis by exerting GJIC. Therefore, we next examined whether GJIC was involved in Cx26-induced EMT and acquired gefitinib resistance of NSCLC cells. First, GJIC in primarily human foreskin fibroblasts (HFFs) as positive control was confirmed, and treatment of these cells with RA (a welldefined GJIC enhancer) significantly enhanced GJIC among these cells. As shown in Figure 3a, no detectable GJIC was found in HCC827, PC9, and their GR cells. To exclude the involvement of undetectable GJIC in these cells, GJIC was further measured in the presence of 10, 20, and 40 μM of RA for 4, 8, 12, 24, and 48 h, respectively. The result showed no enhancement of GJIC in these cells incubated with RA ( Figure 3b). These observations suggest that Cx26 is not functional as a component of gap junctions in these cells. Furthermore, we assessed the subcellular localization of Cx26 in these cells. Figure 3c clearly showed that Cx26 protein accumulated in the cytoplasm of HCC827, PC9 cells, and their GR cells. Although, incubated with the GJIC enhancer RA, Cx26 was still retained in the cytoplasm ( Figure 3d). These results indicate that Cx26 cannot form functional gap junctions between these cells due to the absence of integration into plasma membrane, confirming that GJIC is not implicated in the Cx26-mediated EMT and acquired gefitinib resistance in NSCLC cells.
To explore the role of Cx26 per se in the regulation of EMT and acquired gefitinib resistance in NSCLC, we engineered GJIC-deficient HCC827 and PC9 cells stably expressing chimeric Cx26 with the green fluorescent protein (GFP) fused to the amino-terminal ( Figure 4a). Characterization of this chimeric protein revealed that Cx26 accumulated in the cytoplasm and failed to establish functional GJIC ( Figure 4b). After incubation with RA, Cx26 was still retained in the cytoplasm with no detectable GJIC ( Figure 4c). Despite  Figure 4f). Furthermore, Cx26 overexpression exerted obvious gefitinib insensitivity in these cells ( Figure 4g). Besides, the in vivo data showed that administration of gefitinib (100 mg/kg per day, gavaged orally) led to more significant inhibition of HCC827-mock tumor xenografts than HCC827-Cx26 xenografts, compared with vehicle groups (Figure 4h). These results reinforce the GJICindependent role of Cx26 in the promotion of EMTand gefitinib resistance in NSCLC.
To further confirm the effect of Cx26 on EMT and gefitinib resistance in NSCLC, we transducted Cx26 short hairpin RNA (shRNA, shCx26) or scramble shRNA into HCC827 GR and PC9 GR cells ( Figure 5a). Knockdown of Cx26 expression significantly restored the rounded epithelial-like appearance (Figure 5b), enhanced E-cadherin expression while reduced vimentin and slug expression (Figure 5c), and meanwhile strongly inhibited migratory and invasive potential of HCC827 GR and PC9 GR cells (Figure 5d). Moreover, gefitinib efficacy was substantially increased in shCx26-transduced HCC827 GR and PC9 GR cells (Figure 5e). The capability of Cx26 depletion to restore gefitinib sensitivity of NSCLC was also observed in in vivo tumor model. As shown in Figure 5f, administration of gefitinib (100 mg/kg per day, gavaged orally) triggered more dramatic regression of shCx26-transduced HCC827 GR tumor xenografts than scramble HCC827 GR xenografts, compared with vehicle groups.
Taken together, these results indicate that Cx26 per se, but not the extent of GJIC, corresponds to acquired gefitinib resistance in NSCLC cells via induction of EMT both in vitro and in vivo.
Reciprocal positive regulation between Cx26 and PI3K/ Akt pathway is involved in Cx26-mediated EMT and gefitinib resistance of NSCLC cells. Based on the aforementioned observations, we became interested in exploring the molecular mechanism underlying the GJIC-independent role of Cx26 in the stated effects. PI3K/Akt pathway is known to play a prominent role in driving EMT and drug resistance in cancers. 23 It has been reported that activation of PI3K/Akt signaling could directly increase Cx43 phosphorylation 24 and Cx43 also could contribute to activation of PI3K/Akt signaling. 25 Therefore, we sought to determine whether there exists a reciprocal activation between Cx26 and PI3K/Akt pathway in promoting EMT and acquired gefitinib resistance of NSCLC cells. As shown in Figures 6a-d, treatment of Cx26-overexpressing HCC827 and PC9 cells with a specific PI3K/Akt pathway inhibitor LY294002 (25 μM) for 24 h could apparently antagonize the facilitating effects of Cx26 on EMT and gefitinib resistance of NSCLC cells. However, LY294002 treatment only had little effect on cell invasion and migration, as well as gefitinib efficacy in vitro. Consistent results were obtained from these cells treated with another selective PI3K/Akt pathway inhibitor wortmannin (10 μM) for 4 h (data not shown). Moreover, Cx26 overexpression significantly activated PI3K/Akt pathway as represented by elevated Akt phosphorylation in HCC827 and PC9 cells, while Cx26 depletion caused reduced PI3K/Akt activity in HCC827 GR and PC9 GR cells (Figure 6e). In vivo studies showed that treatment with LY294002 (25 mg/kg, twice a week, i.p.) induced marked tumor regression of Cx26-overexpressing group to the level comparable to that of mock control group (Figure 6f). Together, these findings suggest that activation of PI3K/Akt pathway is sufficient to account for Cx26-promoted EMT and gefitinib resistance in NSCLC cells.
PI3K/Akt pathway is constitutively activated in various cancers including NSCLC. 26,27 Thus, we were interested in whether Akt activation induces Cx26 expression. As shown in Figure 7a, treatment with 25 μM LY294002 for 24 h caused a significant reduced Cx26 expression both in HCC827, PC9 cells, and their GR cells. Moreover, ectopic expression of Akt significantly increased Cx26 expression in these cells (Figure 7b).
In addition, we investigated the biological significance of the mutual positive regulation between Cx26 and PI3K/Akt pathway in EMT and gefitinib resistance of NSCLC cells. As shown in Figures 7c-f, Akt overexpression alone also induced EMT and gefitinib resistance of HCC827 and PC9 cells. Cx26 overexpression strengthened Akt-facilitated EMT and gefitinib resistance, whereas Cx26 depletion rendered impaired Aktpromoted effects in these cells. Collectively, these results indicate that interdependent positive regulation of Cx26 and PI3K/Akt pathway contributes to gefitinib resistance in NSCLC through induction of EMT.

Discussion
We present here that a reciprocal positive regulation exists between Cx26 and PI3K/Akt signaling, thus providing insights into the molecular mechanism underlying the dysregulation of Cx26 and PI3K/Akt in NSCLC cells. Furthermore, the functional interplay between Cx26 and PI3K/Akt signaling contributes to the acquired gefitinib resistance in NSCLC cells by GJIC-independent induction of EMT.
Cxs are frequently deregulated in cancers from different origins, either by reduction, lack of expression, or upregulation. 28,29 In this study, we found that various NSCLC cell lines have high level of Cx26, but moderate level of Cx32 and Cx31.1, and only low level of Cx43. Such aberrant Cx expression is in agreement with accumulating evidences indicating that different Cxs have different facets in cancer chemoresistance. For instance, Yu et al. 30 reported that Cx43 overexpression reversed EMT and cisplatin resistance in cisplatin-resistant NSCLC cell lines. On the contrary, two recent reports showed that Cx43 knockdown could sensitize glioblastoma cells to temozolomide. 16,31 Especially for Cx26, its upregulation improved gemcitabine anticancer efficacy in pancreatic cancer cells. 21 However, in this study, we demonstrate that Cx26 is the predominant Cx isoform expressed in NSCLC cells, and Cx26 upregulation contributes to gefitinib resistance via induction of cell EMT. Together, while these opposing observations underscored the complex role of Cxs in the development of cancer chemoresistance, our results reveal a novel role of Cx26 that implicates in the acquisition of EMT and gefitinib resistance in NSCLC cells.
facilitating roles of Cxs in tumorigenesis and cancer chemoresistance via GJIC-independent manner. For example, Cx43 could promote the resistance to temozolomide or cisplatin in cancer cells in a GJIC-independent manner. 16,35 Moreover, the cytoplasmic Cx32 protein itself, which failed to form GJIC, could facilitate progression of HCC. 15 In this work, 'parachute' dye-coupling assay showed no functional GJIC in HCC827 and PC9 cells with low Cx26 expression, and their GR cells with high Cx26 expression. Immunofluorescence staining revealed that Cx26 is aberrantly accumulated in the cytoplasm but not in the normal cell-cell contact areas in these cells. Pharmacological stimulation using RA, a well-defined GJIC enhancer, has no enhancement effects on GJIC in these cells, and could not change the cytoplasmic localization of Cx26. Thus, these results indicate that Cx26 is incapable of forming functional GJIC between NSCLC cells due to the defects in plasma membrane assembly, excluding the possible involvement of GJIC in the Cx26-mediated EMTand acquired gefitinib resistance in NSCLC cells.
Many studies support a role of Cx26 in tumorigenesis that might be independent of GJIC. Cytoplasmic accumulation of Cx26 has been associated with lung metastasis in colorectal cancer 36 and with poor prognosis in NSCLC and breast carcinoma. 22,37 Actually, in the present study, we found that overexpression of chimeric Cx26, which resulted in a significantly increase of Cx26 in the cytoplasm of HCC827 and PC9 cells, was sufficient to induce EMT phenotypes and gefitinib insensitivity in vitro and in vivo. On the contrary, knockdown of Cx26 reversed EMT and gefitinib resistance in their GR cells and the tumor model. Taken together with the above observations, these results reinforce the GJICindependent role of Cx26 in the promotion of EMTand gefitinib resistance in NSCLC.
PI3K/Akt pathway-dependent EMT has been shown to contribute to cisplatin resistance in HCC cells 38 and gefitinib resistance in head and neck SCC cells. 39 Therefore, in this study, whether EMT and gefitinib resistance in NSCLC cells mediated by Cx26 itself is dependent on PI3K/Akt pathway was determined. We found that inhibition of PI3K/Akt by specific inhibitors LY294002 or wortmannin could reverse EMT and gefitinib resistance in Cx26-overexpressed NSCLC cells. Inhibition of PI3K/Akt also led to tumor regression in Cx26-overexpressed xenografts. Moreover, Cx26 overexpression significantly activated Akt in parental NSCLC cells, while Cx26 depletion reduced PI3K/Akt activity in their GR cells. Consequently, these results indicate that Cx26 contributes to EMT and gefitinib resistance in NSCLC cells mainly through activation of PI3K/Akt pathway. However, the mechanisms by which Cx26 stimulates PI3K/Akt pathway in NSCLC cells have not been explored. Cx43 has been shown to contribute to the activation of PI3K/ Akt signaling possibly as a cofactor of Gβ in cardiomyocytes. 25 Moreover, a positive correlation between Cx26 expression and insulin-like growth factor receptor I (IGF-IR) has been demonstrated in human colorectal cancer. 40 IGF-IR upregulation could mediate resistance to EGFR-TKI therapy in primary human glioblastoma cells through continued activation of PI3K/Akt signaling. 41 These findings combined with ours suggest that the mechanisms for Cx26-stimulated PI3K/Akt pathway are complicated and there may be crosstalk with other signals, such as IGF-IR, to subsequently activate PI3K/ Akt pathway.
Interestingly, herein, we also demonstrated that inhibition of PI3K/Akt pathway results in decreased Cx26 expression, whereas overexpression of Akt increases Cx26 expression in NSCLC cells. Supporting these observations was the involvement of activation of PI3K/AKT pathway in TGF-β1induced Cx43 expression. 42 Besides, activation of PI3K/AKT pathway by shear stress led to increased nuclear accumulation of β-catenin, which could bind to the Cx43 promoter and stimulate Cx43 expression. 43 Therefore, our results demonstrate that there exists a positive feedback regulation between Cx26 expression and PI3K/Akt pathway in NSCLC cells.
In addition, our study showed that overexpression of either Cx26 or Akt alone results in EMT phenotypes and gefitinib resistance in NSCLC cells. Furthermore, Cx26 overexpression enhanced Akt-induced EMT and gefitinib resistance, while Cx26 knockdown led to impaired Akt-mediated effects in these cells. These results indicate that dysfunction of either Cx26 or Akt contributes to acquisition of EMT and gefitinib resistance in NSCLC cells. More importantly, the positive regulatory circuit that mutually reinforces the Cx26 expression and PI3K/Akt activity further augments the EMT and gefitinib resistance in NSCLC cells.
Despite further studies are needed to explore the efficacy of disruption of regulatory network between Cx26 expression and PI3K/Akt pathway in targeted therapy for NSCLC with aberrant Cx26 expression or PI3K/Akt activation, our study shows a positive regulatory loop between Cx26 expression and PI3K/Akt pathway, which confers the acquired gefitinib resistance in NSCLC cells by promoting EMT in a GJICindependent manner. In conclusion, the Cx26 may be an attractive target for overcoming gefitinib resistance in NSCLC therapy.   Cells were maintained at 37°C and 5% CO 2 in RPMI 1640 supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA) and antibiotics (100 units/ml penicillin and 100 μg/ml streptomycin). Two gefitinib-sensitive NSCLC cell lines HCC827 and PC9, both harboring EGFR exon 19 in-frame deletion mutation, 6 were exposed to increasing concentrations of gefitinib and generated GR cells as reported previously. 2,44 The procedures of isolation and culture of HFFs were shown in the Supplementary File.
RNA interference and overexpression of Cx26. The hU6-MCSubiquitin-EGFP-IRES-puromycin lentiviral RNAi vectors (GeneChem, Shanghai, China) containing shRNA against Cx26 and Ubi-MCS-EGFP-IRES-puromycin lentiviral vector (GeneChem) containing chimeric Cx26 where the GFP is tagged to the amino-terminal of Cx26 were constructed as other studies reported previously. 45,46 The shRNA with no target gene (scramble) or empty lentiviral vector (mock) was used as control. Cells were infected by lentiviral supernatant plus 5 μg/ml Polybrene (GeneChem) and selected by 10 μg/ml puromycin for 14 days. Thereafter, the resistant clones were pooled and analyzed for Cx26 knockdown or overexpression by western blotting.
Western blot analysis. Western blot protocol was according to our previous report. 47 The blotted membrane was incubated with primary antibodies at final dilutions ranging between 1/1000 and 1/2000 and then probed with horseradish peroxidase (HRP)-labeled anti-rabbit secondary antibody. Antibody binding was detected by enhanced chemiluminescence (ECL) detection kit (Thermo Fisher Scientific, Waltham, MA, USA) and captured on X-ray film. The densitometry of the bands was quantified by Quantity One software on a GS-800 densitometer (Bio-Rad Laboratories, Hercules, CA, USA).
MTT assay. As we previously reported, 48   Cell migration and invasion assay. Cell invasion assay was performed according to our previous study. 49 The invaded cells were determined as eight highpower fields of cells were counted in each well under an inverted microscope at × 200 magnification. Invasion was presented as the relative invasive rate, which is calculated by the invasive rate of the treated group (number of invaded cells per total cell number) divided by that of the control group. The protocol of the migration assay was the same as that of invasion assay, except that no Matrigel (BD Biosciences, San Jose, CA, USA) was used.
'Parachute' dye-coupling assay. Functional GJIC was detected as described by Goldberg et al. 50 Donor and receiver cells were grown to confluence. Donor cells were double-labeled with 5 μmol/l CM-DiI (Life Technologies, Carlsbad, CA, USA), a membrane dye that does not spread to coupled cells, and with 5 μmol/l calcein acetoxymethyl ester, which is converted intracellularly into the gap junctionpermeable dye calcein. The donor cells were then trypsinized and seeded onto the receiver cells at 1 : 150 ratio. The donor cells were allowed to attach to the monolayer of receiver cells and form GJIC for 4 h at 37°C, and then examined with a fluorescence microscope (Nikon, TF-1, Tokyo, Japan). The average number of receiver cells containing dye per donor cell was considered as a measure of the degree of GJIC. For the studies involving GJIC enhancer, the donor and receiver cells were exposed to RA (10, 20, and 40 μM) for indicated time (4,8,12,24, and 48 h) in which the donor cells were plated onto the receiver cell monolayer.
Immunofluorescence. The cells were grown on coverslips for 24 h and then fixed in cold methanol for 10 min. Cells were blocked in 2% bovine serum albumin for 30 min at room temperature and incubated with goat anti-Cx26 primary antibody (diluted 1 : 50, Santa Cruz, Dallas, TX, USA) overnight at 4°C. The cells were then incubated with FITC-conjugated donkey anti-goat secondary antibody (diluted 1 : 200, Abcam, Cambridge, MA, USA) for 1 h at room temperature followed by counterstaining with 4′,6-diamidino-2-phenylindole (DAPI). Images were acquired on a confocal laser scanning fluorescence microscope (Nikon A1) and analyzed using the NIS-Elements software (Nikon Corporation, Tokyo, Japan).
Xenograft studies in athymic mice. All animal experiments were approved by the Institutional Animal Care and Use Committee of Guangxi Medical University, China. Five-to 6-week-old female nu/nu athymic mice were obtained from the Laboratory Animal Center of Guangxi Medical University. Cells were injected subcutaneously (1 × 10 7 cells per 200 µl PBS per mouse) into the right flank of these mice. When tumor volumes (calculated as (length × width 2 )/2) reached 100 mm 3 , as measured by calipers every 3 days, mice were randomly allocated into groups of six animals to receive gefitinib (100 mg/kg per day) 51 or vehicle by oral gavage, or a combination of gefitinib (100 mg/kg per day) and LY294002 (25 mg/kg, twice a week, i.p.). 52 All mice were killed on day 15 after their tumor size had been measured.
Statistical analysis. Data were presented as means ± S.D. and were statistically analyzed with unpaired Student's t-test at a significance level *Po0.05 and **Po0.01 using Sigma Plot 10.0 software (Jandel Scientific, San Rafael, CA, USA).